Structure-based drug design

Citation preview

netLibrary - eBook Summary

University of North Florida eBook Collection You are here:

Structure-based Drug Design by Veerapandian, Pandi. New York Marcel Dekker, Inc., 1997.

ISBN: 0824798694 eBook ISBN: 0585157448 Subject: Drugs--Design. Drugs--Structure-activity relationships. Drugs--Conformation. Drug Design. Structure-Activity Relationship. Language: English

Log out.

home > eBook summary

Read this eBook Browse this eBook online (borrow for a short time)

Check out and read online (add to "eBookshelf")

Hello, Tony. Go to your ● eBookshelf - 0 ●

My Favorites - 11

●

Bookmarks - 0

●

Notes - 0

Add to your My Favorites list Recommend this eBook to a friend netLibrary eBook Warranty Disclaimer

home | search tools | reading room | about us | help | log out © 2001 - 2004, netLibrary, a division of OCLC Online Computer Library Center, Inc. All rights reserved. privacy statement | terms of use

http://legacy.netlibrary.com/ebook_info.asp?product_id=12640&piclist=19799,20772,39801,42375 [4/9/2004 12:08:55 AM]

Structure-based Drug Design

Page #

Return to: home | | eBook summary

Help

Structure-based Drug Design Table of Contents Structure-Based Drug Design Preface Contents Contributors 1 Inhibitors of HIV-1 Protease 2 Structural Studies of HIV-1 Reverse Transcriptase and Implications for Drug Design 3 Retroviral Integrase: Structure as a Foundation for Drug Design 4 Bradykinin Receptor Antagonists 5 Design of Purine Nucleoside Phosphorylase Inhibitors 6 Structural Implications in the Design of Matrix-Metalloproteinase Inhibitors 7 Structure—Function Relationships in Hydroxysteroid Dehydrogenases 8 Design of ATP Competitive Specific Inhibitors of Protein Kinases Using Template Modeling 9 Structural Studies of Aldose Reductase Inhibition 10 Structure-Based Design of Thrombin Inhibitors 11 Design of Antithrombotic Agents Directed at Factor Xa 12 Polypeptide Modulators of Sodium Channel Function as a Basis for the Development of Novel Cardiac... 13 Rational Design of Renin Inhibitors

http://legacy.netlibrary.com/reader/reader.asp?product_id=12640 (1 of 2) [4/9/2004 12:09:15 AM]

Structure-based Drug Design

14 Structural Aspects in the Inhibitor Design of Catechol OMethyltransferase 15 Antitrypanosomiasis Drug Development Based on Structures of Glycolytic Enzymes 16 Progress in the Design of Immunomodulators Based on the Structure of Interleukin-1 17 Structure and Functional Studies of Interferon: A Solid Foundation for Rational Drug Design 18 The Design of Anti-Influenza Virus Drugs from the X-ray Molecular Structure of Influenza Virus Ne... 19 Rhinoviral Capsid-Binding Inhibitors: Structural Basis for Understanding Rhinoviral Biology and f... 20 The Integration of Structure-Based Design and Directed Combinatorial Chemistry for New Pharmaceut... 21 Structure-Based Combinatorial Ligand Design 22 Peptidomimetic and Nonpeptide Drug Discovery: Impact of StructureBased Drug Design Index

http://legacy.netlibrary.com/reader/reader.asp?product_id=12640 (2 of 2) [4/9/2004 12:09:15 AM]

Document

Page 1

1 Inhibitors of HIV-1 Protease Krzysztof Appelt Agouron Pharmaceuticals, Inc., San Diego, California I. Introduction Since the discovery of human immunodeficiency virus (HIV) as the causative agent of acquired immunodeficiency syndrome (AIDS), perhaps the largest and most powerful consortium of scientists ever assembled to tackle a single disease has been brought to bear on the problem of AIDS and its treatment. From an unprecedented wealth of information regarding the molecular biology and virology of HIV collected in recent years, it became possible to identify numerous intervention points in the viral life cycle that could be exploited in the development of drugs for AIDS therapy (for reviews see Reference 1, 2, and 3). Among these, the virally-encoded enzymes, in particular reverse transcriptase and protease, have emerged as the most popular targets. A separate chapter of this book is dedicated to the description of reverse transcriptase and its inhibitors [4]. For the purpose of introduction only, it should be noted that nucleoside inhibitors of reverse transcriptase (AZT, ddI, ddC, d4T, and 3TC) have been widely used in clinical practice since 1987. Since then it has become apparent that this class of agents, while slowing progression of disease in HIV-infected patients, is limited in both activity and the duration of the clinical responses produced. Therefore in the search for better anti-HIV agents, the focus of effort was expanded to include the search for clinically useful inhibitors of a second viral enzyme, namely the protease. In contrast to reverse transcriptase, for which activity is required prior to the integration of viral genetic information into the host cell chromosomes, the viral protease plays a key role late in the virus life cycle and inhibitors of this enzyme display equal anti-viral activity in chronic and acute infection models in vitro [5]. The HIV protease (HIV PR) is encoded by the 5' portion of the retroviral pol gene, which encodes all replicative enzymes. Viral structural proteins (p24,

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_1.html [2/29/2004 2:14:53 AM]

Document

Page 2

p17, p9, and p7) and replicative enzymes (protease, reverse transcriptase/ RnaseH, and integrase) are translated as either polyprotein P55-GAG, or a larger frameshift product P160-GAG-POL. In the process of virus assembly these polyproteins are proteolytically cleaved by the protease and this processing step, both in its timing and accuracy, is essential for the formation of infectious particles of HIV [6]. It was also shown early on that the inactivation of HIV PR, either by chemical inhibition or certain mutations, leads to the production of immature, noninfectious viral particles [7,8]. Structurally HIV PR is a 99-amino-acid protein translated initially as a central part of the P160-GAGPOL polyprotein precursor. The autocatalytic processing from the 160 kDa precursor is poorly understood, but most likely occurs during the process of budding of pre-formed viral particles from the host cell [9]. After release from the precursor polyprotein, HIV PR forms a homodimer and acts in trans to correctly process GAG and GAG-POL polyproteins—a process required for formation of the viral capsid and nucleoprotein core. Retroviral proteases such as HIV PR are the latest additions to the wellstudied family of aspartic proteases. This family of enzymes, which includes, among others, proteases such as pepsin, renin, and cathepsins D and E, has been intensely studied in the past, and the knowledge gained from studies of these enzymes allowed early inferences as to the structure and function of the dimeric HIV PR. Moreover, the intensive effort over the past two decades to make inhibitors of human renin provided impetus for the early design of inhibitors of HIV PR. In fact, some of the renin inhibitors have turned out to be effective inhibitors of retroviral aspartic proteases as well and have served as the starting point for drug design. As a result of this many early inhibitors of HIV PR were peptidyl in nature and the best known example of such compounds is Ro31-8959, better known as saquinavir, a hydroxyethylaminecontaining mimetic of a hexapeptide substrate [10]. This potent inhibitor of HIV PR was discovered using a substrate-based rational approach to drug design and displays extremely high in vitro activity against clinical isolates and laboratory strains of HIV. Saquinavir has been recently approved by the FDA for the treatment of AIDS in combination with nucleoside inhibitors of reverse transcriptase, and the discovery of this compound was the first breakthrough and the starting point for many other innovative designs. Determination of the crystal structures of HIV PR gave new impetus to the design of novel inhibitors. One measure of the intensity with which new inhibitors were designed or discovered is the total number of crystal structures of inhibitory complexes, currently exceeding 250, that have been determined over the past 5 years. Very detailed crystallographic analysis combined with extensive biochemical characterization and site-specific mutagenesis studies made HIV PR perhaps the best characterized enzyme to date.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_2.html [2/29/2004 2:14:57 AM]

Document

Page 3

Based on the avalanche of papers describing the structure-based design of various HIV PR inhibitors, it would be reasonable to assume that, with the exception of saquinavir, all other HIV PR inhibitors that entered the stage of preclinical or clinical development were discovered using the elements of a structurebased approach. From the long list of more than 30 inhibitors considered as clinical candidates [11], currently there are three compounds (saquinavir, ritonavir, and indinavir) already approved by the FDA as anti-HIV drugs. Many factors that are requisite for in vivo activity in AIDS patients can only be predicted a priori in a very general sense. For instance, erratic oral bioavailability in humans, first-pass metabolism, binding to plasma proteins or tissue distribution may disqualify a perfect in vitro inhibitor of HIV replication and such properties can be very poorly predicted by any process of drug design. A potential answer to these problems is the parallel design of several chemically distinct compounds that may have similar in vitro activity but significantly different in vivo properties. The application of protein structure-based design offers such possibilities and in this text the discovery and optimization of different series of potent inhibitors of HIV PR will be discussed. In order to familiarize the reader with the architecture of HIV PR and the properties of its active site, the first paragraphs are devoted to the detailed description of the x-ray structures of the enzyme followed by several examples of inhibitors in a bound conformation. A. Three-Dimensional Structure of HIV PR Retroviral proteases such as HIV PR were tentatively assigned to the aspartic protease family on the basis of putative active-site sequence homology [12]. Mammalian aspartic proteases are bilobal, singlechain enzymes in which each lobe (or domain) contributes an aspartic acid residue to the active site [13]. The active site itself is formed at the interface on the N- and C-terminal domains and exhibits approximate two-fold symmetry. Since the retroviral proteases are only about one-third the size of the two-domain eukaryotic enzymes, they were hypothesized to function as dimers in which each monomer contributes a single aspartic acid to the active site [14]. Obligate homodimeric proteases, in addition to providing a regulatory mechanism to control activation of the enzyme, represent the most efficient use of genetic information which, in retroviruses, is naturally parsimonious. The crystal structures of HIV PR confirmed the predicted dimeric character of the enzyme [15,16] (Figure 1). In all published crystallographic investigations of the unliganded form of the enzyme, the monomers are related to each other by crystallographic two-fold symmetry and are necessarily identical. The general topology of the HIV PR monomer is similar to that of a single-domain

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_3.html [2/29/2004 2:15:00 AM]

Document

Page 4

Figure 1 Stereo view of the α-carbon backbone of HIV PR dimer. (a) The apoenzyme with flaps in the “open” conformation. (b) Inhibited form of HIV PR with flaps in a “closed” conformation. For clarity, the inhibitor is removed from the active site.

pepsin-like aspartic protease and consists of antiparallel β-strands and a short, two-turn α-helix connected by loops of varying length. The dimer interface is formed by an antiparallel β-sheet comprising two strands from each monomer. The hydrophobic residues from those β-strands and two symmetry-related α-helices form the core of the dimer. The dimer is further stabilized by a net of hydrogen bonds involving the residues around the catalytic aspartic acids. The active site is formed by the dimer interface and is composed of equivalent contributions of residues from each monomer. The substrate-binding cleft is bound on one side by the active site aspartic acid (Asp25/25') and on the other side by a pair of two-fold related, antiparallel β-hairpin structures, commonly referred to as “flaps.”

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_4.html [2/29/2004 2:15:23 AM]

Document

Page 5

The conserved active-site residues (Asp25, Thr26, and Gly27 from both monomers) form a symmetrical and highly hydrogen-bonded arrangement virtually identical to that described for pepsin [17]. The two aspartates are nearly coplanar with the “inner” carboxylate oxygens hydrogen bonded to the amide hydrogens of Gly27/27'. This designation (e.g. Gly 27/27') will be used throughout this text to indicate equivalent residues of the dimer. The two threonines are inaccessible to solvent and are hydrogenbonded to the main-chain amide groups of the other monomer, forming a rigid network called a “fireman's grip” [17]. As in the case of the structures of eukaryotic pepsins, there is electron density for a water molecule bound between the two carboxylates of the active-site aspartates. In the structure of the apo-form of HIV PR, the flaps from both monomers are related by crystallographic two-fold symmetry and can be considered as being in an open conformation. In the structures of related proteases from Rous Sarcoma Virus and HIV-2, the flaps are either crystallographically disordered or in a partly closed conformation [18]. This suggests that, in solution, in the absence of ligands, the flaps are rather flexible and that the stable conformation of the flaps observed in the crystal structure of the apo-enzyme of HIV PR could be considered to result from kinetic trapping during the crystallization process. In the apo-form of HIV PR, the active site residues are located at the bottom of a rather shallow groove. Upon binding an inhibitor, the protease undergoes significant structural changes, particularly apparent in the flap region. As a result, a tunnel-like site is formed, which runs diagonally across the dimer interface. The tunnel has a volume of approximately 1140 Å3 and is 23 Å long. Because of the dimeric nature of HIV PR, the active site has approximate two-fold symmetry with the dyad axis intersecting the plane of the catalytic aspartates. Along the active site tunnel, starting from the central aspartates, there are distinct subsites S1, S2, S3, and S4, and corresponding symmetry related subsites S1', S2' S3', and S4' (Figure 2). It should be noted that in this chapter, the convention of Schechter and Burger [19] will be used to describe enzyme specificity subsites (S1, S1', etc.) and the corresponding side chains of inhibitors (P1, P1', etc.). The boundaries of the subsites are formed by residues from both monomers of HIV PR. All subsites, with the exception of S4/S4', which are exposed to solvent, are bounded by mostly aliphatic side chains and have hydrophobic character. The borders of the S1/S1' subsites are formed by the side chains of Ile23/23', Ile50/50', Ile84/84', Pro81/81', the γ carbon of Thr80/80', carboxylates of the active site Asp25/25', and the carbonyl oxygens of Gly27/27'. The S2/S2' subsites are bounded by Val32/32', Ile50/50', Ile47/47', Leu76/76', Ala28/28', and the carboxylates of Asp30/30'. The S3/S3', subsites are partly exposed to solvent and are bordered by the side chains of Leu23/23', Val82/82', Pro81/81', and the guanidinium groups of Arg8/8', which form a salt bridge with the carboxylates of Asp29/29'. Most of

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_5.html [2/29/2004 2:15:26 AM]

Document

Page 6

Figure 2 Schematic representation of the specificity subsites of the HIV PR active site with bound peptidic inhibitor JG-365. Amino acids forming the boundaries of the particular subsites are shown.

the hydrogen bond donor and acceptor functional groups of the active site are located in an approximate plane that lies along the long axis of the tunnel and is somewhat perpendicular to the plane of the subsites. The hydrogen-bonding functionalities include the carboxylates of the catalytic aspartates, the carbonyl oxygens of Gly27 and Gly48, the amide nitrogens of Asp29' and Gly48, the carboxylate of Asp29', and the dimer symmetry-related groups on the other side of the active site. Additional groups capable of forming hydrogen bonds with ligands are located in the outer part of the S2/S2' subsite and include the amide nitrogens and the carboxylates of Asp30/30'. There are five conserved water molecules in the active site of HIV PR. Four of the waters are symmetrically distributed in the S3/S3' subsites and one, hereafter called Wat301, is located near the two-fold axis of the dimer and, in the presence of most inhibitors, is approximately tetrahedrally coordinated by the hydrogen bonds formed between carbonyl oxygens of the ligand(s) and the amide nitrogens of Ile50/50' of the flaps. In the ligand-bound form of HIV PR, Wat301 is completely inaccessible to solvent, and it has been speculated that its functional substitution could be energetically favorable [18] or at least may lead to discovery of novel nonpeptidic inhibitors [20]. Thus, there are 18 hydrogen bond donors or acceptors in

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_6.html [2/29/2004 2:15:34 AM]

Document

Page 7

the active site of HIV PR-16 that could form hydrogen bonds directly and two in which the interaction is mediated by the conserved Wat301. The total solvent accessible surface area of the eight subsites of the HIV PR active site is approximately 1150 Å2. Because of the large number of groups with hydrogen-bondforming potential, 450 Å2 of the surface has a polar character, and the nonpolar area of the subsites is slightly larger, approximately 700 Å2. B. Structural Flexibility of HIV PR In the process of viral assembly, HIV PR specifically cleaves nine cleavage sites on GAG and GAG-POL polypeptides [21]. Examination of the amino acid composition of the recognized substrate sites (Table 1) indicates their hydrophobic character and significant sequence variability. The loose specificity of HIV PR most likely reflects its functions in a world of reduced complexity within the confines of the budding virion. The length of the viral protein precursors (approximately 1500 amino acids) reduces the number of potential sequences the protease must discriminate from in selecting its nine cleavage sites. Therefore, HIV PR and other retroviral proteases are not enzymes that have evolved to carry out a single reaction at a rapid rate, but rather enzymes with minimum specificity required to cleave the viral precursors in a specific and orderly manner. The loose specificity requirements demonstrated by effective binding and catalytic processing of all nine sequences, albeit at different rates [22], was the Table 1 The Sequences of the Proteolytic Processing Sites of HIV-1 HIV-1 PR Cleavage sites

Scissile bond

P17/P24

V

S

Q

N

Y

P

I

V

Q

N

P24/P2

K

A

R

V

L

A

E

A

M

S

P2/P7

S

A

T

I

M

M

Q

R

G

N

P7/P1

E

R

Q

A

N

F

L

G

K

I

P1/P6

R

P

G

N

F

L

Q

S

R

P

TF/PR

V

S

F

S

F

P

Q

I

T

L

PR/RT

C

T

L

N

F

P

I

S

P

I

RT/RN

G

A

E

T

F

Y

V

D

G

A

RN/IN

I

R

K

V

L

F

L

D

G

I

P5

P4

P3

P2

P1

P1'

P2'

P3'

P4'

P5'

Schechter-Berger notation

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_7.html (1 of 2) [2/29/2004 2:15:46 AM]

Document

TF—transframe, PR—protease, RT—reverse transcriptase, RH—RNAse H, IN—integrase. The location of the processing sites in HIV-1 were determined by protein sequencing of HIV-1 virion proteins.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_7.html (2 of 2) [2/29/2004 2:15:46 AM]

Document

Page 8

first indication that the recognition subsites of the HIV PR can display flexibility upon binding of substrates or inhibitors. Early crystal structures of the HIV PR apo-enzyme and complexes with peptidic inhibitors showed several conformations of the active site forming flaps, which include the residues Met46/46' to Ile54/54' [15,16]. Increased availability of coordinates of HIV PR complexed with various inhibitors and crystallized in different crystallographic space groups allowed for more rigorous examination of domain movements and structural changes in the active site. The alignment of several crystal structures of HIV PR in a common frame of reference, which most commonly includes the region around the symmetryrelated active site triad Asp25/25'-Thr26/26'Gly27/27', will highlight those regions of the backbone where significant displacement occurs upon accommodating the individual inhibitors. Examination of the aligned structures, which included examples of all classes of inhibitors, indicated only small variation of the backbone and limited movements in the two binding loops, comprising residues Leu76-Ile84 from both monomers. The loops form the outer walls of subsites S1/S1' and S3/S3' with inward-facing hydrophobic side chains of isoleucines and valines. The flexibility of these loops, which in some cases can move outward by as much as 2.5 Å, has a significant impact on the volume of the specificity subsites, which in turn can accommodate corresponding P1/P1' and P3/P3' moieties of various sizes. Interestingly, the predominant resistance-causing mutations are located on the same loops and involve changes in residues Val82/82' and Ile84/84' (see below). It should be noted, that while the alignment of several crystal structures provides information about the flexible regions, the extent of flexibility of the residues around the HIV PR active site can be limited by crystal packing forces and may represent a crystallographic artifact. In all characterized crystal forms of HIV PR [23] the loops 76–84 and 46–56 participate in crystal lattice formation and the particular conformation of these loops can be driven by crystallization conditions or interactions with other molecules related by the crystallographic symmetry. C. Inhibitors of HIV PR In general, inhibitors of HIV PR can be divided into three distinct groups. The first group includes peptidic inhibitors that utilize various transition-state dipeptide analogs such as statine, hydroxyethylene, and hydroxyethylamine incorporated into peptidic frameworks of differing lengths. Several crystal structures of this type of inhibitor complexed with HIV PR were solved and the structural information provided a wealth of information as to the minimum size of inhibitors, geometry of hydrogen bonds formed within the active site, and the structural flexibility of the subsites (for reviews see Reference 18 and 23). The second and perhaps largest group of HIV PR inhibitors includes peptidomimetic

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_8.html [2/29/2004 2:15:50 AM]

Document

Page 9

compounds that utilize similar transition-state analogs and retain at least one peptide bond with a side chain corresponding to a naturally occurring amino acid. Several compounds from this group have excellent pharmacokinetic and antiviral properties and, in fact, all three HIV PR inhibitors approved for clinical use (saquinavir, ritonavir, and indinavir) belong to this class of compounds. The last and the smallest group of HIV PR inhibitors has a distinct nonpeptidic character. Compounds from this class were discovered either by screening libraries of existing compounds or by structure-based de novo design. Illustrative examples of inhibitors belonging to all three classes and a brief description of the discovery of selected compounds are presented below. D. Peptidic Inhibitors of HIV PR The concept of peptidic inhibitors of HIV PR can be exemplified by the crystal structure of the statinecontaining peptidic compound AG1002 (Figure 3) [23]. In AG1002, the statine moiety replaces the scissile dipeptide while the flanking amino acids were derived from the natural substrate cleaved by HIV PR. The inhibitor binds to the active site in an extended conformation with the central hydroxyl group of the statine moiety forming hydrogen bonds with the active-site aspartic acids 25/25'. The peptidic backbone and the side chains of the

Figure 3 Stereo view of the peptidic inhibitor AG1002 bound to the active site of HIV PR. The distribution of the specificity subsites S and S' is similar to that shown in Figure 2. The boundaries of the HIV PR active site are indicated by the dotted surface.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_9.html (1 of 2) [2/29/2004 2:16:10 AM]

Document

Page 10

inhibitor form 16 hydrogen bonds and occupy subsites from S4 to S1, S2', and S3'. The carbonyl oxygens of P2 and P1' accept two hydrogen bonds from the flap water Wat301, which in effect is nearly tetrahedrally coordinated. Due to the structural nature of statine, which lacks the P1' side chain, the S1' pocket remains unoccupied. The S1 subsite is only partially filled by the P1 side chain of leucine. The P2 and P2' side chains of asparagine and glutamine form hydrogen bonds with Asp30' and 30, while the aliphatic carbons of both side chains make several hydrophobic contacts in the S2 and S2' pockets respectively. Despite the large number of hydrogen bonds formed within the HIV PR active site, AG1002 has rather low inhibitory potency with a binding constant of 0.55 µM The low binding constant most likely reflects the absence of the P1' group, the free energy required for desolvation of the hydrophilic side chains, and the charged N- and C-termini as well as entropic effects caused by the flexible nature of the heptapeptide. Other interesting examples of peptidic inhibitors are compounds utilizing other transition-state analogs, e.g. reduced amide-containing hexapeptide MVT-101 [24], hydroxyethylene-containing octapeptide U85548e [25], and hydroxyethylamine-containing heptapeptide JG-365 [26]. All these compounds bind to the active site of HIV PR in a similar extended conformation and the small differences in the geometry of hydrogen bonds formed with HIV PR can be attributed to the different character and length of the transition-state analogs. The chemical structures and inhibition constants of these inhibitors are summarized in Table 2. Note that the inhibition constants cited throughout this chapter and in Tables 2, 3, and 6 were determined in different laboratories—often using significantly different assay conditions—and therefore might not be meaningfully comparable. Due to their substantial size and peptidic nature, inhibitors from this class were not suitable for clinical application. Nevertheless, the structural information derived from many crystal structures of peptidic inhibitors bound to the HIV PR active site was critical for subsequent modeling and design of the next generation of peptidomimetic and nonpeptidic inhibitors of HIV PR. E. Peptidomimetic Inhibitors of HIV PR Design and Structure of Ro-31-8959 (Saquinavir) The strategy of designing saquinavir was based on the transition-state mimetic concept, an approach that has been used successfully in the design of potent inhibitors of renin and other aspartic proteases [10]. From the variety of nonscissile transition-state analogs of a dipeptide, the hydroxyethylamine mimetic was selected because it most readily accommodates the amino acid moiety characteristic of the Phe-Pro and Tyr-Pro cleavage sequence of the

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_10.html [2/29/2004 2:16:13 AM]

Document

Page 11

retroviral substrates. In the first step of design, the dipeptide analog consisting of Phe[CH(OH)CH2N]Pro was used to determine the minimum sequence required for potent inhibition. From this study a compound was selected that included benzyloxycarbonyl at the N-terminal side of the inhibitor followed by the P2 asparagine, the hydroxyethylamine isostere with side chains of phenylalanine and proline in the P1 and P1' positions respectively and the NH-t-butyl group at the Cterminal part. In the following design, the side chain of proline was consequently modified to a piperidine and finally to a decahydroisoquino-

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_11.html (1 of 2) [2/29/2004 2:16:45 AM]

Document

Page 12

line moiety, and the N-terminal benzyloxycarbonyl group was replaced by the quinoline-2-carbonyl. The resulting compound, Ro-31-8959, was one of the first peptidomimetic inhibitors with very high antiviral potency and became a benchmark for further design of HIV PR inhibitors [10]. The high-resolution crystal structure of saquinavir bound to the active site of HIV PR was solved in many laboratories [23,27]. The incorporation of decahydroisoquinoline moiety, which can be considered as a conformationally restrained mimic of cyclohexylalanine, has some important consequences. First, the length of the C-terminal part of the inhibitor has been restricted to the P2' residue which, in saquinavir, consists of a NH-t-butyl group. Second, it restrained the conformational freedom of the otherwise peptidic backbone, minimizing the entropic penalty to the free energy of binding. In the crystal structure of saquinavir with HIV PR (Figure 4), the decahydroisoquinoline in the preferred chairchair conformation, makes extended hydrophobic contacts in the S1' subsite. The bond between the methylene carbon and the nitrogen of decahydroisoquinoline is in the low-energy equatorial conformation and the nitrogen, even if protonated, is not in a position to form a hydrogen bond with the active-site residues. The central hydroxyl group is in the R(syn) conformation and is within the hydrogen-bond-forming distance with both carboxylates 12640-0012a.gif Figure 4 Stereo view of the peptidomimetic inhibitor Ro 31-8959 (saquinavir) bound to the active site of HIV PR. The distribution of the specificity subsites S and S' is identical to that shown in Figure 2. Note the stacking interaction between the quinoline moiety and the P1 side chain of phenylalanine.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_12.html [2/29/2004 2:16:50 AM]

Document

Page 13

of the active-site aspartates. Similar to Ag1002 and other peptidic inhibitors, the carbonyl oxygens of the P2 and P1' amides are within hydrogen-bonding distance of the flap water molecule; however, the geometry of the second hydrogen bond is distorted due to the additional spacing between both carbonyl groups. The nitrogen of the t-butylamide is displaced from the normal P2' position by approximately 1.8 Å and, as a result, cannot form a direct hydrogen bond with the carbonyl oxygen of Gly27. Instead the tbutylamide nitrogen interacts via highly ordered water molecules with the amide nitrogen of Asp29 and the carbonyl oxygen of Gly27. The aliphatic t-butyl moiety occupies the S2' subsite and the position of the backbone in this region prohibits any further extension into the S3' pocket. The P1 and P2 side chains of phenylalanine and asparagine, respectively, occupy the corresponding subsites and have a similar conformation to the equivalent groups observed in peptidic inhibitors. In the crystal structure, the N-terminal quinoline-2-carboxylate is moved to the side and, as a result, the carbonyl oxygen forms hydrogen bonds with the ordered water molecule and with the amide nitrogen of Asp29'. The quinoline ring is in a low-energy conformation with respect to the preceding carbonyl oxygen, which places the aromatic nitrogen in unfavorable close contact (3.3 Å) to the carbonyl oxygen of the flap Gly48. Because of the absence of any further contacts with the HIV PR active site residues, the contribution of the quinoline moiety to the free energy of binding remains unclear. Perhaps in solution, a stacking interaction of the P1 phenyl ring and the aromatic quinoline restricts the conformational freedom of Ro31-8959, in effect diminishing the free-energy loss due to the entropic and desolvation effects. Saquinavir, despite its distinct peptidomimetic character is a very potent inhibitor of HIV PR with an inhibition constant of 0.9 nM and an antiviral IC50 in vitro of 0.020 µM [10]. Although it suffers from a low oral bioavailability (5–10% in humans), it became an important starting point for the design of second generation, less-or nonpeptidic inhibitors. Saquinavir became the first HIV PR inhibitor approved by the FDA for treatment of AIDS. Design and Structure of ABT-538 (Ritonavir) An interesting concept for designing specific HIV PR peptidomimetic inhibitors with internal two-fold symmetry was first formulated by John Erickson and his colleagues from Abbott Laboratories [28]. They reasoned that if HIV PR incorporates symmetry into its active site structure, compounds that mimic this symmetry might be novel, more specific, and potent inhibitors and, furthermore, due to the bidirectionality of peptide bonds, might be sufficiently less peptidic in character and pharmacologically superior to the classical peptide-based compounds. The crystal structure of one of the first compounds from this series (A74704) verified the assumption of symmetrical binding conformation in the

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_13.html [4/5/2004 4:44:15 PM]

Document

Page 14

active site of HIV PR. The inhibitor consists of the central diamino alcohol moiety with symmetrically distributed phenylalanine side chains and two flanking, Cbz-blocked, valine residues. Except for the asymmetric hydroxyl group, A74704 binds to the active site in a symmetric mode and the inconsistent distribution of the terminal Cbz groups is most likely caused by crystal lattice contacts and may not reflect the binding mode in solution [28]. The design of symmetrical inhibitors was further extended to include a series of diamino, diol core units, in which the C2 axis bisects the bond connecting the two hydroxy-bearing carbon atoms [29]. Such inhibitors consistently showed greater potency than A74704, but the relative potencies of the diols differed for different diastereomers, and they did not exhibit a uniform dependence on the stereochemistry at the hydroxymethyl position. Surprisingly, high-resolution crystal structures of HIV PR with all possible diol diastereomers, (S,S, R,R and R,S) revealed that most of the inhibitors bind in a clearly asymmetric fashion placing only one of the diol hydroxyl groups on the C2 axis dissecting the active site of HIV PR and the catalytic carboxylates of Asp25/25'. The asymmetric placement of diols causes translation of inhibitors along the long axis of the active site and, as a result, the midpoint of the compounds is displaced by up to 0.9 Å from the two-fold axis of the HIV PR. Nevertheless, the dihedral angles of the symmetry-related bonds are in most cases within 10°, and the inhibitors maintain overall symmetry in the bound conformation [23,29]. The ABT-538 design was a direct consequence of the pioneering work with peptidomimetic compounds with the internal C2 symmetry [30]. Since the high-resolution crystal structures of a family of diolcontaining compounds indicated that in most cases only one of the diol hydroxyls interacts with the catalytic aspartic acids 25/25', in subsequent designs the noninteracting hydroxyl group was replaced by a hydrogen. This substitution reduced the free energy penalty required for desolvation of the hydroxyl group and increased the inhibitory potency without perturbing the binding mode of the compounds [30]. In the further search for related inhibitors with improved oral bioavailability, the focus of effort concentrated on the effect that molecular size, aqueous solubility, and hydrogen-bonding capability has on pharmacokinetic behavior. This study resulted in a smaller compound, A-80987, in which the P2' side chain of valine was eliminated and the terminal 2-pyridinyl group was replaced by 3-pyridinyl moiety that makes van der Waals contacts in the S2' subsite and forms a hydrogen bond with the amide nitrogen of Asp30 [31]. The pharmacokinetic properties of A-80987 were significantly improved over larger, symmetrical compounds from this series and, at the same time, the high antiviral activity typical for these inhibitors was largely unaffected. In subsequent optimization, which focused on the metabolic stability of these inhibitors in vivo, the electron-rich and oxidation-prone pyridinyl groups were replaced by thiazoles.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_14.html [4/5/2004 4:44:17 PM]

Document

Page 15

Thiazoles are less electron-rich isosteres of pyridines and therefore it was speculated that compounds with such substitution may have improved metabolic stability [30]. The modeling of A-82200 in which the N-terminal pyridinyl group was substituted by a 4-thiazolyl moiety indicated that the 5-membered ring binds in the S3 subsite and can be further derivatized at the 2 position by an isopropyl group. The isopropyl functionality makes van der Waals contacts with Val82 and fills the hydrophobic part of the S3 subsite in nearly optimal fashion.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_15.html (1 of 2) [4/5/2004 4:44:40 PM]

Document

Page 16

The resulting compound, ABT-538 (Table 3), binds to the active site of HIV PR in an extended conformation. The central, asymmetric hydroxyl group is within hydrogen-bonding distance of the catalytic aspartates 25/25', and the P1/P1' phenylalanine side chains are symmetrically distributed in the corresponding subsites. The nitrogens of the symmetric amide bonds on both sides of the central aminoalcohol are barely within the hydrogen-bonding distance of the carbonyl oxygens of Gly27/27' (3.4 Å) and the carbonyl oxygens of these amide bonds participate in the tetrahydral coordination of the flap water molecule Wat301. On the N-terminal side of the compound, the P2 side chain of valine fills the S2 subsite and the terminal 2-isopropyl-4-thiazolyl makes hydrohobic contacts with the residues in the S3 pocket and has a stacking interaction with the P1 phenylalanine. On the C-terminal side, the 5thiazole is positioned to interact within the S2' subsite, and the nitrogen on the 5-membered ring is within hydrogen-bonding distance of the amide of Asp30'. Despite two peptide bonds present in ABT-538, this compound has substantial oral availability in humans and a very high antiviral activity in vivo [30]. Recently, ABT-538, better known as ritonavir, has been approved by the FDA for treatment of AIDS in combination with inhibitors of the reverse transcriptase. Design and Structure of L-735,524 (Indinavir) Indinavir is another example of very potent peptidomimetic compound discovered using the elements the crystal structure-based design [32] and SAR (structure activity relationship). The starting point for the design was a series of compounds containing the hydroxyethylene isostere of a scissile dipeptide [33]. An example of compounds from these series is L-685,434, which consists of a tert-butylcarbamate group forming the P2 moiety, symmetrically distributed phenylalanine side chains in the P1/P1', and the indanol group in the P2' portion of the inhibitor. Although very potent, the optimized molecules from this series lacked aqueous solubility and an acceptable pharmacokinetic profile [32]. The Merck group hypothesized that incorporation of a basic amine-containing functionality, such as the decahydroisoquinoline group of saquinavir, into the backbone of L-685,434 series might improve the solubility and bioavailability of this type of compound. Also the replacement of the P2/P1 functionalities, the tert-butylamide and phenylalanine side chain by the decahydroisoquinoline tertbutylamide, would generate a novel class of hydroxylaminepentanamide isostere with potentially improved metabolic stability in vivo. An additional strong argument for using decahydroisoquinoline as an isostere of P1/P2 moieties was the restricted conformational freedom of the enclosed-into-a-ring basic amine, which should decrease the entropy change upon binding to HIV PR in a similar fashion to that observed in saquinavir. In

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_16.html [4/5/2004 4:44:41 PM]

Document

Page 17

the resulting chimeric inhibitor the central hydroxyl group forms hydrogen bonds with the catalytic aspartic acids 25/25' and the hydrophobic side chains of the P1/P1' decahydroisoquinoline and phenylalanine respectively are separated from the central hydroxyl-bearing carbon by the methylene linkers forming a pseudosymmetrical arrangement. In the subsequent optimization of inhibitors from this novel series, a smaller piperazine group was substituted for the decahydroisoquinoline group, which offered a possibility to expand from the N4 position to the partially lipophilic S3 subsite. One of the first compounds from the piperazine series possessed a benzyloxycarbonyl moiety attached to the piperazine ring and the additional hydrophobic interaction in the S3 subsite was reflected by substantial increase in both intrinsic potency and in the ability to inhibit viral spread in infected cells in vitro. Finally, the replacement of the benzyloxycarbonyl group by the 3-pyridylmethyl moiety (Table 3) provided both lipohilicity for binding to the HIV PR active site and a weakly basic nitrogen that increased aqueous solubility and oral bioavailability. The crystal structure of L-735,524 (indinavir) bound to the active site of HIV PR [34] indicates that the 3-pyridylmethyl group attached to the N4 position of the piperazine ring makes hydrophobic contacts with the residues in the S3 and S1 pockets and the tert-butyl moiety fills the S2 subsite in the fashion previously observed in the structure of saquinavir. The positions of the P2 and P1' carbonyls maintain the proper alignment to form hydrogen bonds with the flap water Wat301. The terminal indanol group of indinavir occupies the S2' subsite with the hydroxyl group within hydrogen-bonding distance of the amide nitrogen of Asp29. The high aqueous solubility and largely nonpeptidic character of indinavir may be responsible for the good oral bioavailability, respectable pharmacokinetic profile, and high antiviral activity observed with this compound. Similar to saquinavir and ritonavir, indinavir has been recently approved by the FDA for treatment of AIDS. F. Nonpeptidic Inhibitors of HIV PR The nonpeptidic inhibitors of HIV PR can be divided into two subclasses. Compounds that belong to the first group maintain the general binding mode of the peptidomimetic inhibitors including formation of the key hydrogen bonds with the active site residues. An example of such nonpeptidic inhibitors of HIV PR is AG1343 (nelfinavir). The second group of nonpeptidic HIV PR inhibitors includes compounds with a binding mode significantly different from that described for the peptidomimetic compounds. Most inhibitors in this latter class were initially discovered by screening natural-products libraries or by structurebased de novo design. The most interesting examples of the nonpeptidic inhibitors from this group are the independently discovered but structurally

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_17.html [4/5/2004 4:44:43 PM]

Document

Page 18

related 4-hydroxypyrans and 4-hydroxycoumarins, the cyclic urea-based DMP323 series, and AG1284. Design and Structure of AG1343 (Nelfinavir) Analysis of the crystal structure of saquinavir with HIV PR indicated that while the nonpeptidic components of the ligand, namely the decahydroisoquinoline and the t-butylamide moieties fill the S1' and S2' subsites nearly optimally, the N-terminal portion offered the possibility for remodeling, aimed at the elimination of the peptidic character. Also, the contribution of the quinoline to the binding affinity to HIV PR was difficult to rationalize. Since the removal of quinoline resulted in a nearly 1000-fold loss in binding constant, it was concluded that the stacking interactions of the P1 phenyl ring and the P3 aromatic moiety of quinoline are necessary for the conformational stability of Ro 31-8959. In an attempt to redesign the N-terminal part of the ligand, the nonpeptidic portions of the P1' and P2' were maintained but for reasons of synthetic accessibility, the decahydroisoquinoline moiety was replaced by an orthosubstituted benzylamide [35]. Crystallographic analysis of both compounds showed that saquinovir and the modified LY289612 bind essentially identically to the active site of HIV PR and their inhibition constants and antiviral activity were very similar (Table 4 and Table 3). In the first attempt to functionally substitute the P2 side chain of asparagine, the isophthalic-acidcontaining compound was modeled and the low-energy conformation of the aromatic ring, required for binding in the S2 subsite, was stabilized by a tertiary carboxamide in the P3 region of the inhibitor [36]. The analysis of the binding mode and interactions of the isophthalic ring in the S2 subsite indicated a lipophilic pocket deep on the border between the S2 and S1' subsites, which could be conveniently filled with a methyl group extending from the 2 position of the ring. The resulting compound II in Table 4 lost most of the peptidomimetic character of LY289612 but retained its inhibitory potency. In an independent line of design, the relationship between the P1 phenylalanine side chain and the P3 quinoline was investigated. In the crystal structure of saquinavir bound to the active site of HIV PR (Figure 4), the aromatic ring of the P1 phenylalanine makes several van der Waals contacts with residues forming the S1 subsite. Computer modeling indicated that an extension of the phenylalanine side chain to phenethyl (homophenylalanine) would lead to prohibitive close contacts of the phenyl ring with the aliphatic side chains of HIV PR. On the other hand, replacement of the γ-carbon of the homophenylalanine by sulphur, which has a more acute C-S-C bond angle, would direct the aromatic ring into the neighbouring S3 subsite without changing the desired lipophilic nature of the P1 side chain. The increased area of hydrophobic interactions in the S1 and S3 subsites by compounds with the Sphenylcysteine and

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_18.html [4/5/2004 4:44:45 PM]

Document

Page 19

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_19.html (1 of 2) [4/5/2004 4:44:51 PM]

Document

Page 20

S-naphthylcysteine derived side chains in P1 resulted in a substantial increase in the inhibition constants [37]. The increase in the binding affinity to the low picomolar range in enzyme inhibition assay, allowed for subsequent truncation of the P3 quinoline moiety. The final compound from this miniseries (compound III in Table 4) consisted of the ortho-substituted benzamide in the P1' and P2', Snaphthylcysteine in P1 and asparagine in P2. Despite reduced molecular weight, the inhibition constant of this compound for HIV PR was comparable to LY289612. The observation that a larger, nonpeptidic moiety in the P1 could eliminate the need for the P3 side chain led to hybrid molecules that incorporated ring structures as the P2 component and maintained the P1 S-naphthylcysteine side chain of compound III. In this miniseries several bicyclic functionalities were modeled as the P2 substituents and one example, compound IV utilizing a tetrahydroquinoline group, is shown in Table 4 [38]. In subsequent modeling, it was noticed that the P2 bicyclic functionality might be replaced by 2,3-disubstituted phenyl rings. In particular, a methyl substitution in position 2 would increase the area of hydrophobic interaction in a manner previously observed in the isophthal series. Addition of a hydrophilic functionality attached at position 3 could increase the solubility of the compound and contribute to the binding constant by forming a hydrogen bond with the carboxylate oxygen of Asp30. A compound with a 2-methyl-3-hydroxy substitution pattern was synthesized and showed an improved inhibition constant of 3 nM in the HIV PR enzyme assay (Table 4). The crystal structure of compound V with HIV PR was solved and indicated the predicted binding mode with the possibility of a stacking interaction between the P2 phenyl and the P1 thio-naphthyl groups and the expected hydrophobic and hydrogen-bonding interactions of the P2 moiety with the protein side chains in the corresponding specificity pocket [38]. As with the optimized compounds from other series, compound V suffered from low aqueous solubility. The replacement of the P1' aryl group by the basic amine-containing decahydroisoquinoline dramatically increased the solubility and allowed for truncation of the P1 S-naphthylcysteine side chain to S-phenylcysteine without any loss of inhibitory activity. The resulting compound VI, AG1343 or nelfinavir, has an inhibition constant of 1.9 nM in the HIV PR enzyme assay and respectable antiviral activity with an IC90 of 60 nM [39]. The nonpeptidic character, pH-dependent solubility profile, and the small molecular weight of nelfinavir may contribute to its good pharmacokinetic profile in humans [40,41]. Currently, this compound is being tested and is in the advanced phase of clinical trials. The crystal structure of nelfinavir bound to the active site of HIV PR is shown in Figure 5. The general binding mode of this compound, in particular the path of the backbone, is similar to the binding mode of peptidyl inhibitors. Nevertheless, the lack of any peptide bonds utilizing naturally occurring amino

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_20.html [4/5/2004 4:44:58 PM]

Document

Page 21

Figure 5 Stereo view AG1343 (nelfinavir) bound to the active site of HIV PR.

acids qualifies nelfinavir to be a member of the group of nonpeptidic inhibitors of HIV PR. The unique, and perhaps crucial hydrogen-bonding interaction of the P2 hydroxyl group with the carboxylate oxygen of Asp30, combined with the smaller area of hydrophobic contacts in the S1 and S3 specificity subsites are the principal differences from other clinically active HIV PR inhibitors and may contribute to a distinct resistance pattern and point to additional utility of nelfinavir in the treatment of AIDS. Design and Structure of DMP323 A cyclic urea-containing HIV PR inhibitor, DMP323, was discovered using de novo structure-based design principles. Similar to the concept of Erickson and his co-workers from Abbott Laboratories, the group from DuPont-Merck attempted to take advantage of the two-fold symmetry of HIV PR in designing compounds that maintained the interaction of the diol with the catalytic aspartic acids 25/25' and at the same time were able to functionally displace the ubiquitous flap water molecule Wat301. They hypothesized that incorporation of the binding features of this structural water molecule into an inhibitor would be beneficial because of the entropic gain due to its displacement and because the conversion of a flexible linear inhibitor into a rigid, cyclic structure with restricted conformation should provide an additional, positive entropic effect. In the initial design, a cyclohexanone with the ketone oxygen as the structural

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_21.html [4/5/2004 4:45:10 PM]

Document

Page 22

water mimic was used and in subsequent synthetic targets the cyclohexanone ring was enlarged to a 7membered ring to incorporate a diol functionality. This target was further modified to a cyclic urea, which can be symmetrically substituted from both nitrogens without creating unnecessary stereocenters. The crystal structures of about 10 cyclic-urea-based inhibitors with HIV PR were solved [42]. In all cases, the C2 symmetric inhibitors bind to the HIV PR active site with the diad symmetry axes of the protease and the compounds being nearly coincident. The 7-membered ring of the inhibitors is roughly perpendicular to the plane of the catalytic aspartates 25/25' and both hydroxyl groups of the diol are positioned to interact with their carboxylates. The carbonyl oxygen of the inhibitors accepts hydrogen bonds from backbone amides of symmetrically distributed residues Ile50/50' of the flap. In the structure of DMP323, symmetrically substituted moieties of hydroxymethylbenzyls and phenylalanines extend towards the S2/S2' and S1/S1' subsites respectively and are involved in van der Waals interactions with the hydrophobic residues of these pockets [42]. The interaction of DMP323 with the residues of HIV PR are restricted to the central four specificity subsites of the active site. Despite this limited area of hydrophobic interaction and hydrogen bonding restricted to the central cyclic urea functionality, DMP323 is a very potent inhibitor of HIV PR with good antiviral activity in vitro (Table 5). The limited solubility of this compound was perhaps responsible for erratic oral availability in humans, and after short trials, DMP323 was withdrawn from the clinical investigation. Nevertheless, the discovery of this class of compounds represents a very interesting and, by now, classical example of de novo structure-based drug design. Design and Structure of AG1284 Another compound discovered by the application of de novo structure-based design is AG1284 [43]. In the initial design of a lead compound, the nonpeptidic hydroxyethyl-t-butylbenzylamide portion of LY289612 occupying the S1' and S2' subsites was retained as a “starting module.” In attempting to fill the pockets related by the dimer two-fold symmetry it was discovered that, by extending a two-carbon fragment from the central hydroxyl carbon, the S1 subsite could be accessed by an aromatic ring. The ring was oriented orthogonal to the observed P1 phenyl group of the classical inhibitors and this allowed further extension off the ortho position towards the S2 subsite. In order to maintain the critical hydrogen bond to the flap water Wat301, in the initial compounds an acylated amino group was used, replaced in subsequent designs by a benzamide functionality. In this model, the geometry of the hydrogen bonds to the flap water was somewhat perturbed, and the nitrogens of the t-butyl amides on both sides of the compound were in a position to interact favorably with solvent,

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_22.html [4/5/2004 4:45:29 PM]

Document

Page 23

potentially lowering the desolvation penalty. The absence of hydrogen-bonding interactions with the carbonyl oxygens of Gly27/27' was viewed as a positive factor, since the accumulated structural and mechanistic information suggested that formation of these hydrogen bonds may not be energetically favorable [23]. The compound was synthesized as a racemic mixture of two enantiomers of the central hydroxymethyl group and had the inhibition constant of 24 µM. Despite the modest binding constant of compound II (Table 6) and very low water solubility, the co-crystal structure with HIV PR was solved at 2.3 Å resolution, providing a critical starting point for further design. The inhibitor was found to bind largely as anticipated with the two aromatic rings occupying the S1 and S1' subsites and the two benzamide carbonyls forming hydrogen bonds to the flap water Wat301. Both benzamide nitrogens interact via a string of highly ordered water molecules with the amide nitrogens of Asp29/29'. The crystal structure of the complex indicated that the S enantiomer was the more active component of the racemic mixture and this was confirmed by stereoselective synthesis of subsequent compounds [44]. In the subsequent designs, the ortho-substituted benzyl rings were consecutively replaced by larger naphthyl groups that occupied more of the S1–S3 and S1'–S3' subsites. The increased area of hydrophobic interactions with the residues in these subsites was reflected in a substantial improvement in the

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_23.html [4/5/2004 4:45:41 PM]

Document

Page 24

binding constant and also in reduced aqueous solubility. Also, due to the very tight fit of both naphthyl moieties in the S1 and S1' subsites, subsequent design targeting the S3/S3' subsites proved to be difficult and synthetically challenging [44]. In the search for a simpler solution, the di-tertiary amides were designed using the crystal structure of compound II (Table 6) as a starting model. Branching from the amide nitrogens provided an interesting possibility to access S2–S3/S2'–S3' subsites while simultaneously increasing the solubility and stability of the compounds. In the first design, the hydroxyethyl moieties were fused to the amide nitrogens and the hydroxyl groups were intended to form hydrogen bonds with the amide nitrogens of Asp29/29' (compound III in Table 6). The addition of both hydroxyethyl groups resulted in a rather significant increase in the binding constant, and the racemic mixture had the Ki of 1.1µM. When the crystal structure of compound III complexed with HIV PR was solved at 2.2 Å resolution, it was observed that the inhibitor had undergone an inversion in binding mode relative to the secondary amide series. The phenyl groups of compound III occupied the S2/S2' subsites, switching positions with the t-butyl groups, which were in turn occupying the S1/S1' pockets (Figure 6). Due to this change in binding mode, the R enantiomer would be expected to be preferred relative to S. The final position of the hydroxyethyl moieties was less effected by the change, and both hydroxyls were within hydrogen-bonding distance from the amide nitrogens of Asp 29/29'. In the S2/S2' pockets, the phenyl groups occupied only a fraction of subsites, but the interaction was strengthened by highly ordered water molecules involved in electrostatic interaction with the aromatic rings and by forming hydrogen bonds to Asp30/30'. Interestingly the position of the hydrogen bonds with respect to the flap water was significantly disturbed in the new binding mode, and the conserved Wat301 was no longer tetrahydrally coordinated [43,45]. This unanticipated change in binding mode presented a potential for new avenues of design different from those of the secondary amides. The ability to design into neighboring subsites depends to a large extent on the positions of bond vectors suitable for substitution in the bound conformation of a given inhibitor. These vectors in the crystallographically discovered new binding mode of compound III were positioned ideally to access unfilled space in the S3/S3' pockets. The discovery of this new conformation of compound III highlighted the power of crystallographic feedback in the process of inhibitor design and, without this structural information, further design in this series would have been severely impeded. Inspection of the crystal structure of compound III bound to the active site of HIV PR revealed lipophilic cavities extending off the S1/S1' subsites adjacent to the t-butyl groups of the benzamidine moiety. The cavities are bordered by flexible loops around Pro81/81' and previous crystallographic studies indicated that both loops can move back by up to 2.5 Å, extending the size and

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_24.html [4/5/2004 4:45:43 PM]

Document

Page 25

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_25.html (1 of 2) [4/5/2004 4:46:06 PM]

Document

Page 26

Figure 6 Change of the binding mode of compound III observed during iterative design of AG1284. (a) Crystallographically determined binding mode of compound II. Pseudosymmetrically distributed aryl groups are bound in the S1 and S1' specificity subsites. (b) Crystallographically determined binding mode of compound III. Note the inversion of the binding mode. The ortho-substituted benzyl groups bind in a pseudosymmetric fashion in the S2 and S2' subsites.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_26.html (1 of 2) [4/5/2004 4:46:26 PM]

Document

Page 27

volume of the active site. With this in mind, the dimethylbenzyl group was attached to compound III and the additional phenyl ring was accommodated well in the lipophilic pocket of the S1'/S3' sides. As the S1 pocket was not fully occupied, a Monte Carlo-based De Novo Ligand-Generating program (MCDNLG) [46] was used to identify other amide substituents that might fill this subsite more effectively. From several moieties identified by the MCDNLG program, a larger cyclopentylethyl group showing very good shape complementarity to the S1/S3 subsite was selected for synthesis. In addition, due to the asymmetrical nature of this compound, additional space was identified at the bottom of the S2' pocket that was conveniently filled with either a methyl or a chlorine group on the 5 position of the benzamidine ring. The inhibition constant of the resulting compound (compound IV in the Table 6) was 0.008 µM, which represents approximately a 2500-fold improvement over the first compound from this series. The crystal structure of compound IV or AG1284 complexed with HIV-1 PR was solved, revealing excellent complementarity between the ligand and protein. The ligand forms only 4 hydrogen bonds with either protein functional groups or ordered water molecules, in contrast to the nine hydrogen bonds formed by peptidomimetic LY289612, despite their similar binding affinities. The nonpeptidic character of AG1284 may have contributed to good oral bioavailability and pharmacokinetics in three animal species [43]. Despite very good inhibitory potency on the enzyme level, AG1284 has rather modest antiviral activity in vitro (Table 6). The reason for this discrepancy is unclear but could be related to the low water solubility and higher affinity for membranes, which may effect cell partitioning. A similar lack of correlation between the potency of enzyme inhibition and antiviral activity has been previously observed with other HIV PR inhibitors [11]. Hydroxypyrans and Hydroxycoumarins The lead compounds for the 4-hydroxypyran and 4-hydroxycoumarin series were discovered in biological screens as low potency inhibitors of HIV PR [47–49]. Successful structure solution of both lead compounds with HIV PR enabled rapid optimization of their enzyme inhibitory potencies and antiHIV activities, and one of these compounds, U96988, has already entered Phase I clinical testing [49,50]. The binding mode of this type of inhibitor differs substantially from the classical peptidomimetic compounds and is somewhat similar to de novo-designed compounds from the cyclic urea series. In the case of 4-hydroxycoumarin, the two oxygen atoms of the lactone functionality are positioned within hydrogen-bonding distance of the two NH amides of Ile50/50' on the flap, replacing the ubiquitous water molecule Wat301. The 4-hydroxyl group (Table 5) is located within hydrogenbonding distance of the two catalytic

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_27.html [4/5/2004 4:46:27 PM]

Document

Page 28

aspartic acid residues Asp25/25' and this hydrogen-bonding network of the 4-hydroxycoumarin defines the essential pharmacophore of this new class of inhibitors. In the structure of U96988, this pharmacophore is pseudosymmetrically subsituted by an ethyl and a phenyl group at the C-3a and an ethyl and a benzyl group at the C-6a positions. These four substituents extend into the central core of S2/S2' subsites, where they make van der Waals contacts with the hydrophobic residues of the active site [49]. With a molecular weight of 362 U96988 is the smallest inhibitor of HIV PR in clinical testing. It suffers from rather low antiviral activity (ED90 of ~ 10 µM)but can be considered as the first in a series of this promising class of nonpeptidic HIV PR inhibitors. II. Structural Basis of Resistance of HIV PR Inhibitors The dimeric character and the two-fold symmetry of the active site, in which the monomers contribute equivalent residues to symmetrically distributed specificity subsites, led to early speculations that HIV PR may be less susceptible to resistance than, for example, reverse transcriptase. In the case of retroviral proteases, a single base mutation in the viral genome corresponds to two changes in the threedimensional structure and two structurally identical changes in the active site could result in an enzyme with a drastically modified specificity profile and impaired catalytic activity. Identification of HIV PR variants in cell-culture experiments clearly indicated, however, that this class of drugs is not immune to the challenge of viral resistance. It should be stressed that HIV, unlike other human viruses, is characterized by a dynamic viral turnover in the steady state [51,52]. The rapid replication rate coupled with the lengthy duration of infection will favor the emergence of resistant mutants to targeted antiviral agents [53]. The accumulated data from cell-culture sequential-passage experiments with several structurally different inhibitors and from the resistant variants identified during clinical exposure to four HIV PRtargeting drugs indicate a very complex pattern of mutations in the structure of HIV PR. In contrast to mutations in the reverse transcriptase, which frequently cause multihundredfold resistance [54], single base changes in the HIV PR gene (i.e., two identical substitutions per protease dimer) lead in most cases to 5–10-fold decrease in the antiviral potency of a given drug [11]. It has been shown for the most clinically studied HIV PR inhibitors, such as indinavir and ritonavir, that the clinical manifestation of resistance (increase in the viral load and decrease in the CD4 count) requires the simultaneous appearance of several mutations [55,56]. For example the resistant HIV strain isolated from patients exposed for 40 weeks to indinavir carried mutations at residues 10/10'L > R, 46/46'M > I, 63/63'L > P, 82/82'V > T, and 84/84'I > V [59,60]. However, the combination of these five

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_28.html [4/5/2004 4:46:29 PM]

Document

Page 29

Figure 7 Cartoon representation of the HIV PR dimer. The sites of primary resistance-causing mutations in the active site are indicated. For clarity, the names of the residues are shown for one monomer only.

mutations (ten assuming the dimeric nature of HIV PR) changed the susceptibility of the resistant strain to indinavir by only eight-fold if compared to the wild type HIV. The resistance-causing mutations are localized in a few “hot spots” in the structure of HIV PR and can be divided into two groups. The first group consists of the primary mutations located directly in the active site and includes changes at residues Val82/82', Ile84/84', somewhat less frequently at Gly48/48' and, in the case of nelfinavir, Asp30/30' (Figure 7). Residues 82/82' and 84/84' are located on the flexible loops that form the outer walls of the S3/S3' and S1/S1' subsites, respectively. In the resistant variants, valine 82/82' is most frequently substituted by the smaller side chain of alanine or the larger side chains of phenylalanine or isoleucine [57,58]. The change in position 82/82' is usually accompanied by a substitution of Ile84/84', most commonly to the smaller amino acids alanine or valine [57]. From the clinically tested compounds, ritonavir and indinavir, which were optimized to form strong hydrophobic interactions with the side chain of Val82/82' in the S3 subsite, suffer most significantly from any change at this position. On the other hand, the antiviral activities of saquinavir, and nelfinavir, which do not form any interaction in the S3/S3' subsite are not affected by mutations at Val82/82' and are only marginally cross resistant to changes involving Ile84/84' [57,58,62]. The resistance-causing

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_29.html [4/5/2004 4:46:39 PM]

Document

Page 30

mutation of Asp30/30' to asparagine seems to be specific for nelfinavir and was initially observed in cellculture sequential passage experiments [62]. Recently, the same phenotypic change was confirmed as the predominant mutation in the resistant variants appearing in AIDS patients exposed to low doses of this HIV PR inhibitor [64]. The molecular basis of resistance involving this mutation can be rationalized as follows: in the crystal structure of nelfinavir with the wild type HIV PR, the 3-hydroxyl group of the 2,3-substituted phenyl group is within hydrogen-bonding distance of the carboxylate oxygen of Asp30 in the S2 subsite. Due to the expected coulombic character of this interaction, the hydrogen bond formed with the negatively charged carboxylate of Asp30 would be expected to be a relatively strong one. The change of the negatively charged carboxylate of Asp30/30' to the amide oxygen of the asparagine side chain should reduced the strength of this interaction. Apparently the loss in the enthalpic contribution to the free energy of binding is only partially balanced by the entropic gain caused by the difference in desolvation of a charged vs. neutral side chain of the receptor, leading to decreased binding affinity of nelfinavir and eventually to viral resistance. An additional resistance-causing mutation that qualifies as a primary mutation involves the change of Gly48/48' to valine. This particular mutation seems to be specific for saquinavir and was observed both in cell-culture sequential passage experiments and in AIDS patients exposed to this inhibitor [61,65]. Located on the lower strands of the active-site forming flaps, Gly48/48' can be considered a part of the S4/S4' subsites. The replacement of the glycine hydrogen by the rigid side chain of valine has most likely a dual effect: first it has a direct impact on the interaction of the quinoline moiety of saquinavir with the active site of HIV PR, and second it may change the mobility of the flaps, which in turn will effect the binding kinetics of the natural substrates or inhibitors. Although none of the other clinically tested inhibitors form any interaction with this part of the flap, the HIV variants with mutation of Gly48/48' seem to be cross-resistant to all compounds, which is reflected by a 3–5-fold reduction of their antiviral activity [55,62]. While the effect of primary mutations on reduced binding affinities of inhibitors can be at least partially explained in view of the accumulated structural data, the function of secondary, or compensatory mutations in the resistant HIV PR is difficult to rationalize as yet. The predominant compensatory mutations observed in the resistant variants involve residues Leu63/63', Ala71/71', Met46/46', Asn88/88', Leu10/10', and Leu90/90' (Figure 8) [60,63]. Changes of these residues alone do not confer viral resistance, but their appearance increases the viability of the virus carrying the primary mutations in the active site of protease. All these residues are located far away from the active site of HIV PR do not participate in any apparent way in the inhibitor binding and it seems unlikely that they form a longrange interaction with the natural

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_30.html [4/5/2004 4:46:41 PM]

Document

Page 31

Figure 8 Cartoon representation of the HIV PR dimer. The sites of compensatory mutations are indicated.

substrate. Also, the reported compensatory mutations are conservative in nature and have no effect on the overall distribution of atomic charges on the surface of HIV PR. Sequence polymorphism at the Leu63/63' position, located on the surface at the base of HIV PR, has been observed in clinical isolates of the virus not exposed to any HIV PR inhibitors. Variations of Ala71/71', where the side chains are buried very close to Leu63/63', are less commonly found in clinical isolates. After a prolonged challenge by HIV PR inhibitors, Leu63/63' changes to proline and Ala71/71' to valine. The side chains of Met46/46' are fully exposed to solvent and these residues are located on the βhairpins that form the active side flaps. It has been speculated that the compensatory change of Met46/46' to isoleucine or phenylalanine may affect the dynamics of the flap movement, which in turn could influence the rates of catalytic activity of HIV PR impeded by the primary mutations in the active site [58]. Any changes to Asn88/88' and Leu90/90', buried in the body of HIV PR, most likely affect the structural stability of the enzyme. The side chains of Asn88/88' form buried hydrogen bonds and replacement of this residue by aspartic acid or serine not only eliminates some of these bonds but also introduces unfavorable interactions in the core of HIV PR. Similarly, Leu90/90' is buried in a tight hydrophobic space close to the “fireman's grip” motif, which

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_31.html (1 of 2) [4/5/2004 4:46:51 PM]

Document

Page 32

involves the catalytic asparates 25/25'. The structural effect of a mutation of Leu90/90' to the larger methionine is rather difficult to predict since it can either rigidify or destabilize the HIV PR dimer or it may have an effect on the catalytic efficiency of the “resistant” enzyme. The complicated pattern of HIV resistance to protease inhibitors, in particular the appearance of compensatory mutations that alone do not confer any resistance, suggests that the key to understanding the basis of decreased susceptibility of the virus to a given drug is the kinetics of specific processing of the GAG and GAG-POL polyproteins. The reduction in sensitivity of a mutant HIV PR towards any inhibitor can be conveniently reflected by the ratio of Ki mutant/Ki wild type. However, this reduced inhibitor sensitivity is only one component that distinguishes mutant-form from wild-type proteases. For virus encoding of a mutant HIV PR to be viable, the mutant protease must be capable of a minimal (although not yet quantified) level of enzymatic activity towards all substrates required for maturation of the virions. This proteolytic efficiency is reflected in the specificity constants (Kcat/Km) as determined for mutant and wild type HIV PRs. In order to rationalise these potentially conflicting relationships between enzymes, substrates, and inhibitors, Gulnik and his colleagues [66] introduced the term “Vitality Factor,” in which

In order for the “Vitality Factor” to be predictive for the level of resistance expected for a particular drug or combination of drugs for a given resistant strain of HIV, the determination of the specificity constants (Kcat/Km for mutants) must be repeated for all nine known substrates processed by HIV PR. The inhibition constants of a given compound should not depend on the substrate, but the Kcat/Km ratios do and therefore vitality values will differ for different substrates. It will be expected that the mean for all nine “vitality” values will be predictive for the change in antiviral activity for a particular compound. Although those data will be derived from in vitro experiments and are clearly not without some limitations, they may help in understanding the molecular basis of resistance and may contribute some value to possible multidrug strategies for the clinical management of AIDS. III. Perspective HIV PR inhibitors with acceptable oral availability and pharmacokinetic properties offer great promise for the treatment of HIV infection and AIDS. Efficacy studies of indinavir, ritonavir, or nelfinavir using plasma viral RNA as a marker have demonstrated up to three log reductions in RNA copy numbers that are

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_32.html [4/5/2004 4:46:54 PM]

Document

Page 33

sustained in many patients [67–69]. In contrast, nucleoside antiretroviral therapy that targets reverse transcriptase rarely results in more than one log reduction of viral RNA, indicating fairly poor inhibition of viral replication by this class of compounds. One of the reasons for the apparent greater in vivo antiviral activity of HIV PR inhibitors could be the mechanistic difference of the two enzymes and their respective activities in the viral life cycle. However, growing evidence of retroviral resistance to protease inhibitors remains a concern. The availability of several chemically distinct HIV PR inhibitors, including the second generation of compounds currently under preclinical development, offers a possibility of combining two or more drugs that share little cross-resistance. Also, it seems reasonable to evaluate these compounds in combination with various nucleoside and nonnucleoside reverse transcriptase inhibitors. Early clinical data from such combination therapy indicates reduction of retroviral RNA in plasma to levels lower than the currently available limit of detection [70]. This is the first indication that the application of well-chosen combination therapy can place AIDS patients in prolonged virologic and clinical remission. Undoubtedly, protein crystallography and other elements of structure-based drug design were widely applied in the discovery of HIV PR inhibitors. It will be prudent to assume that, in the absence of structural feedback, rapid discovery of several chemically different and potent inhibitors of HIV PR would have been severely impeded if not even impossible. However, structure-based drug design still remains a new and developing technology. Further success of this drug discovery technique largely depends on the development of methods of computational chemistry. Several computational approaches such as ALADDIN [71], DOCK [72], and MCDNLG [46] have been applied with a limited degree of success in a search for novel inhibitors of HIV PR and these methods will be developed further. The most difficult and challenging computational task required for full implementation of structure-based drug design involves assigning a priority to designed compounds before their synthesis, by computation of the absolute free energy of binding or by prediction of the relative difference in the binding constants of chemically related compounds. While the former approach is technically very difficult, due to the size of configurational space that must be sampled and the limited accuracy of the force field that describes atomic interactions in the molecular system [73], the latter approach has had some successes [74,75]. Nevertheless, owing to the various assumptions and approximations that underlie these techniques, such methods are useful only as order-of-magnitude estimates [74]. Further improvements of these methods heavily depend on the availability of structural and thermodynamic data for several closely related compounds that could be used to calculate parameters required for the implementation of such thermodynamic-integration cycles. The large number of high-resolution crystal structures of HIV PR

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_33.html [4/5/2004 4:46:56 PM]

Document

Page 34

complexed with various inhibitors offers a unique opportunity for the development of such computational methods if the structural data can be coupled with thermodynamic measurements of inhibitor-protein binding. These include direct measurements by microcalorimetry of the association constant, K, and in addition the enthalpy, entropy, heat capacity, and stoichiometry of binding. The combination of such thermodynamic and structural data will lead to a more precise understanding of the factors that influence binding and, ultimately, will lead to new general design principles that can be applied to drug discovery in the area of AIDS as well as other challenging diseases. Acknowledgments I wish to thank all my co-workers from Agouron who contributed to these studies, in particular J. Davies, S. Reich, M. Melnick, V. Kalish, A. Patick, L. Musick, and B-W. Wu. Steven Kaldor from Ely Lilly is acknowledged for his contribution in designing AG1343 (nelfinavir). I would like to thank Richard Ogden for critical reading of the manuscript and D. Olson for expert assistance in preparing the manuscript. References 1. Mitsuya H, Yarchoan R Broder S. Molecular Targets for AIDS therapy. Science 1990; 249:1533–1543. 2. DeClerq E. Toward improved anti-HIV chemotherapy: Therapeutic intervention with HIV infections. J. Med. Chem. 1995; 38:2491–2517. 3. Tomaselli AG, Howe JW, Sawyer TK, Wlodawer A, Henrikson RL. HIV-1 protease as a target for drug design. Chimica Oggi 1991; 9:6–14. 4. Ding J, Das K, Yadav PNS, Hsiou Y, Zhang W, Hughes SH, Arnold E. Structural studies of HIV-1 reverse transcriptase and implications for drug design. In: Structure-Based Drug Design. New York: Marcel Dekker 1996, in press. 5. Perno CF, Bergamini A, Pesce CD. Inhibition of the protease of human immunodeficiency virus blocks replication and infectivity of the virus in chronically infected macrophages. J. Infect. Dis. 1993; 168:1148–1156. 6. Kohl NE, Emini EA, Schleif WA, Davis LJ, Heimbach JC, Dixon RAF, Scolnick EM, Sigal, IS. Active human immunodeficiency virus protease is required forviral infectivity. Proc. Natl. Acad. Sci. USA, 1988; 85:4186–4690. 7. McQuade TK, Tomasselli AG, Liu L, Karacostas V, Moss B, Sawyer TK, Heinrikson RL, Tarpley WG. A synthetic HIV-1 protease inhibitor with antiviral activity arrests HIV-like particle maturation. Science 1990; 247:454–456.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_34.html (1 of 2) [4/5/2004 4:46:59 PM]

Document

8. Kaplan AH, Zack JA, Knigge M, Paul DA, Kempf DJ, Norbeck DW, Swanstrom R. Partial inhibition of the human immunodeficiency virus type 1 protease results in aberrant virus assembly and the formation of noninfectious particles. J. Virol. 1993; 67:4050–4055.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_34.html (2 of 2) [4/5/2004 4:46:59 PM]

Document

Page 35

9. Ratner L. HIV life cycle and genetic approaches. Perspectives in drug discovery and design. 1993; 1:3–22. 10. Roberts NA, Martin JA, Kinchington D, Broadhurst AV, Craig JC, Duncan IB, Galpin SA, Handa BJ, Kay J, Krohn A, Lambert RW, Merrett JH, Mills JS, Parkes KEB, Redshaw S, Ritchie AJ, Taylor DL, Thomas GJ, Machin PJ. Rational design of peptide-based HIV proteinase inhibitors. 1990; 248:358–361. 11. Winslow DL, Otto MJ. HIV protease inhibitors. Current Science 1995; 9:183–192. 12. Toh H, Ono M, Saigo K, Miyata T. Retroviral protease-like sequence in the yeast transposon TY1. Nature 1985; 315:691–693. 13. Davies DR. The structure and function of the aspartic proteinases. Annu. Rev. Biophys. Biophys. Chem. 1990; 19:189–215. 14. Pearl LN, Taylor WR. A structural model for the retroviral protease. Nature 1987; 329:351–354. 15. Navia MA, Fitzgerald PMD, McKeever BM, Leu C-T, Heimbach JC, Herber WK, Sigal IS, Darke PL, Springer JP. Three-dimensional structure of aspartyl protease from human immunodeficiency virus HIV-1. Nature 1989; 333:615–620. 16. Wlodawer A, Miller M, Jaskolski M, Sathyanarayana BK, Baldwin E, Webber IT, Selk LM, Clawson L, Schneider J, Kent SBH. Conserved folding in retroviral proteases: Crystal structure of a synthetic HIV-1 protease. Science 1989; 245:616–621. 17. James MNG, Sielecki A. Structure and refinement of penicillopepsin at 1.8 Å resolution. J. Mol. Biol. 1983; 163:299–361. 18. Wlodawer A, Erickson JW. Structure-based inhibitors of HIV-1 protease. Annu. Rev. Biochem. 1993; 62:543–585. 19. Schechter I, Burger A. On the size of the active site in proteases. Biochem. Biophys. Res. Commun. 1967; 27:157–162. 20. Lam PYS, Jadhav PK, Eyermann CJ, Hodge CN, Ru Y, Bacheler LT, Meek JL, Otto MJ, Rayner MM, Wong N, Chang CH, Webber PC, Jackson DA, Sharpe TR, Erickson-Viitanen S. Rational design of potent bioavailable, nonpeptide cyclic ureas as HIV protease inhibitors. 1994; 263:380–384. 21. Pettit SC, Michael SF, Swanstrom R. The specificity of the HIV-1 protease. Perspectives in Drug Discovery and Design. 1993; 1:69–83.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_35.html (1 of 2) [4/5/2004 4:47:02 PM]

Document

22. Krausslich HG, Ingraham RH, Skoog MT, Wimmer E, Pallai PV, Carter CA. Activity of purified biosynthetic proteinase of human immunodeficiency virus on natural substrates and synthetic peptides. Proc. Natl. Acad. Sci. USA. 1989; 86:807–811. 23. Appelt K. Crystal structures of HIV-1 protease—inhibitor complexes. Perspectives in Drug Discovery and Design. 1993; 1:23–48. 24. Miller M, Schneider J, Sathanarayana BK, Toth MV, Marshall GR, Clawson L, Selk L, Kent SBH, Wlodawer A. Structure of complex of synthetic HIV-1 protease with a substrate-based inhibitor at 2.3 A resolution. Science. 1989; 246:1149–1151. 25. Jaskolski M, Tomasselli AG, Sawyer TK, Staples RL, Heinrikson RL, Schneider J, Kent SBH, Wlodawer A. Structure at 2.5 Å resolution of chemically synthesized Human Immunodeficiency Virus type 1 protease complexed with a hydroxyethylene-based inhibitor. v Biochemistry. 1991; 30:1600–1609.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_35.html (2 of 2) [4/5/2004 4:47:02 PM]

Document

Page 36

26. Swain AL, Miller M, Green J, Rich DH, Schneider J, Kent, SBH, Wlodawer A. X-ray crystallographic structure of a complex between a synthetic protease of Human Immunodeficiency Virus 1 and a substrate-based hydroxyethylamine inhibitor. Proc. Natl. Acad. Sci. USA. 1990; 87:8805–8809. 27. Krohn A, Redshaw S, Ritchie JC, Graves BJ, Hatada MH. Novel binding mode of highly potent HIV proteinase inhibitors incorporating the (R)-hydroxyethylamine isostere. J. Med. Chem. 1991; 34:3340–3342. 28. Erickson J, Neidhart DJ, VanDrie J, Kempf DJ, Wang XC, Norbeck DW, Plattner JJ, Rittenhouse JW, Turon M, Wideburg N, Kohlbrenner WE, Simmer R, Helfrich R, Paul DA, Knigge, M. Design, activity, and 2.8 Å crystal structure of a C2 symmetric inhibitor complexed to HIV-1 protease. Science. 1990; 249:527–533. 29. Hosur MV, Bhat TN, Kempf DJ, Baldwin ET, Liu B, Gulnik S, Wideburg NE, Norbeck DW, Appelt K, Erickson JW. Influence of stereochemistry on activity and binding modes for C2 symmetry-based diol inhibitors of HIV-1 protease. J. Amer. Chem. Soc. 1994; 116:847–855. 30. Kempf DJ, Marsh KC, Denissen JF, McDonald E, Vasavanonda S, Flentge CA, Green BE, Fino L, Park CH, Kong XP, Wideburg NE, Saldivar A, Ruiz L, Kati WM, Sham HL, Robins T, Stewart KD, Hsu A, Plattner JJ, Leonard JM, Norbeck DW. ABT-538 is a potent inhibitor of human immunodeficiency virus protease and has high oral availability in humans. Proc. Natl. Acad. Sci. USA. 1995; 92:2484–2488. 31. Kempf DJ, Marsh KC, Fino LC, Bryant P, Craig-Kennard A, Sham HL, Zhao C, Vasavanonda S, Kohlbrenner WE, Wideburg NE, Saldivar A, Green BE, Herrin T, Norbeck D.W. Design of orally available, symmetry-based inhibitors of HIV protease. Bioorg. Med. Chem. Lett. 1994; 2:847–858. 32. Holloway MK, Wai JM, Halgren TA, Fitzgerald PMD, Vacca JP, Dorsey BD, Levin RB, Thompson WJ, Chen LJ, deSolms SJ, Gaffin N, Ghosh AK, Giuliani EA, Graham SL, Guare JP, Hungate RW, Lyle TA, Sanders WM, Tucker TJ, Wiggins M, Wiscount CM, Woltersdorf OW, Young SD, Darke PL, Zugay JA. A priori prediction of activity fro HIV-1 protease inhibitors employing energy minimization in the active site. J. Med. Chem. 1995; 38:305–317. 33. Vacca JP, Guare JP, deSolms SJ, Sanders WM, Giuliani EA, Young SD, Darke PL, Zugay J, Sigal IS, Schleif W, Quintero J, Emini E, Anderson P, Huff JR. L-687,908, a potent hydroxyethylenecontaining HIV protease inhibitor. J. Med. Chem. 1991; 34:1225–1228. 34. Dorsey BD, Levin RB, McDaniel SL, Vacca JP, Guare JP, Darke PL, Zugay JA, Emini EA, Schleif WA, Quintero LC, Lin JH, Chen I-W, Holloway MK, Fitzgerald PMD, Axel MG, Ostovic D, Anderson PS, Huff JR. L-735,524: the design of a potent and orally available HIV protease inhibitor. J. Med. Chem. 1994; 37:3443–3451.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_36.html (1 of 2) [4/5/2004 4:47:03 PM]

Document

35. Kaldor SW, Hammond M, Dressman BA, Frtiz JE, Crowell TA, Hermann RA. New peptide isosteres useful for the inhibition of HIV—1 protease. Bioorg. Med. Chem. Lett. 1994; 4:1385–1390. 36. Kaldor SW, Dressman BA, Hammond M, Appelt K, Burgess JA, Lubbehausen PP, Muesing MA, Hatch SD, Wiskerchen MA, Baxter AJ. Isophthalic acid derivatives: amino acid surrogates for the inhibition of HIV-1 protease. Bioor. Med. Chem. Lett. 1995; 5:721–726.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_36.html (2 of 2) [4/5/2004 4:47:03 PM]

Document

Page 37

37. Kaldor SW, Appelt K, Fritz JE, Hammond M, Crowell TA, Baxter AJ, Hatch SD, Wiskerchen MA, Muesing MA. A systemic study of P1–P3 spanning sidechains for the inhibition of HIV-1 protease. Bioorg. Med. Chem. Lett. 1995; 5:715–720. 38. Kalish JV, Tatlock JH, Davies JF, Kaldor SW, Dressman BA, Reich S, Pino M, Nyugen D, Appelt K, Musick L, Wu B-W. Structure-based drug design of nonpeptidic P2 substituents for HIV-1 protease inhibitors. Bioorg. Med. Chem. Lett. 1995; 5:727–732. 39. Kaldor SW, Kalish VJ, Davies JF, Appelt K, Tatlock JH, Dressman BA, Campanale KM, Burgess JA, Lubbehusen PL, Muesing MA, Hatch SD, Shetty BV, Patick AK, Kosa MB, Khalil DA. AG1343: a potent orally bioavailable inhibitor of HIV-1 protease. J. Med. Chem. 1996; submitted for publication. 40. Shetty B, Kosa M, Khalil DA, Webber S. Preclinical Pharmacokinetics and Distribution to Tissue of AG1343, an inhibitor of human deficiency virus type 1 protease. Antimicr. Ag. and Chemoth. 1996; 40:110–114. 41. Quart BD, Chapman SK, Peterkin J, Webber S, Oliver S. Phase 1 safety, tolerance, pharmacokinetics and food effect studies of AG1343—a novel HIV protease inhibitor. Abst. LB3. In: Proceedings of the 2nd National Conference on Human Retroviruses and Related Infections. 1995:163. 42. Lam PYS, Jadhaw PK, Eyermann CJ, Hodge CN, Lee YR, Bacheler LT, Meek JL, Otto MJ, Rayner MM, Wong YN, Chang C-H, Weber PC, Jackson DA, Sharpe TR, Erickson-Viitanen S. Rational design of potent, bioavailable, nonpeptide cyclic ureas as HIV protease inhibitors. Science, 1994; 263:380–384. 43. Reich SH, Melnick M, Davies JF, Appelt K, Lewis KK, Fuhry MA, Pino M, Trippe AJ, Nguyen D, Dawson H, Wu B-W, Musick L, Kosa M, Kahil D, Webber S, Gehlhaar DK, Andrada D, Shetty B. Protein structure-based design of potent, orally bioavailable, nonpeptide inhibitors of human immunodeficiency virus protease. Proc. Natl. Acad. Sci. 1995; 92:3298–3302. 44. Reich SH, Melnick M, Pino M, Fuhry MA, Trippe AJ, Appelt K, Davies JF, Wu B-W, Musick L. Structure-based design and synthesis of substituted 2-butanols as nonpeptidic inhibitors of HIV protease: secondary amide series. J. Med. Chem. 1996; 39:2781–2794. 45. Melnick M, Reich SH, Lewis KK, Mitchell L, Ngyen D, Trippe AJ, Dawson H, Davies JF, Appelt K, Wu B-W, Musick L, Gehlhaar DK, Webber S, Shetty B, Kosa M, Kahil D, Andrada D. Bis-tertiary amide inhibitors of the HIV-1 protease generated via protein structure-based iterative design. J. Med. Chem. 1996; 39:2795–2811. 46. Gehlhaar DK, Moerder KE, Zichi D, Sherman CJ, Ogden RC, Freer ST. De novo design of enzyme inhibitors by Monte Carlo ligand generation. J. Med. Chem. 1995; 38:466–472.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_37.html (1 of 2) [4/5/2004 4:47:05 PM]

Document

47. Tummino PJ, Ferguson D, Hupe L, Hupe D. Competitive inhibition of HIV-1 protease by 4-hydroxybenzopyran-2-ones and 4-hydroxyphenylpyran-2-ones. Biochem. Biophys. REs. Commun. 1994; 200:1658–1664. 48. Tummino PJ, Fergusson D. Hupe D. Competitive inhibition of HIV-1 protease by warfarin derivatives. Biochem. Biophys. Res. Commun. 1994; 201:290–294. 49. Thaisrivongs S, Tomich PK, Watenpaugh KD, Chong KT, Howe WJ, Yang K-T, Strohbach JW, Turner SR, McGRath JP, Bohanon JCL, Mulichak AM, Spinelli PA, Hinshaw RR, Pagano PJ, Moon JB, Ruwart MJ, Wilkinson KF, Rush BD,

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_37.html (2 of 2) [4/5/2004 4:47:05 PM]

Document

Page 38

Zipp GL, Dalga RJ, Schwende FJ, Howard GM, Padbury GE, Toth LN, Zhao Z, Koeplinger KA, Kakuk TJ, Cole SL, Zaya RM, Piper RC, Jeffrey P. Structure-based design of HIV protease inhibitors: 4-hydroxycoumarins and 4-hydroxy-2-pyrones as nonpeptidic inhibitors. J. Med. Chem. 1994; 37:3200–3204. 50. Vara Prasad JVN, Para KS, Lunney EA, Ortwine DF, Dunbar Jr. JB, Fergusson D, Tummino PJ, Hupe D, Tait BD, Domagala JM, Humblet C, Bhat TN, Liu B, Guerin DMA, Baldwin ET, Erickson JW, Sawyer TK. Novel series of achiral low molecular weight and potent HIV-1 protease inhibitors. J. Am. Chem. Soc. 1994; 116:6989–6990. 51. Ho DD, Neuman AU, Pereison AS, Chen W, Leonard JM, Markowitz M. Rapid turnover of plasma virions and CD4 lymphocytes in HIV-1 infection. Nature 1995; 373:123–126. 52. Wie X, Ghosh SK, Taylor ME, Johnson VA, Emini EA, Deutsch P, Lifson JD, Bonhoeffer S, Nowak MA, Hahn BH, Saag MS, Shaw GM. Viral dynamics in HIV-1 infection. Nature 1995; 373:117–122. 53. Coffin JM. HIV population dynamics in vivo: implications for genetic variation, pathogenesis and therapy. Science 1995; 267:483–489. 54. De Clercq E. HIV resistance to reverse transcriptase inhibitors. Biochem. Pharmacol. 1994; 47:155–169. 55. Mo H, Markowitz M, Ho DD. Patterns of specific mutations in HIV-1 protease that confer resistance to protease inhibitors in clinical development. Third International Workshop on HIV drug resistance. Kauai, Hawaii, August 1994. Abstract 13. 56. Markowitz M, Mo H, Kempf DJ, Norbeck D. Selection and analysis of human immunodeficiency virus type 1 variants with increased resistance ABT-538, a novel protease inhibitor. J. Virol 1995; 69:701–706. 57. Kaplan AH, Michael SF, Wehbie RS, Knige MF, Paul DA, Everitt L, Kempf DJ, Norbeck DW, Erickson JW, Swanstrom R. Selection of multiple human immunodeficiency virus type 1 variants that encode viral proteases with decreased sensitivity to an inhibitor of the viral protease. Proc. Natl. Acad. Sc. USA 1994; 91:5597–5601. 58. Erickson JW, Baldwin ET, Bhat TN, Gulnik S, Leu B, Yr B. Structural basis of drug resistance to HIV-1 protease inhibitors. Third International Workshop on HIV Drug Resistance. Kuaui, Hawaii, August 1994, Abstract 2. 59. Condra JH, Schleif WM, Blahy OH, Gabryelski LJ, Graham DJ, Quintero JC, Rhodes A, Robbins HL, Roth E, Shivaprakash M, Titus PL, Yang Y, Emini EA. Mutations in HIV protease conferring resistance to inhibitor L735,524. Abstract 187. In: Proceedings of the 2nd National Conference on Human Retroviruses and Related Infections. 1995:88.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_38.html (1 of 2) [4/5/2004 4:47:07 PM]

Document

60. Condra JH, Schleif WA, Blahy OM, Gabryelski LJ, Graham DJ, Quintero JC, Rhodes A, Robbins HL, Roth E, Shivaprakash M. In vivo emergence of HIV-1 variants resistant to multiple protease inhibitors. Nature, 1995; 374:569–571. 61. Jackobsen H, Yasargil K, Winslow JC, Craig JC, Krohn A, Duncan IB, Mous J. Characterization of human immunodeficiency virus type 1 mutants with decreased sensitivity to proteinase inhibitor RO 31–8959. Virology, 1995; 206:527–534. 62. Patick AK, Mo H, Markowitz M, Appelt K, Wu B-W, Musick L, Kalish V, Kaldor SW, Reich SH, Ho D, Webber S. Antiviral and resistance studies of AG1343, an

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_38.html (2 of 2) [4/5/2004 4:47:07 PM]

Document

Page 39

orally bioavailable inhibitor of human immunodeficiency virus protease. Antimicr. Ag. and Chemoth. 1996; 40:292–297. 63. Korant B, Lu Z, Strack P, Rizzo C. HIV protease mutations leading to reduced inhibitor susceptibility. In: Intracellular Protein Catabolism. New York: Plenum Press, 1996:241–250. 64. Patick AK, Duran M, Cao Y, Pei Z, Keller MR, Peterkin J, Chapman S, Anderson B, Markowitz M. Genotypic and phenotypic characterization of HIV-1 variants isolated from in vitro selection studies and from patients treated with the protease inhibitor, nelfinavir. Fifth International Workshop on HIV Drug Resistance, Whistler, Canada, 1996:29. 65. Jacobsen H, Brun-Vezinet F, Duncan I, Hanggi M, Ott M, Vella S, Weber J, Mous J. Genotypic characterization of HIV-1 from patients after prolonged treatment with protease inhibitor saquinavir. In: Abstracts of the 3rd International Workshop on HIV Drug Resistance. London: MediTech Media, 1994:16. 66. Gulnik SV, Suvorov LI, Liu B, Yu B, Anderson B, Mitsuya H, Erickson JW. Kinetic characterization and cross-resistance patters of HIV-1 protease mutants selected under in vitro drug pressure. Biochemistry 1995; 34:9282–9287. 67. Stein DS, Fish DG, Chodakewitz J. A 24-week open-label phase I evaluation of the HIV protease inhibitor L 735,524. Abstract LB1 Second National Conference on Human Retroviruses and Related Infections, Washington D.C., 1995. 68. Markowitz M, Jalil L, Hurley A. Evaluation of the antiviral activity of orally administered ABT-538, an inhibitor of HIV-1 protease. Abstract 185 Second National Conference on Human Retroviruses and Related Infections, Washington D.C., 1995. 69. Gathe Jr. J, Burkhardt B, Hawley P, Conant M, Peterkin J, Chapman S. A randomized Phase II study of Virocept™, a novel HIV protease inhibitor, used in combination with stavudine (D4T) vs. stavudine (D4T) alone. Abstract Mo.B.413 In: XI International Conference on AIDS, Vancouver, 1996:25. 70. Hammer S. Advances in antiretroviral therapy and viral load monitoring. Abstract Mo.01 In: Abstracts of the XI International Conference on AIDS, Vancouver, 1996:2. 71. VanDrie JH, Weininger D, Martin YC. Aladdin: an integrated tool for computer-assisted molecular design and pharmacophore recognition from geometric, steric, and substrate searching of threedimensional structures. Computer-Aided Mol. Design. 1989; 3:225–234. 72. Kuntz ID, Blanley JM, Oatley SJ, Langridge R, Ferrin TE. A geometry approach to macromoleculeligand interactions. J. Mol. Biol. 1982; 161:269–278. 73. Van Gunsteren WF Berendsen HJC. Computer simulation of molecular dynamics: methodology, application and perspectives in chemistry. Angew. Chem. Int. Ed. Eng. 1990; 29:992–996. http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_39.html (1 of 2) [4/5/2004 4:47:09 PM]

Document

74. Reddy RM, Varney MD, Kalish V, Viswanadhan VN, Appelt K. Calculation of relative differences in the binding free energies of HIV-1 protease: a thermodynamic cycle perturbation approach. J. Med. Chem. 1994; 37:1145–1152. 75. Verkhivker G, Appelt K, Freer ST, Villafranca JE. Empirical free energy calculations of ligandprotein crystallographical complexes—knowledge-based ligand-protein interaction potentials applied to the prediction of human immunodeficiency virus protease binding affinity. Protein Engineering, 1995; 8:677–691.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_39.html (2 of 2) [4/5/2004 4:47:09 PM]

http://legacy.netlibrary.com/reader/message.asp?message=811&BookID=12640&FileName=Page_40.html

The requested page could not be found. Return to previous page

http://legacy.netlibrary.com/reader/message.asp?message=811&BookID=12640&FileName=Page_40.html [4/5/2004 4:47:11 PM]

Document

Page 41

2 Structural Studies of HIV-1 Reverse Transcriptase and Implications for Drug Design Jianping Ding, Kalyan Das, Yu Hsiou, Wanyi Zhang, and Edward Arnold Center for Advanced Biotechnology and Medicine, and Rutgers University, Piscataway, New Jersey Prem N. S. Yadav University of Medicine and Dentistry of New Jersey, Piscataway, New Jersey Stephen H. Hughes ABL-Basic Research Program, NCI-Frederick Cancer Research and Development Center, Frederick, Maryland I. Introduction Like all other retroviruses, human immunodeficiency virus type 1 (HIV-1) contains the multifunctional enzyme reverse transcriptase (RT). Retroviral RTs have a DNA polymerase activity that can use either an RNA or a DNA template and an RNase H activity. HIV-1 RT is essential for the conversion of singlestranded viral RNA into a linear double-stranded DNA that is subsequently integrated into the host cell chromosomes [1–4]. In this conversion process HIV-1 RT catalyzes the incorporation of approximately 20,000 nucleotides. Chemotherapeutic agents have been identified that target virtually all stages of the HIV-1 replication cycle (see review [5]). Since both the polymerase and RNase H activities of HIV-1 RT are essential, inhibiting either step blocks viral replication. Therefore, HIV-1 RT is an important target for the treatment of AIDS. Two major classes of antiviral agents that inhibit HIV-1 RT polymerization have been identified; these are nucleoside RT inhibitors (NRTIs) (Figure la) and nonnucleoside RT inhibitors (NNRTIs) (Figure 1b). Nucleoside analogs, such as 3'-azido-2',3'dideoxythymidine (AZT), 2',3'-dideoxyinosine (ddI), 2',3'-dideoxycytidine (ddC), 2',3'-dideoxy-3'thiacyti-

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_41.html [4/5/2004 4:47:13 PM]

Document

Page 42

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_42.html (1 of 2) [4/5/2004 4:47:17 PM]

Document

Figure 1 (a) Chemical structures of representative nucleoside analog inhibitors of HIV-1 RT. AZT: 3'-azido-2',3'-dideoxythymidine; d4T: 2',3'-didehydro-2',3'-dideoxythymidine; ddI: 2',3'-dideoxyinosine; ddC: 2',3'-dideoxycytidine; 3TC: 2',3'-dideoxy-3'-thiacytidine; PMEA: 9-(2-phosphonylmethoxylethyl)adenine.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_42.html (2 of 2) [4/5/2004 4:47:17 PM]

Document

Page 43

(b) Chemical structures of representative nonnucleoside inhibitors of HIV-1 RT. Nevirapine: 11 -cyclopropyl-5,11 -dihydro-4-methyl-6H-dipyrido (3,2-b:2'3'-e)(1,4) diazepin-6-one; α-APA: αanilinophenylacetamine; TIBO: tetrahydroimidazo-(4,5,1-jk) (1,4)-benzo-diazepin-2(1H)-one and thione; pyridinones; HEPT: 1-{(2-hydroxyethoxy) methyl}-6-(phenylthio)thymine; BHAP: bis(heteroaryl)piperazine; TSAO: {2',5'-bis-O-(tert-butyldimethylsilyl)}-3' -spiro-5''-(4''-amino-1",2"-oxathiole)-2", 2"-dioxide; L-737,126: 5-chloro3-(phenylsulfonyl)indole-2-carboxamide; TBA: 1-(2',6'-difluoro-phenyl)-1H,3H-thiozolo -(3,4-a) -benzimidazole; quinoxaline S-2720: 6-chloro-3,3-dimethyl-4-(isopropenyloxycarbonyl) -3-4-dihydroquinoxalin-2(1H)-thione.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_43.html (1 of 2) [4/5/2004 4:47:21 PM]

Document

dine (3TC), and 2',3'-didehydro-2',3'-dideoxythymidine (d4T), have been widely used in the treatment of HIV-1 infections [5–7]. However, the effectiveness of these drugs is limited by their cytotoxicity and the rapid emergence of drug-resistant viral strains [5,8–12]. Nonnucleoside inhibitors, e.g., the HEPT derivatives [13], TIBO derivatives [14], nevirapine [15], pyridinones [16], BHAP derivatives [17], TBA derivatives [18,19], TSAO derivatives [20], α-APA [21], and quinoxalines (HBY) [22,23], are potent inhibitors of HIV-1 RT (see reviews [5,11,12]). While these inhibitors differ considerably in chemical structure, all of them are quite specific for HIV-1 RT and inhibit neither HIV-2 RT nor variety of cellular polymerases. Challenging a virus with these drugs

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_43.html (2 of 2) [4/5/2004 4:47:21 PM]

Document

Page 44

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_44.html [4/5/2004 4:48:34 PM]

Document

Page 45 Figure Continued

rapidly selects viral strains containing drug-resistance mutations [9,24,25]. HIV-1 viral variants are known whose RT is resistant to all of the currently available drugs/inhibitors (see reviews [5,9,11,12,26,27]). In some cases, drug-resistant variants can be selected in very short periods of time [9], a consequence of the high viral load and rapid turnover of viral populations in infected individuals [28–30]. A better understanding of how these viral variants confer resistance should provide insight into the limitations of their genetic flexibility. In the past few years, substantial progress has been made in understanding the three-dimensional structure of HIV-1 RT. This paper will discuss the recent biochemical, genetic, and clinical data of HIV-1 RT in the context of the crystal structure of HIV-1 RT and prospects for development of more effective inhibitors of HIV-1 replication. II. Three-Dimensional Structures of HIV-1 RT Three-dimensional crystal structures of HIV-1 RT have been determined both for the unliganded form of the protein and for complexes with either template-primer substrate or nonnucleoside inhibitors (Figure 2 and Table 1). Structures of HIV-1 RT have been determined in complexes with a series of NNRTIs, including nevirapine [31–33], 1051U91 (a nevirapine analog) [33], α-APA R95845 [34], α-APA R90385 [33], HEPT [33], 8-Cl TIBO (R86183) [35], and 9-Cl TIBO (R82913) [36,37]. The structure of HIV-1 RT in a ternary complex with a 19-mer/18-mer double-stranded DNA (dsDNA) template-primer and an antibody Fab fragment has been described [38]. In addition, structures of unliganded HIV-1 RT have also been solved in multiple crystal forms [39–43]. The structure of a polypeptide corresponding to the fingers and palm subdomains of the HIV-1 RT polymerase domain has also been determined [44]. HIV-1 RT is an asymmetric heterodimer consisting of the p66 (66 kDa) and p51 (51 kDa) subunits. The N-terminal 440 residues of the p66 subunit constitute the polymerase domain and the C-terminal 120 residues of p66 form the RNase H domain; the p51 subunit has the same amino acid sequence as the polymerase domain of the p66 subunit [1,2]. The polymerase domain of the p66 subunit has been likened to a human right hand. On this basis the subdomains of both p66 and p51 have been designated as fingers, palm, thumb, and connection (Figure 2) [31,38]. In the p66 subunit, the fingers, palm, and thumb subdomains form a large cleft that can accommodate the DNA substrate. The polymerase active site, which contains three strictly conserved aspartic acid residues (Asp110, Asp185, and Asp186), is located in the DNA-binding cleft and is part of the p66 palm subdomain (Figure 3) [31,38]. In the p51 subunit, however, the thumb is rotated away from the fingers and the connection subdomain is folded over onto the palm subdomain between the fingers and thumb subdomains. As a

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_45.html [4/5/2004 4:48:36 PM]

Document

Page 46 Table 1 Crystal Structures of HIV-1 Reverse Transcriptase

PDB entry

HIV-1 RT Structure

Resolution Range (Å)

R-Factor/Free R-Factor (%)

Data Completeness (%)

Temperature (°C)

References

85

-165

41

21.9

89.5

-173

42

1hmv

unliganded RT

6.0-3.0

25.4/29.7

1rtj

unliganded RT

25.0-2.35

1dlo

unliganded RT

8.0-2.7

23.0/33.6

99.5

-165

43

1hmi

RT/DNA/Fab

15.0-3.0

26.0

88.1

-10

38

3hvt

RT/nevirapine

8.0-2.9

26.6

95.6

-165

31,32

1vrt

RT/nevirapine

25.0-2.2

18.6

87.1

16

33

1rth

RT/1051U91

25.0-2.2

21.4

81.4

16

33

1hni

RT/α-APA (R95845)

10.0-2.8

25.5/36

78.5

-15

34

1vru

RT/α-APA (R90385)

25.0-2.4

18.7

86.5

16

33

1hnv

RT/8-Cl TIBO (R86183)

10.0-3.0

24.9/35.6

81

-10

35

1rev

RT/9-Cl TIBO (R82913)

25.0-2.6

22.4

80.7

-173

36

1tvr

RT/9-Cl TIBO (R82913)

10.0-3.0

25.9

72

-165

37

1rti

RT/HEPT

25.0-3.0

23.6

86.3

14

33

1hrh

RT (RNase H domain)

20.0-2.4

20.0

93.1

4

94

1rdh

RT (RNase H domain)

20.0-2.8

21.5

95.6

4

95

1har

RT (fingers and palm subdomains)

7.0-2.2

20.8/27.0

96

4

44

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_46.html [4/5/2004 4:48:38 PM]

Document

Page 47

Figure 2 Ribbon diagrams of the three types of HIV-1 RT crystal structures. (a) Structure of the HIV-1 RT/DNA/Fab ternary complex [38]. The bound nucleic acid is shown as double-stranded helices with template strand in black and primer strand in gray. The Fab fragment is not shown. (b) Structure of the HIV-1 RT complexed with the NNRTI TIBO R86183 [34]. For the sake of clarity, the bound TIBO inhibitor is shown as an atomic model. (c) Structure of unliganded HIV-1 RT [41,43]. (d) A schematic diagram showing the resemblance of the HIV-1 RT p66 subunit to a human right hand. The RNase H domain, which has no counterpart for a human hand, is shown as an oval below the thumb. When a template-primer binds to HIV-1 RT, the fingers, palm, and thumb subdomains of p66 form a large cleft to bind the DNA. The polymerase active site (shown as a small circle) lies at the bottom of the DNA-binding cleft. The NNRTI binds in the highly hydrophobic NNIBP (shown as a large circle), which is located in the vicinity of the polymerase active site. The p66 thumb subdomain in the NNRTI-bound HIV-1 RT structures is in an upright position extended beyond that observed in the structure of RT with bound DNA. In the absence of any bound nucleic acid or NNRTI, however, the p66 thumb folds down into the DNA-binding cleft (shown as dashed drawing).

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_47.html [4/5/2004 4:48:52 PM]

Document

Page 48

Figure Continued

consequence, the p51 submit has no cleft for binding nucleic acid substrates and hence no polymerase activity. There is considerable evidence showing that HIV-1 RT is quite flexible and that this flexibility is essential for DNA polymerization. Comparisons among DNA-bound, inhibitor-bound, and unliganded HIV-1 RT structures provide a demonstration of the enzyme's flexibility. When a DNA template-primer binds to HIV-1 RT, structural elements of the fingers, palm, and thumb subdomains of the p66 subunit form a clamp-like structure that holds the nucleic acid (Figure 2) [38]. The template-primer substrate interacts with amino acid residues of the fingers, palm, and thumb subdomains, especially in the regions denoted as “primer grip” and “template grip,” believed to position the template-primer precisely relative to the polymerase active site [38]. The primary contacts between the template-primer and the protein are along the sugar-phosphate backbone of the DNA and thus are not sequence-specific [38]. In the absence of nucleic acid template-primer or NNRTI, the thumb subdomain of p66 is folded down into the DNAbinding cleft and lies near the fingers subdomain (Figure 2) [40,41,43]. As a consequence, the DNAbinding cleft is closed. However, even in the absence of a bound nucleic acid, binding of an NNRTI induces both short-range and long-range structure distortions, including a hinge-like movement near the base of the p66 thumb that constrains the p66 thumb in a conformation that is extended beyond the upright

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_48.html [4/5/2004 4:48:59 PM]

Document

Page 49

Figure 3 Overall structure of HIV-1 RT p66/p51 heterodimer [38] showing the locations of the major target sites for anti-HIV-1 RT inhibitors. The NRTIs target the dNTP-binding site/the polymerase active site (shown as a small striped circle) that lies at the floor of the DNA-binding cleft. The NNRTIs bind to the NNIBP (shown as a large dotted circle), which is near, but distinct from, the polymerase active site. The RNase H domain is located at the C-terminal of the p66 subunit. The RNase H catalytic site (shown as a medium circle) is an attractive target site for anti-HIV-1 drugs. The HIV-1 RT p66/p51 heterodimer interface is shown as a dashed line. Since HIV-1 RT functions as a heterodimer, any inhibitors that could interfere with the dimerization process might also be potential drugs for treating HIV-1 infection.

position of the thumb observed in the HIV-1 RT/DNA/Fab structure [31,33–37,43]. In the unliganded structure of HIV-1 RT reported by Esnouf et al. [42], the p66 thumb subdomain is in an upright conformation different from that observed in the other unliganded HIV-1 RT structures but similar to that found in the DNA-bound HIV-1 RT and NNRTI-bound HIV-1 RT structures. Esnouf et al. [42] contend that the upright conformation of the p66 thumb subdomain in their unliganded RT structure is appropriate for unliganded HIV-1 RT and that the binding of an NNRTI does not affect the conformation of the p66 thumb. However, we believe that the conformation of the p66 thumb in the Esnouf et al. structure may well be a result of the method used to prepare the

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_49.html (1 of 2) [4/5/2004 4:49:12 PM]

Document

Page 50

crystals, which were produced by soaking out a weakly bound NNRTI (HEPT) from pregrown crystals. When the crystals were grown in the presence of HEPT, the p66 thumb was presumably in an upright position similar to that seen in all of the known NNRTI-bound HIV-1 RT structures. As the weakly bound HEPT diffused out, the molecular packing arrangement may have constrained the position of the p66 thumb subdomain. As a consequence, this unliganded HIV-1 RT structure may represent an intermediate between the other unliganded structures and the structure of HIV-1 RT with a bound NNRTI. It is evident that the p66 thumb subdomain has considerable flexibility and can adopt substantially different conformations during the binding of template-primer or inhibitors and, presumably during DNA polymerization as well. III. Polymerase Active Site of HIV-1 RT and the NRTIs Polymerization of DNA by HIV-1 RT involves a sequential stepwise binding of the template-primer and deoxynucleoside triphosphate (dNTP) substrates at the polymerase active site [45,46]. The incoming dNTP is covalently linked via the α-phosphorus to the 3'-oxygen of the primer strand, accompanied by the release of pyrophosphate. An essential requirement for the polymerization reaction is the presence of a 3'-OH group at the end of the primer strand. Nucleoside analogs contain a modified sugar moiety in which the 3'-OH group is replaced by another group (e.g., hydrogen, halogen, or azido) (Figure la). To exert their antiviral activity at the level of RT, the NRTIs must be phosphorylated successively to the 5'monophosphate, 5'-diphosphate, and 5'-triphosphate forms by a series of kinases. Once the NRTI is converted to the triphosphate form and interacts with RT, it can inhibit polymerization in two possible ways. One possibility is that the NRTI binds preferentially to the dNTP-binding site and competitively inhibits the binding of natural dNTPs. Another possibility, which seems to be the predominant mode of inhibition, is that the NRTI is incorporated into the growing chain and acts as a terminator of chain elongation. Once an NRTI is incorporated, no additional nucleotides can be added to the DNA chain since the primer terminal 3'-OH group (the site of phosphodiester bond formation) is absent. Phosphorylation is a crucial step in the intracellular metabolism of the NRTI and often is a limiting factor for the antiviral activity of the NRTI [47]. In an attempt to bypass the first phosphorylation step, several acyclic nucleoside phosphonates have been developed in which the sugar moiety of normal NRTIs is replaced with an acyclic phosphonate group, such as 9-(2-phosphonylmethoxyethyl)adenine (PMEA), (R)-9-(2-phosphonylmethoxypropyl)adenine (PMPA), and (S)-9-(3-fluoro-2phosphonylmethoxyethyl) adenine (FPMPA) (Figure la) (see reviews [5,11]). These acyclic nucleoside phosphonates are dideoxynucleoside

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_50.html [4/5/2004 4:49:13 PM]

Document

Page 51

monophosphate analogs which can easily be converted to the active triphosphate form by adding two additional phosphates [48,49], and can inhibit HIV replication [50,52]. Analysis of the structure of the HIV-1 RT/DNA/Fab complex showed that the dNTP-binding site is composed of both protein and nucleic acid. In addition to the 5'-terminus of the template nucleotide and three carboxylate residues (Asp110, Asp185, and Asp186), the amino acid residues Asp113, Tyr115, Phe116, Gln151, Phe160, and possibly Met184 and Lys219 form part of the putative dNTP-binding site [12,53] (Figure 4 and Table 2). It is important to realize that the precise composition, position, and conformation of the template-primer can influence the recognition and incorporation of incoming nucleotides at the polymerase catalytic site. In the wild-type HIV-1 RT, the dNTP-binding

Figure 4 Stereoview of the polymerase active site of HIV-1 RT [38]. The amino acid residues that compose the putative dNTP-binding site, including the three catalytically essential aspartic acids, are shown with side chains. The double-stranded nucleic acid is shown with the atomic model in the HIV-1 RT/DNA/Fab complex. The dNTP-binding site consists of structural elements from both protein and nucleic acid. The precise composition, position, and conformation of the template-primer can affect the recognition of incoming dNTPs at the polymerase active site.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_51.html (1 of 2) [4/5/2004 4:49:23 PM]

Document

Page 52 Table 2 HIV-1 RT Amino Acid Residues Composing the Putative dNTP-Binding Sites and the Locations of NRTIResistance Mutations (see also [12]) dNTP-Binding Site

Nucleoside Drug-Resistance Mutation Site

Residue

Location

Mutation

Location

Possible Effects

Asp110

β6

Met41Leu

αA

template binding

55

Asp113

β6-αC loop

Ile50Thr

β2

unclear

154

Try115

αC

Lys65Arg

β3–β4

template binding

154

Phe116

αC

Asp67Asn

β3–β4

template binding

155

Gln151

β8-αE

Thr69Asp

β3–β4

template binding

156

Phe160

αE

Lys70Arg

β3–β4

template binding

155

Asp185

β9–β10

Leu74Val

β4

template binding

8

Asp186

β9–β10

Val75Thr

β4

template binding

163

Lys219

β11

Glu89Gly

β5

dsDNA binding

157

Tyr115Phe

αC

dNTP binding

170

Ile135Thr

β7–β8

unclear

Gln151Met

β8-αE

dNTP binding

Met184Val, Ile

β9–β10

dNTP-binding/fidelity

Thr215Tyr, Phe, Cys

β11

indirect effect/dNTP binding

Lys219Gln

β11

dNTP binding

References

158,159

154,160,161 155,162

155

site can accommodate both the normal dNTP substrates and dideoxynucleoside analogs. The majority of mutations that confer resistance to NRTIs are not located at the dNTP-binding site; however, they appear to influence the geometry of the dNTP-binding site indirectly in a way that permits RT to discriminate between a normal dNTP and a modified nucleoside triphosphate (see discussion in the next section).

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_52.html (1 of 2) [4/5/2004 4:49:26 PM]

Document

Analysis of various HIV-1 RT structures has revealed an unusual β-turn geometry for the YMDD motif at the polymerase catalytic site of p66 [33,34,41–43]. The energetically unfavorable main chain conformation of Met184 (torsion angles φ~60° and ϕ~-120°) is stabilized by a hydrogen bond of its carbonyl oxygen to the side chain of either Gln182 [33,41,43] or Gln161 [42]. It has been suggested that this novel β-turn geometry might be required to position the aspartic acids in precisely the correct way for catalysis [41]. IV. Mechanism of NRTI-Resistance Mutations Development of resistance to NRTIs has been a major problem with clinical use of these drugs. Careful analysis of mutations that confer resistance to different

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_52.html (2 of 2) [4/5/2004 4:49:26 PM]

Document

Page 53

NRTIs in light of available structural data might provide information that could be used in the development of improved NRTIs that are more effective against the commonly observed NRTI-resistant HIV-1 variants. A viable drug-resistant RT mutant should be able to recognize and incorporate normal nucleoside triphosphate, yet reject a nucleoside analog. The only difference between normal nucleotide substrates and the NRTIs is the modification of the sugar moiety. This alteration may affect sugar puckering and the conformation of the glycosyl bond. Recognition of these differences could render the triphosphate form of the nucleoside analog a good substrate for wild-type RT but a poor substrate for a drug-resistant variant of RT. Structural analysis of HIV-1 RT has shown that most of the NRTI-resistance mutations are not located close to the putative dNTP-binding site and are unlikely to have a direct impact on the binding of dNTP analogs (Figure 5 and

Figure 5 A close-up view showing the relative locations of the commonly identified drug-resistance mutations for NRTIs (in dark-gray) and for NNRTIs (in light-gray) with respect to the bound DNA. Most of the NRTI-resistance mutations are not located at the putative dNTP-binding site, but are at positions to have potential interactions with the nucleic acid template-primer. Conversely, all the NNRTI-resistance mutations are clustered around the NNIBP and have direct contacts with NNRTIs or have direct effect on inhibitor binding.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_53.html (1 of 2) [4/5/2004 4:49:46 PM]

Document

Page 54

Table 2) [12,54]. For example, none of the five mutations, in HIV-1 RT Met41Leu, Asp67Asn, Lys70Arg, Thr215Tyr, and Lys219Gln [55] (Table 2) consistently associated with resistance to AZT, are at locations close to the dNTP-binding site. However, most (but not all, c.f. [56]) biochemical studies have failed to show that recombinant HIV-1 RT enzymes containing these mutations are more resistant to inhibition by AZT triphosphate than the wild-type HIV-1 RT [27,57,58]. Other mutations that confer resistance to NRTIs have been identified at positions 50, 65, 69, 74, 75, 89, 115, 135, 151, and 184 of HIV-1 RT (Figure 5 and Table 2). Most of these mutations do not lie close to the dNTP-binding site (Met184Val/Ile are the exception), but instead are located at positions where they could interact with the nucleic acid template-primer [12,59,60]. Biochemical data have shown that only when the 5'-template extension length is greater than three nucleotides does the wild-type RT begin to incorporate dideoxynucleotides effectively [54]. If the template extension is less than three nucleotides in length, wild-type HIV-1 RT is resistant to dideoxynucleotides. On the other hand, HIV-1 RT variants containing the mutations Leu74Val or Glu89Gly did not readily incorporate dideoxynucleotides either with short or long template extensions [54]. Based on both structural and biochemical data, it was proposed that mutations that cause HIV-1 RT to have reduced sensitivity to NRTIs exert their effects via interactions with the nucleic acid template-primer, which consequently alter the geometry of the polymerase active site [54]. It has been suggested that mutations that confer resistance to foscarnet might use a similar mechanism [61]. One possible exception to this mechanism might be the mutations of Met184Val and Met184Ile (see review [27]). Part of the highly conserved YMDD motif, Met184 is adjacent to residues Asp185 and Asp186, which are two of the three catalytically essential aspartic acid residues at the polymerase active site. In addition, Met184 appears to interact with the ribose moiety of the 3'-terminal nucleotide of the primer strand [12,38,53] (Figure 4). Therefore, mutations at this position could affect interactions with the incoming dNTP directly and/or alter the positioning of the nucleic acid. These mechanisms are not mutually exclusive and which mechanism is responsible for resistance has not yet been resolved [62]. There are two recent reports suggesting that the Met184Val mutant HIV-1 RT has approximately three-fold higher fidelity than the wild-type enzyme [63,64]. Based on these data, it was suggested that this increase in fidelity might reduce the overall rate of generation of viral variants in patients treated with 3TC or other dideoxynucleosides [64]. However, owing to both theoretical and technical problems with these analyses, these conclusions are controversial. Determination of crystal structures of both wild-type and mutant HIV-1 RT complexed with individual NRTIs in the presence of a variety of template-primers and/or dNTP substrates should provide a better understanding of the mechanisms of dNTP selection and drug resistance.ed at Arial for catalysis [41]. IV. Mechanism of NRTI-Resistance Mutations Development of resistance to NRTIs has been

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_54.html [4/5/2004 4:49:47 PM]

Document

Page 55

V. Drug Design Targeting at the Polymerase Active Site All of the existing NRTIs contain a modified sugar moiety that lacks the 3'-OH group that is essential for incorporation of the next nucleotide. Modification can also be made on other functional groups such as the base and the triphosphate moieties. It may be worthwhile to try to alter the base moiety of the NRTIs to produce compounds that will be more specific to HIV-1 RT (i.e., less cytotoxic to normal cellular polymerases) and more effective against both wild-type or drug-resistant viral variants. Structure-activity analysis indicates that the pyrimidine moiety of the NRTIs can be modified at the C5 position. An AZT derivative that has a 3'-azido group on the sugar moiety and a methyl group at the C5 position of the pyrimidine moiety showed potent antiviral activity [65]. Substitution of the methyl group with a chlorine atom at position C5 of AZT results in a compound that has strong anti-HIV-1 activity [66]. Other possible substitutions at the C5 position include other halogen atoms or an ethyl group. Another possible drug-design strategy would be to devise compounds that can interface with the binding of the metal ions (Mg2+ or Mn2+) at the polymerase active site. Metal ions appear to be important in DNA polymerase catalysis. Based on the structural and biochemical data, a two-metal dependent mechanism of polymerization has been postulated [53,67,68] that is similar to that proposed for other DNA polymerases [69–71]. In this model, the metal ions mediate interactions between the three catalytically essential aspartic acid residues (Asp100, Asp185, and Asp186) and the α-, β-, and γphosphates of the incoming dNTP and promote the nucleophilic attack on the α-phosphate by the oxygen atom of the 3'-OH group of the primer strand. In the structure of the fingers and palm subdomains of the RT of Moloney murine leukemia virus (MuLV), a single Mn2+ ion was found bound to the two aspartic acid residues at the polymerase active site [72]. In the structure of the unliganded HIV-1 RT, an electron density peak was located at the polymerase active site with a good coordination geometry to the Oδ1 atoms of both Asp185 and Asp186 [43]. This electron-density peak is in a position similar to that of the Mn2+ ion observed in the MuLV RT structure. It is possible that this position corresponds to a Mg2+ ion-binding site [43]. It might be useful to design inhibitors that would influence the metal-ion coordination using either computer-based calculations (such as DOCK [73–75]) or based directly on an analysis of HIV-1 RT structure. Crystal structures of HIV-1 RT complexed with Mg2+/Mn2+ ion(s) at the polymerase catalytic site in the presence of template-primer and/or dNTP substrates would be helpful in defining the target sites of inhibitors of this type. Further studies on the structure activity relationship of HIV-1 RT complexes with these inhibitors, if active, might ultimately lead to a new type of HIV-1 RT drug that would not compete with the dNTP binding but would affect the DNA polymerization mechanism.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_55.html [4/5/2004 4:49:49 PM]

Document

Page 56

It is also attractive to consider developing agents that bind to the HIV-1 RT polymerase active site but are not nucleoside analogs. Since the amino acid residues at the dNTP-binding site are highly conserved, viral variants resistant to such inhibitors may be significantly impaired in their polymerase activity. However, there is a good chance that drug resistance could result from mutations in HIV-1 RT that influence the precise positioning of template-primer [54]. Initial attempts to use this approach starting from the crystal structure of the HIV-1 RT/DNA/Fab complex [38] (Ding, et al., in preparation) have uncovered some interesting lead compounds (Kuntz, Kenyon, Arnold, Hughes, et al., unpublished). VI. NNRTIs and the NNIBP Nonnucleoside RT inhibitors (NNRTIs) constitute the other major class of HIV-1 RT inhibitors. Many structurally distinct families of NNRTIs have been identified, including HEPT [13], TIBO [14], nevirapine [15], BHAP [17], TBA [18,19], TSAO [76], α-APA [21], pyridinones [16] and quinoxalines (HBY) [22,23] (Figure 1b). However, development of drug resistance is a major problem when NNRTIs are used to treat AIDS patients. An ideal drug should be able to block replication of all viable strains of HIV-1, but should not inhibit normal cellular enzymes. In this regard, the known NNRTIs may be too specific. While these inhibitors do not inhibit cellular polymerases, they are also inactive against HIV-2 RT (which can be viewed as an extreme variant of HIV-1 RT). In addition, drug-resistant variants of HIV-1 RT emerge rapidly in the presence of most inhibitors. In contrast, the NRTIs inhibit a broad spectrum of polymerases including the host cellular polymerases. Though it appears to be more difficult for the virus to evade NRTIs than NNRTIs (in general, it takes longer for the virus to develop resistance to NRTIs than NNRTIs), NRTI toxicity is a serious problem. Structural and biochemical studies have shown that all NNRTIs bind in a highly hydrophobic pocket in the p66 subunit, located approximately 10 Å away from the polymerase active site (Figures 2 and 3) [31,33–37]. Nevertheless, in all known structures of HIV-1 RT/NNRTI complexes, the bound NNRTIs have not been found to have any direct interactions with residues that compose the putative dNTPbinding site. The nonnucleoside inhibitor binding pocket (NNIBP) contains primarily amino acid residues from the β5–β6 loop (Pro95, Leu100, Lys101, and Lys103), β6 (Val106 and Val108), the β9–β10 hairpin (Val179, Tyr181, Tyr188, and Gly190), and the β12–β13 hairpin (Phe227, Trp229, Leu234, His235, and Pro236) of the p66 palm subdomain, and β15 (Tyr318) of the p66 thumb subdomain, as well as the β7–β8 connecting loop (Glu138) of the p51 fingers subdomain (Figure 6 and Table 3).

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_56.html [4/5/2004 4:49:51 PM]

Document

Page 57

Figure 6 Superposition of the NNRTIs in the HIV-1 RT/TIBO complex [35], HIV-1 RT/α-APA complex [34], and HIV-1 RT/nevirapine complex [32]. The side chains are shown for those amino acid residues that have close contacts with bound inhibitors and the three catalytically essential aspartic acids in the HIV-1 RT/TIBO complex. Most of the amino acid residues that form the NNIBP are hydrophobic. Though the NNRTIs are chemically and structurally diverse, the bound NNRTIs all assume a strikingly common butterfly-like shape.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_57.html [4/5/2004 4:50:13 PM]

Document

Page 58 Table 3 HIV-1 RT Amino Acid Residues Defining the Nonnucleoside Inhibitor-Binding Pocket (NNIBP) and the Locations of NNRTI-Resistance Mutations (see also [12]) NNIBP Residues

Location

Mutation

Possible Effects

References

Pro95

β5

Ala98

β5–β6

Gly

less bulky

9,164

Leu100

β5–β6

Ile

β-branch

9,164-166

Lys101

β5–β6

Glu

charge change

Lys103

β5–β6

Asn, Gln

charge loss, less bulky

9,24,164

Val106

β6

Ala

less bulky

9,58,165

Val108

β6

Ile

bulkier

Glu138

β7–β8(p51)

Lys

charge change

166

Val179

β9

Glu, Asp

charge gain, bulkier

164

Tyr181

β9

Cys, Ile

aromaticity loss, less bulky

24,58,164

Tyr188

β10

His, Cys, Leu

aromaticity loss, less bulky

9,167

Gly190

β10

Glu

charge gain, bulkier

Phe227

β12

Leu228

β12

Phe

aromaticity gain, bulkier

133

Trp229

β12

Glu233

β13

Val

charge loss, less bulky

169

Leu234

β13

Pro236

β13–β14

Leu

increase flexibility, bulkier

133

Lys238

β14

Thr

charge loss, less bulky

133

Tyr318

β15

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_58.html (1 of 2) [4/5/2004 4:50:15 PM]

164

9,164

9,22,168

Document

Most of the amino acid residues that form the binding pocket are hydrophobic and five of them are aromatic residues. The hydrophobic interactions of the side chains of these residues, especially Tyr181, Tyr188, and Trp229, with the hydrophobic moieties of the NNRTIs appear to be important for inhibitor binding [32–34,59]. Since most of the NNRTIs also contain polar group(s), they have the potential to form hydrogen bonds with surrounding amino acid residues either directly or via water bridges [33–36]. In the structures of both liganded HIV-1 RT and the HIV-1 RT/DNA/Fab complex, there are two small surface depressions at the base of the NNIBP that are the putative entrances to the pocket [34,43]. One surface depression is located at the p66/p51 heterodimer interface and is composed of amino acid residues Leu100, Lys101, Lys103, Val179, Tyr181, and Tyr188 of p66, and Glu138 of p51 [34]. This putative entrance is narrow compared to the size of the NNRTIs. Another surface depression has been found at the location near the base of the p66 thumb subdomain between two adjacent structural elements: the β5–β6 connecting loop (Lys101 and Lys103) and the β13–β14 hairpin (Pro236 and Leu238) [43]. Since this site is also exposed to solvent, an NNRTI could approach the NNIBP from it. How-

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_58.html (2 of 2) [4/5/2004 4:50:15 PM]

Document

Page 59

ever, once an NNRTI is bound to RT, only the first putative entrance remains accessible; the second disappears due to the conformational change and repositioning of the β12-β13-β14 sheet [43]. It is evident, from comparison of the various HIV-1 RT structures that the NNIBP has a highly flexible structure that apparently allows the enzyme to accommodate various types of NNRTIs with different shapes and sizes. Despite apparent differences in the structures of the bound inhibitors, comparison of structures of several HIV-1 RT/NNRTI complexes revealed remarkable similarity in the geometry of both the bound inhibitors and the NNIBP [33,35]. All these chemically diverse NNRTIs assume a strikingly similar butterfly-like shape (Figure 6). The binding of NNRTIs in the NNIBP can be likened to a butterfly sitting on the β6-β10-β9 sheet and facing toward the putative entrance to the pocket. The angle between the two wings of the “butterfly” is approximately 112–115° in the TIBO, α-APA, and nevirapine complexes [35]. This angle might be critical in inhibitor binding and could be a crucial parameter in the design of new NNRTIs. There are many other NNRTIs that are significantly larger or smaller in size than α-APA, TIBO, or nevirapine. It is very likely that the NNIBP can adopt other conformations. For example, BHAP appears to be too large to fit into the NNIBP in any of the reported HIV-1 RT/NNRTI complexes. The NNIBP in the HIV-1 RT/BHAP complex would need to be significantly larger than that observed in the structures of the known HIV-1 RT/NNRTI complexes. It is possible that the BHAP inhibitor may not conform to a butterfly-like shape. This underscores the importance of solving crystal structures for as many HIV-1 RT/NNRTI complexes as possible. Additional structural and biochemical data for other HIV-1 RT/NNRTI complexes should provide the insight needed to define the limits of the flexibility of HIV-1 RT in the NNIBP region. VII. Process of NNRTI Binding In crystal structures of unliganded HIV-1 RT [40,41,43] and of HIV-1 RT/DNA/Fab complex [38], the NNIBP does not exist (although a small cavity is found in the region of the NNIBP proximal to the polymerase active site in the unliganded HIV-1 RT structure described by Esnouf et al. [42]). In these structures, the side chains of Tyr181 and Tyr188 in p66 point away from the polymerase active site and toward the hydrophobic core. However, in the HIV-1 RT/NNRTI complex structures, the side chains of Tyr181 and Tyr188 point toward the polymerase active site, and the side chain of Tyr181 is in a position that prevents Trp229 from occupying the position it has in the unliganded or DNA bound HIV-1 RT structures. Binding an NNRTI also moves the β12-β13-β14 sheet away from the hydrophobic core [34,35,37]. These conformational

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_59.html [4/5/2004 4:50:16 PM]

Document

Page 60

changes create the space in the pocket required to accommodate inhibitors. In other words, significant conformational changes occur during the process of inhibitor binding that lead to the formation of the NNIBP [33–35]. This observation also underscores the importance of determining structures of HIV-1 RT with and without bound inhibitors. An obvious question is, what forces initiate this series of conformational changes during NNRTI binding? One possibility is the contacts between the inhibitor and the protein. Though the NNIBP is hydrophobic, there are three hydrophilic amino acid residues (Lys101 and Lys103 of p66, and Glu138 of p51) at the rim of the putative entrance(s) to the pocket. The flexible and polar side chains of these residues could assist in steering an inhibitor into the pocket and/or could block the bound inhibitor from escaping out of the pocket. Mutagenesis studies have shown that these three residues are important in the binding of NNRTIs. Though the importance could be explained in terms of the interactions between these residues and the bound inhibitor in the final complexes, interactions at the initial stages of inhibitor binding might also be crucial. The flexible and polar side chains of these residues might help in directing the inhibitor toward the entrance to the pocket via electrostatic interactions, in part by replacing the original hydrogen bonds between the drug and the solvent molecules. Any initial energy gains from such polar interactions could potentially be replaced by hydrogen bonds or other types of interactions between the inhibitor and alternative residues as the inhibitor moves deeper into the binding pocket. In addition, significant portions of the aromatic rings of both Tyr181 and Tyr188 are exposed at the bottom of the surface depression and offer the potential for early π-π interactions with the inhibitor. This type of π-π interaction might also play an important role in the initial approach of inhibitors to the binding pocket. This hypothesis may provide a kinetic explanation for the ineffectiveness of NNRTIs against viral strains of HIV-1 that carry nonaromatic amino acids at positions 181 and 188. As the solvated inhibitor approaches the enzyme and proceeds to enter the binding pocket, most of the water molecules of solvation are lost. The few water molecules that remain in the NNRTI-bound complex are typically located at the entrance to the pocket, forming water bridges between the inhibitor and one or two polar residues around the entrance [33,35,36]. Once the inhibitor is in place, the surface residues close down around the drug preventing it from escaping by effectively sealing the entrance to the pocket. VIII. Mechanisms of Inhibition by NNRTIS Based on structural, biochemical, and genetic data several hypotheses have been postulated about the mechanism(s) of inhibition of HIV-1 RT by NNRTIs. It is

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_60.html [4/5/2004 4:50:18 PM]

Document

Page 61

now clear that the binding of NNRTIs provokes substantial conformational changes in both secondary structural elements and in side chains of residues in the NNIBP. These conformational changes in the NNIBP could directly or indirectly affect the precise geometry and/or mobility of the nearby polymerase catalytic site, especially the highly conserved YMDD motif and/or the divalent metal ions [31,33,34,42,68]. The binding of NNRTIs appears to lock the flexible hinge-like structure between the palm and thumb subdomains and restrict mobility of the thumb subdomain, placing constraints on the geometry of the DNA-binding cleft [12,31,34,43]. The “primer grip” (i.e., β12-β13-β14 sheet), which has close interactions with the 3'-terminus of the primer strand [38], forms a part of the NNIBP and is involved in the binding of NNRTIs. It has become apparent that binding of NNRTIs can substantially alter the conformation of the primer grip; this could affect the precise positioning of the primer strand relative to the polymerase active site [34,37]. Displacement of the primer grip by NNRTI binding could lead to repositioning of the primer terminus. This could explain the observation that dNTP binding is largely unaffected by NNRTI binding while the rate of the chemical step of DNA polymerization is reduced [77]. Long-range distortions of the HIV-1 RT structure by NNRTI binding can potentially account for NNRTI inhibition of polymerization [39,41,43] and alteration of RNase H cleavage specificity [43,78]. These possible mechanisms are not mutually exclusive and the binding of inhibitors might have multiple influences on HIV-1 RT polymerization. The exact mechanism(s) of inhibition is still under investigation. IX. NNRTI-Resistance Mutations Analyses of the crystal structures of HIV-1 RT complexed with various NNRTIs have indicated that amino acid residues whose mutations confer high levels of resistance to NNRTIs [9,11,12,26,27] are located close to the bound inhibitors (Figure 5 and Table 2). Subunit-specific mutagenesis studies have confirmed that mutations that confer resistance to the NNRTIs act directly through the change in the NNIBP itself [60,79]. In these studies, recombinant HIV-1 RTs that contained amino acid substitutions only in the p66 subunit were resistant to NNRTIs, while those containing the same amino acid substitutions only in the p51 subunit remained susceptible to the drugs. There is one exception: the amino acid substitution of Glu138 to Lys, which confers resistance to inhibitors only when it is present in the p51 subunit. Amino acid residue 138 is located in the β7–β8 connecting loop of the fingers subdomain. In the p51 subunit this residue forms a part of the NNIBP, while its counterpart in the p66 subunit is far away from the pocket [12,60].

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_61.html [4/5/2004 4:50:22 PM]

Document

Page 62

The mechanism(s) of resistance may depend on the specific amino acid change. It is likely that most NNRTI-resistance mutations exert their effects by altering interactions between protein side chains and the inhibitors [12,34,35]. Drug-resistance mutations that result in a decrease or increase in the size of side chains might lead to loss of favorable contacts or steric conflicts with bound inhibitors. Mutations that alter the local electrostatic potential, i.e., gain, loss, or inversion of charge, may change the affinity of the NNIBP for inhibitor binding. These altered interactions could interfere with the binding of NNRTIs to the hydrophobic pocket or conceivably could even relax the geometric distortion that the binding of an inhibitor causes in the vicinity of the polymerase active site. X. Design of Improved NNRTIs Different NNRTIs, even from the same class of compounds, show remarkable variations in their ability to inhibit HIV-1 replication and can give rise to different spectra of resistance mutations [9,11,26,27]. For example, biochemical studies showed that the 8-chloro TIBO derivative R86183 is quite potent in inhibiting an HIV-1 strain containing the Tyr181Cys mutation, which is one of the frequently occurring HIV-1 RT mutations that gives rise to a high level of resistance to almost all NNRTIs, including other TIBO derivatives [80]. There are several other reports of NNRTIs that are also relatively effective in inhibiting the HIV-1 RT Tyr181Cys variant [81–84]. These results suggest that although all the inhibitors appear to bind in the NNIBP, there are differences in their specific interactions with HIV-1 RT. Structural analyses of HIV-1 RT/NNRTI complexes and computer modeling studies confirmed that the exact conformations of the amino acid residues forming the NNIBP appear to vary in different complexes and that there are specific interactions between individual inhibitor and surrounding residues [33,35,36,85]. However, these differences have not been sufficiently large to allow a successful combination therapy to be developed using two or more of the currently available NNRTIs (discussed in more detailed in a later section) [9,11,26,27,86]. Systematic analysis of wild-type and drug-resistant mutant HIV-1 RT structures in complexes with various NNRTIs should provide additional insights about constraints that could be used to optimize the design of NNRTIs. This knowledge could guide development of more effective inhibitors for AIDs treatment. As discussed earlier, the bound NNRTIs in HIV-1 RT complexes determined so far conform to a common butterfly-like shape (Figure 6). A close inspection of interactions between inhibitors and protein reveals that though

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_62.html [4/5/2004 4:50:38 PM]

Document

Page 63

most of the amino acid residues forming the pocket adjust their side chains to make close contacts with the inhibitor, the inhibitor is not sufficient to fill all of the space in the pocket. There is space for additional nonpolar, polar, or charged groups. Modification of the inhibitor would result in adjustment of the orientation of the side chains and could improve interactions between the inhibitor and surrounding residues such as Leu100, Lys101, Lys103, Val106, or Leu234. Inhibitors designed to have more extensive interactions with essential elements in the pocket should minimize the chances of selecting resistant HIV-1 RT variants. From this point of view, NNRTIs that interact with the relatively conserved residues of the pocket, such as Trp229, Leu234, and Tyr318, may reduce the risk of encountering resistance mutations that do not have significant costs for the enzyme. In addition, compounds could be designed to contain functional groups (for example charged or polar groups) able to fill more of the available space of the NNIBP and also capable of specific hydrophilic interactions with the polar or charged side chains and/or with polypeptide backbone atoms of the NNIBP (for example the main chain amide nitrogens and carbonyl oxygens). The hydrophilic interactions between inhibitors and protein backbone atoms should be advantageous because mutations to any amino acid other than proline would not affect such contacts. In the structures of HIV-1 RT/NNRTI complexes, the bound inhibitors are located very close to the polymerase active site composed of the three catalytically essential aspartic acids Asp110, Asp185, and Asp186. It might be useful to design compounds that have a long and branched aliphatic group or a substituted aromatic group that could not only produce hydrophobic interactions with Tyr181, Tyr188, and Trp229, but could also be able to interact with the three aspartic residues or interfere with the metal ion(s) binding at the polymerase active site. XI. RNase/H-Active Site as a Potential Drug Target Site HIV-1 RT contains RNase H, which is responsible for degradation of viral RNA and removal of RNA primers for minus- and plus-strand DNA synthesis (see reviews [87–89]). The absolute requirement for virus-associated RNase H function [90–93] offers an additional target for antiretroviral drugs. The RNase H domain of HIV-1 RT is located at the C-terminus of the p66 subunit (Figures 2 and 3). In contrast to the polymerase domain of HIV-1 RT, the structure of the RNase H domain is quite similar in all known HIV-1 RT structures and conforms quite well with the structure of the isolated HIV-1 RNase H domain [94–95]. The relative stability of the structure of the RNase H domain suggests that the RNase H active site could be a relatively well-defined target for drug

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_63.html [4/5/2004 4:50:40 PM]

Document

Page 64

design. Mutagenesis studies have demonstrated the interdependence of DNA polymerase and RNase H activities. Mutations that disrupt one of the two enzymatic activities of HIV-1 RT often also impair the second activity [96–99]. Indeed, according to the crystal structure of HIV-1 RT, the polymerase active site and the RNase H active site are separated by approximately 17–18 nucleotides [38] and the RNase H domain has many contacts with the polymerase domain, especially with the connection subdomain of p66 and the thumb and connection subdomains of p51 [31,38,100]. Interactions between the polymerase domain and nucleic acid can modulate RNase H activity. Because the predominant contacts of HIV-1 RT with template-primer occur in the vicinity of the polymerase active site, precise placement of the template strand relative to the RNase H active site may be regulated by the sequence and composition of the template-primer. Mutagenesis experiments showed that mutations located at or near the “template grip” in the polymerase domain of HIV-1 RT can have a greater effect on RNase H than on polymerase activity [99,101,102]. It was also reported that binding of the NNRTI nevirapine alters the cleavage specificity of RNase H [78]. Structural distortions in the position and conformation of template-primer induced by NNRTI-binding may account for alteration of the cleavage specificity of RNase H [43]. Divalent metal ions such as Mg2+ or Mn2+ are essential for the RNase H activity [103–106]. The structure of the isolated RNase H domain crystallized in the presence of MnCl2 revealed two tightly bound Mn2+ ions in close proximity to four catalytically essential acidic residues, Asp443, Glu478, Asp498, and Asp549, that form the active site [94]. Biochemical data have shown that mutations of these conserved residues could either disrupt RNase H activity or lead to a highly unstable enzyme [107–109]. Based on the crystal structures, a two-metal ion-dependent catalytic mechanism for RNase H activity has been postulated [101], which is similar to that proposed for phosphoryl transfer reactions catalyzed by polymerases and their associated nucleases [67,69–71]. In contrast, in the structure E. coli RNase H reported by Katayanagi et al. [111] only one Mg2+ ion was observed, and that led to the proposal of a single metal-ion catalyzed hydrolysis [112]. Interestingly, in the structure of unliganded HIV-1 RT reported by Rodgers et al. [41] and Hsiou et al. [43] only one Mg2+ ion was found at the RNase H active site. The mechanism of RNase H cleavage and the exact role of metal ion(s) in the hydrolysis and formation of phosphodiester bonds are still under investigation (see review [89]). Very few inhibitors specifically target HIV-1 RNase H activity. Illimaquinone, a natural marine product, was shown to preferentially inhibit the HIV-1 RNase H activity [113,114]. However, this compound appears to react with a sulfhydryl group in the polymerase domain and not with RNase H itself. It may be possible to use the available information on structural and biochemi-

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_64.html [4/5/2004 4:50:41 PM]

Document

Page 65

cal properties of the polymerase and RNase H of HIV-1 RT to design compounds that would bind at the RNase H active site or interfere with the metal ion(s) binding and inhibit the RNase H activity of HIV-1 RT. However, for optimal utility, these compounds should selectively inhibit the RNase H activity of HIV-1 and not the RNase H activity of the host cells. XII. Other Possible Target Sites for HIV-1 RT Inhibitors Both polymerase and RNase H activities of HIV-1 RT require that the enzyme be in a dimeric form [115–118] (Figure 3). The exact role(s) of the p51 subunit in the enzymatic activities of HIV-1 RT are not yet known. The three-dimensional structure of HIV-1 RT shows that the interface between p66 and p51 primarily involves interactions between the p66 palm and the p51 fingers subdomains, between the p66 connection and the p51 connection and fingers subdomains, and between the RNase H and the p51 thumb and connection subdomains [34,100,119] (Figure 3). A compound that would interfere with dimerization would be a potential candidate for an anti-AIDS drug. As discussed earlier, the flexibility of HIV-1 RT permits the enzyme to adopt different conformations. In the absence of bound DNA, the thumb and the fingers subdomains come together and close a major portion of the DNA-binding cleft [40,41,43]. Synthetic oligonucleotides that could interact with the specific or conserved regions of the DNA-binding cleft could potentially block binding of templateprimer substrates. An RNA pseudoknot has been reported to bind and specifically inhibit HIV-1 RT [120]. Chemical modification and substitution of specific groups in RNA ligands can change the structure of the pseudoknot, which could result in considerably more effective pseudoknot inhibitors with high binding specificity [121]. Studies employing the phosphorodithioate analogs of the primer sequence recognized by HIV-1 RT showed that these compounds can act as inhibitors and that inhibition is a function of both the sequence and length of these novel single-stranded nucleic acid oligomers [122,123]. A series of natural products, i.e., trihydroxyquinolone compounds isolated from Red Sea marine organisms, were reported to inhibit the DNA polymerase activity of HIV-1 RT [124,125]. This type of inhibitor appears to have a mechanism of inhibition that is different from either the NRTI inhibition mechanism or the NNRTI inhibition mechanism. The inhibition is reversible and noncompetitive with respect to both dNTP and template-primer [125]. This result indicates that there are other potential binding sites for inhibitors of HIV-1 RT.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_65.html [4/5/2004 4:50:43 PM]

Document

Page 66

XIII. Useful Tools in Structure-Based Drug Design Several computer modeling algorithms have been developed for structure-based drug design. Among them, DOCK [73–75] and 3D SEARCH [126] have been successfully applied in the design of HIV-1 protease inhibitors. These programs search a target protein for invaginations, grooves, and recognition surfaces that could bind a potential receptor molecule. Compounds complementary to the putative receptor binding site in both shape and chemical properties can be identified through searching databases of small molecules, such as the Cambridge Crystallographic Database, the Fine Chemicals Directory, or other commercially available databases. An important issue in analyzing HIV-1 RT is the flexibility of the enzyme. Comparison of structures of unliganded HIV-1 RT and NNRTI-bound HIV-1 RT complexes has shown that the NNIBP is not present in the unliganded form [34,41,43]. This underscores the importance of searching both the unliganded HIV-1 RT and the HIV-1 RT complexes with inhibitors and substrates in order to identify any potential inhibitor-binding sites. Many other approaches have been and are being developed for computeraided design of inhibitors. For example, pharmacophore analysis can identify the spatial arrangement of groups or atoms common to all active inhibitor molecules and then incorporate these elements into a single molecule [127,128]. Detailed analysis of the volumes occupied by different inhibitors bound to the same binding site could also provide new suggestions for inhibitor design. For example, the volume union of all known NNRTIs such as nevirapine, TIBO, α-APA, HEPT, and 1051U91 can be calculated. This type of analysis could be used to screen for new NNRTIs. Since the coordinates for a number of HIV-1 RT/NNRTI complex structures are now available in the Protein Data Bank, these approaches can be applied to the design of new or improved NNRTIs. Given the relatively high flexibility of the NNIBP region and the diversity of NNRTI structures, the NNIBP of HIV-1 RT could be a methodologically challenging yet extremely important target for structure-based drug design. XIV. Enzymatic Efficiency of Drug-Resistant HIV-1 RT Variants Analyses of viral population dynamics indicated that, although drug resistance cannot be seen as a positive outcome of chemotherapy, clinical progress can be made through the development of drugresistant viral variants (see review [30]).alyzed hy Arial H itself. It may be possible to use the available information on structural and biochemi-

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_66.html [4/5/2004 4:50:44 PM]

Document

Page 67

Biochemical data show that HIV-1 replicates extremely rapidly in infected individuals and that the viral load is low in the early stages of the disease because the host immune system is initially successful in limiting viral replication [28,29]. When patients are treated with either RT and/or protease inhibitors, wild-type HIV-1 is rapidly replaced with drug-resistant variants. In fact, even in patients who have not received treatment with any anti-RT drugs, HIV-1 variants that contain residues corresponding to both NRTI- and NNRTI-resistance mutations in RT can be found as minor components of the viral population [129]. Similarly, viral variants that contain residues in protease sequence corresponding to protease inhibitor drug-resistance mutations have also been observed in patients prior to drug therapy (see review [130]). Enzymatic components found in a wild-type virus, such as RT or protease, are optimized for efficient viral replication [30]. In the absence of selective pressure (drug), the wild-type virus has a fitness advantage over drug-resistant viral variants. However, in the presence of drugs, drugresistant variants have a fitness advantage over the wild-type because the drug impairs efficiency of the target enzyme in the wild-type virus [30]. Binding of an NRTI or an NNRTI to wild-type HIV-1 RT interferes with the polymerization reaction. However, the presence of resistant variants in the population allows the virus to escape, and the variants to rapidly replace the wild-type virus. Nevertheless, this escape has a price. When the optimized wildtype virus is replaced by the less fit drug-resistant variants, the relative fitness of the virus decreases. In other words, the enzymatic efficiency of a drug-resistant HIV-1 RT variant is impaired relative to the wild-type enzyme (see review [131]). If the enzymatic efficiency of a drug-resistant viral variant is sufficiently impaired, the replication of the variant virus would be significantly decreased. Thus, an antiviral drug will be useful not because it would completely stop the growth of HIV-1 but because it selects viral variants whose replication is significantly impaired. Positive clinical benefit results from the fact that the viral load is decreased owing to reduced replication of the variant virus. As predicted by this model, some HIV-1 RT and protease inhibitors seem to select for relatively less fit drug-resistant variants. For example, treatment of HIV-1 infection with HBY 097, a quinoxaline inhibitor, induces development of an HIV-1 RT variant containing the Gly190Glu mutation that appears to have substantially decreased polymerase activity and replicates relatively slowly [22,84]. Replacement of the hydrogen atom of Gly190 with an acidic side chain of Glu190 in the hydrophobic NNIBP apparently interferes with the stability of the enzyme as well as the ability of the NNIBP to bind a hydrophobic inhibitor. The relative inefficiency of HIV-1 variant containing the Gly190Glu mutation in RT can be viewed as a positive outcome of the selection pressure provided by this particular inhibitor. However, most of the HIV-1 variants selected by

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_67.html [4/5/2004 4:50:46 PM]

Document

Page 68

currently available antiviral agents are not significantly less fit than the wild-type virus and the clinical benefits are not obvious. In this regard, new drugs should be designed that would be intended to select HIV-1 RT variants that are significantly less fit and do not replicate efficiently. XV. Combination Therapy Using Multiple ANTI-HIV-1 Drugs Monotherapy using either NRTIs or NNRTIs has led to the emergence of drug-resistant viral strains of HIV-1. Though many drug-resistance mutations confer cross-resistance to other inhibitors belonging to the same class, there are indications that some mutations conferring resistance to certain inhibitors are incompatible (see reviews [5,11,131]). A multidrug clinical trial with HIV-1 infected patients has shown that AZT resistance can be reversed by mutations that confer resistance to ddI [8]. The Leu74Val mutation appears to suppress the effects of the Thr215Tyr mutation that confers resistance to AZT [8,27]. The Met184Val mutation, which causes resistance to 3TC or other oxathiolane-cytosine analogs, also appears to reverse the effects of the AZT-resistance mutations [27]. Recent clinical studies have shown that a combination of AZT and 3TC led to a considerable decrease in viral load and a substantial increase of CD4 cells when compared with monotherapy using AZT alone, even after emergence of the Met184Val mutation [132]. Another example is the Pro236Leu mutation that confers resistance to BHAP. The sensitivity of this HIV-1 RT variant to TIBO, nevirapine, and pyridinone is increased ten fold [133]. Although the NRTIs and NNRTIs target two distinct binding sites of HIV-1 RT and lead to different sets of resistance mutations, some of the NRTI- and NNRTI-resistance mutations also appear to be incompatible. For example, the NNRTI-resistance mutations Leu100Ile and Tyr181Cys have been shown to suppress the effects of some AZT-resistance mutations [11,134]. This has led to the suggestion that a combination of anti-HIV-1 drugs would be more effective in inhibiting HIV-1 replication than using individual drugs alone. In fact, both clinical and in vitro studies have shown that combination therapy has considerable advantages over monotherapy. At least in some cases, the effectiveness of the therapy increases with an increase in the number of drugs in the combination [5,135]. Combination therapy may, in addition to increasing antiviral activity, also slow emergence of drug-resistant variants and may have the added benefit that reducing the dosage of individual drugs can reduce toxicity. It is generally believed that synergistic drug interactions arise from the fact that certain combinations of drugresistance mutations are particularly detrimental for the enzyme (and, by extension, the virus). This has focused attention on

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_68.html [4/5/2004 4:50:47 PM]

Document

Page 69

determining the mechanism(s) underlying drug resistance and, from this understanding, to devise ways for identifying combinations of drugs which might provoke drug-resistance mutations incompatible with viral survival. Several protocols have been designed for combination therapy using a variety of anti-HIV-1 drugs (see reviews [5,11]). Combinations of different drugs that interact with the same binding site of the same viral protein but lead to mutually antagonistic or suppressive resistance mutations have been studied extensively, especially for the combined uses of different but structurally related NRTIs (see for example [136–138]) or NNRTIs [139–141]. Combinations of drugs or inhibitors that target different sites of the same viral protein, primarily the combination of NRTIs and NNRTIs of HIV-1 RT, show enhanced inhibition of HIV-1 RT polymerase activity and suppression of the emergence of drugresistance mutations (for example [142–146]). Experiments have also been conducted with combinations of drugs that target different viral proteins, e.g., inhibitors of virus adsorption, virus-cell fusion, and/or uncoating proteins have been tested in combination with protease inhibitors and/or RT inhibitors. Combinations of AZT with the glycosylation inhibitor castanospermine [147], or with the Tat inhibitor Ro 24-7429 [148], or with the protease inhibitor Ro 31-8959 [149] have been shown to potently inhibit HIV-1 viral replication in vitro. Combination therapy can increase the effectiveness of inhibition and significantly impair efficiency of viral replication. However, both NRTI- and NNRTI-resistance mutations can affect the positioning of the nucleic acid and/or the overall structure of HIV-1 RT [23]. These two sets of resistance mutations can communicate with each other and can result in cross resistance. Moreover, new drug-resistance mutations that confer cross-resistance to both NRTIs and NNRTIs can be selected, which reduce the effectiveness of some drug combinations (see reviews [5,11,26]). Biochemical studies showed that both HIV-1 RT mutants [150] and viral variants [151] could be obtained that are resistant to the combination of AZT, ddI, and nevirapine. In clinical trials, treatment with AZT and ddI or AZT and ddC led to a different spectrum of NRTI-resistance mutations [152,153]. The most notable of these new mutations is Gln151Met, which is located at a position close to the dNTP-binding site. Structural analysis of the HIV1 RT/DNA/Fab complex suggests that the side chain of Gln151 in the wild-type enzyme may interact with the first unpaired template nucleotide. The side chain of this residue may play a role in selecting the correct base for the incoming nucleotide [72]. Since the RT mutant containing only the Gln151Met mutation can confer high-level resistance to a number of NRTIs, including AZT, ddI, and ddC, it is not clear why this mutation did not emerge in monotherapy of these NRTIs. However, Gln151 is relatively well conserved and mutations at this position may have an unfavorable impact on HIV-1 RT.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_69.html [4/5/2004 4:50:49 PM]

Document

Page 70

XVI. Perspective Substantial progress has been made in understanding the structure and function of HIV-1 RT and in the development of anti-HIV-1 inhibitors. However, the genetic flexibility of HIV-1 will continue to make development of a truly effective antiviral therapy for AIDS an exceptionally difficult task. We are beginning to understand how to circumvent drug resistance. The accumulated evidence has shown that the ability of the virus to develop drug resistance is limited and that the drug-resistant viral variants are less efficient than the wild-type virus. If the selection pressure provided by antiviral drugs makes the virus pay a sufficiently high price, then the viral load can be decreased and there will be a measurable clinical benefit. Based on a better understanding of the structure-function relationships of HIV-1 RT, we are now coming to grips with the mechanisms of polymerization, drug inhibition, and drug resistance. This information should make it possible to develop new or improved HIV-1 RT inhibitors that have different properties and provoke different patterns of drug-resistance mutations. Though it is likely that there will be no single drug which would be effective against all HIV-1 variants, we have reasons to believe that new or improved drugs or, more likely, new drug combinations, will be designed that are broadly effective against all of the HIV-1 variants that can grow efficiently. Detailed analysis of the conformational changes among the various HIV-1 RT structures may reveal additional sites (in addition to the currently known NRTI- and NNRTI-binding sites) for binding new inhibitors able to interfere with the polymerization and/or the flexibility of the enzyme required for its activity. The considerable physical and genetic flexibility of HIV-1 RT suggests that more effective anti-RT drugs should be designed to target the conserved portions of HIV-1 RT that the virus cannot easily afford to change. Such conserved elements can be identified by comparing the sequences of RTs from different retroviruses; the functions and relative importance of these conserved elements can be determined by mutagenesis and biochemical and structural analyses. It is our hope that application of structure-based drug design strategies may aid in the development of novel HIV-1 RT inhibitors for a more effective treatment of HIV-1 infection. Acknowledgments We thank the other members of the Arnold and Hughes laboratories and our collaborators for their helpful discussions and assistance, including Koen Andries, Gail Ferstandig Arnold, Paul Boyer, Arthur Clark, Jr., Paul Janssen, Jörg-Peter Kleim, Luc Koymans, Tack Kuntz, Karen Lentz, Chris Michejda, Henri Moereels, Manfred Roesner, Marilyn Kroeger Smith, Rick Smith, Jr., and

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_70.html [4/5/2004 4:50:51 PM]

Document

Page 71

Chris Tantillo. The work in Edward Arnold's laboratory has been supported by Janssen Research Foundation and NIH grants (AI 27690 and AI 36144). Research in Stephen Hughes' laboratory is sponsored in part by the National Cancer Institute, DHHS, under contract with ABL, and by NIGMS. References 1. Goff SP. Retroviral reverse transcriptase: synthesis, structure, and function. J Acquired Immune Deficiency Syndromes 1990; 3:817–831. 2. Jacobo-Molina A, Arnold E. HIV reverse transcriptase structure-function relationships. Biochemistry 1991; 30:6351–6361. 3. Whitcomb JM, Hughes SH. Retroviral reverse transcription and integration: progress and problems. Ann Rev Cell Biol 1992; 8:275–306. 4. Le Grice SFJ. Human immunodeficiency virus reverse transcriptase. In: Skalka AM, Goff SP, eds. Reverse Transcriptase. Plainview, New York: Cold Spring Harbor Laboratory Press, 1993:163–191. 5. De Clercq E. Toward improved anti-HIV chemotherapy: therapeutic strategies for intervention with HIV infections. J Med Chem 1995; 38:2491–2517. 6. De Clercq E. HIV inhibitors targeted at the reverse transcriptase. AIDS Res Human Retroviruses 1992; 8:119–134. 7. Larder BA. Inhibitors of HIV reverse transcriptase as antiviral agents and drug resistance. In: Skalka AM, Goff SP, eds. Reverse Transcriptase. Plainview, New York: Cold Spring Harbor Laboratory Press, 1993:205–222. 8. St. Clair MH, Martin JL, Tudor-Williams G, Bach MC, Vavro CL, King DM, et al. Resistance to ddI and sensitivity to AZT induced by a mutation in HIV-1 reverse transcriptase. Science 1991; 253:1557–1559. 9. Richman DD. Resistance of clinical isolates of human immunodeficiency virus to antiretroviral agents. Antimicrob Agents Chemother 1993; 37:1207–1213. 10. Schinazi RF. Competitive inhibitors of human immunodeficiency virus reverse transcriptase. Perspectives in Drug Discovery and Design 1993; 1:151–180. 11. De Clercq E. HIV resistance to reverse transcriptase inhibitors. Biochem Pharmacol 1994; 47:155–169.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_71.html (1 of 2) [4/5/2004 4:50:53 PM]

Document

12. Tantillo C, Ding J, Jacobo-Molina A, Nanni RG, Boyer PL, Hughes SH, et al. Locations of antiAIDS drug binding sites and resistance mutations in the three-dimensional structure of HIV-1 reverse transcriptase: implications for mechanisms of drug inhibition and resistance. J Mol Biol 1994; 243:369–387. 13. Miyasaka T, Tanaka H, Baba M, Hayakawa H, Walker RT, Balzarini J, et al. A novel lead for specific anti-HIV-1 agents: 1-[(2-hydroxyethoxy)methyl]-6-(phenylthio)thymine. J Med Chem 1989; 32:2507–2509. 14. Pauwels R, Andries K, Desmyter J, Schols D, Kukla MJ, Breslin HJ, et al. Potent and selective inhibition of HIV-1 replication in vitro by a novel series of TIBO derivatives. Nature 1990; 343:470–474. 15. Merluzzi VJ, Hargrave KD, Labadia M, Grozinger K, Skoog M, Wu JC, et al. Inhibition of HIV-1 replication by a nonnucleoside reverse transcriptase inhibitor. Science 1990; 250:1411–1413.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_71.html (2 of 2) [4/5/2004 4:50:53 PM]

Document

Page 72

16. Goldman ME, Nunberg JH, O'Brien JA, Quintero JC, Schleif WA, Freund KF, et al. Pyridinone derivatives: specific human immunodeficiency virus type 1 reverse transcriptase inhibitors with antiviral activity. Proc Natl Acad Sci USA 1991; 88:6863–6867. 17. Romero DL, Busso M, Tan C-K, Reusser F, Palmer JR, Poppe SM, et al. Nonnucleoside reverse transcriptase inhibitors that potently and specifically block human immunodeficiency virus type 1 replication. Proc Natl Acad Sci USA 1991; 88:8806–8810. 18. Chimirri A, Grasso S, Monforte A-M, Monforte P, Zappala M. Anti-HIV agents. I. Synthesis and in vitro anti-HIV evaluation of novel 1H,3H-thiazolo[3,4-a]benzimidazoles. Farmaco 1991; 46:817–823. 19. Chimirri A, Grasso S, Monforte A-M, Monforte P, Zappala M. Anti-HIV agents. II. Synthesis and in vitro anti-HIV activity of novel 1H3H-thiazolo[3,4-a]benzimidazoles. Farmaco 1991; 46:925–933. 20. Balzarini J, Perez-Perez M-J, San-Felix A, Schols D, Perno C–F, Vandamme A–M, et al. 2',5'-bis-O(tert-butyldimethylsilyl)-3'-spiro-5''-(4''-amino-1",2"-oxathiole-2",2"-dioxide)pyrimidine (TSAO) nucleoside analogues: highly selective inhibitors of human immunodeficiency virus type 1 that are targeted at the viral reverse transcriptase. Proc Natl Acad Sci USA 1992; 89:4392–4396. 21. Pauwels R, Andries K, Debyser Z, Van Dable P, Schols D, Stoffels P, et al. Potent and highly selective human immunodeficiency virus type 1 (HIV-1) inhibition by a series of αanilinophenylacetamide derivatives targeted at HIV-1 reverse transcriptase. Proc Natl Acad Sci USA 1993; 90:1711–1715. 22. Kleim J–P, Bender R, Billhardt U-M, Meichsner C, Riess G, Roesner M, et al. Activity of a novel quinoxaline derivative against human immunodeficiency virus type 1 reverse transcriptase and viral replication. Antimicrob Agents Chemother 1993; 37:1659–1664. 23. Kleim J-P, Roesner M, Winkler I, Paessens A, Kirsch R, Hsiou Y, et al. Selective pressure of a quinoxaline class nonnucleoside inhibitor of human immunodeficiency virus type 1 (HIV-1) reverse transcriptase (RT) on HIV-1 replication results in the emergence of nucleoside RT inhibitor specific (RT L74->V/I, V75->L/I) HIV-1 mutants. Proc Natl Acad Sci USA 1996; 93:34–38. 24. Nunberg JH, Schleif WA, Boots EJ, O'Brien JA, Quintero JC, Hoffman JM, et al. Viral resistance to human immunodeficiency virus type 1-specific pyridinone reverse transcriptase inhibitors. J Virol 1991; 65:4887–4892. 25. Mellors JW, Dutschman GE, Im G-J, Tramontano E, Winkler SR, Cheng Y-C. Rapid emergence of HIV-1 resistant to non-nucleoside inhibitors of reverse transcriptase. Antiviral Research 1992; 17 S1:48. 26. Larder BA. Interactions between drug resistance mutations in human immunodeficiency virus type 1 reverse transcriptase. J Gen Virol 1994; 75:951–957.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_72.html (1 of 2) [4/5/2004 4:50:55 PM]

Document

27. Kimberlin DW, Coen, DM, Biron KK, Cohen JI, Lamb RA, McKinlay M, et al. Molecular mechanisms of antiviral resistance. Antiviral Res 1995; 26:369–401. 28. Wei X, Ghosh SK, Taylor ME, Johnson VA, Emini EA, Deutsch P, et al. Viral dynamics in human immunodeficiency virus type 1 infection. Nature 1995; 373:117–122. 29. Ho DD, Neumann AU, Perelson AS, Chen W, Leonard JM, Markowitz M. Rapid turnover of plasma virions and CD4 lymphocytes in HIV-1 infection. Nature 1995; 373:123–126.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_72.html (2 of 2) [4/5/2004 4:50:55 PM]

Document

Page 73

30. Coffin JM. HIV population dynamics in vivo: implications for genetic variation, pathogenesis, and therapy. Science 1995; 267:483–489. 31. Kohlstaedt LA, Wang J, Friedman JM, Rice PA, Steitz TA. Crystal structure at 3.5 Å resolution of HIV-1 reverse transcriptase complexed with an inhibitor. Science 1992; 256:1783–1790. 32. Smerdon SJ, Jager J, Wang J, Kohlstaedt LA, Chirino AJ, Friedman JM, et al. Structure of the binding site for nonnucleoside inhibitors of the reverse transcriptase of human immunodeficiency virus type 1. Proc Natl Acad Sci USA 1994; 91:3911–3915. 33. Ren J, Esnouf R, Garman E, Somers D, Ross C, Kirby I, et al. High resolution structures of HIV-1 RT from four RT-inhibitor complexes. Nature Struct Biol 1995; 2:293–302. 34. Ding J, Das K, Tantillo C, Zhang W, Clark AD Jr., Jessen S, et al. Structure of HIV-1 reverse transcriptase in a complex with the nonnucleoside inhibitor α-APA R 95845 at 2.8 Å resolution. Structure 1995; 3:365–379. 35. Ding J, Das K, Moereels H, Koymans L, Andries K, Janssen PAJ, et al. Structure of HIV-1 RT/TIBO R 86183 reveals similarity in the binding of diverse nonnucleoside inhibitors. Nature Struct Biol 1995; 2:407–415. 36. Ren J, Esnouf R, Hopkins A, Ross C, Jones Y, Stammers D, et al. The structure of HIV-1 reverse transcriptase complexed with 9-chloro-TIBO: lessons for inhibitor design. Structure 1995; 3:915–926. 37. Das K, Ding J, Hsiou Y, Clark AD Jr., Moereels H, Koymans L, et al. Crystal structures of 8-Cl and 9-Cl TIBO complexed with wild-type HIV-1 RT and 8-Cl TIBO complexed with the Tyr181Cys HIV-1 RT drug-resistant mutant. J. Mol Biol 1996; 264:1085–1100. 38. Jacobo-Molina A, Ding J, Nanni RG, Clark AD Jr., Lu X, Tantillo C, et al. Crystal structure of human immunodeficiency virus type 1 reverse transcriptase complexed with double-stranded DNA at 3.0 Å resolution shows bent DNA. Proc Natl Acad Sci USA 1993; 90:6320–6324. 39. Jager J, Smerdon S, Wang J, Boisvert DC, Steitz TA. Comparison of three different crystal forms shows HIV-1 reverse transcriptase displays an internal swivel motion. Structure 1994; 2:869–876. 40. Raag R, Clark AD Jr., Ding J, Jacobo-Molina A, Lu X, Nanni RG, et al. 3.0 Å crystal Structure of HIV-1 reverse transcriptase without dsDNA reveals largescale motion of p66 thumb subdomain. Am Crystallogr Assoc Mtg Abstr, Ser 2 1994; 18:44. 41. Rodgers DW, Gamblin SJ, Harris BA, Ray S, Culp JS, Hellmig B, et al. The structure of unliganded reverse transcriptase from the human immunodeficiency virus type 1. Pro Natl Acad Sci USA 1995; 92:1222–1226.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_73.html (1 of 2) [4/5/2004 4:53:09 PM]

Document

42. Esnouf R, Ren J, Ross R, Jones Y, Stammers D, Stuart D. Mechanism of inhibition of HIV-1 reverse transcriptase by non-nucleoside inhibitors. Nature Struct Biol 1995; 2:303–308. 43. Hsiou Y, Ding J, Das K, Clark AD Jr., Hughes SH, Arnold E. Structure of unliganded HIV-1 reverse transcriptase at 2.7 Å resolution: implications of conformational changes for polymerization and inhibition mechanisms. Structure 1996; 4:853–860. 44. Unge T, Knight S, Bhikhabhai R, Lovgren S, Dauter Z, Wilson K, et al. 2.2 Å resolution structure of the amino-terminal half of HIV-1 reverse transcriptase (fingers and palm subdomains). Structure 1994; 2:953–961.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_73.html (2 of 2) [4/5/2004 4:53:09 PM]

Document

Page 74

45. Majumadar C, Abbotts J, Broder S, Wilson SH. Studies on the mechanism of human immunodeficiency virus reverse transcriptase: Steady-state kinetics, processivity and polynucleotide inhibiton. J Biol Chem 1988; 263:15657–15665. 46. Kati WM, Johnson KA, Jerva LF, Anderson KS. Mechanism and fidelity of HIV reverse transcriptase. J Biol Chem 1992; 267:25988–25997. 47. Balzarini J, Herdewijn P, De Clercq E. Differential patterns of intracellular metabolism of 2',3'didehydro-2',3'-dideoxthymidine and 3'-azido-2',3'- dideoxythymidine: two potent anti-human immunodeficiency virus compounds. J Biol Chem 1989; 264:6127–6133. 48. Balzarini J, De Clercq E. 5' -Phosphoribosyl 1-pyrophosphate synthetase converts the acyclic nucleoside phosphonates 9-(3-hydroxy-2-phosphonyl-methoxypropyl) adenine and 9-(2phosphonylmethoxyethyl)adenine directly to their antivirally active diphosphate derivatives. J Biol Chem 1991; 266:8686–8689. 49. Merta AI, Votruba J, Jindrich J, Holy A, Cihlar T, Rosenberg I, et al. Phosphorylation of 9-(2phosphonylmethoxyethyl)adenine and (S)-9-(3-fluoro-2-phosphonylmethoxypropyl)adenine by AMP (dAMP) kinase from L1210 cells. Biochem Pharmacol 1992; 44:2067–2077. 50. De Clercq E. Broad spectrum anti-DNA virus and anti-retrovirus activity of phosphonylmethoxyalkyl-purines and -pyrimidines. Biochem Pharmacol 1991; 42:963–972. 51. Balzarini J, Hao Z, Herdewijn P, Johns DG, De Clercq E. Intracellular metabolism and mechanism of anti-retrovirus action of 9-(2-phosphonyl-methoxyethyl) adenine, a potent anti-human immunodeficiency virus compound. Proc Natl Acad Sci USA 1991; 88:1499–1503. 52. Balzarini J, Holy A, Jindrich J, Dvorakova H, Hao Z, Snoeck R, et al. 9-[(2RS)-3- fluoro-2phosphonylmethoxyethyl] derivatives of purines: a class of highly selective antiretroviral agents in vitro and in vivo. Proc Natl Acad Sci USA 1991; 88:4961–4965. 53. Patel PH, Jacobo-Molina A, Ding J, Tantillo C, Clark AD Jr., Raag R, et al. Insights into DNA polymerization mechanisms from structure and function analysis of HIV-1 reverse transcriptase. Biochemistry 1995; 34:5351–5363. 54. Boyer PL, Tantillo C, Jacobo-Molina A, Nanni RG, Ding J, Arnold E, et al. The sensitivity of wildtype human immunodeficiency virus type 1 reverse transcriptase to dideoxynucleotides depends on template length; the sensitivity of drug-resistant mutants does not. Proc Natl Acad Sci USA 1994; 91:4882–4886. 55. Kellam P, Boucher CA, Larder BA. Fifth mutation in human immunodeficiency virus type 1 reverse transcriptase contributes to the development of high-level resistance to zidovudine. Proc Natl Acad Sci USA 1992; 89:1934–1938.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_74.html (1 of 2) [4/5/2004 4:53:11 PM]

Document

56. Martin JL, Wilson JE, Haynes RL, Furman PA. Mechanism of human immunodeficiency virus type 1 resistance to dideoxyinosine. Proc Natl Acad Sci USA 1993; 90:6135–6139. 57. Lacey SF, Reardon JE, Furfine ES, Kunkel TA, Bebenek K, Eckert KA, et al. Biochemical studies on the reverse transcriptase and RNase H activities from human immunodeficiency virus strains resistant to 3'-azido-3'-deoxythymidine. J Biol Chem 1992; 267:15789–15794. 58. Larder BA. 3'-Azido-3'-deoxythymidine resistance suppressed by a mutation conferring human immunodeficiency virus type 1 resistance to nonnucleoside

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_74.html (2 of 2) [4/5/2004 4:53:11 PM]

Document

Page 75

reverse transcriptase inhibitors. Antimicrob Agents Chemother 1992; 36:2664–2669. 59. Nanni RG, Ding J, Jacobo-Molina A, Hughes SH, Arnold E. Review of HIV-1 reverse transcriptase three-dimensional structure: implications for drug design. Perspectives in Drug Discovery and Design 1993; 1:129–150. 60. Boyer PL, Ding J, Arnold E, Hughes SH. Drug resistance of human immunodeficiency virus type 1 reverse transcriptase: subunit specificity of mutations that confer resistance to nonnucleoside inhibitors. Antimicrob Agents Chemother 1994; 38:1909–1914. 61. Mellors JW, Bazmi HZ, Schinazi RF, Roy BM, Hsiou Y, Arnold E, et al. Novel mutations in reverse transcriptase of human immunodeficiency virus type 1 reduce susceptibility to foscarnet in laboratory and clinical isolates. Antimicrob Agents Chemother 1995; 39:1087–1092. 62. Boyer PL, Hughes SH. Analysis of mutations of position 184 in the reverse transcriptase of human immunodeficiency virus type 1. Antimicrob Agents Chemother 1995; 39:1624–1628. 63. Pandey VN, Kaushik N, Rege N, Sarafianos SG, Yadav NS, Modak MJ. Role of M184 of human immunodeficiency virus type-1 reverse transcriptase in the polymerase function and fidelity of DNA synthesis. Biochem 1996; 35:2168–2179. 64. Wainberg MA, Drosopoulos WC, Salomon H, Hsu M, Borkow G, Parniak MA, et al. Enhanced fidelity of 3TC-selected mutant HIV-1 reverse transcriptase. Science 1996; 271:1282–1285. 65. Mitsuya H, Yarchoan R, Broder S. Molecular targets for AIDS therapy. Science 1990; 249:1533–1544. 66. Balzarini J, Van Aerschot A, Herdewijn P, De Clercq E. 5-Chloro-substituted derivatives of 2',3'didehydro-2',3'-dideoxyuridine, 3'-fluoro-2',3'-dideoxyuridine and 3'-azido-2',3'-dideoxyuridine as antiHIV agents. Biochem Pharmacol 1989; 38:869–874. 67. Steitz TA, Steitz JA. A general two-metal-ion mechanism for catalytic RNA. Proc Natl Acad Sci USA 1993; 90:6498–6502. 68. Spence RA, Kati WM, Anderson KS, Johnson KA. Mechanism of inhibition of HIV-1 reverse transcriptase by nonnucleoside inhibitors. Science 1995; 267:988–993. 69. Beese LS, Steitz TA. Structural basis for the 3'-5' exonuclease activity of Escherichia coli DNA polymerase I: a two metal ion mechanism. EMBO J 1991; 10:25–33. 70. Steitz TA. DNA- and RNA-dependent DNA polymerases. Curr Opin Struct Biol 1993; 3:31–38. 71. Joyce CM, Steitz TA. Function and structure relationships in DNA polymerases. Ann Rev Biochem 1994; 63:777–822. http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_75.html (1 of 2) [4/5/2004 4:53:12 PM]

Document

72. Georgiadis MM, Jessen SM, Ogata CM, Telesnitsky A, Goff SP, Hendrickson WA. Mechanistic implications from the structure of a catalytic fragment of Moloney murine leukemia virus reverse transcriptase. Structure 1995; 3:879–892. 73. Kuntz ID, Blaney JM, Oatley SJ, Langridge R, Ferrin TE. A geometric approach to macromolecularligand interactions. J Mol Biol 1982; 161:269–288. 74. Kuntz ID. Structure-based strategies for drug design and discovery. Science 1992; 257:1078–1082.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_75.html (2 of 2) [4/5/2004 4:53:12 PM]

Document

Page 76

75. Ring CS, Sun E, McKerrow JH, Lee GK, Rosenthal PJ, Kuntz ID, et al. Structure-based inhibitor design by using protein models for the development of antiparasitic agents. Proc Natl Acad Sci USA 1993; 90:3583–3587. 76. Balzarini J, Perez-Perez M-J, San-Felix A, Velazquez S, Camarasa M-J, De Clercq E. 2',5'-bis-O(tert-butyldimethylsilyl)-3'-spiro-5"-(4"-amino-1",2"-oxathiole-2",2"-dioxide) (TSAO) derivatives of purine and pyrimidine nucleosides as potent and selective inhibitors of human immunodeficiency virus type 1. Antimicrob Agents Chemother 1992; 36:1073–1080. 77. Rittinger K, Divita G, Goody R. Human immunodeficiency virus reverse transcriptase substrateinduced conformational changes and the mechanism of inhibition by nonnucleoside inhibitors. Proc Natl Acad Sci USA 1995; 92:8046–8049. 78. Palaniappan C, Fay PJ, Bambara RA. Nevirapine alters the cleavage specificity of ribonuclease H of human immunodeficiency virus 1 reverse transcriptase. J Biol Chem 1995; 270:4861–4869. 79. Jonckheere H, Taymans J-M, Balzarini J, Velazquez S, Camarasa M-J, Desmyter J, et al. Resistance of HIV-1 reverse transcriptase against [2',5'-bis-O-(tert-butyldimethylsilyl)-3'-spiro-5"-(4"-amino-1",2"oxathiole-2",2"-dioxide)] (TSAO) derivatives is determined by the mutation Glu138->Lys on the p51 subunit. J Biol Chem 1994; 269:25255–25258. 80. Pauwels R, Andries K, Debyser Z, Kukla MJ, Schols D, Breslin HJ, et al. New TIBO derivatives are potent inhibitors of HIV-1 replication and are synergistic with 2',3'-dideoxynucleoside analogues. Antimicrob Agents Chemother 1994; 38:2863–2870. 81. Kashman Y, Gustafson KR, Fuller RW, Cardellina JH II, McMahon JB, Currens MJ, et al. The calanolides, a novel HIV-inhibitory class of coumarin derivatives from the tropical rainforest tree, Calophyllum langericum. J Med Chem 1992; 35:2735–2743. 82. Boyer PL, Currens MJ, McMahon JB, Boyd MR, Hughes SH. Analysis of nonnucleoside drugresistant variants of human immunodeficiency virus type 1 reverse transcriptase. J Virol 1993; 67:2412–2420. 83. Goldman ME, O'Brien JA, Ruffing TL, Schleif WA, Sardana VV, Bynes VW, et al. A nonnucleoside reverse transcriptase inhibitor active on human immunodeficiency type 1 isolates resistant to related inhibitors. Antimicrob Agents Chemother 1993; 37:947–949. 84. Kleim J-P, Bender R, Kirsch R, Meichsner C, Paessens A, Riess G. Mutational analysis of residue 190 of human immunodeficiency virus type 1 reverse transcriptase. Virology 1994; 200:696–701. 85. Smith MBK, Rouzer CA, Taneyhill LA, Smith NA, Hughes SH, Boyer PL, et al. Molecular modeling studies of HIV-1 reverse transcriptase nonnucleoside inhibitors: Total energy of complexation as a predictor of drug placement and activity. Protein Science 1995; 4:2203–2222.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_76.html (1 of 2) [4/5/2004 4:53:14 PM]

Document

86. De Clercq E. Antiviral therapy of human immunodeficiency virus infections. Clin Microbiol Rev 1995; 8:200–239. 87. Champoux JJ. Roles of ribonuclease H in reverse transcription. In: Skalka AM, Goff SP, eds. Reverse Transcriptase. Plainview, New York: Cold Spring Harbor Laboratory Press, 1993:103–117. 88. Telesnitsky A, Goff SP. Strong-stop strand transfer during reverse transcription. In: Skalka AM, Goff SP eds. Reverse Transcriptase. Plainview, New York: Cold Spring Harbor Laboratory Press, 1993:49–83.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_76.html (2 of 2) [4/5/2004 4:53:14 PM]

Document

Page 77

89. Hughes SH, Arnold E, Hostomsky Z. RNase H of retroviral reverse transcriptases. Plainview, New York: Cold Spring Harbor Laboratory Press, 1996; in press. 90. Schatz O, Mous J, Le Grice SFJ. HIV-1 RT-associated ribonuclease H displays both endonuclease and 3' 5' exonuclease activity. EMBO J 1990; 9:1171–1176. 91. Tanese N, Goff SP. Domain structure of the Moloney murine leukemia virus reverse transcriptase: mutational analysis and separate expression of the DNA polymerase and RNase H activities. Proc Natl Acad Sci USA 1988; 85:1777–1781. 92. Tanese NT, Telesnitsky A, Goff SP. Abortive reverse transcription by mutants of Moloney murine leukemia virus deficient in the reverse transcriptase-associated RNase H function. J Virol 1991; 65:4387–4397. 93. Tisdale M, Schulze T, Larder BA, Moelling K. Mutations within the RNase H domain of human immunodeficiency virus type 1 reverse transcriptase abolish virus infectivity. J Gen Virol 1991; 72:59–66. 94. Davies JF, Hostomska Z, Hostomsky Z, Jordan SR, Matthews DA. Crystal structure of the ribonuclease H domain of HIV-1 reverse transcriptase. Science 1991; 252:88–95. 95. Chattopadhyay D, Finzel BC, Munson SH, Evans B, Sharma SK, Strakalattis NA, et al. Crystallographic analysis of an active HIV-1 ribonuclease H domain show structural features that distinguish it from the inactive form. Acta Crystallogr 1993; D49:423–427. 96. Prasad VR, Goff SP. Linker insertion mutagenesis of the human immunodeficiency virus transcriptase expressed in bacteria: definition of the minimal polymerase domain. Proc Natl Acad Sci USA 1989; 86:3104–3108. 97. Hizi A, Hughes SH, Shaharabany M. Mutational analysis of the ribonuclease H activity of human immunodeficiency virus 1 reverse transcriptase. Virology 1990; 175:575–580. 98. Hizi A, Tal R, Hughes SH. Mutational analysis of the DNA polymerase and ribonuclease H activities of human immunodeficiency virus type 2 reverse expressed in Escherichia coli. Virology 1991; 180:339–346. 99. Boyer PL, Ferris AL, Clark P, Whitmer J, Frank P, Tantillo C, et al. Mutational analysis of the fingers and palm subdomains of human immunodeficiency virus type-1 (HIV-1) reverse transcriptase. J Mol Biol 1994; 243:472–483. 100. Ding J, Jacobo-Molina A, Tantillo C, Lu X, Nanni RG, Arnold E. Buried surface analysis of HIV-1 reverse transcriptase p66/p51 heterodimer and its interaction with dsDNA template/primer. J Mol Recognition 1994; 7:157–161.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_77.html (1 of 2) [4/5/2004 4:53:17 PM]

Document

101. Boyer PL, Ferris AL, Hughes SH. Cassette mutagenesis of the reverse transcriptase of human immunodeficiency virus type 1. J Virol 1992; 66:1031–1039. 102. Boyer PL, Ferris AL, Hughes SH. Mutational analysis of the fingers domain of human immunodeficiency virus type-1 reverse transcriptase. J Virol 1992; 66:7533–7537. 103. Crouch RJ, Dirksen ML, Ribonuclease H. In: Linn SM, Roberts RJ, eds. Nuclease. Plainview, New York: Cold Spring Harbor Laboratory Press, 1982:211–241. 104. Kanaya S, Kohara A, Miura Y, Sekiguchi A, Iwai S, Inoue H, et al. Identification of the amino acid residues involved in an active site of Eschericia coli ribonuclease H by site-directed mutagenesis. J Biol Chem 1990; 265:4615–4621. 105. Smith JS, Roth MJ. Purification and characterization of an active human immunodeficiency virus type 1 RNase H domain. J Virol 1993; 67:4037–4049.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_77.html (2 of 2) [4/5/2004 4:53:17 PM]

Document

Page 78

106. Cirino NM, Cameron CE, Smith JS, Rausch JW, Roth MJ, Benkovic SJ, et al. Divalent cation modulation of the ribonuclease functions of human immunodeficiency virus reverse transcriptase. Biochemistry 1995; 34:9936–9943. 107. Schatz O, Cromme FV, Gruninger-Leitch F, Le Grice SFJ. Point mutations in conserved amino acid residues within the C-terminal domain of HIV-1 reverse transcriptase specifically repress RNase H function. FEBS Lett 1989; 257:311–314. 108. Volkmann S, Wohrl BM, Tisdale M, Moelling K. Enzymatic analysis of two HIV-1 reverse transcriptase mutants with mutations in carboxyl-terminal amino acid residues conserved among retroviral ribonucleoside H. J Biol Chem 1993; 268:2674–2683. 109. Mizrahi V, Brooksbank RL, Nkabinde NC. Mutagenesis of the conserved aspartic acid 443, glutamic acid 478, asparagine 494, and aspartic acid 498 residues in the ribonuclease H domain of p66/p51 human immunodeficiency virus type 1 reverse transcriptase. J Biol Chem 1994; 269:19245–19249. 110. Yang W, Hendrickson WA, Crouch RJ, Satow Y. Structure of ribonuclease H phased at 2 Å resolution by MAD analysis of the selenomethionyl protein. Science 1990; 249:1398–1405. 111. Katayanagi K, Miyagawa M, Matsushima M, Ishikawa M, Kanaya S, Ikehara M, et al. Threedimensional structure of ribonuclease H from E. coli. Nature 1990; 347:306–309. 112. Katayanagi K, Okumura M, Morikawa K. Crystal structure of Escherichia coli RNase HI in complex with Mg2+ at 2.8 Å resolution. Proteins 1993; 17:337–346. 113. Loya S, Tal R, Kashman Y, Hizi A. Illimaquinone, a selective inhibitor of the RNase H activity of human immunodeficiency virus type 1 reverse transcriptase. Antimicrob Agents Chemother 1990; 34:2009–2012. 114. Loya S, Hizi A. The interaction of illimaquinone, a selective inhibitor of the RNase H activity, with the reverse transcriptases of human immunodeficiency and murine leukemia retroviruses. J Biol Chem 1993; 268:9323–9328. 115. Restle T, Muller B, Goody RS. Dimerization of human immunodeficiency virus reverse transcriptase. J Biol Chem 1990; 265:8986–8988. 116. Muller B, Restle T, Kuhnel H, Goody RS. Expression of the heterodimeric form of human immunodeficiency virus type 2 reverse transcriptase in Escherichia coli and characterization of the enzyme. J Biol Chem 1991; 266:14709–14713. 117. Restle T, Muller B, Goodys RS. RNase H activity of HIV reverse transcriptase is confined exclusively to the dimeric forms. FEBS Letters 1992; 300:97–100.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_78.html (1 of 2) [4/5/2004 4:53:19 PM]

Document

118. Divita G, Rittinger K, Geourjon C, Deleage G, Goody RS. Dimerization kinetics of HIV-1 and HIV2 reverse transcriptase: A two step process. J Mol Biol 1994; 145:508–521. 119. Wang J, Smerdon SJ, Jager J, Kohlstaedt LA, Rice PA, Friedman JM, et al. Structural basis of asymmetry in the human immunodeficiency virus type 1 reverse transcriptase heterodimer. Proc Natl Acad Sci USA 1994; 91:7242–7246. 120. Tuerk C, MacDougal S, Gold L. RNA pseudoknots that inhibit human immunodeficiency virus type 1 reverse transcriptase. Proc Natl Acad Sci USA 1992; 89:6988–6992. 121. Green L, Waugh S, Binkley JP, Hostomska Z, Hostomsky Z, Tuerk C. Comprehensive chemical modification interference and nucleotide substitution analysis of an RNA pseudoknot inhibitor to HIV-1 reverse transcriptase. J Mol Biol 1995; 247:60–68.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_78.html (2 of 2) [4/5/2004 4:53:19 PM]

Document

Page 79

122. Marshall WS, Beaton G, Stein CA, Matsukura M, Caruthers MH. Inhibition of human immunodeficiency virus activity by phosphorodithioate oligodeoxycytidine. Proc Natl Acad Sci USA 1992; 89:6265–6269. 123. Marshall WS, Caruthers MH. Phosphorodithioate DNA as a potential therapeutic drug. Science 1993; 259:1564–1570. 124. Loya S, Hizi A. The inhibition of human immunodeficiency virus type 1 reverse transcriptase by avarol and avarone derivatives. FEBS Lett 1990; 269:131–134. 125. Loya S, Rudi A, Tal R, Kashman Y, Loya Y, Hizi A. 3,5,8-Trihydroxy-4-quinolone, a novel natural inhibitor of the reverse transcriptase of human immunodeficiency viruses type 1 and type 2. Arch Biochem Biophys 1994; 309:315–322. 126. Sheridan RP Rusinko A, Nilkantan R, Venkatraghavan R. Searching for pharmacophores in large coordinate data bases and its use in drug design. Proc Natl Acad Sci USA 1989; 86:8165–8169. 127. Marshall GR, Barry CD, Bosshard HE, Dammkoehler RA, Dunn DA. The conformational parameter in drug design: the active analog approach. In: Olson EC, Christoffersens RE eds. ComputerAssisted Drug Design. Honolulu, Hawaii: American Chemical Society, 1979:205–226. 128. Marshall GR. Computer-aided drug design. Ann Rev Pharmacol Toxicol 1987; 27:193–213. 129. Najera I, Holguin A, Quinones-Mateu ME, Munoz-Fernandez MA, Najera R, Lopez-Galindez C, et al. Pol gene quasispecies of human immunodeficiency virus: mutations associated with drug resistance in virus from patients undergoing no drug therapy. J Virol 1995; 69:23–31. 130. Erickson JW. The not-so-great escape. Nature Struct Biol 1995; 2:523–529. 131. Arnold E, Das K, Ding J, Yadav PNS, Hsiou, Y, Boyer PL, Hughes SH. Targeting HIV reverse transcriptase for anti-AIDS drug design. Drug Design and Discovery 1996; 13:29–47. 132. Larder BA, Kemp SD, Harrigan PR. Potential mechanism for sustained antiretroviral efficacy of AZT-3TC combination therapy. Science 1995; 269:696–699. 133. Dueweke TJ, Pushkarskaya T, Poppe SM, Swaney SM, Zhao JQ, Chen ISY, et al. A mutation in reverse transcriptase of bis(heteroaryl) piperazine-resistant human immunodeficiency virus type 1 that confers increased sensitivity to other nonnucleoside inhibitors. Proc Natl Acad Sci USA 1993; 90:4713–4717. 134. Larder BA. AZT resistance suppressed by a mutation conferring HIV-1 resistance to nonnucleoside reverse transcriptase inhibitors. Antimicrob Agents Chemother 1992; 36:1171–1174.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_79.html (1 of 2) [4/5/2004 4:53:21 PM]

Document

135. Mazzulli T, Rusconi S, Merrill DP, D'Auilla RT, Moonis M, Chou T-C, et al. Alternating verse continuous drug regimens in combination chemotherapy of human immunodeficiency virus type 1 infection in vitro. Antimicrob Agents Chemother 1994; 38:656–661. 136. Smith MS, Brian EL, De Clercq E, Pagano JS. Susceptibility of human immunodeficiency virus type 1 replication in vitro to acyclic adenosine analogs and synergy of the analogs with 3'-azido-3'deoxythymidine. Antimicrob Agents Chemother 1989; 33:1482–1486. 137. Dornsife RE, St. Clair MH, Huang AT, Panella TJ, Koszalka GW, Burns CL, et al. Anti-human immunodeficiency virus synergism by zidovudine (3'-azidothymi

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_79.html (2 of 2) [4/5/2004 4:53:21 PM]

Document

Page 80

dine) and didanosine (dideoxyinosine) contrasts with their additive inhibition of normal human marrow progenitor cells. Antimocrob Agents Chemother 1991; 35:322–328. 138. Antonelli G, Dianzani F, Bellarosa D, Turriziani O, Riva E, Gentile A. Drug combination of AZT and ddI: synergism of action and prevention of appearance of AZT-resistance. Antiviral Chem Chemother 1994; 5:51–55. 139. Balzarini J, Karlsson A, Perez-Perez MJ, Camarasa MJ, Tarpley WG, De Clercq E. Treatment of human immunodeficiency virus type 1 (HIV-1)-infected cells with combinations of HIV-1-specific inhibitors results in a different resistance pattern than does treatment with single-drug therapy. J Virol 1993; 67:5353–5359. 140. Balzarini J, Perez-Perez M-J, Velazquez S, San-Felix A, Camarasa M-J, De Clercq E, et al. Suppression of the breakthrough of human immunodeficiency virus type 1 (HIV-1) in cell culture by thiocarboxanilide derivatives when used individually or in combination with other HIV-1-specific inhibitors (i.e., TSAO derivatives). Proc Natl Acad Sci USA 1995; 92:5470–5474. 141. Fletcher RS, Arion D, Borkow G, Wainberg MA, Dmitrienko GI, Parniak MA. Synergistic inhibition of HIV-1 reverse transcriptase DNA polymerase activity and virus replication in vitro by combinations of carboxanilide nonnucleoside compounds. Biochemistry 1995; 34 10106–10112. 142. Baba M, Ito M, Shigeta S, Tanaka H, Miyasaka T, Ubasawa M, et al. Synergistic inhibition of human immunodeficiency virus type 1 replication by 5-ethyl-1-ethoxymethyl-6-(phenylthio)uracil (EEPU) and azidothymidine in vitro. Antimocrob Agents Chemother 1991; 35:1430–1433. 143. Richman DD, Rosenthal AS, Skoog M, Echner RJ, Chou T-C, Sabo JP, et al. BIRG-587 is active against zidovudine-resistant human immunodeficiency virus type 1 and synergistic with zidovudine. Antimicrob Agents Chemother 1991; 35:305–308. 144. Buckheit RW Jr., White EL, Germany-Decker J, Allen LB, Ross LJ, Shannon WM, et al. Cellbased and biochemical analysis of the anti-HIV activity of combinations of 3'-azido-3'-deoxythymidine and analogues of TIBO. Antiviral Chem Chemother 1994; 5:35–42. 145. Chong K-T, Pagano PJ, Hinshaw RR. Bis(heteroacyl)piperazine reverse transcriptase inhibitor in combination with 3'-azido-3'-deoxythymidine or 2',3'-dideoxycytidine synergistically inhibits human immunodeficiency virus type 1 replication in vitro. Antimicrob Agents Chemother 1994; 38:288–293. 146. Brennan TM, Taylor DL, Bridges CG, Leyda JP, Tyms AS. The inhibition of human immunodeficiency virus type 1 in vitro by a non-nucleoside reverse transcriptase inhibitor MKC-442 alone and in combination with other anti-HIV compounds. Antiviral Res 1995; 26:173–187.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_80.html (1 of 2) [4/5/2004 4:53:24 PM]

Document

147. Johnson VA, Walker BD, Barlow MA, Paradis TJ, Chou T-C, Hirsch MS. Synergistic inhibition of human immunodeficiency virus type 1 and type 2 replication in vitro by castanospermine and 3'-azido-3'deoxythymidine. Antimicrob Agents Chemother 1989; 33:53–57. 148. Connell EV, Hsu M-C, Richman DD. Combinative interaction of a human immunodeficiency virus (HIV) tat antagonist with HIV reverse transcriptase inhibitors and an HIV protease inhibitor. Antimicrob Agents Chemother 1994; 38:348–352.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_80.html (2 of 2) [4/5/2004 4:53:24 PM]

Document

Page 81

149. Craig JC, Duncan IB, Whittaker L, Roberts NA. Antiviral synergy between inhibitors of HIV proteinase and reverse transcriptase. Antiviral Chem Chemother 1993; 4:161–166. 150. Emini EA, Graham DJ, Gotlib L, Condra JH, Byrnes VW, Schleif WA. HIV and multidrug resistance. Nature 1993; 364:679. 151. Larder BA, Kellam P, Kemp SD. Convergent combination therapy can select viable multidrug resistant HIV-1 in vitro. Nature 1993; 365:451–453. 152. Shafer RW, Kozal MJ, Winters M, Iversen AKN, Katenstein DA, Ragni MV, et al. Combination therapy with zidovudine and didanosine selects for drug-resistant human immunodeficiency virus type 1 strains with unique patterns of pol gene mutations. J Infect Dis 1994; 169:722–729. 153. Shirasaka T, Kavlick MF, Ueno T, Gao W-Y, Kojima E, Alcaide ML, et al. Emergence of human immunodeficiency virus type 1 variants with resistance to multiple dideoxynucleosides in patients receiving therapy with dideoxynucleosides. Proc Natl Acad Sci USA 1995; 92:2398–2402. 154. Gu Z, Gao Q, Fang H, Salomon H, Parniak MA, Goldberg E, et al. Identification of a mutation of codon 65 in the IKKK motif of reverse transcriptase that encodes human immunodeficiency virus resistance to 2',3'-dideoxycytidine and 2',3'-dideoxy-3'-thiacytidine. Antimicrob Agents Chemother 1994; 38:275–281. 155. Lander BA, Kemp SD. Multiple mutations in HIV-1 reverse transcriptase confer high-level resistance to zidovudine (AZT). Science 1989; 246:1155–1158. 156. Fitzgibbon JE, Howell RM, Haberzettl CA, Sperber SJ, Gocke DJ, Dubin DT. Human immunodeficiency virus type 1 pol gene mutations which cause decreased susceptibility to 2',3'dideoxycytidine. Antimicrob Agents Chemother 1992; 36:153–157. 157. Prasad VR, Lowy I, Santos TDL, Chiang L, Goff SP. Isolation and characterization of a dideoxyguanosinetriphosphate-resistant mutant of human immunodeficiency virus reverse transcriptase. Proc Natl Acad Sci USA 1991; 88:11363–11367. 158. Havlir D, Murphy R, Saag M, Kaul I, Johnson V, Richman DD. Nevirapine: further dose escalation of monotherapy (600 mg/daily) and combination therapy with zidovudine. In: The First National Conference on Human Retroviruses and Related Infections, Washington D.C., 1993:101. 159. Saag MS, Sommadossi JP, Rainey D, Myers M, Cort S, Hall D, et al. A pharmacokinetic and antiretroviral activity study of nevirapine in combination with zidovudine plus zalcitabine (ZDV/ddC), zidovudine plus didanosine (ZDV/ddI), or didanosine (ddI) Alone. In: The First National Conference on Human Retroviruses and Related Infections, Washington D.C., 1993:102.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_81.html (1 of 2) [4/5/2004 4:53:26 PM]

Document

160. Gu Z, Gao Q, Li X, Parniak MA, Wainberg MA. Novel mutation in the human immunodeficiency virus type 1 reverse transcriptase gene that encodes cross-resistance to 2',3'-dideoxyinosine and 2',3'dideoxycytidine. J Virol 1992; 66:7128–7135. 161. Tisdale M, Kemp SD, Parry NR, Larder BA. Rapid in vitro selection of human immunodeficiency virus type 1 resistant to 3'-thiacytidine inhibitors due to a mutation in the YMDD region of reverse transcriptase. Proc Natl Acad Sci USA 1993; 90:5653–5656. 162. Larder BA, Coates KE, Kemp SD. Zidovudine-resistant human immunodeficiency virus selected by passage in cell culture. J Virol 1991; 65:5232–5236.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_81.html (2 of 2) [4/5/2004 4:53:26 PM]

Document

Page 82

163. Lacey SF, Larder BA. A novel mutation (V75T) in the HIV-1 reverse transcriptase confers resistance to 2',3'-didehydro-2',3'-dideoxythymidine (D4T) in cell culture. Antimicrob Agents Chemother 1994; 38:1428–1432. 164. Byrnes VW, Sardana VV, Schleif WA, Condra JH, Waterbury JA, Wolfgang JA, et al. Comprehensive mutant enzyme and viral variant assessment of human immunodeficiency virus type 1 reverse transcriptase resistance to nonnucleoside inhibitors. Antimicrob Agents Chemother 1993; 37:1576–1579. 165. Vasudevachari MB, Battista C, Lane HC, Psallidopoulos MC, Zhao B, Cook J, et al. Prevention of the spread of HIV-1 infection with nonnucleoside reverse transcriptase inhibitors. Virology 1992; 190:269–277. 166. Balzarini J, Karlsson A, Perez-Perez M-J, Vrang L, Walbers J, Zhang H, et al. HIV-1-specific reverse transcriptase inhibitors show differential activity against mutant strains containing different amino acid substitutions in the reverse transcriptase. Virology 1993; 192:246–253. 167. Balzarini J, Velazquez S, San-Felix A, Karlsson A, Perez-Perez M-J, Camarasa M-J, et al. Human immunodeficiency virus type 1-specific [2',5'-bis-O-[tert-butyldimethylsilyl] - β - D - ribofuranosyl] -3' spiro-5'' -(4'' -amino - 1",2" - oxathiole-2", 2'-dioxide)-purine analogues show a resistance spectrum that is different from that of the human immunodeficiency virus type-1-specific nonnucleoside analogues. Mol Pharmacol 1993; 43:109–114. 168. Bacolla A, Shih C-K, Rose JM, Piras G, Warren TC, Grygon CA, et al. Amino acid substitutions in HIV-1 reverse transcriptase with corresponding residues from HIV-2. J Biol Chem 1993; 268:16571–16577. 169. Demeter L, Resnick L, Nawaz T, Timpone JG Jr., Batts D, Reichman RC. Phenotypic and genotypic analysis of ateviridine (ATV) susceptibility of HIV-1 isolates obtained from patients receiving ATV monotherapy in a phase I clinical trial (ACTG 187): comparison to patients receiving combination therapy with ATV and zidovudine. In: Third Workshop on Viral Resistance. Gaithersburg, Maryland, 1993. 170. Tisdale M, Parry NR, Cousens D, St. Clair MH and Boone LR. Anti-HIV activity of (lS-4R)-4-[2amino-6-cyclopropylamino-9H-purin-9-yl]-2-cyclopentene-1-methanol (1592U89). In: Abstracts of the 34th Interscience Conference on Antimicrobial Agents and Chemotherapy. Orlando, Florida, 1994:92.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_82.html [4/5/2004 4:53:28 PM]

Document

Page 83

3 Retroviral Integrase: Structure as a Foundation for Drug Design Alison B. Hickman and Fred Dyda National Institutes of Health, Bethesda, Maryland I. Introduction A. Retroviral Lifecycle The human immunodeficiency virus (HIV) is one of only a few retroviruses known to infect humans. It is estimated that approximately twenty-two million people are now infected worldwide [1]. With only a tiny number of exceptions, infection ultimately leads to the development of the lethal condition of acquired immunodeficiency syndrome, or AIDS. To date, only a handful of drugs have been shown to have any effect on the course of the disease. These are, in general, relatively ineffective at significantly prolonging life, and drug resistance develops rapidly. Equally discouraging, vaccines have not yet been developed to prevent infection. The retroviral lifecycle presents several steps that can be targeted as possible sites of intervention by inhibitors. As shown in Figure 1, when a retrovirus encounters a host cell, specific recognition between proteins on the surface of the virus and receptors on the host cell surface leads to membrane fusion. The viral core then enters the cell cytoplasm where the process of reverse transcription begins. The requirement of the conversion of viral RNA to double-stranded DNA is a feature unique to retroviruses. With the recent exception of the protease inhibitor saquinavir, ritonavir, and indinavir, the drugs approved to date by the U.S. Food and Drug Administration (FDA) for the treatment of HIV infection have been nucleoside analogs targeted against the viral enzyme that carries out this conversion, reverse transcriptase.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_83.html [4/5/2004 4:53:30 PM]

Document

Page 84

Figure 1 Retroviral lifecycle as summarized in Reference 67. Reprinted by permission of Springer-Verlag Publishing Co., New York, NY.

Although details of the timing of reverse transcription, nuclear localization, and integration are not yet clear, it is generally recognized that the movement of double-stranded viral DNA across the nuclear membrane is followed by insertion, or integration, of the viral genome into a host-cell chromosome. The viral DNA moves as part of a larger “preintegration complex,” a high-molecular-weight aggregate whose composition has not yet been completely defined. The end result of integration is the incorporation of the viral DNA into the DNA of the host cell. Once there, the provirus can serve as a template for the production of mRNA, allowing for the synthesis of viral proteins. These are assembled at the cell membrane to produce new viral particles, which then bud off to seek out new cells to infect. The integrated viral DNA is also necessarily copied whenever the host cell undergoes cell division. The insidious nature of

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_84.html (1 of 2) [4/5/2004 4:54:27 PM]

Document

Page 85

the virus arises because, once integrated, the viral DNA can no longer be distinguished from host cell DNA and has become a permanent fixture of the host cell genome. B. Rationale for Drug Design Against Integrase to Fight HIV and AIDS It has been demonstrated that the chemical steps that comprise DNA integration are carried out by the viral protein, integrase (IN). Integrase is encoded by the 3' end of the viral pol gene, which also codes for two other viral enzymes, the protease and reverse transcriptase. These three enzymes are initially synthesized as part of a larger polyprotein that is subsequently cleaved by the protease to the individual proteins. Why is integrase a good target for drug-design efforts to prevent infection by halting the viral replication cycle? First, integration is required for replication. In the absence of integration, the virus is unable to continue to make copies of itself. Secondly, the enzyme that carries out integration is virally encoded, and when the viral genome is disrupted so that functional integrase is no longer made, sustained viral replication does not occur [2]. This demonstrates that if viral integrase can be effectively inhibited, there is no protein encoded by the host cell that can replace it and carry out viral integration. Finally, since mammalian cells do not have enzymes capable of integrating HIV DNA, there are no vital host cell analogs of integrase carrying out essential reactions whose function would be blocked by integrase inhibitors. Effective inhibition of HIV integrase would add to the number of sites at which the virus replication cycle can be halted. One can imagine treatment protocols in which a mixture of inhibitors, each aimed at a different viral protein, could be administered. This is known as divergent combination therapy. As structural details are a necessary starting point for rational drug design, we present here our recent results on the high-resolution three-dimensional structure of the catalytic core domain of HIV-1 integrase [3]. We also review the current literature discussing integrase inhibitors and present thoughts on ways in which knowledge of the chemical reactions carried out by integrase and its structure might direct the development of effective inhibitors. II. Biochemical Reactions Catalyzed By HIV Integrase A. In Vivo Integration Details of the initial chemical reactions that occur during HIV integration are now well understood (for reviews, see References 4,5). Once linear double-stranded DNA is available for integration, (Figure 2a) integrase then removes

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_85.html [4/5/2004 4:54:29 PM]

Document

Page 86

Figure 2 In vivo reactions carried out by HIV integrase.

two nucleotides from each 3' end of the viral DNA (Figure 2b). The two nucleotides are removed as a dinucleotide rather than in two individual steps. The specificity for this reaction is conferred by the third and fourth nucleotides from each 3' end, a -CA sequence that is absolutely conserved. Once two nucleotides have been removed, leaving recessed 3' hydroxyl groups, the next step is the joining of the 3' ends to target DNA (Figure 2c, d). This process, known as double-ended integration, occurs on opposite strands such that the joining sites on each of the target DNA strands are separated by five base pairs. The final step in integration is the repair of the single-stranded gaps generated by the staggered insertion of the viral 3' ends on opposite strands; this regenerates an intact double-stranded DNA molecule (Figure 2e and f). Gap repair is probably carried out by host cell DNA repair systems. One necessary consequence of retroviral integration is the duplication of five base pairs of host cell DNA on either side of the integrated provirus.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_86.html [4/5/2004 4:54:34 PM]

Document

Page 87

Another is the loss from the ends of the viral DNA of the original two base pairs that preceded the conserved 3'-CA. B. In Vitro Assays to Monitor Integrase Activity In contrast to the in vivo reaction, concerted integration in vitro of two HIV DNA ends into a target DNA molecule separated by a 5 base-pair stagger occurs very inefficiently. However, in vitro systems have been developed [6,7] using recombinant HIV integrase that have allowed the chemistry of the single-ended integration event to be studied in fine detail. It is possible and routine to use short, doublestranded synthetic oligonucleotides that mimic the viral ends to monitor the removal of two nucleotides from 3' ends (denoted 3' processing or cutting) and the subsequent insertion of one 3' processed DNA molecule into another (known as strand transfer or joining). Typical reactions are depicted in Figure 3. The stereochemical mechanism of 3' processing and strand transfer has been investigated using DNA substrates that incorporate phosphorothioate link-ages [8]. For both reactions, the introduced chiral centers are inverted in the products, implying that the reactions occur via a one-step in-line displacement mechanism rather than via a covalent intermediate. A third assay of integrase activity, termed disintegration, has more recently been developed [9] that monitors the apparent reversal of strand-

Figure 3 Reactions carried out by integrase in vitro, using short oligonucleotide substrates.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_87.html [4/5/2004 4:54:38 PM]

Document

Page 88

transfer (Figure 3). While disintegration probably has no physiological significance, it has been useful in defining aspects of integrase biochemistry. The three in vitro activities of integrase require divalent metal ions as cofactors. The only two metals that support these activities are Mn2+ and Mg2+. Since quite high metal concentrations must be added to assays (1–10 mM for optimal activity), it has been presumed that Mg2+ is the ion used in vivo. C. Evidence for a Multimer as the Active Unit of Integrase Several lines of evidence demonstrate that the active unit of integrase is a multimer. It is clear, as an isolated protein in solution, that integrase forms dimers [6,10–12], and it has been shown by sedimentation equilibrium studies that Rous sarcoma virus (RSV) integrase exists in reversible equilibrium between monomeric, dimeric, and tetrameric forms [13]. Protein-protein cross-linking studies of HIV-1 [14] and RSV [15] integrases confirm the existence of protein dimers and tetramers in solution, and in vivo, the yeast GAL4 two-hybrid system has demonstrated that HIV-1 integrase can interact with itself [16]. Complementation studies using mutant proteins in vitro provide compelling evidence that the active form of integrase must be at least a dimer [14, 17]. This can be inferred from the result that when certain inactive forms of integrase—generated either by truncation or point mutation—are mixed, robust activity can be reconstituted. This indicates that different monomers in a multimer are capable of providing different essential functions in the context of an active complex. Collectively, these studies suggest that integrase acts as a multimer. This would also seem the most straight-forward model to explain the observation that viral integration requires two coordinated cutting and joining reactions on the target DNA during strand transfer. However, physical studies have not yet addressed what form of integrase actually binds to DNA and carries out the chemical reactions of integration. III. Properties of HIV-1 Integrase A. Domain Structure of Retroviral Integrases A consistent view of the domain structure of retroviral integrases has emerged by combining the results from biochemical studies using deletion and site-specific mutants, limited proteolysis experiments, and sequence comparisons among the family of retroviral integrases. The organization of the domains of integrase is shown schematically in Figure 4. The central domain of HIV-1 integrase, consisting approximately of residues 50 to 200, is largely conserved among retroviral integrases, and forms

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_88.html [4/5/2004 4:54:40 PM]

Document

Page 89

Figure 4 Schematic domain structure of HIV-1 integrase as adapted from Engelman et al. [19]. Structures of two domains, the catalytic core extending from residues 50 to 212 [3] and the nonspecific DNA-binding domain from residues 220 to 270 [28,29], have recently been determined by x-ray crystallography and NMR spectroscopy, respectively.

the protease-resistant core of the protein [18,19]. Within this domain are three invariant residues that comprise the “D,D-35-E motif” (see alignment in Figure 5). These are residues Asp64, Asp116, and Glu152. Even conservative substitution of any of these residues leads to loss of all three in vitro activities of integrase in parallel [19–21]. The D,D-35-E motif is also observed in retrotransposons and some prokaryotic transposases. A truncated form of HIV-1 integrase consisting of residues 50 to 212 is capable of disintegration [22], implying that the catalytic site is contained within this domain. These observations and the absolute requirement for metals for in vitro activity have led to the proposal that the three acidic residues constitute a divalent metal-binding site capable of binding one or two Mg2+ or Mn2+ ions to form a catalytically active enzyme. As will be seen in later sections, the three-dimensional structure of the core domain of HIV-1 integrase is consistent with this hypothesis. The catalytic mechanism may be, therefore, similar to the one proposed by Beese and Steitz for the 3'–5' exonuclease of E. coli DNA polymerase I [23]. It is proposed that for phosphate bond cleavage, one metal ion helps form the attacking hydroxide ion while the other stabilizes a pentacovalent intermediate around the phosphorus. The C-terminus of HIV-1 integrase, consisting approximately of residues 210 to 288, includes the dominant nonspecific DNA binding domain [24, 25], which has been more finely mapped to residues 220–270 [26]. The C-terminus is the least conserved region of retroviral integrases; only one residue, Trp235, is absolutely invariant. However, it has been reported that removal of only five amino acids from the C-terminus of HIV-1 integrase is enough to severely reduce its 3' processing and strand transfer activities [27]. One notable feature of the C-terminus is its high proportion of positively charged residues. As discussed in Section IV.E, the structure of part of this region has recently been determined using NMR spectroscopic methods [28,29].

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_89.html (1 of 2) [4/5/2004 4:54:43 PM]

Document

Page 90

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_90.html [4/5/2004 4:54:55 PM]

Document

Page 91

Figure 5 Alignment of amino acid sequences of retroviruses. See Engelman et al. [19] for details.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_91.html [4/5/2004 4:55:13 PM]

Document

Page 92

The role of the N-terminus of integrase, residues 1 to 50, is still unclear. Within this region are four strictly conserved amino acids: two His and two Cys residues. In HIV-1 integrase, the spacing is His-X3His-X23-Cys-X2-Cys. This cluster of His and Cys residues is reminiscent of a zinc-binding motif, and it has been demonstrated that the full-length protein binds Zn2+ [22,30], and that the separately expressed domain consisting only of residues 1 to 55 also binds Zn2+ stoichiometrically [31]. However, it has not been shown that either the structural integrity or the enzymatic activities of integrase require Zn2+. While truncation of residues from the N-terminus of HIV integrase results in loss of 3' processing and strand transfer activities [22,25], in the case of RSV integrase, the N-terminal region can be replaced by unrelated sequences, and the enzyme is still capable of all three in vitro activities [32]. B. Biophysical Properties of Full-Length Recombinant HIV-1 Integrase It has been known for some time that recombinant HIV-1 integrase is a particularly poorly behaved protein in solution. Its solubility in most usual buffers is limited to approximately 1 mg/mL, and even then only in the presence of high concentrations of NaCl. At ~1 mg/mL, HIV-1 integrase slowly precipitates out of solution, revealing one of its characteristic features, a tendency towards aggregation. These properties of the protein are not unreasonable, since in its viral environment integrase is probably never required to be a soluble protein. To maintain the integrity of preintegration complexes, it may even be advantageous for the protein to have the properties of being rather insoluble and sticking to itself, nucleic acid, and perhaps other proteins. C. Properties of Truncated Versions of HIV-1 Integrase It has been our approach to protein structure determination by x-ray crystallography that it is imperative to begin with well-characterized and well-behaved protein. In particular, it is important that the protein be reasonably soluble and monodisperse in solution. Unfortunately, as discussed above, recombinant HIV-1 integrase satisfies neither of these conditions. One approach we and others have taken to circumvent these problems has been to examine truncated versions of HIV-1 integrase to determine if removal of amino acids from either terminus or both affects solubility and aggregation properties. Although we observed that two proteins we constructed, IN213–288 and IN50–288, were more soluble than the full-length HIV-1 integrase, IN1–288 [33, and unpublished observations], our first target protein for crystallization efforts was the core domain of HIV-1 integrase consisting of residues 50 to 212, IN50–212. We reasoned that this protein domain was likely to be compact and

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_92.html [4/5/2004 4:55:15 PM]

Document

Page 93

well folded since it is relatively resistant to proteolysis. As it is also active for disintegration, we concluded that it contained the enzyme active site. We exploited the convenience of histidine-tag (HT) technology to develop methods to purify large quantities of IN50–212 [12]. A 20-amino-acid histidine-containing tag was added to the N-terminus of HIV-1 IN50–212 [22] to allow rapid purification on nickel affinity columns. It was subsequently removed by thrombin cleavage. Biophysical studies showed that although buffer conditions could be identified where the protein was soluble to ~ 4 mg/mL, under these conditions the protein was highly aggregated (unpublished observations). Although the aggregation problem could be largely avoided by the addition of high concentrations of the zwitterionic detergent CHAPS, conditions could not be identified under which protein crystals formed in the presence of CHAPS. D. Systematic Mutation of Hydrophobic Residues to Improve Protein Solubility As it became clear that IN50–212 was crystallographically challenged, a condition readily understood in terms of its aggregation problems and low solubility, a more radical approach was undertaken to try and improve its biophysical properties. Hydrophobic residues in the catalytic core were targeted for sitespecific mutation according to the following criteria: where two or more hydrophobic residues were encountered close together in the primary amino acid sequence, they were each changed to an alanine residue. When a hydrophobic residue stood alone, it was mutated to lysine. In this way, 29 different mutant proteins of IN50–212 were rapidly generated using the overlapping polymerase chain reaction (PCR) and screened for improved solubility properties [34]. Three mutated proteins were identified that were more soluble at lower NaCl concentration than the unmutated core (V165K, F185K, and the double mutation of W131A/W132A). However, one of these in which Phe185 was mutated to Lys had dramatically improved solubility and was ultimately crystallized and its three-dimensional structure determined [3]. The remarkable biophysical properties of this single point mutant of IN50–212 have recently been described [34]. IV. Structure Of The Catalytic Core Domain Of HIV-1 Integrase A. Description of the Structure The Overall Protein Fold The three-dimensional structure of the catalytic core domain of HIV-1 integrase is centered on a mixed five stranded β sheet flanked by several helices forming

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_93.html [4/5/2004 4:55:17 PM]

Document

Page 94

Figure 6 Molscript stereo figure of the three-dimensional structure of the catalytic core of HIV-1 integrase. The two catalytically essential aspartic acid residues (D64 and D116) visible in the x-ray structure are highlighted.

an α-β meander sandwich (see Figure 6) [35]. In the crystal structure, interpretable electron density starts at Cys56, leading to a short loop. The first β strand starts at Gly59 and runs until Val68. The two central residues of a type I' reverse β turn, Glu69 and Gly70, change the polypeptide chain direction to form the second β strand between residues Lys71 and His78, which runs antiparallel with the first strand. A type I'β turn follows, with Val79 and Ala80 changing the chain direction again to form the third β strand between Ser81 and Ile89, which runs antiparallel with the second strand. A short loop between Pro90 and Glu92 leads to the first α helix (helix A) between Thr93 and Trp108. This helix packs against the bottom face of the sheet formed by the first three antiparallel strands by several hydrophobic interactions. A short loop (Pro109 and Val110) leads to the fourth β strand between Lys111 and His114. This strand is parallel with the first. A short loop starting at Thr115 leads to helix B, a one turn helix between Gly118 and Thr122, followed by helix C between Ser123 and Ala133. This helix runs parallel to and packs against helix A. The residues Gly134 and Ile135 form a short loop prior to the fifth and last β strand of the structure between Lys136 and Ala138. This short strand is parallel with the first and the fourth. There is no interpretable electron density due to disorder between Gly140 and Met154. At Met154, the fourth α helix (helix D) starts and runs until Ala169 on the top face of the sheet formed by the first three β strands. The residue Glu170 leads into the next helix (helix E) running between His171 and Lys186, the first residue of a short basic sequence (Lys186, Arg187, and Lys188). Together with

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_94.html (1 of 2) [4/5/2004 4:55:35 PM]

Document

Page 95

Gly193 and Tyr194, Lys188 and Gly189 form two short antiparallel β strands separated by a turn of three residues (Gly190, Ile191, and Gly192) not involved in main chain hydrogen bonds. The first residue of the last helix, (helix F), which runs until Asp212, is Ser194. Three Conserved Acidic Residues at the Enzyme Active Site There are four amino acids in the core domain sequence that are absolutely conserved among retroviral integrases: Asp64, Asp116, Glu152, and Lys159. The three acidic residues form the conserved D,D-35E motif and have been shown to be essential for catalysis (see Section III.A). The role of Lys159 in retroviral integrases is not obvious; its replacement with Val does not abolish catalytic activity, although there is a decrease in strand transfer activity [20]. The first essential acidic residue, Asp64, is located in the middle of the first β strand, while Asp116 is in a loop region right after the fourth β strand. These two residues define the active-site area and they are right next to each other three-dimensionally with their α-carbons separated by only 6.7 Å. The closest approach is 3.4 Å between Oδ1 of Asp64 and Cβ of Asp116. These residues are on the surface of the molecule, not part of any obvious substrate-binding cleft. The third catalytically essential acidic residue, Glu152, is in the disordered and hence crystallographically invisible region between Gly140 and Met154. Its location therefore must be inferred from other parts of the structure and from available threedimensional structures of related proteins. The location of Met154, the residue only two positions upstream from Glu152, is known because of interpretable electron density. The distance between the αcarbons of Glu152 and Met154 cannot be larger than about 7.3 Å, which constrains Glu152 to the neighborhood of the two other essential carboxylates, allowing it to contribute to the formation of a divalent metal-binding site. More recently, a crystal structure of the avian sarcoma virus (ASV) integrase core domain was solved [36]. Within this domain, ASV integrase has 24% sequence identity to the HIV-1 integrase core and, as expected, its three-dimensional structure is remarkably similar. The ASV integrase core in its native form has much better solution properties than the HIV-1 integrase core, and did not require any point mutations to render it crystallizable. Due to this fact and perhaps also due to its different crystal packing interactions, the crystal lattice of the ASV integrase core domain is somewhat more ordered than that of HIV-1. The two three-dimensional structures can be aligned quite well, using 74 α-carbons, with an rms deviation of only 1.4 Å in these α-carbon positions. The most remarkable difference between the two structures is that in the ASV structure the electron density is interpretable in all parts of the molecule. This is not to say, however, that serious disorder is not present. For example, in one particular loop, the temperature factors are above 70 Å2 for the α carbons, indicating larger than 1 Å mean displacement value for these atoms. The corresponding

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_95.html [4/5/2004 4:55:37 PM]

Document

Page 96

region in the HIV-1 structure is the uninterpretable stretch from residues 140 to 153, showing that beyond the three-dimensional similarity, the molecules also share a similar disorder pattern despite their different crystal packing interactions. It is clear that in the apoenzyme (metal-free) form of the HIV-1 integrase core, disorder is present in parts of the active site. However, in the holo form, structural stability must be necessary to form a metal-binding site. Position of the Third Essential Carboxylate Does the structure of ASV integrase give us a hint about the likely conformation of the polypeptide chain around Glu152 in the holoenzyme form of the HIV-1 integrase core? The answer is probably yes, considering the overall similarity of the structures. The first residue after the disordered part in the HIV1 integrase core is Met154, which is also the first residue in helix C. The corresponding helix in the ASV core is longer, running between Gln153 and Gly175. The residue analogous to Glu152 is Glu157 in the ASV integrase core structure, located a half-turn upstream from Ala159, which corresponds to Met154 in the HIV-1 integrase structure. It is plausible to assume, therefore, that the polypeptide chain in the holoenzyme form of HIV-1 integrase would also be in a helical conformation around Glu152, and its location would be very close to the one that can be inferred from the location of Glu157 in the ASV integrase core. Secondary structure prediction also supports this assumption, assigning α-helical structure around Glu152. Why does this helical turn show significant disorder in the HIV-1 integrase structure? The answer might be found in the amino acid sequence: Pro145 is a highly conserved residue among retroviral integrases, the only exception being ASV integrase where it is substituted with a Ser. Since the main chain nitrogen of a proline is not capable of participating in hydrogen bonding, it is very rarely found in helices. It is likely that if the polypeptide chain around Glu152 were helical in the holoenzyme form of the HIV-1 integrase core, then this helix would start after Pro145. There is no such restriction in the ASV integrase core, and it is possible that this is why helix C is longer in ASV than in HIV. This may also explain the disorder around Glu152 in the HIV-1 integrase core, since it is closer to the end of the helix and more susceptible to disordering effects. A longer helix and therefore a more ordered active site in the apo form may be a unique feature of the ASV integrase core. B. Similarity to Other Polynucleotidyl Transferases Overall Protein Folds The catalytic core domain of HIV-1 integrase has a topologically identical fold with the RNase H domain of HIV-1 reverse transcriptase [37], the RuvC Holli-

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_96.html [4/5/2004 4:55:39 PM]

Document

Page 97

Figure 7 Structures of the catalytic core of HIV-1 integrase, HIV-1 RNase H, RuvC, and the core domain of MuA transposase demonstrating similarities in folding topology. The catalytically essential residues are highlighted.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_97.html (1 of 2) [4/5/2004 4:55:55 PM]

Document

day junction resolving enzyme [38], and also the core domain of the phage MuA transposase [39] (see Figure 7). In the case of HIV-1 integrase, RNase H, and RuvC, definite three-dimensional similarity extends only to the ends of the last β strand; from this point, the structures diverge. In HIV-1 RNase H, there is only one more α helix corresponding to helix D in the HIV-1 integrase core but in a 40° different orientation. In RuvC there are three more helices, with the last one running parallel to helix D of HIV-1 integrase, but in the opposite direction and also 4.6 Å closer to the β sheet. In contrast, the homology between HIV-1 inte-

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_97.html (2 of 2) [4/5/2004 4:55:55 PM]

Document

Page 98

grase and the MuA transposase extends until the carboxyl termini of their respective catalytic domains with three very similarly positioned and oriented α helices. Limited three-dimensional alignment between the four molecules can be accomplished by identifying structurally homologous stretches along the polypeptide chains if this search is restricted to between the first well-ordered amino terminal residue and the end of the last β strand. For the core domain of HIV-1 integrase, this corresponds to the region between Ile60 and Gln137. Using the corresponding region in the MuA transposase, the two structures can be aligned with an rms deviation of 1.7 Å over 69 α-carbon positions. The main differences between these structures are two insertions in the transposase core: an 11-residue β-stranded extension replacing the turn between the first and the second strand in the HIV-1 integrase core, and a 15-residue extension before helix B with no secondary structure. Both of these extensions interact with the downstream nonspecific DNA-binding domain of the transposase. For HIV1 RNase H, the alignment results in a rms deviation of 2.0 Å over 48 alignable α-carbon positions. The position of helix A is significantly different in RNase H, as it shifts more than 5 Å toward the β sheet. There is also an additional 2.5 turn helix following the fourth β strand and a 5-residue loop after this helix. For RuvC the alignment yields an rms deviation of 2.0 Å over 50 alignable α-carbon positions. In this case, the differences are mostly the result of longer secondary structure elements in RuvC. Of all the molecules compared, the HIV-1 integrase core is the smallest, with the most compact design in the region where these alignments were performed. For comparison, let us mention again that the homologous ASV integrase core can be aligned with an rms deviation of 1.4 Å over 74 α-carbon positions in this region. Both topological similarity and three-dimensional homology with the MuA transposase was expected based on the similarity of the reactions the enzymes catalyze, but the relationship with RNase H and RuvC was a surprise. This discovery led to the proposal of a new polynucleotidyl transferase superfamily. All the members of the superfamily are divalent metal ion-dependent endonucleases, and they all leave 3'-OH and 5'-phosphate groups at the site of cleavage. All the members of the superfamily display their catalytically essential acidic residues at the same general location. There are three such residues in HIV-1 integrase, RNase H, and the MuA transposase, while there are four in RuvC. Two of these residues are always located on the same three-dimensional structural elements, while the location of the third varies. The Asp64 residue HIV-1 integrase corresponds to Asp443 in HIV-1 RNaseH, Asp7 in RuvC, and Asp269 in the MuA transposase. Based on the three-dimensionally aligned structures, the α-carbon positions of these residues cluster quite well around that of HIV-1 integrase, with an rms deviation of 0.84 Å. All these residues are located in the middle of first β strand. The side chain torsion angle, Chi 1, is

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_98.html [4/5/2004 4:55:56 PM]

Document

Page 99

-62° for HIV-1 integrase, a frequently observed rotamer. For the other three molecules this Chi 1 value varies between -142° and -156°, a common range for rotamer angles. The preference for the first rotamer of the HIV side chain is probably due to the 2.73 Å-long hydrogen bond between Oο°2 of Asp64 and Nε2 of Gln62. There is no such interaction in the other molecules. The different rotamer causes a 2.46 Å rms separation of the Cγ positions around the HIV-1 Cγ, compared to only 1.63 Å around the Cγ of Asp443 in HIV-1 RNase H, indicting the carboxylate of Asp64 in HIV-1 integrase as the outlier. Position of the Second Carboxylate Residue In all the analogous structures, the second essential carboxylate resides just after the end of the fourth β strand. The main-chain atoms are not involved in strand-forming direct hydrogen bonds, therefore the chain diverts from running parallel with the first strand, forming a small cleft. The equivalent residues are Asp116 in HIV-1, Asp498 in HIV-1 RNase H, and Glu66 in RuvC. The clustering is weaker than for Asp64; the rms deviation is 1.77 Å in α-carbon position around the HIV-1 integrase residue. Interestingly, by including the structurally otherwise highly homologous ASV integrase core, the rms deviation increases to 2.15 Å due to the 3 Å distance between the Cα of Asp116 of HIV-1 integrase and that of the corresponding residue, Asp121 of ASV integrase. The rms separation between the ASV position and the rest of the cluster (now excluding HIV-1 integrase) is 2.2 Å, which is rather high, identifying the ASV residue as the outlier. For the Chi 1 torsion angles, all three preferred rotamers are present: Chi 1 is -86° for HIV-1 integrase, 73° for HIV-1 RNase H, and -173° for the MuA transposase. The RuvC Chi 1 value is not included in this comparison because it has a Glu in this position. The different Chi 1 values combined with the variation in α-carbon positions leads to a somewhat more scattered Cγ (or Cδ for Glu66 in RuvC) position with an rms deviation of 2.71 Å around Cγ of Asp116 of HIV-1 integrase. By including the ASV molecule, the scatter increases to 3.82 Å due to the 6 Å distance between Cγ of Asp116 in HIV-1 integrase and Cγ of Asp121 of ASV integrase. The Third Essential Carboxylate The location of the third essential catalytic carboxylate varies between different members of the superfamily. For HIV-1 integrase, Glu152 is in a disordered region with no interpretable electron density. Based on the location of the equivalent residue in the ASV integrase, its position is assumed to be on helix D, as discussed above. For RNase H, Glu478 is located on helix A, with its side chain pointing toward the other two carboxylates to complete the divalentmetal-binding site. Such metal binding has been observed crystallographically

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_99.html [4/5/2004 4:51:54 PM]

Document

Page 100

[37]. For RuvC, four residues have been shown to be essential for catalysis [40]. The third of these, Asp138, is at the amino end of the last helix of the structure. In the three-dimensionally aligned structures, this helix is parallel with the helix D in HIV-1 integrase, although it is running in the opposite direction. The position of helix D in RuvC is also significantly different, mainly due to a 13 Å shift along its axis toward the active site, placing Asp138 close to the other two carboxylates. The fourth essential residue, Asp141, is on the first turn of the same helix, very close in the aligned structures to the essential Glu157 of ASV integrase, their α carbons separated by only 1.6 Å. For the MuA transposase, its third essential carboxylate, Glu392, is in a loop region, just one residue upstream from the amino end of a helix, the topological equivalent of helix D. Unexpectedly, this residue turns away from the region defined by the two other carboxylates to a position where it clearly cannot contribute to the formation of a metal-binding site. It is likely, therefore, that the conformation of the polypeptide chain around Glu392 in the transposase core observed in the crystal structure belongs to an inactive form. In this case, a conformational change upon transposase tetramer assembly or perhaps upon substrate binding is required for activity. Significance of the Disordered Region From the point of view of HIV-1 integrase, it is interesting to note that the apparently flexible part of the MuA transposase structure is topologically equivalent with the disordered and uninterpretable part of the integrase. Similarly, in the crystal structure of the isolated RNase H domain of HIV-1 reverse transcriptase, a five-residue loop in a topologically equivalent location is disordered and therefore uninterpretable. In the ASV integrase core, the corresponding loop is visible but with rather high mobility. It seems that some kind of disorder or flexibility in this region is a common feature of the superfamily. Crystal structures of enzyme-substrate or enzyme-inhibitor complexes will tell us the functional significance of this flexibility as well as the exact configuration of the active site. C. The Dimer Interface HIV-1 integrase is active as a multimer, and the catalytic core domain alone forms dimers in solution, even at low protein concentration (see Section II.C). In the crystal structure, a roughly spherical dimer of about 45 Å diameter was observed, formed by a crystallographic two-fold axis. The dimer has a large solvent-excluded surface of 1300 Å2 per monomer. This area is close to what is expected for dimers in this molecular weight range [41]. Therefore, we are convinced that in the crystal structure the authentic dimer is present. This was subsequently confirmed by the structure of the ASV integrase core. Although

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_100.html [4/5/2004 4:51:56 PM]

Document

Page 101

crystallized under different conditions, forming crystals that are in a different space group with different crystal-packing interactions, the dimer observed for the ASV core is essentially identical with that of HIV-1 integrase, although the solvent-excluded surface is smaller (only 740 Å2). This difference is largely due to the absence of helix F in the ASV structure. The core domain dimer of HIV-1 integrase is held together by several hydrophobic and polar interactions. Hydrophobic interactions dominate the interface between helix E from one monomer and helices A and B from the other. There is a buried salt bridge between Glu87 of the third β strand and Lys103 on helix A. There are also some water-mediated polar interactions between these two secondary structure elements. There are direct hydrogen bonds between residues on helix A and residues on helix E across the interface including one between Lys185 (the substitution responsible for the improved solubility and therefore crystallizability) and the main-chain carbonyl on Ala105. In the ASV core, His198 is in this position, forming a very similar hydrogen bond with the carbonyl oxygen of Ala110. There are about 10 water molecules buried in the interface, all involved in hydrogen bonds. The part of the solvent-accessible surface of the monomer which becomes buried upon dimer formation displays a high degree of shape compatibility with itself: by rotating it 180° around the crystallographic two-fold axis, the resulting surface will fit the original one without forming large pockets. It is possible that the core domain of HIV-1 integrase has evolved to optimize this compatibility in order to increase its stability. It would be interesting to see the effect on protein activity of site-directed mutations aimed at disrupting this interface and hence the dimer (or possibly the higher order multimers in the context of the full-length protein). D. Implications of Crystallographic Dimer for the Chemistry of Catalysis The nearly spherical nature of the dimer formed by two monomers of the integrase catalytic core places active sites on respective monomers on opposite sides of the dimer: approximately 35 Å separates the carboxylate oxygens of Asp64 of each monomer. While we are convinced that the observed dimer is not an artifact or consequence of crystallization, it would seem difficult to reconcile this distance with the observation that, during in vivo strand transfer, cuts on the target DNA occur with a separation of five base pairs, corresponding to 15–20 Å in B-form DNA. How can a single dimer accomplish this? One possibility is that the cuts do not occur simultaneously. One end of the viral DNA could be joined by a reaction at one active site, followed by carefully controlled movement of DNA and protein relative to one another such that the second active site is now positioned five base pairs away from the initial site of strand transfer. It has been proposed, in a variation on this theme, that the first strand-transfer

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_101.html [4/5/2004 4:51:57 PM]

Document

Page 102

reaction is followed by DNA relaxation (unwinding, rotation, etc.), resulting in the site on the target DNA for the second transfer reaction site now being located close to the active site of the second monomer [42]. An alternate possibility is that a multimer larger than a dimer is responsible for the coordinated cutting and strand-transfer reactions. For example, two contacting dimers can be modeled such that two active sites of the resulting tetramer are located 15–20 Å apart. It is also possible that even higher order multimers are involved. There is, as yet, no convincing evidence in support of any one model. The observation that RSV integrase cuts target DNA with a six-base-pair stagger rather than the five observed for HIV correlates intriguingly with the apparently longer distance (~ 38 Å vs. 35 Å) between active sites in the RSV dimer. However, understanding the coordinated cutting and joining reaction awaits three-dimensional information on the arrangement of monomers within an integrase multimer binding to DNA. E. Three-Dimensional Structures of Other Domains of HIV-1 Integrase Three-dimensional structural information has not yet been obtained for a full-length integrase protein. In its absence, attempts have been made to determine the structure of the smaller domains consisting of the separately expressed N- and C-termini that flank the core whose structure is now known. The Amino Terminus of Integrase While the N-terminus of HIV-1 integrase, consisting of residues 1 to 55, has been separately expressed, purified, and biophysically characterized [31], structural data has not yet been obtained. This protein domain binds metal ions such as Zn2+, Co2+, and Cd2+ stoichiometrically, and is monomeric at low protein concentrations. Dramatic changes in helix content (from 14% to 32%) are observed in the circular dichroism (CD) spectrum upon addition of metal. Analysis of CD spectral features led researchers to conclude that it is highly probably that integrase contains a zinc finger that folds in much the same way as the TFIIIA-like DNA binding proteins, with two His residues located on an α helix and two cysteines part of a β sheet [31]. However, confirmation of such a model awaits structure determination by x-ray crystallography or NMR spectroscopy. The Carboxy Terminus of Integrase When the C-terminal domain is expressed as a separate polypeptide, IN213–288 can be purified from the initial soluble fraction from cell lysates [33]. This small protein fragment, therefore, was an attractive target for structure determination.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_102.html [4/5/2004 4:51:59 PM]

Document

Page 103

Figure 8 Molscript stereo figure of the structure of the nonspecific DNA-binding domain of HIV-1 integrase, IN220–270, determined by heteronuclear NMR spectroscopy [28].

The structure of this domain is of particular interest as it represents the dominant nonspecific DNAbinding region of integrase. Gel filtration and sedimentation equilibrium results indicated that purified IN213–288 partitioned between dimers and highly aggregated material (unpublished observations). However, a smaller domain consisting of residues 220 to 270 maintains the DNA-binding properties of the longer C-terminal domain and is better behaved in solution. Recently, two groups have reported the structure of this smaller fragment, IN220–270, determined using multidimensional heteronuclear NMR spectroscopic methods [28,29]. As shown in Figure 8, the overall structure of IN220–270, is that of a β sandwich formed by two threestranded β sheets. As anticipated by biophysical studies, the polypeptide is a dimer in solution. The interface between monomers is formed by the antiparallel interaction of three β strands from each subunit and is stabilized predominantly by hydrophobic interactions. There is a long loop between strands β1 and β2 which, in the dimer, defines the sides of a cleft that is of the appropriate dimensions (about 24 × 24 × 12 Å) to accommodate double-stranded DNA. The folding topology is very similar to that of SH3 domains that are found in several proteins involved in signal transduction, despite the lack of significant sequence homology. This is rather unusual since SH3 domains are generally involved in protein binding rather than interactions with DNA. V. Prospects for Inhibitors A. Overview of Inhibitor Studies to Date The investigation of HIV integrase inhibitors has been largely restricted to testing available compounds that inhibit other enzymes with similar substrates or

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_103.html (1 of 2) [4/5/2004 4:52:07 PM]

Document

Page 104 Table 1 Representative Inhibitors of HIV-1 Integrase and IC50 values of compounds that inhibit 3' processing and strand transfer activities of HIV-1 integrase

Compound

IC50 of 3' processing (µM)

IC50 of strand transfer (µM)

Active against IN50–212?

Reference

I aurintricarboxylic acid monomer

10–50

n.d.

n.d.

44

II cosalane

n.d.

25

n.d.

45

III DHNQ

5.7

2.5

yes

47

IV primaquine

15

3.6

n.d.

46

V CAPE

220

19

yes*

46

VI quercetin

24

14

n.d.

47

VII quercetagetin

0.8

0.1

yes

47

VIII AG1717

0.4

0.16

yes

50

IX β-conidendrol

0.5

0.5

n.d.

51

X suramin

0.25

0.11

n.d.

53

XI curcumin

95

40

yes

55

XII (neocuproine)2-Cu+

3

3

yes

56

XIII AZT-monoPi

100–150

100–150

yes

57

XIV GT 17-mer

0.092

0.046

n.d.

59

XV HCKFWW

2

2

yes

43

The third column indicates if these compounds also inhibit disintegration by IN50–212. n.d. = not determined.* = only if pre-incubated.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_104.html (1 of 2) [4/5/2004 4:52:09 PM]

Document

proposed mechanisms. For example, as will be seen, many topoisomerase II inhibitors also inhibit HIV integrase. While screening of chemical databases is likely underway at several pharmaceutical companies, results have either not been made available or are discouraging (see below). One foray into integrase inhibitor design—rather than discovery—has recently been described using a peptide combinatorial library approach [43]. We summarize below published reports to date (March 1996) in which compounds have been identified that inhibit integrase in in vitro assays with IC50 values of 100 µM or less (IC50 is the concentration at which the measured activity is inhibited by 50%). In vitro inhibition data is compiled in Table 1; structures of selected compounds are shown in Figure 9. An Effective Pharmacophore: Multiple Hydroxyl Groups on Aromatic Rings The first report of a class of compounds that inhibits HIV integrase appeared in 1992 [44]. Aurintricarboxylic acid (I) and its derivatives, known to inhibit other enzymes that process nucleic acids, were determined to inhibit 3' processing with moderate IC50 values of 10–50 µM. As shown in Figure 9, a recurring structural theme was established early on in which integrase inhibitors often possess aromatic rings with multiple hydroxy substituents that are either located

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_104.html (2 of 2) [4/5/2004 4:52:09 PM]

Document

Page 105

Figure 9 Chemical structures of reported inhibitors of HIV-1 integrase. (I) aurintricarboxylic acid monomer; (II) cosalane; (III) dihydronaphthoquinone or DHNQ; (IV) primaquine; (V) caffeic acid phenethyl ester or CAPE; (VI) quercetin; (VII) quercetagetin; (VIII) AG1717; (IX) β-conidendrol; (X) suramin; (XI) curcumin.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_105.html (1 of 2) [4/5/2004 4:52:14 PM]

Document

Page 106

on the same ring or can potentially be positioned close together in three-dimensional space if rings stack on top of each other. More recently it was shown that cosalane (II), a steroid-substituted derivative of (I), was no better in inhibiting integrase in vitro than the parent compound, but showed promise in HIV cytopathicity cell-culture assays [45]. Although cosalane and a number of related analogues inhibit both HIV protease and integrase in vitro, the primary site of action is believed to be inhibition of gp120 binding to CD4 receptors. In 1993, the effects of selected topoisomerase II inhibitors, antimalarial agents, DNA binders, naphthoquinones, and various other agents on integrase activity in vitro were investigated [46]. Although certain effective topoisomerase inhibitors are also good HIV integrase inhibitors, this is not a generalizable correlation. Since many topoisomerase inhibitors also bind DNA, it is difficult to assess whether the observed in vitro effects result from specific interaction with integrase or from the sequestering or distorting of the DNA substrate. However, several compounds were identified that are not known to be DNA binders but that inhibited integrase with reasonable IC50 values. These included dihydroxynaphthoquinone (III), primaquine (IV), caffeic acid phenethyl ester (CAPE, V), and quercetin (VI). Motivated by the structural similarities between compounds III–VI, a more intensive structure-function study of flavones was undertaken in which approximately 50 related compounds were screened for inhibition of in vitro integrase activity [47]. Flavones are planar compounds containing three aromatic rings substituted with various polar groups such as hydroxy substituents. General trends were observed relating structure to inhibitory effectiveness; for example, inhibition required at least three hydroxy groups, the most favorable arrangement being when they were located ortho to one another. The most effective compound, quercetagetin (VII), is a potent topoisomerase II inhibitor and a known DNA intercalator. It was noted by the authors that many flavones are not integrase-specific; rather, they inhibit a broad range of enzymes including DNA polymerases, ATPases, and NAPDH-monooxygenases. They are also, in general, capable of DNA intercalation. It has not been established that their inhibitory effects are due to direct interactions with integrase. A subsequent detailed structure-activity relationship study of CAPE (V) revealed that while the ortho hydroxys were important for in vitro integrase inhibition, both the caffeic acid and phenethyl moieties could be substantially modified [48]. Ortho hydroxyl groups in the context of other classes of compounds also confer anti-integrase potency. For example, several semisynthetic compounds derived from arctigenin, a topoisomerase I inhibitor that itself is not active against integrase, have been shown to inhibit HIV integrase [49]. More compelling evidence of the importance of ortho hydroxyl groups was provided by the tyrphostins, a group of synthetic compounds that are tyrosine kinase inhibitors. Several of these also inhibit integrase in the submicromolar range

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_106.html [4/5/2004 4:52:16 PM]

Document

Page 107

(for example, compound VIII, aka AG1717). In cell-based screening assays, AG1717 demonstrated some antiviral activity [50]. Another study that demonstrated a role for ortho hydroxyl groups in in vitro integrase inhibition identified β-conidendrol (IX) via random screening as an inhibitor with an IC50 value of less than 1 µM [51]. Although β-conidendrol did not inhibit several other nucleic-acid processing enzymes, indicating some specificity for integrase, it was not active in cell-based antiviral assays at concentrations as high as 100 µM. Other Classes of Integrase Inhibitors Several compounds and their derivatives that do not contain adjacent hydroxy groups on a phenyl ring have recently been identified as HIV integrase inhibitors. These include suramin (X), curcumin (XI), phenanthroline-Cu+ complexes (XII), and 3'-azido-3'-deoxythymidylate (AZT) monophosphate (XIII). Suramin (X) is a known inhibitor of DNA and RNA polymerases, retroviral reverse transcriptases, and topoisomerase II. It has also been shown to prevent the infection of T lymphocytes by HIV in vitro [52]. Its six sulfonic acid groups confer a strong negative charge, and it was reasoned that there might be an inhibitory electrostatic interaction with the positive residues of the HIV C-terminus domain. Suramin was shown to be an effective inhibitor of 3' processing and strand transfer, with IC50 values of 0.25 µM and 0.11 µM, respectively [53]. It was not demonstrated, however, that the mechanism of inhibition does involve binding to the C-terminus of integrase, although this could be readily addressed using Cterminal truncated mutants active for disintegration. Curcumin (XI), the coloring dye in the spice turmeric, is structurally related to CAPE (V). It has also been shown to inhibit HIV replication by inhibiting p24 antigen production and tat-mediated transcription [54]. As shown in Table 1, it also has moderate integrase inhibitory properties [55]. Although its twoOH groups are neither adjacent to each other nor on the same phenyl ring, its conformations can be modeled to bring the hydroxy groups into close proximity by stacking the two phenyl rings on each other. Several tetrahedral cuprous phenanthroline complexes, known inhibitors of transcription, were tested against integrase and shown to be reasonably effective inhibitors [56]: IC50 values in the range of 1–10 µM were determined (for example, the neocuproine-Cu+ complex, XII). Analyses of the mode of inhibition demonstrated that these compounds act noncompetitively, and that inhibition does not correlate with inhibition of DNA binding. Thus, it has been proposed that these metal chelates may act at a site distant from the active site, or perhaps in the context of an enzyme-DNA complex. 3'-Azido-3'-deoxythymidine, or AZT, is a nucleoside analog approved for use to treat AIDS. Its metabolites, the mono-, di- and triphosphate forms, accu-

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_107.html (1 of 2) [4/5/2004 4:52:18 PM]

Document

Page 108

mulate during treatment; in particular, AZT-monoPi (XIII) accumulates in cells to millimolar levels. These metabolites were investigated as possible integrase inhibitors and it was shown that all three phosphate derivatives inhibit with IC50 values of 100–150 µM, although AZT itself is not inhibitory [57]. These results suggest that despite the weak inhibition by these particular compounds, nucleoside analogs may serve as lead compounds for the development of integrase inhibitors. It was recently observed that oligonucleotides that form guanosine quartet structures inhibit HIV replication [58]. In light of the AZT results that suggested that there may be a nucleotide binding site on integrase—and because integrase binds DNA—these oligonucleotides (for example, 5'GTGGTGGGTGGGTGGGT-3', XIV) were investigated as possible inhibitors [59]. These compounds have the lowest inhibition constants reported to date (see Table 1), and suggest an exciting new avenue for integrase inhibitor development. In a completely different approach to integrase inhibitors, a synthetic peptide combinatorial library was used to select a hexapeptide capable of inhibiting integrase proteins [43]. The first two amino acids were selected using a library of 400 dipeptides, and the remaining amino acids selected one-by-one in an iterative process. The optimal hexapeptide, HCKFWW (XV), inhibits HIV-1 integrase with an IC50 of 2 µM. The peptide does not compete for DNA binding to viral DNA, nor does it represent a sequence present in integrase itself. Although it is not expected that a peptide consisting of L-amino acids would be a suitable drug in itself, the use of D-amino acids or peptidomimetic backbones may be fruitful directions to pursue. Summary Many of the compounds identified to date that inhibit HIV integrase in vitro have common structural features as illustrated in Figure 9. Most notably, these include hydroxy-substitued phenyl rings. However, these compounds may inhibit in vitro activity for reasons unrelated to enzyme binding. For example, it is difficult to know what to make of inhibition studies where the compounds added to the assays are known to bind DNA. Do the compounds affect in vitro activity because they bind and sequester the substrate? It is possible that they distort the DNA or intercalate between base pairs, preventing appropriate binding to the enzyme. While this is itself is a valid basis for the design of inhibitors against HIV infection, particularly if it targets DNA specific to the virus such as the LTR sequences [60], it is not an approach that builds on knowledge of the three-dimensional structure of the protein. To this end, valuable information could be obtained from studies in which direct binding of compounds to integrase is measured. This is also true for those inhibitors that do not contain the hydroxysubstituted phenyl pharmacophore. Binding studies could be carried

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_108.html [4/5/2004 4:52:19 PM]

Document

Page 109

out, for example, using radiolabeled inhibitors or possibly by monitoring any UV-spectral shifts in the case of aromatic compounds. B. What We Need to Know before We can Start Structure-Based Drug Design The review of known integrase inhibitors presented in the previous section demonstrates the paucity of effective inhibitors reported in the literature. A limited number of structural types have been investigated, focused heavily on compounds with aromatic ring systems and hydroxy substituents. Most inhibitors reported to date are only moderately effective in in vitro assays, with IC50 values residing in the low µM range (see Table 1). To build on this knowledge base, and to use known molecules as lead compounds for the development of more effective inhibitors, it would be extremely valuable to obtain a high-resolution crystal structure of an integrase-inhibitor complex. We and others are vigorously pursuing this goal. It is not clear if the lack of success to date is because the inhibitors identified so far manifest their effects in in vitro assays predominantly at the level of the DNA, or if some aspect of the structure of the HIV-1 integrase core itself—for example, its mobility in certain regoins—prevents the formation of a tight complex. It is also possible that binding is weaker to the HIV-1 core domain than to the full-length protein because of missing enzyme-inhibitor contacts. Co-crystallization attempts would benefit from in vitro studies to determine relative binding constants as a guide in selecting the most tightly bound inhibitors. It would also be useful to obtain information on the effect of variables, such as Mn2+ or Mg2+, on binding constants. It may be futile at this stage to attempt to model the binding of known inhibitors to the catalytic core domain of HIV-1 integrase in the absence of more complete information. It cannot a priori be assumed that the site of action of all these inhibitors is the enzyme active site identified by the constellation of conserved acidic residues. For example, certain very effective nonnucleoside inhibitors of HIV reverse transcriptase bind not to the enzyme active site, but rather to a small pocket adjacent to it. There are no obvious structural features of the integrase core—such as the deep trough surrounding active site residues in the case of HIV protease [61,62]—that can be readily identified as a potential inhibitor binding site. Furthermore, since part of the region defining the active site of HIV-1 integrase is disordered in our crystal structure, this prevents a surface rendering of the region around the conserved acidic residues. It also seems unlikely, given the known differences in active-site geometries between the HIV and ASV integrase (see Section IV.B), that the structure of the related ASV integrase core domain by itself would be particularly useful in this regard. As the active form of the enzyme presumably binds divalent metal ions, it will be important to determine how or if the structure of HIV integrase changes

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_109.html [4/5/2004 4:52:21 PM]

Document

Page 110

when metals are bound. Finally, there is no evidence to rule out the possibility that the two termini contribute to part of the active site, occlude some of it, or restrict access to it in as yet undetermined ways. For this and other obvious reasons, three-dimensional structures of larger versions of HIV-1 integrase, such as IN1–212, IN50–288, and the full-length protein, IN1–288, will be required. We also lack a clear picture of how the enzyme substrates, the viral DNA ends and the target DNA, bind to integrase. The DNA must at some point approach the region defined by the three conserved acidic residues so that bond cleavage and joining can occur. However, the dominant DNA binding domain is defined by residues in the C-terminus. It would be extremely valuable to determine the relative orientation of these domains in the context of a larger version of integrase. Even more revealing would be the structure of the full-length protein with bound DNA. Once we possess this information, it should be possible to rapidly progress with structure-based drug design. C. Possible Approaches to the Design of Effective Integrase Inhibitors There are a variety of approaches to the design of integrase inhibitors that are obvious and do not depend on knowing its three-dimensional structure. However, the rational implementation and refinement of these approaches will require high-resolutional structural data, much of which, as indicated above, is not yet available. It still may be useful to discuss here different classes of inhibitors that can be envisioned. Preventing DNA Binding One approach to the inhibition of integrase would be to prevent binding of the DNA substrate. Unfortunately, we do not yet know how or where DNA binds. There are likely to be several sites on the enzyme that contact DNA, including the C-terminus and the region around the active site. In the absence of the structure of an integrase-DNA complex, structures of related enzymes (RNase H, the MuA transposase, and RuvC) with their DNA substrates would be useful guides in suggesting ways in which DNA could interact with integrase. However, there is no guarantee that modes of DNA binding are conserved among members of this polynucleotidyl transferase family. The overall structure of the Cterminus fragment suggests that it should be possible to develop compounds that bind specifically in the cleft formed by dimers of IN220–270 (see Section IV.E) which may prevent DNA binding. Inhibition at the Active Site It may be possible to inhibit integrase by preventing the binding of the required metal cofactor(s) to the active site acidic residues that are presumed to provide

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_110.html [4/5/2004 4:52:23 PM]

Document

Page 111

coordinating ligands to the metal(s). This could be accomplished by sterically blocking access to the active site or by specifically binding the acidic residues themselves. Such a mechanism has been suggested to explain the inhibition of integrase by curcumin [55], which could bind to the active site aspartates or glutamates via its hydroxy groups. Compounds could also be devised that chelate the metal(s) once bound, preventing access of the active site to the substrate or distorting the active site geometry preventing phosphate bond cleavage. An intriguing approach to disrupting metal binding has recently been reported for the Zn2+-binding HIV nucleocapsid (NC) protein [63]. In this case, compounds were developed that specifically eject Zn2+ from the zinc-finger region of NC, interfering with viral replication. Such an approach might be applied to Mn2+ or Mg2+ binding at the active site, although the Zn2+-binding domain at the N terminus of integrase also suggests itself as a target. It is curious that approaches have not been devised for mechanism-based inhibitors, particularly since the stereochemical mechanism of integration has been understood for some time now [8]. Interfering with Multimerization The structures of the domains of HIV-1 integrase determined to date both reveal dimers [3,28,29,36]. It may be possible to develop compounds that bind specifically to the dimer interfaces, preventing interactions between monomers that may be necessary for activity. The success of this approach presumes that during the retroviral lifecycle there is a point where the monomer surfaces are accessible. It is not clear that integrase in the context of preintegration complexes is in equilibrium between the monomeric and higher order forms. This approach need not be restricted to preventing dimer formation if higher order interactions (e.g. formation of a tetramer) are also mechanistically relevant. To this end, the structure of an integrase tetramer, such as that formed by the full-length protein, would be useful in identifying dimer-dimer interface(s). Other Ways to Confound Integrase There are other parts of the retroviral life cycle involving integrase that could be targeted for inhibition. For example, integrase can bind other proteins such as human Ini1 [64], and most likely interacts with the viral proteins that are part of the preintegration complex [65]. Although it is not known if these interactions are essential for viral replication, preventing protein-protein binding would provide another site at which to attack integrase. It is possible that interactions between integrase and other proteins in the preintegration complex are crucial for maintaining the integrity of this complex and its ability to migrate to the cell nucleus. Interfering with these protein-protein or protein-nucleic acid interactions may be another approach to halting viral replication. For example, prein-

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_111.html [4/5/2004 4:52:24 PM]

Document

Page 112

cubation of phosphotyrosine with integrase has been shown to inhibit interaction with the MA protein [65]. Therefore, phosphotyrosine analogs could be a unique approach to antiviral development. D. Concluding Statement It is known from drug design studies with HIV reverse transcriptase and protease that the virus is able to escape from the pressures of inhibitors by mutation of the drug targets [66]. Although integrase is, a priori, a reasonable target for drug-design efforts, it must be anticipated that integrase will also be able to rapidly mutate and thereby avoid inhibition. By analogy with recent approaches to reverse transcriptase inhibitors, it may be possible to design a set of integrase inhibitors that act at slightly different binding sites and from which the virus cannot simultaneously escape. That is, the combination of mutations required to avoid inhibition may be severe enough to prevent integrase from carrying out its required chemistry. (In the absence of direct structural information on the sites of inhibitor binding to integrase, the generation of escape mutants in vitro and their subsequent sequencing may be an indirect way to identify inhibitor binding sites.) It will also be important to determine if the virus will be able to simultaneously mutate reverse transcriptase, integrase, and the protease in response to a combination of inhibitors targeted against all three pol gene products. The answers to these questions will require the development of suitable inhibitors and the beginning of in vitro testing. To this end—while large-scale screening and the development of combinatorial chemistry methods should continue—the structure of the catalytic core domain of HIV-1 integrase is a starting point for the rational design of integrase inhibitors. There is much more structural information that must be obtained for the full-length protein and its complexes with metals, inhibitors, and substrates. We and others are aggressively pursuing results on these fronts. E. Recent Developments Several studies published since March 1996 have expanded the list of in vitro integrase inhibitors effective at IC50 values below 100 µM. These include two dicaffeoylquinic acids obtained from medicinal plants and a synthetic analog, L-chicoric acid [68], the HIV protease inhibitors NSC 117027 and NSC 158393 [69], certain anthraquinone derivatives [70], coumermycin, and pyridoxal phosphate [71]. In addition to exhibiting in vitro inhibition, the dicaffeoylquinic acids effectively inhibited HIV-1 replication in T-lymphoblastoid cell lines [68].

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_112.html [4/5/2004 4:52:26 PM]

Document

Page 113

Follow-up studies were also reported for two previously identified classes of integrase inhibitors. Several nucleotides that were more effective inhibitors than the originally tested AZT nucleotides were identified [71]. For example, the L-enantiomers of 5-fluoro-2',3'-dideoxycytidine monophosphate and triphosphate inhibit 3' processing and strand transfer with IC50 values of ~40 µM. A structure-function study on GT-containing oligonucleotides showed that both the number of quartets formed and the loop sequences between the quartets are important for activity, and that inhibitors of this type may function by interacting with the N-terminus of integrase [72]. A particularly important contribution was the demonstration that preintegration complexes isolated from HIV-infected lymphoid cells can be used in assays to screen for inhibition of integration [70]. Intriguingly, many compounds previously identified as in vitro inhibitors of 3' processing and strand transfer had no effect on integration carried out by either crude or partially purified preintegration complexes. Thus, such an assay may be a valuable method of screening out “false positives” identified using in vitro oligonucleotide assays, or corroborating the evidence that a particular compound is indeed active against integrase. Acknowledgments Our work described here was carried out in the laboratories of R. Craigie and D. R. Davies. We express our gratitude to our co-workers who, over the years, participated in the effort to determine the structure of HIV integrase. In particular, we acknowledge the contributions of F. D.Bushman, M. Carmichael, A. Engelman, S. Hosseini, T. Jenkins, K. Mizuuchi, I. Palmer, P. Rice, P. Sun, and P. Wingfield. We would also like to thank D. R. Davies and T. Jenkins for their comments on the manuscript and A. Mazumder for alerting us to the most recent work on integrase inhibitors and for his contributions to Section V.A. References 1. AIDS WEEKLY Plus. Key KK, ed. Atlanta, Charles Henderson Publisher, 1996: Feb. 5 & 12, p. 14. 2. Cara A, Guarnaccia F, Reitz, Jr. MS, Gallo RC, Lori F. Self-limiting, cell type-dependent replication of an integrase-defective human immunodeficiency virus type 1 in human primary macrophages but not T lymphocytes. Virol 1995, 208:242–248. 3. Dyda F, Hickman AB, Jenkins TM, Engelman A, Craigie R, Davies DR. Crystal structure of the catalytic domain of HIV-1 integrase: Similarity to other polynucleotidyl transferases. Science 1994; 266:1981–1986.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_113.html [4/5/2004 4:52:27 PM]

Document

Page 114

4. Vink C, Plasterk RHA. The human immunodeficiency virus integrase protein. Trends Genet 1993; 9:433–437. 5. Katz RA, Skalka AM. The retroviral enzymes. Annu Rev Biochem 1994; 63:133–173. 6. Sherman PA, Fyfe JA. Human immunodeficiency virus integration protein expressed in Escherichia coli possesses selective DNA cleaving activity. Proc Natl Acad Sci USA 1990; 87:5119–5123. 7. Bushman FD, Craigie R. Activities of human immunodeficiency virus (HIV) integration protein in vitro: Specific cleavage and integration of HIV DNA. Proc Natl Acad Sci USA 1991; 88:1339–1343. 8. Engelman A, Mizuuchi K, Craigie R. HIV-1 DNA integration: Mechanism of viral DNA cleavage and DNA strand transfer. Cell 1991; 67:1211–1221. 9. Chow SA, Vincent KA, Ellison V, Brown PO. Reversal of integration and DNA slicing mediated by integrase of human immunodeficiency virus. Science 1992; 255:723–726. 10. Grandgenett DP, Vora AC, Schiff RD. A 32,000-dalton nucleic acid-binding protein from avian retravirus cores possesses DNA endonuclease activity. Virol 1978; 89:119–132. 11. Vincent KA, Ellison V, Chow SA, Brown PO. Characterization of human immunodeficiency virus type 1 integrase expressed in Escherichia coli and analysis of variants with amino-terminal mutations. J. Virol 1993; 67:425–437. 12. Hickman AB, Palmer I, Engelman A, Craigie R, Wingfield P. Biophysical and enzymatic properties of the catalytic domain of HIV-1 integrase. J Biol Chem 1994; 269:29279–29287. 13. Jones KS, Coleman J, Merkel GW, Laue TM, Skalka AM. Retroviral integrase functions as a multimer and can turn over catalytically. J Biol Chem 1992; 267:16037–16040. 14. Engelman A, Bushman FD, Craigie R. Identification of discrete functional domains of HIV-1 integrase and their organization within an active multimeric complex. EMBO J 1993; 12:3269–3275. 15. Andrake MD, Skalka AM. Multimerization determinants reside in both the catalytic core and C terminus of avian sarcoma virus integrase. J Biol Chem 1995; 270:29299–29306. 16. Kalpana GV, Goff SP. Genetic analysis of homomeric interactions of human immunodeficiency virus type 1 integrase using the yeast two-hybrid system. Proc Natl Acad Sci USA 1993; 90:10593–10597. 17. Van Gent DC, Vink C, Oude Groeneger AAM, Plasterk RHA. Complementation between HIV integrase proteins mutated in different domains. EMBO J 1993; 12:3261–3267.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_114.html (1 of 2) [4/5/2004 4:52:29 PM]

Document

18. Johnson MS, McClure MA, Feng D–F, Gray J, Doolittle RF. Computer analysis of retroviral pol genes: Assignment of enzymatic functions to specific sequences and homologies with nonviral enzymes. Proc Natl Acad Sci USA 1986; 83:7648–7652. 19. Engelman A, Craigie R. Identification of conserved amino acid residues critical for human immunodeficiency virus type 1 integrase function in vitro. J Virol 1992; 66:6361–6369. 20. Van Gent DC, Oude Groeneger AAM, Plasterk RHA. Mutational analysis of the integrase protein of human immunodeficiency virus type 2. Proc Natl Acad Sci USA 1992; 89:9598–9602.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_114.html (2 of 2) [4/5/2004 4:52:29 PM]

Document

Page 115

21. Kulkosky J, Jones KS, Katz RA, Mack JPG, Skalka AM. Residues critical for retroviral integrative recombination in a region that is highly conserved among retroviral/retrotransposon integrases and bacterial insertion sequence transposases. Mol Cell Biol 1992; 12:2331–2338. 22. Bushman FD, Engelman A, Palmer I, Wingfield P, Craigie R. Domains of the integrase protein of human immunodeficiency virus type 1 responsible for polynucleotidyl transfer and zinc binding. Proc Natl Acad Sci USA 1993; 90:3428–3432. 23. Beese LS, Steitz TA. Structural basis for the 3'-5' exonuclease activity of Escherichia coli DNA polymerase I: a two metal ion mechanism. EMBO J 1991; 10:25–33. 24. Woerner AM, Marcus-Sekura CJ. Characterization of a DNA binding domain in the C-terminus of HIV-1 integrase by deletion mutagenesis. Nucl Acids Res 1993; 21:3507–3511. 25. Vink C, Oude Groeneger AAM, Plasterk RHA. Identification of the catalytic and DNA-binding region of the human immunodeficiency virus type 1 integrase protein. Nucl Acids Res 1993; 21:1419–1425. 26. Puras Lutzke RA, Vink C, Plasterk RHA. Characterization of the minimal DNA-binding domain of the HIV integrase protein. Nucl Acids Res 1994; 22:4125–4131. 27. Drelich M, Wilhelm R, Mous J. Identification of amino acid residues critical for endonuclease and integration activities of HIV-1 IN protein in vitro. Virol 1992; 188:459–468. 28. Lodi PJ, Ernst JA, Kuszewski J, Hickman AB, Engelman A, Craigie R, Clore GM, Gronenborn AM. Solution structure of the DNA binding domain of HIV-1 integrase. Biochem 1995; 34:9826–9833. 29. Eijkelenboom APAM, Puras Lutzke RA, Boelens R, Plasterk RHA, Kaptein R, Hard K. The DNAbinding domain of HIV-1 integrase has an SH3-like fold. Nature Struct Biol 1995; 2:807–810. 30. Haugan IR, Nilsen BM, Worland S, Olsen L, Helland DE. Characterization of the DNA-binding activity of HIV-1 integrase using a filter binding assay. Biochem Biophys Res Commun 1995; 217:802–810. 31. Burke CJ, Sanyal G, Bruner MW, Ryan JA, LaFemina RL, Robbins HL, Zeft AS, Middaugh CR, Cordingley MG. Structural implications of spectroscopic characterization of a putative zinc finger peptide from HIV-1 integrase. J Biol Chem 1992; 267:9639–9644. 32. Bushman FD, Wang B. Rous sarcoma virus integrase protein: Mapping functions for catalysis and substrate binding. J Virol 1994; 68:2215–2223. 33. Engelman A, Hickman AB, Craigie R. The core and carboxyl-terminal domains of the integrase protein of human immunodeficiency virus type 1 each contribute to nonspecific DNA binding. J Virol 1994; 68:5911–5917. http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_115.html (1 of 2) [4/5/2004 4:52:31 PM]

Document

34. Jenkins TM, Hickman AB, Dyda F, Ghirlando R, Davies DR, Craigie R. Catalytic domain of human immunodeficiency virus type 1 integrase: Identification of a soluble mutant by systematic replacement of hydrophobic residues. Proc Natl Acad Sci USA 1995; 92:6057–6061. 35. Orengo CA, Thornton JM. Alpha plus beta fold revisited: some favoured motifs. Structure 1993; 1:105–120.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_115.html (2 of 2) [4/5/2004 4:52:31 PM]

Document

44. Cushman M, Sherman P. Inhibition of HIV-1 integration protein by aurintricarboxylic acid monomers, monomer analogs, and polymer fractions. Biochem Biophys Res Commun 1992; 185:85–90. 45. Cushman M, Golebiewski WM, Pommier Y, Mazumder A, Reymen D, De Clercq E, Graham L, Rice WG. Cosalane analogues with enhanced potencies as inhibitors of HIV-1 protease and integrase. J Med Chem 1995; 38:443–452. 46. Fesen MR, Kohn KW, Leteurte F, Pommier Y. Inhibitors of human immunodeficiency virus integrase. Proc Natl Acad Sci USA 1993; 90:2399–2403. 47. Fesen MR, Pommier Y, Leteurtre F, Hiroguchi S, Yung J, Kohn KW. Inhibition of HIV-1 integrase by flavones, caffeic acid phenethl ester (CAPE) and related compounds. Biochem Pharmacol 1994; 48:595–608. 48. Burke Jr TR, Fesen MR, Mazumder A, Wang J, Carothers AM, Grunberger D, Driscoll J, Kohn K, Pommier Y. Hydroxylated aromatic inhibitors of HIV-1 integrase. J Med Chem 1995; 38:4171–4178. 49. Eich E, Pertz H, Kaloga M, Schulz J, Fesen MR, Mazumder A, Pommier Y. (-)-Arctigenin as a lead structure for inhibitors of human immunodeficiency virus type-1 integrase. J Med Chem 1996; 39:86–95. 50. Mazumder A, Gazit A, Levitzki A, Nicklaus M, Yung J, Kohlhagen G, Pommier Y. Effects of tyrphostins, protein kinase inhibitors, on human immunodeficiency virus type 1 integrase. Biochem 1995; 34:15111–15122. 51. LaFemina RL, Graham PL, LeGrow K, Hastings JC, Wolfe A, Young SD, Emini EA, Hazuda DJ. Inhibition of human immunodeficiency cirus integrase by bis-catechols. Antimicrob Agents Chemother 1995; 39:320–324. 52. Mitsuya H, Popovic M, Yarchoan R, Matsushita S, Gallo RC, Broder S. Suramin protection of T cells in vitro against infectivity and cytopathic effect of HTLV-III. Science 1984; 226:172–174.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_116.html (2 of 2) [4/5/2004 4:52:33 PM]

Document

Page 117

53. Carteau S, Mouscadet JF, Goulaouic H, Subra F, Auclair C. Inhibitory effect of the polyanionic drug suramin on the in vitro HIV DNA integration reaction. Arch Biochem Biophys 1993; 305:606–610. 54. Li CJ, Zhang LJ, Dezube BJ, Crumpacker CS, Pardee AB. Three inhibitors of type 1 immunodeficiency virus long terminal repeat-directed gene expression and virus replication. Proc Natl Acad Sci USA 1993; 90:1839–1842. 55. Mazumder A, Raghavan K, Weinstein J, Kohn KW, Pommier Y. Inhibition of human immunodeficiency virus type-1 integrase by curcumin. Biochem Pharmacol 1995; 49:1165–1170. 56. Mazumder A, Gupta M, Perrin DM, Sigman DS, Rabinovitz M, Pommier Y. Inhibition of human immunodeficiency virus type 1 integrase by a hydrophobic cation: The phenanthroline-cuprous complex. AIDS Res Human Retro 1995; 11:115–125. 57. Mazumder A, Cooney D, Agbaria R, Gupta M, Pommier Y. Inhibition of human immunodeficiency virus type 1 integrase by 3' -azido-3' -deoxythymidylate. Proc Natl Acad Sci USA 1994; 91:5771–5775. 58. Ojwang J, Elbaggari A, Marshall HB, Jayaraman K, McGrath MS, Rando RF. Inhibition of human immunodeficiency virus type 1 activity in vitro by oligonucleotides composed entirely of guanosine and thymidine. J Acqu Immune Defic Syn 1994; 7:560–570. 59. Ojwang JO, Buckheit RW, Pommier Y, Mazumder A, de Vreese K, Esté JA, Reymen D, Pallansch LA, Lackman-Smith C, Wallace TL, de Clercq E, McGrath MS, Rando RF. T30177, an oligonucleotide stabilized by an intramolecular guanosine octet, is a potent inhibitor of laboratory strains and clinical isolates of human immunodeficiency virus type 1. Antimicrob Agents Chemother 1995; 39:2426–2435. 60. Carteau S, Mouscadet JF, Goulaouic H, Subra F, Auclair C. Inhibition of the in vitro integration of Moloney murine leukemia virus DNA by the DNA minor groove binder netropsin. Biochem Pharmacol 1994; 47:1821–1826. 61. Navia MA, Fitzgerald PMD, McKeever BM, Leu C-T, Heimbach JC, Herber WK, Sigal IS, Darke PL, Springer JP. Three-dimensional structure of aspartyl protease from human immunodeficiency virus HIV-1. Nature 1989; 337:615–620. 62. Wlodawer A, Erickson JW. Structure-based inhibitors of HIV-1 protease. Annu Rev Biochem 1993; 62:543–585. 63. Rice WG, Supko JG, Malspeis L, Buckheit Jr RW, Clanton D, Bu M, Graham L, Schaeffer CA, Turpin JA, Domagala J, Gogliotti R, Bader JP, Halliday SM, Coren L, Sowder II RC, Arthur LO, Henderson LE. Inhibitors of HIV nucleocapsid protein zinc fingers as candidates for the treatment of AIDS. Science 1995; 270:1194–1197. 64. Kalpana GV, Marmon S, Wang W, Crabtree GR, Goff SP. Binding and stimulation of HIV-1 integrase by a human homolog of yeast transcription factor SNF5. Science 1994; 266:2002–2006. http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_117.html (1 of 2) [4/5/2004 4:56:01 PM]

Document

65. Gallay P, Swingler S, Song J, Bushman F, Trono D. HIV nuclear import is governed by the phosphotyrosine-mediated binding of matrix to the core domain of integrase. Cell 1995; 83:569–576. 66. Coffin JM. HIV population dynamics in vivo: Implications for genetic variation, pathogenesis, and therapy. Science 1995; 267:483–489. 67. Williams KJ, Loeb LA. Retroviral reverse transcriptases: Error frequencies and mutagenesis. Curr Top Microbiol Immunol 1992; 176:165–180.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_117.html (2 of 2) [4/5/2004 4:56:01 PM]

Document

Page 118

68. Robinson Jr WE, Reinecke MG, Abdel-Malek S, Jia Q, Chow SA. Inhibitors of HIV-1 replication that inhibit HIV integrase. Proc Natl Acad Sci USA 1996; 93:6326–6331. 69. Mazumder A, Wang S, Neamati N, Nicklaus M, Sunder S, Chen J, Milne GWA, Rice WG, Burke Jr TR, Pommier Y. Antiretroviral agents as inhibitors of both human immunodeficiency virus type 1 integrase and protease. J Med Chem 1996; 39;2472–2481. 70. Farnet CM, Wang B, Lipford JR, Bushman FD. Differential inhibition of HIV-1 preintegration complexes and purified integrase protein by small molecules. Proc Natl Acad Sci USA 1996; 93:9742–9747. 71. Mazumder A, Neamati N, Sommadossi J, Gosselin G, Schinazi RF, Imbach J, Pommier Y. Effects of nucleotide analogues on human immunodeficiency virus type 1 integrase. Mol Pharmacol 1996; 49:621–628. 72. Mazumder A, Neamati N, Ojwang JO, Sunder S, Rando RF, Pommier Y. Inhibition of the human immunodeficiency virus type 1 integrase by guanosine quartet structures. Biochem 1996; 35:13762–13771.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_118.html [4/5/2004 4:56:03 PM]

Document

Page 119

4 Bradykinin Receptor Antagonists Donald J. Kyle Scios Nova Inc., Sunnyvale, California I. Introduction The term “kinins” is generally made in reference to either the nonapeptide bradykinin (Arg1-Pro2-Pro3Gly4-Phe5-Ser6-Pro7-Phe8-Arg9) or the decapeptide kallidin (Lys1-Arg2-Pro3-Pro4-Gly5-Phe6-Ser7-Pro8Phe9-Arg10). In rats another kinin, Ile1-Ser2-Arg3-Pro4-Pro5-Gly6-Phe7-Ser8-Pro9-Phe10-Arg11 (T-kinin) is produced under certain circumstances and binds to the same receptors as bradykinin [1,2]. A schematic of the human kinin-kallikrein system is shown in Figure 1. The release of kinins from precursor proteins (known as kininogens) is mediated by enzymes called kininogenases [3–5]. The predominant enzymes responsible are kallikreins, but others, which include trypsin, plasmin, and some snake venoms, also release kinins. Kininogens are primarily synthesized in the liver and represent an abundant source of the precursors that are required for kinin generation. These proteins are produced from alternative splicing of a single gene product and there are two forms: high molecular weight kininogen (HMWK) and low molecular weight kininogen (LMWK) [6]. Unlike HMWK, which exists in the circulation as a complex with plasma pre-kallikrein, LMWK circulates freely. During immunological reaction, charged surfaces—which may be derived from bacterial lipopolysaccharide, oligosaccharides, connective tissue proteoglycans, or damaged basement membranes—facilitate the conversion of factor XII to factor XIIa. Once factor XIIa is present, prekallikrein can be cleaved to its active form, known as plasma kallikrein. This enzyme acts upon its preferred substrate, HMWK, to release bradykinin. Plasma kallikrein is further able to convert inactive factor XII to active XIIa, thereby participating in a positive

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_119.html [4/5/2004 4:56:06 PM]

Document

Page 120

Figure 1 Diagram of the human kinin-kallikrein system including the native ligands for B1 and B2 receptor subtypes.

feedback loop. The cleavage of bradykinin from HMWK is highly localized since pre-kallikrein and substrate (HMWK) circulate as a complex. Another kinin, Lys-bradykinin (also known as kallidin), is produced via the action of the tissuekallikrein enzyme on LMWK. This enzyme is found in many tissues, either in the form of a precursor requiring activation or as an active enzyme. In contrast to plasma kallikrein, which preferentially acts upon HMWK, tissue kallikrein can release kallidin from either HMWK or LMWK. Through the action of aminopeptidases, kallidin can subsequently be converted directly into bradykinin. This enzyme is present in both the plasma and on the surface of epithelial cells. Both bradykinin and kallidin can be degraded by a variety of plasma and cell surface enzymes (kininases) [7]. The most widely recognized of these enzymes are kininase I, kininase II (angiotensin converting enzyme, ACE), and carboxypeptidase N. In plasma, kininase I cleaves the C-terminal arginine from both bradykinin and kallidin to form [des-Arg9] kinins. These [des-Arg9] kinins are known to act as agonists of B1 receptors that are present in some species and have been implicated in the pathophysiology associated with prolonged inflammation [8–10].

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_120.html (1 of 2) [4/5/2004 4:56:27 PM]

Document

Page 121

Nearly all cells express kinin receptors that mediate the activities of both bradykinin and kallidin. The activation of these G-protein coupled receptors causes relaxation of venular smooth muscle and hypotension, increased vascular permeability, contraction of smooth muscle of the gut and airway leading to increased airway resistance, stimulation of sensory neurons, alteration of ion secretion of epithelial cells, production of nitric oxide, release of cytokines from leukocytes, and the production of eicosanoids from various cell types [11,12]. Because of this broad spectrum of activity, kinins have been implicated as an important mediator in many pathophysiologies including pain, sepsis, asthma, rheumatoid arthritis, pancreatitis, and a wide variety of other inflammatory diseases. Moreover, a recent report demonstrated that bradykinin B2 receptors on the surface of human fibroblasts were upregulated three-fold beyond normal in patients with Alzheimer's disease, implicating bradykinin as a participant in the peripheral inflammatory processes associated with that disease [13]. In contrast to the adverse physiologies associated with bradykinin release, there is a growing body of literature that implicates bradykinin as a protective agent during periods of cardiac or renal stress [14–16]. In this regard there is substantial evidence that the cardioprotective effects afforded by ACEinhibitor treatment are a result of metabolically preserving bradykinin and are therefore mediated by bradykinin B2 (and possibly B1) receptors [17–18]. These results point to a possible therapeutic role for a kinin receptor agonist. Overall, the kinins are an important part of a well-organized physiological system. The various aspects and interdependencies of the kinin system have been, and continue to be, the focus of intensive research efforts in many laboratories. Many pharmaceutical companies have identified this system as an ideal site for therapeutic intervention in many inflammatory diseases. Hence, there have been many diverse approaches taken toward the discovery of antagonists (peptide and nonpeptide) of B2 and B1 receptors. This review focuses on the structure-based design strategies pursued in our laboratories during the past several years. II. Ligand-Based Investigations A. The Solution Conformation of Bradykinin In the late 1980s when we began the pursuit of bradykinin receptor antagonists, information of relevance to medicinal chemists was scarce. For example, not one nonpeptide antagonist of this receptor was known, nor were any series upon which to base a structure-activity relationship. Moreover, all publications described bradykinin as being highly flexible in an aqueous environment, such that no structural mimetics could be rationalized. Of course the receptors had not been cloned at that time so nothing was known about the primary sequence of the receptor or the three-dimensional structure.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_121.html [4/5/2004 4:56:29 PM]

Document

Page 122

Initially our approach was to complete a detailed examination of bradykinin using two-dimensional NMR methods in combination with empirical energy calculations [19]. Our strategy was derived on the basis of spectral data, biological results from conformationally restricted analogs, as well as the relationship between ordering in bradykinin and the dielectric environment of the solvent. Our guiding hypothesis was that, although in aqueous solution bradykinin is conformationally random, the biologically active form of the peptide is likely ordered and stabilized within the lipid bilayer of the cell membrane prior to binding with its receptor. Alternatively, the receptor binding environment might also be hydrophobic and thereby lead to similar conformational biases in the ligand. We presumed that an appropriate solvent environment should be able to stimulate, at least in terms of hydrophobicity and dielectric constant, the nature of a cell membrane, and a 90:10 d8-dioxane-H2O mixture was selected for NMR experiments. It was anticipated that under these nonsolvating conditions the conformational diversity of bradykinin might be severely restricted. The ultimate analysis of the two-dimensional NMR data collected at 500 MHz supported a single major conformational species. There were five HN-CαH connectivities, one for each amide. This was confirmed in the 13C NMR spectrum where only nine carbonyl resonances, one for each amino acid, were present. In order to provide a starting point for subsequent molecular dynamics simulations the assumption was made, based on multiple observed long-range amide-to-amide nuclear overhauser effects (NOEs), that it was indeed a single major conformational species. Although bradykinin contains three proline residues, the absence of any strong CαHi-CαHj+, cross peaks in the nuclear overhauser enhancement spectroscopy (NOESY) spectrum was taken as proof that all peptide bonds were trans. In total, 35 interproton distances were extracted from the NOESY spectrum and, whenever possible, stereospecific assignments for pro-R and pro-S hydrogens were made explicitly. A temperature-dependent study of the chemical shifts of the amide protons resulted in a near-linear dependence suggesting no major conformational changes were coinciding with the temperature change and thereby allowing a comparison of slopes (∆δ/∆t). The lowest values obtained for these slopes corresponded to Phe8 and Arg9 suggesting solvent sequestering for these amides. Given the high Chou and Fasman probability of β-turns in the sequences Pro2-Pro3-Gly4-Phe5 and Ser6Pro7-Phe8-Arg9 (3.79 × 10-4 and 1.99 × 10-4, respectively), the computational strategy employed was to begin from two initial structures: (a) an extended β strand, and (b) a structure containing these two predicted β turns. Utilizing custom routines written using the program CHARMm, version 21 [21], the interproton distances were incorporated into the potential-energy expression in the form of an additional potential-energy term. During the 3-ps heating step of the molecular dynamics, the temperature was raised from 0K to 300K in steps of 20K every 0.2 ps. Since the target distances

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_122.html [4/5/2004 4:56:31 PM]

Document

Page 123

were poorly satisfied in the starting structures, the potential-energy term corresponding to the imposed NOE data (ENOE) was applied gradually by increasing the scale factor in a nonlinear fashion such that it was 0.0 after 0.2 ps, 0.1 after 1.2 ps, and 1.0 after the full 3.0 ps. Following 15 ps of equilibration, 7 ps of incremental production dynamics was completed. During this stage the NOE scale factor was raised from 1.0 to 4.5. By slowly raising the force constants for the NOE restraints as the target distances became better satisfied, no dramatic increase in temperature was observed. Finally the NOE scale factor was set to 5.0, 10 ps of production dynamics were completed, and an average structure was extracted from the last 5 ps of the coordinate trajectory. Analysis of the two average structures obtained from the two unique starting points demonstrated convergence to a similar conformational species. In each, the sum of the NOE restraint energy was less than 4.7 kcal/mol and the RMS deviation from the target distances was below 0.25 Å. Similar results were obtained for each simulation when they were repeated without the electrostatic term being included in the total potential-energy function. This important data lends credence to the hypothesis that the final structures are derived from the NOE restraints and not by poorly represented electrostatic interactions. The average dynamic structures are characterized as having all trans peptide bonds and hydrophobic side chain groups oriented outward into solution, perhaps ready to interact with the receptor. There is a possible 1–3 hydrogen-bonded γ turn bridging Phe5, although it is not explicitly defined by the NOE data set. If present, then the preferred overall geometry would be U shaped and, if absent, an S shaped geometry is possible based on coincident conformational analyses. According to the dihedral angle values for Pro7 and Phe8, a type-II β turn extending from Ser6 to Arg9 also exists. A variety of reports have subsequently appeared that are in agreement with the conformation we described in this work. A similar C-terminal turn structure was observed in an analogous NMR study of a first-generation kinin antagonist, NPC 567 (DArg0-Arg1-Pro2-Hyp3-Gly4-Phe5-Ser6-DPhe7-Phe8-Arg9), although the type of turn was not the same. Our initial speculation was that this slight structural difference might partially account for the functional differences of bradykinin and NPC 567. These solution conformations, one of an agonist and the other of an antagonist, were subsequently used to focus the design and synthesis of conformationally constrained peptide analogues of NPC 567. B. Conformationally Constrained Bradykinin Antagonist Peptides The ligand-based approach of conformationally constrained peptides has been widely used. The process involves the incorporation of conformational constraints into known peptides, either agonist or antagonist, which enforce a

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_123.html [4/5/2004 4:57:26 PM]

Document

Page 124

predictable geometry. A series of peptides containing these types of constraints can be useful for extrapolating the steric and electronic environment of a given binding site. This structural information can be derived regardless of whether or not the constrained peptide binds to the target receptor. Since peptides can be prepared rapidly, it is typical to establish a structure-activity relationship using them and then at some later time transpose that information onto a nonpeptide lead molecule in an attempt to improve its potency. As part of an expansion upon the hypothesis that a C-terminal β turn was a structural prerequisite to highaffinity antagonist binding, a novel series of constrained decapeptides was prepared [22–24]. These peptides are of the sequence DArg0-Arg1-Pro2-Hyp3-Gly4-Phe5-Ser6-DHype7-Y8-Arg9, where Y is either tetrahydroisoquinoline-3-carboxylic acid (Tic), or octahydroindole-2-carboxylic acid (Oic). DHype denotes an organic ether of D-4-hydroxyproline in either the cis or trans geometric form. The C-terminal portion of a representative member of this class of peptides was shown—first by empirical calculation [22], then by NMR at 600 MHz—to adopt a β turn nearly unambiguously (figure 2) [25,26]. Moreover, it was shown by calculation that the turn was adopted regardless of the nature of the ether group (alkyl, aryl, etc.) or its geometry (cis or trans). Hence, a diverse series of these peptides was initially used as a tool to probe the steric and electrostatic topology of an antagonist

Figure 2 Lowest 5 kcal of the calculated overall potential energy surface for a model peptide of Ser-DHype(trans propyl)-Oic-Arg. The contour interval is 0.5 Kcalmol-1 and the highest (outermost) and lowest contour energy values are labeled. Superimposed on the contour plots are values for ψi+1 and ψi+2 from each of the thirty structures generated from the NMR data corresponding to the tetrapeptide Ser-DHype(trans propyl)-Oic-Arg. mol-1

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_124.html (1 of 2) [4/5/2004 4:57:43 PM]

Document

Page 125

binding site on the bradykinin B2 receptor in the guinea pig ileum. The cis ethers, in all cases, bound to the receptor with significantly lower affinity than did the trans. A more complete listing of the peptides used in the study is shown in Table 1. These results support the hypothesis that the domain of the receptor that binds these antagonist ligands is partly made up of a hydrophobic cavity about one side of the C-terminal turn. However, adjacent to the other side of the

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_125.html (1 of 2) [4/5/2004 4:58:44 PM]

Document

Page 126

Figure 3 Receptor binding curves for the binding of NPC 17410 and NPC 17643 to B2 receptors from the guinea pig ileum and cloned rat and human B2 receptors. Legends are noted on the figure.

turn, there appears to be some type of steric interference (or lack of a pocket) that might otherwise accommodate the ethers of the cis configuration. More recently, bradykinin B2 receptors have been cloned from both rat and human sources [27,28]. In receptor-binding experiments using these new receptors, selected members of the DHype-containing decapeptides were used to probe these receptors [24], a representative sample of the data is shown in Figure 3. Specifically, NPC 17643 (a trans propyl ether of D-4-hydroxyproline at position 7) and NPC 17410 (a cis propyl ether of D-4-hydroxyproline at position 7) were used. Although the trans ethercontaining decapeptide behaved similarly in binding assays directed toward the bradykinin B2 receptors in guinea pig, rat, and human, the cis ether-containing decapeptide, NPC 17410, displayed an interesting pharmacology. In particular, NPC 17410 bound with similar affinity to both the guinea pig and rat bradykinin B2 receptors, but had an appreciably higher affinity for the human B2 receptor. This result strongly suggests that there are slight structural differences in the antagonist binding sites of the rat and human B2 receptors. With clones available for the rat and human bradykinin B2 receptors, this prompted a systematic search using NPC 17410 binding to rat/human bradykinin B2 receptor chimeras and point mutations in an attempt to discover residues on the receptor that comprise this antagonist site. The details and results of the subsequent application of these novel receptor-probing ligands is fully described later in this chapter.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_126.html [4/5/2004 4:58:51 PM]

Document

Page 127

The des-Arg9 forms of these peptides have also been shown to have high affinity for the recently cloned human B1 receptor [24]. An extension of the work described herein would be to use a more complete series of des-Arg9 DHype-containing nonapeptides to probe the binding site of this new receptor where other interesting pharmacological differences are likely to exist since the B1 receptor is only 33% homologous to the human B2 [29]. In summary, the DHype-containing decapeptides have been useful in many regards. First, they incorporate a novel β-turn mimetic that was alternatively functionalized and used to probe the unknown topology of the guinea pig, rat, and human bradykinin B2 receptors. In this role, one of these tools showed differential pharmacology between rat and human forms of the receptor. This tool was used in a synergistic fashion with subsequent molecular biological and computational procedures in the elucidation of an antagonist binding site. Second, these peptides, together with another potent decapeptide antagonist with similar conformational constraints [30,31], provide the first strong experimental evidence that high-affinity decapeptide bradykinin receptor antagonists adopt a C-terminal β turn in the receptor-bound conformation. Third, certain members of this series of decapeptides contain alkyl ethers of D-4-hydroxyproline at position seven. In this regard, they are the very first examples of decapeptide bradykinin receptor antagonists that do not contain a D-aromatic amino acid at the seventh position as had been previously deemed to be essential. Commercially, this renders the series patentably distinct from all other known bradykinin receptor antagonists. Finally, several members of the series (i.e., NPC 17731, NPC 17761, NPC 17974) are among the most potent antagonists for this receptor yet reported. Hence, there may be applications for these compounds as human therapeutics. Several “second generation” decapeptide antagonists have been reported, but the prototype from the class, which was first to be reported, is HOE 140 (DArg0-Arg1-Pro2-Hyp3-Gly4-Thi5-Ser6-DTic7-Oic8Arg9) [30,31]. This decapeptide has also been shown to preferentially adopt a C-terminal β turn, consistent with the previous discussion [26,32,33]. The side chain of DTic at position seven is, however, flexible. While side-chain rotational movement is not allowed, the saturated six-membered ring easily undergoes an endo/exo ring-flipping motion. Hence, the β turn predominates about the backbone dihedral angles, but the side chain of DTic could be either endo or exo ring flipped in the receptor-bound conformation. In the absence of an appropriate x-ray crystallographic structure, there is no definitive means of establishing which possibility is correct. This type of ring-flipping conformational change serves to orient the bulky hydrophobic side chain of the DTic residue to either one side of the β turn, or the other. The data collected from the decapeptides containing either cis or trans ethers of D-4hydroxyproline at the analogous position in the

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_127.html [4/5/2004 4:58:53 PM]

Document

Page 128

sequence (discussed above) support a hypothesis that the DTic is exo ring flipped in the receptor-bound state. There are two factors that must be considered when applying structure-activity-relationship (SAR) information from a series of peptides toward the design of nonpeptide mimetics and putative library scaffolds. One is in regard to the backbone conformation that primarily serves as a structural scaffold upon which the various functionalities (side chains) are attached. The other factor is the side chains themselves whose spatial positions are primarily dictated by the backbone structure. Usually, the threedimensional arrangement of these differing chemical groups are responsible for affinity and triggering of the receptor. Knowledge of the relative importance of the individual side chains and amide bonds for receptor affinity is therefore a critical aspect of small molecule design from a peptidic structure-activity relationship. Conformationally constrained derivatives of HOE 140 have been prepared in continuing efforts to elucidate the ideal backbone conformation peptide antagonists must adopt for bradykinin B2 receptor interaction. One such series made use of Cα- or N-methyl substituted amino acids, incorporated at position(s) Gly4, Phe5, or both, in the peptide D-Arg0-Arg1-Pro2-Hyp3-Gly4-Phe5-Ser6-D-Tic7-Oic8-Arg9 (NPC 18545) [34]. An N-methyl substitution in the backbone of an L-amino acid is known to disfavor helical, or twisted, backbone conformations while favoring an extended backbone. The contrasting Cαmethyl modification tends to favor a helical (twisted), rather than extended, conformation [35,36]. These conformational preferences apply only to the backbone φ, ψ angles (where φi and ψi correspond to backbone dihedral angles for residue i, defined by the four adjacent amino acid backbone atoms Ci-1-NiCαi-Ci band Ni-Cαi-Ci-Ni+1, respectively) of the amino-acid residues bearing the modification. Receptor binding assays were performed in membrane preparations of the guinea pig ileum, a source of B2 receptors, wherein these constrained peptides were evaluated for their abilities to compete with bradykinin binding. With the exception of the Cα-methyl-Phe5-containing peptide (NPC 18540), each conformational constraint caused a significant, at least 1000-fold, loss in binding affinity with respect to the unconstrained parent peptide, NPC 18545. There are several factors that could contribute to the poor receptor affinities measured for these peptides. In addition to the possible induction of an adverse conformation via the N-methyl substitution, this modification also eliminates an amide proton that might be an important hydrogen-bond donor during ligand-receptor interaction. Furthermore, the N-methyl substitution enhances the likelihood of trans-cis amide bond isomerization, which could also disrupt an optimal ligand-receptor interaction by altering the spatial display of the local side chains. The Cα-methylPhe5 substitution of NPC 18540 is well

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_128.html [4/5/2004 4:58:58 PM]

Document

Page 129

tolerated by the receptor as evidenced by only a seven-fold loss in receptor affinity (Ki=0.54 nM) with respect to the parent of the series, NPC 18545. This implies that the φ, ψ backbone dihedral angles about Phe5 are on the order of -60°, -60° in the biologically active conformation. This combination of dihedral angles represents a helical twist or “kink” in the midsection of the peptide. Since the original submission of the manuscript describing these constrained linear and cyclic peptides, bradykinin B2 receptors have been cloned from other species including human [27,28]. Toward the goal of designing a nonpeptide antagonist as a human therapeutic agent, it would be interesting to evaluate these analogues against these newly reported receptor homologues. This would likely be fruitful and valuable given that there is evidence, including that presented in this chapter, showing that guinea pig and rat B2 receptors differ structurally from the human B2 receptor at the antagonist binding site. Hence, a structure-activity relationship established against the guinea pig or rat receptors could be misleading in the context of potential human therapeutics. A systematic study of the relative importances of amides and side chains in a prototypical second generation antagonist, NPC 18545 (DArg0-Arg1-Pro2-Hyp3-Gly4-Phe5-Ser6-DTic7-Oic8-Arg9) has recently been described [37,38]. The D-Arg0 and Ser6-DTic7-Oic8-Arg9 segments were left intact in all peptides on the assumption that N-terminal positive charge(s) and a hydrophobic C-terminal β turn are minimally required for binding. In a systematic fashion, the amino acids in the core of the peptide (Arg1Pro2-Hyp3-Gly4-Phe5) were substituted with glycine, an amino acid bearing no chirality or side chain. Binding assays, either in membranes from the guinea pig ileum or in membranes from a stable cell line expressing the human B2 receptor, were performed on each peptide and the results compared with the parent, NPC 18545, which has a Ki against [3H]-bradykinin of 0.08 nM. The elimination of all chirality and sidechain moieties in the segment Arg1-Pro2-Hyp3-Gly4-Phe5 via replacement by Gly1-Gly2-Gly3Gly4-Gly5 (NPC 18152), led to a peptide that no longer binds the receptor. This demonstrated that one or more of the side chains in this segment are critical during ligand-receptor interaction. Incorporation of either the Arg1 or Phe5 side chains led to improved potency (285 nM and 483 nM, respectively). Constructing a peptide with side chains of both Arg1 and Phe5 in place (NPC 18149) yielded a peptide with good affinity, Ki of 13.7 nM. Overall, this study shows that to maintain potency in the low picomolar range, peptides in this series require Arg1, Phe5, and either Pro2 or Hyp3 (but not both). The Nterminal charged moieties and the hydrophobic C-terminal β turn are also required. Potencies in the low nanomolar range are attainable without including the side chains or chirality associated with Pro2 and Pro3. These data have

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_129.html [4/5/2004 4:59:04 PM]

Document

Page 130

subsequently been successfully applied toward the design and synthesis of several nopeptide scaffolds and mimetics. Ultimately these mimetics were assembled in a combinatorial fashion as discussed in Section IV. In a related study, wherein NPC 18149 (DArg0-Arg1-Gly2-Gly3-Gly4-Phe5-Ser6-DTic7-Oic8-Arg9; Ki = 13.7 nM; Guinea pig ileum) was taken as the lead peptide, the relative contributions to binding affinity from each amide bond in the segment Arg1-Gly2-Gly3-Gly4-Phe5 were examined. Aminovaleric acid was used in a systematic fashion as a surrogate for any pair of adjacent Gly-Gly residues in the peptide. Aminovaleric acid is atomically identical to Gly-Gly with the exception that the amide bond linking the two glycines is replaced by two methylenes. The synthesis of Gly4-Phe5 required a special Gly-Phe mimic that has since been reported [39]. Since this substitution introduces flexibility into the peptide, it is a means of probing the structural role a given amide bond plays during receptor interaction. Potential hydrogen-bond donor and acceptor groups in the amide bond are removed via this substitution, which yields additional insights into potential electrostatic interactions that may also be important during ligand-receptor interactions. The conclusions drawn from the data are that in terms of structural or electrostatic interactions with this antagonist site on the receptor, the amide bond linking residues two and three may not be as critical as those linking residues three to four and four to five. Each of these investigations was aimed toward an understanding of either the backbone conformation of this prototypical decapeptide or the relative importance of the functional groups in the side chains that make significant contributions to receptor affinity. From the former, nonpeptide frameworks and scaffolds can be imagined. From the latter, insights into which functionality is required for high-affinity binding is derived. The remaining challenge is to reassemble these fragments onto synthetically feasible nonpeptide frameworks as potential new lead compounds. Our approach toward addressing this challenging problem is described later in this chapter. III. Receptor Structure-Based Investigations A. Elucidation of an Agonist Binding Site on the B2 Receptor In addition to the deductions one might make about a receptor binding site on the basis of receptor binding data from conformationally constrained ligands as previously described, models of bradykinin and bradykinin antagonists bound to their respective sites on the receptor as complimentary aspects of the overall strategy are also valuable. Unfortunately, due to the nature of the bradykinin

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_130.html [4/5/2004 4:59:07 PM]

Document

Page 131

receptor, it has not yet been obtained in crystalline form, nor is it likely to be in the near future. The bradykinin receptor is a member of a family of receptors for which an intracellular interaction with a G-protein is a critical part of the signal transduction pathway following agonist binding. Structurally, these G-protein-coupled receptors extend from beyond the extracellular boundary of the cell membrane into the cytoplasm. The tertiary structure is such that the protein crosses the bilayer of the cell membrane seven times, thus forming three intracellular loops, three extracellular loops, and giving rise to cytoplasmic C-terminal and extra-cellular N-terminal strands. It is generally presumed that the transmembrane domains of these receptors exist as a bundle of helical strands. This assumption is derived primarily from the known structure of the trans-membrane portions of a structurally related protein, bacteriorhodopsin [40]. G-Protein-coupled receptors do not lend themselves to analysis by either NMR or x-ray crystallography due to their structural dependence on an intact cell membrane. In our laboratories we pursued this valuable structural information by utilizing a combination of structural homology modeling, molecular dynamics, systematic conformational searching methods, and mutagenesis experiments. The combination of these techniques led to a proposed model of bradykinin bound to the agonist site on its receptor [41]. A hydrophobicity (Kyte-Doolittle) calculation [42] on the amino acid sequence of the rat bradykinin receptor yielded seven segments, each of which were 21 to 25 contiguous residues with predominantly hydrophobic side chains. These were presumed to be the seven transmembrane portions of the receptor. Cartesian coordinates of the backbone atoms within each of these seven segments were built by structural homology from the cryomicroscopic structure of the analogous segments of bacteriorhodopsin. Subsequently, side chains were added to these seven segments as appropriate for the rat bradykinin receptor, and the resulting geometry was optimized via constrained energy minimization to alleviate bad contacts. Extracellular and intracellular loops were extracted from the Protein Data Bank library, following a geometric search based upon a vector defined by terminal alpha carbons in adjacent helices. The model was subsequently subjected to a series of constrained and unconstrained energy minimizations as well as molecular dynamics simulations. The resulting structure of the receptor was used in a novel two-step docking procedure. Following our hypothesis that bradykinin adopts a C-terminal β turn upon complexation with the receptor, the φ, ψ backbone dihedral angles in the tetrapeptide corresponding to the C-terminus of bradykinin (Ser-Pro-Phe-Agr) were constrained in a harmonic fashion (force constant = 15 Kcal Å-1 mol1) to values that define a type II' β-turn [43]. This tetrapeptide probe was then systematically translated about the interior of a theoretical box inscribing the rat

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_131.html [4/5/2004 4:59:09 PM]

Document

Page 132

receptor model. The translations were such that the tetrapeptide probe molecule was incrementally repositioned within the receptor by following a 3 Å × 3 Å × 3 Å grid pattern. At each new position, both the probe and receptor were reset to their initial conformations, then the geometry of the complex was optimized using 200 steps of steepest descent followed by 500 steps of Adopted-Basis Newton-Raphson energy minimization. Subsequently, the sum of the steric and electrostatic contributions to the overall potential energy (interaction energy)—as measured only between the tetrapeptide probe molecule and the atoms of the receptor—were calculated. Slices through the receptor illustrating the energy of interaction as grayscale contour lines (darker gray = lower interaction energy) for that portion of receptor that was sampled by the tetrapeptide probe molecule is shown as an edge-on, frontal view in Figure 4. In this Figure it is qualitatively clear where the transmembrane domains are located (white), as well as where the most favorable sites of probe interaction are located (black).

Figure 4 Complete group of contour plots showing energy of interaction between probe and receptor. Each contour plot corresponds to a different horizontal slice as part of the first stage in the conformational search. Darker gray indicates most favorable interaction and the light shades represent least favorable interactions.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_132.html [4/5/2004 4:59:26 PM]

Document

Page 133

This initial stage of the docking process was used to reduce the computational difficulties that would be inherent in “tumbling” a complete bradykinin molecule (which has great flexibility) about the receptor in a similar fashion. However, following this initial stage, insight into those regions of the receptor capable of accommodating the C-terminal portion of the bradykinin molecule was obtained. On the basis of energetics, and as qualitatively shown in Figure 4, those particular regions are clustered in the central part of the receptor near to the extracellular domain. Using this information as a steering device to limit the size of the problem, an exhaustive conformational search was performed using the entire nineresidue sequence of bradykinin as a probe molecule, again enforcing a C-terminal β turn via dihedral angle constraints. Specifically, 24 unique geometric orientations (eight on each of three axes) of the bradykinin molecule were sampled at each of 100 grid points identified during the initial stage as likely zones to bind the tetrapeptide probe molecule. Bradykinin-receptor complexes within the lowest 150 kcal mol-1 interaction energy with respect to the lowest found (17 complexes out of 2400) were grouped into sets of related conformational families, of which there were five. Computationally, each of the five complexes were presumed to be equally likely. All of these simulations were accomplished using custom routines written using the program CHARMm [21]. To guide the selection of which of the five bradykinin-receptor complexes to consider a “lead” model, supporting experimental evidence was sought from site-directed mutagenesis experiments. This support was taken primarily from work describing bradykinin binding assays performed of mutant rat B2 receptors [44,45]. The underlying strategy of the mutation studies was based on the hypothesis that, since bradykinin has positive charges at either end of its sequence (Arg1 and Arg9), separated by a group of rather hydrophobic amino acids (Pro2-Pro3-Gly4-Phe5-Ser6-Pro7-Phe8), it was likely that some acidic residues in the receptor participated during ligand binding. Several mutant receptors were made such that each contained either a point mutation or a small cluster of point mutations, wherein native residues, having negatively charged side chains (Asp, Glu), were replaced by alanine(s). Table 2 lists the initial cluster mutations (rat) that were prepared as well as the follow-up single point mutations (rat). Figure 5 shows a stereoview of the selected ligand-receptor complex chosen on the basis of best agreement with the results of these mutagenesis studies. None of the other four putative complexes were in agreement with this experimental data and were not considered further. Of particular significance in this work was that the trans-membrane residue Glu49, when mutated to alanine, showed no adverse effect on bradykinin receptor affinity with respect to rat wild type. A similar result was reported for the Glu196 rarrow.gif Ala196 mutation. These residues are remotely situated with respect to the proposed site of bradykinin

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_133.html [4/5/2004 4:59:30 PM]

Document

Page 134

binding and are colored light gray in Figure 5. In contrast, the [Asp175, Glu178,179] rarrow.gif Ala175,178,179 cluster mutation showed a 12-fold loss in bradykinin binding affinity, and the [Glu282, Asp286] rarrow.gif Ala282,286 cluster mutation lost 17-fold with respect to the wild type receptor. The Asp268 rarrow.gif Ala268 and Asp286 rarrow.gif Ala286 point mutations caused 19-and 28-fold respective losses in affinity for bradykinin. Close inspection of the bradykinin Arg1 side chain location and surrounding receptor interactions led to the suspicion that Asp286 and Asp268

Figure 5 Proposed model of bradykinin bound to the rat B2 receptor at the agonist binding site. Only the upper portion of the receptor is shown as gray helical ribbons. Bradykinin backbone and side chain atoms are shown as thick white licorice. Positions of point mutations having no significant adverse effects on bradykinin binding are shown as light gray spheres. Positions of mutations affecting bradykinin binding are shown as dark gray spheres.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_134.html (1 of 2) [4/5/2004 4:59:38 PM]

Document

Page 135

might be jointly interacting either with the guanidino group in the side chain of Arg1 or the N-terminal amino group in bradykinin. Therefore a receptor containing a double mutation (Asp268,286 rarrow.gif Ala268,286) would be expected to show a much more dramatic loss in affinity for bradykinin than would receptors containing the individual point mutations. The appropriate double mutation experiment confirmed this by causing a 500-fold loss in affinity for bradykinin, as predicted (Table 2). The mutagenized residues of this double mutant B2 receptor are colored dark gray in Figure 5. This type of an ionic interaction is also precedented by the body of literature that exists supporting the requirement of an N-terminal arginine residue and a free N-terminal amino group in both bradykinin peptide agonists and antagonists for high affinity binding. All of these mutant receptors were demonstrated to be functional receptors on the basis of bradykinin-induced membrane depolarization in a Xenopus oocyte expression system [44,45]. The selected agonist site model is characterized by an overall twisted S-shape ligand, similar to the conformation of bradykinin determined previously in a hydrophobic environment by NMR [19,46]. Overall, the model suggested that the N-terminal amino and guanidine groups of Arg1 interact directly with negatively charged amino acids in extracellular loop three, and the C-terminal end is in a β-turn conformation buried just below the extracellular boundary of the trans-membrane domain of the receptor. Noteworthy is the presence of a hydrophobic cavity in our receptor model located adjacent to Pro7 of the bradykinin ligand. This cavity is made up, in part, by the residues Phe261, Leu104, Val108, and Ile112. Given the historical significance of position seven in peptide bradykinin-like ligands, these residues represent interesting targets for further mutagenesis experiments. One such result, the mutation of Phe261 to Ala261, has already been described, and the results were supportive of this proposed model [47]. Antibodies to the extracellular loops two and three have also been shown to compete with bradykinin binding, lending further experimental support for an extracellular domain on this agonist binding site. More recently, chemical crosslinking combined with site-directed mutagenesis was used to analyze the bradykinin binding site in the human B2 bradykinin receptor [48]. Previous studies using the bovine B2 receptor showed that heterobifiunctional reagents reactive to amines and free sulfhydryls crosslink the bound bradykinin N-terminus to a sulfhydryl(s) in the receptor [49]. To identify this sulfhydryl(s), two conserved candidate residues in the human B2 receptor—Cys20 in the N-terminal domain and Cys277 in extracellular loop 3—were mutated to serine residues. Single and double mutants were expressed in Cos 7 cells. All mutants bound [3H]bradykinin with typical B2 receptor specificity. The heterobifunctional reagent m-maleimidobenzoyl-N-hydroxysuccinimide ester crosslinked bradykinin to wild-type and mutants with maximum efficiencies of 35% (wild type), 40% (Ser20), 20% (Ser277), and 0%

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_135.html [4/5/2004 4:59:40 PM]

Document

Page 136

(Ser20, Ser277). This clearly demonstrated that Cys20 and Cys277 are the only sulfhydryls available for crosslinking receptor-bound bradykinin. These results provided direct biochemical evidence that the Nterminus of bradykinin, when bound to the B2 receptor, is adjacent to extracellular loop 3 and the Nterminal domain in the receptor. Further consideration of the model led to the hypothesis that agonist peptides may minimally require an intact C-terminal β-turn structure with appropriate side chains in place and N-terminal amino and guanidine groups for primary electrostatic interaction(s) with Asp286 and Asp268 in extracellular loop 3. As a test of this hypothesis, the prototypical second generation antagonist, NPC 18545, (DArg0-Arg1Pro2-Hyp3-Gly4-Phe5-Ser6-DTic7-Oic8-Arg9) was modified such that residues 2–5 were replaced by a simple twelve carbon chain spacer (12-aminotridecanoic acid). The resulting compound, NPC 18325, contains only the appropriately charged moieties at the N-terminus, separated by a simple organic spacer moiety from a known β turn forming tetrapeptide [25,26]. This pseudopeptide was tested in the human bradykinin B2 receptor binding assay and found to have a Ki of 44 nM against [3H]bradykinin binding [50]. Functionally, this pseudopeptide was an agonist as measured by its ability to stimulate IP production in a stable CHO cell line expressing the human B2 receptor and in WI-38 cells. Since it was designed on the basis of an agonist site on the receptor, this result was not completely surprising despite the incorporation of the DTic-Oic pair at the C-terminus that previously had been shown to be critical in high affinity antagonists. Subsequently the length of the linear carbon chain was varied to further explore the hypothesis that the agonist site on the receptor had two domains, a hydrophilic site in the extracellular loop area and a hydrophobic domain in the transmembrane area [50]. Presumably, the two terminal portions of NPC 18325 can only simultaneously interact with each putative domain of the receptor binding site when the carbon chain is 12–13 methylenes long. But if the carbon chain length is shortened too far, this ligand might be unable to simultaneously interact with both domains, resulting in an affinity loss. This series of pseudopeptides and their respective human bradykinin B2 receptor affinities are presented in Table 3. The data are consistent with the hypotyhesis since the receptor affinity decreases as the carbon chain length is shortened. An alternative explanation of the data is that a certain hydrophobicity profile is required of the compounds in this series for good receptor affinity. These results indicated that there may be additional hydrophobic or flexibility prerequisites to binding in this series of pseudopeptides. One noteworthy observation was that NPC 18325 showed divergent behavior when evaluated against different species homologues of the bradykinin B2 receptor. Notably, in the guinea pig ileal membrane preparation assay, the affinity for the receptor was approximately 10-fold less than what had been

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_136.html [4/5/2004 4:59:41 PM]

Document

Page 137

observed for the human B2 [37]. Furthermore, in contrast to the functional activity of NPC 18325 at the human B2 receptor, the compound is a functional antagonist as measured against bradykinin-induced contraction of the isolated guinea pig ileum (pA2 = 5.5). These findings are in agreement with the concept that as a ligand is made smaller (i.e., fewer contact points possible with the receptor), the subtle structural differences in the binding sites on species variants of the same receptor become amplified. This observation further supports a cautionary posture toward developing nonpeptide antagonists for use in human diseases on the basis of results obtained in some animals including the guinea pig. Taking this new molecule as a lead structure, together with the receptor model and structure-activity relationship associated with related peptides including cyclic antagonists, the pursuit of several related pseudopeptides was undertaken. B. Elucidation of an Antagonist Site on the B2 Receptor There have been a variety of single alanine point mutations experimentally introduced into both rat and human bradykinin B2 receptors. Several of these have been shown to decrease the affinity of bradykinin to the receptor and have been implicated structurally near the agonist binding site. In contrast, at the time of this manuscript, there have been no mutations reported that adversely affect the ability of any peptide antagonists to bind to the receptor. Furthermore, antibodies raised against the certain extracellular domains of the kinin receptor compete with bradykinin for binding to the receptor but have no inhibitory

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_137.html [4/5/2004 4:59:52 PM]

Document

Page 138

action on the binding of antagonist peptides. In addition, it has been shown that bradykinin can be covalently crosslinked to the B2 receptor while antagonists cannot. These observations have fostered the belief that the agonist and antagonist binding sites of the receptor are not the same. At best, they may be partially overlapping, although there is no direct evidence for this. The ultimate identification of the amino acid residues that make up the antagonist site would be another important step toward the goal of structure-based design of novel nonpeptide antagonists. As described in a previous section of this chapter, characterizations of the bradykinin B2 receptors from rat and human using NPC 17410 (Figure 3) revealed different pharmacologies. Specifically, it showed a higher affinity for the human B2 receptor than it did for the rat B2 (human IC50 = 0.95 nM, rat IC50 = 48.0). This ligand “tool” provided a means for evaluating a series of bradykinin rat/human B2 receptor chimeras [51–53]. Several different chimeras were prepared in a systematic fashion and the affinity of NPC 17410 was determined for each. The chimeras are depicted schematically in Figure 6 together with the IC50 values determined for NPC 17410. Chimeras I through III sample the N-and C-terminal sections of the receptor for any contributions to an antagonist binding site. The remaining chimeras sample the core transmembrane domains of the receptor. Each chimera was shown to induce a membrane depolarization similar to wild type receptor in response to bradykinin when expressed in Xenopus oocytes. For each NPC 17410 assay, [3H]-NPC 17731 was used as the radioligand. From this systematic approach, specific groups of contiguous residues within the receptor were identified as possible contributors to an antagonist binding site. The NPC 17410 binding to chimeras III, IV, and VIII showed rat-like pharmacology (low NPC 17410 affinity). The NPC 17410 binding to chimeras I, II, VI, and VII showed human-like NPC 17410 pharmacology (high receptor affinity). Binding to chimeras V and VIII, however, was similar to rat-like NPC 17410 pharmacology, but the affinity of the compound was slightly shifted back toward human-like results. Comparisons of rat and human receptor sequences in the regions sampled by the chimeras reveals that only two clusters of residues differ between rat and human B2 receptors. Specifically, TM2 has the same sequence in rat and human receptors so it is unlikely that the differential pharmacology associated with NPC 17410 binding can be attributed to residues there. However, TM3 has a cluster of 3 residues that differ (rat rarrow.gif human: Thr110 rarrow.gif Ala108, Met111 rarrow.gif Ile109, Tyr113 rarrow.gif Ser111) and TM6 has a cluster of 5 residues that differ (rat rarrow.gif human: Phe259 rarrow.gif Leu257, Leu256 rarrow.gif Ile254, Val255 rarrow.gif Ile253, Gly252 rarrow.gif Leu250, Ala249 rarrow.gif Val247) in rat and human receptors. These differences represent important targets for follow-up point (and cluster) mutation experiments. Our current thinking is that the largest effects on NPC 17410 pharmacology, if any, might be derived from the TM3 mutants since

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_138.html [4/5/2004 4:59:54 PM]

Document

Page 139

Figure 6 Rat and human B2 receptor chimera constructs and affinity data for binding NPC 17410. Also shown is the affinity NPC 17410 to rat and human wild type receptors.

between rat and human, these are quite diverse. However, it is also possible that the cluster of residues identified in TM6, while not radically dissimilar, may as a group create different hydrophobic environments between these species homologues. The most significant individual difference within the TM6 zone is the Phe259 (rat) rarrow.gif Leu257 (human) swap and might therefore be most significant

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_139.html (1 of 2) [4/5/2004 5:00:00 PM]

Document

Page 140

Figure 7 Schematic of the primary amino sequence of the human B2 receptor. Shown in black are residues experimentally identified as contributing to an agonist binding site. The dark gray residues are suspect positions for contributing to an antagonist site. The residues colored light gray have been mutagenized only in the rat B2 receptor, but they are conserved in the human. The Thr263 rarrow.gif Ala mutation interferes with agonist binding only, while Gln260 partially interferes with agonist and first generation antagonist binding.

within the context of these TM6 residues. Currently, we have prepared these mutant receptors, but at this time binding to NPC 17410 remains unfinished. A summary of the amino acids in the human B2 receptor implicated in comprising either agonist or antagonist sites are highlighted in Figure 7. Marked in dark black are the residues of extracellular loop 3, TM 6, and the extracellular N-terminal segment that have been shown to participate in agonist binding. Marked in dark gray in TM 6 and TM 3 are residues likely to partially comprise an antagonist binding site based primarily on the chimeric receptor studies described previously, although there is no explicit experimental evidence as yet. Shown in light gray are two residues that are conserved between rat and human B2 receptors. Mutagenesis experiments have been done on this pair in the rat B2

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_140.html (1 of 2) [4/5/2004 5:00:06 PM]

Document

Page 141

receptor with interesting results [54]. Mutations in Thr263 only affect agonist binding, not antagonist. Mutations in Gln260 affect binding of bradykinin and first generation antagonist peptides. As depicted in the figure, it is possible that the agonist and antagonist binding sites have domains on opposite sides of the helix that makes up TM 6, with Gln260 being situated partly in both. IV. Design and Combinatorial Synthesis of Nonpeptidic Antagonists As was previously described, a significant body of information was generated that provides insights into the key structural features of bradykinin receptor binding sites and the residues that participate in ligand binding. In addition, from the ligand-based studies, knowledge about relevant structure-activity relationships was acquired. Our modular synthetic strategy was based primarily upon the recognition that high-affinity ligands appear to be comprised of three domains. These domains are (1) a positively charged N-terminal segment, (2) a midsection containing a bend or twist with some hydrophobic substituent attached and, (3) a C-terminal segment of appropriate hydrophobicity and structurally simulating a type II' β turn. Models of potent cyclic and linear peptide bradykinin receptor antagonists (described previously) were used in a comparative fashion to select nonpeptide ring systems from a database of chemical structures fine chemicals database. For each, some degree of chemical diversity was achieved by altering one of several parameters including, o, m, or p substitution of an aromatic ring or nature of alkyl substituent(s) as well as point(s) of synthetic attachment [55,56]. Each nonpeptide fragment was designed within the framework of several criterion. First, a given scaffold must closely match the known SAR and be compatible with the putative ligand binding site structure. Second, each scaffold must be a relatively simple synthetic target, having readily available starting material, no chiral centers and having a total synthesis of not more than 4–5 steps. Finally, each template must have a “C-terminal” carboxylate and an “N-terminal” amino group with no interfering functionality such that it could be readily used in a solid phase synthetic strategy. Given that each nonpeptide we identified was a viable surrogate for either the second or third domain of high-affinity ligands (as described above) our goal was to rapidly explore the receptor affinities of all possible combinations of these nonpeptide templates at position X and Y of the sequence DArg-Arg-X-Y-Arg, hence a combinatorial synthetic approach was taken. In this study, there were four linear aminoalkanoic acids [50], four different cinnamic acids, three different carbolines, three different phenanthridinones, and five different spirocyclics. The variability in the phenanthridinione series

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_141.html [4/5/2004 5:00:07 PM]

Document

Page 142

was that the central ring could be opened or cleaved and the amino group could be meta or para substituted on the latter. In the carboline series, the cyclic amino group was either at the β or γ position of the cyclohexenyl ring and the methylene chain bearing the “C-terminal” carboxylate could be of variable length. The spirocyclic series was varied by alkyl, cycloalkyl, and aryl substitution on the fivemembered ring amine nitrogen. The cinnamic acids had two carbon chains that could be of varying length, one of which had the further possibility of containing a double bond(s). Rather than perform individual syntheses of all possible combinations of these nonpeptide units, members of each ring type or scaffold family were pooled in equimolar amounts prior to incorporation into the sequence DArg-Arg-X-Y-Arg. Since each individual member of a given pool was constructed on a similar carbocyclic scaffold, the chemical environment of the N-terminal amino group and Cterminal carboxylate groups were expected to follow similar kinetic and thermodynamic controls during the attachment of the nonpeptide residue to the growing peptide chain. The use of these smaller, directed libraries made it readily practical to obtain HPLC and mass spectral data for each and therefore confirm the composition of the library. Ultimately, 10 libraries of novel nonpeptidic structures were synthesized following typical solid-phase methodologies. Each library contained from nine to thirty-six different compounds in approximately equimolar amounts. Unpurified libraries were tested in a receptor binding assay utilizing membrane preparations from a stable CHO cell line expressing the human B2 receptor. Each library was tested at concentrations between 10 nM and 1µM. The ability of each library to inhibit[3H]-bradykinin binding was assessed and the results are presented in Figure 8a. Although this type of screening is highly qualitative, certain libraries appear in Figure 8a that show higher affinity to the receptor than other libraries. Library one (of the series DArg-Arg-PH-CN-Arg) was ultimately selected for further deconvolution. This library was further broken down (decoded) in order to determine which compound(s) were responsible for the apparent activity. It is important to note that breaking these libraries down to elucidate the structure of the hit(s) was feasible due to the inherently small size of each library. Library one contained 12 different structures (recall that there were originally three different phenanthridinones and four different cinnamic acids). The first deconvolution step of the approach is shown in Figure 9. Here, only the CN position was randomized, and the PH moieties were specific. This led to the preparation of three new libraries of 4 compounds each. Receptor binding was again performed as before and only one of these three new sublibraries showed affinity for the receptor at 10 µM (Figure 8b). The final step in the process required to elucidate the active component(s) was to synthesize and purify each of the 4 members of this library as shown in Figure 9. Receptor binding on these

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_142.html [4/5/2004 5:00:09 PM]

Document

Page 143

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_143.html (1 of 2) [4/5/2004 5:00:18 PM]

Document

Figure 8 (a) Binding assay results for 10 original nonpeptidic libraries of the sequence DArg-Arg-X-Y-Arg, where X and Y are defined as PH = phenanthridinone, CB = carboline, SP = spirocycle, SC = straight chain, CN = cinnamic acid. Each library was tested at 1 µnM and 10 nM. Results were compared to cold bradykinin binding, which was tested at two lower concentrations, 0.1 nM and 1 nM. Panels (b) and (c) correspond to the receptor binding results obtained using the two breakdown steps from original library number 1, as shown in Figure 9.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_143.html (2 of 2) [4/5/2004 5:00:18 PM]

Document

Page 144

Figure 9 Composition of ten original nonpeptidic libraries of the sequence DArg-Arg-X-Y-Arg. X and Y were selected from the set of scaffolds shown in Table 1. Also shown are the subsequent breakdown libraries from original library number 1. Two-letter codes used in the figure correspond to the different nonpeptide moieties described in Table 1. Specifically, PH = phenanthridinone, CB = carboline, SP = spirocycle, SC = Straight chain, CN = cinnamic acid.

four novel nonpeptidic structures showed that only one of the four had affinity to the receptor (Figure 8c). This new compound, I, was subsequently shown to be an antagonist in a cellular assay measuring bradykinin-stimulated IP turnover [18]. Overall, there were 285 possible structures to survey due to the number of structure-based scaffolds that were prepared. This was rapidly accomplished via 19 synthetic couplings, 19 assays, and 4 purifications. Not surprisingly, compound I showed divergent potency when assayed in different species homologues of the bradykinin B2 receptor. In particular, in a model of bradykinin-induced hypotension in rats and rabbits, it showed no activity. Likewise, it did not block bradykinin-induced contraction of the isolated guinea pig ileum. Since compound I is considerably smaller than previously reported decapeptide antagonists, subtle structural differences (which are http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_144.html (1 of 2) [4/5/2004 5:00:24 PM]

Document

Page 145

known to exist in species homologs of the bradykinin B2 receptor) are likely amplified. A more comprehensive pharmacological analysis of compound I is currently underway. A. Lead Optimization We have previously reported that the C-terminal guanidinyl moiety of Arg [9] in prototypical peptide bradykinin antagonists is likely to behave more as an aromatic functional group rather than a hydrogenbond donor/acceptor. This speculation was based on proposed models of the agonist and antagonist binding sites of this receptor that have been elucidated using molecular biological and computational procedures. On this premise, the newly discovered lead compound, I, was altered such that the Cterminal arginine was replaced by 3',5'-dimethylpyrimidylornithine in an attempt to increase potency. This known mimetic of arginine contains an aromatic 3',5'-dimethylpyrimidyl ring in the side chain rather than the guanidino group on naturally occurring arginine. The results of the receptor binding assay performed using this compound, IA, are shown in Table 4 where it is clear that affinity to the human B2 receptor is improved with respect to compound I. This data is supportive of the notion that the C-terminal residue(s) in this new series of bradykinin antagonist compounds interact with a hydrophobic environment, perhaps within the transmembrane domain of the receptor as previously suggested. The discovery of I and IA is significant in many regards. First, they are highly nonpeptidic lead compounds that could be further modified to improve potency and/or reduce molecular weight. Such improvements might lead to novel therapeutic agents for the treatment of inflammatory diseases. Thus far in the kinin antagonist literature there is significant evidence showing that, for compounds containing a C-terminal arginine residue, removal of that arginine generally yields compounds that are antagonists of the B1 subtype of the bradykinin receptor. Following a similar strategy with compound I could lead to the discovery of a novel series of nonpeptidic B1 receptor antagonists, although this remains to be demonstrated. V. Conclusions There has been a significant effort invested toward the discovery of novel bradykinin receptor antagonists during the past decade. In that time, several generations of peptide antagonists have been developed and a few are in human clinical trials. The pursuit of nonpeptide antagonists of the human bradykinin B2 receptor continues and incorporates a wide range of strategic approaches. The approach described herein is an early and very good example of a combinatorial synthesis of nonpeptide building blocks that mimic peptide structure,

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_145.html [4/5/2004 5:00:27 PM]

Document

Page 146

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_146.html [4/5/2004 5:00:51 PM]

Document

Page 147

ultimately tested in a nontagged, solution-phase form. Perhaps more significant is that the success described here demonstrates a possible synergy between structure-based design and combinatorial methodology. This approach has many merits, but the most significant is the application of structurally directed libraries toward target binding-site structures which, for one reason or another, may not be fully characterized. This method serves to aim the combinatorial syntheses in a logical direction, rather than attempt to prepare libraries of vast diversify (and numbers). Finally, since the libraries of compounds that were prepared contained few members, it was possible to analytically characterize each of the pools to assess integrity of their composition. Overall, there are many important advantages in the paradigm we have adopted that make the strategy generally viable in the context of structure-based lead compound discovery. References 1. Greenbaum LM. Adv Exp Med Biol 1986; 198A:55. 2. Okamoto H, Greenbaum LM. Biochem Biophys Res Commun 1983; 112:701. 3. Muller-Esterl W. Thromb Haemost 1989; 61:2. 4. Bhoola KD, Figueroa CD, Worthy K. Pharmacol Rev 1992; 44:1. 5. Proud D, Perkins M, Pierce JV, Yates KN, Highet PF, Mangkornkanok/Mark M, Bahu R, Carone F, Pisano JJ. J Biol Chem 1981; 256:10634. 6. Kitamura N, Takagaki Y, Furoto S. Nature 1983; 305:545. 7. Ward PE. In: Burch RM, ed. Bradykinin Antagonists: Basic and Clinical Research. New York: Marcel Dekker, 1991:147–170. 8. Perkins MN, Campbell EA, Davis A, Dray A. Br J Pharmacol 1992; 107:237P. 9. Perkins MN, Campbell EA, Dray A. Pain 1993; 53:191. 10. Perkins MN, Kelly D. Br J Pharmacol 1993; 110:1441. 11. Kyle DJ, Burch RM. Curr Opin Invest Drugs 1993; 2:5. 12. Kyle DJ, Burch RM. Drugs of the Future 1992; 17(4):305. 13. Huang HM, Lin TA, Sun GY, Gibson GE. J Neurochem 1995; 64:761. 14. Rubin LE, Levi R. Circ Res 1995; 76:430. 15. Bao G, Gohlke P, Unger T. J Cardiovasc Pharmacol 1992; 20(Supplement 9):S96. http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_147.html (1 of 2) [4/5/2004 5:00:59 PM]

Document

16. Martorana PA, Kettenbach B, Breipol G, Linz W, Scholkens BA. Eur J Pharmacol 1990; 182:395. 17. McDonald KM, Mock J, D' Ajoia M, Parrish T, Hauer K, Francis G, Stillman A, Cohn JN. Circulation 1995; 91(7):2043. 18. Schwieler JH, Kahan T, Nussberger J, Hjemdahl P. Am J Physiol 1993; 264:E631. 19. Kyle DJ, Blake PR, Hicks RP. In: Burch RM, ed. Bradykinin Antagonists: Basic and Clinical Research. New York: Marcel Dekker, 1991:131–146. 20. Chou PY, Fasman GD. Biophys J 1979; 26:367. 21. (a) Brooks BR, Bruccoleri RE, Olafson BD, States DJ, Swaminathon S, Karplus MJ. CHARMM: a program for macromolecular energy minimization, and dynamics calculations. J Comp Chem 1983; 4:187–217.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_147.html (2 of 2) [4/5/2004 5:00:59 PM]

Document

Page 148

(b) Molecular Simulations, Inc., 16 New England Executive Park, Burlington, MA 01803-5297. 22. Kyle DJ, Martin JA, Burch RM, Carter JP, Lu S, Meeker S, Prosser JC, Sullivan JP, Togo J, Noronha-Blob L, Sinsko JA, Walters RF, Whaley LW, Hiner RN. J Med Chem 1991; 34:2649. 23. Kyle DJ, Burch RM. Drugs of the Future 1992; 17(4):305. 24. Hiner RA, Chakravarty S, Lu S, et al. J Med Chem 1996; manuscript in preparation. 25. Kyle DJ, Green LM, Blake PR, Smithwick D, Summers MF. Pep Res 1992; 5:206. 26. Kyle DJ, Blake PR, Smithwick D, Green LM, Martin JA, Sinsko JA, Summers MF. J Med Chem 1993; 36:1450–1460. 27. McEachern AE, Shelton ER, Shakta S, Obernolte R, Bach C, Zuppan P, Fujisaki J, Aldrich RW, Jarnagin K. Proc Natl Acad Sci USA 1991; 88:7724. 28. Hess JF, Borkowski JA, Young GS, Strader CD, Ransom RW. Biochem Biophys Res Commun 1992; 184(1):260. 29. Menke JG, Borowski JA, Bierilo KK, et al. J Biol Chem 1994; 269:21583–21586. 30. Hock FJ, Wirth K, Albus U, Linz W, Gerhards HJ, Wiemer G, Henke S, Breipohl G, Knoig W, Knolle J, Schölkens BA. Br J Pharmacol 1991; 102:769. 31. Wirth K, Hock FJ, Albus U, Linz W, Alpermann HG, Anagnostopoulos H, Henke S, Breipohl G, Knoig W, Knolle J, Schölkens BA. Br J Pharmacol 1991; 102:774. 32. Sawutz DG, Salvino JM, Seoane PR, Douty BD, Houck WT, Bobko MA, Doleman MS, Dolle RE, Wolfe HR. Biochem 1994; 33:2373. 33. HOE SDS MICELLES 34. Chakravarty S, Wilkens D, Kyle DJ. J Med Chem 1993; 36:2569. 35. Momany FA. In: Metzger, RM, ed. Topics in Current Physics. Vol. 26; New York: Springer Verlag, 1981:41–79. 36. Momany FA, Chuman H. Meth Enz 1986; 124:3–17. 37. Kyle DJ. Brazil J Med Bio Res 1994; 27:1757. 38. Chakravarty S, Mavunkel BJ, Lu S, Goehring R, Wu JP, Connolly M, Valentine H, Liu YW, Tam C, Andy R, Kyle, DJ. 23rd European Peptide Symposium, Braga: Sept 4–10, 1994.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_148.html (1 of 2) [4/5/2004 5:01:01 PM]

Document

39. Mavunkel BJ, Lu Z, Kyle DJ. Tett Lett 1993; 34:2255. 40. Henderson R, Baldwin JM, Ceska TA, Zemlin F, Beckmann E, Downing KH. J Mol Biol 1990; 213:899. 41. Kyle DJ, Chakravarty S, Sinsko JA, Stormann TM. J Med Chem 1994; 37:1347. 42. Kyte J, Doolittle RF. J Mol Biol 1982; 157:105. 43. Rose GD, Gierash LM, Smith JA. Adv Prot Chem 1985; 37:1. 44. Novotny E, Bednar D, Connolly M, Connor J, Stormann T. In: Burch RM, ed. Molecular Biology and Pharmacology of Bradykinin Receptors. Austin, TX: R. G. Landes Company, 1993:19–30. 45. Novotny EA, Bednar DL, Connolly MA, Connor JR, Stormann TM. BBRC 1994; 201:523. 46. Lee SC, Russell AF, Laidig WD. Int. J Pept Prot Res 1990; 35(5):367. 47. Freedman R, Jarnagin K. Cloning of a B2 Bradykinin Receptor: Recent Progress on Kinins. Basel: Birkhauser Verlag, 1992:487–496. 48. Herzig MCS, Leeb-Lundberg F, Nash N, Connolly M, Kyle DJ. Kinin '95. International conference on kallikreins and kinins, Denver, CO, Sept, 1995. 49. Herzig MCS, Leeb-Lundberg F. J Biol Chem 1995; in press. 50. Chakravarty S, Connolly MA, Kyle DJ. Peptide Research 1995; 8:16.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_148.html (2 of 2) [4/5/2004 5:01:01 PM]

Document

Page 149

51. Nash N, Connolly MA, Stormann TM, Kyle DJ. 14th Am Pep Sym, June 18, 1995. 52. Nash N, Connolly MA, Stormann TM, Kyle DJ. Mol Pharm 1996; manuscript in preparation. 53. Burch RM, Kyle DJ, Stormann TM. In: Molecular Biology and Pharmacology of Bradykinin Receptors. Austin, Texas: R. G. Landes Company, 1993:19–32. 54. Nardone J, Hogan PG. PNAS 1994; 91:4417. 55. Chakravarty S, Mavunkel B, Andy R, Kyle DJ. Network Science 1995; 1:1. 56. Chakravarty S, Mavunkel BJ, Andy R, Kyle DJ. 14th American Peptide Symposium, Columbus, Ohio, June, 1995.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_149.html [4/5/2004 5:01:02 PM]

http://legacy.netlibrary.com/reader/message.asp?message=811&BookID=12640&FileName=Page_150.html

The requested page could not be found. Return to previous page

http://legacy.netlibrary.com/reader/message.asp?message=811&BookID=12640&FileName=Page_150.html [4/5/2004 5:01:05 PM]

Document

Page 151

5 Design of Purine Nucleoside Phosphorylase Inhibitors Y. Sudhakara Babu, John A. Montgomery, and Charles E. Bugg BioCryst Pharmaceuticals, Inc., Birmingham, Alabama W. Michael Carson, Sthanam V. L. Narayana, and William J. Cook The University of Alabama at Birmingham, Birmingham, Alabama Steven E. Ealick Cornell University, Ithaca, New York Wayne C. Guida and Mark D. Erion* Ciba-Geigy Corporation, Summit, New Jersey John A. Secrist, III Southern Research Institute, Birmingham, Alabama I. Introduction A. Enzymology Purine nucleoside phosphorylase (PNP, E.C. 2.4.2.1) catalyzes the reversible phosphorylysis of ribonucleosides and 2'-deoxyribonucleosides of guanine, hypoxanthine, and related nucleoside analogs [1]. It normally acts in the phosphorolytic direction in intact cells, although the isolated enzyme catalyzes the nucleoside synthesis under equilibrium conditions. Figure 1 shows the chemical reaction. The enzyme has been isolated from both eukaryotic and prokaryotic organisms [2] and functions in the purine salvage pathway [1,3]. Purine nucleoside phosphorylase isolated from human erythrocytes is specific for the 6-oxypurines and many of their analogs [4] while PNPs from other organisms vary in their specificity [5]. The human enzyme is a trimer with identical subunits and a total molecular mass of about 97,000 daltons [6,7]. Each subunit contains 289 amino acid residues. *Current

affiliation: Gensia, Inc., San Diego, California.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_151.html [4/5/2004 5:01:07 PM]

Document

Page 152

Figure 1 The reaction catalyzed by PNP.

B. Pharmacology Interest in PNP as a drug target arises from its ability to rapidly metabolize purine nucleosides and from its role in the T-cell branch of the immune system. Unfortunately, PNP can also cleave certain anticancer and antiviral agents that are synthetic mimics of natural purine nucleosides, thus interfering with therapy. One such substance is ddI (2'3'-dideoxyinosine), which the Food and Drug Administration approved as a treatment for AIDS in 1991. Another is the potential anticancer agent 2'-deoxy-6thioguanosine [8]. Our goal was to develop a compound that when administered with the nucleoside analogs would inhibit PNP while the anticancer and antiviral agents accomplished their therapeutic missions. The combination of purine nucleoside analogs and a PNP inhibitor might prove to be a more effective treatment. The PNP inhibitors alone have potential therapeutic value based on the importance of PNP to the immune system. Patients lacking PNP activity exhibit severe T-cell immunodeficiency while maintaining normal or exaggerated B-cell function [9]. We, like other researchers, quickly recognized that PNP inhibitors might selectively suppress the T-cell proliferation associated with an array of autoimmune disorders such as rheumatoid arthritis, psoriasis, systemic lupus erythematosus, multiple sclerosis, and insulin-dependent (juvenile-onset) diabetes [10]. This profile also suggests that PNP inhibitors might be useful in the treatment of T-cell proliferative diseases—such as T-cell leukemia or Tcell lymphoma—and in the prevention of organ transplant rejection. C. Drug Design Strategy Recent advances in biotechnology, macromolecular crystallography, computer graphics, and related fields have led to a new approach in drug discovery called structure-based drug design. Structure-based drug design requires a detailed structural knowledge of the target (enzyme or receptor) and the interaction of small molecules with it.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_152.html [4/5/2004 5:01:10 PM]

Document

Page 153

Figure 2 Structure-based drug design strategy.

A tight fit is necessary for potency and specificity. A drug that binds to its target and inactivates it for a long time can be administered in lower doses than one that rapidly separates from its target. A substance designed to mesh perfectly with a particular binding site of one target is unlikely to interact well with any other molecule, minimizing unwanted interactions and side effects. Having chosen PNP as the target, we followed a systematic strategy for designing inhibitory compounds. Figure 2 outlines the overall strategy of this approach. To serve as a drug, an inhibitor has to readily cross cell membranes to the interior of cells, where PNP is located.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_153.html (1 of 2) [4/5/2004 5:01:14 PM]

Document

We determined the structure of human PNP by x-ray crystallography and used these results in combination with computer-assisted molecular modeling to design inhibitor candidates. We examined how well the shape and chemical

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_153.html (2 of 2) [4/5/2004 5:01:14 PM]

Document

Page 154

structure of a candidate would complement the active site of PNP. We used computational chemistry to estimate the strength of the attractive and repulsive forces between a candidate and the enzyme. We synthesized only those candidates suggested by chemical intuition and computer simulation to have high affinity for the target. Then we measured the inhibition of PNP and compared the proposed with the actual fit. Because modeling programs and expert opinion are imperfect, certain compounds did not meet expectations. After exploring the reasons for the successes and the failures, we returned to interactive computer graphics to propose modifications that might increase the effectiveness of drug candidates. The resulting compounds were evaluated by determination of their IC50 values (the inhibitor concentration causing 50% inhibition of PNP) and by x-ray diffraction analysis using difference Fourier maps. This iterative strategy—modeling, synthesis, and structural analysis—led us to a number of highly potent compounds that tested well in whole cells and in animals. D. Previously Known Inhibitors At the beginning of our studies several PNP inhibitors had been reported with Ki values in the 10-6 to 107 range, including 8-aminoguanine [11], 9-benzyl-8-aminoguanine [12], and 5'-iodo-9-deazainosine [13]. Acyclovir diphosphate had been shown to have a Ki near 10-8 if assayed at 1 mM phosphate rather than the more frequently used value of 50 mM phosphate [14]. During our studies, the synthesis of 8-amino9(2-thienylmethyl)guanine was reported with a Ki of 6.7 × 10-8 M [15]. Figure 3 illustrates some of these structures. Despite the potential benefits of PNP inhibitors and the large number of PNP inhibitors that had been synthesized, no compound had reached clinical trials. None of these compounds were potent enough to be useful for therapy and also capable of crossing the cell membrane intact. Although potencies for the best compounds had affinities 10–100 fold higher than the natural substrate (Km = 20 µM), it is expected that T-cell immunotoxicity will only occur with very tight binding inhibitors (Ki < 10 nM) due to the high level of in vivo PNP activity and competition with substrate. II. Crystallography At the present time, x-ray crystallography is the preferred technique for obtaining the required atomic resolution structural data. In the late 1970s when this project was first conceived, determining the structure of a protein was far from routine. The x-ray structural determination occupied a team of crystallographers led by Steven E. Ealick, then at the University of Alabama at Birmingham through most of the 1980s.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_154.html [4/5/2004 5:01:16 PM]

Document

Page 155

Figure 3 Previously known inhibitors.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_155.html (1 of 2) [4/5/2004 5:01:20 PM]

Document

Page 156

The stumbling block did not lie with obtaining pure PNP or converting the protein into crystals. Robert E. Parks, Jr. and Johanna D. Stoeckler of Brown University had already isolated the enzyme from human cells. They supplied quantities of protein to William J. Cook, who succeeded in preparing the well-ordered crystals required for x-ray studies [16]. We established that PNP crystals function normally as a catalyst. Thus crystalline PNP is essentially identical to PNP in the body. If it were profoundly different, one would have no justification for basing drug design on the crystal structure. In the early years we had to depend on a relatively low-intensity x-ray source. High-resolution data was obtained through collaboration with John R. Helliwell and his group at the Daresbury Laboratory Synchrotron Radiation Source in England. Today greatly improved equipment and more synchrotron facilities are available for protein crystallography. The three-dimensional structure was determined by multiple isomorphous replacement techniques using synchrotron radiation [17]. The native and guanine-PNP complex structures have been refined to 2.8 Å resolution [18,19]. A. Structure of the Enzyme Crystals of human PNP are grown from ammonium sulfate solution and stored in artificial mother liquor solution made of 60% ammonium sulfate in 0.05 M citrate buffer at pH 5.4. The space group is R32 with hexagonal cell parameters a=142.9(1) Å and c=165.2(1) Å. The PNP crystals contain about 76% solvent and diffract to around 2.8 Å resolution. The x-ray data established that PNP crystals contain a high percentage of water. This feature proved very useful; proposed drugs could easily be soaked into the active site without disrupting the crystal packing. Figure 4A shows the

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_156.html [4/5/2004 5:01:21 PM]

Document

Page 157

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_157.html (1 of 2) [4/5/2004 5:01:43 PM]

Document

Figure 4 Structure of PNP as stereo drawings. (A) Crystal packing. The low-resolution surfaces of six trimers are shown. The top level is related to the bottom level by a 2-fold axis along X. The distance between the 3-fold axes is 143 Å. A drug molecule is shown in the solvent channel near the entrance to an active site. (B) Ribbon drawing. The lowermost trimer of Figure 4A is shown. This is the native structure; the guanine and phosphate are shown to mark the active site. (C) The swinging gate. The trimer of Figure 4B is rotated about 30° counterclockwise in the plane, followed by a roughly 90° rotation about X to view the entrance to the active site. A model of the transition state is shown as a line drawing. Conformational changes of the protein on binding guanine are shown. Arrows are drawn from the Cα positions in the native structure to their positions in the complex. (D) Active site of PNP. The orientation is approximately that of Figure 4C, but enlarged and clipped to focus on the substrate. Key side-chain residues are labeled. Residue 159'F, in the center of figure toward the viewer, is the only residue from the adjacent subunit. The guanosine and phosphate are shown with thicker bonds. Oxygen and sulfur atoms are shown as white spheres, nitrogen and phosphorus as black spheres. (E) Purine binding pocket. The style is the same as Figure 4D, but the figure is rotated slightly and enlarged. The key group interacting with bound guanine are highlighted. (F) The best inhibitor. The style is the same as Figure 4D, but the figure has been enlarged and rotated to place the phosphate binding site far from the viewer in the upper left.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_157.html (2 of 2) [4/5/2004 5:01:43 PM]

Document

Page 158

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_158.html (1 of 2) [4/5/2004 5:02:02 PM]

Document

Figure 4 (Continued) Only one nitrogen atom from Arg 84 is visible, which—along with Ser 220 and His 86—interact with the acetyl group branching from the benzylic carbon. The chlorinated phenyl group is in the center of the figure, interacting with the aromatic groups in the sugar binding site. The guanine group interactions are the same as seen in Figure 4E. Figure prepared with ribbons (http://www.cmc.uab.edu/ribbons).

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_158.html (2 of 2) [4/5/2004 5:02:02 PM]

Document

Page 159

large solvent channels and the position of the active site. The x-ray analysis confirmed the trimeric nature of the enzyme, as the subunits are related by the crystallographic three-fold axis. A ribbon diagram of the trimer is shown in Figure 4B. Each monomer contains an eight-stranded β sheet and a five-stranded β sheet that join to form a distorted β barrel. Seven α helices surround this β sheet structure. The active site is an irregular indentation on the surface of the enzyme, located from the position of a tightly bound sulfate ion and various substrate analogs. These investigations revealed the identity of the exact amino acids constituting the active site region; such detail was a prerequisite to drug design. Information of greater import emerged from analyses of the complexes formed when synthetic nucleosides, including previously discovered inhibitors, were diffused into the active site. B. The Active Site The structural determinations also yielded a surprise. The shape of the enzyme changes when a purine is bound. The famous lock-and-key analogy [20] has a fallacy; the shape of the lock is not static, but flexible. Awareness of these conformational changes critically aided our modeling efforts, allowing prediction of which parts of PNP could change shape to interact with a proposed inhibitor. A “swinging gate” consisting of residues 241–260 controls access to the active site (Figure 4C). These residues in the native structure had poorly defined electron density with high thermal motion. The gate opens in the native enzyme to accommodate the substrate or inhibitor. The maximum movement caused by substrate or inhibitor binding occurs at His 257, which is displaced outwards by several angstroms. After binding, the electron density becomes well defined. The gate is anchored near the central β sheet at one end and near the C-terminal helix at the other end. The gate movement is complex and appears to involve a helical transformation near residues 257–261. Consequently, initial inhibitor modeling attempts using the native PNP structure were far less successful than subsequent analyses in which coordinates for the guanine-PNP complex were used. Because of the magnitude of the changes that occur during substrate binding, it is unlikely that modeling studies based on the native structure alone would have accurately predicted the structure of PNP/inhibitor complexes. The active site is located near the subunit-subunit boundary within the trimer and involves seven polypeptide segments from one subunit and a short loop from the adjacent subunit (Figure 4D). The purine binding site employs residues Glu 201, Lys 244, and Asn 243 to form hydrogen bonds with N1, O6, and N7 of purine. The remainder of the purine binding pocket is largely

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_159.html [4/5/2004 5:02:05 PM]

Document

Page 160

hydrophobic, composed of residues Ala 116, Phe 200, and Val 217. The phosphate binding site uses residues Ser 33, Arg 84, His 86, and Ser 220 with the phosphate positioned for nucleophilic attack at C1' of the nucleoside. The sugar binding site is mostly hydrophobic consisting of residues Tyr 88, Phe 200, His 257 from one subunit and Phe 159 of the adjacent subunit. This hydrophobic pocket orients the sugar to facilitate nucleophilic attack by phosphate and subsequent inversion of C1'. C. Initial Inhibitor Complexes In order to understand the interaction of inhibitors with the active site residues, the previously known inhibitors were obtained and crystallographic analyses were carried out. The most important findings were (1) 8-amino substituents enhance binding of guanines by forming hydrogen bonds with Thr 242 and possibly the carbonyl oxygen atom of Ala 116; (2) substitution by hydrophobic groups at the 9position of a purine enhances binding through interaction with the hydrophobic region of the ribose binding site; and (3) acyclovir diphosphate is a multisubstrate inhibitor with the acyclic spacer between the purine N9 and the phosphate of near optimal length to accommodate these two binding sites. Based on these results, a number of starting compounds were proposed that incorporated these and other features predicted to enhance inhibitor binding. III. Molecular Modeling Structural information in combination with graphical methods for displaying accessible volume, electrostatic potential, and hydrophobicity of the active site of the target macromolecule greatly facilitates the drug design process. Accurate prediction of binding affinities and protein conformational changes are currently not routinely possible, although significant advances are being made. Proposed compounds were screened by modeling the enzyme-inhibitor complex using interactive computer graphics. Macromodel [21] and AMBER [22] based molecular energetics were used along with Monte Carlo/energy minimization techniques [23] to sample the conformational space available to potential inhibitors docked into the PNP active site. Methods based on the work of Goodford [24] employing custom software were also used. Qualitative evaluation of the enzyme-inhibitor complexes by molecular graphics and semiquantitative evaluation of the interaction energies between the inhibitors and the enzyme aided in the prioritization of compounds for chemical synthesis.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_160.html [4/5/2004 5:02:06 PM]

Document

Page 161

IV. Drug Design Progression We focused initially on filling the purine binding region of the active site. That done, we planned to fill the sugar binding region and, finally, the phosphate binding site. We expected that each successive step, moving the compound closer toward fully occupying the active site, would enhance the affinity of the drug candidate for the enzyme. A. The Purine Site From our crystallographic examinations, we knew that three amino acids in the purine binding pocket of PNP formed hydrogen bonds with purines and their mimics. Such linkages are among the strongest reversible chemical bonds that exist. In proposing inhibitor candidates, we concentrated on compounds that would at least form hydrogen bonds with the same three amino acids. Figure 4e shows a close-up of the purine site. We favored exchanging a carbon atom for the nitrogen atom that normally occupies position nine, since there was no interaction of this nitrogen with the active site and earlier studies showed such a change promotes binding to PNP. Guanine modified in this way is called 9-deazaguanine. The first structures selected for synthesis were 9-deazaguanines substituted by an arylmethyl group at the 9 position. These compounds were prepared by adaption of a literature procedure [25]. We further expected that attaching an amino group to the carbon atom in position eight on 9-deazaguanine would enhance affinity, since 8aminoguanine was the first significant inhibitor of PNP. Both 8-aminoguanine analogs and 9-deazaguanine analogs are good inhibitors of PNP. However, introduction of an 8-amino group into the 9-deazaguanine derivatives resulted in decreased potency. To understand this poor binding, we undertook crystallographic analysis of PNP complexes with four compounds having the 9-thienyl substituent attached to guanine (G), 8-aminoguanine (8AG), 9deazaguanine (DG), and 8-amino-9-deazaguanine (8ADG). The results of this analysis are summarized in Figure 5. These data show one mode of binding for compounds that accept a hydrogen bond from Asn 243 at N7 (G and 8AG) and another for compounds that donate hydrogen to Asn 243 from N7 (DG and 8ADG). The 8AG analogs make use of the Thr 242 side chain to form an additional hydrogen bond, which improves binding affinity. In the 9-deazaguanine series, where N7 has an attached hydrogen atom, Asn 243 undergoes a shift that is clearly seen in difference Fourier maps. This shift is caused by the formation of the N7-H…OD(243) hydrogen bond. A concomitant shift by Thr 242 prevents it from hydrogen bonding to the 8-amino group of 8ADG. Furthermore, the shift moves the methyl group of Thr 242 towards the 8-amino

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_161.html [4/5/2004 5:02:11 PM]

Document

Page 162

Figure 5 Comparison of 8-amino and 9-deaza substitutions on guanine. Details are extensively discussed in the text. Cross-hatching schematically indicates the enzyme. Ball and stick diagrams show the inhibitor and key side-chain residues. Nitrogens are dark gray spheres, oxygens are light gray. Arrows indicate hydrogen bonding, with the arrow size showing relative strength.

amino group, generating a hydrophobic environment for the group and decreasing binding affinity. The carbon-for-nitrogen switch in the 9-deaza variant favors association with PNP by substituting a strong hydrogen bond for the relatively weak one occurring between Asn 243 and guanine. Formation of a simple 8-aminoguanine variant leads to tight binding by giving rise to an extra hydrogen bond between the purine derivative and Thr 242. The combination of the two “improvements”—the carbon-for-nitrogen substitution and the addition of the amino group to position eight—was counterproductive because the carbon in position nine prevented the amino group at

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_162.html (1 of 2) [4/5/2004 5:02:21 PM]

Document

Page 163

position eight from forming the extra bond with Thr 242. In fact, it set up an unfavorable, repulsive clash between the threonine and the added amino group. In the absence of detailed structural information, it would have been extremely difficult to explain why affixing the amino group to the carbon in position eight proved unhelpful. But crystallography quickly provided the explanation. 9-Deazaguanine itself would be a better choice for the purine component of an inhibitor. This experience underscores the wonderful economy of the structure-based approach. Without crystallographic data, we might have pursued a logical but unproductive avenue of research much longer than we did. B. Ribose Site The next task was to fill the sugar binding site. The sugar in a nucleoside does not attach to PNP primarily by forming hydrogen bonds, but through hydrophobic attractions. The sugar binding pocket of PNP consists of three hydrophobic amino acids: Phe 200 and Tyr 88 from the same monomer that binds guanine and Phe 159 from the adjacent monomer. Several known inhibitors carried a benzene group attached to position 9 of the purine in place of the sugar in the nucleoside. An initial series of compounds was synthesized to exploit the hydrophobic region in the ribose binding site. A number of 9-substituted 9-deazapurine analogs were prepared with various aromatic, heteroaromatic, and cycloaliphatic substituents. The first 9-deazaguanine derivatives synthesized, such as 9-benzyl-9deazaguanine, were three to six times more potent than the most potent known inhibitor, 8-amino-9-(2thienylmethyl)guanine. The optimum spacer between the purine base and the aromatic substituent proved to be a single methylene group. Crystallographic data showed that generally the planes of the aromatic rings tend to orient in a reproducible conformation. The aromatic groups optimize their interaction with Phe 159 and Phe 200, which results in the classic “herringbone” arrangement reported in a variety of aromatic systems [26]. Inhibitors with cycloaliphatic substituents at N9 of deazaguanine were also as potent as the aromatic analogs. The cycloaliphatic substituents occupied the same general volume as the aromatic groups. As with the aromatic series, the optimum spacer between the 9-deazaguanine and the hydrophobic substituent is one carbon atom. X-ray analysis of the PNP complexes of 9-cyclohexyl-9-deazaguanine, a relatively poor inhibitor, and the complex of 9-cyclohexylmethyl-9-deazaguanine, a potent inhibitor, showed the two cyclohexyl groups occupy approximately the same space in the active site with the purine base pulled out of its optimal position in the former. The chemistry is more straightforward with the aromatic series. From modeling studies, we saw that the sugar binding pocket could be filled more

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_163.html [4/5/2004 5:02:22 PM]

Document

Page 164

completely by adding any of several chemical groupings to the benzene ring. The best fit came from adding a chlorine atom to position 3 of the benzene ring. C. Phosphate Site The final step added a group that would interact with the phosphate binding site either directly or via electrostatic interactions. We could not use phosphate itself, because phosphate-containing compounds are not metabolically stable and have difficulty passing through cell membranes intact. Acyclovir diphosphate, which is not membrane permeable and is subject to extracellular metabolism, is a good example. Our results suggested that an ideal PNP inhibitor in the 9-deazapurine series would contain an aromatic group and a substituent with affinity for the phosphate site interlinked by spacers with optimum lengths. Crystallographic and modeling studies suggested a two-to-four-atom spacer. Initial modeling studies encouraged us to prepare several structures, but they failed to improve the binding affinity of our twopart structure. Crystallographic analysis of a number of PNP inhibitor complexes revealed significant displacement of the inhibitors. These displacements appear to be the result of close contacts between the inhibitor and the ion in the phosphate binding site. Sulfate ions occupy the phosphate site in PNP crystals as they are grown from ammonium sulfate solution. These inhibitors were more potent when the binding was measured in 1 mM phosphate solution rather than in 50 mM phosphate. Kinetic studies showed that these inhibitors were competitive not only with inosine but also with phosphate, in keeping with the above observation. These results, summarized in Table 1, show that the IC50 (50 mM) is equal to or larger than the IC50 (1 mM), in some cases by as much as 100-fold. The ratio and the dimension of the 9-substituent show some correlation. Compounds such as 8-aminoguanosine and 8-amino-9-(2-thienylmethyl)guanine show no difference. Since the concentration of phosphate in intact cells is 1 mM, we routinely used this assay condition for all PNP inhibitors. Starting with a model of the 9-benzyl-9-deazaguanine/PNP complex, we concluded that two of the positions on the 9-benzyl group, namely the 2-position of the phenyl ring and one of the benzylic sites, appeared to be oriented so that a group attached to either one could interact favorably with the phosphate binding site. The first compound made in this series, 9-[2-(3-phosphonopropoxy)benzyl]guanine, turned out to be a poor PNP inhibitor. Subsequent crystallographic analysis revealed that the plane of the aromatic ring had rotated approximately 90° from its optimum position in the hydrophobic pocket. This reorientation of the ring was presumably necessary to accommodate the four atom spacer between the phenyl ring and the phosphonate group. A compound

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_164.html [4/5/2004 5:02:25 PM]

Document

Page 165 Table 1 Inhibition Data for Selected PNP Inhibitors by Increasing IC50 Value

IC50, µM R2a

R2

(S)-3-Chlorophenyl 3-Chlorophenyl

CH2CO2H CH2CN

50 mM phosphateb

Ratiod 1 mM phosphatec

0.031

0.0059

5.3

1.8

0.010

180

2-Tetrehydrothienyl

H

0.22

0.011

20

3,4-Dichlorophenyl

H

0.25

0.012

21

3-Thienyl

H

0.08

0.020

4

3-Trifluoromethylcyclohexyl

H

0.74

0.020

37

Cyclopentyl

H

1.8

0.029

62

Cycloheptyl

H

0.86

0.030

29

Pyridin-3-yl

H

0.20

0.030

2-(Phosphonoethyl)phenyle

H

0.45

0.035

13

Cyclohexyl

H

2.0

0.043

47

2-Furanyl

H

0.31

0.085

3.6

CH2CO2H

0.90

0.16

5.6

H

42

1.0

(R)-3-Chlorophenyl 2-Phosphonopropoxyphenyle aCompounds

with R2 not equal to H are racemic mixtures unless the R or S isomer is designated.

bCalf

spleen PNP assayed in 50 mM phosphate buffer.

cCalf

spleen PNP assayed in 1 mM phosphate buffer.

dIC

at 50 mM phosphate divided by IC50 at 1 mM phosphate.

50

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_165.html (1 of 2) [4/5/2004 5:02:28 PM]

7.3

42

Document

eGuanine

base.

Source: Ref. 27.

with a two-carbon spacer was a much better PNP inhibitor; however, it was clear from x-ray analysis that the aromatic ring was unable to form the ideal “herringbone” packing interaction. Alternatively, compounds were modeled in which the spacer to the phosphate binding site branched from the benzylic carbon, thus placing no restrictions on the tilt of the aromatic ring. Examination of the 9-benzyl-9-deazaguanine/PNP complex indicated that of the two benzylic positions, one (pro-R) pointed into a sterically crowded area within the active site, whereas the other (pro-S) pointed into a relatively empty space adjacent to the phosphate binding site. This analysis led to the synthesis of racemic 9-[1-(3chlorophenyl)-2-carboxyethyl]-9-deazaguanine. This compound adds an acetate group (CH2COO-) to the methylene carbon atom that joined 9-deazaguanine to the chlorinated benzene ring. This compound was resolved into its (S) and (R) enantiomers.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_165.html (2 of 2) [4/5/2004 5:02:28 PM]

Document

Page 166

As predicated, the (S) acid was a 30-fold more potent inhibitor of PNP than the (R) form. X-ray crystallographic analysis of the complexes revealed that the (S) acid was oriented properly for optimal interactions with all three subsites (Figure 4F), whereas the (R) acid was not. This series of compounds contains the most potent membrane-permeable inhibitors of PNP yet reported [27]. V. Summary Recently, scientists at BioCryst have successfully completed a project to design and synthesize potent inhibitors of the enzyme Purine Nucleoside Phosphorylase (PNP) using the three-dimensional structure of the active site. Crystallographic and modeling methods have been combined with organic synthesis to produce inhibitors. Our experience in creating a set of potential drugs—one of which (BCX-34) is now in human trials for treating psoriasis and a form of T-cell lymphoma—illustrates the process and the power of structure-based design. This structure-based inhibitor design approach led to a number of inhibitors more than 100 times more potent than any membrane-permeable inhibitor available at the beginning of this project. During the two and half years of this project, about 60 active compounds were synthesized. This is a remarkably small number compared with the extensive synthesis programs generally involved in drug discovery by trial and error techniques. The large number of active compounds and the enhancement of inhibitor potency stand as proof that crystallographic and modeling techniques are now capable of playing a critical role in the rapid discovery of novel therapeutic agents. The entire protocol, from choosing the target to creating a drug suitable for clinical trials, can probably be accomplished today in two or three years. A. Obstacles Encountered and Lessons Learned Crystallographic analysis was based primarily on the results of difference Fourier maps in which the interactions between residues in the active site and the inhibitor could be characterized. During these studies, about 35 inhibitor complexes were evaluated by x-ray crystallographic techniques. It is noteworthy that the resolution of the PNP model extends to only 2.8 Å and that all of the difference Fourier maps were calculated at 3.2 Å resolution, much lower than often considered essential for drug design. Crystallographic analysis was facilitated by the large solvent content that allowed for free diffusion of inhibitors into enzymatically active crystals. Initial inhibitor modeling attempts using the native PNP structure were far less successful than subsequent analyses in which coordinates for the guanine-

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_166.html [4/5/2004 5:02:29 PM]

Document

Page 167

PNP complex were used, mainly because of the magnitude of the changes that occur during substrate binding. We found that computer modeling required significant tuning in order to provide useful results. Crystallographic results were useful in testing and modifying modeling parameters. The most useful modeling results were achieved after incorporation of the conformational searching techniques described earlier and when the coordinates for the PNP-guanine complex model were used. Visual inspection and chemical intuition were very important. B. Perspectives of Treating Targeted Disease One of the inhibitors designed during the drug discovery process, 9-(3-pyridylmethy)-9-deazaguanine (BCX-34), was selected for initial clinical development. Current clinical trials utilize both topical and oral formulations of the drug. Researchers at the University of Alabama at Birmingham and Washington University School of Medicine have recently completed small Phase II clinical trials of two indicated applications, cutaneous T-cell lymphoma (CTCL) and psoriasis, using a topical formulation of BCX-34. Although patients showed improvement in both trials, the duration of each was too short (six weeks) to adequately assess the efficacy of the drug. Subsequently 80% of the patients from the CTCL trial (24 patients) entered an open label trial for treatment of their disease for up to twelve months. At the end of the first six months of treatment, seven of the patients were in complete remission (verified by biopsy), two patients showed a clinical complete response, and nine patients had shown definite improvement. The other six patients had shown no change or progression of disease. No serious, drug-related adverse events were reported during the study. The process of structure-based drug design helped to ensure that the inhibitor would be highly selective for the PNP enzyme, and thus far no other targets for the drug have been identified. The mechanism of action of BCX-34 appears to be entirely related to its effect on the proliferation of human T-cells. This high degree of specificity probably also contributes to the high safety profile of the drug. Although longterm studies in more patients will be necessary to substantiate these results, it appears likely that BCX34 will have a significant clinical effect on at least some T-cell mediated diseases. Based on the results from these three trials, BioCryst has initiated a multicenter Phase III trial for the treatment of CTCL, as well as a large, multicenter Phase II trial for psoriasis. In addition to the two clinical trials using the topical formulation, a Phase I clinical trial in CTCL and T-cell lymphoma/leukemia has begun using an oral formulation of BCX-34. In the future, a number of other Tcell mediated diseases or processes are possible targets for BCX-34, including rheumatoid arthritis, multiple sclerosis, inflammatory bowel disease, and organ transplant rejection.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_167.html [4/5/2004 5:02:31 PM]

Document

Page 168

References 1. Parks RE Jr., Agarwal RP. In: Boyer PD, ed. The Enzymes. 3rd Ed., New York: Academic, 1972; 7:483–514. 2. Stoeckler JD. In: Glazer, RE, ed. Developments in Cancer Chemotherapy. Florida: CRC, Baco Raton, 1984; 35–60. 3. Friedkin M, Kalckar H. In: Boyer PD, Lardy H, Myrback K, eds. The Enzymes. 2nd Ed. New York: Academic, 1961; 5:237–55. 4. Agarwal KC, Agarwal RP, Stoeckler JD, Parks RE Jr. Purine nucleoside phosphorylase. Microheterogeneity and comparison of kinetic behavior of the enzyme from several tissues and species. Biochemistry 1975; 14:79–84. 5. Bzowska A, Kulikowska E, Shugar D. Properties of purine nucleoside phosphorylase (PNP) of mammalian and bacterial origin. Z Naturforschung C Biosci 1990; 45:59–70. 6. Stoeckler JD, Agarwal RP, Agarwal KC, Schmid K, Parks RE Jr. Purine nucleoside phosphorylase from human erythrocytes: physiocochemical properties of the crystalline enzyme. Biochemistry 1978; 17:278–83. 7. Williams SR, Goddard JM, Martin DW Jr. Human purine nucleoside phosphorylase cDNA sequence and genomic clone characterization. Nucleic Acids Res 1984; 12:5779–87. 8. LePage GA, Junga IG, Bowman B. Biochemical and carcinostatic effects of α'-deoxythiguanosine Cancer Res 1964; 24:835–40. 9. Giblett ER, Ammann AJ, Wara DW, Sandman R, Diamond LK. Nucleoside-phosphorylase deficiency in a child with severely defective T-cell immunity and normal B-cell immunity. Lancet 1975; 1:1010–3. 10. Otterness I, Bilven M. In: Rainsford K, Velo G, eds. New Developments in Antirheumatic Therapy. Inflammation and Drug Therapy Series. Norwell, MA: Kluwer Academic, 1989; 2:277–304. 11. Stoeckler JD, Cambor C, Kuhns V, Chu SH, Parks RE Jr. Inhibitors of purine nucleoside phosphorylase, C(8) and C(5') substitutions. Biochemical Pharmacology 1982; 31:163–71. 12. Shewach DS, Chern JW, Pillote KE, Townsend LB, Daddona PE. Potentiation of 2'-deoxyguanosine cytotoxicity by a novel inhibitor of purine nucleoside phosphorylase, 8-amino-9-benzylguanine. Cancer Res 1986; 46:519–23. 13. Stoeckler JD, Ryden JB, Parks RE Jr, Chu MY, Lim MI, Ren WY, Klein RS. Inhibitors of purine nucleoside phosphorylase: effects of 9-deazapurine ribonucleosides and synthesis of 5'-deoxy-5'-iodo-9deazainosine. Cancer Res 1986; 46:1774–8. http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_168.html (1 of 2) [4/5/2004 5:02:33 PM]

Document

14. Tuttle JV, Krenitsky TA. Effects of acyclovir and its metabolites on purine nucleoside phosphorylase. J Biol Chem 1984; 259:4065–9. 15. Gilbertsen RB, Scott ME, Dong MK, Kossarek LM, Bennett MK, Schrier DJ, Sircar JC. Preliminary report on 8-amino-9-(2-thienylmethyl) guanine (PD 119,229), a novel and potent purine nucleoside phosphorylase inhibitor. Agents and Actions 1987; 21:272–4. 16. Cook WJ, Ealick SE, Bugg CE, Stoeckler JD, Parks RE Jr. Crystallization and preliminary X-ray investigation of human erythrocytic purine nucleoside phosphorylase. J Biol Chem 1981; 256:4079–80.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_168.html (2 of 2) [4/5/2004 5:02:33 PM]

Document

Page 169

17. Ealick SE, Rule SA, Carter DC, Greenhough TJ, Babu YS, Cook WJ, Habash J, Helliwell JR, Stoeckler JD, Parks RE Jr, Chen SF, Bugg CE. Three-dimensional structure of human erythrocytic purine nucleoside phosphorylase at 3.2 Å resolution. J Biol Chem 1990; 265:1812–20. 18. Narayana SVL, Bugg CE, Ealick SE. Refined structure of purine nucleoside phosphorylase at 2.75 Å resolution. Acta Cryst D 1996; accepted. 19. Babu YS, Refined structure of guanine: purine nucleoside phosphorylase at 2.8 Å resolution. In preparation. 20. Koshland DE Jr. The key-lock theory and the induced fit theory. Angew Chem Int Ed Engl 1994; 33:2375–8. 21. Mohamadi F, Richards NGJ, Guida WC, Liskamp R, Lipton M, Caufield C, Change G, Hendrickson T, Still WC. MacroModel—An integrated software system for modeling organic and bioorganic molecules using molecular mechanics. J Comput Chem 1990; 11:440–67. 22. Weiner SJ, Kollman PA, Case DA, Singh UC, Ghio C, Alagona S, Profeta S, Weiner P. New force field for molecular mechanical calculations simulations of proteins and nucleic acids. J Am Chem Soc 1984; 106:765–84. 23. Chang G, Guida WC, Still WC. An internal coordinate monte carlo method for searching conformational space. J Am Chem Soc 1989; 111:4379–86. 24. Goodford P. A computational procedure for determining energetically favorable binding sites on biologically important macromolecules. J Med Chem 1985; 28:849–57. 25. Montgomery JA, Niwas S, Rose JD, Secrist JA 3d., Babu YS, Bugg CE, Erion MD, Guida WC, Ealick SE. Structure-based design of inhibitors of purine nucleoside phosphorylase. 1. 9-(arylmethyl) derivatives of 9-deazaguinine. J Med Chem 1993; 36:55–69. 26. Burley SK, Petsko GA. Aromatic-aromatic interaction: a mechanism of protein structure stabilization. Science 1985; 229:23–8. 27. Ealick SE, Babu YS, Bugg CE, Erion MD, Guida WC, Montgomery JA, Secrist JA 3d. Application of crystallographic and modeling methods in the design of purine nucleoside phosphorylase inhibitors. PNAS USA 1991; 88:11540–4.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_169.html [4/5/2004 5:02:34 PM]

http://legacy.netlibrary.com/reader/message.asp?message=811&BookID=12640&FileName=Page_170.html

The requested page could not be found. Return to previous page

http://legacy.netlibrary.com/reader/message.asp?message=811&BookID=12640&FileName=Page_170.html [4/5/2004 5:02:36 PM]

Document

Page 171

6 Structural Implications in the Design of Matrix-Metalloproteinase Inhibitors John C. Spurlino 3-Dimensional Pharmaceuticals, Inc., Exton, Pennsylvania I. Matrix-Metalloproteinases The matrix metalloproteinases (MMPs) are a family of ubiquitous enzymes that are involved in extracellular matrix degradation and remodeling. They are critical for the processes of morphogenesis and wound healing, but are also implicated in many human diseases including arthritis, metastasis, and cancer tumor growth [1, 2]. This family includes matrilysin, fibroblast collagenase (HFC), neutrophil collagenase (HNC), stromelysin 1 (HFS), stromelysin-2, stromelysin-3, gelatinases A and B, collagenase3, and the membrane type MMP. In addition to the destructive involvement in diseases, MMPs play a critical role in the remodeling of the extracellular matrix [3]. The MMP enzyme family is part of the superfamily of metzincins. The metzincin superfamily is distinguished by a conserved zinc binding motif for the catalytic zinc and a Met-turn region [4]. The MMPs are unique in that they also contain a second structural zinc, however this zinc may be absent in the intact full-length enzyme [5]. The presence of one to four structural calcium ions has been detected in the MMPs that have been characterized to date. The importance of the zinc ions and at least one of the structural calcium ions to enzymatic activity has been proven [6]. The MMPs are secreted as inactive proenzymes, which are activated by proteolytic cleavage. Once activated they are subject to control by tissue inhibitors of metalloproteinases (TIMPs). It is the imbalance between the active enzymes and the TIMPs that leads to destructive tissue degradation that potent directed pharmaceuticals can overcome. These enzymes have been the target of

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_171.html [4/5/2004 5:02:38 PM]

Document

Page 172

drug design since the late 1970s [7]. Batimastat, [{4-N-hydroxyamino}-2R-isobutyl-3S-{thienylthiomethyl} succinyl]-L-phenyl-alanine-N-methylamide, a potent nonspecific MMP inhibitor from British Bio-Tech, is now in Phase III trials. The information gained from current studies indicate that there is some efficacy in the treatment of disease states by MMP inhibitors [8–10]. There is still much debate about whether a broad spectrum or directed MMP inhibitor is the best course of treatment for a variety of diseases, partly because the exact role of individual MMPs is still unclear. Both collagenase-3 and HFC are suspected to cause osteoarthritis [11]. It is currently believed that gelatinase and collagenase-3 have a role in breast cancer [12]. Gelatinase A and B have been implicated in hemorrhagic brain injury [13]. Gelatinase A and B, HFC, and stromelysin may all be involved in gastric cancer [14]. Matrilysin may be implicated in colon cancer [15]. Increased gelatinase A and B activity has also been seen in response to beta-amyloid production [16]. The role that individual MMPs play in causing these diseases, however, remains unclear. This uncertainty underscores the need to develop selective inhibitors of individual MMPs to ferret out the roles each play in the development of specific disease states. The MMPs consist of one or more structural domains (Figure 1). The first domain, the propeptide domain, confers a self-inhibitory action on the full-length MMP. The second domain contains the active site residues and is referred to as the catalytic domain. The catalytic domain is characterized by the conserved zinc-binding sequence (HEXGHXXGXXHS), which also contains the glutamate residue that is essential for activity [17]. The MMPs are activated by cleavage of the prodomain. All MMPs contain these first two domains. Matrilysin, the simplest of the MMPs, is an example of a two-domain enzyme, where the active enzyme consists solely of the catalytic domain. The remainder of the MMPs also contain a hemopexin-like domain connected to the catalytic domain by a proline-rich linker. This domain is involved in the interaction of the collagenases and stromelysins with collagen and, in the case of the collagenases, is essential for activity against collagen [18]. The cleavage of the proline-rich linker region in HFC and HNC is another route to control collagenase activity. The hemopexin domain of the gelatinases is not necessary for collagen binding, but may be involved in receptor recognition [19]. The gelatinases also contain a fibronectin-like insert in the catalytic domain that is involved in binding collagen [20]. The fibronectin domain has also been shown to be essential for elastolytic activity [21]. A structural picture of these additional domains is essential for an understanding of the mode of action for these larger MMPs, but is not necessary for a structure-based drug design strategy. Differences in the catalytic domains of the MMPs can be used to drive a targeted drug discovery effort.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_172.html [4/5/2004 5:02:39 PM]

Document

Page 173

Figure 1 A ribbon model of the full-length collagenase structure (1fbl.pdb). The prodomain would precede and include the portion labeled in the figure. The catalytic domain is shown, with a highlighted region where the fibronectin-like domain of the gelatinases is inserted.

II. 3-Dimensional Structure of MMPs Catalytic domain structures for fibroblast collagenase [22–25], neutrophil collagenase [26,27], matrilysin [28], and stomelysin [29, 30] have all been determined and deposited in the Protein Data Bank [31]. The catalytic domains of MMPs, as seen in the archetypal collagenase structure (shown as a ribbon drawing [32] in Figure 1), consist of an upper 5-stranded β sheet flanked by two α helices on one side of the active site cleft and a long loop that contains the Met-turn flanked by a single α helix on the other side of the cleft. The active-site groove as seen in the solvent-accessible surface [33] is an obvious structural feature (Figure 2). The top wall of the cleft (as seen in Figure 2) is formed by the top strand of the β sheet and the loop that contains the calcium binding site. The lower wall of the cleft is formed from the residues on either side of the Met-turn. These residues can be considered as an interrupted strand which, together with the substrate, complete the twisted β sheet of the catalytic domain. The bottom of the cleft is formed by the second helix, which

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_173.html (1 of 2) [4/5/2004 5:02:46 PM]

Document

Page 174

Figure 2 The accessible surface of HFC with a modeled substrate from human collagen showing the binding sites.

contains the HExxH motif, the catalytic zinc, and the S1' pocket. Substrates bind in an extended conformation that approximates an antiparallel strand. The cleft, however, is not large enough to accommodate a triple helix collagen molecule. A structure for the full-length active porcine synovial collagenase [34] has been determined. The structure of the catalytic domain of this full-length enzyme is equivalent to the structures of the isolated catalytic domains of HFC, HNC, and matrilysin. The flexible linker domain between the catalytic and hemopexin domains is disordered and the orientation of the hemopexin domain in the structure offers no real clue as to the mode of action for the full-length collagenases. Furthermore, the matrilysin structure of the full-length active enzyme has almost identical secondary structural features (a Ca overlap of

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_174.html (1 of 2) [4/5/2004 5:02:54 PM]

Document

Page 175

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_175.html (1 of 2) [4/5/2004 5:03:13 PM]

Document

Figure 3 The sequence alignment of MMPs with the catalytic domain region highlighted. The residues that line the subtrate pockets are marked: S3 (3), S2 (2), S1 (1), S1 (*), S2' (@), and S3' (#). The highlighted catalytic domain alignment was dominated by the structural alignment of the determined structures of HFC, HNC, and matrilysin.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_175.html (2 of 2) [4/5/2004 5:03:13 PM]

Document

Page 176

0.43 Å) as the catalytic-domain structure of HFC. This demonstrates that the absence or presence of the hemopexin domain does not affect the overall structure of the catalytic domain. The sequence homology of the catalytic domains of the collagenases is 62%. This can be extended to the other members of the MMP family as seen in Figure 3. An understanding of the structural features of the target enzyme is essential for structure-based drug design. In this example we will be looking at inhibiting MMPs by binding to the active site. The numbering system used throughout this chapter will be in regard to the HFC sequence used in 1hfc.pdb. III. Surface Features The surface features of matrilysin, fibroblast collagenase, and neutrophil collagenase are all similar (Figure 4). The active-site groove can be plainly seen on the surface: two main pockets punctuated by the active-site zinc. Modeling

Figure 4 The accessible surfaces of HFC (a), HNC (b), and matrilysin (c) are shown with a bound P'-side hydroxamate inhibitor.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_176.html (1 of 2) [4/5/2004 5:03:26 PM]

Document

Page 177

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_177.html (1 of 2) [4/5/2004 5:03:36 PM]

Document

Page 178

studies show that there is not sufficient room in the S1'-S3' cleft to accommodate the native coiled triple collagen bundle. The active site consists of a series of subsites on either side of the catalytic zinc. These subsites are numbered starting at the catalytic zinc and preceding from N to C as S1', S2', etc., corresponding to the residues (P1', P2', etc.) of the substrate that is bound (Figure 2). Likewise the subsites are numbered S1, S2, etc. outward from the other side of the catalytic zinc. While the interaction of the substrate (and the inhibitor) with the catalytic zinc is the most important interaction, the remainder of the substrate (inhibitor) also forms hydrogen bonds with residues from the top strand of the β sheet and the loop region posterior to the Met-turn. These interactions with the substrate in the binding pockets of the MMPs are the prime targets for engineering specific MMP inhibitors. An in-depth understanding of the differences of the properties of these pockets in the different MMPs and the interactions of specific residues within these pockets is essential for structure-based design of inhibitors. IV. Main-Chain Substrate Interactions Most of the hydrogen bonds between the substrate and the MMP occur with the top strand of the β sheet. The P3 residue does not make any direct hydrogen bonds with the MMP. The P2 residue makes two hydrogen bonds with residue 184, which is a conserved alanine residue in all the aligned MMP sequences. Residue 183 is a conserved histidine, which is bound to the structural zinc, further stabilizing the conformation of the top strand. The carbonyl oxygen of P1 is liganded to the catalytic zinc. The lefthand side of the substrate is thus held in place by only two hydrogen bonds with the enzyme and one interaction with the catalytic zinc. Although the P3 residue does not make any hydrogenbond contributions to substrate binding, it is essential for catalytic activity [43]. The right-hand side of the substrate is held much tighter. The P1' residue's carbonyl oxygen makes a hydrogen bond with the amide nitrogen of residue 181, which is a conserved leucine residue. The amide nitrogen of the P1' residue is hydrogen bonded to the conserved alanine-182 carbonyl oxygen. The P2' substrate residue is held in place by hydrogen bonds to proline 238 and tyrosine 240, two more conserved residues. The amide nitrogen of P3' makes a hydrogen bond with the carbonyl oxygen of residue 179, the only nonconserved residue, making a main-chain interaction with the substrate. The use of conserved residues to maintain the main-chain interactions along the substrate backbone makes differentiation of the MMPs through these interactions difficult. Instead, differences in the regions where the side chains of the substrate interact can be used to drive the discovery of specific MMP inhibitors. The hydrogen-bonding pattern also indicates that right-hand side (P'-

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_178.html [4/5/2004 5:03:38 PM]

Document

Page 179

side) inhibitors will bind with greater affinity. Indeed, most of the structures of MMPs were determined with right-hand side inhibitors, and most of the pharmaceuticals currently in development are also righthand side binders. A closer look at the binding pockets themselves also demonstrates the reasons for the preference of researchers for the right-hand side. V. Substrate-Binding Pockets The nonprimed or left-hand side of the cleft consists of a large shallow depression. The S1 pocket consists of a shallow ridge that complements the glycine residue of the collagen strands. Most of the interactions with the glycine residue are brought about due to its interaction with the catalytic zinc. Asparagine 180 approaches the P1 residue in HFC. Crystallographic evidence indicates an interaction of the thiophene ring of batimastat via electrostatic interactions of the p orbitals and the catalytic zinc and the possibility of a water-mediated hydrogen bond to the carbonyl O of residue 184 [35]. Larger substituents can be accommodated in regions adjacent to the P1 pocket, possibly in the large pocket above the S1 site (see Figure 2). Increased potency was noted for several compounds with cyclic imido P1 substituents that could bind here [2]. The S2 pocket is a large shallow depression offering no real binding cavities. One side of the pocket is made up from the conserved histidine at position 228 and the main chain from residue 227. The bottom of the pocket is formed by histidine 222. Both of these histidines are liganded to the catalytic zinc. The other side consists mostly of the residue 186 side chain with some hydrogen-bonding contacts possible from the tip of the glutamine side chain in the case of HFC and HNC. The S3 pocket offers a shallow cavity to bind the conserved proline. The proline residue of the substrate would lie between the side chains of His 183, Phe 185, and Ser172 [36]. Residues 183 and 185 are conserved among the MMPs with the minor exception of a tyrosine replacing phenylalanine 164 in stromelysin. Residue 172 shows some variability among the MMPs existing as a serine in HNC and HFC and a tyrosine in the remainder of the aligned MMPs. The primed or right-hand side of the active site exists as a narrow canyon with a large well at the beginning. The S1' pocket is a narrow, deep cavity providing an ideal binding site for inhibitor design. The S1' pocket is the most significant feature of the surface, extending as a tunnel completely through the enzyme in the case of neutrophil collagenase and stromelysin. This feature makes the S1' pocket an ideal candidate for use in designing an inhibitor with specificity for HNC, HFC, or HFS. The volumes of the S1' pockets vary greatly. Matrilysin has the smallest S1' pocket at 111 Å3. The fibroblast collagenase pocket is not much larger at

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_179.html [4/5/2004 5:03:39 PM]

Document

Page 180

Figure 5 (a) A cut away view of the S1' pocket of HFC showing the termination of the pocket by Arg214. Matrilysin also has a truncated pocket. (b) A cut away view of the S1' pocket of HNC showing the clear path through to the other side of the enzyme. The gelatinases are also likely to have large extended S1' pockets.

123 Å3. The pocket of the neutrophil collagenase that travels through the enzyme has a volume of 305 Å3. Figure 5 demonstrates the differences in the relative sizes of the S1' pockets of HFC and HNC. Stromelysin has a pocket that should be of similar size as that of neutrophil collagenase [21]. The gelatinases and collagenase-3, based on sequence alignment, should also posses long tunnel-like S1' pockets. The residues that line the S1' pocket are mostly hydrophobic residues (see Figure 3) and show a remarkable overall similarity. The specificity of a number of inhibitors of MMPs can be linked to differences in the S1' pockets. The S1' pocket of HFC is terminated by arginine 214, while matrilysin has a tyrosine residue that accomplishes the same thing. The remainder of the aligned MMPs have leucine residues at that position. In addition the conformation of the leucine residue is swung back, forming the tunnel. There are three additional residues that are significant in their differences within the S1' pocket: residues 239 and a two-residue insert, relative to HFC, after residue 242. These residues

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_180.html (1 of 2) [4/5/2004 5:03:48 PM]

Document

Page 181

form the lower end of the tunnel. The major effect of the different residues found here is on the diameter of the exit hole in the S1' tunnel. However, there are some residues that could be targeted for hydrogenbond formation. The S2' binding cavity is a narrow cleft that can easily accommodate a peptide backbone, but with no room for a side chain. The interaction of the P2' side chain is made with the exterior surface of the enzyme. The S2' site is exposed to solvent and presents two possible interaction sites for bound inhibitors that are related by a rotation about χ1. These sites consist of residues 179–180 on one side and residues 238–240 on the other. The S3' binding cavity opens up out of the canyon and consists almost entirely of surface interactions. The side chain of P3' interacts with residues 210 and 240, which are mostly conserved tyrosine residues. Additional interactions could be formed with some of the additional residues found in the insertions after residue 242. VI. Structure-Based Design Structure-based drug design is an iterative process that starts with a lead compound, a structural model of the target, and a structure-activity relationship

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_181.html (1 of 2) [4/5/2004 5:03:56 PM]

Document

Page 182

(SAR) model. The lead compound can come from compound screening, previously discovered inhibitors, or it can be based on a known substrate. The model can be obtained from x-ray crystallography, high-field nuclear magnetic resonance spectroscopy, or from homology-based model building. Inhibitor structures are developed and docked into the model of the binding site of interest, typically the active site of the enzyme. The interactions of the inhibitor-enzyme complex are evaluated and ranked. The most promising compounds are then synthesized and tested. Based on the results of the testing, additional enzyme-inhibitor structures are determined, the SAR model is updated, and the process beings again. As a model case of structure-based drug design for MMPs we will look at the design of a right-handed inhibitor based on the x-ray structures of HFC and HNC. VII. Zinc-Binding Group The design of active-site inhibitors based on the natural substrate of the collagenases has produced a variety of zinc-binding groups to anchor the inhibitor to the catalytic zinc. These group include hydroxamates, thiols, phosphorous acid derivatives (phosphinate, phosphonate, phosphoramidate), and carboxylates. The selection of a suitable zinc-binding group has been studied in depth [37–40]. The most potent zinc-binding group found for the collagenases to date is the hydroxamate. The structural comparison of hydroxamate, carboxylate, and sulfodiimine in matrilysin provided information on the contribution of the zinc ligand to the overall potency of the inhibitor [19]. The potency of the zinc-bind group can be directly related to the number of bonds in which it is involved for this instance. The hydroxamate is the perfect bidentate ligand to the zinc with both oxygens being within 2.2 Å of the zinc. The hydroxamate group also is involved in hydrogen bonds with Glu219 and the carbonyl oxygen of Ala182. The carboxylate group is also a bidentate ligand to zinc, however the oxygens are not equidistant from the zinc. The carboxylate forms only one additional hydrogen bond with Glu219 of the enzyme. The sulfodiimine bound to matrilysin is a monodentate zinc ligand and the weakest of the zinc-binding groups. The comparison of several inhibitors with both carboxylate and hydroxamate zinc-binding groups demonstrates this property in fibroblast and neutrophil collagenases as well. While the potency of inhibitors with different zinc-binding groups maps directly to the number of bonds formed by the zinc-binding group, some of the increase in potency of the hydroxamate group over charged groups most likely is a result of the decreased energetic cost of the desolvation of the neutral hydroxamate.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_182.html [4/5/2004 5:03:58 PM]

Document

Page 183

VIII. S1' Interactions Matrix metalloproteinase structural studies of the P'-side inhibitors to date show a common set of inhibitor-enzyme interactions. This can be attributed primarily to the strong directional zinc-binding forces. Further stabilizing forces from the backbone hydrogen-bonding patterns common to a β sheet allow for minor adjustments due to the zinc interactions to be made while maintaining a common pharmacophore. The fairly rigid constraints of binding based on the known hydroxamate inhibitors allows the use of computer-aided modeling to play a useful role in the design of specific MMP inhibitors. Exploration of the S1' pocket was carried out by docking the P1' group within the cavity followed by rounds of energy minimization. In order to maintain the integrity of the MMP structure several limitations were used. All MMP atoms that are greater than 8 Å from the docked inhibitor were frozen. The Cα atoms of all residues within 8 Å of the inhibitor were initially constrained to their original position by a 20 kcal/mol Å2 force constant that was gradually relaxed to 1 kcal/mol Å2. Strong initial constraints were also placed on the conserved hydrogen bonds and zinc-ligand interactions. This method has several advantages and disadvantages over the common static treatment of target structures. The advantages are that it more closely approximates the actual dynamic state of protein structure and is not computationally prohibitive. The disadvantages include the increased computational cost over a static enzyme target and the fact that gross structural rearrangements can still not be accounted for. The structural similarity of the active site of the MMP family allows structure-based drug design to effectively be used for those enyzmes whose structures have not been determined yet. Examination of the S1' cavities of HFC and HNC clearly indicates a path for designing inhibitors that bind preferentially (Figure 5). The cavity of HFC is mostly filled by the leucine side chain of the preferred substrate, while the S1' pocket of HNC [26] and HFS [30] remains unfilled. The sequence similarities of the gelatinases with HNC indicate that they too can accommodate a much larger P1' group. A series of compounds was designed to explore filling the long S1' tunnel of gelatinase B [41, 42]. The optimum length for binding to gelatinase B was found. There was an increase in affinity for the phenolic ethers versus the benzylic ethers of the same overall length for binding to gelatinase B, but there was no preference seen in binding to HFS. Surprisingly the phenolic ethers also showed potent binding to HFC. The size of the bottom of the S1' pocket and the differences in the preferred torsion angle for the bond to the aromatic ring both play a role in this differential binding.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_183.html [4/5/2004 5:03:59 PM]

Document

Page 184

IX. S2' Interactions The interactions at the S2' site display an interesting structure-activity relationship. While the interactions do not include any hydrogen bonds, favorable stacking interactions do play a significant role in binding. A glycine residue at P2' results in a loss of three orders of magnitude in potency for otherwise identical inhibitors [41]. There is a preference for an aromatic ring at this position in natural peptide substrates [43]. Structural considerations also allow the placement of a t-butyl group here. The potency of a t-butyl glycine P2' is less than that of a phenylalanine (Table 1), but the expected gain in bioavailability brought about by shielding the amide bond from solvation effects should compensate for the loss in potency. X. Conclusions High resolution x-ray crystallographic structure determination is an essential step in structure-based drug design. The need for high resolution structural data

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_184.html [4/5/2004 5:04:31 PM]

Document

Page 185

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_185.html (1 of 2) [4/5/2004 5:04:42 PM]

Document

Figure 6 (a) The pocket of HFC with a leucine residue in the S1' pocket. (b) The volume of the S1' pocket of HFC can change when there are favorable interactions. The binding of the (CH2)4OPh can cause Arg214 to move, thereby making room for the extended side chain.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_185.html (2 of 2) [4/5/2004 5:04:42 PM]

Document

Page 186

to develop an appropriate SAR is demonstrated in the case of the inhibitors shown in Table 1. The unusual potency of a benzylic ether for HFC was unexpected and would not have been predicted with standard docking and minimization studies (Figure 6). The differences in the potency of the various 4-substituted analogs of inhibitor 10 against HFC suggest ππ interactions are the driving force for the displacement of arginine 214. The electron-withdrawing Cl substitution decreases the affinity for HFC while increasing the affinity for HFS. The leading 4-pentyl group can not effectively interact with arginine 214 in HFC; therefore, the rearrangement does not occur. The open channel that is present in HFS and gelatinase B presents no such impediment to binding and the affinity is essentially equal to the unsubstituted form. Not all structure-based design experiments are successful. Attempts to displace the arginine residue that caps the S1' pocket of HFC by forming a salt link with a P1' carboxylate or hydroxyl moiety were unsuccessful [42]. However, these failed attempts offer some redeeming features in the refinement of parameters that can be used to evaluate the energetic potentials for displacing buried water molecules as well as the inherent desolvation energies for polar compounds. The outlook for structure-based drug design is good. The advancement in both x-ray area detectors and computer hardware will make the determination of a series of compounds bound to a target enzyme for use in SAR development commonplace in drug-discovery efforts. The continued explosion of structural studies will lead to an increased understanding of the dynamics of protein interactions, which will, in turn, lead to better docking algorithms. The combination of structural information and greater computational power will also make more accurate predictions of protein—ligand interactions possible. References 1. Birkedal-Hansen H, Moore WGI, Bodden MK, Windsor LJ, Birkedal-Hansen B, DeCarlo A, Engler JA. Matrix metalloproteinases: a review. Crit. Rev. Oral. Biol. Med. 1993; 4:197–250. 2. Beckett RP, Davidson AH, Drummond AH, Huxley P, Whittaker M. Recent advances in matrix metalloproteinase inhibitor research. DDT 1996; 1:16–26. 3. Birkdell-Hansen H. Proteolytic remodeling of extracellular matrix. Cur. Op. Cell Biology 1995; 7:728–735. 4. Stocker W, Grams F, Baumann U, Gomis-Ruth F-X, McKay DB, Bode W. The metzincins topological and sequential relations between the astacins, adamalysins, serralysins, and matrixins (collagenases) define a superfamily of zinc peptidases. Protein Sci. 1995; 4:823–840.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_186.html [4/5/2004 5:04:45 PM]

Document

Page 187

5. Willenbrock F, Murphy G, Phillips IR, Brocklehurst K. The second zinc atom in the matrix metalloproteinase catalytic domain is absent in the full-length enzymes: a possible for the C-terminal domain. FEBS Lett. 1995; 358:189–192. 6. Lowry CL, McGeehan G, Levine H. Metal ion stabilization of the conformation of a recombinant 19kDa catalytic fragment of human fibroblast collagenase. Proteins 1992; 12:42–48. 7. Hodgson J. Remodeling MMPIs. Bio Technology 1995; 13:554–557. 8. Brown PD. Matrix metalloproteinase inhibitors: a novel class of anticancer agents. Adv. Enzyme Regul. 1995; 35:293–301. 9. Watson SA, Morris TM, Robinson G, Crimmin MJ, Brown PD, Hardcastle JD. Inhibition of organ invasion by matrix metalloproteinase inhibitor batimastat (BB-94) in two human colon carcinoma metastasis models. Cancer Res. 1995; 16:3629–3633. 10. Sledge GW Jr, Qulali M, Goulet R. Bone EA, Fife R. Effect of matrix metalloproteinase inhibitor batimastat on breast cancer regrowth and metastasis in athymic mice. J. Natl. Cancer Inst. 1995; 87:1546–1150. 11. Mitchel PG, Magna HA, Reeves LM, Lopresti-Morrow LL, Yocum SA, Rosner PJ, Geoghegam KF, Hambor JE. Cloning, expression, and type II collagenolytic activity of matrix metalloproteinase-13 from human osteoarthritic cartilage. J. Clin. Invest. 1996; 3:761–768. 12. Freije JMP, Diez-Ita I, Balbin M, Sanchez LM, Blaco R, Toliva J, Lopez-Otin C. Molecular cloning and expression of collagenase-3, a novel human matrix metalloproteinase produced by breast carcinomas. J. Biol. Chem. 1994; 269:16766–16773. 13. Rosenberg GA. Matrix metalloproteinases in brain injury. J. Neurotrauma 1995; 12:833–842. 14. Nomura H, Fujimoto N, Seiki M, Mai M, Okada Y. Enhanced production of matrix metalloporteinase and activation of matrix metalloproteinase 2 (gelatinase A) in human gastric carcinomas. Int. J. Cancer 1996; 69:9–16. 15. Itoh F, Yamamoto H, Hinoda Y, Imai K. Enhanced secretion and activation of matrilysin during malignant conversion of human colorectal epithelium and its relationship with invasive potential of colon cancer cells. Cancer 1996; 77:1717–1721. 16. Deb S, Gottschall PE. Increased production of matrix metalloproteinases in enriched astrocyte and mixed hippocampal cultures treated with beta-amyloid peptides. J. Neurochem. 1996; 66:1641–1647. 17. Crabbe T, Zucker S, Cockett MI, Willenbrock F, Tickle S, O'Connell JP, Scothern JM, Murphy G, Docherty AJP. Mutation of the active site glutamic acid of human gelatinase A: effects on latency, catalysis and the binding of tissue inhibitor of metalloproteinase-1. Biochemistry 1994; 33:6684–6690. http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_187.html (1 of 2) [4/5/2004 5:04:47 PM]

Document

18. Murphy G, Allan JA, Willenbrock, F, Cockett MI, Docherty AJP. The role of the C-terminal domain in collagenase and stromelysin specificity. J. Biol. Chem. 1992; 267:9612–9618. 19. Murphy G, Willenbrock F, Ward RV, Cockett MI, Eaton D, Docherty AJP. The C-terminal domain of 72 kDa gelatinase A is not required for catalysis, but is essential for membrane activation and modulates interactions with tissue inhibitors of metalloproteinase. J. Biochem. 1992; 328:637–641.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_187.html (2 of 2) [4/5/2004 5:04:47 PM]

Document

Page 188

20. Murphy G, Docherty AJP. Assessment of the role of fibronectin-like domain of gelatinase A by analysis of a deletion mutant. J. Biol. Chem. 1994; 269:6632–6636. 21. Shipley JM, Doyle GA, Fliszar CJ, Ye QZ, Johnson LL, Shapiro SD, Welgus HG, Senior RM. The structural basis for the elastolytic activity of the 92-kDa and 72-kDa gelatinases. Role of the fibronectin type II-like repeats. J. Biol. Chem. 1996; 271:4335–4341. 22. Spurlino, J, Smallwood AM, Carlton DC, Banks TM, Vavra KJ, Johnson JS, Cook EW, Falvo J, Wahl RC, Pulvino TA, Wendoloski JJ, Smith, DL. 1.56 Å structure of mature truncated human fibroblast collagenase. Proteins 1994; 19:98–109. 23. Lovejoy B, Cleasby A, Hassell AM, Longley K, Luther MA, Weigl D, McGeehan G, McElroy AB, Drewry D, Lambert MH, Jordan SR. Structure of the catalytic domain of fibroblast collagenase complexed with an inhibitor. Science 1994; 263:375–377. 24. Lovejoy B, Hassell AM, Luther MA, Weigl D, Jordan SR. Crystal structures of recombinant 19-kDa human fibroblast collagenase complexed to itself. Biochemistry 1994; 33:8207–8217. 25. Borkakoti N, Winkler FK, Williams DH, D'Arcy A, Broadhurst MJ, Brown PA, Johnson WH, Murray EJ. Structure of the catalytic domain of human fibroblast collagenase complexed with an inhibitor. Struct. Biol. 1994; 1:106–110. 26. Stams T, Spurlino JC, Smith DL, Wahl RC, Ho TF, Qoronfleh MW, Banks TM, Rubin B. Structure of human neutrophil collagenase reveals large S1' specificity pocket. Struct. Biol. 1994; 1:119–123. 27. Bode W, Reinemer P, Huber R, Kleine T, Schnierer S, Tschesche H. The X-ray crystal structure of the catalytic domain of human neutrophil collagenase inhibited by a substrate analogue reveals the essentials for catalysis and specificity. EMBO J. 1994; 13:1263–1269. 28. Browner MF, Smith WW, Castelhano AL. Matrilysin-inhibitor complexes: common themes among metalloproteinases. Biochemistry 1995; 34:6602–6610. 29. Wetmore DR, Hardman KD. Roles of the propeptide and metal ions in the folding and stability of the catalytic domain of stromelysin (matrix metalloproteinase 3). Biochemistry 1996; 35:6549–6558. 30. Dhanaraj V, Ye Q–Z, Johnson LL, Hupe DJ, Ortwine DF, Dunbar JB, Rubin JR, Pavvlovsky A, Humblet C and Blundell TL. X-ray structure of a hydroxamate inhibitor complex of stromelysin catalytic domain and its comparison with members of the zinc metalloproteinase superfamily. Structure 1996; 4:375–386. 31. Bernstein FC, Koetzle TF, Williams GJB, Meyer EF, Brice MD, Rodgers JR, Kennard O, Shimanouchi T, Tasumi M. The Protein Data Bank: a computer-based archival file for macromolecular structures. J. Mol. Biol. 1977; 112:535–542.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_188.html (1 of 2) [4/5/2004 5:04:49 PM]

Document

32. Carson M. Ribbon models of macromolecules. J. Mol. Graphics 1987; 5:103–106. 33. Connolly ML. The molecular surface package J. Mol. Graphics 1993; 11:139–141. 34. Li J, O'Hare MC, Skarzynski T, Lloyd LF, Curry VA, Clark IM, Bigg HF, Hazleman BL, Cawston TE, Blow DM. X-ray structure of a hydroxamate inhibitor complex of stromelysin catalytic domain and its comparison with members of the zinc metalloproteinase superfamily. Structure 1996; 4:375–386. 35. Grams F, Crimmin M, Hinnes L, Huxley P, Pieper M, Tschesche H, Bode W. Structure determination and analysis of human neutrophil collagenase complexed with a hydroxamate inhibitor. Biochemistry 1995; 34:14012–14020.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_188.html (2 of 2) [4/5/2004 5:04:49 PM]

Document

Page 189

36. Bode W, Reinemer P, Huber R, Kleine T, Schnierer S, Tschesche H. The X-ray crystal structure of the catalytic domain of human neutrophil collagenase inhibited by a substrate analogue reveals the essentials for catalysis and specificity. EMBO J. 1994; 13:1263–1269. 37. Schwartz MA, Van Wart HE. In: Ellis GP, Luscombe DK, eds. Progress in Medicinal Chemistry, Vol. 29. London: Elsevier Publishers, 1992: Chapter 8. 38. Johnson WH, Roberts NA, Borkakoti N. J. Enzyme Inhibition 1987; 2:1–22. 39. Wahl RC, Dunlop RP, Morgan BA. In: Bristol JA, ed. Annual Reports in Medicinal Chemistry. New York: Academic Press, 1990: Chapter 19. 40. Henderson B, Docherty AJP, Beeley NRA. Drugs of the Future 1990; 15:495–408. 41. Wahl RC, Pulvino TA, Mathiowetz AM, Ghose AK, Johnson JS, Delecki D, Cook ER, Gainer JA, Gowravaram MR, Tomczuk BE. Hydroxamate inhibitors of human gelatinase B (92kDa). Biorg. and Med. Chem. Lett. 1995; 5:349–352. 42. Gowravaram MR, Tomzcuk BE, Johnson JS, Delecki D, Cook ER, Ghose AK, Mathiowetz AM, Spurlino JC, Rubin B, Smith DL, Pulvino T, Wahl RC. Inhibition of matrix metalloproteinases by hydroxamates containing heteroatom-based modifications of the P1' group. J. Med. Chem. 1995; 38:2570–2581. 43. Netzel-Arnett S, Fields G, Birkdal-Hansen H, Avan Wart HE. Sequence specificities of human fibroblast and neutrophil collagenases. J. Biol. Chem. 1991; 206:6747–7855.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_189.html [4/5/2004 5:04:50 PM]

http://legacy.netlibrary.com/reader/message.asp?message=811&BookID=12640&FileName=Page_190.html

The requested page could not be found. Return to previous page

http://legacy.netlibrary.com/reader/message.asp?message=811&BookID=12640&FileName=Page_190.html [4/5/2004 5:04:52 PM]

Document

Page 191

7 Structure—Function Relationships in Hydroxysteroid Dehydrogenases Igor Tsigelny and Michael E. Baker University of California, San Diego, La Jolla, California I. Introduction Steroid hormones regulate a multitude of physiological processes in humans. Androgens and estrogens regulate sexual development and reproduction; glucocorticoids are important in the response to stress; vitamin D is important in bone growth; progestins are important for a viable fetus during pregnancy; mineralocorticoids regulate sodium and potassium balance to maintain normal blood pressure. Moreover, the growth of some breast and prostate tumors depends on steroids. With this multitude of medically important steroid-dependent actions, much research has gone into understanding their mode of action, with most of the effort concerned with the receptors that mediate the actions of steriods. A. High Blood Pressure and 11β-Hydroxysteroid Dehydrogenase It is only in the last decade that the role of hydroxysteroid dehydrogenases (Figure 1) in regulating the actions of steroids has been appreciated [1–6]. This mechanism for regulating steroid hormone action was uncovered in several laboratories studying various aspects of high blood pressure. One source was the study in the 1970s that identified the syndrome, Apparent Mineralocorticoid Excess (AME), a genetic disease that results in high blood pressure in children [7–10]. Also important is the work from laboratories investigating paradoxes in the mechanism of action of aldosterone in the kidney [1–4,9,10]. These studies identified 11β-hydroxysteroid dehydrogenase (11β-HSD) as a key enzyme.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_191.html [4/5/2004 5:04:54 PM]

Document

Page 192

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_192.html (1 of 2) [4/5/2004 5:04:59 PM]

Document

Figure 1 Reactions catalyzed by 11β-hydroxysteroid and 17β-hydroxysteroid dehydrogenases. (a) 11 β-hydroxysteroid dehydrogenase type 1, an NADPH-dependent enzyme, catalyzes the conversion of the inactive steroid, cortisone, to cortisol, which is the biologically active glucocorticoid. 11β-hydroxysteroid dehydrogenase type 2, an NAD+-dependent enzyme, catalyzes the reverse direction. (b) 17β-hydroxysteroid dehy-drogenase type 1, an NADPH-dependent enzyme, catalyzes the reduction of estrone to estradiol. Type 2, an NAD+-dependent enzyme, catalyzes the oxidation of estradiol to estrone. Type 3, an NADPH-dependent enzyme, catalyzes the reduction of androstene dione to testosterone. Type 4, an NAD+-dependent enzyme, catalyzes the oxidation of estradiol to estrone, and androstenediol to dehydroepiandrosterone.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_192.html (2 of 2) [4/5/2004 5:04:59 PM]

Document

Page 193

The enzyme 11β-HSD interconverts the active glucocorticoid cortisol and cortisone, an inactive metabolite (Figure 1a). By the oxidation of cortisol to cortisone, 11β-HSD prevents glucocorticoids from deleterious actions in certain cell types. For example, excess glucocorticoids in Leydig cells inhibit testosterone synthesis [3,11]. Expression of 11β-HSD in Leydig cells prevents this effect of glucocorticoids. In this way, 11β-HSD is important in androgen action. The enzyme 11β-HSD is also important in aldosterone action in the distal tubule of the kidney. Glucocorticoids have high affinity for the mineralocorticoid receptor [12] and can stimulate the mineralocorticoid response—uptake of sodium from urine—one effect of which is to increase blood pressure. Local expression of 11β-HSD in the distal tubule prevents this effect of glucocorticoids. The steroid aldosterone, which is not metabolized by 11β-HSD, can bind to the mineralocorticoid receptor and regulate sodium and potassium balance. Thus, 11β-HSD has an important role in regulating the biological actions of both glucocorticoids and mineralocorticoids. As would be expected, interference with 11β-HSD activity due to a mutation [13,14] or by an inhibitor such as licorice (Figure 2) [1–3,15] has a variety of physiological effects including high blood pressure due to mineralocorticoid actions of glucocorticoids in the kidney's distal tubule. Thus, studies to unravel genetic hypertension in children and the actions of aldosterone in the kidney yielded the general insight that, at specific times, altered expression of 11β-HSD in specific tissues is an important mechanism for regulating glucocorticoid, mineralocorticoid, and androgen action. A similar mechanism has been found for 17β-hydroxysteroid dehydrogenase (17β-HSD), the enzyme that regulates the concentrations of estradiol and testosterone in human [5,16,17] (Figure 1b). Genetics diseases associated with mutations in this enzyme lead to developmental abnormalities [18]. Enzymes that regulate the concentrations of retinoids [19] and prostaglandins [20] may also have a similar role [6]. B. Multiple Divergent 11β-Hydroxysteroid and 17β-Hydroxysteroid Dehydrogenases The cloning and sequencing of 11β-HSD [21–25] and 17β-HSD [16–18,26] revealed two 11β-HSDs and four 17β-HSDs with very different sequences

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_193.html [4/5/2004 5:05:01 PM]

Document

Page 194

Figure 2 Structure of licorice and carbenoxolone. Glycyrrhizic acid, a constituent of licorice extract, is found in the root of Glycyrrhiza glabra. The glycosidic group at C3 is cleaved by bacteria in the small intestine to form glycyrrhetinic acid, the compound that inhibits 11β-hydroxysteroid dehydrogenase. Carbenoxolone, a water soluble synthetic analog of glycyrrhetinic acid, is widely used to regulate 11β-HSD in vitro and in vivo.

(Figure 3, Table 1). This was surprising, as one would expect the two 11β-HSDs to be similar because they recognize the same substrates. Instead, the two 11β-HSDs have less than 20% sequence identity, after including gaps in the alignment (Table 1). The same degree of sequence divergence is found in the four 17β-HSDs [6]. This sequence divergence is reflected in differences in their catalytic properties. For example 11β-hydroxysteroid dehydrogenase-type 1 (11β-HSD-1) is an NADPH-dependent enzyme that converts cortisone to cortisol, and 11β-hydroxysteroid dehydrogenase-type 2 is an NAD+-dependent enzyme that oxidizes cortisol to cortisone. The enzyme 17βHSD-1 is an NADPH-dependent enzyme that converts estrone to estradiol, and 17βHSD-2 is an NAD+-dependent enzyme that oxidizes estradiol to estrone and testosterone to androstenedione.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_194.html (1 of 2) [4/5/2004 5:05:06 PM]

Document

Page 195

Figure 3 Alignment of 11β-and 17β-hydroxysteroid dehydrogenases. As seen in this Figure and Table 1, the sequences of the two 11β-HSDs and four 17β-HSDs are very divergent. Boxes denote sites where either 5 or 6 residues are conserved, which are likely to be functionally important.

There is considerable interest in understanding the structural bases for these differences because this information would be very useful in designing steroids and other compounds to selectively regulate the activity of one or more steroid dehydrogenases as a means of treating hormone-responsive diseases. There is precedent for this kind of treatment in the use of licorice extract from the root of the plant Glycyrrhiza glabra [15,27] to treat Addison's disease, http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_195.html (1 of 2) [4/5/2004 5:05:15 PM]

Document

Page 196

Figure 3 (Continued)

which is characterized by insufficient levels of cortisol. Licorice inhibits 11β-HSD-2, which raises the circulating levels of cortisol and provides some relief from the symptoms of Addison's disease. This use of licorice is an example of a plant-derived compound having important uses in mammalian steroid hormone physiology and indicates another reason why elucidation of the structure of steroid dehydrogenases is of medical interest. Plants contain a wide variety of compounds, many of which have been purified and had their structures determined; however, we don't know which of these compounds inhibit steroid

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_196.html (1 of 2) [4/5/2004 5:05:23 PM]

Document

Page 197 Table 1 Percent Identity Between Hydroxysteroid Dehydrogenase Sequences Shown in Figure 3

11β-HSD-2

11β-HSD-2

17β-HSD-2

17β-HSD-3

17β-HSD-1

11β-HSD-1

17β-HSD-4

0.00

46.1

20.3

28.6

20.6

18.1

0.0

21.1

21.2

17.7

18.9

0.0

19.1

19.6

17.5

21.1

20.4

0.0

17.1

17β-HSD-2 17β-HSD-3 17β-HSD-1 11β-HSD-1

0.0

17β-HSD-4

0.0

dehydrogenases. Knowledge of structure-activity relationships for the binding site on steroid dehydrogenases will be helpful in identifying novel compounds from plants and other sources that could be useful in regulating steroid dehydrogenases. At this time, we are just beginning to work on this ambitious goal. Structural information is limited. The 3-D structure of 17β-HSD type 1 has been determined [28], but without the steroid or cofactor in the binding site. Fortunately, 11β-HSD and 17β-HSD belong to a large family of enzymes that are called short-chain alcohol dehydrogenases [29–31] or sec-alcohol dehydrogenases [32]. The structures of dehydrogenase homologs in bacteria, plants, and animals have been determined [33–37] and we used them as templates for modeling 11β-HSD and 17β-HSD [38,39]. There also is information about the effects of mutations on catalytic activity in 11β-HSD-1 [40] and 17β-HSD-1 [41] and in homologs, especially for Drosophila alcohol dehydrogenase (ADH) [42–48]. Together they enable us to begin to understand the relationship between structure and function in hydroxysteroid dehydrogenases, as we discuss in this chapter. II. Methods A. Molecular Modeling

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_197.html (1 of 2) [4/5/2004 5:05:25 PM]

Document

Important for the validity of the models that we constructed is the evidence from models of other proteins indicating that two proteins can have as little as 20 to 25% sequence identity and still have very similar 3D structures, especially in α helices and β stands [49–52]. Variation is found in the loops and coiled structures. A relevant example for this chapter is the comparison of the tertiary structure of rat dihydropteridine reductase [33] and Streptomyces hydrogenans 20β-hydroxysteroid dehydrogenase [34]. As noted by Varughese et al. [33], despite less than 20% sequence identity between dihydropteridine reductase and

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_197.html (2 of 2) [4/5/2004 5:05:25 PM]

Document

Page 198

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_198.html (1 of 2) [4/5/2004 5:05:32 PM]

Document

Figure 4 Amino acids important in cofactor and catalysis in human 11b-hydroxysteroid dehydrogenase types 1 and 2. (a) 11b-HSD type 1. Preference of 11b-HSD type 1 for NADPH resides in lysine-44 and arginine-66, which have positively charged side chains that stabilize the binding of the 2'-phosphate on NADPH. These residues also counteract the repulsive interaction between glutamic acid 69 and the phosphate group. (b) 11b-HSD type 2. Preference of 11b-HSD type 2 for NAD+ is due to favorable bonds with aspartic acid-91, serine-92, and threonine-112. Moreover there is a coulombic repulsion between aspartic acid-91 and NADP+, which destabilizes binding of NADP+ 11b-HSD type 2 lacks a nearby amino acid with a positively charged side chain that could diminish the repulsive interaction between NADP+ and aspartic acid-91. Also shown are threonine residues that could hydrogen bond to nicotinamide's carboxamidemoiety.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_198.html (2 of 2) [4/5/2004 5:05:32 PM]

Document

Page 199

20β-hydroxysteroid dehydrogenase, the root mean square deviation for the two tertiary structures is 2 Å over 160 Cα carbon atoms. We aligned human 11β-HSD-1 [21] and 11β-HSD-2 [22] with S. hydrogenans 20β-hydroxysteroid dehydrogenase and Escherichia coli 7α-hydroxysteroid dehydrogenase [37] for 3D modeling. Human 11β-HSD has extra segments at the amino terminus and carboxyl terminus. Previously reported alignments [30,31,53] were used to find the core structure consisting of about 225 residues that are structurally similar to the template. The first 190 residues of the 255 residues are reasonably well conserved among the hydroxysteroid dehydrogenases. Alignment of the C-terminal 65 residues is less certain as this part contains gaps and insertions. Fortunately, the core 190 residues contains the catalytic site and the cofactor binding site. We also superimposed the two 11β-HSD structures on mouse carbonyl reductase [37]. The 11β-HSD 3D structures superimpose nicely on α helices E and F and other helices and strands that are important in binding of cofactor and substrate. Then, we extracted NADPH from carbonyl reductase and NAD+ from 7α-hydroxysteroid dehydrogenase and inserted the cofactors into the structures of the two 11β-HSDs. The α helix F in 17β-HSD-1 [16], 17β-HSD-2 [17], 17β-HSD-3 [18], and porcine 17β-HSD-4 [26] was constructed by modeling on α helix F in 20β-hydroxysteroid dehydrogenase. Comparisons with other 3D structures [33–37] have demonstrated that this α helix is highly conserved. The modeled dimers were not minimized as a dimer complex to avoid the artifactual adjustment of α helix F side chains. The Homology program (Biosym Technologies, Inc., 1995) was used to model a 255-residue segment of 11β-HSD and α helix F of 17β-HSD on the S. hydrogenans 20β-hydroxysteroid dehydrogenase template. To produce the final model (Figures 4 and 5) this program finds an optimal configuration of the residues when arranged in the template structure by minimizing unfavorable interactions between amino acid side chains. The side chains of each monomer were then minimized (1,000 iterations of the conjugate gradient) using the Discover program (Biosym Technologies Inc., 1995).

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_199.html [4/5/2004 5:05:34 PM]

Document

Page 200

Figure 5 Structure of α helix F interface of human 11β-hydroxysteroid dehydrogenase types 1 and 2. The α helix F part of the dimer interface on 11β-HSD-1 and -2 is shown along with side chains of the highly conserved tyrosine and lysine residues and other residues that are oriented into the cavity that binds substrate and nucleotide cofactor.

III. Results and Discussion A. NADPH Binding Site on 11β-Hydroxysteroid Dehydrogenase Types 1 and 2 Several lines of evidence—sequence analysis, mutagenesis studies, and the solved 3D structure of homologs of 11β-HSD—indicate that the nucleotide binding site in these enzymes has many similarities to that in other classes of dehydrogenases. For many dehydrogenases, the nucleotide binding domain

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_200.html (1 of 2) [4/5/2004 5:05:58 PM]

Document

Page 201

consists of a β strand α helix, β strand in a fold that provides a hydrophobic pocket for the adenosine monophosphate (AMP) part of the nucleotide cofactor [51,54,55]. In short-chain alcohol dehydrogenases, this βαβ fold is at the amino terminus. The turn between the first β strand and the α helix is a glycine-rich segment of the form Gly-X-X-X-Gly-X-Gly. This glycine-rich segment forms a hydrophobic pocket that allows close association of the AMP part of the cofactor. However, this glycine-rich segment has other functions in short-chain alcohol dehydrogenases. Tanaka et al. [37] and our 3D modeling [60] indicate that this glycine rich segment has an important role in cofactor specificity and binding of the nicotinamide moiety to 11β-HSD. B. 11β-HSD-1 Preference for NADPH Figure 4 shows our 3D model of human 11β-HSD types 1 and 2. These models identify residues important in preference of 11β-HSD-1 for NADPH and 11β-HSD-2 for NADH. In 11β-HSD type 1, lysine-44 and arginine-66 have favorable coulombic interactions with the 2'-phosphate on NADP+ that stabilize binding (Figure 4a). Moreover, their positively charged side chains compensate for the negative interaction between glutamic acid-69 and the 2'-phosphate group. Tanaka et al. [37] found a similar function for lysine-14 and arginine-39 in the preference of mouse carbonyl reductase for NADPH. C. 11β-HSD-2 Preference for NAD+ The 3D structure of 11β-HSD-2 shows that NAD+ has stabilizing interactions between the ribose hydroxyl and aspartic acid-91, serine-92, and threonine-112. Replacement of NAD+ with NADP+ reveals a coulombic repulsion between the 2'-phosphate group and aspartic acid-91. However, 11β-HSD type 2 lacks a nearby amino acid with a positively charged side chain that could compensate for the negative charge on aspartic acid-91. This explains the preference of 11β-HSD-2 for NAD+. D. Amino Acids Important in Binding the Nicotinamide Ring and Carboxamide Moiety Both 11β-HSD types 1 and 2 contain residues in the C-terminal half that interact with the nicotinamide ring and carboxamide moiety to limit rotations about the N-glycosidic bond. These intersections are important in positioning the cofactor for proS hydride transfer at C4. In 11β-HSD-1, cysteine-213 stabilizes the nicotinamide ring; threonine-220 and threonine-222 stabilize the carboxamide moiety. In 11β-HSD-2, there are more interactions: proline-262, phenylalanine-265, threonine-267, serine-

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_201.html [4/5/2004 5:06:00 PM]

Document

Page 202

269, and valine-270 are close to either the nicotinamide ring or the carboxamide moiety. In addition, the face of the side chain of phenylalanine-94 is below the nicotinamide ring and its carboxamide group. This latter interaction is unusual because phenylalanine-94 is between the two canonical glycine residues in the βαβ fold, which is usually thought of as interacting mainly with the AMP part of the cofactor. In 11β-HSD-2, there is an interesting configuration of amino acids with aromatic side chains that are below the nicotinamide ring and which provide a hydrophobic cushion for NAD+. E. Catalytic Site Comparison of 11β-HSD-1 with homologs identifies tyrosine-183 and lysine-187 as being highly conserved residues. Mutagenesis of these residues [40] and the homologous tyrosine and lysine in 17βHSD-1 and Drosophila alcohol dehydrogenase [44,45] shows that these residues are important for catalytic function. The 3D model of 11β-HSD-1 presented in Figure 4 shows that tyrosine-183 is 3.6 Å from the nicotinamide C4, where hydride transfer occurs. Similarly, in 11β-HSD-2, tyrosine-232 is 4 Å from C4 on NAD+. Their positions support the notion that tyrosine is the catalytically active residue. However, a problem with this model is that the pKa of tyrosine is about 10, which would make this residue a poor nucleophile at neutral pH. To resolve this problem for the homologous tyrosine in Drosophila alcohol dehydrogenase, Chen et al. [44] proposed that the pKa of tyrosine is lowered by a nearby positively charged lysine. The 3D structure of the two 11β-HSDs shows that lysine-187 and lysine-236 are close to the proposed catalytically active tyrosine residues, which supports the hypothesis of Chen et al. [44]. F. Dimer Interface and the Catalytic Site Most short-chain alcohol dehydrogenases are active as either dimers or tetramers. Analysis of rat dihydropteridine reductase by Varughese et al. [33] indicates that the dimer interface consists of α helix E and α helix F from each monomer arranged in a four α helix bundle, a structure in which the hydrophobic surfaces on the helices form a core that yields very stable structure in a wide variety of proteins [56–59]. A four-helix bundle also appears to stabilize S. hydrogenans 20β-hydroxysteroid dehydrogenase, a tetrameric enzyme [34]. The α helix F contains the conserved tyrosine and lysine residue, which adds a constraint to changes in the sequence of this helix. It has at least two functions: stabilizing the dimer and orienting tyrosine and lysine and other residues for optimal interaction with substrate and nucleotide cofactor. The role of a specific site on the outer hydrophobic surface of α helix F in dimerization was suggested recently when a Drosophila ADH mutant that does

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_202.html [4/5/2004 5:06:02 PM]

Document

Page 203

not form stable dimers was sequenced [48]. This ADH mutant has alanine-159 replaced with threonine. A 3D model of Drosophila ADH shows alanine-159 on the opposite surface of α helix F from tyrosine153 and lysine-157 [48]. Alanine-159 along with alanine-158 form a hydrophobic anchor that stabilizes the dimer interface. These two residues of ADH and the homologous residues in other short-chain alcohol dehydrogenases have been overlooked in sequence analyses because they are not absolutely conserved. In fact, at least five amino acids are found in these positions among the different sec-alcohol dehydrogenases. G. Dimer Interface in 11β-and 17β-Hydroxysteroid Dehydrogenases Because α helix F at the dimer interface also contains the catalytic tyrosine and the nearby lysine residue, any structural analysis of the catalytic site must also consider the structure of this part of the dimer interface. For this reason, we modeled α helix F on 11β-HSD-1 and -2 and 17β-HSD-1, -2, -3, and -4 to gain an insight into stabilizing interactions and how they may affect the catalytic site. H. Human 11β-HSD-1 Figure 5 shows the modeled structure for the α helix F interface in human 11β-HSD-1, in which phenylalanine-188 and alanine-189 form an anchor. Alanine-189 is 3.5 Å and 4.7 Å from alanine-189 and alanine-185, respectively, on the other subunit. The phenylalanine-188 side chain is 3.2 Å from glycine-192. There is a hydrogen bond between serine-185 and serine-196, which are 3.2 Å apart. Alanine-185 is 4.7 Å from phenylalanine-193. There also is a hydrophobic interaction between phenylalanine-193 and alanine-181, which are 3.9 Å apart. This web of interaction between side chains on the outer surface of α helix F on each subunit influences residues that have side chains oriented to the interior where catalysis occurs. Alanine-185, which is stabilized by interactions with phenylalanine-193, and serine-184, which interacts with serine-196, are between the conserved tyrosine-183 and lysine-187. Phenylalanine-193 is adjacent to phenylalanine194, which is positioned into the catalytic site. I. Human 11β-HSD-2 Figure 5 shows the α helix F interface of human 11β-HSD-2. Alanine-237 is about 3 Å from leucine241; alanine-238 is about 3.7 Å from both alanine-238 and threonine-234. Threonine-234 has a stabilizing hydrophobic interaction

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_203.html [4/5/2004 5:06:03 PM]

Document

Page 204

Figure 6 Structure of α helix F interface of mammalian 17β-hydroxysteroid dehydrogenases. The α helix F part of the dimer interface on 17β-hydroxysteroid dehydrogenases is shown along with side chains of the highly conserved tyrosine and lysine residues and three other residues that are oriented into the cavity that binds substrate and nucleotide cofactor. (a) Modeled structure of human 17β-hydroxysteroid dehydrogenase type 1. (b) Modeled structure of human 17β-hydroxysteroid dehydrogenase type 2. (c) Modeled structure of human 17β-hydroxysteroid dehydrogenase type 3. (d) Modeled structure of porcine 17β-hydroxysteroid dehydrogenase type 4.

with leucine-242, which is 4.2 Å distant. Threonine-245 is 4.2 Å from the Cα carbon of glycine-230.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_204.html (1 of 2) [4/5/2004 5:06:14 PM]

Document

The type 1 and type 2 enzymes preferentially catalyze the reduction of the 11-keto group and the oxidation of the 11-hydroxyl group, respectively, on glucocorticoids. The chemistry of the side chains on methionine-243 in the type-2

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_204.html (2 of 2) [4/5/2004 5:06:14 PM]

Document

Page 205

enzyme and of phenylalanine-194 on 11β-hydroxysteriod dehydrogenase type 1 is quite different; it may be important in the different catalytic properties of these two enzymes. J. Human 17β-HSD-1 Figure 6a shows the modeled α helix F interface in human 17β-hydroxysteroid dehydrogenase type 1 in which phenylalanine-160 and alanine-161 form an anchor. Both residues have important stabilizing interactions across the dimer interface. Alanine-161 is 4.1 Å from alanine-161 on the other subunit. Alanine-161 has a hydrophobic interaction with alanine-157, which is in the segment between the conserved tyrosine-155 and lysine-159. There is a hydrophobic

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_205.html [4/5/2004 5:06:24 PM]

Document

Page 206

interaction between alanine-157 and leucine-165, which are about 3.8 Å apart. Phenylalanine-160 is 4 Å from glycine-164. There also is a hydrogen bond between cysteine-156 and serine-168, which are 3.2 Å apart. This is an interesting structural property of residues in the segment between the conserved tyrosine and lysine residues: this segment is important in stabilizing dimers. This pattern is repeated in the other 11β- and 17β-hydroxysteroid dehydrogenases suggesting conservation of this stabilizing structure, although the residues are not as well conserved as the tyrosine and lysine. K. Human 17β-HSD-2 Figure 6b shows the modeled α helix F interface in human 17β-hydroxysteroid dehydrogenase type 2. Alanine-237 is 3 Å from the hydrophobic part of the side chain of methionine-241 on the other subunit. Methionine-241 is 3.2 Å from serine-234. Alanine-230 is 3.7 Å from phenylalanine-242 and 4.5 Å from valine-245. Alanine-238, the other anchoring residue, is 4.1 Å from alanine-238 on the other subunit. L. Human 17β-HSD-3 Figure 6c shows the modeled α helix F interface in human 17β-hydroxysteroid dehydrogenase-type 3. Alanine-203 is 3.1 Å from alanine-207. Phenylalanine-204 is 3.2 Å from the other phenylalanine-204 and alanine-200. These are the only stabilizing interactions that we find in our analysis. Human 17βhydroxys-teroid dehydrogenase type 3 has the weakest interactions across the α helix F interface among the four types of 17β-hydroxysteroid dehydrogenases. The conformation of this part of 17βhydroxysteroid dehydrogenase type 3 could change upon binding of substrate, leading to other stabilizing interactions. And, of course other parts of the protein may have intersubunit interactions that stabilize the dimer. Alternatively, the hydrophobic surface of α helix F may interact with another protein or a membrane surface, a potentially important regulatory mechanism that we discuss later in this paper. M. Pig 17β-HSD-4 Figure 6d shows the modeled α helix F interface in pig 17β-hydroxysteroid dehydrogenase type 4. Leucine-169 is 2.9 Å from glycine-173. Leucine-174 is 4.3 Å from alanine-162 and 3.3 Å from alanine166. There also is a hydrogen bond between serine-165 and serine-175, which are 3 Å apart. N. Prospects for the Application of Structure-Function Analysis of Steroid Dehydrogenases in Hormone Therapy In the last two years there has been impressive progress in understanding the structure of steroid dehydrogenases that are important in regulating blood pres

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_206.html [4/5/2004 5:06:25 PM]

Document

Page 207

sure and the actions of reproductive hormones. This progress has come from several directions. First, the cloning and sequencing the dehydrogenases that regulate the actions of aldosterone, cortisol, estradiol, and testosterone. Second, determination of the 3D structure of 17β-HSD-1 and several homologs. Analyses of their 3D structures confirm a general principle that structural similarity is much higher than sequence similarity. This supports proposed molecular models of medically important steroid dehydrogenases using the alignment of their sequences onto the templates of 3D structural homologs. Models of 11β-HSD-1 and -2 are beginning to reveal important properties about these enzymes. We now have a good picture of the structural basis for specificity for NADPH and NADH in 11β-HSD-1 and -2. With this information, we can now turn our attention to modeling cortisol in these two enzymes. This information will open up the possibility for developing analogs to regulate the actions of these two enzymes for use in regulating blood pressure and other physiological processes. The development of carbenoxolone, a water-soluble synthetic analog of glycyrrhetinic acid, shows that chemists can create compounds that have high affinity for 11β-HSD. The next task is to synthesize compounds that are specific for 11β-HSD-1 or 11β-HSD-2. Molecular modeling can contribute important information to solving this kind of problem. Similarly important information will come from the 3D models of 17β-HSD-1, -2, -3 and -4. These models will be useful in developing compounds to regulate estrogen and androgen action, which have important application in reproductive medicine and in treating estrogen-dependent breast tumors and androgen-dependent prostatic tumors. Many compounds in plants have estrogenic and androgenic activity; some of these compounds are likely to work via inhibition of one of the 17β-HSD enzymes [27]. Analogous to the development of carbenoxolone to regulate 11β-HSD, we can seek synthetic compounds that regulate specific types of 17β-HSD, which may be useful in reproductive medicine and in treating cancers. Considering the explosive pace of biomedical research and the new developments in computers for sophisticated structural analyses, the next few years promise to yield important advances in design of new hormone therapies based on the knowledge of the structure of steroid dehydrogenases. Acknowledgments We thank Drs. Tanaka, Nonaka, Nakanishi, Deyashiki, Hara, and Mitsui for providing us with the x-ray crystallographic coordinates of carbonyl reductase and 7α-hydroxysteroid dehydrogenase. The support of the Supercomputer Center of the University of California, San Diego is gratefully acknowledged.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_207.html [4/5/2004 5:06:27 PM]

Document

Page 208

References 1. Funder JW, Pearce PT, Smith R, Smith AI. Mineralocorticoid action: target tissue specificity is enzyme, not receptor, mediated. Science 242; 1988:583–586. 2. Edwards CRW, Stewart PM, Burt D, Brett L, McIntyre MA, Sutanto WS, De Kloet ER, Monder C. Localization of 11β-hydroxysteroid dehydrogenase-tissue specific protector for the mineralocorticoid receptor. Lancet 2; 1988:986–989. 3. Monder C. Corticosteroids, receptors, and the organ-specific functions of 11β-hydroxysteroid dehydrogenase. FASEB J5; 1991:3047–3054. 4. White PC. Disorders of aldosterone biosynthesis and action. New Eng J Med 331; 1994:250–258. 5. Andersson S. 17β-hydroxysteroid dehydrogenase: isozymes and mutations. J Endocrinol 146; 1995:197–200. 6. Baker ME. Unusual evolution of 11β- and 17β-hydroxysteroid and retinol dehydrogenases. Bioessays 18; 1996:63–70. 7. New MI, Levine LS, Biglieri EG, Pareira J, Ulick S. Evidence for an unidentified steroid in a child with apparent mineralocorticoid hypertension. J Clin Endocrinol Metab 44; 1977:924–933. 8. Ulick S, Levine LS, Gunczler P, Zanconato G, Ramirez LC, Rauh W, Rosler A, Bradlow HL, New MI. A syndrome of apparent mineralocorticoid excess associated with defects in the peripheral metabolism of cortisol. J Clin Endocrinol Metab 49; 1979:757–764. 9. Funder JW, Pearce PT, Myles K, Roy LP. Apparent mineralocorticoid excess, pseudohypoaldosteronism, and urinary electrolyte excretion: toward a redefinition of mineralocorticoid action. FASEB J 4; 1990:3234–3238. 10. Stewart PM, Edwards CRW. The cortisol-cortisone shuttle and hypertension. J Steroid Biochem Molec Biol 40; 1991:501–509. 11. Monder C. Comparative aspects of 11β-hydroxysteroid dehydrogenase. Testicular 11βhydroxysteroid dehydrogenase: development of a model for the mediation of Leydig cell function by corticosteroids. Steroids 59; 1994:69–73. 12. Arriza JL, Weinberger C, Cerelli G, Glaser TM, Handelin BL, Housman DE, Evans RM. Cloning of the human mineralocorticoid receptor complementary DNA: structural and functional kinship with the glucocorticoid receptor. Science 237; 1987:268–275.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_208.html (1 of 2) [4/5/2004 5:06:29 PM]

Document

13. Wilson RC, Krozowski ZS, Li K, Obeyesekere VR, Razzaghy-Azar M, Harbison MD, Wei JQ, Shackleton CHL, Funder JW, New MI. A mutation in the HSD11B2 gene in a family with apparent mineralocorticoid excess. J Clin Endocrinology Metab 80; 1995:2263–2266. 14. Mune T, Rogerson FM, Nikkila H, Agarwal AK, White PC. Human hypertension caused by mutations in the kidney isozyme of 11β-hydroxysteroid dehydrogenase. Nature Genetics 10; 1995:394–399. 15. Baker ME. Licorice and enzymes other than 11β-hydroxysteroid dehydrogenase. Steroids 59; 1994:136–141. 16. Peltoketo H, Isomaa V, Vihko R. Genomic organization and DNA sequences of human 17βhydroxysteroid dehydrogenase genes and flanking regions. Eur J Biochem 209; 1992:459–466. 17. Wu L, Einstein M, Geissler WM, Chan HK, Elliston KO, Andersson S. Expression cloning and characterization of human 17β-hydroxysteroid dehydrogenase type 2,

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_208.html (2 of 2) [4/5/2004 5:06:29 PM]

Document

Page 209

a microsomal enzyme possessing 20α-hydroxysteroid dehydrogenase activity. J Biol Chem 268; 1993:12964–12969. 18. Geissler WM, Davis DL, Wu L, Bradshaw KD, Patel S, Mendonca BB, Elliston KO, Wilson JD, Russell DW, Andersson S. Male pseudohermaphrodites, caused by mutations to testicular 17βhydroxysteroid dehydrogenase-3. Nature Genetics 7; 1994:34–39. 19. Napoli JL, Boerman MHEM, Chai X, Zhai Y, Fiorella PD. Enzymes and binding proteins affecting retinoic acid concentrations. J Ster Biochem Molec Biol 55; 1995:589–600. 20. Baker ME. Evolution of enzymatic regulation of prostaglandin action: novel connections to regulation of human sex and adrenal function, antibiotic synthesis and nitrogen fixation. Prostaglandins 42; 1991:391–407. 21. Tannin GM, Agarwal AK, Monder C, New MI, White PC. The human gene for 11β-hydroxysteroid dehydrogenase. J Biol Chem 266; 1991:16653–16658. 22. Albiston AL, Obeyesekere VR, Smith RE, Krozowski ZS. Cloning and tissue distribution of the human 11β-hydroxysteroid dehydrogenase type 2 enzyme. Mol Cell Endocrinol 105; 1994:R11–R17. 23. Agarwal AK, Mune T, Monder C, White PC. NAD+-dependent isoform of 11β-hydroxysteroid dehydrogenase. Cloning and characterization of cDNA from sheep kidney. J Biol Chem 269; 1994:25959–25962. 24. Naray-Fejes-Toth A, Fejes-Toth G. Expression cloning of the aldosterone target cell-specific 11βhydroxysteroid dehydrogenase from rabbit collecting duct cells. Endocrinology 136; 1995:2579–2586. 25. Cole TJ. Cloning of the mouse 11β-hydroxysteroid dehydrogenase type 2 gene: tissue specific expression and localization in distal convoluted tubules and collecting ducts of the kidney. Endocrinology 136; 1995:4693–4696. 26. Leenders F, Adamski J, Husen B, Thole, HH, Jungblut PW. Molecular cloning and amino acid sequence of the porcine 17β-estradiol dehydrogenase. Eur J Biochem 222; 1994:221–227. 27. Baker ME. Endocrine activity of plant-derived compounds: an evolutionary perspective. Proc Soc Exper Biol Med 208; 1995: 131–138. 28. Ghosh D, Pletnev VZ, Zhu D-W, Wawrkak Z, Duax WL, Pangborn W, Labrie F, Lin S–X. Structure of human estrogenic 17β-hydroxysteroid dehydrogenase at 2.20 Aº resolution. Structure 3; 1995:503–513.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_209.html (1 of 2) [4/5/2004 5:06:31 PM]

Document

29. Baker ME. Genealogy of regulation of human sex and adrenal function, prostaglandin action, snapdragon and petunia flower colors, antibiotics, and nitrogen fixation: functional diversity from two ancestral dehydrogenases. Steroids 56; 1991:354–360. 30. Persson B, Krook M, Jornvall H. Characteristics of short-chain alcohol dehydrogenases and related enzymes. Eur J Biochem 200; 1991:537–543. 31. Krozowski Z. 11β-hydroxysteroid dehydrogenase and the short chain alcohol dehydrogenase (SCAD) superfamily. Mol Cell Endocrinol 84; 1992:C25–C31. 32. Baker ME. Protochlorophyllide reductase is homologous to human carbonyl reductase and pig 20βhydroxysteroid dehydrogenase. Biochem J 300; 1994:605–607. 33. Varughese KI, Xuong NH, Kiefer PM, Matthews DA, Whiteley JM. Structural and mechanistic characteristics of dihydropteridine reductase: a member of the Tyr- (Xaa)3-Lys-containing family of reductases and dehydrogenases. proc Natl Acad Sci USA 91; 1994:5582–5586.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_209.html (2 of 2) [4/5/2004 5:06:31 PM]

Document

Page 210

34. Ghosh D, Wawrzak Z, Weeks CM, Duax WL, Erman M. The refined three-dimensional structure of 3α,20β-hydroxysteroid dehydrogenase and possible roles of the residues conserved in chart-chain dehydrogenases. Structure 2; 1994:629–640. 35. Dessen A, Quemard A, Blanchard JS, Jacobs Jr, WR, Sacchettini JC. Crystal structure and function of the isoniazid target for Mycobacterium tuberculosis. Science 267; 1995:1638–1641. 36. Rafferty JB, Simon JW, Baldock C, Artymiuk PJ, Stuitje AR, Slabas AR, Rice DW. Common themes in redox chemistry emerge from the X-ray structure of oilseed rape, Brassica napus, enoyl acyl carrier protein reductase. Structure 3; 1995:927–938. 37. Tanaka N, Nonaka T, Nakanishi M, Deyashiki Y, Hara A, Mitsui Y. Crystal structure of the ternary complex of mouse lung carbonyl reductase at 1.8 Å resolution: the structural origin of coenzyme specificity in the short-chain dehydrogenase/reductase family. Structure 4; 1996:33–45. 38. Tsigelny I, Baker ME. Structures stabilizing the dimer interface on human 11β-hydroxysteroid dehydrogenase-types 1 and 2 and human 15-hydroxyprostaglandin dehydrogenase and their homologs. Biochem Biophys Res Commun 217; 1995:859–868. 39. Tsigelny I, Baker ME. Structures important in mammalian 11β-and 17β-hydroxysteroid dehydrogenases. J Ster Biochem Molec Biol 55; 1995:589–600. 40. Obeid J, White PC. Tyr-179 and lys-183 are essential for enzymatic activity of 11β-hydroxysteroid dehydrogenase. Biochem Biophys Res Comm 188; 1992:222–227. 41. Puranen TJ, Poutanen MH, Peltoketo HE, Vihko PT, Vihko RK. Site-directed mutagenesis of the putative active site of human 17β-hydroxysteroid dehydrogenase type 1. Biochem J 304; 1994:289–293. 42. Chen Z, Lu L, Shirley M, Lee WR, Chang SH. Site-directed mutagenesis of glycine-14 and two “critical” cysteinyl residues in Drosophila alcohol dehydrogenase. Biochemistry 29; 1990:1112–1118. 43. Chen Z, Lin Z-G, Lee WR, Chang SH. Role of aspartic acid-38 in the cofactor specificity of Drosophila alcohol dehydrogenase. Eur J Biochem 202; 1991:263–267. 44. Chen Z, Jiang JC, Lin Z-G, Lee WR, Baker ME, Chang SH. Site-specific mutagenesis of Drosophila alcohol dehydrogenase: evidence for involvement of tyrosine-152 and lysine-156 in catalysis. Biochemistry 32; 1993:3342–3346. 45. Cols N, Marfany G, Atrian S, Gonzalez-Duarte R. Effect of site-directed mutagenesis on conserved positions of Drosophila alcohol dehydrogenase. FEBS Lett 319; 1993:90–94.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_210.html (1 of 2) [4/5/2004 5:06:33 PM]

Document

46. Ribas dePoplana L, Fothergill-Gilmore LA. The active site architecture of a short chain dehydrogenase defined by site-directed mutagenesis and structure modeling. Biochemistry 33; 1994:7047–7055. 47. Chen Z, Tsigelny I, Lee WR, Baker ME, Chang SH. Adding a positive charge at residue 46 of Drosophila alcohol dehydrogenase increases cofactor specificity for NADP+. FEBS Lett 356; 1994:81–85. 48. Chenevert S, Fossett N, Lee WR, Tsigelny I, Baker ME, Chang SH. Amino acids important in enzyme activity and dimer stability for Drosophila alcohol dehydrogenase. Biochem J 308; 1995:419–423. 49. Chothia C, Lesk AM. The relation between the divergence of sequence and structure in proteins. EMBO J 5; 1986:823–826.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_210.html (2 of 2) [4/5/2004 5:06:33 PM]

Document

Page 211

50. Greer J. Comparative modeling of homologous proteins. Methods Enzymol 202; 1991:239–252. 51. Branden C, Tooze J. Introduction to protein structure. New York: Garland Publishing, 1991. 52. Ring CS, Cohen FE. Modeling protein structures: construction and their applications. FASEB J 7; 1993:783–790. 53. Baker ME. Sequence analysis of steroid-and prostaglandin- metabolizing enzymes: application to understanding catalysis. Steroids 59; 1994:248–258. 54. Wierenga RK, De Maeyer MC, Hol WGJ. Interaction of pyrophosphate moieties with α-helixes in dinucleotide binding proteins. Biochemistry 24; 1985:1346–1357. 55. Wierenga RK, Terpstra PP, Hol WGJ. Prediction of the occurrence of the ADP-binding βαβ-fold in proteins, using an amino acid sequence fingerprint. J Mol Biol 187; 1986:101–107. 56. Weber P, Salemme FR. Structural and functional diversity in four-α-helical proteins. Nature 287; 1980:82–84. 57. Chou K-C, Maggiora GM, Nemethy G, Scheraga HA. Energetics of the structure of the four-α-helix bundle in proteins. Proc Natl Acad Sci USA 85; 1988:4295–4299. 58. Presnell SR, Cohen FE. The topological distribution of four-α-helical proteins. Proc Natl Acad Sci USA. 86; 1989:6592–6596. 59. Harris NL, Presnell SR, Cohen FE. Four helix bundle diversity in globular proteins. J Mol Biol 236; 1994:1356–1368.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_211.html [4/5/2004 5:06:34 PM]

http://legacy.netlibrary.com/reader/message.asp?message=811&BookID=12640&FileName=Page_212.html

The requested page could not be found. Return to previous page

http://legacy.netlibrary.com/reader/message.asp?message=811&BookID=12640&FileName=Page_212.html [4/5/2004 5:06:37 PM]

Document

Page 213

8 Design of ATP Competitive Specific Inhibitors of Protein Kinases Using Template Modeling Janusz M. Sowadski,* Charles A. Ellis,* Rolf Karlsson* University of California, San Diego, La Jolla, California Madhusudan Scripps Research Institute, La Jolla, California I. Protein Kinases and Diseases The protein kinase family encompasses more than three hundred members of critically important enzymes, each one with a specific role or function within the cell. These enzymes, ATPphosphotransferases, recognize target proteins and through the phosphorylation of specific sites either activate or deactivate a particular pathway of signal transduction. Many of these signaling pathways are associated with cell surface receptors, which are located in the membranes that surround cells. The difference between the families of protein kinases is that they have different targets and generally fall into two major classes: The serine/threonine protein kinases transfer a phosphate from ATP to a serine or threonine residue in the target protein. This class of enzymes are generally associated with cytoplasmic signaling events. The tyrosine protein kinases transfer phosphate from ATP to tyrosine residues in the target protein and are generally associated with receptors that become activated after binding a growth factor or other ligand. Protein kinases are significant targets for therapeutic drug development and have been implicated as the disease causing components of numerous tumor *Current

affiliation: Tufts University, Boston, Massachusetts.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_213.html [4/5/2004 5:06:38 PM]

Document

Page 214

viruses. Specifically, it is the deregulation of the activity of protein kinases that leads to disease by tumor viruses. The importance of this deregulation can be dramatically illustrated by the large number of viral oncogenes (or cancer causing genes) that encode structurally modified protein kinases. These deregulated enzymes are able to bypass the normal tightly regulated processes of growth control, leading to acute malignant transformation. These oncogenes are one of the first examples of the identification of disease-causing genes. Many of these viral genes have subsequently been implicated in human diseases. Malignant tissues also share the common characteristic of an acquired independence from controls. The receptor—for example, PDGF and EGFR—can be stimulated by a ligand coming either from the cell itself (autocrine) or from nearby tissues (paracrine). Regardless of the mechanism leading to receptor activity, the resulting kinase activity results in a cascade of signals that turn on cellular proliferation programs. Therefore, selective inhibition of receptor tyrosine kinase will block tyrosine kinase driven cell proliferation resulting in antitumor activity. In addition to cancer, a growing number of nonmalignant proliferative diseases, (e.g., psoriasis, atherosclerosis, restenosis, fibrosis, etc.) or inflammatory responses (e.g., septic shock, asthma, osteo and rheumatoid arthritis, etc.) involve dysfunctional signaling pathways. Successful development of drugs that target this class of enzymes will depend on the discovery of selective inhibitors designed for the appropriate protein kinase within the family. In the past several years there has been an explosion of structural studies within the protein kinase family [1–8]. These studies, initiated by the crystal structure of Protein Kinase A [9–12] (CAPK) have shown that all members of the protein kinase family fold into a uniform three-dimensional catalytic core. Yet this uniform three-dimensional fold exhibits both different surface charges and at least two major conformations. II. Protein Kinase Template The stereo view of the ribbon diagram of cAPK is presented in Figure 1a. The overall topology of the core extending from strand 1 through helix h, Figure 1b, is identical (except helix B) with the eight other structures of protein kinases determined to this point. Furthermore, Figure 2 presents an overall structural comparison of the catalytic cores of the five kinases, cAPK, CDK2, CDK2-CYCLIN, IR, and MAP. The N-terminal helix A, which is present only in the cAPK crystal structure, is anchored by myristic acid in the mammalian bovine heart of cAPK. Myristic acid inserts itself into the hydrophobic pocket of the lower lobe of the enzyme, which results in the structural ordering of helixA and Ser10[13], one of the four autophosphorylation sites.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_214.html [4/5/2004 5:06:40 PM]

Document

Page 215

Figure 1 Diagram of cAPK fold. (a) Stereo MOLSCRIPT diagram. (b) The key loops are as follows: phosphate anchor located between strand 1 and strand 2, catalytic loop located between strands 6 and 7, DFG motif located between strands 8 and 9, activation loop including P+1 site between strand 9 and helix F. Phosphorylation site Thr197 is indicated by a large circle, inhibitor PKI (5–24) is colored in dark, and the P site in the peptide is shown in the dark circle. This figure has been generated using MOLSCRIPT [27].

Following the connectivity diagram, helix A is connected to β-strand 1, then to the phosphate anchor encompassing signature motif Gly50XGly 52XXGly55. The β-strand 2 is followed by β-strand 3 carrying invariant Lys72. Three antiparallel beta strands create the unique fold of the nucleotide binding site of the protein kinase. The β-strands 3 is followed by helix B, which is present only in cAPK, helix C, and β-strands 4 and 5. Helix C shows the largest displacements among many different protein kinase structures and consists of invariant Glu91, which forms a salt bridge with Lys72. This salt bridge is absent in the inactive CDK-2 structure [1] but present in the crystal structure of the complex of CDK-2 and its activator-cyclin [7]. Displacement of helix C is perhaps most pronounced in the case of the insulin receptor tyrosine kinase structure (IRK) [3]. In PKA, Phe185 resides in the hydrophobic pocket formed by the hydrophobic residues of helix C (upper lobe) and Tyr164 (lower lobe). In the IRK crystal structure, the hydrophobic residues of helix C, which provide the pocket for invariant Phe185 (of DFG motif), no longer interact with

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_215.html [4/5/2004 5:06:47 PM]

Document

Page 216

Figure 1 (Continued)

this residue. The DFG motif in IRK occupies the ATP site, which blocks the access of ATP. The division between the upper and lower lobes of the enzyme is well defined by the two major conformations of the upper lobe observed in the crystal structures of cAPK. One conformation has been observed in the orthorhombic crystals of recombinant cAPK [9,10,14] and another in the cubic

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_216.html (1 of 2) [4/5/2004 5:06:54 PM]

Document

Page 217

Figure 2 The diagrams of the Cα trace of the five kinases, cAPK, CDK2, CDK2-CYCLIN, IR, MAP. The thin line represents crystallographically determined homologous regions of the five kinases (R. Karlsson and J.M. Sowadski, personal communication).

crystals of bovine heart mammalian cAPK [15,16]. A comparison between the two structures has shown that there is rotation of the upper lobe by 15 degrees and a translation of 1.9 Å in the mammalian structure, which results in the opening of the nucleotide binding cleft. The motion of this lobe, which includes His87, one of the ligands to Thr197 [15], indicates that the phosphorylation of this site will be important for conformational diversity of the upper lobe. This is confirmed by the varying degrees of displacement of the upper lobe of all structures of the inactive unphosphorylated protein kinases (see review [17]). The lower lobe of the enzyme starts with helix D, which is followed by helix E and β-strands 6 and 7. The catalytic loop connecting both strands consists of a critical set of residues with Tyr164 and Arg165 at the beginning of the loop. The Tyr164 residue forms a hydrogen bond with invariant Asp220. The Arg165 residue, which is present in a great majority of protein kinases, provides two hydrogen bonds to the oxygens of the phosphate of Thr197. Invariant Asp166 (the catalytic base) and Asn171 (the ligand to one of the metal sites) are also located within this loop.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_217.html (1 of 2) [4/5/2004 5:07:00 PM]

Document

Page 218

The catalytic loop is the region of divergence between Ser/Thr and Tyr kinases. In cAPK and all Ser/Thr Kinases, Lys168 interacts with the γ phosphate of ATP during catalysis [12]. The role of Lys is replaced by Arg [9] and the insulin receptor tyrosine kinase structure [3] shows Arg1136 in a similar position as Lys168 in the active site of cAPK. The catalytic loop and β-strand 7 are followed by β-strand 8 and a short DFG conserved motif. This conserved motif consists of invariant Asp184, a ligand to the metal site, and invariant Phe185. The DFG motif is followed by β-strand 9, an activation loop that includes Thr197. The activation loop differs considerably among unphosphorylated kinases, CDK-2, ERK-2, and insulin receptor kinase. Furthermore, two crystal structures, twitchin protein kinase [2] and phosphorylase kinase [5], both lack a phosphorylation site in this region. In the first structure, twitchin protein kinase, Thr197 and Arg165, are replaced by hydrophobic residues, Val6098 and Leu6062 respectively [2]. This was predicted using modeling [18] and subsequently confirmed by the three-dimensional structure. In the second structure, phosphorylase kinase, the Thr197 is replaced by Glu182, which interacts with Arg148 [5]. Hence, in both structures the regulatory function of the phosphorylation site is replaced by a stable scaffold secured by either hydrophobic or electrostatic interactions. Since it has been shown that this loop provides a stable template for PKI(5–24) binding in cAPK, the status of phosphorylation of the activation loop critically affects the substrate binding. This is demonstrated in c-Src, a homolog of the Rous Sarcoma virus oncogene by mutation of Arg385, which is predicted to interact with the phosphate of Tyr416, and results in loss of activity toward the exogenous substrate [19]. The activation loop is followed by a P+1 loop which accommodates the P+1 site of the substrate. The P+1 loop is followed by invariant Glu208, which forms a salt bridge with invariant Arg280. This conserved pair plays a structural role and as the structure of CK-1 [16] shows, it can be replaced by other charged residues that maintain the same fold of the lower lobe. The P+1 loop and Glu 208 are followed by helix F consisting of invariant Asp220, followed by helices G, H, and I. The helix J and the C-terminal tail of cAPK, which are absent in other protein kinases, undergo a large motion during the cleft opening. The opening of the cleft results in loss of hydrogen bonds provided by the γ phosphate of ATP and the peptide that would bridge the lower and upper lobe of the enzyme. The motion of the upper domain increases the accessibility of the ATP binding site and one can envision that in the “open” conformation ATP binds. Yet, in the “closed” conformation ATP and its γ phosphate are positioned for a nucleophilic attack on the substrate. The motion of this lobe—which includes His87, one of the ligands to Thr197 [15]—indicates that the phosphorylation of this site will be important for conformational diversity of the upper lobe. This is confirmed by the varying degrees of displacement of the upper lobe

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_218.html [4/5/2004 5:07:03 PM]

Document

Page 219

of all structures of the inactive unphosphorylated protein kinases (see review [17]). The various displacements of the conserved upper domain of the catalytic cores of various kinases documented by crystallographic work suggest that this is the important underlying mechanism of catalysis. Analysis of crystal contacts of various kinases is however required to define the extent of displacement due to the lattice forces. In the case of cAPK, the displacement as observed for mammalian cAPK in the cubic crystal form is due to the intermolecular interaction in the lattice [20]. Analysis of the two crystal structures of the cell cycle-controlling kinases clearly shows two binding modes of ATP. In the inactive state without cyclin, ATP binding of its triphosphate moieties is different from that in the active form with cyclin bound. The major difference is the re-arrangement upon cyclin binding of the conserved Lys33-Glu51 pair, which is responsible for the binding of the α and β phosphates of ATP. III. Crystallographic Analysis of Substrate Specificities of Individual Kinases The most important contribution of subsequent crystallographic studies has been the confirmation of the structural homology extending through the members of this family of enzymes. The crystal structures of CDK-2 [1], ERK-2 [5], twitchin [2], insulin receptor kinase [3], phosphorylase kinase, CK-1 [6], along with structure of calcium/calmodulin-dependent protein kinase I [8] provide solid proof for the structural conservation of the catalytic core in the family. This is further confirmed by the recent structure of the active complex of CDK2/cyclin, which shows that Lys33-Glu51 pair is at a distance of 3.0 Å [7] as predicted in the model of CDK-2 based on the cAPK structure [21]. The structure of the complex has also confirmed the binding of cyclin to helix C and to the upper lobe, demonstrating the mechanism of activation by cyclin that results in bridging the invariant residues into the common network of distances required by structural homology of the protein kinase catalytic core. The crystallographic analysis of the structural homology of protein kinases can now be carried out using structures of various kinases to find a common search model to be used in molecular replacement methods (J.M. Sowadski and R. Karlsson private communication). The structures of various kinases have been used as search models to solve the structure of the cAPK using cAPK diffraction data. The best search model consists of fragments of the catalytic core excluding the activation loop, inserts, and upper lobe due to rotational motion observed in each structure. A structure solution has been found for several protein kinases using this selected model as shown in Figure 2. One of the most critical aspects of this analysis is the presence of the structurally

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_219.html [4/5/2004 5:07:06 PM]

Document

Page 220

conserved substrate-binding cleft as observed in the crystal structure of cAPK:PKI complex (Figure la,b). This finding allows the charges within this cleft to be predicted using the amino acid sequence of any given kinase. In the analysis of the structural data of other protein kinases, it is noted that only cAPK has been crystallized with its specific peptide inhibitor. Nevertheless, three other structures of protein kinases compared with the structure of the cAPK-PKI complex provide substantial evidence for the conservation of the substrate binding cleft. The substrate binding cleft of the phosphorylase kinase structure has been analyzed in detail and it is clear that all amino acids of the known specific substrate can be built into the PKI model and all required corresponding charges can be found in the cleft of the phosphorylase kinase structure. In the CK-1 structure determined without a peptide, the requirement of the peptide specificity resides on the P-3 site, which has to be phosphorylated. An analysis of the surface charges of the cleft of the CK-1 structure reveals the exact correspondence of the residues required to interact with a phosphorylated substrate at this site. Finally, the tripeptide of the pseudosubstrate site of IRK consisting of the Asp-Tyr-Tyr motif has corresponding charges in the structure of the enzyme's substrate cleft and confirms the data obtained from the degenerated peptide library for the unique sequence motifs of nine tyrosine kinases [22]. It is becoming increasingly clear that the wealth of structural data of protein kinases with cAPK as a prototype provides evidence for two important features concerning substrate binding. First, the substrate binding cleft is structurally conserved and second, the surface charges of this cleft and hydrophobic cavities on the surface are very diverse and correspond to the specificity requirement of the substrate for individual protein kinases (see Figure la). It is now possible to use the structural conservation of the substrate binding cleft to predict the charges and hydrophobic residues of the cleft to define substrate specificities for individual kinases. IV. Crystallographic Analysis of the ATP Binding Site Reveals Distinct Differences Utilized for the Further Design of Specific Inhibitors The diagram elucidating detailed interactions of ATP with the enzyme is presented in Figure 3a,b. Six out of nine invariant residues of the catalytic core of protein kinase are involved in ATP binding and catalysis. The key residues that hold the β and γ phosphates in position are the phosphate anchor, the metal sites, and Lys168. The amides of the residues—Phe54, Gly55, and Ser53—are essential for the position of the γ and β phosphates. The metal site coordinated by invariant Asp184 is also sequestered by the β and γ phosphates and the metal

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_220.html [4/5/2004 5:07:08 PM]

Document

Page 221

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_221.html (1 of 2) [4/5/2004 5:07:18 PM]

Document

Figure 3 (a) Ternary complex of MnATP with the inhibitor peptide PKI (5–24). Glu121 and Val123 of the conserved linker region of the protein kinase catalytic core form the bidentate hydrogen bond with the 6-amino group and N1 nitrogen of the purine base. Thr183, non-conserved, forms a hydrogen bond with the N7 position of purine. 2'-OH of ribose interacts with the side chain of Glu127 of helix D and P-3 Arg of the specific inhibitor while the 3'-OH interacts with Glu170 of the catalytic loop. (b) The local environment of serine nucleophile at P site (left site) and local environment of phosphorylated P site serine (right side). The side chain of the catalytic base is at hydrogen bond distance from-OH of the Ser nucleophile which defines the conserved substrate binding P site.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_221.html (2 of 2) [4/5/2004 5:07:18 PM]

Document

Page 222

site coordinated by invariant Asn171 is also sequestered by the α and γ phosphates. The residue Lys168 is at hydrogen-bond distance from the γ phosphate. The postulated in-line mechanism of phosphotransfer in cAPK [23] can be examined through an analysis of the MnATP and PKI (5–24), Ser substrate peptide and ADP complexes [24]. A comparison between cAPK and IRK structures indicates that Ser versus Tyr specificity is obtained by displacement of the substrate binding site in such a way that the hydroxyl of nucleophiles of both Ser and Tyr fall in the same point in the active site facing corresponding catalytic bases Asp1132 and Asp166 for IRK and cAPK respectively. The purine base of ATP is anchored to the enzyme by three hydrogen bonds, two of them involve the 6amino group and N1 nitrogen, which interact with the backbone atoms of Glu121 and Val123. the 6amino group and N1 nitrogen of the purine base form the hydrogen bonds also in the structure of CK-1 and in the structure of phosphorylase kinase. Yet, in the structure of inactive CDK-2, the purine base forms only one hydrogen bond via its N6 position. While N7 nitrogen interacts directly in cAPK with the side chain of Thr183 in CK-1, this nitrogen interacts directly with Glu55 and Tyr59 via two hydrogen-bonded water molecules and in phosposhorylase kinase via one water molecule. Hence, the N7 nitrogen and its interaction with the enzyme is a region for potential modification of ATP competitive inhibitors. Ribose is held by both enzyme (Glu127 and Glu170) and inhibitor (P-3 Arg). While the side chain of Glu127 interacts with 2'-OH of ribose in cAPK, in CK-1 the 2'-OH interacts via two water molecules with Ser91 and Asp94. In ERK-2, 2'-OH interacts with Asp109. This region has been utilized to design ATP-based specific inhibitors by modifications of the ATP-competitive nonspecific inhibitor staurosporine [25], see Figure 4. In the model cAPK with bound staurosporine inhibitor, the lactam amide group of the inhibitor functions as a bidentate hydrogen bond donor-acceptor

Figure 4 Chemical structures of staurosporine inhibitor, left, and CGP52411.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_222.html [4/5/2004 5:07:22 PM]

Document

Page 223

Figure 5 (a) Staurosporine molecule docked in the ATP binding site of PKA. hydrogen bonds anchoring staurosporine molecule in the active site of PKA consists of the 6-amino group and N1 nitrogen and carbonyl of Glu121 and the amide hydrogen atom of Val123. This bidentate hydrogen bond formation has been observed in all complexes of protein kinases and ATP solved so far. (b) Inhibitor of CGP 52411 and ATP docked on the active site of PKA. Residues of Glu127 and Glu170 are also shown and these are not conserved in the EGFR kinase.

(Figure 5a). This key observation is supported by chemical data of lactam amide derivatives, which provide a plausible model of staurosporine inhibition. This is in the protonated boat-type conformation found to fit in the ATP binding cleft with minimal steric hindrance. In this model, the 4-amino group forms hydrogen bonds with the backbone carbonyl of Glu170 and the carboxylate group of Glu127 of cAPK. A model of the EGFR kinase shows that Glu127 and 170 are replaced by Cys and Arg, respectively (Figure 5b). Replacement of Glu127 by Cys is critical according to the model and explains the several-fold decrease of potency of staurosporine inhibitor toward the EGFR kinase (IC50=630 nM

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_223.html (1 of 2) [4/5/2004 5:07:30 PM]

Document

Page 224

Figure 5 (Continued)

versus IC50=15 nM for cAPK). Yet enhancement of specificity utilizing the inhibitor CGP52411 (Figure 4), whose selectivity originates in the occupancy by one of the anilino moieties of the inhibitor in the region of the enzyme cleft that normally binds the ribose ring of ATP, is considerable. The inhibitor CGP52411 inhibits the EGFR tyrosine kinase with an IC50 value of 300 nM while it is less active by at least two orders of magnitude on a panel of protein kinases including cAPK, phosphorylase kinase, casein kinase, protein kinase C (most isoforms), and v-abl, c-lyn, c-fgr tyrosine kinases. Hence, the three-dimensional model of EGFR tyrosine kinase rationalizes the specificity of the CGP52411 inhibitor. This model suggests that analysis of the putative regions of the ribose binding of ATP in other kinases through template modeling would provide the required chemical modification of pharmacophores to enhance their selectivity. In modeling, the use of the bidendate hydrogen bonding provided by the linker region (Figure 3a) of the conserved

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_224.html [4/5/2004 5:07:34 PM]

Document

Page 225

protein kinase core is essential. Yet to predict the region of the structure of the inhibitor that will form these strong Watson-Crick hydrogen bonds remains difficult. The use of the 6-amino group and N1 nitrogen of the purine base to model the pharmacophore, as seen with staurosporine, is not possible in two other adenine-type inhibitors. Both isopentenyl adenine, a nonspecific inhibitor of protein kinases, and olomoucin, a more specific inhibitor of Ser/Thr protein kinases, are modified only at the 6-amino group position. Thus, bidentate hydrogen-bond formation as seen in the ATP purine base is not possible. Furthermore, there are inhibitors that do not contain the chemical structure of adenine, for example des-chloro-flavopyridol, a potent inhibitor of cdc-2 cell cycle kinase. In crystallographic analysis of the binding of three inhibitors, olomoucin (OLO), isopentenyl adenine (ISO) and des-chloro-flavopyridol (DFP) to inactive CDK-2 cell cycle protein kinase, Kim and coworkers [26] have provided additional insight into the binding of adenine-and nonadenine-based inhibitors. Inhibitors with purine rings (OLO and ISO) bind in relatively the same area of the binding cleft as the adenine ring of ATP. Relative orientation of each purine ring with respect to the protein is different for all three ligands. This is most likely due to the fact that the 6-amino group of adenine in ATP is replaced by an isopentenylamino group in ISO and by the bulky benzylamino group in OLO. In the case of the third inhibitor, which is not an adenine derivative, the benzopyran ring occupies approximately the same region as the purine ring of ATP. The two ring systems overlap in the same plane but benzopyran is rotated about 60 degrees relative to the adenine of ATP. In this orientation, two strong bidendate hydrogen bond are formed with the oxygens in the 4th and 5th positions of the inhibitor. Furthermore, these bonds are the same ones formed by the 6-amino group and N1 nitrogen of the adenine ring. Crystallographic analysis has shown that both the substrate and ATP-binding clefts are structurally conserved yet differ in the surface charges between individual protein kinases. The structural template of the protein kinase family as discovered in the structure solution of cAPK predicts these differences. Template modeling provides a rational basis for the design of specific inhibitors for protein kinases based on the ATP binding site [25]. This is the first significant step in the design of specific inhibitors targeted at the ATP site. Recent work has now shown the conformational diversity of inhibitors binding in the interdomain ATPbinding cleft [26]. Although the residues of the protein kinase catalytic core that form the bidentate donor-acceptor bond with inhibitors are identical throughout different structures, the residues of the inhibitors vary greatly. All inhibitors use this common bidentate bond yet the specificity lies in several other bonds formed between the inhibitor and specific regions of the individual protein kinases. Furthermore, it is difficult to model

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_225.html [4/5/2004 5:07:35 PM]

Document

Page 226

the bidentate bond-forming residues of the inhibitor and it is through protein crystallography that these are determined. Template modeling can then be used to determine the specific surface charges of the modeled kinase that can be exploited by the inhibitor for its specificity. Modeling and focused combinatorial chemistry are the tools to achieve the goal of inhibitor specificity. Acknowledgments This work was supported by CTR grant 4237. We thank Drs. Furet, P. Traxler, and N. Lydon of Ciba for their drawings presented in Figure 5. We also thank Dr. Nikola P. Pavletich for the coordinates of the CDK2-CYCLIN complex used in Figure 2. References 1. De Bondt HL, Rosenblatt J, Jancarik J, Jones HD, Morgan DO, Kim SH. Crystal structure of cyclindependent kinase2. Nature 1993; 363:595–602. 2. Hu S-H, Parker MW, Lei JY, Wilce MCJ, Benian GM, Kemp BE. Insights into autoregulation from the crystal structure of twitchin kinase. Nature 1994; 369:581–584. 3. Hubbard SR, Wei L, Ellis L, Hendrickson WA. Crystal structure of the tyrosine kinase domain of the human insulin receptor. Nature 1994; 372:746–754. 4. Zhang F, Strand A, Robbins D, Cobb MH, Goldsmith EJ. Atomic structure of the MAP kinase ERK2 at 2.3 Å resolution. Nature 1994; 367:704–710. 5. Owen DJ, Noble MEM, Garman EF, Papageorgiou AC, Johnson LN. Two structures of the catalytic domain of phosphorylase kinase; an active protein kinase complexed with substrate analogue and product. Structure 1995; 3:467–482. 6. Xu R-M, Carmel G, Sweet RM, Kuret J, Cheng X. Crystal structure of casein kinase-1, a phosphatedirected protein kinase. The EMBO Journal 1995; 14:1015–1023. 7. Jeffrey PD, Russo AA, Polyak K, Gibbs E, Hurwitz J, Massague J, Pavletich NP. Mechanism of CDK activation revealed by the structure of a cyclinA-CDK2 complex. Nature 1995; 376:313–320. 8. Goldberg J, Nairn AC, Kuryian J. Structural basis for the autoinhibition of calcium/calmodulindependent protein kinase I. Cell 1996; 84:875–887. 9. Knighton DR, Zheng J-H, Ten Eyck LF, Xuong N-H, Taylor SS, Sowadski JM. Structure of a peptide inhibitor bound to the catalytic subunit of cyclic adenosine monophosphate-dependent protein kinase. Science 1991; 253:414–420.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_226.html (1 of 2) [4/5/2004 5:07:38 PM]

Document

10. Knighton DR, Zheng J-H, Ten Eyck LF, Ashford VA, Xuong N-H, Taylor SS, Sowadski JM. Crystal structure of the catalytic subunit of cyclic adenosine monophosphate dependent protein kinase. Science 1991; 253:407–414. 11. Zheng J-H, Trafny EA, Kninghton DR, Xuong N-H, Taylor SS, Ten Eyck LF, Sowadski JM. 2.2Å refined crystal structure of the catalytic subunit of cAMP-dependent protein kinase complexed with MnATP and a peptide inhibitor. Acta Cryst 1993; D49:362–365.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_226.html (2 of 2) [4/5/2004 5:07:38 PM]

Document

Page 227

12. Zheng J-H, Knighton DR, Ten Eyck LF, Karlsson R, Xuong N-H, Taylor SS, Sowadski JM. Crystal structure of the catalytic subunit of cAMP-dependent protein kinase complexed with MgATP and peptide inhibitor. Biochemistry 1993; 32:2154–2161. 13. Sowadski JM, Ellis C, Madhusudan. Detergent binding to unmyristylated protein kinase A—structural implications for the role of myristate. Journal of Bioenergetics and Biomembranes 1996; 28:7–12. 14. Knighton DR, Bell S, Xuong N-H, Ten Eyck LF, Taylor SS, Sowadski JM. 2.0Å refined crystal structure of catalytic subunit of cAMP-dependent protein kinase complexed with a peptide inhibitor and detergent. Acta Cryst 1993; D49:357–361. 15. Karlsson RF, Zheng J-H, Xuong N-H, Taylor SS, Sowadski JM. Crystal structure of the mammalian catalytic subunit of cAMP-dependent protein kinase and an inhibitor peptide displays an open conformation. Acta Cryst 1993; D49:381–388. 16. Zheng J-H, Knighton DR, Xuong N-H, Taylor SS, Sowadski JM, Ten Eyck LF. Crystal structures of the myristylated catalytic subunit of cAMP-dependent protein kinase reveal open and closed conformations. Proteins Science 1993; 2:1559–1573. 17. Taylor SS, Radzio-Andzelm E. Three protein kinase structures define a common motif. Structure 1994; 2:345–355. 18. Knighton DR, Pearson RB, Sowadski JM, Means AR, Ten Eyck LF, Taylor SS, Kemp BE. Structural basis of the intrasteric regulation of myosin light chain kinase. Science 1992; 258:130–135. 19. Senften M, Schenker G, Sowadski JM, Ballmer-Hofer K. Catalytic activity and transformation potential of v-Src require arginine 385 in the substrate binding pocket. Oncogene 1995; 10:199–203. 20. Karlsson RF, Madhusudan Taylor SS, Sowadski JM. Intermolecular contacts in various crystal forms related to the open and closed conformational states of the catalytic subunit of cAMP-dependent protein kinase. Acta Cryst 1994; D50:657–662. 21. Marcote MJ, Knighton DR, Basi G, Sowadski JM, Brambilla P, Draetta G, Taylor SS. A threedimensional model of the cdc2 protein kinase: identification of cyclin and suc1 binding regions. Molecular and Cellular Biology 1993; 13:5122–5133. 22. Songyang, Z et al., Catalytic specificity of protein tyrosine kinase is critical for selective signalling. Nature 1991; 373:536–539. 23. Ho M-F, Bramson HN, Hansen DE, Knowles JR, Kaiser ET. Stereochemical course of the phospho group transfer catalyzed by cAMP-dependent protein kinase. J Am Chem Soc 1988; 110:2680–2681.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_227.html (1 of 2) [4/5/2004 5:07:40 PM]

Document

24. Madhusudan Xuong N-H, Ten Eyck LF, Taylor SS, Sowadski JM. cAMP-dependent protein kinase: Crystallographic insights into substrate recognition and phosphotransfer. Protein Science 1994; 3:176–187. 25. Furet P, Caravattti G, Priestle J, Sowadski J, Trinks U, Traxler P. Modeling study of protein kinase inhibitors: Binding mode of staurosporine-origin of the selectivity of CGP 52 411. J Comp Aid Mol Design 1995; 9:465–472. 26. Azevedo WF Jr, Mueller-Diechmann H-J, Schulze-Gahmen U, Worland PJ, Sausville E, Kim S-H. Proc Natl Acad Sci 1996; In press. 27. Kraulis PJ, MOLSCRIPT—A program to produce both detailed and schematic plots of protein structures. Journal of Applied Crystallography 1991; 24:946–950.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_227.html (2 of 2) [4/5/2004 5:07:40 PM]

http://legacy.netlibrary.com/reader/message.asp?message=811&BookID=12640&FileName=Page_228.html

The requested page could not be found. Return to previous page

http://legacy.netlibrary.com/reader/message.asp?message=811&BookID=12640&FileName=Page_228.html [4/5/2004 5:07:46 PM]

Document

Page 229

9 Structural Studies of Aldose Reductase Inhibition David K. Wilson and Florante A. Quiocho Baylor College of Medicine, Houston, Texas J. Mark Petrash Washington University School of Medicine, St. Louis, Missouri I. Introduction Aldose reductase (ALR2; EC 1.1.1.21) is an ~36 kDa enzyme that catalyzes the reduction of a wide range of carbonyl-containing compounds to their corresponding alcohols. It is a member of an extensive aldo-keto oxidoreductase enzyme family, a collection of structurally similar proteins expressed in both animals and plants. Most members of the enzyme family possess similarities in molecular mass, pH optimum, coenzyme dependence, and demonstrate overlapping specificity for many substrates and inhibitors. While no essential physiological function has been established for ALR2, extensive experimental evidence suggests that it plays an important role in the development of diabetic complications affecting the visual, nervous, and renal systems [1]. The linkage between ALR2 and pathogenesis of diabetic complications lies in the polyol pathway of glucose metabolism (Figure 1). In hyperglycemic tissues such as in diabetes mellitus, the capacity of hexokinase to shunt glucose to glycolysis and other major pathways of glucose metabolism is exceeded. Consequently, enhanced flux of glucose through the polyol pathway occurs. The enzyme ALR2 catalyzes the first step in this pathway, producing sorbitol, an active osmolyte. The polyol pathway is completed by the NAD+-dependent oxidation of sorbitol to fructose, mediated by sorbitol dehydrogenase. Extensive evidence exists to suggest a linkage between the pathogenesis of diabetic complications and enhanced glucose metabolism via the polyol pathway. The polyol pathway functions in all tissues susceptible to clinically

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_229.html [4/5/2004 5:07:49 PM]

Document

Page 230

Figure 1 Schematic of the polyol pathway showing the NADPH-dependent reduction of open chain D-glucose to sorbitol, which is catalyzed by ALR2. This step is followed by the NAD+-dependent oxidation of sorbitol by sorbitol dehydrogenase to yield D-fructose.

significant diabetic complications. Transgenic animals overexpressing ALR2 in target tissues of diabetic complications are more prone to development of experimentally induced diabetic complications [2,3]. The most extensive body of evidence linking ALR2 to the pathogenesis of diabetic complications comes from numerous successes in the treatment of experimental animals with a variety of ALR2 inhibitors (ARI) [4]. Many of these studies demonstrated that ARIs substantially delay or in some cases prevent the onset of complications. Clinical trials of ARIs have yielded encouraging results in alleviating painful symptoms of diabetic complications. However, unacceptable side effects related to toxicity or inadequate pharmocokinetic profiles have rendered most of the drug candidates undesirable. Nevertheless, several ARIs are commercially available in some countries and more appear to be in the pipeline. The therapeutic rationale for treatment of human diabetics with ARIs to delay or prevent onset of diabetic complications is compelling. Animal models with experimentally induced hyperglycemia develop complications that are morphologically and functionally similar to that seen in the human diabetic patient. Many structurally

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_230.html (1 of 2) [4/5/2004 5:07:56 PM]

Document

Page 231

diverse ARIs have been shown to substantially delay or completely prevent the onset of such complications in experimental animal models. While some studies indicate that ALR2 may play a functional role in osmotic homeostasis in the kidney, evidence from animal studies suggests that it is metabolically dispensible. Long-term complications exact a terrible toll of morbidity and mortality on patients with diabetes mellitus. For example, patients with diabetes have about a 25-fold increased risk for becoming blind over that of the general population. Diabetic retinopathy is one of the most common causes of visual loss and accounts for about 12% of new cases of blindness each year in the United States alone [5]. II. Drug Design Prior to Structural Data Many inhibitors have been developed over the past two decades without the advantage of a structural understanding of the enzyme [4,6,7]. Significant improvement has been made since the discovery of the first such orally active compound to show in vivo activity, alrestatin (Figure 2), which had an IC50 in the low micromolar range [8]. Many high-affinity inhibitors with IC50s in the low nanomolar range are now under study. Recent drug-design efforts have yielded compounds usually with one of two chemical motifs: carboxylates or spirohydantoins. A number of these compounds such as tolrestat [9], ponalrestat [10], epalrestat [11], sorbinil [12], and zopolrestat [13] have progressed to the point of clinical trials. Unfortunately, clinical ineffectiveness and/or unacceptable side effects have limited the usefulness of most of those that had been shown to be effective in vitro. The latter problems may be associated with a lack of specificity since many aldose reductase inhibitors inhibit both ALR2 and aldehyde reductase [14]. For this reason, ALR2 as well as the other members of the aldo-keto reductase family have been the subject of crystallographic studies with the hope of determining a structural basis for inhibitor specificity and ultimately to provide a basis for enhancing binding affinity. III. Structural Studies of Aldose Reductase The first crystal structures available for ALR2 were those of the porcine form complexed with the NADPH analog 2'-monophosphoadenosine-5'-diphosphoribose [15] and the human enzyme complexed with the NADPH cofactor [16]. Further studies have been conducted on mutants of the human enzyme [17] and ternary complexes of the human enzyme with an inhibitor [18]. All of these

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_231.html [4/5/2004 5:08:00 PM]

Document

Page 232

Figure 2 A number of ALR2 inhibitors that have entered clinical trails.

structures show the protein to fold into a (β/α)8 barrel (Figure 3). This fold has emerged as the most common enzyme motif [19] although most of the proteins adopting this structure share no sequence homology. The ALR2 enzyme is, however, the first NAD(P)H binding protein to adopt this fold. It contains an extra β hairpin preceding the first β strand, which caps the N-terminal end of the barrel. It also has two helices that are not part of the regular barrel. One precedes α7 and the other follows α8. A. Cofactor Binding The NADPH cofactor is bound in an extended conformation across the C-terminal end of the β barrel. The catalytically active nicotinamide moiety is located

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_232.html (1 of 2) [4/5/2004 5:08:09 PM]

Document

Page 233

Figure 3 Cα trace of the ALR2 holoenzyme looking down the (β/α)8 barrel. The NADPH cofactor is seen bound across the carboxy-terminal end of the β-barrel with the active nicotinamide moiety in the center. Figure produced using the MOLSCRIPT program [48].

at the center of the barrel while the adenosine extends away to bind between α7 and α8. A belt composed of residues 213 to 227 folds over the pyrophosphate of the NADPH to sequester a large part of the cofactor from the solvent. It is fastened to the other side of the NADPH binding site via Asp216 on the loop that forms bifurcated salt links with Lys21 and Lys262. The dominant interactions holding the coenzyme in place are directional hydrogen bonds and salt links from positively charged side chains to the phosphates. The interaction between the 2' phosphate on the NADPH and the side chains from Lys262 and Arg268 account for the enzyme's preference for NADPH over NADH. Earlier biochemical studies had shown that it is the 4-pro-R hydride that is transferred from the nicotinamide to the substrate [20]. This is ensured by a hydrogen-bonding network using side chains from Ser159 and Asn160 and the main chain of Gln183 to orient the amide. It is also determined by the stacking interactions with Tyr209, which is adjacent to the 4-pro-S side of the nicotinamide. B. Mechanism The catalytic site was unambiguously identified using the location of the nicotinamide moiety of the NADPH cofactor in the holoenzyme structure. The region surrounding the catalytic site is a 12-Å deep groove that measures

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_233.html (1 of 2) [4/5/2004 5:08:15 PM]

Document

Page 234

approximately 7 by 13 Å and is lined primarily with hydrophobic side chains. This is entirely consistent with earlier experiments showing that the enzyme has a marked preference for lipophilic substrates versus polar substrates such as sugars [21]. The catalytic site also suggested a model for the enzyme's chemical mechanism, which is substantially similar to most other NAD(P)H-dependent oxido reductases. Upon binding of the reduced cofactor, the enzyme is able to form a ternary complex with the substrate. The pro-R hydride from the C-4 of the nicotinamide is transferred to the carbonyl carbon of the substrate, which in turn causes the carbonyl oxygen to abstract a proton from a general acid, which is presumably located on the protein, to form the alcoholic product. Three proton-donating side chains are located within 6 Å of the C-4 atom in the NADPH cofactor that could potentially fulfill this role: Tyr48, His110, and Cys298. Since it is not conserved in other members of the aldo-keto reductase family that exhibit enzymatic activity (Figure 8) Cys298 was unlikely as a candidate proton donor. The histidine is surrounded by several hydrophobic residues including Val47, Trp79, and Trp111, which would serve to lower the pKa of the side chain, making it less effective as a proton donor at physiological pHs. The tryrosine, which ordinarily has a pKa of approximately 11 engages in an interaction with the charged ammonium group of Lys77, which in turn charge-pairs with Asp43. This network serves to depress the pKa of the phenolic oxygen, increasing the exchangability of the proton. Subsequent activity studies involving site-directed mutants support this model [17,22]. The Tyr48 rarrow.gif Phe mutation shows a complete lack of activity while the Asp43 rarrow.gif Asn, Lys77 rarrow.gif Met, His110 rarrow.gif Asn, and Cys298 rarrow.gif Ser showed losses in catalytic efficiency of approximately 100-, 1000-, 106-, and 10-fold respectively when compared with the wildtype enzyme. These results correlate well with the functions predicted for each residue with the exception of the histidine. The structure of the ALR2 holoenzyme showed that the catalytic site was situated atop the nicotinamide moiety of the NADPH cofactor. The substrate binding site, which would determine the enzyme's specificity and also presumably bind inhibitors, appeared to be composed of a deep cleft (Figures 3 and 4). It extended away from the catalytic site towards the loop composed of residues between β4 and α4 and the last 20 residues of the carboxy-terminal meander. This hypothesis was supported by the appearance of poorly resolved density that occupied this region, which suggested the presence of an endogenously bound substrate or inhibitor in the structure of the holoenzyme [16]. Subsequent studies indicate that this electron density may be a citrate molecule, one of the components included in the crystallization mixture. Activity studies indicate that citrate is indeed one of the many inhibitors of the enzyme with a Ki in the millimolar range [23].

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_234.html [4/5/2004 5:08:18 PM]

Document

Page 235

Figure 4 Surface representation of the ALR2 holoenzyme in an orientation similar to Figure 3. The nicotinamide moiety that defines the active site of the enzyme is seen in the center. The groove that extends down from it is highly hydrophobic and was initially assumed to be the inhibitor binding site. Figure prepared using the GRASP program [49].

IV. Aldose Reductase Complexed with Inhibitor While a large number of high-potency inhibitors for ALR2 have been developed [4], a structural understanding of the exact molecular features that foster this affinity have been only vaguely understood. Several general chemical motifs such as hydrophobic ring systems, a spirohydrantoin group or carboxylate group are seen repeatedly when examining a list of known inhibitors (Figure 2) but little was known about the specific role for each in inhibitor binding. The structure of the ALR2/NADPH/zopolrestat ternary complex [18] has provided some answers about the mode of binding of zopolrestat (Figure 2), a high-affinity, carboxylate-containing compound developed by Pfizer, Inc. [13]. While the overall structure was preserved, the inhibitor binding induced a conformational change of the enzyme. This change, which involved the movement of several loops in the active site of the molecule, was large enough to cause a change in crystal packing relative to the holoenzyme. As a consequence of the shifting of the active-site loops, a cavity is created inside the protein in which the benzothiazole ring is seated and the groove that was implicated in substrate and inhibitor binding by the holoenzyme structure vanishes. This illustrates the unpredictability of conformational changes within a protein in response to substrate or inhibitor binding. It also implies that modeling compounds

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_235.html [4/5/2004 5:08:24 PM]

Document

Page 236

in the active site of an enzyme or drug-design algorithms such as inhibitor docking may be inappropriate in some cases if they do not adequately model considerable plasticity in the binding site. A closer look at the binding of zopolrestat shows that it is dominated by extensive hydrophobic contacts between the protein and the inhibitor. These include side chains from Trp20, Tyr48, Trp79, Trp111, Phe115, Phe122, Trp219, Ala299, Leu300, Tyr309, and Pro310 (Figure 7). This is not surprising given the apolar nature of the enzyme's active site as determined by the holoenzyme structure. What is surprising is that the inhibitor created the part of its own binding site that the benzothiazole rings occupy by “burrowing” into the hydrophobic core of the protein to carve out a region with very good steric complementarity to this moiety. It does this rather than binding in the solvent-exposed hydrophobic binding groove that is seen in the holoenzyme structure. The remaining interactions involving hydrogen bonds and salt links also appear to be very important in inhibitor binding. With the exception of one of the fluorine atoms, all atoms that are able to engage in hydrogen bonding do so. The carboxylate, which is seen in so many aldose reductase inhibitors, is saltlinked to His110, which is located very near the catalytic site (Figure 7). Presumably, the carboxylate in the other inhibitors plays the same role and could be used as an anchor when modeling these into the active site. Inhibition studies involving ALR2 have indicated noncompetitive inhibition for virtually all compounds examined to date when the forward (reduction) reaction is monitored. This mode of inhibition is often interpreted as meaning that the inhibitor binds to a site on the enzyme that is independent of the catalytic site. Kinetic and competition studies have both led to this conclusion in the case of ALR2 [24,25]. The crystal structure of the enzyme complexed with both the NADPH cofactor and zopolrestat, however, clearly shows the inhibitor occupying the region directly above the nicotinamide of the NADPH and, therefore, the active site (Figures 5, 6, and 7). Most previous inhibition studies reported noncompetitive and/or uncompetitive inhibition patterns when aldose reductase inhibitors were examined in the forward direction, i.e. inhibition of NADPH-dependent aldehyde reduction. With the finding that the overall rate-limiting step in the direction of aldehyde reduction is at the level of structural isomerization following alcohol product release [26,27], it is not surprising that lack of competitive inhibition would be observed in such standard double reciprocal plots. To further complicate matters, many aldose reductase inhibitors were not recognized in previous studies as tight-binding inhibitors and were inappropriately evaluated using Michaelis-Menten kinetics. Thus, noncompetitive or uncompetitive inhibition patterns were previously reported for inhibitors that were subsequently shown to bind directly at the active site. Recent structure-function and kinetic studies have revealed important details concerning the structural basis for the catalytic

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_236.html [4/5/2004 5:08:26 PM]

Document

Page 237

Figure 5 Schematic representation of the ALR2/NADPH/zopolrestat ternary complex. The NADPH is bound across the enzyme from the center to the right while the zopolrestat binds atop the nicotinamide and extends to the lower left. Conformational changes are seen in the C-terminal loop below the zopolrestat in the picture and the loop to the right of the inhibitor between β4 and α4.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_237.html (1 of 2) [4/5/2004 5:08:37 PM]

Document

Figure 6 A surface representation of the ternary complex as seen in Figure 5. Note that the inhbitor creates part of its binding site by “burrowing” into the protein rather than binding entirely in the groove seen Figure 4.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_237.html (2 of 2) [4/5/2004 5:08:37 PM]

Document

Page 238

Figure 7 Stereo of zopolrestat binding to the active site of ALR2. The salt link made by the carboxylate of the inhibitor and hydrogen bonds are depicted with dashed lines. The remainder of the interactions are apolar with the residues shown.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_238.html [4/5/2004 5:08:48 PM]

Document

Page 239

and inhibition mechanisms and provided a clarification for the mechanism by which inhibitors of both the carboxylate [28] and spirohydantoin [29] classes bind at the active site of ALR2. V. Aldo-Keto Reductase Family A. Effects of Sequence on Drug Design The problem of structure-based drug design for ALR2 and the drug-design effort in general is compounded by the fact that this enzymes is a member of a large family of aldo-keto reductases with overlapping substrate specificity. In humans at least three such enzymes have been found: ALR2 [30], aldehyde reductase (ALR1) [30], and chlordecone reductase [31]. Other members of the family that have been isolated in other species include rat 3α-hydroxysteroid dehydrogenase [32], murine fibroblast growth factor induced protein [33], bovine prostagladin F synthase [34], murine vas deferens protein [35], frog ρ-crystallin [36], the P100/11E gene product in Leishmania major [37], and Corynebactium diketogluconate reductase [38]. This large number suggests that there may be more such enzymes to be found in humans. The similarities between the proteins with respect to both the sequence and substrate specificity implies that the nature of the substrate binding sites are similar across the family. This has indeed been the case in all the structures determined from this family to date (see below). While detailed binding studies of various inhibitors with all the different enzymes have not been conducted, it is likely that drugs intended for ALR2 are likely to “cross react” with many of the other enzymes within the family. One such case that has recently been studied both crystallographically and biochemically is the murine FR-1 protein described below. The binding sites of all of these enzymes are characterized by their large size and hydrophobicity suggesting that ideal substrates may be steroids or molecules of a similar size and nature. Sequence comparisons of all the proteins, including those whose structure has not yet been determined, show that there is a large amount of similarity involving residues implicated in substrate binding (see Figure 8). One region that diverges somewhat is the 15-amino acid segment at the carboxy terminus of the protein. This segment is likely to be responsible for what little differences in substrate specificity exhibited by the enzymes. It is the same segment that is seen adopting a different conformation upon zopolrestat binding to ALR2. It may then be possible that it is not only the chemical nature of this loop that—in making positive and negative interactions with the substrate/inhibitor—modulates specificity, but also the flexibility conferred by the amino acid sequence. Such a difference is seen when contrasting the structures of ALR2 and FR-1 bound to zopolrestat.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_239.html [4/5/2004 5:09:05 PM]

Document

Page 240

Figure 8 Sequence alignment of several members of the aldo-keto reductase family. Abbreviations used are HALR2, human aldose reductase (30); HALR1, human aldehyde reductase [30]; 3α-HSD, 3α-hydroxysteroid dehydrogenase [32]; FR-1, murine FR-1 [33]; BPGFS, bovine prostaglandin F synthase [34]; CCDR12, human chlordecone reductase [31]; CDGR, Corynebacterium diketogluconate reductase [38]; MVDP, murine vas deferens protein [35], JFRC, Japanese frog ρ crystallin [36].

B. Structures Structures of several other members of the aldo-keto reductase family have also been determined. These include aldehyde reductase [39,40], FR-1 [41] and 3α-hydroxysteroid dehydrogenase [42]. Since each of these proteins retain a large amount of sequence identity and homology with human ALR2, it is not surprising to note that the overall tertiary structures are very similar. Root-meansquare Cα deviations between the human ALR2 holoenzyme structure and the rest of the family are in the range of 1–2Å

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_240.html (1 of 2) [4/5/2004 5:09:22 PM]

Document

Page 241

Closer examination of the active site shows that the residues involved in catalysis (most notably the residues analogous to Asp43, and Tyr48, and Lys77 in ALR2) are structurally conserved among each of these proteins (Figure 8), suggesting that the mechanism is also conserved throughout the family. Many of the residues found in the binding site (defined as those making contact with the zopolrestat in the ternary complex) are also largely conserved with the exception of a number of residues in the carboxy terminus of the protein. This is indeed where the most structure variation appears to be concentrated among the proteins. These residues compose a loop that is the same loop that shifts upon binding of zopolrestat in ALR2. Inhibitors with improved specificity will very likely take advantage of the subtle structural differences that are introduced by the variation in sequences in this area. VI. Future Design of Aldose Reductase Inhibitors The availability of structural data for ALR2 in its holoenzyme and different ternary forms is likely to lead to improvements in the affinity of future generations of inhibitors. As the architecture and plasticity of the binding site are better understood, increasingly potent inhibitors may be designed to occupy it. Although these structures provide a positive target for drug design, there are a number of negative targets. Increased in vivo potency is likely to be derived from the specific inhibition of ALR2 that would entail the avoidance of other members of the aldo-keto reductase family. Determination of the structures of other members of the family may increase the specificity of compounds by providing structures of targets to avoid. While the incorporation of the negative targets in the drug-design process relies on the determination of other structures and is likely to be complicated, conventional computational techniques may be applied to the problem of the positive target. Two such methods that may hold promise are docking [43] and computational thermodynamic perturbation [44]. A. Inhibitor Docking to the Enzyme Our initial efforts to exploit the ALR2 holoenzyme structure for drug design utilized the program DOCK [43]. This program is capable of finding depressions on the surface of the enzyme that could serve as binding sites for substrates or inhibitors. Once the correct area is defined, the program rotates structures of candidate compounds within this space and scores each compound based upon its steric complementarity with the binding site. However, the program does not include the potential polar interactions between the inhibitor and protein when scoring. The search was further constrained by the inability to include conformational variations both in the test compounds as well as the protein, due to computational limitations.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_241.html [4/5/2004 5:09:24 PM]

Document

Page 242

This method was used to screen a significant portion of the contents of the Cambridge Structural Database [45] against the ALR2 holoenzyme binding site (D.K. Wilson, J. M. Petrash, and F. A. Quiocho, unpublished data). Among the approximately 30 highest scoring compounds were several aromatic aldoximes that had inhibition constants in the micromolar range. These were similar to aldoximes such as benzaldoxime, which has been previously observed to have similar inhibition constants [46]. A disappointing result was that this search did not “rediscover” any of the known high-affinity ALR2 inhibitors that were contained in the search library. Before the determination of the ternary complex of the enzyme with zopolrestat, this was interpreted as meaning that these compounds bound to the enzyme in a conformation somewhat different than the one adopted in the crystal structure used for the search. The structure of the ternary complex showed this to be a wrong assumption; it was the protein that changed conformation upon inhibitor binding, creating a pocket that did not exist in the holoenzyme structure. When bound to the protein, zopolrestat is actually quite similar in conformation to its small molecule x-ray structure. It is therefore very possible that different ALR2 inhibitors and substrates may cause the enzyme to flex in different ways, creating binding sites that may be different in size and chemical nature. B. Computational Thermodynamic Perturbation Computational thermodynamic perturbation is a powerful, albeit computationally expensive, group of techniques that are designed to estimate relative binding affinities of two closely related drugs, given the structure of at least one of them complexed with the target protein [44]. This approach has the potential to assay candidate compounds in the computer for improvements in inhibitor binding, thereby removing the necessity to sythesize and assay these compounds in the lab. For a number of reasons, ALR2 promises to be a good system for the application of this technique and the experimental verification of the results. The structure is very well determined in complex with zopolrestat, a high-affinity inhibitor. A number of zopolrestat derivatives with various functional groups decorating the compound have been sythesized and partially characterized with respect to ALR2 inhibition [13,47]. These compounds could serve as a sort of basis set of controls for the theoretical calculations. If parameters used in these calculations can be selected such that the computationally derived binding energies agree even qualitatively with the experimentally determined binding energies, serious consideration should be given to new compounds that are predicted to bind with enhanced affinity. Since ALR2 is crystallizable with zopolrestat bound, there is every reason to believe that crystals of the enzyme complexed with similar compounds will be obtainable. Such structures could provide the basis for further rounds of drug improvement.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_242.html [4/5/2004 5:09:26 PM]

Document

Page 243

VII. Conclusions The observation of a protein undergoing a conformational change when binding to an inhibitor, as seen with ALR2 and zopolrestat, illustrates a common problem associated with structure-based drug design. It is tempting to view proteins as static structures since their crystal structures are static. Attempts to design drugs to fit the apparent active site of an enzyme may fail when the plasticity of the protein is not taken into account. While the disorder associated with amino acid side chains can be modeled with a moderate computational effort, larger conformational changes—such as the loop movement seen in ALR2—are virtually impossible to predict. Until this becomes possible, x-ray crystal structures of complexes will continue to be indispensible. Finally, it can be easy to forget that a compound's affinity for the protein is not the only consideration when designing inhibitors of enzymes from a structural point of view. The structures of aldose reductase and the FR-1 protein complexed with the drug zopolrestat, a compound with a very high affinity for ALR2, can serve as a reminder of how specificity can also be a very important factor. This is particularly true when a protein is a member of a family of proteins that share sequence homology and are apt to have overlapping specificities. Structure may then play a key role in the determination of features that are unique to the target protein and therefore prime considerations when designing inhibitors. Acknowledgments We thank T. Reynolds who assisted with the production of the figures. This work was supported by a grant from Research to Prevent Blindness, Inc. and grants EY05856, EY02687, and DK20579 to J. Mark Petrash. Florante A. Quiocho is an investigator of the Howard Hughes Medical Institute. References 1. Kinoshita JH, Nishimura C. The involvement of aldose reductase in diabetic complications. DiabetesMetebolism Rev 1988; 4:323–337.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_243.html (1 of 2) [4/5/2004 5:09:27 PM]

Document

2. Lee AYW, Chung SK, Chung SSM. Demonstration that polyol accumulation is responsible for diabetic cataract by the use of transgenic mice expressing the aldose reductase gene in the lens. Proc Natl Acad Sci USA 1995;92:2780–2784. 3. Yamaoka T, Nishimura C, Yamashita K, Itakura M, Yamada T, Fujimoto J, Kokai Y. Acute onset of diabetic pathological changes in transgenic mice with human aldose reductase cDNA. Diabetologia 1995;38:255–261. 4. Sarges R, Oates P. Aldose reductase inhibitors: Recent developments. Prog Drug Res 1993; 40:99–161.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_243.html (2 of 2) [4/5/2004 5:09:27 PM]

Document

Page 244

5. National Advisory Eye Council (1994). In: Vision Research: A National Plan 1994–1998. National Institutes of Health Publication No. 93–3186. 6. Kador PF. The role of aldose reductase in the development of diabetic complications. Med Res Rev 1988; 8:325–352. 7. Dvornik D. Aldose Reductase Inhibition: An Approach to the Prevention of Diabetic Complications. New York: Biomedical Information Corporation 1987. 8. Dvornik D, Simard-Duquesne N, Krami M, Sestanj K, Gabbay KH, Kinoshita JN, Varma SD, Merola LO. Polyol accumulation in galatosemic and diabetic rats: control by an aldose reductase inhibitor. Science 1973; 182:1146–1148. 9. Sestanj K, Bellini F, Fung S, Abraham N, Treasurywala A, Humber L, Simard-Duquesne N, Dvornik D. N-[[5-(trifluoromethyl)-6-methoxy-1-naphthalenyl]thioxomethyl]-N-methylglycine (Tolrestat), a potent, orally active aldose reductase inhibitor. J Med Chem 1984;27:255–256. 10. Ward WHJ, Sennitt CM, Ross H, Dingle A, Timms D, Mirrlees DJ, Tuffin DP. Ponalrestat: a potent and specific inhibitor of aldose reductase. Biochem Pharmacol 1990; 39:337–346. 11. Terashima H, Hama K, Yamamoto R, Tsuboshima M, Kikkawa R, Hatanaka 1, Shigeta Y. Effects of a new aldose reductase inhibitor on various tissues in vitro. J Pharmacol Exp. Ther 1984;229:226–230. 12. Peterson MJ, Sarges R, Aldinger CD. CP-45634: a novel aldose reductase inhibitor that inhibits polyol pathway activity in diabetic and galactosemic rats. Metabolism 1979; 28(supp 1):456–461. 13. Mylari BL, Larson ER, Beyer TA, Zembrowski WJ, Aldinger CE, Dee MF, Siegel TW, Singleton DH. Novel, potent aldose reductase inhibitors: 3,4-dihydro-4-oxo-3-[[5-(trifluoromethyl)-2benzothiazolyl]methyl]-1-phthalazine-acetic acid (zopolrestat) and congeners. J Med Chem 1991; 34:108–122. 14. Srivastava SK, Petrash JM, Sadana AJ, Partridge CA. Susceptibility of aldose and aldehyde reductases to aldose reductase inhibitors. Curr Eye Res 1982; 2:407–410. 15. Rondeau JM, Tete-Favier F, Podjarny A, Reymann JM, Barth P, Biellmann JF, Moras D. Novel NADPH-binding domain revealed by the crystal structure of aldose reductase. Nature 1992; 355:469–472. 16. Wilson DK, Bohren KM, Gabbay KH, Quiocho FA. An unlikely sugar substrate site in the 1.65 Å structure of the human aldose reductase holoenzyme implicated in diabetic complications. Science 1992; 257:81–84. 17. Borhani DW, Harter TM, Petrash JM. The crystal structure of the aldose reductase-NADPH binary complex. J Biol Chem 1992; 267:24841–24847. http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_244.html (1 of 2) [4/5/2004 5:09:29 PM]

Document

18. Wilson DK, Tarle I, Petrash JM, Quiocho FA. Refined 1.8 Å structure of human aldose reductase complexed with the potent inhibitor zopolrestat. Proc Natl Acad Sci USA 1993; 90:9847–9851. 19. Branden CI. The TIM barrel—the most frequently occurring folding motif in proteins. Curr Opin Struct Biol 1991; 1:978–983. 20. Feldman HB, Szczepanik PA, Havre P, Corrall RJM, Yu LC, Rodman HM, Rosner BA, Klein PD, Landau, BR. Stereospecificity of the hydrogen transfer catalyzd by human placental aldose reductase. Biochim Biophys Acta 1997; 480:14–20. 21. Wermuth B, Buergisser HB, Bohren KM, von Wartburg JP. Purification and characterization of human brain aldose reductase. Eur J Biochem 1982; 127:279–284.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_244.html (2 of 2) [4/5/2004 5:09:29 PM]

Document

Page 245

22. Tarle I, Borhani DW, Wilson DK, Quiocho FA, Petrash JM. Probing the active site of human aldose reductase: site directed mutagenesis of Asp-43, Lys-77 and His-110. J Biol Chem 1993; 268:25687–25693. 23. Harrison DH, Bohren KM, Ringe D, Petsko GA, Gabbay KH. An anion binding site in human aldose reductase: mechanistic implications for the binding of citrate, cacodylate, and glucose 6-phosphate. Biochemistry 1994; 33:2011–2020. 24. Kador PF, Sharpless NE. Pharmacophor requirements of the aldose reductase inhibitor site. Mol Pharmacol 1983; 24:521–531. 25. Kador PF, Goosey JD, Sharpless NE, Kolish J, Miller DD. Stereospecific inhibition of aldose reductase. Eur J Med Chem 1981; 16:293–298. 26. Kubiseski TJ, Hyndman DJ, Morjana NA, Flynn TG. Studies on pig muscle aldose reductase. Kinetic mechanism and evidence for a slow conformational change upon coenzyme binding. J Biol Chem 1992; 267:6510–6517. 27. Grimshaw CE, Shahbaz M, Putney CG. Mechanistic basis for nonlinear kinetics of aldehyde reduction catalyzed by aldose reductase. Biochemistry 1990; 29:9947–9955. 28. Grimshaw CE, Bohren KM, Lai CJ, Gabbay KH. Human aldose reductase: pK of tyrosine 48 reveals the preferred ionization state for catalysis and inhibition. Biochemistry 1995; 34:14374–14384. 29. Liu SQ, Bhatnagar A, Srivastava SK. Does sorbinil bind to the substrate binding site of aldose reductase? Biochem Pharmacol 1992; 44:2427–2429. 30. Bohren KM, Bullock B, Wermuth B, Gabbay KH. The aldo-keto reductase superfamily: cDNAs and deduced amino acid sequences of human aldehyde and aldose reductases. J Biol Chem 1989; 264:9547–9551. 31. Winters CJ, Molowa DT, Guzelian PS. Isolation and characterization of cloned cDNAs encoding human liver chlordecone reductase. Biochemistry 1990; 29:1080–1087. 32. Pawlowski JE, Huizinga M, Penning TM. Cloning and sequencing of the cDNA for rat liver 3αhydroxysteroid/dihydrodiol dehydrogenase. J Biol Chem 1991; 266:8820–8825. 33. Donohue PJ, Alberts GF, Hampton BS, Winkles JA. A delayed-early gene activated by fibroblast growth factor-1 encodes a protein related to aldose reductase. J Biol Chem 1994; 269:8604–8609. 34. Watanabe K, Fujii Y, Nakayama K, Ohkubo H, Kuramitsu S, Kagamiyama H, Nakanishi S, Hayaishi O. Structural similarity of bovine lung prostaglandin F synthase to lens ε crystallin of the European common frog. Proc Natl Acad Sci U S A 1988; 85:11–15.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_245.html (1 of 2) [4/5/2004 5:10:06 PM]

Document

35. Pailhoux EA, Martinez A, Veyssiere GM, Jean CG. Androgen-dependent protein from mouse vas deferens: cDNA cloning and protein homology with the aldo-keto reductase superfamily. J Biol Chem 1990; 265:19932–19936. 36. Fujii Y, Watanabe K, Hayashi H, Urade Y, Kuramitsu S, Kagamiyama H, Hayashi O. Purification and characterization of p-crystallin from Japanese common bullfrog lens. J Biol Chem 1990; 265:9914–9923. 37. Samaras N, Spithill TW. The developmentally regulated P100/11E gene of Leishmania major shows homology to a superfamily of reductase genes. J Biol Chem 1989; 264:4251–4254. 38. Anderson S, Marks CM, Lazarus R, Miller J, Stafford K, Seymour J, Light D, Rastetter W, Estell D. Production of 2-keto-L-gulonate, an intermediate in L-

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_245.html (2 of 2) [4/5/2004 5:10:06 PM]

Document

Page 246

ascorbate synthesis, by a genetically modified Erwinia herbicola. Science 1985; 230:144–149. 39. El-Kabbani O, Green NC, Lin G, Carson M, Narayanam SVL, Moore K, Flynn TG, DeLucas LJ. Structures of human and porcine aldehyde reductase: an enzyme implicated in diabetic complications. Acta Crystallogr D 1994; 50:859–868. 40. El-Kabbani O, Judge K, Ginell SL, Myles Daa, DeLucas LJ, Flynn TG. Structure of porcine aldehyde reductase holoenzyme. Nat Struct Biol 1995; 2:687–692. 41. Wilson DK, Nakano T, Petrash JM, Quiocho FA. 1.7 Å structure of FR-1, a fibroblast growth factorinduced member of the aldo-keto reductase family complexed with coenzyme and inhibitor. Biochemistry 1995; 34:14323–14330. 42. Hoog SS, Pawlowski JE, Alzari PM, Penning TM, Lewis M. Three-dimensional structure of rat liver 3α-hydroxysteroid/dihydrodiol dehydrogenase: a member of the aldo-keto reductase superfamily. Proc Natl Acad Sci USA 1994; 91:2517–2521. 43 Schoichet B, Bodian D, Kuntz I. Molecular docking using shape descriptors. J Comp Chem 1992; 13:380–397. 44 Straatsma TP, McCammon JA. Computational alchemy. Annu Rev Phys Chem 1992 43:407–435. 45 Allen FG, Bellar SA, Brice MD, Cartwright BA, Doubleday A, Higgs H, Hummelink T, HummelinkPeters BG, Kennard O, Motherwell WDS, Rodgeres JR, Watson DG. The Cambridge Crystallographic Data Centre: Computer based search retrieval, analysis and display of information. Acta Crystallogr B35:2331–2339. 46 Shen C, Sigman DS. New inhibitors of aldose reductase: anti-oximes of aromatic aldehydes. Arch Biochem Biophys 1991; 286:596–603. 47 Mylari BL, Beyer TA, Scott PJ, Aldinger CE, Dee MF, Siegel TW, Zembrowski WJ. Potent, orally active aldose reductase inhbitors related to zopolrestat: surrogates for benzothiazole side chain. J Med Chem 1992; 35:457–465. 48 Kraulis PJ. MOLSCRIPT: a program to produce both detailed and schematic plots of protein structures. J Appl Crystallog 1991; 24:946–950. 49 Nicholls A, Sharp KA, Honig B. Protein folding and association: insights from the interfacial and thermodynamic properties of hydrocarbons. Proteins 1991; 11:281–296.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_246.html [4/5/2004 5:10:07 PM]

Document

Page 247

10 Structure-Based Design of Thrombin Inhibitors Patricia C. Weber and Michael Czarniecki Schering-Plough Research Institute, Kenilworth, New Jersey I. Roles of Thrombin in Hemostasis and the Therapeutic Utility of Thrombin Inhibitors Thrombin is a serine protease that plays critical roles in both blood clot formation and anticoagulation. In the penultimate step of the coagulation cascade, thrombin cleaves soluble fibrinogen to form insoluble fibrin. Thrombin also activates other coagulation factors including Factor XIII, the enzyme responsible for crosslinking fibrin to further stabilize the thrombus. Additional clot-promoting functions include stimulation of platelet aggregation by cleavage of the thrombin receptor. In contrast to its roles in clot formation, thrombin participates in anticoagulant functions. For example, thrombin-mediated activation of protein C, a protease involved in anticoagulation, is enhanced when thrombin is complexed with thrombomodulin, and in this complex, thrombin can neither cleave fibrinogen nor activate platelets. The interrelationship among thrombin's many roles in hemostasis is complex and presents several mechanisms for inhibition of thrombus formation. For recent reviews see References 1 through 5. Most drug discovery efforts focus on thrombin inhibition as a means to prevent the serious consequences of thrombus formation in myocardial infarction and stroke. Thrombin inhibitors may also prevent clot formation in patients prone to deep vein thrombosis or repeat heart attack. In combination with thrombus dissolution therapies, thrombin inhibitors may decrease the incidence of reocclusion due, in part, to the release of active clot-bound thrombin. In this article, recent examples of small molecule inhibitors interacting at the fibrinogen primary specificity pocket and with residues of the catalytic triad

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_247.html [4/5/2004 5:10:09 PM]

Document

Page 248

are given. Inhibitors designed to make more extended interactions with thrombin are also presented. II. Structure of Thrombin Thrombin consists of two polypeptides, an A chain of 36 residues and a 259-residue B chain, linked by a disulfide bond. The crystallographic structure of thrombin reveals a globular protein organized about two β barrels with the overall folding pattern of the chymotrypsin serine protease family [6,7]. The catalytic triad and nearby oxyanion hole are located roughly between the β barrels and adopt the geometric arrangement required for serine-protease-assisted, peptide bond cleavage (Figure 1). Thrombin's multifunctionality and regulation of activity are achieved by specialized subsites on the enzyme's surface (Figure 2). Fibrinogen cleavage, for example, involves interactions at the primary specificity pocket, the extended fibrinogen recognition exosite, and an additional specificity pocket. Subsite interactions differ for cleavage of other thrombin substrates including the thrombin receptor and protein C. Additional and overlapping subsites exist for thrombin effector molecules including heparin, antithrombin III, and heparin cofactor II [8,9].

Figure 1 Stereoscopic view of the crystallographic structure of thrombin complexed with N-acetyl-(D-Phe)-Pro-boroArg-OH. Helical regions are represented in the standard way and arrows indicate regions of β sheet. Solid lines show the thrombin bound conformation of N-acetyl-(D-Phe)-Pro-boroArg-OH (taken from Reference 10). Active-site residues, His57 and Ser195, are shown with a ball-and-stick representation. The authors thank Dr. C. L. Strickland for the drawing.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_248.html [4/5/2004 5:10:15 PM]

Document

Page 249

Figure 2 Schematic representation of Subsite Utilization in Thrombin Complexes (after Reference 8). Fibrinogen interacts with three thrombin subsites (here thrombin is represented by a large oval and the interconnected subsites by an irregular three-armed shape). Physiological effectors of thrombin and thrombin inhibitors form distinct interactions at these subsites. Additional subsites, such as the heparin-binding site, exist on the thrombin surface and are not indicated here. The catalytic triad is represented by three circles at the vertices of a triangle.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_249.html (1 of 2) [4/5/2004 5:10:27 PM]

Document

Page 250

III. Thrombin Inhibitors Directed at the Fibrinopeptide a Binding Pocket The majority of synthetic thrombin inhibitors interact at the fibrinopeptide A binding pocket, which includes the catalytic residues Ser195 and His57, hydrogen-bonding capabilities within the oxyanion hole, peptide backbone functional groups that hydrogen bond with the peptide backbone of the substrate, and residues involved in amino acid recognition (Figure 3). Many of these binding determinants are utilized by N-acetyl-(D-Phe)-Pro-Arg-chloromethylketone (PPACK [7]) and its boronic acid analog (DUP714 [10]). The crystallographic structures of these molecules complexed with thrombin have both served as starting points for structure-based drug design and as reference structures for comparison of binding modes of other inhibitors. The use of arginine boronate esters as transition-state mimetics results in potent peptidyl thrombin inhibitors. These inhibitors, however, exhibit significant affinity for other serine proteases that have in common a specificity for substrates with basic residues at P1 (e.g. trypsin, Factor Xa, and plasmin). Earlier work demonstrated that neutral side chains of P1 boronate esters impart greater selectivity for thrombin. The boropeptide shown in Figure 4 was investigated as the prototype of neutral side chain, tripeptide thrombin inhibitors [11]. It had a Ki against thrombin of 7 nM and shows selectivity relative to other trypsin-like plasma proteases. Since these inhibitors have a neutral residue at the P1 site, Deadman and coworkers [11] sought to demonstrate the mode of binding to thrombin in the absence of a salt bridge with Asp189.

Figure 3 Schematic diagram of binding determinants within the fibrinopeptide A binding pocket of thrombin and their utilization by N-acetyl-(D-Phe)-Pro-boroArg-OH.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_250.html [4/5/2004 5:10:30 PM]

Document

Page 251

Figure 4 Schematic representation of the active-site orientation of a “neutral” P1 boronic acid thrombin inhibitor.

Boron-11-NMR, a sensitive probe of the chemical environment around boronate esters, can distinguish between trigonal and tetrahedral forms of boron. The 11B-NMR spectrum of this inhibitor complexed with thrombin showed a single peak at -17 ppm that remained constant for 7 hours. The chemical shift suggests boron adopts a tetrahedral geometry on binding to thrombin and is consistent with the orientation of the inhibitor in the active site shown in Figure 4. While the 11B-NMR revealed an interaction within the catalytic site, it could not distinguish between bonding with Ser195 or His57. Kahn and coworkers [12] recently investigated the application of synthesized peptidomimetics as novel inhibitors of thrombin. Fibrinogen peptide A mimetic (FPAM, Figure 5) incorporates a bicyclic peptidomimetic within the turn region of fibrinogen peptide A. The bicyclic peptidomimetic confers conformational stability to the turn region as suggested by x-ray crystal structures of fibrinogen peptide complexes as well as complexes of BPTI with thrombin. X-ray crystallographic studies of FPAM complexed with thrombin (Figure 5) showed that the S1 subsite is occupied by the arginine guanidinium [12]. The Val group of FPAM makes extensive hydrophobic contacts within the S2 apolar binding site. The Gly at P3 interacts with thrombin via a β-sheet-type hydrogen bond with the carbonyl group of Gly216 and appears to be important in the positioning of the bicyclic ring corresponding to the β bend. This bicyclic ring, although not aromatic, forms an edge-toface contact with Trp215. One of the phenyl rings shows hydrophobic contact with lle174, while the other shows no significant interactions with thrombin. By comparison to other inhibitors complexed to thrombin, FPAM appears to have a new binding mode that differs from that of substrate, or hirudin, or argatroban.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_251.html (1 of 2) [4/5/2004 5:10:32 PM]

Document

Page 252

Figure 5 Schematic representation of the principal intermolecular interactions of a fibrinogen peptide A mimetic within the active site of thrombin.

Obst et al. [13], departing from the peptide template, designed and synthesized novel nonpeptide inhibitors of thrombin. They began with a cyclic template having attachment sites for three side chains that would be complementary to the S1, S2, and S3 sites in thrombin. Important to the design was a rigid template that would avoid hydrophobic collapse of the side chains and loss of conformational degrees of freedom upon complex formation with thrombin. Using computational approaches (Insight II/Discover/CVFF force field), possible templates were modeled within the active site of thrombin. These studies resulted in the synthesis of thirteen analogs that shared a common template. The most active molecule (Ki = 90 nM, 8-fold selective versus trypsin) was studied further by x-ray crystallography (Figure 6). The positively charged benzamidine binds into the S1 pocket of thrombin forming a bidentate hydrogen bond with Asp189. The proximal carbonyl of the rigid template acts as a hydrogen-bond acceptor for the amide NH of Gly216. The methylene dioxybenzyl group at P3 interacts with thrombin in two ways. An edge-to-face interaction was observed with Trp215, and an oxygen of the methylenedioxy group acts as an acceptor for a hydrogen bond with the OH hydrogen of Tyr60A. Recent communications from Bristol-Myers Squibb [14,15] describe peptidomimetic inhibitors (Figure 7) that were designed to bind thrombin with an N- to C-polypeptide chain sense opposite that of the substrate and form interactions similar to those made by the first three residues of hirudin (lle1, Thr2, Tyr3). In the x-ray crystal structure of BMS-183507 (Ki = 17.2 nM) with thrombin [15], the N terminus is facing the catalytic site while the methyl ester is

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_252.html (1 of 2) [4/5/2004 5:10:35 PM]

Document

Page 253

Figure 6 Schematic representation of the principal intermolecular interactions of a nonpeptide bicyclic inhibitor within the active site of thrombin.

exposed to solvent. No specific interactions were observed with the catalytic triad. A bound water molecule hydrogen bonded to the Ser195 hydroxyl. The complex is stabilized by a network of hydrogen bonds as well as hydrophobic interactions. The Phe1-O and the Phe3-NH form hydrogen bonds with Gly216, and the Phe1-NH hydrogen bonds to the backbone carbonyl of Ser214. The Phe1 phenyl group occupies the S2 site, while Phe3 interacts within the S3 site. The retro-inhibitors contain a 4-guanidinobutanoyl group that extends into the S1 specificity site. Rather than forming two hydrogen bonds between the guanidine and Asp189 in a manner similar to PPACK, BMS-183507 forms only one, with the second hydrogen bond being directed to the carbonyl oxygen of Gly219. Binding affinity, as evidenced by loss of more than two orders of magnitude in affinity on addition of one or two methylene groups, was sensitive to chain length at this position. The allo-Thr hydroxyl oxygen accepts a hydrogen bond from the backbone NH of Gly219. This additional interaction accounts, at least in part, for the increase in affinity when compared to the inhibitor with Leu in this position. Comparison of the crystal structures of thrombin complexed with BMS-183507 and with hirudin reveals that the hirudin residue, Thr2, and the allo-Thr of BMS-183507 interact differently with thrombin. The hirudin Thr2 binds at S2,

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_253.html (1 of 2) [4/5/2004 5:10:38 PM]

Document

Page 254

Figure 7 Schematic representation comparing the principal inhibitor-to-thombin interactions of related inhibitors with either Leu (7a) or allo-Thr (7b) at P3.

whereas the allo-Thr sidechain is oriented toward the protein exterior and is partially exposed to solvent.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_254.html (1 of 2) [4/5/2004 5:10:42 PM]

Document

Cyclotheonamide A (CtA), a macrocyclic marine natural product derived from the Japanese sponge, Theonella sp., inhibits thrombin with an IC50 value of

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_254.html (2 of 2) [4/5/2004 5:10:42 PM]

Document

Page 255

Figure 8 Schematic representation of the principal intermolecular interactions of cyclotheonamide A within the active site of thrombin.

100 nM and represents a novel structural class of serine protease inhibitors. An x-ray crystal structure of CtA complexed with thrombin was used to determine the molecular basis for this inhibition (Figure 8 [16]). The Arg-Pro unit binds to the S1 and S2 sites in a manner similar to the Arg-Pro of PPACK. The Arg guanidinium group forms a bidentate hydrogen bond with Asp189 while the Pro establishes a βsheet interaction with the Ser214-Gly216 backbone. The α-ketoamide acts as a transition-state mimetic forming a tetrahedral hemiketal with the hydroxyl of Ser195. Within the complex, CtA adopts a relatively open conformation with the Pro orthogonal to the macrocycle and confined by a hydrophobic pocket defined by Tyr60A, Trp60D, and Leu99. Two aromatic residues are involved in stacking interactions with Tyr60A and Trp60D. Cyclotheonamide A, however, does not effectively match the S3 interactions provided by the D-Phe group found in PPACK. In CtA, the formamide group is too polar to effectively complement the S3 site adjacent to Trp215. The authors note that the complex of CtA with thrombin does not appear optimal and suggest that synthetic analogs could significantly improve both potency and selectivity. Starting with the known thrombin inhibitors Argatroban and Nα(2-naphthyl-sulfonyl-glycyl)-DL-pamidinophenylalanyl-piperidine (NAPAP), a group at Roche initiated a medicinal chemistry program to develop thrombin inhibitors with reduced toxicity and an improved hemodynamic profile [17].

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_255.html (1 of 2) [4/5/2004 5:10:45 PM]

Document

Page 256

The discovery program proceeded in four iterative phases which are shown in Table 1. Initial screening of low molecular weight organic bases led to the discovery of 1-amidinopiperidine (1–1) as a new surrogate for the guanidine and amidine functionality in Argatroban and NAPAP, respectively. A distinct advantage of 1-amidinopiperidine is its intrinsic selectivity for thrombin over

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_256.html (1 of 2) [4/5/2004 5:10:50 PM]

Document

Page 257

trypsin. Application of three-fold iterative strategy of design involving synthesis, x-ray crystallography, and molecular modeling, this group elaborated the 1-amidinopiperidine from structures that inhibited in the micromolar range to some inhibiting in the picomolar range. In doing so significant improvements in the selectivity of thrombin relative to trypsin were also achieved. In the case of the D-amino acid series (1–2), a “second inhibitor binding mode” that differed from that of Argatroban was identified. In this new and unexpected binding mode, the S2 pocket is unoccupied and the napthalenesulfonyl group fills the S3 site and overlaps the front of the S2 site. The benzyl group of the phenylalanine is oriented toward the protein surface and is partially exposed to solvent. The Argatroban or “inhibitor binding” mode was favored by the more potent L-amino acid series (1–3 and 1–4) where the piperidide (1–3) or Nbenzyl (1–4) binds to the S2 site and the aryl groups are found in the S3 site. IV. Bivalent Thrombin Inhibitors Directed at the Fibrinopeptide a Binding Pocket and the Fibrinogen Recognition Site A strategy to prepare highly selective thrombin inhibitors involves linkage of molecules capable of interacting at distinct subsites. This approach should result in inhibitors more specific for thrombin: while serine proteases possess common structural features related to catalysis and some serine proteases—including the coagulation enzyme Factor Xa—also exhibit primary substrate specificity for positively charged residues, only thrombin possesses recognition subsites for fibrinogen and effector molecules such as thrombomodulin. Nature has used this strategy in the evolution of hirudin, the anticoagulant protein produced by the medicinal leech. When this effective anticoagulant binds thrombin [18–20], the N-terminal domain blocks the primary specificity pocket while the C-terminal residues adopt an extended conformation and make multiple interactions within the fibrinogen recognition exosite. Guided by structural and biochemical information, small molecules capable of simultaneous interactions with both the primary specificity pocket and the fibrinogen recognition exosite were designed and synthesized. These bivalent inhibitors are composed of three regions: a group to block the primary specificity pocket, a sequence to bind the fibrinogen recognition site, and a chemical linker. The bivalent inhibitor approach was first executed with peptides [21–22]. In 1990, DiMaio et al. (3–3 [22]) used the peptide sequence from hirudin to link (d-Phe)-Pro-Arg-Pro, known to bind at the primary specificity pocket [23], with hirudin C-terminal residues, known to bind at the fibrinogen recognition site. Polyglycine linkers were also used to connect these sequences (Maraganore

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_257.html [4/5/2004 5:10:52 PM]

Document

Page 258

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_258.html [4/5/2004 5:11:22 PM]

Document

Page 259

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_259.html [4/5/2004 5:11:26 PM]

Document

Page 260

et al. [21]). Among these hirudin analogs, the tetraglycine linker appeared optimal (3–8, Ki = 2.3 nM, Table 2). Most of the peptide-based bivalent inhibitors were slowly cleaved by thrombin. Incorporation of a ketomethylene pseudo peptide bond (3–4) resulted in a noncleavable bivalent inhibitor that retained high thrombin affinity [24]. Decreased proteolysis in bivalent inhibitors increasingly nonpeptide in character continues to be observed. Chemically simpler linkers were made using multiple methylene-containing glycine variants [25]. The dependence of affinity on placement of amide linkage within linkers containing the same number of atoms indicated some specific thrombin-to-linker interactions (3–13,14,15,16). This was confirmed in the crystal structure of hirutonin-6:thrombin complex (3–26 [26]) where continuous electron density was observed for the entire bivalent inhibitor including the linker region. The extended nature of the fibrinogen recognition site complicates attempts to reduce inhibitor molecular weight while maintaining affinity. Although of similar molecular weight, substitution of the sequence -Asp-Tyr-Glu-Pro-lle-Pro-Glu-Glu-Ala-cyclohexylalanine-(D-Glu) for -Asp-Phe-Glu-Glu-llePro-Glu-Glu-Tyr-Leu-Gin increases affinity an order of magnitude (compare 3–17 and 3–18). Within a series of bivalent inhibitors, inclusion of sulfated tyrosine, the naturally occurring residue of hirudin, increases affinity 5 to 6 fold (3–8 compared to 3–11, and 3–1 to 3–2). Only seven residues are present in one of the smallest bivalent inhibitors (3–26). Increasingly nonpeptide substituents have been incorporated into the primary specificity pocket binding portion of the bivalent inhibitors. Higher affinity for thrombin was achieved by replacement of the (DPhe)-Pro-Arg with either dansyl-Arg-(D-pipecolic acid) (3–17, [27]) or 4-tert-butylbenzenesulfonyl-Arg(D-pipecolic acid) (3–18, [27]). While the arginine side chain of these and the (D-Phe)-Pro-Argcontaining inhibitors make similar interactions with the aspartic acid within the S1 specificity pocket, the dansyl-Arg-(D-pipecolic acid) inhibitors bind in a nonsubstrate mode [27]. This initial result suggests that other nonpeptide thrombin inhibitors may be successfully incorporated into bivalent inhibitors. Recently, a pyridinium methyl ketone bivalent inhibitor capable of forming a reversible covalent complex with thrombin was synthesized (3–26, [28]). Crystallographic analysis of its complex with thrombin showed the ketone carbonyl becomes tetrahedrally coordinate by bonding to the side chain of thrombin's active site residue, Ser195. Substitutions of cyclohexylalanine for phenylalanine (3–4 compared to 3–5) and the cyclohexylalanine-containing fibrinogen recognition peptide for the hirudin sequence (3–17 compared to 3–18) also contribute to the increased affinity of this bivalent inhibitor.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_260.html [4/5/2004 5:11:28 PM]

Document

Page 261

V. Inactivated Thrombin as an Inhibitor of Clot Formation A means to selectively inhibit thrombin's role in coagulation while preserving its anticoagulant functions involves site-directed mutagenesis of thrombin itself. By introduction of a single mutation, Gibbs et al. [29] altered thrombin's relative specificity for fibrinogen and protein C. The engineered thrombin's increased activation of protein C over fibrinogen cleavage offers the possibility of inhibiting clot formation with a modified human protein, a molecule likely to exhibit few side effects. VI. The Role of Structural Information The discovery of thrombin inhibitors has benefited from available protein structural information. Models of the thrombin overall structure and its active site geometry, constructed from available structures of related serine proteases [30], aided in the design of the mechanism-based inhibitors such as PPACK [31] and its boroarginine analog [10]. The unexpected, nonsubstrate binding mode of early thrombin inhibitors such as NAPAP was revealed by x-ray crystallographic analyses [32]. Iterative structurebased design methods have been critical in the optimization of bivalent inhibitors and inhibitors directed at the primary specificity pocket. Structures of inhibitor:thrombin complexes are essential for the optimization of substitutents forming interactions within the aryl-binding site of the primary specificity pocket. In some cases (e.g. Table 1), seemingly minor alterations of the inhibitor can result in dramatic changes in the inhibitor's overall interactions with thrombin [17]. Drug discovery efforts have also been strongly influenced by results of structural studies of thrombin complexed with effectors and substrate peptides. For example, recently the structures of thrombin complexed with fibrinopeptide A [33] and human prothrombin fragment F1 [34] have been determined. In addition to their role in design of high-affinity inhibitors, these structures provide valuable insights for design of drugs specific for the various subsites and conformational states of thrombin. VII. Conclusion Discovery of therapeutically effective thrombin inhibitors involves issues such as affinity and selectivity, bioavailability, and formulation. In addition to these relatively common concerns, the complex in vivo mechanisms designed to

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_261.html [4/5/2004 5:11:30 PM]

Document

Page 262

balance its pro- and anticoagulant activities present additional challenges in the discovery of therapeutically effective thrombin inhibitors. References 1. Tapparelli C, Metternich R, Ehrhardt C, Cook NS. Synthetic low-molecular weight thrombin inhibitors: molecular design and pharmacological profile. TIPS 1993; 14:366–376. 2. Stubbs MT, Bode W. Structure and specificity in coagulation and its inhibition. Trends Cardiovasc Med 1995; 5:157–166. 3. Stone SR. Thrombin Inhibitors: A new generation of antithrombotics. Trends Cardiovasc. Med. 1995; 5:134–140. 4. Harker LA. Strategies for inhibiting the effects of thrombin. Blood Coagulation and Fibrinolysis 1994; 5:47–58. 5. Claeson G. Synthetic peptides and peptidomimetics as substrates and inhibitors of thrombin and other proteases in the blood coagulation system. Blood Coagulation and Fibrinolysis 1994; 5:411–436. 6. Bode W, Mayr I, Baumann U, Huber R, Stone SR, Hofsteenge J. The refined 1.9 Å crystal structure of human α-thrombin: interaction with D-Phe-Pro-Arg chloromethylketone and significance of the TyrPro-Pro-Trp insertion segment. EMBO J. 1989; 8:3467–3475. 7. Bode W, Turk D, Karshikov A. The refined 1.9-Å X-ray crystal structure of D-Phe-Pro-Arg chloromethylketone-inhibited human α-thrombin: structure analysis, overall structure, electrostatic properties, detailed active-site geometry, and structure-function relationships. Protein Science 1992; 1:426–471. 8. Stubbs MT, Bode W. A Player of many parts: the spotlight falls on thrombin's structure. Thrombosis Research. Vol. 69. Pergamon Press, 1993; 1–58. 9. Whinna HC, Church FC. Interaction of thrombin with antithrombin, heparin cofactor II and protein C inhibitor. Journal of Protein Chemistry 1993; 12:677–688. 10. Weber PC, Lee S-L, Lewandowski FA, Schadt MC, Chang C“ H, Kettner C. Kinetic and crystallographic studies of thrombin with Ac-(D)Phe-Pro-boroArg-OH and its lysine, amidine, homolysine and ornithine analogs. Biochemistry 1995; 34:3750–3757. 11. Deadman JJ, Elgendy S, Goodwin C, Green D, Baban J, Patel G, Skordalakes E, Chino N, Claeson G, Kakkar V, Scully M. Characterization of a class of peptide boronates with neutral P1 side chains as highly selective inhibitors of thrombin. J Med Chem 1995; 38:1511–1522.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_262.html (1 of 2) [4/5/2004 5:11:34 PM]

Document

12. Wu T-P, Yee V, Tulinsky A, Chrusciel R, Nakanishi H, Shen R, Priebe C, Kahn M. The structure of a designed peptidomimetic inhibitor complex of α-thrombin. Protein Engineering 1993; 5:471–478. 13. Obst U, Gramlich V, Diederich F, Weber L, Banner DW. Design of novel, nonpeptidic thrombin inhibitors and structure of a thrombin-inhibitor complex. Angew Chem Int Ed Engl 1995; 34:1739. 14. Iwanowicz EJ, Lau WF, Lin J, Roberts DGM, Seiler SM. Retro-binding tripeptide thrombin active inhibitors: discovery, synthesis and molecular modeling. J Med Chem 1994; 37:2122–2124.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_262.html (2 of 2) [4/5/2004 5:11:34 PM]

Document

Page 263

15. Tabernero L, Chang CÁ, Ohringer SL, Lau WF, Iwanowicz EJ, Han W-C, Wang T, Seiler S, Roberts D, Sack JS. Structure of a retro-binding peptide inhibitor complexed with human α-thrombin. J Mol Biol 1995; 246:14–20. 16. Maryanoff BE, Qiu Z, Padmanabhan KP, Tulinsky A, Almond HR, Andrade-Gordon P, Greco M, Kauffman J, Nicolaou KC, Liu A, Brung P, Fusetani N. Molecular basis for the inhibition of human αthrombin by the macrocyclic peptide cyclotheonamide A. Natl Academy Sci USA 1993; 90:8048–8052. 17. Hilpert K, Ackermann J, Banner DW, Gast A, Gubernator K, Hadvary P, Labler L, Muller K, Schmid G, Tschopp T, Van de Waterbeemd H. Design and synthesis of potent and highly selective thrombin inhibitors. J Med Chem 1994; 37:3889–3901. 18. Rydel TJ, Tulinsky A, Bode W, Ravichandran KG, Huber R, Roitsch R, Fenton JW, II. The structure of a complex of recombinant hirudin and human α-thrombin. Science 1990; 249:277–280. 19. Grutter MG, Priestle JP, Rahuel J, Grossenbacher H, Bode W, Hofsteenge J, Stone SR. Crystal structure of the thrombin-hirudin complex: a novel mode of serine protease inhibitor. EMBO J 1990; 9:2361–2365. 20. Rydel TJ, Tulinsky A, Bode W, Huber R. Refined structure of the hirudin-thrombin complex. J Mol Biol 1991; 221:583–601. 21. Maraganore JM, Bourdon P, Jablonski J, Ramachandran KL, Fenton JW, II. Design and characterization of hirulogs: a novel class of bivalent peptide inhibitors of thrombin. Biochemistry 1990; 29:7095–7101. 22. DiMaio J, Gibbs B, Munn D, Lefebvre J, Ni F, Konishi Y. Bifunctional thrombin inhibitors based on the sequence of hirudin. J Biol Chem 1990; 265:21698–21703. 23. Kettner C, Shaw E. D-Phe-Pro-Arg CH2C1—A selective affinity label for thrombin. Thromb Res 1979; 14:969–973. 24. DiMaio J, Ni F, Gibbs B, Konishi Y. A new class of potent thrombin inhibitors that incorporates a scissile pseudopeptide bond. FEBS 1991; 282:47–52. 25. DiMaio J, Gibbs B, Lefebvre J, Konishi Y, Munn D, Yue SY. Synthesis of a homologous series of ketomethylene arginyl pseudodipeptides and application to low molecular weight hirudin-like thrombin inhibitors. J Med Chem 1992; 35:3331–3341. 26. Zdanov A, Wu S, DiMaio Y, Konishi Y, Li Y, Wu X, Edwards B, Martin P, Cygler M. Crystal structure of the complex of human α-thrombin and nonhydrolyzable bifunctional inhibitors, hirutonin-2 and hirutonin-6. PROTEINS: Structure, Function and Genetics 1993; 17:252–265.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_263.html (1 of 2) [4/5/2004 5:11:39 PM]

Document

27. Tsuda Y, Cygler M, Gibbs BF, Pedyczak A, Fethiere J, Yue SY, Konishi Y. Design of potent bivalent thrombin inhibitors based on hirudin sequence: incorporation of nonsubstrate-type active site inhibitors. Biochemistry 1994; 33:14443–14451. 28. Rehse PH, Steinmetzer T, Li Y, Konishi Y, Cygler M. Crystal structure of a peptidyl pyridinium methyl ketone inhibitor with thrombin. Biochemistry 1995; 34:11537–11544. 29. Gibbs CS, Coutre SE, Tsiang M, Li WX, Jain AK, Dunn KE, Law VS, Tao CT, Matsumura SY, Mejza SJ, Paborsky LR, Leung LLK. Conversion of thrombin into an anticoagulant by protein engineering. Nature 1995; 378:413–416. 30. Greer J. Comparative model-building of the mammalian serine proteinases. J Mol Biol 1981; 153:1027–1042.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_263.html (2 of 2) [4/5/2004 5:11:39 PM]

Document

Page 264

31. Kettner C, Shaw E. D-PHE-PRO-ARGCH2Cl-1 selective affinity label for thrombin. Thrombosis Research 1979; 14:969–973. 32. Brandstetter H, Turk D, Hoeffken W, Grosse D, Sturzebecher J, Martin PD, Edwards BFP, Bode W. X-ray crystal structure of thrombin complexes with the benzamidine- and arginine-base inhibitors NAPAP, 4-TAPAP and MQPA: a starting point for elaborating improved antithrombotics. J Mol Biol 1992; 226:1085–1099. 33. Martin PD, Robertson W, Turke D, Bode W, Edwards BFP. The structure of residues 7–16 of the Aα-chain of human fibrinogen bound to bovine thrombin at 2.3 Å resolution. J Biol Chem 1992; 267:7911–7920. 34. Arni RK, Padmanabhan K, Padmanabhan KP, Wu TP, Tulinsky A. The structure of the non-covalent complex of prothrombin kringle 2 with PPACK-thrombin. Chem Phys Lipids; 1994; 67–68:59–66. 35. Stone SR, Hofsteenge J. Kinetics of the inhibition of thrombin by hirudin. Biochemistry 1986; 25:4622–4628. 36. Witting JI, Bourdon P, Maraganore JM, Fenton JW II. Hirulog-1 and -B2 thrombin specificity. Biochem J 1992; 287:663–664. 37. Bourdon P, Jablonski J, Chao BH, Maraganore JM. Structure-function relationships of hirulog peptide interactions with thrombin. FEBS 1991; 294:163–166. 38. Szewczuk Z, Gibbs BF, Yue SY, Purisima E, Zdanvo A, Cygler M, Konishi Y. Design of a linker for trivalent thrombin inhibitors: interaction of the main chain of the linker with thrombin. Biochemistry 1993; 32:3396–3404.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_264.html [4/5/2004 5:11:41 PM]

Document

Page 265

11 Design of Antithrombotic Agents Directed at Factor Xa William C. Ripka Corvas International, Inc., San Diego, California I. Introduction Serine proteases have long been recognized as important players in a number of biochemical processes and their specific and selective inhibition provides multiple therapeutic opportunities [1]. In particular, the blood coagulation process is the result of an amplified cascade of proteolytic events in which several specific zymogens of serine proteases in blood are activated sequentially by selective cleavages to produce active enzymes [2,3]. This process, in pathological circumstances, may lead to the formation of a thrombus—an insoluble matrix of fibrin and platelets. Thrombosis is a serious medical problem in the United States and Europe as exemplified by the fact that half the people who die each year die of cardiovascular related problems. While much recent work in antithrombotic therapeutic approaches has focused on inhibition of thrombin, the central role that Factor Xa plays in the coagulation response to vascular injury also makes it an ideal pharmacological target for antithrombotic drug development. The recent report of the x-ray crystal structure of native Factor Xa [4] allows, for the first time, a wellfounded structure-based drug design approach for inhibitors. A number of reviews describing the biology [5–10] and chemistry [11,12] of Factor Xa inhibitors have appeared. II. Coagulation Cascade In the coagulation cascade (Figure 1), a highly amplified process leads to the formation of thrombin, which is the primary mediator for the conversion of fibrinogen to fibrin, as well as activation of platelets through the thrombin

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_265.html [4/5/2004 5:11:42 PM]

Document

Page 266

Figure 1 The coagulation cascade.

receptor. Thrombin generation is itself, however, the result of Factor Xa in complex with Factor Va on a phospholipid surface (the prothrombinase complex) acting on prothrombin. Vascular injury is the initiating event in the coagulation process, causing the activation of Factor Xa by the Factor VIIa/tissue factor complex. Factor Xa is, therefore, a central and crucial enzyme directly leading to the production of thrombin and its inhibition should be effective in blocking thrombogenesis. As a consequence of its key role early in the coagulation cascade process Factor Xa represents a potentially valuable therapeutic target for potent and specific inhibition. III. Proof of Principle for a Factor Xa Inhibitor In recent years the method by which certain hematophageous organisms maintain blood flow during feeding has been determined. Interestly, several of these organisms utilize Factor Xa inhibitors to prevent coagulation [13–15]; the tick anticoagulant peptide (TAP), a small protein isolated from the Ornithidoros moubata tick [13], and antistasin isolated from the Haementeria officinalis leech [14] are both potent and selective inhibitors of Factor Xa. As expected, these molecules are effective antithrombotics in several animal models of thrombosis (Table 1) and provide an important proof of principle with regard to the potential effectiveness of Factor Xa inhibitors as therapeutic anticoagulants.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_266.html [4/5/2004 5:11:49 PM]

Document

Page 267 Table 1 Factor Xa Inhibitors (TAP, Antistasin) in Experimental Models of Thrombosis Rat and rabbit models of venous thrombosis [25] Canine model of high shear, coronary arterial thrombosis [6,26] Canine model of femoral arterial thrombosis [27,28] Rhesus monkey model of acute disseminated intravascular coagulation [29,30] Baboon model of platelet dependent arterial thrombosis [9,31,32]

In an alternate approach, it has been shown that a covalently blocked, activesite-modified Factor Xa (DEGR-Xa) [16] as well as a catalytically impaired recombinant form [17] can be effective anticoagulants in models of deep-vein thrombosis [18] and in canine arterial thrombosis models [8]. In these examples, the active-site-inactivated Factor Xa competes with the active form for incorporation into the active prothrombinase complex since the binding of Factor Xa to Factor Va in this complex is independent of the active site [20]. These studies with Factor Xa inhibitors suggest that inhibiting earlier in the coagulation cascade, as well as inhibiting the production of thrombin by inactivating Factor Xa in the prothrombinase complex, may have certain therapeutic advantages. IV. Factor Xa—Structure and Function Factor Xa is a 59 kilodalton protein synthesized in the liver and secreted into the blood as an inactive zymogen (Figure 2) [21]. Prior to secretion the singlechain molecule undergoes co- and posttranslational modifications including removal of a signal sequence [22–24], gamma carboxylation of several glutamic acids (Gla) in the N-terminus [33], beta hydroxylation of Asp63 [34], N-glycosylation at two sites [35], and cleavage at two sites, Arg139 and Arg142, to give a two-chain molecule [36]. The mature form of Factor X consists of a light chain (139 amino acids) and a heavy chain (303 amino acids) held together by a single disulfide (Figure 2). The Gla residues are responsible for calcium and phospholipid binding and the second EGF domain is thought to mediate binding to Factor VIIIa and Factor Va [37,38]. The heavy chain contains the catalytic domain with the prototypic serine protease active site triad, His226, Asp279, and Ser376. During coagulation, Factor X is converted to the active protease, Factor Xa, by a complex of Factor VIIa/tissue factor or a complex of Factor IXa/Factor, VIIIa/phospholipid, and calcium, both of which cleave a specific Arg-Ile bond to release an activation peptide (Figure 2) [39]. Similar to the activation of chymotrypsin, trypsin, and thrombin, the newly formed N-terminal Ile folds into the interior of the protein to form an ion pair at the active site with Asp]375 [39,40]. In the presence of calcium ions the newly formed Factor Xa associates with Factor Va on a phospholipid membrane surface to form the

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_267.html (1 of 2) [4/5/2004 5:11:51 PM]

Document

Page 268

Figure 2 Factor Xa structure. Residues of the catalytic triad (His226, Asp279, Ser376) are circled.

prothrombinase complex that rapidly converts prothrombin (PT) to thrombin by cleavage at two sites in PT, Arg271–Thr272 and Arg320–Ile321 [41]. The x-ray structure of human des(1–45) Factor Xa at 2.2 Å resolution has now been reported [4] (Figure 3). V. Natural Inhibitors of Factor Xa Several small, potent, and naturally occurring Factor Xa inhibitors—tick antico-agulant protein (TAP)[3], tissue factor pathway inhibitor (TFPI) [42], antistasin (ATS) [43], Ecotin [44,45]—have been isolated and characterized (Table 2). All but TAP apparently inhibit the enzyme in the extended substrate conformation referred to as the standard mechanism of inhibition (Figure 4) [47]. In this standard mechanism the inhibitor presents a conformationally constrained binding loop with a partial beta sheet motif to the target enzyme that mimics the required substrate conformation and, after binding to the enzyme, can undergo a reversible proteolytic hydrolysis at the reactive site peptide bond (P1–P1').

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_268.html [4/5/2004 5:11:59 PM]

Document

Page 269

Figure 3 Stereo ribbon diagram of Factor Xa [4]. Residues of the catalytic triad are shown (His57, Asp102, Ser195) as well as residues of specific interest for the binding of small molecules to the active site: Glu192; S4 pocket residues Tyr99, Phe174, Trp215; and S1 pocket residue, Asp189. Residues are designated with the chymotrypsin numbering.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_269.html [4/5/2004 5:12:37 PM]

Document

Page 270 Table 2 Naturally Occurring Inhibitors of Factor Xa Inhibitor

Source

Ki

Structural information

TFPI

Human

3 pM [49] 90 nM [50]

Ecotin

Escherichia coli

50 pM [44]

TAP

Ornithidoros moubata (tick)

Antistasin

Haementeria officinalis (leech)

61 pM [51]

(X-ray in progress) [52]

AcAP5

Ancylostoma caninum

43 pM [19]

homology to Ascaris lumbricoides var.suum [79,80]

135 pM [13]

X-ray; complex with trypsin [46] 2D-NMR [57,58]

Studies of these natural inhibitors can be useful in defining the active site requirements for Factor Xa inhibition, and importantly, can indicate the level of inhibition that may be necessary for an effective Factor Xa inhibitor, recognizing that TAP and antistasin have evolved to yield functional, in vivo antithrombotics. Table 3 shows the reactive-site sequences of these substratelike inhibitors as well as the cleavage site sequences recognized by Factor Xa in the activation of prothrombin (PT), Factor VII, and Factor V. A. Tissue Factor Pathway Inhibitor (TFPI) The mature tissue factor pathway inhibitor (TFPI) is a 276-residue protein consisting of three tandom domains with homology to the Kunitz-like protease

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_270.html (1 of 2) [4/5/2004 5:12:45 PM]

Document

Figure 4 Extended binding modes for substrates and inhibitors. Sites in the enzyme (S) and in the inhibitor (P) are designated by the Schecter-Berger notation [48].

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_270.html (2 of 2) [4/5/2004 5:12:45 PM]

Document

Page 271 Table 3 Active Site Sequences for Factor Xa Substrates and Inhibitors Substrates

P4

P3

P2

P1

P'1

P'2

P'3

P'4

PT271,272

Ile

Glu

Gly

Arg

Thr

Ala

Thr

Ser

PT320,321

Ile

Asp

Gly

Arg

Ile

Val

Glu

Gly

FVII

Pro

Gln

Gly

Arg

Ile

Val

Gly

Gly

FV

Lys

Lys

Tyr

Arg

Ser

Leu

His

Leu

Antistasin

Val

Arg

Cys

Arg

Val

His

Cys

Pro

Ecotin

Val

Ser

Thr

Met

Met

Ala

Cys

Pro

TFPI-II

Gly

Ile

Cys

Arg

Gly

Tyr

Ile

Thr

AcAP5

Cys

Arg

Ser

Arg

Gly

Cys

Leu

Leu

AcAP6

Cys

Arg

Ser

Phe

Ser

Cys

Pro

Gly

Inhibitors

inhibitors [42]. A potent inhibitor of both Factor VIIa and Xa as well as trypsin, TFPI does not, however, have significant activity against leukocyte elastase, urokinase, activated Protein C, tissue factor plasminogen activator, thrombin, or kallikrein [53,54]. The second Kunitz domain from the Nterminus of TFPI has been identified as primarily responsible for the Factor Xa inhibition while both the first and second domains contribute to inhibition of Factor VIIa [42] (Figure 5). The proposed mechanism for this Factor-Xa-dependent inhibition of FVIIa/tissue factor involves the formation of a quarternary FXa-TFPI-FVIIa/TF complex [42]. The recombinant, isolated second domain, TFPI-II, has a Ki for Factor Xa of 90 nM [50] compared to 3 pM [49] for the intact protein. The sequence of the P4–P5' region of the Factor Xa inhibitory second Kunitz domain (Table 3) has been incorporated into a prototypic Kunitz inhibitor, bovine pancreatic inhibitor (BPTI), to produce potent and selective Factor Xa inhibitors [75,76]. B. Antistasin (ATS) Antistasin is one of several anticoagulants isolated from the Mexican leech, Haementeria officinalis [14]. It is a 119-amino-acid cysteine-rich protein with a primary structure that shows a two-fold sequence symmetry suggesting the molecule possesses two separate and distinct domains [51]. Mutagenesis studies have shown that ATS binds to Factor Xa in a substratelike manner in the P3(Arg32–P'3(Cys37) regions and is cleaved only in the first domain [55].

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_271.html (1 of 2) [4/5/2004 5:12:47 PM]

Document

While the x-ray structure of antistasin has not been reported, it is known that a Factor-Xa-induced cleavage occurs between Arg34 and Val35 suggesting this peptide loop conforms to the conformationally rigid substratelike conformation suggested by other known protein inhibitors of serine proteases [55]. A hallmark of this mode of inhibition is the rigid structure around the cleaved

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_271.html (2 of 2) [4/5/2004 5:12:47 PM]

Document

Page 272

Figure 5 Predicted secondary structure of tissue factor pathway inhibitor (TFPI) showing the Factor Xa and Factor VIIa inhibitory domains. The arrows point to the P1 sites.

bond, often imposed by cysteine crosslinks constraining the two ends of the cleavage site to be in close proximity even after cleavage [47]. Antistasin has cysteines at the P2(Cys33) and P'3 (Cys37) positions. C. Tick Anticoagulant Peptide (TAP) The tick anticoagulant peptide (TAP) is a 60-amino-acid polypeptide isolated from the soft tick Ornithodorus Moubata and is a potent (Ki = 2–200 pM) and selective inhibitor of Factor Xa, both as the free enzyme and in the prothrombinase complex [13]. The TAP anticoagulant does not inhibit trypsin or other trypsinlike serine proteases and, importantly, is not cleaved by Factor Xa. The mechanism by which TAP inhibits Factor Xa appears to be unique and it apparently does not utilize the substratelike binding modes characteristic of antistasin and the Kunitz inhibitors. Mutagenesis studies have shown that the primary interaction of TAP with Factor Xa occurs at the N-terminous where Arg3 appears to play a key role [56]. The solution structure of TAP has been deter-

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_272.html [4/5/2004 5:12:58 PM]

Document

Page 273

Figure 6 Stereo diagram of the NMR solution structure of the tick anticoagulant peptide. The crucial N-terminus Arg3 is indicated along with the pattern of cysteine bonds.

mined by 2D-NMR studies [57,58] (Figure 6). With the exception of the region neighboring the Cys15—Cys39 bond in TAP, these studies support the originally proposed idea that there are significant structural similarities between TAP and Kunitz proteinase inhibitors [10]. Nevertheless, it is clear that TAP inhibitors FXa by a fundamentally different, and as yet, not fully understood mechanism. D. AcAP's In addition to ticks and leeches, other hematophagous organisms such as hook- worms have also evolved potent and selective Factor Xa inhibitors as anticoagulant strategies. Two such proteins, AcAP5 and AcAP6, have been isolated from the Ancylostoma caninum hookworm [19]. The AcAP5 protein is a 77amino-acid polypeptide with 10 cysteine residues. It inhibits the amidolytic activity of Factor Xa with a Ki of 43 ± 5 pM. Incubation of rAcAP5 with its target enzyme Factor Xa results in partial cleavage of the Arg40–Gly41 peptide bond suggesting the sequence around this cleavage site can adopt the restricted conformational requirements of substrates [47]. The AcAP6 protein is a 75-amino-acid polypeptide, also with cysteines, with a Ki for Factor Xa inhibition of 996 ± 65 pM. Alignment of the sequences of AcAP5 and AcAP6 suggest the P1 residue in AcAP6 is Phe38 and not the basic residue usually associated with Factor Xa specificity [19]. Substitution of Phe38 in AcAP6 with Arg resulted in a mutant that inhibited Factor Xa with a potency similar to rAcAP5. Both AcAP6 and ecotin suggest that an Arg or

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_273.html (1 of 2) [4/5/2004 5:13:03 PM]

Document

Page 274

Lys in the P1 position is not an absolute requirement for potent and selective activity. E. Ecotin Ecotin, a protein isolated from Escherichia coli, is a promiscuous protease inhibitor that potently inhibits kallikrein, urokinase, Factor XIIa, granzyme B, trypsin, chymotrypsin, and elastase (reviewed in Reference 46). As with most protein inhibitors (hirudin [59] and TAP [10] being the exceptions) ecotin presents a ‘baitlike’ substrate sequence to the target protease resulting in a reversible cleavage of the P1–P1' peptide bond. However, unlike other Factor Xa inhibitors that require basic residues such as Arg or Lys in the P1 position, ecotin is cleaved between two hydrophobic amino acids, Met84 and Met85 [44]. Three preliminary crystal structures of ecotin with the serine proteases chymotrypsin, trypsin, and fiddler crab collagenase have been described [46]. The conformation of the sequence around the reactive site is similar to the bovine pancreatic trypsin inhibitor (BPTI) but differs in that the Cys at P' (not P2 as in BPTI) provides the rigidifying function for the reactive-loop sequence. Interestingly, antistasin has cysteines at both the P2 and P3' positions. The Pro at P4' is commonly found in FXa inhibitors including antistasin. In its complexes with the serine proteases for which structures are available [46] there is a sub van der Waals contact between Met84-C and the enzyme Ser195-O. The Met84-O faces the oxyanion hole and forms hydrogen bonds with Ser195 and Gly193. In the trypsin structure the Met84 side chain extends into the S1 site in a manner similar to Lys15 in the BPTI-trypsin complex. Ecotin also forms both beta sheet hydrogen bonds to the enzyme Gly216, Ser82-N and O to Gly216 O and N. VI. Small Molecule Inhibitors of Factor Xa While the x-ray structure of native Factor Xa has been reported [4] the nature of its crystal packing, specifically the fact that the active site of one Factor Xa molecule is blocked by the N-terminus of a second resulting in a “continuous polymeric structure,” apparently has precluded diffusing inhibitors into the preformed crystals to obtain complexes. Complexes with inhibitors cocrystallized with Factor Xa also have not been reported [81]. Thus, efforts to do structure-based design with this enzyme have relied on molecular modeling. Since, to date, it has not been possible to directly obtain x-ray structures of inhibitor complexes with Factor Xa, the substantial information available with respect to how serine proteases, particularly thrombin, bind inhibitors can

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_274.html [4/5/2004 5:13:05 PM]

Document

Page 275

be utilized to model known Factor Xa inhibitors using the x-ray coordinates of native Factor Xa. Preliminary modeling of the several inhibitors described below into the Factor Xa active site was accomplished by the author [69] by first superimposing the backbone atoms (N, Ca, C) of the catalytic triads (His57, Asp102, Ser195 in the benzamidine:thrombin and PPACK:thrombin x-ray structures on the corresponding residues in Factor Xa. This allowed an excellent fit of the P1 basic groups, arylbenzamidine in the case of benzamidine and arginine for PPACK, into the S1 pocket of Factor Xa. These groups were then used as templates for positioning the appropriate P1 basic groups of the various synthetic inhibitors [69]. Holding these docked P1 groups fixed, the remaining rotatable bonds were manipulated to allow a reasonable and complementary fit of the inhibitor atoms to the solvent accessible surface of the Factor Xa active site. In those cases where it was possible, hydrogen bonds, particularly to Gly216, were formed. To fit extended peptide sequences such as that for antistasin and the antistasinderived peptides described below, the backbone atoms of the residues around the cleavage site (e.g., P4–P4') were positioned using the corresponding BPTI backbone atoms as a template. This was done after first aligning the trypsin catalytic triad backbone in the BPTI:trypsin x-ray complex (vida infra) to Factor Xa. Where the side chains of the bound peptide segments differed from BPTI, their orientation was either modeled for maximum complementarity to the Factor Xa molecular surface or set by an algorithmic approach [78]. In some cases the structures obtained were energy minimized initially by steepest descent followed by conjugate gradient minimization. Recently, compounds based on a bisamidine motif (e.g., DX-9065a) have been reported as potent and selective Factor Xa inhibitors (1, DX-9065a) [60–63].

The position of the amidino group makes little difference to the Factor Xa potency of these compounds but, interestingly, has a dramatic effect on the selectivity towards thrombin [62]. It was also observed that one carboxylic acid isomer (CX-9065a) was 7 times more potent on Factor Xa than the other. A second set of analogs shows a similar SAR [60].

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_275.html [4/5/2004 5:13:08 PM]

Document

Page 276

In both cases the acids are much more selective for Factor Xa over thrombin. An initial modeling study using a homology-built Factor Xa structure proposed a fit of DX9065a to Factor Xa in which the amidinoary1 group occupies the S1 pocket and the acetimidoyl group is directed out of the S4 pocket [60]. Lin et al. [64] have provided a more systematic study of possible fits of compound 2 and DX9065a using the recently available Factor Xa coordinates [4]. After aligning the His57, Ser195, and Asp102 backbone atoms for Factor Xa and thrombin (in the benzamidine:thrombin x-ray structure [65]) the arylamidino group of 2 was superimposed on the benzamidine template and a systematic conformational search was performed on the rotatable bonds of inhibitor 2. Energetics and complementarity to the Factor Xa surface determined a saved set, about 300 low-energy conformations, for further study. The final result was an optimized structure in which the acetimino group of 2 fits into

Figure 7 Molecular modeling fit of compound 2 with the arylamidino group positioned in the S1 pocket and the acetimino group in the S4 cation-π site [69].

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_276.html (1 of 2) [4/5/2004 5:13:12 PM]

Document

Page 277

the S4 pocket formed by the aromatic residues Trp215, Tyr99, and Phe174 (Figure 7). It was proposed that this collection of aryl residues forms the basis for a cation-π interaction that has recently been well documented in other cases [66]. Interestingly, this cation-π site is absent in thrombin's equivalent “aryl binding site” where the corresponding residues are Trp215, Leu99, and Ile174—residues that cannot provide the π-electrons necessary for stabilization of the cation in the inhibitor. The apparent preference for FXa inhibitors to have cations in the P3,P4 position may be directly related to the ability of the Factor Xa S4 site to stabilize these cations via a cation-π interaction. Factor Xa appears to be unique among the coagulation factors in providing this electron-rich S4 pocket (Table 4). The initial discovery of bisamidine structures as potential Factor Xa inhibitors was actually made much earlier with the finding that compounds such as 4 showed an almost 300-fold preference for Factor Xa over thrombin with a Ki of 13 nM (FXa) [67],

As with the DX-9065 analogs and compounds 2 and 3, the Factor Xa potency was relatively insensitive to the positioning of the amidino groups (4,4' versus 3,3') while replacing the 7-membered cycloalkyl ring with the 5 or 6 membered ring analogs reduced potency by about 10 fold. Model building compound 4 and docking into Factor Xa [69], again using the x-ray benzamidine:thrombin complex [65] as a template, shows that the second aryl amidino group can be positioned into the S4 aromatic pocket of Factor Xa in a conformation closely related to the mode of binding proposed for DX9065a (Figure 8) [64]. In an effort to compare the relative efficacy of thrombin versus Factor Xa inhibitors, Markwardt et al. [67] synthesized a set of amidinoaryl compounds with moderate potency as Factor Xa inhibitors (5). Table 4 S4 Residues in Selected Serine Proteases Residue Positiona

Factor Xa

Thrombin

Factor VIIa

Trypsin

99

Tyr

Leu

Thr

Leu

174

Phe

Ile

Pro

Gly

215

Trp

Trp

Trp

Trp

aChymotrypsin

numbering system. Sequence alignments by comparison of x-ray structures (sequence for Factor VIIa).

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_277.html (1 of 2) [4/5/2004 5:13:15 PM]

Document

Page 278

The n=3 chain length was optimal while the nature of the tosyl group on the α- nitrogen was relatively nonspecific, with tosylGly and Nα-β-naphthylsulfonyl- Gly being of similar potency. While the phenyl group on the amide nitrogen was best, other groups were also tolerated. Although the authors did not speculate on how these compounds were bound to Factor Xa, it is reasonable to suggest that the amidino phenyl group fits in the S1 pocket similar to the orientation determined by x-ray crystallography for benzamidine in the benzamidine:thrombin x-ray crystal structure. If the aryl amidino group of 5 is matched to that of benzamidine after alignment of the catalytic triad backbone atoms (N,Cα,C) for thrombin and Factor Xa, a proposed fit of 5 to Factor Xa can be made (Figure 9) [69]. This mode of binding is consistent with the structure activity relationships observed but does not suggest the reasons for the observed small preference for Factor Xa over thrombin.

Figure 8 Molecular modeling fit of compound 4 in the Factor Xa active site [69]. The carbonyl of the cycloheptanone makes a hydrogen bond (3.12 Å) to N-Gly216.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_278.html (1 of 2) [4/5/2004 5:13:20 PM]

Document

Page 279

Figure 9 Molecular modeling fit of compound 5 in the Factor Xa active site.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_279.html [4/5/2004 5:13:25 PM]

Document

Page 280

Figure 10 Proposed model of dansyl-Glu-Gly-Arg-chloromethyl ketone in Factor Xa [4,69].

Tulinsky and coworkers [4] have proposed a model for the complex of dansyl-Glu-Gly-Argchloromethyl ketone using the thrombin-PPACK crystal structure as a template for the fit to Factor Xa (Figure 10). Using antistasin as a starting point, Ohta et al. [68] have synthesized a series of cyclic peptides based on the antistasin sequence. Three of these peptides are shown below and represent the most potent in the series. Ki (FXa) ATS29-47

NH2-Ser-Gly-Val-Arg-Cys*-Arg-Val-His-Cys*-Pro-His-Gly-Phe-Gln-ArgSer-Arg-Tyr-Gly-OH

ATS29-40

NH2-Ser-Gly-Val-Arg-Cys*-Arg-Val-His-Cys*-Pro-His-Gly-OH

11.8 µM

dR-ATS32-38

NH2-dArg-Cys-Arg-Val-His-Cys-Pro-OH

0.96 µM

(The Cys*-Cys* are joined in disulfide bonds to form cyclic structures)

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_280.html (1 of 2) [4/5/2004 5:13:29 PM]

0.035 µM

Document

These molecules are cleaved by Factor Xa suggesting they bind in a similar manner to antistasin itself. Assuming the sequence around the cleavage site occupies the FXa active site locally in a manner similar to BPTI in the BPTI:trypsin complex, a modeled structure of the complex can be constructed

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_280.html (2 of 2) [4/5/2004 5:13:29 PM]

Document

Page 281

Figure 11 dArg-ATS32-38 modeled into the active site of Factor Xa utilizing the BPTI: trypsin x-ray structure as a template [69].

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_281.html [4/5/2004 5:13:34 PM]

Document

Page 282

using BPTI residues 11–19 as a template [69]. The modeled dR-ATS32-38:FXa complex is shown in Figure 11. Interestingly, these peptides do not inhibit trypsin even at 1000-fold higher concentrations than their FXa inhibitory concentrations. Antistasin, on the other hand, inhibits trypsin with a Ki of 10 nM. Preliminary reports have appeared describing a pentapeptide that is a potent (Ki=3 nM) and selective inhibitor of Factor Xa (SEL2711) [70,71].

Two possible modes of binding can be envisioned for this compound with either the methylpyridinium group occupying the S1 site in a “substratelike mode” or the p-amidinophenyl group in the S1 pocket, which would require a reversed binding reminiscent of the hirudin-thrombin interaction [59]. Figure 12 shows the case for the amidinophenyl group in the Factor Xa S1 pocket [69]. In this mode of binding the methylpyridinium group easily fits the S4-aryl binding site and is well positioned for a π-cation interaction. Of interest from a drug-design viewpoint is the finding that cyclotheonamide, a compound isolated from a marine sponge and originally reported as a thrombin inhibitor, has been found to also inhibit Factor Xa with a Ki of 50 nM [72]. Cyclotheonamide possesses a novel α-ketoamide transition state functionality and x-ray structures of cyclotheonamide with trypsin [73] and thrombin [74] provide templates for modeling this inhibitor into Factor Xa [69]. In the resulting fit (Figure 13) cyclotheonamide does not project functionality into the S4-cation-π site and would not be expected to show Factor Xa selectivity. VII. Defining the Requirements for Factor Xa Inhibition by Mutagenesis of BPTI It has been known for some time that many examples of naturally occurring Kunitz inhibitors exist, both isolated and as domains in larger proteins, which inhibit a variety of serine proteases [47]. This strongly suggests that this molecular framework is compatible with inhibition of this general class. The contact region between these inhibitors and their protease targets is known from a

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_282.html (1 of 2) [4/5/2004 5:13:38 PM]

Document

Page 283

Figure 12 Modeled fit of SEL2711 in Factor Xa with the arylamidino group positioned in the S1 pocket and the methylpyridinium group in the cation-π S4 site.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_283.html [4/5/2004 5:13:43 PM]

Document

Page 284

Figure 13 Cyclotheonamide modeled into Factor Xa utilizing the x-ray structure of cyclotheonamide:trypsin [73] and cyclotheonamide:thrombin [74] as templates [69].

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_284.html [4/5/2004 5:13:51 PM]

Document

Page 285

number of x-ray structures of complexes. Appropriate site-specific and random mutagenesis, particularly of the P3-P4' residues of a prototypic Kunitz inhibitor, bovine pancreatic trypsin inhibitor (BPTI), has been shown to result in potent and selective Factor Xa inhibitors [75,76]. A potent inhibitor of trypsin, kallikrein, and plasmin, BPTI does not inhibit Factor Xa. It binds to serine proteases such as trypsin in an extended substrate mode from residue 13 (P3) through 17 (P'2) [47]. A second loop from BPTI also extends into the active site bringing residues 34, 39, and 46 into contact with the protease-active site. In terms of spatial proximity of residues three clusters can be defined: cluster 1 (13,39); cluster 2 (11,17,19,34); and cluster 3 (16,18,20,46). While residue 39 is approximately in the same region of space as residue 13 (9.4 Å CB-CB) the CA rarrow.gif CB vectors are directed in different directions and substitution at 39 would not be expected to have a cooperative effect with residue 13. Residue 34 on the other hand is in a key position. It is centrally located between residues 11,17, and 19 with CB-CB distances of 5.6, 5.7, and 6.5 Å respectively, and its CA rarrow.gif CB vector converges with the corresponding vectors from these residues to a common point in space. This residue is therefore expected to have a substantial cooperative effect with the other residues of cluster 2. Finally residue 46 is close to residue 20 (CB—CB of 6.5 Å) although the CA rarrow.gif CB vectors are approximately parallel and cooperative effects are expected to be minimal. The BPTI residues 11,12,13,15–20, 34,39, and 46 were therefore the focus of the site-directed and random mutagenesis studies. Residue 14 is Cys in BPTI and was not modified in the mutants since it is required for structural reasons. A. Site Specific Mutagenesis As a starting point for the design of BPTI-based Factor Xa inhibitors, the second domain of TFPI (TFPIII) was used as a template [75,76]. Table 5 shows the results of site-directed mutagenesis of BPTI. Mutant 50cl is a direct analog of TFPI-II with the exception of the Lys at position 46. The finding that 4c2 and 4c10 are essentially equivalent in potency (Ki 2.8 versus 1.8 nM) and are identical Table 5

Site-Directed BPTI Mutants with Factor Xa Inhibition Ki(nM)

12

13

14

15

16

17

18

19

20

34

39

46

r-TFPI-II

90

Gly

Ile

Cys

Arg

Gly

Tyr

Ile

Thr

Arg

Lys

Leu

Glu

50cl

205

Gly

Ile

Cys

Arg

Ala

Tyr

Ile

Thr

Arg

Lys

Leu

Lys

4c2

2.8

Gly

Ile

Cys

Arg

Ala

Tyr

Ile

Thr

Arg

Val

Leu

Glu

4c10

1.8

Gly

Ile

Cys

Arg

Ala

Tyr

Ile

Thr

Arg

Val

Leu

Lys

57c1

1.6

Gly

Ile

Cys

Arg

Ala

Tyr

Ile

Ile

Arg

Val

Leu

Lys

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_285.html (1 of 2) [4/5/2004 5:13:54 PM]

Document

w.t. BPTI

>1 mM

Gly

Pro

Cys

Lys

Ala

Arg

Ile

Ile

Arg

Val

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_285.html (2 of 2) [4/5/2004 5:13:54 PM]

Arg

Lys

Document

Page 286

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_286.html [4/5/2004 5:14:20 PM]

Document

Page 287

cal in sequence with the exception of the Glu46 Lys switch, suggests this residue is of minor importance. Therefore, the potency of 50cl (within a factor of 3) is consistent with r-TFPI-II. On the other hand, the incorporation of Val for Lys at position 34 leads to a dramatic increase in potency (˜100 fold; cf. 4c10 and 4c2 with 50c1). From cluster 2 above this suggests the importance of P5, P'2,P'4 for potency. The changes in wild type BPTI that result in a potent Factor Xa inhibitor are Lys15 rarrow.gif Arg15, Arg17, rarrow.gif Tyr17, and Arg39 rarrow.gif Leu39. B. Random Mutagenesis 11Libraries of mutant BPTI were created by inserting mutagenic cassettes in the BPTI gene of filamentous phage PIII coat proteins [76]. These libraries produced large numbers of mutants (~106) with randomized amino acids in positions 11, 13, 16, 17, 18, 19, 20, 34, and 39. The mutants were panned against Factor Xa, which was affixed to a solid support by a nonneutralizing antibody and the most potent inhibitors were separately expressed as soluble proteins. By this process it was possible to determine consensus sequences at the reactive sites and to define the pharmacophore requirements of inhibitors of Factor Xa in both a functional and conformational sense from the P4 to the P5 positions. Inhibitor amino acid preferences from both site directed and random mutagenesis studies are shown in Table 6. VIII. Positional Requirements of Factor Xa Inhibitors (Table 6) Examination of models of BPTI-mutants bound to Factor Xa show the L-amino acids in the P3 position project into solvent. In the Factor Xa cleavage sites in thrombin these residues are polar and acidic (Glu, Asp); they are polar and basic in antistasin (Arg), and polar and neutral in Ecotin (Ser). The exception is TFPI- II, with this position occupied by Ile. The BPTI random mutant results are consistent with the TFPI-II case and show a preference for aliphatics or aromatics in this position. There is a hydrophobic pocket in the enzyme, formed by Trp215, Tyr99, and Phe174, that would be accessible to a D-residue in this position. The accessibility of the S2 pocket of the enzyme by P2 groups would be expected to be influenced by the orientation of Tyr99. As the x-ray structure shows its position this residue puts severe limitations on the size of the P2 group. Consistent with this is the observation that Gly is the sole residue in the FXa:thrombin cleavage sites. Little information is available from the BPTI mutants, TFPI-II, or antistasin, which all have a structural requirement for Cys at this position. Synthetic compounds show, however, that in potent inhibitors large bulky aromatics are, in fact, allowed at this position, a situation that requires Tyr99 to move out of the way [75].

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_287.html [4/5/2004 5:14:31 PM]

Document

Page 288

Additionally, the binding of BPTI mutants also requires Tyr99 to move because of potential severe interactions with the Cys14–Cys38 bridge of the BPTI mutants. While larger groups can be accommodated, structural requirements are fairly rigid in agreement with the limited mobility expected of the Tyr99. The apparent requirement for a P2-Gly in the larger substrates may suggest that in these cases extended binding occurs that does not permit movement of Tyr99. In the P1 position, as expected, a basic group is preferred. Interestingly, of the two naturally occuring basic amino acids, arginine is preferred over lysine. This is seen in both the BPTI mutants as well as the Arg rarrow.gif Lys switch in antistasin. This may be due to Ala190 in the S1 pocket of FXa, which cannot orient and stabilize the BPTI lysine analog as Ser190 does in trypsin. An interesting exception to the need for a basic group is in Ecotin where a methionine occupies this site. The x-ray of Ecotin with trypsin clearly shows this neutral residue in the P1 pocket, aligned very closely to that seen for lysine in the BPTI:trypsin complex, and proximal to the charged Asp189 [77]. Apparently, extended binding over the rest of the site compensates for this energetically unfavorable situation. In the P'1 position, the natural cleavage sites use Thr and Ile while Ecotin has Met. In contrast to thrombin, FXa lacks the 60-insertion loop and can accommodate large groups at this position. The BPTI mutants, however, are forced to use a small residue (Ala) because of steric hindrance from the Cys58Cys42 group residue 61 in the enzyme. The inhibitor TFPI-II has a Gly at P'1. The BPTI mutants, TFPI-II, and antistasin all show a preference for aromatic groups at P'2. In the BPTI panning experiments Tyr was selected more than 80% of the time at this position. It can be seen from Table 5 that the Arg to Tyr change at position 17 is one of three significant changes that converts wild type BPTI from a non-Factor Xa inhibitor to a ˜1.6 nM inhibitor. There is a possible hydrogen-bond interaction between Tyr17 of the inhibitor and Gln192 of the enzyme, which may explain the strong preference. While models suggest the P'3 residue is directed at solvent and the FXa thrombin cleavage sites have polar residues at this position (Thr, Glu), the BPTI mutant results show a clear preference for a hydrophobic group. It is possible that aromatic groups can pack to Phe41 of the enzyme. Mutant results show Ile is favored over Phe, His, which in turn is selected over Tyr.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_288.html (1 of 2) [4/5/2004 5:15:07 PM]

Document

Page 289

IX. Conclusions Factor Xa is clearly an important component of the coagulation process and inhibition of this enzyme can lead to potent anticoagulant effects. Recently, a number of naturally occurring anti-Factor-Xa polypeptides have been isolated from several hematophagous organisms including ticks, leeches, and hook- worms. With the exception of TAP, these molecules appear to bind to Factor Xa by the standard mechanism of inhibition proposed earlier [49]. The sequence information derived from these inhibitors as well as the natural cleavage sites of substrates of Factor Xa can be used along with the conformational constraints imposed by the proposed substrate-like binding to define the pharmacophore requirements of the active site of Factor Xa. The structurally rigid BPTI mutants, which have been found to be potent Factor Xa inhibitors, also provide important conformational information particularly with regard to the specific binding interactions on the P' side of the Factor Xa active site. A number of small molecule inhibitors have also recently been reported which appear to take advantage of a unique cation-π S4-site available in Factor Xa to achieve good selectivity with moderate potency. The availbility of the X-ray structure of native Factor Xa has allowed molecular modeling approaches to suggest possible fits of these inhibitors to the Factor Xa active site. Note Added in Proof After this review was written, the x-ray structure of Factor Xa with DX-9065a was reported [81]. References 1. Proteinase inhibitors. In: Barrett AJ, Salvesen G, eds. Research Monographs in Cell and Tissue Physiology. Vol 12. New York: Elsevier, 1986. Design of Enzyme Inhibitors as Drugs. Sandler M, Smith HJ, eds. New York: Oxford University Press, 1989. 2. Colman RW, Hirsh J, Marder VJ, Salzman EW. Hemostasis and Thrombosis. Basic Principles and Clinical Practice. Second Edition. Philadelphia: J. B. Lippincott Company, 1987. 3. Hathaway, WE, Goodnight, Jr SH. Disorders of Hemostasis and Thrombosis. New York: McGrawHill, 1993. 4. Padmanabhan K, Padmanabhan KP, Tulinsky A, Park CH, Bode W, Huber R, Blankenship DT, Cardin AD, Kisiel W. Structure of human Des(1–45) factor Xa at 2.2 Å resolution. J Mol Biol 1993; 232:947–966.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_289.html [4/5/2004 5:15:45 PM]

Document

Page 290

5. Kaiser B, Hauptmann J. Factor Xa inhibitors as novel antithrombotic agents: facts and perspectives. Cardiovascular Drug Reviews 1994; 12 (3):225–236. 6. Lynch JJ, Sitko GR, Mellott MJ, Nutt EM, Lehman ED, Friedman PA, Dunwiddie CT, Vlasuk GP. Maintenance of canine coronary artery patency following thrombolysis with front loaded plus low dose maintenance conjunctive therapy. A comparison of factor Xa versus thrombin inhibition. Cardiovasc Res 1994; 28:78–85. 7. Hollenback S, Sinha U, Lin P-H, Needham K, Frey L, Hancock T, Wong A, Wolf D. A comparative study of prothrombinase and thrombin inhibitors in a novel rabbit model of non-occlusive deep vein thrombosis. Thromb Haemost 1994; 71:357–362. 8. Benedict CR, Ryan J, Todd J, Kuwabara K, Tijburg P, Cartwright Jr J, Stern D. Active site-blocked factor Xa prevents thrombus formation in the coronaryvasculature in parallel with inhibition of extravascular coagulation in a canine thrombosis model. Blood 1993; 81:2059–2066. 9. Schaffer LW, Davidson JT, Vlasuk GP, and Siegl PKS. Antithrombotic efficacy of recombinant tick anticoagulant peptide. A potent inhibitor of coagulation factor Xa in a primate model of arterial thrombosis. Circulation 1991; 84:1741– 1748.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_290.html (1 of 2) [4/5/2004 5:16:10 PM]

Document

10. Vlasuk GP. Structural and functional characterization of tick anticoagulant peptide (TAP): a potent and selective inhibitor of blood coagulation factor Xa. Thrombosis and Haemostasis 1993; 70:212–216. 11. Scarborough RM. Anticoagulant strategies targeting thrombin and factor Xa. Ann Reports Med Chem 1995; 30:71–80. 12. Wallis RB. Inhibitors of coagulation factor Xa: from macromolecular beginnings to small molecules. Current Opinion in Therapeutic Patents Aug, 1993. 13. Waxman L, Smith DE, Arcuri KE, Vlasuk GP. Tick anticoagulant peptide (TAP) is a novel inhibitor of blood coagulation factor Xa, Science 1990; 248:593. 14. Hauptmann J, Kaiser B, Vowak G, Struzebecher J, Markwardt F. Comparison of the anticoagulant and antithrombotic effects of synthetic thrombin and factor Xa inhibitors. Throm Haemostas 1990; 63:220–223. 15. Jacobs JW, Cupp EW, Sardana M, Friedman PA. Isolation and characterization of a coagulation factor Xa inhibitor from black fly salivary glands. Thromb Haemost (GERMANY) 1990; 64:235–238. 16. Taylor Jr FB, Chang ACK, Peer GT, Mather T, Blick K, Catlett R, Lockhart MS, Esmon CT Blood 1991; 78:364. 17. Sinha U, Hancock T, Lin P-H, Hollenback S, Wolf D. Expression, purification, and characterization of inactive human coagulation factor Xa (Asn322Ala419). Protein Expr Purif 1992;3:518–524. 18. Dunwiddie CT, Waxman L, Vlasuk GP, Friedman PA. Purification and characterization of inhibitors of blood coagulation factor Xa from hematophagous organisms. Methods in Enzymol 1993; 233:291–312. 19. Stanssens P, Bergum PW, Gansemans Y, Jespers L, LaRoche Y, Huang S, Maki S, Messeno J, Lauwereys M, Cappello M, Hotez PJ, Lasters I, Vlasuk GP. Anticoagulant repertoire of the hookworm ancyclostoma caninum. Proc Natl Acad Sci 1996; 93:2149–2154. 20. Nesheim ME, Kettner C, Shaw E, Mann KG. Cofactor dependence of factor Xa incorporation into the prothrombinase complex. J Biol Chem 1981; 256:6537– 6540.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_290.html (2 of 2) [4/5/2004 5:16:10 PM]

Document

Page 291

21. Davie EW, Fujikawa K, Kisiel W. The coagulation cascade: interaction, maintenance, and regulation. Biochemistry 1991; 30:10363–10370. 22. Leytus SP, Foster DC, Kurachi K, Davie EW. Gene for human factor X: a blood coagulation factor whose gene organization is Essentially identical with that of factor IX and protein C. Biochemistry 1986; 25:5098–5102. 23. Fung MR, Hay CW, MacGillivray RTA. Characterization of an almost full length cDNA coding for human blood coagulation Factor X. Proc Nat Acad Sci USA 1985; 82:3591–3595. 24. Blanchard RA, Faye KLM, Barrett JM, William B. Isolation and characterization of profactor X from the liver of a steer treated with sodium warfarin. Blood 1985; 66(suppl.1):331a. 25. Vlasuk GP, Ramjit D, Fujita T, Dumwiddie CT, Nutt EM, Smith DE, Shebuski RJ. Comparison of the in vivo anticoagulant properties of standard heparin and the highly selective factor Xa inhibitors antistasin and tick anticoagulant peptide (TAP) in a rabbit model of venous thrombosis. Thromb Haemostas 1991; 65:257– 262. 26. Sitko GR, Ramjit DR, Stabilito II, Lehman D, Lynch JJ, Vlasuk GP. Conjunctive enhancement of enzymatic thrombolysis and prevention of thrombotic reocclusion with the selective factor Xa inhibitor, tick anticoagulant peptide. Comparison to hirudin and heparin in a canine model of acute coronary artery thrombosis. Circulation 1992; 85:805–815. 27. Mellot MJ, Strainieri MT, Sitko GR, Stabilito II, Lynch JJ, Vlasuk GP. Enhancement of recombinant tissue plasminogen activator-induced reperfusion by recombinant tick anticoagulant peptide, a selective factor Xa inhibitor, in a canine model of femoral arterial thrombosis. Fibrinolysis 1993; 7:195–202. 28. Mellot JJ, Holahan MA, Lynch JJ, Vlasuk GP, Dunwiddie CT. Acceleration of recombinant tissuetype plasminogen activator-induced reperfusion and prevention of reocclusion by recombinant antistasin, a selective factor Xa inhibitor, in a canine model of femoral arterial thrombosis. Circ Res 1992; 70:1152–1160.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_291.html (1 of 2) [4/5/2004 5:16:13 PM]

Document

29. Neeper MP, Waxman L, Smith DE, Schulman CA, Sardana M, Ellis RE, Schaffer LW, Siegel PKS, Valsuk GP. Characterization of recombinant tick anticoagulant peptide. A highly selective inhibitor of blood coagulation factor Xa. J Biol Chem 1990; 265:17746–17752. 30. Dunwiddie CT, Nutt EM, Vlasuk GP, Siegel PDS, Schaffer LW. Anticoagulant efficacy and immunogenicity of the selective factor Xa inhibitor antistasin following subcutaneous administration in the rhesus monkey. Thromb Haemostas 1992; 67:371–376. 31. Schaffer LW, Davidson JT, Vlasuk GP, Dunwiddie CT, Siegel PKS. Selective factor Xa inhibition by recombinant antistasin prevents vascular graft thrombosis in baboons. Arteriosclerosis and Thromb 1992; 12:879–885. 32. Kelly AP, Hanson SR, Dunwiddie CT, Harker LA. Circulation 1992; 86:411. 33. Suttie JW. Vitamin K-dependent carboxylase. Annu Rev Biochem 1985; 54:459– 477. 34. Fernlund P, Stenflo J. β-Hydroxy-aspartic acid in vitamin K-dependent proteins. J Biol Chem 1983; 258:12509–12512. 35. DiScipio RG, Hermodson MA, Davie EW. Activation of human factor X (Stuart factor) by a protease from Russell's viper venom. Biochemistry 1977; 16:5253– 5260.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_291.html (2 of 2) [4/5/2004 5:16:13 PM]

Document

Page 292

36. Leytus SP, Chung DW, Kisiel W, Kurachi K, Davie EW. Characterization of a cDNA coding for human factor X. Proc Nat Acad Sci USA 1984; 81:3699–3702. 37. Skogen WF, Esmon CT, Cox AC. Comparison of coagulation factor Xa and des(1–44) factor Xa in the assembly of prothrombinase. J Biol Chem 1984; 259:2306–2310. 38. Hertzberg MS, Ben-Tal O, Furie BC. Construction, expression, and characterization of a chimera of Factor IX and Factor X: the role of the second epidermal growth factor domain and serine protease domain in factor Xa binding. J Biol Chem 1992; 267:14759–14766. 39. Sigler PB, Blow DM, Matthews BW, Henderson R. Structure of crystalline α- chymotrypsin. II. A preliminary report including a hypothesis for the activation mechanism. J Mol Biol 1968; 35:143–164. 40. Bode W, Turk D, Karshikov A. The refined 1.9 Å X-ray crystal structure of D- Phe-Pro-Arg chloromethylketone-inhibited human α-thrombin. Structure analysis, overall structure, electrostatic properties, detailed active site geometry, structure function relationships. Protein Sci 1992; 1:426–471. 41. Mann KG, Nesheim ME, Church WR, Haley P, Krishnaswamy S. Surface-dependent reactions of the vitamin K-dependent enzyme complexes. Blood 1990; 76:1– 16. 42. Girard TJ, Warren LA, Novotny WF, Likert KM, Brown SG, Miletich JP, Broze Jr GJ. Functional significance of the Kunitz-type inhibitory domains of lipoprotein- associated coagulation inhibitor. Nature 1989; 338:518–520. 43. Nutt E, Gasic T, Rodkey J, Gasic G, Jacobs J, Friedman P, Simpson E. The amino acid sequence of antistasin. J Biol Chem 1988; 263:10162–10167. 44. Lauwereys M, Stanssens P, Lambier AM, Messens J, Dempsey E, Vlasuk GP. Ecotin as a potent factor Xa inhibitor. Thromb Haemostasis 1993; 69:864. 45. Seymour JL, Lindquist RN, Dennis MA, Moffat B, Yansura D, Reilly D, Wessinger ME, Lazurus RA. Ecotin is a potent anticoagulant and reversible tightbinding inhibitor of factor Xa. Biochemistry 1994; 33:3949–3958. 46. McGrath ME, Gillmor SA, Fletterick RJ. Ecotin: lessons on survival in a proteasefilled world. Protein Sci 1995;4:141–148. 47. Laskowski M, Kato I. Protein inhibitors of proteinases. Annu Rev Biochem 1980; 49:593–626. 48. Schechter I, Berger A. On the size of the active site in proteases. I. Papain. Biochem Biophys Res Commun 1967; 27:157–162. 49. Peterson LC. Progress in vascular biology, haemostasis and thrombosis. Abstracts, 1992 Zimmerman Conference. San Diego, California, Feb. 27–29, 1993. http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_292.html (1 of 2) [4/5/2004 5:16:15 PM]

Document

50. Broze Jr GJ. Trends Cardiovasc Med 1994; 2:72. 51. Hofmann KJ, Nutt EM, Dunwiddie CT. Site-directed mutagenesis of the leechderived factor Xa inhibitor antistasin. Biochem J 1992; 287:943. 52. Schreuder H, Arkema A, deBoer B, Kalk K, Dijkema R, Mulders J, Theunissen H, Hol W. Crystallization and preliminary crystallographic analysis of antistasin, a leech-derived inhibitor of blood coagulation factor Xa. J Mol Biol 1993; 231:1137–1138. 53. Broze Jr GJ, Warren LA, Novotny WF, Huguchi DA, Girard JJ, Miletich JP. The lipoproteinassociated coagulation inhibitor that inhibits the factor VII-tissue factor complex also inhibits factor Xa: insight into its possible mechanism of action. Blood 1984; 71:335–343.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_292.html (2 of 2) [4/5/2004 5:16:15 PM]

Document

Page 293

54. Broze Jr GJ, Girard TJ, Novotny WF. Regulation of coagulation by a multivalent Kunitz-type inhibitor. Biochemistry 1990; 29:7539. 55. Dunwiddie CT, Vlasuk GP, Nutt EM. The hydrolysis and resynthesis of a single reactive site peptide bond in recombinant antistasin by coagulation factor Xa. Arch Biochem Biophys 1992; 294:647–653. 56. Dunwiddie CT, Neeper MP, Nutt EM, Waxman L, Smith DE, Hoffman KJ, Lumma PK, Garsky VM, Vlasuk GP. Site-directed analysis of the functional domains in the factor Xa inhibitor tick anticoagulant peptide: identification of two distinct regions that constitute the enzyme recognition sites. Biochemistry 1992; 31:12126–12131. 57. Lim-Wilby MSL, Hallenga K, DeMaeyer M, Lasters I, Vlasuk GP, Brunck TK. NMR structure determination of tick anticoagulant peptide (TAP). Protein Sci 1995; 4:178–186. 58. Antuch W, Guntert P, Billeter M, Hawthorne T, Grossenbacher H, Wuthrich K. NMR solution structure of the recombinant tick anticoagulant protein (rTAP), a factor Xa inhibitor from the tick ornithodoros moubata. FEBS Lett 1994; 325:251–257. 59. Rydel TJ, Tulinsky A, Bode W, Huber R. Refined structure of the hirudin-thrombin complex. J Mol Biol 1991;221:583–601. 60. Katakura S-I, Nagahara T, Hara T, Iwamoto M. A novel Factor Xa inhibitor: structure-activity relationships and selectivity between Factor Xa and thrombin. Biochem Biophys Res Comm 1993; 197:965–972. 61. Hara T, Yokoyama A, Ishihara H, Yokoyama Y, Nagahara T, Iwamoto M. DX- 9065a, a new synthetic, potent anticoagulant and selective inhibitor for Factor Xa. Thrombosis and Haemostasis 1994; 71:314–319. 62. Nagahara T, Yokoyama Y, Inamura K, Katakura S-I, Komoriya S, Yamaguchi H, Hara T, Iwamoto M. J Med Chem 1994; 37:1200–1207. 63. Nagahara T, Kanaya N, Inamura K, Yokoyama Y. Aromatic amidine derivatives and salts thereof. Eur Pat App 0-540-051-A1. 64. Lin Z, Johnson ME. Proposed cation-π mediated binding by Factor Xa: a novel enzymatic mechanism for molecular recognition. FEBS Lett 1995; 370:1–5. 65. Banner DW, Hadvary P. Crystallographic analysis of 3.0 Å resolution of the binding to human thrombin of four active-site directed inhibitors. J Biol Chem 1991; 266:20085–20093. 66. Dougherty DA. Cation-π interactions in chemistry and biology: a new view of benzene, Phe, Tyr, and Trp. Science 1996; 271:163–167.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_293.html (1 of 2) [4/5/2004 5:17:35 PM]

Document

67. Sturzebecher J, Sturzebecher U, Vieweg H, Wagner G, Hauptmann J, Markwardt F. Synthetic inhibitors of bovine factor Xa and thrombin comparison of their anticoagulant efficiency. Thrombosis Res 1989; 54:245–252. 68. Ohta N, Brush M, Jacobs JW. Interaction of antistasin-related peptides with factor Xa: identification of a core inhibitory sequence. Thromb Haemostasis (GERMANY) 1994; 72:825–830. 69. The preliminary modeled structures of the synthetic inhibitors described in this review were constructed and energy minimized by the author using HyperChem (1995, Hypercube, Inc., Release 4.5). Docking of these inhibitors to the active site of Factor Xa was accomplished by the author using the x-ray coordinates of native Factor Xa [4] and INSIGHT II (Biosym Technologies, Inc.) and the approach outlined in Section 6.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_293.html (2 of 2) [4/5/2004 5:17:35 PM]

Document

Page 294

70. Seligmann B, Stringer SK, Ostrem JA, Al-Obeidi F, Wildgoose P, Walser A, Safar P, Safarova A, LoCascio A, Spoonamore J, Thorpe DS, Kasireddy P, Ashmore B, Strop P. SEL 2711: a specific, orally available, active-site inhibitor of Factor Xa discovered using synthetic combinatorial chemistry. Abstract, Sixth IBC International Symposium on Advances in Anticoagulants and Antithrombotics. Washington, D. C., Oct. 23–24, 1995. 71. Al-obeidi F, Lebl M, Safar P, Stierandova A, Strop P, Walser A. Factor Xa inhibitors, Patent Appl. WO 95/29189; 1995. 72. Lewis SD, Ng AS, Balwin JJ, Fusetani N, Naylor AM, Shafer JA. Inhibition of thrombin and other trypsin-like serine proteinases by cyclotheonamide A Thrombosis Research 1993; 70:173–190. 73. Lee AY, Hagihara M, Karmacharya R, Albers MW, Schreiber, SL, Clardy J. Atomic structure of the trypsin-cyclotheonamide A complex: lessons for the design of serine protease inhibitors. J Am Chem Soc 1993; 115:12619. 74. Marynoff BE, Qui X, Padmanabhan KP, Tulinsky A, Almond Jr HR, Andrade- Gordon P, Greco MN, Kauffman JA, Nicolaou KC, Liu A, Brungs PH, Fusetani N. Proc Natl Acad Sci USA 1993; 90:8048. 75. Ripka W, Brunck T, Stanssens P, LaRoche Y, Lauwereys M, Lambeir A-M, Lasters I, DeMaeyer M, Vlasuk G, Levy O, Miller T, Webb T, Tamura S, Pearson D. Strategies in the design of inhibitors of serine proteases of the coagulation cascade—factor Xa. Eur J Med Chem 1995; 30 (Suppl):88s–100s. 76. Lasters I, DeMaeyer M, Ripka W. Bovine pancreatic trypsin inhibitor derived inhibitors of Factor Xa. Pat Appl WO 94/01461; 1994. 77. McGrath ME, Erpel T, Bystroff C, Fletterick RJ. Macromolecular chelation as an improved mechanism of protease inhibition: structure of the ecotin-trypsin complex. EMBO 1994; 13:1502–1507. 78. Desmet J, DeMaeyer M, Hazes B, Lasters I. Nature 1992; 356:539. 79. Grasberger BL, Clore AM, Gronenborn GM. Structure 1994; 2:669–678. 80. Huang K, Strynadka NCJ, Bernard VD, Peanasky RJ, James MNG. Structure 1994; 2:679–689.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_294.html (1 of 2) [4/5/2004 5:17:42 PM]

Document

81. Brandstetter H, Kuhne A, Bode W, Huber R, von der Saal W, Wirthensohn K, Engh RA. J Biol Chem 1996; 47:29988–29992.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_294.html (2 of 2) [4/5/2004 5:17:42 PM]

Document

Page 295

12 Polypeptide Modulators of Sodium Channel Function as a Basis for the Development of Novel Cardiac Stimulants Raymond S. Norton Biomolecular Research Institute, Parkville, Victoria, Australia I. Introduction Cardiovascular diseases remain one of the major causes of premature death in western societies. Chronic congestive heart failure (CHF) in particular is a common disease with a poor prognosis, median survival times after the onset of heart failure being 1.7 years in men and 3.2 years in women [1]. Current treatment relies on diuretics to reduce fluid volume, vasodilators to decrease the work load of the heart, and positive inotropic agents to increase cardiac contractility [2]. The most commonly prescribed of the positive inotropes is the cardiac glycoside digoxin (Figure 1) [3]. Although this drug has been in therapeutic use for over two hundred years, its efficacy in patients with a sinus rhythm has remained controversial, and evidence for its beneficial effects is quite recent [3–5]. It is also possible that these beneficial effects are not due solely to the positive inotropic activity of digoxin and that its neurohormonal effects may also be important [2, 5–7] Nevertheless, digoxin remains a widely used drug [3] and it follows that a suitable replacement or adjunct would find access to a significant market worldwide. The incentive to develop such a replacement follows from the low therapeutic index of digoxin [8,9] and the relatively common occurrence of side effects due to digitalis toxicity. In the 1960s and 1970s, 20–30% of patients receiving digitalis experienced serious toxicity and about one quarter of this group died [6, 10]. Digitalis toxicity is manifest in CNS side-effects such as

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_295.html [4/5/2004 5:17:59 PM]

Document

Page 296

Figure 1 Structures of the positive inotropes digoxin [3,4], DPI 201-106 [15], and BDF 9148 [15–17]. In digoxin the R group is (O-2,6-dideoxy-β-D-ribo-hexopyranosyl-(1 rarrow.gif 4)-O-2, 6-dideoxy-β-D-ribo-hexopyranosyl-(1 rarrow.gif 4)-2, 6-dideoxy-β-D-ribo-hexopyranosyl)oxy. In DPI 201-106 the configuration at the hydroxyl-bearing carbon influences cardiac activity.

fatigue, visual disturbances, and anorexia, and in cardiac side-effects that depend on the nature and extent of the underlying heart disease [3]. Careful monitoring of digoxin serum levels and bioavailability have reduced the incidence of digitalis toxicity [3] and the recent introduction of digoxin-binding antibodies or antibody fragments has provided an effective means of treating severe digitalis toxicity [3,7]. Nevertheless the quest continues for a substitute for the cardiac glycosides in the treatment of chronic CHF. Positive inotropic compounds can be classified into three groups: cAMP generators, intracellular calcium regulators, and modulators of ion channels or pumps [11]. The cAMP generators such as dopamine, dobutamine, and milrinone (a phosphodiesterase inhibitor) may worsen ischemia, cause arrhythmias, and increase mortality [2,6]. Intracellular calcium modulators have not reached clinical use, possibly because of additional effects such as vasoconstriction,

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_296.html (1 of 2) [4/5/2004 5:18:07 PM]

Document

Page 297

whereas, calcium sensitizers such as EMD 57033 may be useful positive inotropic compounds, even in the diseased myocardium [12]. Ion channel modulation represents another approach to positive inotropy [13]. Sodium channel modulators increase Na+ influx and prolong the plateau phase of the action potential; sodium/calcium exchange then leads to an increase in the level of calcium available to the contractile elements, thus increasing the force of cardiac contraction [13,14]. Synthetic compounds such as DPI 201-106 and BDF 9148 (Figure 1) increase the mean open time of the sodium channel by inhibiting channel inactivation [15]. Importantly, BDF 9148 remains an effective positive inotropic compound even in severely failing human myocardium [16] and in rat models of cardiovascular disease [17]. Modulators of calcium and potassium channel activities also function as positive inotropes [13], but in the remainder of this article we shall focus on sodium channel modulators. II. The Anthopleurins Two decades ago “drugs from the sea” were the subject of high expectations and a good deal of effort in various centers around the world. The number of therapeutically useful compounds to have emerged from that effort has been rather limited, but with the advent of high-throughput screening it is likely that useful new leads will be found, even from species investigated previously. Notwithstanding, some valuable leads did emerge from work carried out in the 1970s, amongst which were the polypeptide cardiac stimulants known as the anthopleurins. These were isolated from sea anemones, where they are components of the animal's venom and are believed to have a function in defense and the capture of prey. The work that led to the isolation and characterization of these and related polypeptides from sea anemones is covered in earlier reviews [18,19] and will not be reiterated here. The best characterized of the anthopleurins is anthopleurin-A (AP-A), which was isolated from the northern Pacific sea anemone Anthopleura xan- thogrammica and consists of 49 residues cross-linked by three disulfide bonds [18,20]. It is active as a cardiac stimulant at nanomolar concentrations in vitro, making it some 200 fold more potent on a molar basis that digoxin. Its positive inotropic activity is not associated with any significant effects on heart rate or blood pressure [21], and in conscious dogs its therapeutic index is 7.5, which is about three-fold higher than that of digoxin [8]. Anthopleurin-A is active under conditions of stress and hypocalcaemia [18,22], as well as in ischemic myocardium where many other positive inotropes give equivocal results [23]. The profile of activity for AP-A suggests that it is a potentially valuable lead in the development of an alternative positive inotrope to digoxin

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_297.html [4/5/2004 5:18:14 PM]

Document

Page 298

for use in the treatment of chronic CHF. This chapter describes how this development is being tackled using the approach of structure-based drug design. A. Related Sea Anemone Toxins Anthopleurin-A is a member of a family of sea anemone polypeptides [24] (Figure 2) that is steadily increasing in number. These polypeptides have been classified into two groups, designated Types 1 and 2 [27], which are similar with respect to the locations of their disulfide bridges and a number of residues thought to play a role in biological activity or maintenance of the tertiary structure [24,27], but are distinguishable on the basis of sequence similarity (>>60% within each type but NH group (pepsin numbering) in much the same way as the P1' >C=O group of other isosteres (Figure 2). Aminoalcohols In principle, a good analog of the putative intermediate would be the aminal –CHOH–NH– group but this would be in equilibrium with the aldehyde and amino fragments. Interposing a methylene group between the hydroxymethyl and amino groups stabilizes the analog and may still allow tight binding to the enzyme. Such aminoalcohols (-CHOH-CH2-NH-) have been synthesized [25] and were shown to be potent inhibitors. Cocrystallisation of two such compounds extending from P1 to P3' with endothiapepsin allowed their bound structures to be solved at high resolution [26]. The bound structures revealed that despite the insertion of a methylene group in the analog a frameshift in the binding mode does not occur since the residue following the aminoalcohol occupies the S1' pocket. In contrast, the single amino acid, statine, replaces two residues of the substrate. The hydroxyl of the aminoalcohol (S-enantiomer) is bound symmetrically to both essential carboxyls as is the case for the hydroxyl of the statine and hydroxyethylene analogs. Glycols Incorporation of glycol or vicinal diol analogs of the peptide bond (-CH(OH)-CH(OH)-) has led to potent inhibitors and the x-ray structure for one such compound complexed with endothiapepsin is available [27]. The first hydroxyl in

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_326.html [4/5/2004 5:24:56 PM]

Document

Page 327

this analog interacts with the catalytic aspartate carboxyls in the same manner as statine or hydroxyethylene moieties; whereas, the second hydroxyl forms a hydrogen bond with the NH group of Gly 76, thereby mimicing the carbonyl oxygen at P1' of other analogs (see Figure 2). Phosphorus-Containing Analogs Aspartic proteinase inhibitors in which the scissile bond is replaced by a phosphinic acid group (shown below) have been reported [28].

These may mimic the tetrahedral intermediate more closely than statine or hydroxyethylene analogs. One of the oxygens binds to the carboxyl diad and the other resides adjacent to Tyr75 (pepsin numbering) forming a hydrogen bond with the outer oxygen of Asp32 [27]. This isostere is very effective against pepsin. However, it ionizes at physiological pH and the resulting anion is ineffective as an inhibitor of renin [29]. Fluoroketone Analogs and Implications for Catalysis Fluoroketone analogs (-CO-CF2-) have been reported [16, 30] and found to be substantially more potent than the unhalogenated statone molecules, presumably due to the ease of hydration and greater complementarity of the resulting hydrated gem-diol with the catalytic site. The structure of a difluorostatone inhibitor complexed with endothiapepsin [31] revealed interactions that indicate how the catalytic intermediate is stabilized by the enzyme (Figure 3). One hydroxyl of the hydrated fluoroketone associates tightly with the aspartate diad in the same position as the statine hydroxyl or the native solvent molecule and the other hydroxyl is positioned such that it hydrogen bonds to the outer carboxyl oxygen of Asp32. It has been suggested that the tetrahedral intermediate is uncharged, because if the carboxyl of Asp32 carries a negative charge instead, the latter can be stabilized by a full complement of hydrogen bonds donated by the gem-diol intermediate and surrounding protein atoms [31]. The current mechanistic proposals are based on the key suggestion by Suguna et al. [32] that, although transition state analogs appear to displace the active-site water molecule located between the two catalytic aspartate carboxyls, the more weakly bound substrate may not. Instead as the substrate binds, the water may be partly displaced to a position appropriate for nucleophilic attack on the scissile bond carbonyl. Details of the proposed mechanism are given in Figure 3.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_327.html [4/5/2004 5:25:04 PM]

Document

Page 328

Figure 3 The catalytic mechanism proposed by Veerapandian et al. [31] based on the x-ray structure of a difluoroketone (geminal-diol) inhibitor bound to endothiapepsin. A water molecule tightly bound to the aspartates in the native enzyme is proposed to nucleophilically attack the scissile-bond carbonyl. The resulting geminal-diol intermediate is stabilised by hydrogen bonds with the negatively charged carboxyl of aspartate 32. Fission of the scissile C-N bond is accompanied by transfer of a proton from Asp215 to the leaving amino group.

B. Complementarity of the Inhibitor Optimizing the fit of a ligand to its binding site improves the potency by burying lipophilic residues and by maximizing the number of van der Waals contacts, hydrogen bonds, and charge—charge interactions. The principles that apply to ligand binding are similar to those involved in protein folding. Inhibitor

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_328.html (1 of 2) [4/5/2004 5:25:14 PM]

Document

Page 329

binding involves displacement of hydrogen-bonded water molecules from both the ligand and the binding cleft. This process is entropy favored since waters in the solvent lattice are more disordered than those bound to protein. The change in enthalpy on forming NH…CO bonds in the complex from >CO…HOH and >NH…OH2 is favorable but relatively small [33]. The hydrophobic effect is therefore thought to play a dominant role in the energetics of binding with hydrogen bonds providing precise alignment of the ligand with respect to the catalytic apparatus. The main chain >CO and >NH groups from P3 to P3' of aspartic proteinase inhibitors are nearly always satisfied by hydrogen-bond interactions on formation of the complex. Therefore, given that polypeptides can form the same hydrogen bonds to the binding cleft regardless of amino acid sequence, differences in affinity for ligands of equal length must be due to other interactions at the specificity pockets, presumably those between the ligand's side chains and the enzyme. One example of optimizing these interactions for renin is the use of the cyclohexylmethyl side chain at P1, which has been shown to improve the potency by two orders of magnitude relative to the equivalent leucine-containing inhibitor [13]. Structure/Activity Relationship (SAR) studies have shown that in many inhibitor types, the cyclohexylmethyl group is optimal for the S1 pocket of human renin; whereas, other analogs such as cyclohexyl, cyclohexylethyl, and the very bulky dicyclohexyl and adamantyl rings generally have significantly reduced potency [10]. The use of a cyclohexylmethyl appears to introduce selectivity for renin versus other human aspartic proteinases. This has been partly rationalized for endothiapepsin where it was shown by x-ray analysis that the cyclohexylmethyl group at P1 can force the Phe at P3 to adopt a less energetically favorable X2 angle. Hence, differences at the S3 pocket in renin may allow the P3 Phe to adopt a more favorable X2 angle in the presence of a cyclohexyl at P1. In contrast the S2 site is able to accommodate a wide variety of side chains depending on inhibitor type, e.g., Phe and His are equipotent in some analogs [34]. The x-ray structures of a number of bound renin inhibitors complexed with endothiapepsin have shown that His at P2 can adopt different X1 angles separated by about 120 degrees [8]. In one conformation the imidazole is lying partly in the S1' pocket, which has a definite hydrophobic character. In the other conformation, the His side chain is in a more polar environment. The ability of aspartic proteinases to accept a variety of both polar and hydrophobic groups at the P2 position may be due to this bifurcation. Many inhibitors possess naphthylalanine side chains at P3 and P4 [14,24,35]. Compounds of this type are potent renin inhibitors with binding constants in the nanomolar range. Cocrystallisation of such an inhibitor with endothiapepsin revealed that one naphthalic ring is accommodated in the S3 pocket by significant conformational changes of local enzyme side chains (Asp77 and Asp114). The other naphthalene lies in the S4 binding region [36].

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_329.html [4/5/2004 5:25:18 PM]

Document

Page 330

C. Rigidification The number of conformations that a peptide can adopt in solution is reduced by cyclization. This can be optimized, at least in theory, to lock the peptide in the conformation that has the highest affinity for the receptor, resulting in a gain of affinity, primarily for entropic reasons. The structures of the fungal aspartic proteinases reveal that the binding cleft is a wide channel with no obvious division between the pockets, e.g., S1 and S3 are contiguous. The bound structures of numerous inhibitors have shown that alternate side chains are in van der Waals contact due to the extended conformation that these ligands adopt. In addition, at certain positions, e.g., P2, the side chains are allowed very different conformations due to the permissiveness of the pocket. Hence, the cross-linking of certain side chains may, at least, not be detrimental to inhibitory potency and may also reduce the susceptibility to degradation in the gut or plasma. It might therefore be expected that oligopeptide renin inhibitors would be suitable' candidates for cross-linking experiments. A similar philosophy of rigidification was pursued in the development of the ACE inhibitor cilazapril [37]. A number of statine-containing inhibitors possessing disulphide links between P2 and P5, and P2 and P4' have been synthesized [13] although the best potencies were slightly less than for the linear peptides. An alkyl cycle of varying length was introduced between the hydroxyl of a serine residue at P1 and the main chain nitrogen of P2 in a series of reduced-bond inhibitors [53]. Potencies similar to the uncrosslinked molecule were achieved but none were greater. This was attributed to the cis isomerisation of the P3—P2 peptide bond giving a conformation that cannot fit the active site of the enzyme. Difficulties in achieving more potent cyclic inhibitors may be due to the tight binding environment provided by some pockets (especially S1 and S3), and the possibility that other unproductive conformations of the inhibitor become favorable. More recently similar findings have been reported for cyclic analogs of pepstatincontaining alkyl crosslinks of variable length between the P1 and P3 side chains [38]. D. In Vivo Stability Peptides, when administered orally, are susceptible to degradation in the stomach by gastric enzymes and the proteinases of the pancreas and brush border of the small intestine. Their lifetimes in the plasma are often short due to rapid proteolysis and other metabolic processes. Early efforts were made to improve the resistance of renin inhibitors to hydrolysis in vivo by the use of blocking groups at the Nand C-terminii [39] and replacement of susceptible peptide bonds other than the renin cleavage site. Studies of SAR have shown that various N- and C-terminal groups, some based on the morpholine nucleus and derivatives of it, have a favorable effect on the duration of inhibition in the

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_330.html [4/5/2004 5:25:25 PM]

Document

Page 331

plasma. This may arise from reduced nonspecific plasma binding due to the relatively polar nature of these blocking groups [24]. Resistance of inhibitors to gut proteolysis has been improved by various methods such as replacement of phenylalanine at P3 with O-methyl tyrosine (or naphthylalanine), which was shown to abolish chymotrypsin cleavage and yet retain high inhibitory potency for renin [40]. III. Structural Studies of Rennin Complexed With Inhibitors The three-dimensional structures of renin-inhibitor complexes had long been sought as an aid to the discovery of clinically effective antihypertensives [41]. X-ray analyses of recombinant human renin [42] and mouse submandibulary renin [43] have given an accurate picture of active-site interactions and largely confirm the predictions of models based on homologous aspartic proteinases [4]. A large number of questions concerning the specificities of renins have been answered by these x-ray analyses. The renin-inhibitor structures also make an important contribution towards the rational design of effective antihypertensive agents. A. X-Ray Analysis of Mouse and Human Renin Complexes For both of these renins multiple copies of the molecules have been independently defined in the x-ray analysis and shown to have very similar structures. These x-ray structures were refined to final agreement factors and correlation coefficients of 0.19 and 0.91 for human renin at 2.8 Å resolution and 0.18 and 0.95 for mouse renin at 1.9 Å resolution. As expected from the high degree of sequence identity of human and mouse renins (approximately 70%), they have very similar three-dimensional structures as shown in Figure 1. The active-site cleft has a less open arrangement in renins than in the other aspartic proteinases. Many loops as well as the helix hc (residues 224–236) belonging to the C-domain (residues 190–302) are significantly closer to the active site in the renin structures compared to those of endothiapepsininhibitor complexes. This is partly due to a difference in relative position of the rigid body comprising the C-domain. For instance, there is a domain rotation of ~4° and translation of ~0.1 Å in the human renin complex with respect to the endothiapepsin-difluorostatone complex. The entrance to the active-site cleft is made even narrower in renins as a consequence of differences in the positions and composition of several well-defined loops and secondary structure elements. Unique to the renins is a cis proline, Pro111, which caps a helix (hN2) and contributes to the subsites S3 and

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_331.html [4/5/2004 5:25:30 PM]

Document

Page 332

S5. This helix is nearer to the active site in renins than in other aspartic proteinases. On an equivalent loop in the C-lobe (related by the intramolecular pseudo 2-fold axis), there is a sequence of three prolines—the Pro292–Pro293– Pro 294 segment. This structure is also unique to the renins among the aspartic proteinases with Pro294 and Pro297 in a cis configuration. Such a proline-rich structure provides an effective means of constructing well-defined pockets from loops that would otherwise be more flexible. This rather rigid poly-proline loop, together with the loop comprised of residues 241–250, lies on either side of the active site “flap” formed by residues 72–81. Hence, in the renins, the cleft is covered by the flaps from both lobes rather than from the N-lobe alone as in other pepsin-like aspartic proteinases. This gives renin a superficial similarity to the dimeric, retroviral proteinases where each subunit provides an equivalent flap that closes down on top of the inhibitor [44,45]. B. The Role of Hydrogen Bonds in Inhibitor Recognition Whereas the mouse renin inhibitor extends from P6 to P4', the human renin inhibitor extends only from P4 to P1'. The cyclohexyl norstatine residue at P1 in the human renin inhibitor mimics a dipeptide analog with its isopropyloxy group occupying the subsite for the side chain of P1'. The mouse renin inhibitor (CH-66) possesses a Leu-Leu hydroxyethylene transition state analog [12]. Both inhibitors are bound in the extended conformation that is found in other aspartic proteinase-inhibitor complexes. Both inhibitors make extensive hydrogen bonds with the enzymes as shown in Figure 4. In general the two renininhibitor complexes described here demonstrate that a similar pattern of hydrogen bonding is probably used in the substrate recognition of all aspartic proteinases although their specificities differ substantially. There is also great similarity between aspartic proteinases in terms of interactions with the transitionstate analog inhibitors at the catalytic center. The catalytic aspartyl side chains and the inhibitor hydroxyl group are essentially superimposable in both renin complexes. The isostere C-OH bonds lie at identical positions when the structures of inhibitor complexes of several aspartic proteinases are superposed, in spite of the differences in the sequence and secondary structure. Most of the complex array of hydrogen bonds found in endothiapepsin complexes are formed in renin with the exception of that to the threonine or serine at 218, which is replaced by alanine in human renin. The similarity can be extended to all other pepsin-like aspartic proteinases and even to the retroviral proteinases [44,45]. This implies that the recognition of the transition state is conserved in evolution, and the mechanisms of this divergent group of proteinases must be very similar.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_332.html [4/5/2004 5:25:32 PM]

Document

Page 333

Figure 4 The inhibitors complexed with human (top) and mouse (bottom) renin showing the putative hydrogen-bond interactions made with the enzyme moieties.

C. Specificity If the main-chain hydrogen bonding of substrates is conserved among aspartic proteinases, how are the differences in specificities achieved? Table 1 defines the enzyme residues that line the specificity pockets for both mouse and human renin. In modeling exercises (e.g., Reference 4) it was assumed that specificities derive from differences in the sizes of the residues in the specificity pockets (Sn) and their ability to complement the corresponding side chains at positions Pn in the substrate/inhibitor. A detailed analysis now shows that this simple assumption only partly accounts for the steric basis of specificity.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_333.html (1 of 2) [4/5/2004 5:25:42 PM]

Document

Page 334

For example, in the specificity subsite S3 the phenyl rings of Phe P3 occupy almost identical positions in both renin inhibitor complexes. Modeling studies have predicted the specificity subsite S3 to be larger in renins than in other aspartic proteinases [4] due to substitution of smaller residues, Pro 111, Leu114, and Ala115, in place of larger ones in mammalian and fungal proteinases. However, a compensatory movement of a helix (hN2) makes the pocket quite compact and complementary to the aromatic ring as shown in Figure 5. Thus, the positions of an element of secondary structure differ between renin and other aspartic proteinases with a consequent important difference in the specificity pocket. The differing positions of secondary structural elements may also account for the specificities at P2'. Mouse submaxillary and other nonprimate renins do not appreciably cleave human angiotensinogen or its analogs [46], which have an isoleucine at P2', although they do cleave substrates with a valine at this position. In contrast, human renin not only cleaves the human and nonprimate substrates but also the rat angiotensinogen with tyrosine at P2', albeit rather slowly [47]. This can be explained in terms of the threedimensional structures. In the mouse renin complex, the P2' tyrosyl ring is packed parallel to an adjoining helix (h3) in a narrow pocket and there is only limited space available beyond the Cβ methylene group. This appears to be able to accommodate a valine, but not the larger isoleucine at P2', which will suffer greater steric interference from several residues that are conserved in identity and position in the two renins. On the other hand, in human renin differences in the orientation and position of

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_334.html [4/5/2004 5:26:06 PM]

Document

Page 335

Figure 5 The S3 specificity pocket of human renin occupied by phenylalanine in the cyclohexylnorstatine inhibitor.

helix h3 bring it closer by (by ˜0.5Å) to the substrate-binding site than in mouse renin. It is orientated in such a fashion in human renin that, although it can accommodate the isobutyl side chain of isoleucine at P2', aromatic rings on substituents such as phenylalanine and tyrosine will have severe short contacts with the side chain of Ile130 (valine in mouse renin). Thus the reorientation of a helix, coupled with subtle differences in the shapes of the side chains, makes significant changes in the substrate specificity at this subsite. It is interesting to note that in pepsin this helix is in a similar position with respect to the active site as in human renin. This provides a structural rationale for the negative influence of peptides containing phenylalanine [48], tyrosine, or histidine [49] at this subsite (S2') on the rate of proteolytic pepsin cleavage, while isoleucine and valine enhance catalysis. Differences in the specificity subsites at S1' in the human and mouse renins have a more complicated explanation. At first sight the situation appears to be explained by complementarity of the subsites to the valine and leucine at http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_335.html (1 of 2) [4/5/2004 5:26:14 PM]

Document

Page 336

P1' in human and mouse angiotensinogens. Accordingly residue 213 is leucine in human renin and valine in mouse renin. The S1' pockets of chymosin, pepsin, and endothiapepsin have an aromatic side chain at residue 189 while the renins have amino acids with smaller side chains (valine in human and serine in mouse renins). This would be expected to make the pocket larger in renins. However the structure of the mouse renin complex shows that the substrate moves closer to the enzyme in renins as a result of the smaller residue at 189 and the pocket is made even more compact due to a compensatory change in the position and composition of the polyproline loop (residues 290–297). Thus, the specificity difference at this site arises not only from a compensatory movement of a secondary structure, in this case a loop region, but also from the substitution of an enzyme residue that allows the substrate to come closer to the body of the enzyme. Elaboration of loops on the periphery of the binding cleft in renins also influences the specificity. This is most marked at P3' and P4', for which it has been particularly difficult to obtain complexes with welldefined conformations for other aspartic proteinases. In endothiapepsin, which has been the subject of the greatest number of studies, different conformations are adopted at P3' and the residue at P4' is generally disordered. In contrast these residues are clearly defined in mouse renin. This is mainly a consequence of the polyproline loop, illustrated in Figure 6, which occurs uniquely in renins. The x-ray analysis of the mouse renin complex shows that the S3' and S4' subsites are formed by the polyproline loop together with residues of the flap, and a similar situation is likely to occur in human renin. The welldefined interactions of P3' described in the mouse renin complex explains the significant affinity when inhibitors have phenylalanine or tyrosine at P3' as well as the importance of a P3' residue for catalytic cleavage of a substrate by renin [50]. Hydrogen bonds between the side chains of the inhibitor and the enzyme do not play a major role in most specificity pockets. However, S2 is an exception. This subsite is large and contiguous with S1', so that in human renin the S-methyl cysteine (SMC) side chain of P2 is oriented towards the S1' pocket, which is only partly filled by the isopropyloxy group of the putative P1' residue. The carbonyl oxygen of P2 accepts a hydrogen from the Oγ of Ser76, which is unique to human renin; residue 76 is a highly conserved glycine in all the other aspartic proteinases, including mouse renin. In mouse renin the P2 histidyl group has a different orientation and forms a hydrogen bond with the Oγ of Ser222. If such a conformation were adopted by the human angiotensinogen in complex with human renin, the two imidazole nitrogens would be hydrogen bonded to the Oγ of both Ser76 and Ser222. The observed reduction in the rate of cleavage of a human angiotensinogen analog containing a 3-methyl histidine substituent at P2 [51] could be explained on the basis of the hydrogen bonding scheme proposed above.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_336.html [4/5/2004 5:26:17 PM]

Document

Page 337

Figure 6 The P3' tyrosine residue of the mouse renin inhibitor complex showing the unique polyproline loop on the right. Specificity of this and neighboring subsites in renins must derive partly from this rigid loop region.

IV. Rational Drug Design The pioneering work of Burton, Szelke, and others in developing peptide-based renin inhibitors has been followed by a worldwide commercial effort to elaborate such compounds into therapeutically active antihypertensives. The twin problems of insufficient oral bioavailability and rapid clearance has seemingly presented major obstacles to success. In addition, the possible advantages of renin inhibitors compared with ACE inhibitors remain questionable. Never-theless information from human-renin crystallographic studies—such as the more recent high resolution analyses [52] and algorithms for analysing voids in the complexes as potential sites for elaborating the drug molecule (e.g. Figure 7)—may yet provide leads for compounds with suitable therapeutic characteristics. The detailed analyses of renin-inhibitor complexes reported here confirm the general structural features thought to contribute to renin's specificity but

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_337.html (1 of 2) [4/5/2004 5:26:24 PM]

Document

Page 338

Figure 7 Schematic illustration of the voids between enzyme and inhibitor in the crystal structure of human renin complexed with a norstatine inhibitor. The figure was produced using the GapE software (Dr. Roman Laskowski). The inhibitor (dark bonds) is enclosed by a net surface and the gaps (where the enzyme and inhibitor are not in contact) are represented by solid surfaces [42].

demonstrate the need for careful, high-resolution x-ray analyses for more confidence in drug design. In particular, they show that even minor alternations in the positions of secondary structural elements can lead to major changes in the disposition of the subsites and thus the recognition of substrates. Since such molecular recognition defines the species specificity and determines the catalytic efficiency of the enzymes, a through understanding is indispensable for the synthesis of suitable inhibitors. The specificity pockets—the molecular recognition sites—are modified by elaboration, particularly of surface loops, which can be disordered in the uncomplexed enzymes and difficult to model with precision from homologous structures. These data establish a new foundation for the rational design of renin inhibitors and have provided a rational base for development of clinically successful HIV proteinase inhibitors. References 1. Ondetti MA, Cushman DW. Enzymes of the renin-angiotensin system and their inhibitors. Annu Rev Biochem 1982; 51:283–308.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_338.html (1 of 2) [4/5/2004 5:27:09 PM]

Document

2. Blundell TL, Cooper J, Foundling SI, Jones DM, Atrash B, Szelke M. On the rational design of renin inhibitors: X-ray studies of aspartic proteinases complexed with transition state analogues. Biochemistry 1987; 26:5585–5590.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_338.html (2 of 2) [4/5/2004 5:27:09 PM]

Document

Page 339

3. Pearl LH, Blundell TL. The active sites of aspartic proteinases. FEBS Lett 1984; 174:96–101. 4. Sibanda BL, Blundell TL, Hobart PM, Fogliano M, Bindra JS, Dominy BW, Chirgwin JM. Computer graphics modeling of human renin. FEBS Lett 1984; 174:102–111. 5. Skeggs LT, Dover FE, Levine M, Lentz KE, Kahn JR. In: Johnson JA, Anderson RR, ed. The ReninAngiotensin System. New York: Plenum. 6. Bone R, Silen JL, Agard DA. Structural plasticity broadens the specificity of an engineered protease. Nature 1989; 339:191–195. 7. Haber E, Burton J. Inhibitors of renin and their utility in physiologic studies. Fedn Proc 1979; 38:2768–2773. 8. Cooper JB, Bailey D. A structural comparison of 21 inhibitor complexes of the aspartic proteinase from Endothia parasitica. Protein Science 1994, 3:2129–2143. 9. Foundling SI, Cooper, J, Watson FE, Cleasby A, Pearl LH, Sibanda BL, Hemmings A, Wood SP, Blundell TL, Valler MJ, Norey CG, Kay J, Boger J, Dunn BM, Leckie BJ, Jones DM, Atrash B, Hallett A, Szelke M. High resolution X-ray analyses of renin inhibitor-aspartic proteinase complexes. Nature (London) 1987; 327:349–352. 10. Luly JR, Bolis G, Bamaung N, Soderquist J, Dellaria JF, Stein H, Cohen J, Thomas JP, Greer J, Plattner JJ. New inhibitors of human renin that contain novel replacements. Examination of the P1 site. J Med Chem 1988; 31:532–539. 11. Hofmann T, Fink AL. Cryoenzymology of penicillopepsin. Biochemistry 1984; 23:5249–5256. 12. Szelke M, Leckie B, Hallett A, Jones DM, Sueiras-Diaz J, Atrash B, Lever AF. Potent new inhibitors of human renin. Nature 1982; 299:555–557. 13. Boger J. Renin inhibitors. Design of angiotensin transition state analogues containing statine. In: Kostka V. ed. Aspartic Proteinases and Their Inhibitors. Berlin: Walter de Gruyter, 1985:401–420. 14. Kokubu T, Hiwada K, Murakami E, Imamura Y, Matsueda R, Yabe Y, Koike H, Iijima Y. Highly potent and specific inhibitors of human renin. Hypertension 1985; 7 (suppl.1):8–11. 15. Kokubu T, Hiwada K, Nagae A, Murakami E, Morisawa Y, Yabe Y, Koike H, Iijima Y. Statine containing dipeptide and tripeptide inhibitors of human renin. Hypertension (Suppl.II) 1986; 8:1–5. 16. Gelb MH, Svaren JP, Abeles RH. Fluoroketone inhibitors of hydrolytic enzymes. Biochemistry 1985; 24:1813–1817.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_339.html (1 of 2) [4/5/2004 5:27:13 PM]

Document

17. Szelke M. Chemistry of renin inhibitors. In: Kostka V, ed. Aspartic Proteinases and Their Inhibitors. Berlin: Walter de Gruyter, 1985:421–441. 18. Sham HL, Stein HH, Rempel CA, Cohen J and Plattner JJ. Highly potent and specific inhibitors of human renin. FEBS Lett 1987; 220:299–301. 19. Cooper JB, Foundling SI, Blundell TL, Boger J, Jupp R, Kay J. X-ray studies of aspartic proteinasestatine inhibitor complexes. Biochemistry 1989; 28:8596–8603. 20. Arrowsmith RJ, Carter K, Dann JG, Davies DE, Harris CJ, Morton JA, Lister P, Robinson JA, Williams DJ. Novel renin inhibitors: synthesis of aminostatine and comparison with statine-containing analogues. J Chem Soc Chem Commun 1986; 10:755–757. 21. Jones M, Sueiras-Diaz J, Szelke M, Leckie B, Beattie S. Renin inhibitors containing the novel amino-acid 3aminodeoxystatine. In: Deber CM, Hruby VJ, Kopple

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_339.html (2 of 2) [4/5/2004 5:27:13 PM]

Document

Page 340

KD eds. Peptides: Structure and Function. Rockford:Pierce Chemical Company, 1985:759–762. 22. Rich DH, Sun ETO, Ulm E. Synthesis of analogues of the carboxyl proteinase inhibitor pepstatin. Effect of structure on the inhibition of pepsin and renin. J Med Chem 1980; 23:27–33. 23. Sali A, Veerapandian B, Cooper JB, Founding SI, Hoover DJ, Blundell TL. High resolution X-ray diffraction study of the complex between endothiapepsin and an oligopeptide inhibitor: the analysis of inhibitor binding and description of the rigid body shifts in the enzyme. EMBO J 1989; 8:2179–2188. 24. Iizuka K, Kamijo T, Kubota T, Akahane K, Umeyama H, Kiso Y. New human renin inhibitors containing an unnatural amino acid, norstatine. J Med Chem 1988; 31:701–704. 25. Dann JG, Stammers DK, Harris, CJ, Arrowsmith RJ, Davies DE, Hardy GW, Morton JA. Human renin: an new class of inhibitors. Biochem Biophys Res Commun 1986; 134:71–77. 26. Cooper JB, Foundling SI, Blundell TL, Arrowsmith RJ, Harris CJ, Champness JN. A rational approach to the design of antihypertensives: X-ray studies of complexes between aspartic proteinases and aminoalcohol inhibitors. In: Leeming PR, ed. Topics in Medicinal Chemistry. London: Royal Society of Chemistry, 1988; 308–313. 27. Lunney EA, Hamilton HW, Hodges JC, Kaltenbrohn JS, Repine JT, Badasso M, Cooper J, Dealwis C, Wallace B, Lowther WT, Dunn BM, Humblet C. Analyses of ligand binding in five endothiapepsin crystal complexes and their use in the design and evaluation of novel renin inhibitors. J Med Chem 1993; 36:3809–3820. 28. Bartlett PA, Kezer WB. Phosphinic acid dipeptide analogues: potent, slowbinding inhibitors of aspartic proteinases. J Amer Chem Soc 1984; 106:4282–4283. 29. Greenlee WJ. Renin inhibitors. Pharm Res 1987; 4(5):364–374. 30. Thaisrivongs S, Pals DT, Harris DW, Kati WM, Turner SR. Design and synthesis of potent and specific renin inhibitors containing difluorostatine, difluorostatone and related analogues. J Med Chem 1986; 29:2088–2093. 31. Veerapandian B, Cooper JB, Sali A, Blundell TL. Direct observation by X-ray analysis of the tetrahedral “intermediate” of aspartic proteinases. Protein Science 1992; 1:322–328. 32. Suguna K, Padlan EA, Smith CW, Carlson WD, Davies DR. Binding of a reduced peptide inhibitor to the aspartic proteinase from Rhizopus chinensis: implications for a mechanism of action. Proc Natl Acad Sci USA 1987; 84:7009–7013. 33. Ptitsyn OB. Pure Appl Chem 1973; 31:227–244.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_340.html (1 of 2) [4/5/2004 5:27:15 PM]

Document

34. Rosenberg SH, Plattner JJ, Woods KW, Stein HH, Marcotte PA, Cohen J, Perun TJ. Novel renin inhibitors containing analogues of statine retro-inverted at the C-termini: specificity of the P2 histidine site. J Med Chem 1987; 30:1224–1228. 35. Luly JR, Yi N, Soderquist J, Stein H, Cohen J, Perun TJ, Plattner JJ. New inhibitors of human renin that contain novel Leu-Val replacements. J Med Chem 1987; 30:1609–1616. 36. Cooper J, Quail W, Frazao C, Foundling SI, Blundell TL. X-ray crystallographic analysis of inhibition of endothiapepsin by cyclohexyl renin inhibitors. Biochemistry 1992; 31:8142–8150.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_340.html (2 of 2) [4/5/2004 5:27:15 PM]

Document

Page 341

37. Attwood MR, Hassall CH, Krohn A, Lawton G, Redshaw S. The design and synthesis of angiotensin converting enzyme inhibitor cilazapril and related bicyclic compounds. J Chem Soc Perkin 1986; I:1011–1019. 38. Szewczuk Z, Rebholz KL, Rich DH. Synthesis and biological activity of new conformationally restricted analogues of pepstatin. Int J Pept Res 1992; 40:233–242. 39. Wood JM, Fuhrer W, Buhlmayer P, Riniker B, Hofbauer KG. Protection groups increase the in vivo stability of a statine-containing renin inhibitor. In: Kostka V, ed. Aspartic Proteinases and Their Inhibitors. Berlin: Walter de Gruyter, 1985:463–466. 40. Bolis G, Fung AKL, Greer J, Kleinert HD, Marcotte PA, Perun TJ, Plattner JJ, Stein HH. Renin inhibitors. Dipeptide analogues of angiotensinogen incorporating transition-state, nonpeptidic replacements at the scissile bond. J Med Chem 1987; 30:1729–1737. 41. Greenlee, WJ Renin inhibitors. Med Res Rev 1990; 10:173. 42. Dhanaraj V, Dealwis C, Frazao C, Badasso M, Sibanda BL, Tickle IJ, Cooper JB, Driessen HPC, Newman M, Aguilar C, Wood SP, Blundell TL, Hobart PM, Geoghegan KF, Ammirati MJ, Danley DE, O'Connor BA, Hoover DJ. X-ray analyses of peptide-inhibitor complexes define the structural basis of specificity for human and mouse renins. Nature 1992; 357:466–472. 43. Dealwis CG, Frazao C, Badasso M, Cooper JB, Tickle IJ, Driessen H, Blundell TL, Murakami K, Miyazaki H, Sueiras-Diaz J, Jones DM, Szelke M. X-ray analysis at 2.0 Å resolution of mouse submaxillary renin complexed with a decapeptide inhibitor CH-66, based on the 4–16 fragment of rat angiotensinogen. J Mol Biol 1994; 236:342–360. 44. Wlodawer A, Miller M, Jaskolski M, Sathyanarayana BK, Baldwin E, Weber IT, Selk LM, Clawson L, Schneider J, Kent S. Conserved folding in retroviral proteinases: crystal structure of synthetic HIV-1 proteinase. Science 1989; 245:616–621. 45. Lapatto R, Blundell TL, Hemmings A, Overington J, Wilderspin A, Wood SP, Merson JR, Whittle PJ, Danley DE, Geoghegan KF, Hawrylik SJ, Lee SE, Scheld KG, Hobart PM. X-ray analysis of HIV-1 proteinase at 2.7 Å resolution confirms structural homology among retroviral enzymes. Nature (Lond) 1989; 342:299–302. 46. Poe M, Wu JK, Lin TY, Hoogsteen K, Bull HG, Slater EE. Renin cleavage of a human-kidney renin substrate analogous to human angiotensinogen that is human renin specific and resistant to cathepsin D. Analyt Biochem 1984; 140:459–467. 47. Cumin F, Lenguyen D, Castro B, Menard J, Corvol P. Comparative enzymatic studies of human renin acting on pure natural or synthetic substrates. Biochim Biophys Acta 1987; 913:10–19.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_341.html (1 of 2) [4/5/2004 5:27:17 PM]

Document

48. Powers JC, Harley AD, Myers DV. Subsite specificity of porcine pepsin. In: Tang J, ed. Acid Proteases-Structure, Function and Biology. New York: Plenum Press, 1977:141–157. 49. Antonov VK. In: Tang J, ed. Acid Proteases-Structure, Function and Biology. New York: Plenum Press, 1977:179. 50. Skeggs LT, Lentz KE, Kahn JR, Hochstrasser H. Kinetics of the reaction of renin with nine synthetic peptide substrates. J. Exp Med 1968; 120:130–34. 51. Holzman TF, Chung CC, Edalji R, Egan DA, Martin M, Gubbins EJ. Krafft GA, Wang GT, Thomas AM, Rosenberg SH, Hutchins C. Characterisation of recombi-

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_341.html (2 of 2) [4/5/2004 5:27:17 PM]

Document

Page 342

nant human renin kinetics, pH-stability, and peptidomimetic inhibitor binding. J Protein Chem 1991; 10:553–563. 52. Tong L, Pav S, Lamarre D, Pilote L, Laplante S, Anderson PC, Jung G. High resolution crystalstructures of recombinant human renin in complex with polyhydroxymonoamide inhibitors. J Mol Biol 1995; 250:211–222. 53. Sham HL, Bolis G, Stein HH, Fesik SW, Marcott PA, Plattner JJ, Rempel CA, Greer J. Renin inhibitors. Design and synthesis of a new class of conformationally restricted analogues of angiotensinogen. J Med Chem 1988; 31:284–295.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_342.html [4/5/2004 5:27:19 PM]

Document

Page 343

14 Structural Aspects in the Inhibitor Design of Catechol O-Methyltransferase Jukka Vidgren and Martti Ovaska Orion Corporation Orion Pharma, Espoo, Finland I. Introduction Catechol O-methyltransferase (COMT) plays an important role in the catabolic inactivation of catecholamines. It is present both in extracerebral tissues and in the central nervous system. During the last few years there has been a remarkable interest in COMT. Basic biochemical and molecular biology research has given detailed insights into the function and nature of the enzyme. The knowledge of the crystallographic structure has allowed researchers to analyze the molecular mechanism of the catalytic reaction and to accomplish the structure-based design of inhibitors. The development of potent and selective inhibitors has provided effective pharmacological tools to investigate the physiological role of the enzyme. The main clinical interest has been the possible application of COMT inhibitors as adjuncts in the L-dopa therapy of Parkinson's disease. Parkinson's disease is a dopamine deficiency disorder. The dopamine-producing neurons in striatum are destroyed. The medication strategy is to replenish the missing dopamine. L-Dopa, given together with a peripheral inhibitor of dopa decarboxylase (DDC), for example, carbidopa, is a standard therapy in Parkinson's disease. While dopamine does not penetrate into the brain, L-dopa penetrates the blood-brain barrier and is decarboxylated into dopamine in the brain. The half-life of L-dopa is short and in the presence of DDC inhibitor a large amount of the drug is eliminated by COMT. The COMT enzyme produces the metabolite 3-methoxytyrosine (3-OMD), which has no benefit in the treatment of Parkinson's disease, but has a long elimination half-life and may be harmful during chronical treatment. Also a gradual loss of the efficacy of L-dopa occurs during longterm medication. Since the early 1980s active

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_343.html [4/5/2004 5:27:21 PM]

Document

Page 344

Figure 1 The rationale of COMT inhibition as adjunct in the L-dopa therapy of Parkinson's disease (reproduced by permission from COMT News, Issue 1, Orion Corporation, Orion Pharma, 1994).

research in pharmaceutical companies has been carried out to develop new potent, selective, and orally active COMT inhibitors. Some of them (e.g., entacapone), are now in final clinical trials and the results have been promising. The rationale of COMT inhibition can be seen in Figure 1. It can be concluded that COMT inhibition in peripheral tissues improves the brain entry of L-dopa and decreases the formation of 3-OMD. The dose of L-dopa can be lowered and the dose interval prolonged. Also a decrease of the fluctuations of dopamine formation has been observed. The inhibition of COMT seems to be the next step in improving the L-dopa therapy of Parkinson's disease. This paper discusses the structure-based approach for the understanding of the enzyme function and inhibitor design. II. The Enzyme A. Physiological Role of COMT Catechol O-methyltransferase (COMT, EC 2.1.1.6) was originally detected in rat liver extracts [1]. Since then, COMT has been found in many species: http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_344.html (1 of 2) [4/5/2004 5:27:34 PM]

Document

Page 345

Figure 2 The reaction catalyzed by catechol O-methyltransferase. Dopamine: R=CH2-CH2-NH2; L-dopa: R=CH2-CH(NH2)-COOH.

animals, plants, and procaryotes [2]. In mammals the highest COMT activities have been found in the liver and kidney, but COMT is common in almost all mammalian tissues [2–4]. The COMT enzyme catalyzes the transfer of the methyl group from the coenzyme S-adenosyl-Lmethionine (AdoMet) to one of the phenolic hydroxyl groups of a catechol or substituted catechol [1] (Figure 2). The presence of magnesium ions is required for the catalysis. The reaction products are Omethylated catechol and S-adenosyl-L-homocysteine (AdoHcy). Physiological substrates of COMT are catecholamine neurotransmitters, dopamine, noradrenaline, and adrenaline, and some of their metabolites. The COMT enzyme inactivates catecholic steroids such as 2-hydroxyestradiol, drugs with a catechol structure such as L-dopa, and a large number of other catechol compounds [1,2,5–7]. The general physiological function of COMT is the inactivation of biologically active or toxic catechols. A schematic view of the major catecholamine pathways in the brain is shown in Figure 3. L-Dopa is the dopamine precursor used in the treatment of Parkinson's disease [8]. B. Primary Structures There are no isoenzymes of COMT known in different mammalian tissues. Two distinct forms of COMT have been found: one is soluble (S-COMT) and the other membrane bound (MB-COMT) [9,10]. Both soluble and membrane-bound COMT have been cloned and characterized [11–16]. The soluble and membrane-bound COMT are coded by one gene using two separate promoters [17]. The soluble COMT contains 221 amino acids, whereas the membrane-bound form has a 50-(human) or 43-(rat) residueslong amino-terminal extension containing the hydrophobic membrane anchor region. The sequences of COMT enzymes from different species are highly similar (see Figure 4). The soluble human protein is 81% identical with the rat enzyme. The 165-amino-acids-long fragment of porcine COMT has 82% homology with the human

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_345.html [4/5/2004 5:27:37 PM]

Document

Page 346

Figure 3 The main metabolic routes of dopamine and noradrenaline in the brain. COMT, Catechol O-methyltransferase; MAO, monoamino oxidase; DDC, dopa decarboxylase; DBH, dopamine β-hydroxylase; 3-OMD, 3-methoxytyrosine; Dopac, dihydroxyphenyl acetic acid.

enzyme [13]. The existence of a thermolabile low-activity and a thermostable high-activity COMT in human population has been reported [18]. Interestingly, the two published sequences of human soluble COMT differ in only one amino acid. Recent kinetic studies have shown that this difference affects unambiguously the thermostability of the enzyme [19]. C. Kinetics of Human COMT The kinetic mechanism of the methylation reaction of human COMT has been studied exhaustively using recombinant enzymes [19]. The mechanism is sequential ordered: AdoMet binding first, then Mg2+ and the catechol substrate as the last ligand. Human S-COMT and MB-COMT have similar kinetic properties. The main difference is the one-order lower Km value of MB-COMT for dopamine as substrate (S-COMT 207 µM and MB-COMT 15 µM). The COMT enzyme is a rather slow enzyme with a low catalytic number. At saturating substrate levels S-COMT has a double efficiency compared with MB-COMT (kcat=37 and kcat =17, respectively). At low substrate concentrations (7.8a

131

1027

0.04

1021

>9.3

142

300

0.13

1156

Run 2 2-Nitrobenzoic acid hydrazide Dopamine

L-histidinol aCould

not be tested at higher concentrations due to solubility problems.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_382.html (2 of 2) [4/5/2004 5:40:52 PM]

Document

Page 383

Figure 10 Binding mode of 2-guanidinobenzimidazole to T.brucei GAPDH as predicted by DOCK. The benzimidazole moiety occupies roughly the position occupied by adenine in the holo-enzyme, whereas the guanidino group a salt bridge to Asp37.

the basis of structural rigidity, chemical stability, solubility, and electrostatic complementarity. The latter property was evaluated with the program DELPHI [85]. Each of the sixteen compounds was tested for GAPDH inhibition (Table 7). Half of them were inactive while the other ones showed inhibition in a range between 1.2 and 25 mM. Unfortunately, there appears to be no correlation between the DOCK scores and the IC50 values. For examples, norepinephrine and 1,3-diphenylguanidine are inactive while they have a better score than 2- guanidinobenzimidazole (Figure 10), the compound with the best IC50. Also, it appeared that the two different receptor descriptions used led to almost completely different lists of compounds. Only 156 molecules occurred in both lists of top-scoring molecules. In the modeled binding mode all of the inhibitors occupy roughly the same position as the purine ring of NAD in the crystal structure of GAPDH. While the values obtained for IC50 are indicative of poor inhibition, one has to keep in mind that adenosine exhibits an IC50 of 50 mM [13]. By using the program DOCK we were able to discover ligands that have a substantially higher affinity for GAPDH than the natural ligand. C. Phosphoglycerate Kinase: Leads from the Past The starting point for drug design in the case of PGK is quite different from TIM or GAPDH because a number of nonsubstrate-like inhibitors have been

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_383.html (1 of 2) [4/5/2004 5:41:28 PM]

Document

Page 384

Figure 11 Two-dimensional structure of SPADNS, a micromolar inhibitor of T.brucei PGK.

discussed in the literature. Suramin inhibits glycosomal PGK of T. brucei with a Ki of 8.0 µM [69]. In addition a number of yeast PGK inhibitors are known: gallic acid with a Ki = 0.4 mM [86], hydroxyethylidene biphosphate with a Ki = 24 mM [87]. 1,3,6-naphthalenetrisulphonic acid with a Ki = 5.5 mM, and 2-(p- sulphophenylazo)-1,8-dihydroxy-3,6-disulphonic acid, also known as SPADNS (Figure 11), with a Ki = 126 µM [87]. None of the four yeast PGK inhibitors are potent, but, for SPADNS, the binding mode has been further characterized. Studies by Williams et al. [87] have demonstrated that SPADNS is directly competitive with both enzyme substrates, 3-phosphoglycerate and ATP. Moreover, by 600 MHz 1H-NMR it was shown that SPADNS interacts with the nucleotide binding site while the conformation of the enzyme changes substantially [87]. Since the four yeast PGK inhibitors are commercially available it was logical to test them for T. brucei PGK inhibition. The first three compounds were active in the millimolar range. However, SPADNS exhibited a Ki of 10.0 µM in these in preliminary tests [88]. Moreover, when assayed against a commercially available rabbit muscle PGK, SPADNS had no influence on the enzyme kinetics up to a concentration of 250 µM [88]. In conclusion, SPADNS appears to be an excellent lead because of its potency and selectivity. Crystallographic experiments to determine its binding mode to T. brucei PGK are underway. IV. Lead Optimization: Glycosomal Gapdh From the selectivity point of view the adenosine binding site of GAPDH is attractive for drug design, as we explained in Section II.B. Unfortunately, inhibition studies on T. brucei and L. mexicana GAPDH revealed the poor affinity of our natural lead adenosine with IC50 values of 100 mM and 50 mM, respectively. Moreover, adenosine is an “antiselective” lead because the IC50

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_384.html [4/5/2004 5:41:34 PM]

Document

Page 385

for human GAPDH is better than for the parasite enzymes, namely 35 mM [13]. Despite the fact that these IC50 values are about ten thousand times higher than what would be considered a lead in the pharmaceutical industry we decided to optimize the affinity and selectivity of adenosine. Each of the three areas where differences occur between the parasite and the human enzyme are hydrophobic. Therefore, we modeled hydrophobic substituents at positions C2, C8, and O2' of adenosine under the constraint that they were conformationally compatible with the C2'-endo pucker of the ribose sugar. Designing derivatives at O2' was a problem, however. Each of the two ribosyl hydroxyls forms a hydrogen bond with the carboxylate of Asp37. Since making direct derivatives of the hydroxyl, such as ethers or esters, would deprive the Asp of a hydrogen-bond partner while burying the carboxylate, resulting molecules would have a dramatically reduced affinity. Moreover, an alignment of 47 GAPDH sequences made it clear that the Asp is highly conserved [89]. An elegant way to overcome this problem was to replace the 2' -hydroxyl by a 2'- amino function. Moreover, coupling with carboxylic acids was appealing from a synthetic point of view while the conformational properties of the amido-substituted system would ensure the correct orientation of substituents into the selectivity cleft. The modeled inhibitors were evaluated for the quality of their fit to the protein surface and subsequently synthesized. From Table 8 it can be seen that our predictions were successful. The addition of a methyl group at C2 of the adenine ring, which is close to Val36, increased the affinity for parasite GAPDH by an order of magnitude. The effect of a thienyl substitution on C8, targeted to Leu112, was even bigger, namely two orders of magnitude. However, both substitutions are only mildly selective (Table 8). As expected, the greatest gain in selectivity was obtained by modifying the 2'-position of the ribose, so that the selectivity cleft is filled up (Figure 12). The 2'-deoxy-2'-(3-methoxybenzamido) adenosine compound (Figure 13) bound at least 48 times better to L. mexicana GAPDH than to the human enzyme. The selectivity versus T. brucei GAPDH appeared to be smaller. This has to be ascribed to a difference in residues contacting the 3-methoxy moeity. The residue Asn39 of T. brucei GAPDH has a Ser equivalent in the L. mexicana Table 8 Inhibition Gains of Designed GAPDH Inhibitors with Respect to Adenosine C2-subst

C8-subst

CH3

H

H

T. brucei

L. mexicana

human

OH

12.5

6.25

10-4. Examples of these compounds are 1,5-bis(4-amidophenoxy) pentane (pentamidine), in the form of pentamidine isothionate, and an imidazoline in the form of 1,5-di(4-imidazolinophenoxy)pentane. Tricycli- cylidene-acetic acid [92] and its 2-chloro derivative [93] were found to be inhibitors of IL-1 release, claiming clinical improvement in patients with psoriasis, periodontal disease, and Alzheimer's disease. In vitro this compound blocks the synthesis of prostaglandins and inhibits the release of IL-1α and IL-1β from human monocytes and murine macrophages. Probucol, a hypocholesterolemic drug that possesses antioxidant activity, inhibits the ex vivo release of IL-1 from LPS-stimulated macrophages of mice pretreated orally with 100 mg/kg/day of this compound [94,95]. This compound has been shown to inhibit LPS-induced zinc-lowering effect, is cited as direct evidence for the inhibition of IL-1 release, and may be useful candidate for the treatment of atherosclerosis [95,96]. An amino-dithiol-one derivative (RP 54745) blocked the proliferative action of IL-1β on murine thymocytes in vitro and also inhibited the production of IL-1 in mouse peritoneal macrophages in vitro and in vivo. The compound RP 54745 selectively inhibited the expression of IL1α and IL-1β mRNA while TNFα mRNA was unaffected [97, 98]. Administration of a cocktail containing eicosapentenoic acid and docosahexenoic acid to volunteers for up to 6 weeks, resulted in a significant depression in IL-1β (61%), IL-1α (39%), and TNF (40%) synthesis. These levels returned to normal after a few weeks [99]. In vitro studies indicate that Pentoxifylline can block the effects of IL-1 and TNF on neutrophils [100]. It is a phosphodiesterase (PDE) inhibitor that causes increased capillary blood flow by decreasing blood viscocity and is used clinically in chronic occlusive arterial disease of the limbs with intermittent claudication. Denbufylline, a closely related xanthine, has been patented as a functional inhibitor of cytokines and exhibits a similar profile to Pentoxifylline [101]. Romazarit (Ro-31-3948) derived from oxazole and isoxazole propionic acids has been shown to block IL- 1-induced activation of human fibroblasts in vitro and in animal models reduces inflammation [102,103,104]. By using a spontaneous autoimmune MRL/lpr mouse model, a significant efficacy was shown [105]. Two-dimensional structures of some of these molecules are shown in Figure 14. Even though the above mentioned small molecules exhibit IL-1 inhibition none of them were discovered based on defined functional or structural aspects. An understanding of the three-dimensional structure of IL-1s and their receptors, by themselves or in complexes, will form a very strong foundation for structure-based design of more specific and potent IL-1-based immunomodulators.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_427.html [4/5/2004 5:50:49 PM]

Document

Page 428

VI. Conclusion The design of novel compounds to inhibit or manipulate the IL-1 system remains a daunting task. At this time, the design of immunomodulators for the IL-1 system is still in its infancy and has largely been confined to the use of whole or fragmented proteins or the identification of nonspecific small molecules. In addition, newer approaches have also been initiated and these include the use of antisense oligonucleotides, small molecules designed to compete with IL-1's binding to its receptor, ICE inhibitors, and intracellular signaling inhibitors. All such strategies show promise. Structure-based design has not been explicitly used in the design of agonists and antagonists of IL-1. But as of now we have the structures of IL-1α, IL-1β, and IL-1Ra. A new insight may be forthcoming once the complex crystallographic structure of one of the interleukin-1 molecules and its corresponding receptor molecule is available. This structural information, coupled with the anticipated IL-1 + IL-1R complex structure, will form the foundation for rational design of inhibitors with improved selectivity for the treatment of various IL-1-mediated diseases. Acknowledgements Our special thanks to Professor Russell Doolittle for his encouragement and support, and also to Dr. Mitch Lewis for providing us with the very high-resolution coordinates of IL-1α. We gratefully acknowledge San Diego Super Computing Center for their assistance and support in providing valuable software and high-power computing time. We thank Dr. Donald Kyle for his valuable comments. We also thank Dr. Per Kraulis for providing us the latest version of MOLSCRIPT, Dr. Anthony Nicholls for the program GRASP, Dr. Rob Russell for the structural alignment, and Professor Lynn Ten Eyck and Dr. Jerry Greenberg for their help. References 1. Dinarello CA. Biological basis for interleukin-1 in disease. Blood 1996; 87- 6:2095–2147. 2. Dinarello CA. Interleukin-1 is produced in response to infection and injury. Rev infect Dis 1984;6:51–56. 3. Oppenheim JJ, Kovacs EJ, Matsushima K, Durum SK. There is more than one interleukin 1. Immunol Today 1986;7:45–56. 4. Durum KS, Oppenheim JJ, Neta R. Immunophysiologic role of interleukin-1. In: Oppenheim JJ, Shevach EM, eds. Immunophysiology. New York: Oxford University Press, 1990:210–225.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_428.html [4/9/2004 12:10:04 AM]

Document

Page 429

5. Dinarello CA. On the Biology of Interleukin-1. Adv Immunology 1989; 44:153– 205. 6. Meager A. Cytokines. 1990; Open University Press. 7. Henderson B, Blake S. Therapeutic Advantage of Cytokine Manipulation. TIPS 1992; 13:145–152. 8. Dower SK, Fanslow W, Jacobs C, Waugh S, Sims JE, Widmer MB. Interleukin-1 antagonists. Therapeutic Immunol 1994; 1:113. 9. Burger DD, J. Inhibitory Cytokines and Cytokine Inhibitors. Neurology 1995; 45(supp 16):s39–43. 10. Abraham DJ. X-ray crystallography and drug design. In: Perun TJ, Propst CL, eds. Computer Aided Drug Design. New York: Marcel Dekker, Inc., 1989:93–132. 11. Fesik SW. Approaches to drug design using nuclear magnetic resonance spectroscopy. In: Perun TJ, Propst CL, eds. Computer-Aided Drug Design. New York: Marcel Dekker, Inc., 1989:133–184. 12. Veerapandian B. Structure aided drug design. In: Wolf M, ed. Buerger's Drug Discovery and Medicinal Chemistry. New York: John Wiley and Sons, 1995:303– 348. 13. Whittle PJ, Blundell TL. Protein Structure based drug design. Ann Rev Biophy and Biomol Structure 1994; 23:349–375. 14. Carter DB, Deibel MRJ, Dunn CJ, Tomich C-SC, Laborde AL, Slightom JL. Berger AE, Bienkowski MJ, Sun FF, McEwan RN, Hams PKW, Yem AW, Waszak GA, Chosay JG, Sieu LC, Hardee MM, Zucher-Neely HA, Reardon IM, Heinnckson RL, Truesdell SE, Shelly JA, Eessalu TE, Taylor BM, Tracey DE. Purification, cloning, expression, and biological characterization of an interleukin-1 receptor antagonist protein. Nature 1990; 344:633–638. 15. Eisenberg SP, Evans RJ, Arend WP, Verderber E, Brewer MT, Hannum CH, Thompson RC. Primary structure and functional expression from complementary DNA of a human interleukin-1 receptor antagonist. Nature, 1990; 343:341–346. 16. Sims JE, March CJ, Cosman D, Widmer MB, MacDonald HR, McMahan CJ, Grubin CE, Wignall JM, Jackson JL, Call SM, et al. cDNA expression cloning of the IL-1 receptor, a member of the immunoglobulin superfamily. Science 1988; 241:585. 17. McMahon CJ, Slack JL, Mosley B, Cosman D, Lupton SD, Brunton LL, Grubin CE, Wignall JM, Jenkins NA, Brannan C1, Copeland NG, Huebner K, Croce CM, Cannizzaro LA, Benjamin D, Dower S, Spriggs MK, Sims JE. A novel IL-1 receptor cloned from B cells by mammalian expression is expressed in many cell types. EMBO J 1991; 10:2821.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_429.html (1 of 2) [4/9/2004 12:10:07 AM]

Document

18. Greenfeder SA, Nunes P, Kwee L, Labow M, Chizzonite RA, Ju G, Molecular cloning and characterization of a second subunit of the Interleukin-1 receptor complex. J Biol Chem 1995;270:13757–13765. 19. Thornberry NA, Bull HG, Calaycay JR, Chapman KT, Howard AD, Kostura MJ, Miller DK, Molineaux SM, Weidner JR, Aunins J, Schmidt JA, Tocci M.A novel heterodimeric cysteine protease is required for interleukin-1β processing in monocytes. Nature 1992; 356:768. 20. Cerretti DP, Kozlosky CJ, Mosley B, Nelson N, Van Ness K, Greenstreet TA, March CJ, Kronheim SR, Druck T, Cannizzaro LA, Huebner K, Black RA. Molecular cloning of the IL-1β processing enzyme. Science 1992; 256:97.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_429.html (2 of 2) [4/9/2004 12:10:07 AM]

Document

Page 430

21. Seckinger P, Lowenthal JW, Williamson K, Dayeri M, MacDonald HR. A urine inhibitor of interleukin-1 activity that blocks ligand binding. J Immunol 1987; 139:1546. 22. Hannum CH, Wilcox CJ, Arend WP, Joslin FG, Dripps DJ, Heimdal PL, Armes LG, Sommer A, Eisenberg SP, Thompson RC. Interleukin-1 receptor antagonist activity of a human interleukin-1 inhibitor. Nature 1990; 343:336. 23. Mazzei GJ, Seckinger PL, Dayer JM, Shaw AR. Purification and characterization of a 26-kDa competitive inhibitor of interleukin 1. Eur J Immunol, 1990; 20:683. 24. Haskill S, Manin M, VanLe L, Morris J, Peace A, Bigler CF, Jaffe GJ, Sporn SA, Fong S, Arend WP, Ralph P. cDNA cloning of a novel form of the interleukin-1 receptor antagonist associated with epithelium. Proc Natl Acad Sci USA 1991; 88:3681. 25. Arend WP. Interleukin-1 receptor antagonist. Adv Immunol 1993; 54:167. 26. Muzio M, Polentarutti N, Sironi M, Pli G, DeGioia L, Introna M, Mantovani A, Colotta F. Characterization of intracellular interleukin-1 receptor antagonist II. Cytokine 1995; 7:632. 27. Kroggel R, Martin M, Pingoud V, Dayer J-M, Resch K. Two Chain structure of the interleukin-1 receptor. FEBS Lett 1988; 229:59. 28. Sims JE, Painter SL, Gow IR. Genomic organization of the type I and type II IL-1 receptors. Cytokine 1995, 7:483–490. 29. Sims JE, GMA, Slack JL, Alderson MR, Bird TA, Gin JG, Colotta F, Re F, Mantovani A, Shanebeck K, Grabstein KH Dower SK. Interleukin-1 signaling occurs exclusively via the type I receptor. Proc Natl Acad Sci USA 1993; 90:6155–6159. 30. Colotta F, DSK, Sims JE, Mantovani A. The type II “decoy” receptor: A novel regulatory pathway for interleukin-1. Immunol Today 1994; 15:562. 31. Symons JA, Young PA, Duff GW. The soluble interleukin I receptor: Ligand binding properties and mechanisms of release. Lymphokine Cytokine Res 1993; 12:381. 32. Arend WP, Malyak M, Smith MF, Whisenand TD, Slack JL, Sims JE, Giri JG, Dower SK. Binding of IL-1α, IL-1β, and IL-1 receptor antagonist by soluble IL-1 receptors and levels of soluble IL-1 receptors in synovial fluids. J Immunol 1994; 153:4766. 33. Symons JA, Young PA, Duff GW. Differential release and ligand binding of type II IL-1 receptors. Cytokine 1994; 6:555. 34. Hopp TP. Evidence from sequence information that the interleukin-I receptor is a transmembrane GTPase. Protein Sci 1995; 4:1851. http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_430.html (1 of 2) [4/9/2004 12:10:13 AM]

Document

35. Bendtzen K, Svenson M, Jonsson V, Hippe E. Autoantibodies to cytokines-Friends or foes? Immunol Today 1990; 11:167. 36. Svenson M, Hensen MB, Bendtzen K. Distribution and characterization of autoantibodies to interleukin-1α in normal human sera. Scand J Immunol 1990; 32:695. 37. Svenson M, Hensen MB, Kayser L, Rasmussen AK, Reimert CM, Bendtzen K. Effects of human anti-IL-1α autoantibodies on receptor binding and biological activities of IL-1. Cytokine 1992; 4:125. 38. Satoh H, Chizzonite R, Ostrowski C, Ni-Wu G, Kim H, Fayer B, Mae N, Nadeau R, Liberato DJ. Characterization of antiIL-1α autoantibodies in the sera from healthy humans. Immunopharmacology 1994; 27:107.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_430.html (2 of 2) [4/9/2004 12:10:13 AM]

Document

Page 431

39. Saurat JH, Schfferli J, Steiger G, Dayer J-M, Didierjean L. Anti-interleukin-1α auto antibodies in humans. J Allergy Clin Immunol 1991; 88:244. 40. Preistle JP, Schar H-P, Grutter MG. Crystallographic Refinement of Interleukin 1β at 2.0 Å Resolution. Proc Natl Acad Sci USA 1989; 86:9667–9671. 41. Finzel BC, Clancy LL, Holland DR, Muchmore SW, Watenpaugh K, Einspahr HM. Crystal structure of recombinant human interleukin-1β at 2.0 Å resolution. J Mol Biol, 1989; 209:779–791. 42. Veerapandian B, Gilliland GL, Raag R, Svensson AL, Masui Y, Hirai Y, Poulos TL. Functional implications of interleukin-1β bases on the three-dimensional structure. Proteins: Struct, Funct Genet 1992; 12:10–23. 43. Clore GM, Wingfield PT, Gronenborn AM. High resolution three dimensional structure of interleukin-1β in solution by three and 4 dimensional nuclear magnetic spectroscopy. Biochemistry 1991; 30:2315–2319. 44. Ohlendorf DHT, A, Weber PC, Wondolski JJ, Salemme FR, Lischwe M, Newton RC. A comparison of the high resolution structures of human and marine interleukin-1β to be published. 45. Graves BJ, Hatada MH, Hendnckson WA, Miller JK, Madison VS, Satow Y. Structure of interleukin1α at 2.7 A resolution. Biochemistry 1990; 29:2679– 2684. 46. Stockman BJS, TA, Strakalaitis NA, Brunner DP, Yem AW, Deibel MR Jr. Solution structure of human interleukin-1 receptor antagonist protein. FEBS Letters 1994; 349:79–83. 47. Vigers GPA, Caffes P, Evans RJ, Thompson RC, Eisenberg SP, Brandhuber BJ. X-ray structure of interleukin-1 receptor antagonist at 2.0 Å resolution. J of Biol Chem 1994; 269:12874–12879. 48. Schreuder HAR, J-M, Tardif C, Soffientini A, Sarubbi E, Akeson A, Bowlin TL, Yanofsky S, Barrett RW. Crystal structure of the interleukin-1 receptor antagonist. To be published. 49. Spraggon G, Singh O, Stuart DI, Jones EY. The crystal structure of intact interleukin-1 receptor antagonist. To be published. 50. Wilson KPB, JF, Thomson JA, Kim EE, Griffith JP, NAvia MA, Murcko MA, Chambers SP, Aldape RA, Raybuck SA, Livingston DJ. Structure and mechanism of interleukin-1β converting enzyme. Nature 1994; 370:270–275. 51. McLachlan AD. Three-fold structural pattern in the soybean trypsin inhibitor (Kunitz). J Mol Biol 1979; 133:557–563.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_431.html (1 of 2) [4/9/2004 12:10:17 AM]

Document

52. Murzin AG, Lesk AM, Chothia C. b- Trefoil Fold: patterns of structure and sequence in the kunitz inhibitors interleukins-1β and 1α and fibroblast growth factors. J Mol Biol 1992; 223:531–543. 53. Ohelndorf DH. Accuracy of refined proteins structures II. Comparison of four independantly refined models of interleukin-1β. Acta Cryst 1994; D50:808–812. 54. Yuan J, Shaam S, Ledoux S, Ellis HM, Horvitz HR. The C. elegans cell death gene ced-3 encodes a protein similar to mammalian interleukin-1β converting enzyme. Cell 1993; 75:641–652. 55. Labriola-Tompkins E, Chandran C, Kaffka K L, Biondi D, Graves BJ, Hatada M, Madison VS, Karas J, Klian PL, Ju G. Identification of the discontinuous binding site in human interleukin-1β for the type I interleukin-1 receptor. Proc Natl Acad Sci USA 1991; 88:11182–11186. 56. Ju G, Labriola-Tompkins E, Campen CA, Benjamin WR, Karas J, Plocinski J, Biondi D, Kaffka KL, Klian PL, Eisenberg SP, Evans RJ. Conversion of the IL-1

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_431.html (2 of 2) [4/9/2004 12:10:17 AM]

Document

Page 432

receptor antagonist into an agonist by a single amino acid substitution. Proc Natl Acad Sci USA 1991; 88:2658–2662. 57. Kawashima H, Yamagishi J, Yamayoshi M, Ohue M, Fukwi T, Kotani H, Yamada M. Structureactivity relationships in human interleukin-1α: identification of key residues for expression of biological activities. Protein Eng 1992; 5–2:171–176. 58. Labriola-Tompkins E, Chandran C, Varnell TA, Madison VS, Ju G. Structure-function analysis of human IL-1α: Identification of residues required for binding to the human type I IL-1 receptor. Protein Eng 1993; 6–5:535–539. 59. Guinet F, Guitton J, Gault N, Folliard F, Touchet N, Cherel J, Crespo A, Destourbe A, Bertrand P, Denefle P, Mayaux J, Bousseau A, Duchesne M, Terlain B, Cartwright T. Interleukin-1β specific partial agonists defined by site-directed mutagenesis studies. Eur J Biochem 1993; 211:583–590. 60. Grutter MG, van Oostrum J, Pnestle JP, Edelmann E, Joss U, Feige U, Vosbeck K, Schmitz A. Protein Eng, 1994; 7:663–671. 61. Greenfeder SA, Varnell T, Powers G, Lombard-Gillooly K, Shuster D, McIntyre KW, Ryan DE, Leven W, Madison V, Ju G. Insertion of a structural domain of interleukin-1β confers agonist activity to the IL-1 receptor antagonist. J Biol Chem 1995; 270:22460–22466. 62. Wolfson AJ, Kanaoka M, Lau F, Ringe D, Young P, Lee J, and Blumenthal J. Biochemistry 1993; 32:5327–5331. 63. Simoncsits A, Bnstulf J, Tjornhammar ML, Cserzo M, Pongor S, Rybakina E, Gatti S, Bartfai T. Cytokine 1994; 6:206–214. 64. Antoni G, Presentini R, Penn F, Tagliabue A, Ghiara P, Censini S, Volpini G, Villa L, Boraschu D. J Immunol 1986; 137:3201. 65. Boraschi D, Nencioni L, Villa L, Censini S, Bossu P, Ghiara P, Presentini R, Perin F, Frasca D, Dona G, Forni G, Musso T, Giovarelli M, Ghezzi R, Bertini R, Besedovsky H, Del Rey A, Sipe J, Anotoni G, Silvestn S, Tagliabue A. J Exp Med 1988; 168:675–686. 66. Frasca D, Boraschi D, Baschien S, Bossu P, Tagliabue A, Adorini L, Dona GJ Immunol 1988; 141:2651–2655. 67. Beckers W, Villa L, Gonfloni S, Castagnoli L, Newton SMC, Cesareni, Ghiara P. J Immunol 1993; 151:1757–1764. 68. Joss UR, Schmidli I, Vosbeck K. Mapping the receptor binding domain of interleukin-1β by means of binding studies using overlapping fragments: Why did it fail? J Recept Res 1991; 11:275–282.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_432.html (1 of 2) [4/9/2004 12:10:20 AM]

Document

69. Obal F, Opp M, Cady AB, Johannsen L, Postlethwaite AE, Poppleton HM, Seyer JM, Krueger JM. Interleukin-1β and an interleukin-1β fragment are somnogenic. Am J Physiol, 1990; 259:R439. 70. Antoni G, Presentini R, Perin F, Tagliabue A, Ghiara P, Censini S, Volpini G, Villa L, Boraschi D. Peptide analogues of IL-1 and biochemical assay of their binding to its receptors. J Immunol 1986; 137:3201–3204. 71. Slack J, McMahan CJ, Waugh S, Schooley K, Spriggs MK, Sims JE, Dower SK. Independent binding of interleukin-1α and interleukin-1β to type I and type II interleukin-1 receptors. J Biol Chem 1993; 268:2513. 72. Alcami AS, Smith GL. A soluable receptor for interleukin-1β encoded by Vaccinia virus: A novel mechanism of virus modulation of the host response to infection. Cell 1992; 71:153–167. 73. Burger D, Dayer JM. Inhibitory cytokines and cytokine inhibitors. Neurology 1995; 45 (suppl 6):S39–S43.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_432.html (2 of 2) [4/9/2004 12:10:20 AM]

Document

Page 433

74. Bender PE, Lee JC., eds. Pharmacological modulation of interleukin-1 (1989). Annual reports in medicinal Chemistry-25. Johns. Section IV-Metabolic diseases and endocrine function. Chapter 20. 75. Roche Y, Fay M, Gougerot-Pocidalo MA. Antimicrob Chemother 1988; 21:597. 76. Bailly S, Mahe Y, Ferrua B, Fay M, Wakasugi H, Tursz T, Gougerot-Pocidalo MA. Cytokine 1989; 1:303. 77. Otterness IG. Abstracts, 3rd Interscience World Conference on Inflammation. Monte-Carlo, 1989:371. 78. Otterness IG, Bliven ML, Downs JT, Manson DC. Arthritis Rheum. Abstracts, 1988; 31–4:S90, C55. 79. McDonald B, Loose L, Rosenwasser LJ. Arthritis Rheum. Abstracts, 1988; 31– 4:S52, A88. 80. Kadin, U.S. Patent 4,730,004 (1988). 81. Jurgensen CH, Wolberg G, Zimmerman TP. Agents Actions 1989; 27:398. 82. Schmidt JA, Bomford R, Gao XM, Rhodes J. Int J Immunopharmacol 1990; 12:89. 83. Lee JC, Griswold DE, Votta B, Hanna N. Int J Immunopharmacol, 1988; 10:835. 84. Lee JC, Votta B, Griswold DE, Hanna N. Agents Actions 1989; 27:280. 85. Bender PE, Griswold DE, Hanna N, Lee JC. 1988; U.S. Patent 4,794,114. 86. Bender PE, Griswold DE, Hanna N, Lee JC. 1988; U.S. Patent 4,780,470. 87. Bender PE, Griswold DE, Hanna N, Lee JC. 1988; U.S. Patent 4,778,806. 88. Shirota H, Goto M, Hashida R, Yamatsu I, Katayama D. Agents Actions 1989; 27:322. 89. Goodacre J, Carson WD. Allison in Immunopathogenetic Mechanisms of Arthritis. Boston: MTP Press, 1988:211. 90. Rainford KD. J Pharm Pharmacology 1989; 41:112. 91. Shinmei M, Kikuchi T, Masuda K, Shimomura Y. Drugs, 1988; 35 (Suppl. 1):33. 92. Seibel MJ, Bruckle W, Respondek M, Beveridge T. Schnyder J, Muller W, Rheumatol Z. 1989; 48:147. 93. Bollinger P, Gubler HU, Schnyder J. 1989; Derwent 89–138880–B2; DE 38 36 329 Al.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_433.html (1 of 3) [4/9/2004 12:10:25 AM]

Document

94. Ku G, Doherty N. 1988; Derwent 88-314770; AU-A-13160/88. 95. Ku G, Doherty NS, Wobs JA, Jackson RL. A, J Cardiol 1988; 62:778. 96. Marx JL. Science 1988; 239:257. 97. Folliard F, Terlain B. Abstracts, 3rd Inter-science World Conference on Inflammation. MonteCarlo; 1989; 415. 98. Folliard F, Bousseau A, Terlain B. Cytokine 1989; 1:108. 99. Endres S, Ghorbani R, Keliey VE, Georgilis K, Lonnemann G, van der Meer JWM, Cannon JG, Rogers TS, Klempner MS, Weber PC, Schaefer EJ, Woldf SM, Dinarello CA. N Engl J Med 1989; 320:265. 100. Sullivan GW, Carper HT, Novick Jr. WJ, Mandell GL. Infect Immun 1988; 56:1722. 101. Mandell GL, Sullivan GW, Novick Jr. WJ, 1989; Derwent 89–191 551–B2; WO 89 05145.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_433.html (2 of 3) [4/9/2004 12:10:25 AM]

Document

102. Machin PJ, Osbond JM, Sqix CR, Smithen CE, Tong BP. U.S. Patent 4,774,253 (1988). 103. Bloxham DP, Bradshaw D, Cashin CH, Dodge BB, Lewis EJ, Westmacott D, Barber W.E, Machin PJ, Osbond JM, Self CR, Smithen CE, Tong BP. Brit J Rheumatol 1987; 26 (Suppl. 2):2.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_433.html (3 of 3) [4/9/2004 12:10:25 AM]

Document

Page 434

104. Bradshaw D, Dodge BB, Franz PH, Lee SC, Wilson SE. Abstracts, 3rd Interscience World Conference on Inflammation. Monte-Carlo, 1989, 183. 105. Sedgwick AD. Abstracts, 3rd Interscience World Conference on Inflammation. Monte-Carlo; 1989:183. 106. Kraulis PJ, MOLSCRIPT: a program to produce both detailed and schematic plots of protein Structures. J Appl Cryst 1991; 24:946–950. 107. Merrit EM, M RASTER 3D version 2.0: a program for photorealistic molecular graphics. Acta Crystallogr 1994; D50:869–873. 108. Nicholls A, Sharp KA, Honig B. Protein folding and association: insights from the interfacial and thermodynamic properties of hydrocarbons. Proteins 1991; 11:281–296. 109. Russell RB, Barton CJ, Proteins, 1992; 14:309–323. 110. Barton CJ, Protein Engineering, 1989; 6:37–40.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_434.html [4/9/2004 12:10:26 AM]

Document

Page 435

17 Structure and Functional Studies of Interferon: A Solid Foundation for Rational Drug Design Michael A. Jarpe Cambridge NeuroScience, Inc., Cambridge, Massachusetts Carol H. Pontzer University of Maryland, College Park, Maryland Brian E. Szente* and Howard M. Johnson University of Florida, Gainesville, Florida I. Introduction The interferons (IFNs) were discovered in 1957 by Isaacs and Lindenman when they observed that a substance secreted by virally infected cells could protect other cells from viral infection [1a]. They called this substance interferon and found that it was a protein that caused uninfected cells to produce other proteins that made them resistant. Researchers since then have been finding a growing family of structurally related molecules: the interferons. Through the years, the interferons have been given many different names including immune, fibroblast, leukocyte, Type I, and Type II interferons. The recognized nomenclature includes alpha, beta, omega, tau, and gamma (α, β, ω, τ, and γ) interferons. Alpha, beta, omega, and tau all belong to the similar Type I subclass. Gamma is the sole member of the Type II or immune interferon class. The Type I interferons all share a greater sequence homology to each other than they do to IFN-γ (for a recent general review of the IFNs, see Reference 1b). The IFNs exert their actions on cells via cell surface receptors. Type I IFNs share the IFN Type I receptor (IFN-R1) while IFN-γ has its own unique Type II receptor. The signal transduction pathways of Type I and Type II *

Current affiliation: Brigham and Women's Hospital, Boston, Massachusetts.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_435.html [4/9/2004 12:10:29 AM]

Document

Page 436

Figure 1 Interferon activity.

receptor activation are similar. These pathways involve ligand and receptor binding followed by the activation of tyrosine kinases and the phosphorylation of various proteins and their subsequent interaction with transcription elements on DNA. The two receptor activation pathways differ at the level of ligand binding. Type I IFNs bind their receptor as a complex of a single ligand, ligand binding element, and an accessory molecule. The Type II binding event occurs as a dimer of IFN-γ binding to two identical receptor molecules leading to receptor dimerization and activation. The IFNs of all subclasses posses antiviral activity. Additionally, they produce cellular responses that are distinct from antiviral activity, including antiproliferative and immunomodulatory activities (Figure 1). These activities have led to an interest in their use as potential therapeutics to combat viral disease, cancer, and autoimmune disease. Currently, the IFNs have a worldwide market in excess of 2 billion dollars annually. There are six FDA-approved indications in the United States with several more in clinical trials (Table 1). In fact, two of the top-ten grossing biotechnology-based drugs on the U.S. market are IFNs. Intron A is an IFN-α used for immune protection and has an annual U.S. market of $570 million. Roferon-A is another IFN-α used for hairy-cell leukemia and Kaposi's sarcoma with an annual market of $170 million. These sales are despite profound negative side effects associated with IFN treatment. High doses are required to achieve positive clinical results and can lead to severe flu-like symptoms including nausea, vomiting, and fever. These side effects can cause patients to drop out of treatment before beneficial effects are seen. Another drawback of IFNs as drugs is that they require parenteral delivery. The IFNs are protein drugs that must be administered by injection and

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_436.html [4/9/2004 12:10:34 AM]

Document

Page 437 Table 1 Approval Indications for IFNs FDA Approved IFN-α

Clinical Trials

Chronic hepatitis

HIV infection

Kaposi's sarcoma

Colon tumors

Genital warts (papillomavirus)

Kidney tumors Bladder cancer

Hairy cell leukemia

Malignant melanoma Non-Hodgkin's lymphoma Chronic myelogenous leukemia Throat wartz (papillomavirus)

IFN-β

Relapsing remitting multiple sclerosis

Basal cell carcinoma

IFN-γ

Chronic granulomatous disease

Kidney tumors Leishmaniasis

cannot be given orally. For many of the clinical indications, treatments of many months are needed requiring repeat injections. These drawbacks, coupled with the market value of IFN-related treatments, now and in the future, have created an interest in producing second-generation molecules that can mimic IFN activity. These “mimetics” could potentially have greater specificity with fewer side effects. They may also have the advantages of reduced manufacturing costs and more versatile delivery. The design of mimetics can be achieved through structure-based drug design methodologies that are currently being developed. However, in order to apply structure-based drug design to a protein, a solid understanding of the structure/function relationship is needed. A three-dimensional structure, taken alone, gives little insight into the activity of a protein. Structure/function studies must be done for the full potential of structurebased drug design to be realized.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_437.html (1 of 2) [4/9/2004 12:10:37 AM]

Document

Structure/function studies can take a variety of forms and use a number of techniques including the use of molecular biology, synthetic peptides, and antibodies, or combinations of these methods. Molecular biology is a powerful tool for structure/function analysis. Mutagenesis of cDNAs to produce mutant proteins with point mutations, truncations, or deletions can identify functional sites. One drawback to this approach, especially with large proteins, is the proverbial “needle in a haystack” problem. One has difficulty determining where to begin placing mutations. The synthetic peptide approach can be equally as powerful. One can synthesize individual domains or segments of proteins and test them for agonist or antagonist activities thereby identifying functional domains. Synthetic peptides can be used to map the epitope specificity of antibodies that block the activity of a protein. Peptides can also be used to produce monospecific antisera to a defined region of a protein. The antibody approach has also proved quite useful in determining functional sites of

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_437.html (2 of 2) [4/9/2004 12:10:37 AM]

Document

Page 438

proteins. One potential disadvantage to using antibodies is the possibility of over interpreting the blocking results because of steric hindrance. A large antibody molecule may inhibit function through binding to a distant site and covering up a functional site. These approaches have all been used with success on a variety of proteins, but are best used in combination. For example, the information obtained from synthetic peptides and antibodies can significantly narrow down the region for site-directed mutagenesis studies. The entire sequence is narrowed to a segment, which reduces the size of the “haystack” in which the needle is hidden. Even though these approaches are powerful methods for determining functional sites on proteins, they are limited if not coupled with some form of structural determination. As Figure 2 illustrates, molecular biology and synthetic peptide/antibody approaches are not only interdependent, they are tied in with structural determination. Structural determination methods can take many forms, from the classic x-ray crystallography and NMR for three-dimensional determination, to two-dimensional methods such as circular dichroism and Fourier Transformed Infrared Spectroscopy, to predictive methods and modeling. A structural analysis is crucial to the interpretation of experimental results obtained from mutational and synthetic peptide/antibody techniques.

Figure 2 Flow diagram of structure/function studies.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_438.html [4/9/2004 12:10:42 AM]

Document

Page 439 Table 2 Three-Dimensional Structure Studies of the IFNs Type of Study

Reference

X-ray studies IFN-β

2

IFN-γ human

3

IFN-γ bovine

C.T. Samudzi, J.R. Rubin, unpublished data

IFN-γ rabbit

4

IFN-γ human + receptor

5

NMR studies IFN-γ human

6

IFN-γ mouse N-terminal peptide (1–39)

7

Models IFN-α2a

8

IFN-α8

9

IFN-τ sheep

10 T. Senda, S.I. Saitoh, Y. Mitsui, J.Li, and R.M. Roberts, unpublished data

Note: This is not meant to be an exhaustive list of all structural studies of the IFNs. It only highlights some of the three-dimensional studies that have been conducted.

While there are no hard-and-fast rules for conducting structure/function studies, the approaches taken for studying the IFNs can be used to illustrate some of the methods that have been successful. Over the years, a large body of work has accumulated on the IFNs, including a number of structural studies. Table 2 summarizes some of the studies exploring the three-dimensional structure of the IFNs. The following sections review some of the structure/function studies that have begun to elucidate important features of IFN activity and form a basis for future rational drug design.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_439.html (1 of 2) [4/9/2004 12:10:45 AM]

Document

II. Type IFNs A great deal of structure/function analysis has been done on the Type I IFNs. One Type I IFN in particular, IFN-τ has received attention recently because of its lower cytotoxicity compared to the IFNαs. Structure/function studies have concentrated on comparing IFN-τ with IFN-α. Therefore, IFN-τ provides an excellent example of structure/function studies of the Type I IFNs. First isolated from the conceptuses of sheep, IFN-τ is the major conceptus secretory protein responsible for signaling maternal recognition of pregnancy in ruminants [11]. it is produced in large quantities (200 µg in 30 h from a day 16 conceptus culture). The protein was purified using a combination of anion exchange and molecular sieve chromatography.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_439.html (2 of 2) [4/9/2004 12:10:45 AM]

Document

Page 440

A. IFN-τ Synthetic Peptide Studies A sheep blastocyst library was screened with a probe based on the N-terminal sequence of the IFN-τ protein and the cDNA obtained (Table 3). Surprisingly, it exhibited 45–55% homology with various IFNs from human, mouse, rat, and pig and 70% homology with bovine IFN-ω [12]. It shared both molecular weight (19 kDa) and pI (5.4–5.6) with IFN-αs, while its length, 172 amino acids, was equivalent to the IFN-ωs. In competition studies, IFN-τ was found to compete with IFNs α, β, and ω for binding to the Type I IFN receptor [13]. In contrast, IFN-τ exhibited several unique properties such as its reproductive function, its poor inducibility by virus, and its apparent reduced cytotoxicity. Thus, IFNτ conceptus protein appears to be a novel IFN. Structural studies began with production of overlapping synthetic peptides, each 30–35 amino acids in length, corresponding to the entire

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_440.html (1 of 2) [4/9/2004 12:11:03 AM]

Document

Page 441

sequence of the molecule [14]. The peptides were used in competition assays with the native molecule. Peptide inhibition of a particular function would implicate the region of the molecule that it represented in the elicitation of that function. The effect of the IFN-τ peptides on the antiviral activity of ovine IFNτ was examined in a dose/response assay using Madin Darby bovine kidney (MDBK) cells challenged with vesicular stomatitis virus. The carboxy-terminal peptide oIFN-τ(139–172) was found to be the most effective inhibitor of antiviral activity. Three additional peptides, oIFN-τ(1–37), (62–92), and (119–150), also reduced IFN-τ antiviral activity. This suggested that multiple regions of the IFN-τ molecule interact with the Type I IFN receptor and elicit antiviral activity. These regions are underlined in Table 3. The data were consistent with studies of antiviral activity and receptor binding with IFN-α analogs demonstrating that 3 distinct sites, located in the amino-terminal, internal, and carboxy-terminal regions of the molecule, influenced human IFN-α activity [15]. To verify functional results using synthetic peptides, antipeptide antisera were produced [14]. All antipeptide antisera were reactive with the native molecule. Interestingly, antisera titers correlated with the hydropathic index of the peptide, rather than with the predicted surface accessibility of the specific region in the 3-D configuration. Consistent with the peptide studies, antisera against the same four regions of the molecule inhibited IFN-τ activity while antisera to other regions did not. Since IFN-τ and IFN-α bind to the same receptor, the ability of the IFN-τ synthetic peptides to block both bovine and human IFN-α was examined. Interestingly, only three of the four inhibitory peptides were effective competitors of IFN-α. Cross-inhibition of IFN-α by the internal and carboxy-terminal peptides was observed and suggested that these residues may adopt a similar conformation in both molecules and bind to a common site on the receptor. The aminoterminal peptide failed to reduce IFN-α function entirely. Thus, either the IFN-α amino-terminus has a much higher affinity for receptor or the IFN-τ aminoterminus binds a unique site on the receptor complex that may be associated with its unique properties. As expected, none of the peptides blocked the antiviral activity of IFN-τ, which interacts with a different receptor. Next, it was determined whether the same active regions of IFN-τ were involved in additional systems. The Type I IFN receptor on cells has been reported to be somewhat more promiscuous than on other cell types [16]; therefore, vesicular stomatitis challenge of Fc-9 cells was performed. Only the carboxyterminal peptide inhibited IFN-τ activity in this system [17]. This suggested that it was the carboxyterminus that was crucial to receptor interaction. In studies examining IFN-τ-treated feline immunodeficiency virus infected FeT-1 cells and human immunodeficiency virus-infected peripheral

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_441.html [4/9/2004 12:11:10 AM]

Document

Page 442

blood lymphocytes, peptide inhibition of IFN-τ antiretroviral activity implicated both the amino- and carboxy-termini as functionally important [17]. The structural basis of the antiproliferative activity of IFN-τ was also investigated. While multiple regions were again involved in IFN-τ antiproliferative activity, it was the area adjacent to the carboxy terminus, rather than the carboxy-terminus itself, which was the most crucial for antiproliferative activity, inhibiting cell division by blocking entry into the S phase of the cell cycle [18]. Since, for a particular IFN-α subtype, antiviral potency does not necessarily correlate with antiproliferative potency, localization of these functions in different domains of the molecule is not unexpected [19]. Within all known IFN-αs, the 8 amino acids from 139 to 147 are highly conserved. These residues are contained in both carboxy-terminal peptides, but while they may be involved in antiviral activity, they do not appear to be solely responsible for antiproliferative activity since the two peptides are not equivalent inhibitors of IFN-τ antiproliferative activity. This observation is consistent with inhibition of antiviral activity but not antiproliferative activity by a monoclonal antibody in this conserved region in human IFN-α and with the requirement for tyrosine at position 123 for human IFN-α1 antiproliferative activity [20,21]. It has also been reported that mutations around Arg33 affected both antiviral and antiproliferative activity of human IFN-α4 on human cells [22], while the amino-terminus did not appear to be as important in IFN-τ antiproliferative activity on bovine cells. B. IFN-τ Monoclonal Antibodies Another approach to structure/function analysis of IFN-τ involved generation of anti-IFN-τ monoclonal antibodies. Four monoclonal antibodies were produced that reacted with the native IFN-τ protein. They were epitope mapped using the available IFN-τ peptides. Two of the antibodies were directed against the carboxy-terminus of the molecule, one against a region adjacent to the aminoterminus, and the final one appeared to react with a conformational, rather than a linear determinant (C. Pontzer, unpublished data). When these antibodies were used as competitors in binding assays, all four inhibited IFN-τ binding to the Type I IFN receptor on MDBK cells. That anti-IFN-τ carboxy-terminal antibodies would inhibit binding is not unexpected, but the inhibitory activity of the monoclonal antibodies directed against the more amino-terminal region was not anticipated. There is the caveat that warns that results using monoclonal antibodies to delineate function sites must be interpreted with caution since their size may cause significant steric hindrance. To proceed further with structural studies of IFN-τ, access to larger quantities of pure protein was required. The obvious route to this end entailed production of recombinant protein. A synthetic gene for IFN-τ was designed

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_442.html [4/9/2004 12:11:12 AM]

Document

Page 443

that allowed for optimal expression in both bacterial and yeast systems [23]. In addition, restriction sites were incorporated at intervals throughout the length of the sequence to allow for cassette mutagenesis. Using the Pichia pastoris expression system, 50 mg of purified IFN-τ were obtained from a one-liter culture. C. IFN-τ Binding and Signal Transduction Detailed receptor-binding studies were performed comparing recombinant human IFN-α and IFN-τ [24]. The Kd of 125I-IFN-τ and 125I-IFN-αA for receptor on MDBK cells was 3.9 × 10-10 M and 4.45 × 10-11 M, respectively. Consistent with the higher binding affinity, IFN-αA was several fold more effective than IFN-τ as a competitive inhibitor. Functionally, the two IFNs had similar specific antiviral activities, but IFN-τ was 30 fold less toxic to MDBK cells at high concentrations. Phosphorylation of the signal transduction proteins, Tyk2, Stat1a, and Stat2 did not appear to be involved in the cellular toxicity associated with IFN-α relative to IFN-τ. Excess IFN-τ did not block the cytotoxicity of IFN-αA, suggesting that they recognize the receptor differently. While maximal IFN antiviral activity required only fractional receptor occupancy, toxicity was associated with maximal occupancy. Thus, “spare” receptors may exist with respect to certain biological properties, and IFNs may induce a concentrationdependent selective association of receptor subunits. D. Structural Biology of IFN-τ In order to better interpret the information derived from the above studies, an understanding of the 3-D structure of IFN-τ is required. Prior to resolution of the crystal structure, modeling techniques were employed for structural predictions [10]. For IFN-τ, the homology it shares with the other IFNs can be exploited. Since the x-ray coordinates for IFN-β are known [2] (Figure 3), it was used as a template for predicting the topology of IFN-τ. When the sequences of IFN-τ and IFN-β are aligned, the overall homology is approximately 30%. When residues are compared on the basis of conservative substitutions, the similarity rises to about 50%, and if only the location of hydrophobic residues is compared, the similarity is approximately 75%. This is important because hydrophobicity is thought to be a critical factor in driving protein folding. The interferons IFN-β, IFN-α-2, and several other cytokines including IL-2, IL-4, growth hormone, and GM-CSF belong to a family in which all share a four-helix bundle structural motif. Four-helix bundles exhibit a characteristic apolar periodicity in the α helices where every third or fourth residue is apolar, forming a hydrophobic strip down one side of the helix, which facilitates packing. The aligned helical regions of IFN-τ show the same apolar periodicity, suggesting a four-helix bundle motif.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_443.html [4/9/2004 12:11:16 AM]

Document

Page 444

Figure 3 Stereo view of IFN-β crystal structure [2].

The structure of IFN-τ was also examined by CD [10]. Analysis of the IFN-τ spectra predicts that the secondary structural elements derived from CD spectra indicate approximately 70% α-helix. The remainder of the molecule is either predicted to be random or a combination of β sheet and turn. Since it is known that algorithms that predict secondary structures from CD spectra are most accurate at identifying α helices, we are confident that IFN-τ is mainly α helical. The CD spectra for the synthetic peptides of IFN-τ were also obtained. The peptides IFN-τ(1–37), IFN-τ(62–92), IFN-τ(119–150), and IFN-τ(139–172) all show the presence of α helix, while IFN-τ(34–64) and IFN-τ(90–122) are mainly random. The presence of an α helix in the peptides supports the CD analysis of the intact protein and also roughly indicates the location of helical segments. The secondary structure of IFN-τ, including the location of the α helices and loop region, was then predicted using a neural network-based computer program called PHD that relies on sequence alignments of all proteins related to the target sequence [25,26]. When this prediction is correlated with the CD data, peptides that possess considerable α helicity are predicted to contain entire helical segments, and conversely, peptides with little helicity are predicted to be within loop regions. A model of the 3-D structure of IFN-τ was constructed using a distance geometry-based homology modeling method with mouse IFN-β acting as a template. The distance constraints were generated between residues within IFN-τ that are homologous to residues of IFN-β. Dihedral-angle restraints of α helices were generated from the secondary-structure prediction of IFN-τ. No constraints were applied to the 13-residue carboxy tail of IFN-τ, which is absent in IFN-β, since it is likely to be flexible in a manner similar to other proteins such as IFN-γ. Additional distance constraints were added from putative disul-

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_444.html (1 of 2) [4/9/2004 12:11:24 AM]

Document

Page 445

Figure 4 Stereo view of IFN-τ model. Highlighted sequences are from 1–37, 62–92, and 139–172.

fide bridges between residues 1 and 99 and residues 29 and 139. Several structures were generated using distance geometry routines, and the energy was minimized and averaged to yield a final model [16]. A similar model was built by Senda et al. (unpublished results) using a homology modeling method. This model was also built using the x-ray coordinates of IFN-β and shows a similar topology to the IFN-β three-dimensional structure (Figure 4). The most striking feature of both models is that those discontinuous regions, previously determined to be functionally important, are localized to one side of the molecule and found to be spatially contiguous (Figure 4). This observation is consistent with multiple binding sites on IFN-τ interacting simultaneously with the Type I IFN receptor and emphasizes the importance of structural modeling in the understanding and interpretation of functional data. III. Type II IFN A. Functional Sites on the IFN-γ Molecule The production of IFN-γ-neutralizing antibodies specific for an N-terminal peptide of human IFN-γ provided the first evidence that the N-terminus of IFN-γ contained an important functional site [27]. A similar approach was used to produce N-terminus-specific neutralizing antisera against murine IFN-γ [28]. Subsequent studies using IFN-γ synthetic peptides to map the epitope specificity of monoclonal antibodies to murine IFN-γ showed that N-terminal specific monoclonal antibodies neutralize IFN-γ antiviral activity [29]. In receptor-

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_445.html [4/9/2004 12:11:30 AM]

Document

Page 446

competition studies, murine IFN-γ N-terminal peptide consisting of residues 1–39 [IFN-γ (1–39)] blocked both binding to receptor and antiviral activity of IFN-γ [30]. Overlapping peptides of other regions of the IFN-γ molecule failed to block binding and function of IFN-γ [31]. Thus the combination of peptide mapping of epitope specificities and receptor competition using peptides has identified the Nterminus as a structurally and functionally important region of the IFN-γ molecule. This region is highlighted in the sequence of human IFN-γ found in Table 3. Interestingly, site-specific antibodies to the C-terminus of murine IFN-γ, which were induced using the peptide consisting of residues 95–133 [IFN-γ (95–133)], also neutralized IFN-γ activity, however IFN-γ (95–133) failed to block binding of IFN-γ to receptor and IFN-γ activity simultaneously. Antibodies to internal peptides failed to block both antiviral activity and binding of IFN-γ to receptor. In studies with recombinant murine IFN-γ receptor, which consisted of the entire α chain except for the transmembrane domain, the C-terminal peptide did block binding of IFN-γ to receptor [32]. Thus we have the interesting paradox wherein the IFN-γ C-terminal peptide blocked binding of IFN-γ to the recombinant, soluble receptor and yet did not block binding to the cell-surface receptor. One interpretation of these findings has allowed us to formulate the “velcro-key” model of binding to receptor that involves both N- and Cterminal domains of IFN-γ (Figure 5). The N-terminus binds in the “lock and key” manner characterized by specific ligand-receptor binding. The hydrophilic C-terminus binds to a region of the receptor distinct from that for the N-terminus, most likely through its polycationic region, which is conserved across species barriers. Binding of this type would exhibit high affinity and low specificity, similar to a piece of velcro. The C-terminal peptide of IFN-γ would therefore act as a poor competitor for cell-surface binding due to its low specificity alone. This interaction becomes specific in the context of the whole IFN-γ molecule and may increase the affinity of receptor binding. An alternative explanation that may also account for the inability of the C-terminal peptide to compete for cell-surface interactions is that its binding site is located not on the extracellular domain of the receptor, but rather on the intracellular domain. The primary differences between the cell-surface form of the IFN-γ receptor and (2) the accessibility of the recombinant receptor's cytoplasmic domain. A synthetic peptide corresponding to the membrane proximal region of the cytoplasmic domain of the murine IFN-γ receptor was able to bind IFN-γ and specifically compete with the binding of IFN-γ (95–133) to fixed/permeabilized cells [33]. Studies by others have reaffirmed the importance of both the N- and C-terminal regions of IFN-γ in function. Using recombinant DNA techniques, it has been shown that deletion of residues from the Nterminus of the molecule

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_446.html [4/9/2004 12:11:32 AM]

Document

Page 447

Figure 5 Velcro-key model of IFN-γ binding to its receptor. (From Reference 25. Copyright 1992. The American Association of Immunologists.)

results in decreased receptor binding [34]. Deletions or substitutions at the C- terminus have a direct effect on function of the molecule [35–38]. Epitope mapping of neutralizing monoclonal antibodies has also revealed an internal region of the molecule (from residues 84–94) as being functionally important [39]. This sequence bears strong homology to the nuclear localization sequence (NLS) of the SV40 large T antigen and has recently been demonstrated to be fully functional as an NLS for IFN-γ [40]. Thus, internal regions of the IFN-γ molecule are also likely to play an important functional role. B. IFN-γ Receptor α Chain Sites of Interaction with IFN-γ Both the human and the murine IFN-γ receptors consist of a ligand-binding subunit and a speciesspecific cofactor molecule. It is through interaction with this cell-surface receptor complex that IFN-γ exerts its biological effects. The IFN-γ molecule and its N-terminal peptide IFN-γ (1–39) bind specifically to the http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_447.html (1 of 2) [4/9/2004 12:11:36 AM]

Document

Page 448

cell-surface receptor and to a recombinant, soluble form of the ligand-binding chain of the receptor [25]. Synthetic peptides corresponding to the sequence of the extracellular domain of the ligand-binding subunit were used to define the region of the receptor to which the N-terminus of IFN-γ binds. Receptor peptide MIR (95–120) competed most strongly with IFN-γ binding to both cell-surface and recombinant, soluble receptor [41]. Additionally, antisera to this peptide and the adjacent overlapping peptide, MIR (118–143), inhibited the binding of IFN-γ to the recombinant, soluble receptor. Therefore, the receptor domain responsible for binding the N-terminus of IFN-γ is defined by the region encompassing residues 95–120 of the ligand-binding subunit of the IFN-γ receptor and may extend further into the neighboring sequence. Antibodies to the C-terminal region of IFN-γ have been shown to be potent neutralizers of IFN-γ activity [29]. However, no cell-surface binding site for the C-terminus of IFN-γ could be localized using either antisera or synthetic peptides. Furthermore, as indicated above, the C-terminal IFN-γ peptide, IFN-γ (95–133), competed specifically with the intact IFN-γ molecule for binding to a recombinant, soluble form of the receptor, which consists of both the extracellular and the intracellular domains [42]. It was hypothesized that since the intracellular portion of the soluble receptor was accessible, in contrast to that of the cell-surface receptor, the C-terminus of IFN-γ might indeed be binding to this region. In studies using synthetic peptides corresponding to the cytoplasmic domain of the murine IFN-γ receptor, only peptide MIR (253–287) specifically bound both murine IFN-γ and its C-terminal peptide, MuIFN-γ (95–133) [33]. This peptide corresponds to the membrane proximal region of the receptor's cytoplasmic region. Antibodies to this receptor peptide inhibited the binding of the C-terminus of murine IFN-γ to the receptor in cells which had been fixed and permeabilized. Analogous binding studies with human IFN-γ and its C- terminal peptide, HuIFN-γ (95–134), yielded a similar result [43]. Surprisingly, the binding of the IFN-γ C-terminal peptides to their cytoplasmic binding sites is not species restricted, which is in contrast to the binding of the whole molecule at the cell surface. Both human and murine IFN-γ and their C-terminal peptides bound equally well to receptor peptides of either human or murine origin [43]. Thus, a receptor binding site for the C-terminus of the IFN-γ molecule has been localized to the membrane proximal region of the ligand-binding subunit's cytoplasmic domain (Figure 6). Previously, there have been several reports of human IFN-γ having activity on murine cells when administered cytoplasmically [44–46]. With the identification of a cytoplasmic binding site for IFN-γ, which is not species restricted, the question arose as to whether this might be the basis for these earlier observations. Thus, C-terminal IFN-γ peptides of both human and murine origin were used to stimulate murine macrophage lines P388D1 WEHI-3. Macrophages were chosen particularly for their capacity to nonspecifically endocytose material,

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_448.html [4/9/2004 12:11:38 AM]

Document

Page 449

Figure 6 Proposed receptor activation pathway for IFN-γ. (From Reference 53. Copyright 1995. The American Association of Immunologists.)

and we took advantage of this as a means of introducing the IFN-γ peptides into these cells. The IFN-γ Cterminal peptides induced a potent antiviral state in the murine macrophages and upregulated expression of MHC class II molecules, both in a dose-dependent fashion [43]. These effects were demonstrated to be sequence specific, as a scrambled version of the murine C-terminal peptide lacked activity. Furthermore, a truncated form of the murine C-terminal peptide, lacking the sequence of basic amino acids (RKRKR), was also without activity. The absence of activity of this truncated peptide was linked directly to a loss of its ability to bind to the receptor [43]. Therefore, interaction of IFN-γ, via its Cterminus, with its cytoplasmic binding site is important for function and requires the presence of a region of basic amino acids near the C-terminus of the molecule. C. Structural Biology of IFN-γ and the IFN-γ Receptor Structure-function studies of IFN-γ carried out using the synthetic peptide approach and site-specific antibodies indicated that both the N- and C-terminal regions of the protein were not only functionally important, but also accessible at the surface of the molecule. The x-ray crystal structure of human IFN-γ has been determined and reveals that in the IFN-γ homodimer both the N- and the

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_449.html [4/9/2004 12:11:51 AM]

Document

Page 450

Figure 7 Stereo view of IFN-γ [40]. Regions 1–39 of chain A and 95–119 of chain B of the dimer are highlighted.

C-terminus were indeed accessible [3]. Figure 7 illustrates the close proximity of the N- and C-termini of IFN-γ. The subunits of the homodimer are oriented head to tail, such that the N-terminal helix-loophelix (corresponding to residues 1–39) of one IFN-γ molecule interacts with the C-terminus of the second IFN-γ molecule. As mentioned above, synthetic peptides were also instrumental in identifying the region of the receptor to which the N-terminus of IFN-γ binds. Recently, the crystal structure of a complex between murine IFN-γ and the murine IFN-γ Rα subunit has been determined [5]. The synthetic peptides and corresponding antisera had predicted an interaction of murine IFN-γ residues (1–39) with receptor region (95–143). The crystal structure confirmed these observations, indicating an interaction of IFN-γ residues (1–42) with receptor residues (108–132). However, the crystal structure did not define an extracellular binding site for the C-terminus of IFN-γ. There has been some speculation that the basic amino acid residues of the C-terminus may interact with an acidic patch on the receptor's extracellular domain, which would support the previously mentioned “velcro-key” model, but that the crystallization conditions precluded this interaction [5]. It is quite possible that such an interaction may occur as a transitional state prior to the internalization of the C-terminal portion of IFN-γ and interaction with the cytoplasmic region of the receptor. An alternative explanation for the apparent lack of a binding site for the IFN-γ C-terminus on the receptor's extracellular face is that its primary site of interaction is

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_450.html (1 of 2) [4/9/2004 12:11:58 AM]

Document

Page 451

with the cytoplasmic portion of the receptor as described above. Thus, the orientation of the C-terminal portion of the IFN-γ molecules in the receptor complex should be such that they are situated near to the cell membrane. When one examines the structure of the receptor-ligand complex, it is easy to see that this is indeed the case. Studies are currently under way to determine the crystal structure of the complex between the cytoplasmic domain of the IFN-γ receptor and the IFN-γ molecule, in particular the Cterminus. With regards to the definition of sites of interaction between receptors and ligands, the synthetic peptide approach has repeatedly proven to be an accurate indicator of structurally important regions of the IFN-γ/IFN-γ R system. D. Signal Transduction by IFN-γ Within the past several years some of the immediate-early signal transduction events initiated in response to IFN-γ stimulation have been elucidated. Treatment of cells with IFN-γ leads to the rapid activation of two protein tyrosine kinases, JAK1 and JAK2 [47]. The JAK kinases are a newly emerging family of protein kinases important in signaling via cytokines and growth factors. These proteins are unrelated to the src family of tyrosine kinases and are characterized as being larger, having two putative phosphotransferase domains and containing no characteristic SH2 or SH3 domains [48–51]. Members of the Janus kinase family are found associated with the cytoplasmic domains of cytokine and growth factor receptors at or near to the membrane proximal region [51]. In the resting cell, JAK1 and JAK2 are found associated with the α and β/AF-1 chains of the IFN-γ receptor, respectively [52]. These kinases as well as the ligand-binding chain of the IFN-γ receptor are tyrosine phosphorylated in response to IFN-γ treatment [47,53,54]. This leads in turn to the tyrosine phosphorylation of a latent cytoplasmic transcription factor, known variously as p91, Stat 91, or Stat 1 α on tyrosine residue 701 [55]. It is interesting to note that the IFN-α signal-transduction pathway partially overlaps with that of IFN-γ. Stimulation of cells by IFN-α leads to the activation of JAK1 and another Janus family kinase, Tyk2 [56,57]. In turn, this cascade leads to phosphorylation of two latent cytoplasmic transcription factors, p84 (Stat 1β) and p113 (Stat 2β) in addition to the p91 (Stat 1α) activated by IFN-γ. The identification of tyrosine kinases that directly associate with the subunits of the IFN-γ receptor lead to the question of how the binding of IFN-γ might affect these proteins. Recently, the synthetic peptide method was used to identify two regions of the murine IFN-γ receptor's α chain as being important for interaction with the kinase JAK2 [58]. One of these regions lies in the distal portion of the cytoplasmic tail (residues 404–432), while the other (residues 283–309) is nearer to the membrane proximal region to which the C-terminal part of IFN-γ binds (residues 253–287). The fact that there are adjacent binding sites for JAK2 and IFN-γ implied a potential for interaction between the IFN-γ

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_451.html [4/9/2004 12:12:00 AM]

Document

Page 452

ligand and the machinery of signal transduction, namely JAK2. It was found that both intact murine IFNγ and its C-terminal peptide (95–133) are capable of specifically mediating an increase in the degree of association between the recombinant, soluble IFN-γ receptor and JAK2. These findings were further supported as IFN-γ and IFN-γ (95–133) caused an increase in the amount of JAK2 coprecipitating with the receptor from intact murine macrophages [58]. This has been the first such demonstration of an extracellular cytokine ligand participating directly in interaction with cytoplasmic signaling elements. E. IFN-γ as a Candidate for Rational Drug Design The IFN-γ molecule is a potentially attractive model for the rational application of drug-design strategies. Reagents exist that are capable of either positively or negatively modulating the in vivo effects of IFN-γ. An initial target is quite simply at the level of receptor-ligand interaction. Synthetic peptide analogs of the N-terminal region have been successfully applied in vitro to inhibit interaction of intact IFN-γ with cell-surface receptors [30]. Interestingly, it has been observed that the N-terminal region of mouse IFN-γ has the ability to interact with the human receptor [59]. It was shown that the mouse peptide IFN-γ (1–39) had a 10-fold greater ability to block the binding of human IFN-γ to cellsurface receptors. This was shown to be correlated with a more stable structure in solution for the murine peptide and illustrates the importance of stable structure to receptor binding, which may be exploited when designing peptide mimetics. The solution structure of this peptide has also been determined and could provide the beginning steps for determining the structural requirements of an antagonist [7]. Future studies could focus on cocrystallization of the peptides with receptor or NMR studies of peptide domains of the receptor and IFN-γ. Recombinant, soluble forms of the extracellular domain of the ligand-binding subunit of the receptor have also been used in analogous fashion both in vitro to inhibit cell-surface binding and in vivo to interfere with disease progression [60]. Therefore prevention of potentially deleterious effects of IFN-γ may be achieved by preventing initial interactions with receptor molecules at the cell surface. A second candidate region lies within amino acid residues 84–94 of human IFN-γ and the corresponding region of its murine homologue. This portion of the molecule functions as a nuclear localization signal and, therefore, is also an attractive target for drug design. In cells treated with IFN-γ, the IFN-γ molecule traffics rapidly to the nucleus of the cell, usually within one to two minutes. When the IFN-γ molecule is crosslinked to its receptor, the resultant receptor-ligand complex migrates to the nucleus [40]. The implication is that this sequence may therefore be of use in artificially targeting proteins from the cytoplasm directly to the nucleus. This is potentially a very attractive method

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_452.html [4/9/2004 12:12:03 AM]

Document

Page 453

for directed subcellular targeting of synthetic transcriptional activators or repressors that might not otherwise be directed to the nucleus. Finally, the C-terminal region of IFN-γ offers another possibility. With respect to the induction of an antiviral state, or the upregulation of MHC class II molecule expression, synthetic peptides corresponding to the C-terminal 39 amino acids of either human or murine IFN-γ function as potent agonists. This activity is due to the interaction of these peptides with the cytoplasmic domain of the IFNγ receptor and the associated protein tyrosine kinases. Furthermore, in contrast to the stringent species specificity of intact IFN-γ, the action of the C-terminal peptides agonists is not limited by species constraints. This property therefore renders the C-terminal portion of the IFN-γ molecule an attractive model for the development of IFN-γ agonists and antagonists. The identification of the C-terminus of IFN-γ as that part of the molecule that contacts the cytoplasmic portion of the receptor implies that in developing IFN-γ agonists, the primary focus should be on this region of the protein. Corresponding antagonists may be developed based upon the portion of the receptor to which the IFN-γ C-terminus binds. IV. Conclusion The interest in IFNs as therapeutics has existed from their initial discovery in 1957. Since then scientists have been trying to understand the mechanism of their action and apply that knowledge to the treatment of many different diseases, meeting with some success. The effort now is to understand how IFNs work at the molecular level, with the goal being to design better, more specific therapeutics. Through structure/function studies, we now know where the functional sites lie on many of the IFNs. We also know the sites of interaction with their receptors and second messenger systems. From these studies, initial candidates for structure-based drug design have been identified. Although more work is needed to further characterize the IFNs and their receptor systems, the challenge now is to apply our existing knowledge and create second generation molecules that can modulate the many activities of these fascinating proteins. References 1a. Isaacs A and Lindenmann J. Virus Interference. I. The interferon. Proc R Soc London Ser B 1957; 147:258. 1b. Johnson HM, Bazer FW, Szente BE, Jarpe MA. How interferons fight disease. Scientific American 1994; 270:40–47. 2. Senda T, Shimazu T, Matsuda S, Kawano G, Shimizu H, Nakamura KT, Mitsui Y. Three-dimensional crystal structure of recombinant murine interferon-β. EMBO J 1992; 11:3193–3201.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_453.html (1 of 2) [4/9/2004 12:12:07 AM]

Document

Page 454

3. Ealick SE, Cook WJ, Vijay-Kumar S, Carson M, Nagabhushan TL, Trotta PP, Bugg CE. Threedimensional structure of recombinant human interferon-g. Science 1991; 252:698–702. 4. Samudzi CT, Burton LE, Rubin JR. Crystal structure of recombinant rabbit interferon-gamma at 2.7 A resolution. J Biol Chem 1991; 266:21791–21797. 5. Walter MR, Windsor WT, Nagabhushan TL, Lundell DJ, Lunn CA, Zavodny PJ, Narula SK. Crystal structure of a complex between interferon-g and its soluble high-affinity receptor. Nature 1995; 376:230–235. 6. Grzesiek S, Dobeli H, Gentz R, Garotta G, Labhardt AM, Bax A. 1H, 13C, and 15N NMR backbone assignments and secondary structure of human interferon-gamma. Biochemistry 1992; 31 (35):8180–90. 7. Sakai TT, Jablonski MJ, DeMuth PA, Krishna NR, Jarpe MA, Johnson HM. Proton NMR sequence specific assignments and secondary structure of a receptor binding domain of mouse γ-interferon. Biochemistry 1993; 32:5650. 8. Murgolo NJ, Windsor WT, Hruza A, Reichert P, Tsarbopoulos A, Baldwin S, Huang E, Pramanik B, Ealick S, Trotta PP. A homology model of human interferon alpha-2. Proteins 1993; 17(1):62–74. 9. Seto MH, Harkins RN, Adler M, Whitlow M, Church WB, Croze E. Homology model of human interferon-alpha-8 and its receptor complex. Protein Science 1995; 4:655–70. 10. Jarpe MA, Johnson HM, Bazer FW, Ott TL, Curto EV, Rama Krishna N, Pontzer CH. Predicted structural motif of IFN-τ. Protein Engineering 1994; 7:863–867. 11. Bazer FW. Mediators of maternal recognition of pregnancy in mammals. Proc Soc Exp Biol Med 1992; 199:373–384. 12. Imakawa K, Anthony RV, Kazemi M, Marotti KR, Polites HG, Roberts RM. Interferon-like sequence of ovine trophoblast protein secreted by embryonic trophectoderm. Nature 1987; 330:377–379.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_454.html (1 of 2) [4/9/2004 12:12:10 AM]

Document

13. Stewart HJ, McCann SHE, Barker PJ, Lee KE, Lamming GE, Flint APF. Interferon sequence homology and receptor binding activity of ovine trophoblast antileuteolytic protein. J Endocrinol 1987; 115:R13–R15. 14. Pontzer CH, Ott TL, Bazer FW, Johnson HM. Structure/function studies with interferon tau: Evidence for multiple active sites. J Interferon Res 1994; 14:133–141. 15. Fish EN, Banerjee K, Stebbing N. The role of three domains in the biological activity of human interferon-α. J Interferon Res 1989; 9:97–114. 16. Novick D, Cohen B, Rubinstein M. The human interferon α/β receptor: characterization and molecular cloning. Cell 1994; 77:391–400. 17. Pontzer CH, Yamamoto JK, Bazer FW, Ott TL, Johnson HM. Potent anti-feline immunodificiency virus and anti-human immunodeficiency virus effect of interferon tau. J Immunol 1995; (in press). 18. Pontzer CP, Bazer FW, Johnson HM. Antiproliferative activity of a pregnancy recognition hormone, ovine trophoblast protein-1. Cancer Res 1991; 51:5304–5307. 19. Pestka S, Langer JA, Zoon KC, Samuel CE. Interferons and their actions. Annu Rev Biochem 1987; 56:727–777. 20. Barasoain I, Portolès A, Aramburu JF, Rojo JM. Antibodies against a peptide representative of a conserved region of human IFN-α. Differential effects on the antiviral and antiproliferative effects of IFN. J Immunol 1989; 143:507–512.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_454.html (2 of 2) [4/9/2004 12:12:10 AM]

Document

Page 455

21. McInnes B, Chambers PJ, Cheetham BF, Beilharz MW, Tymms MJ. Structure-function studies of interferons-α: Amino acid substitutions at the conserved residue tyrosine 123 in human interferon-α1. J Interferon Res 1989; 9:305–314. 22. Waine GJ, Tymms MJ, Brandt ER, Cheetham BF, Linnane AW. Structure-function study of the region encompassing residues 26–40 of human interferon-α4: Identification of residues important for antiviral and antiproliferative activities. J Interferon Res 1992; 13:42–48. 23. Ott TL, Van Heeke G, Johnson HM, Bazer FW. Cloning and expression in Saccharomyces cerevisiae of a synthetic gene for the type-1 trophoblast interferon ovine trophoblast protein-1: Purification and antiviral activity. J Interferon Res 1990; 11:357–364. 24. Subramaniam PS, Khan SA, Pontzer CH, Johnson HM. Differential recognition of the type In IFN receptor by IFN-τ and IFN-α is responsible for their differential cytotoxicities. 1995; (submitted). 25. Sander C, Schneider R. Database of homology-derived structure and the structural meaning of sequence alignment. Proteins 1991; 9:56–68. 26. Rost B, Sander J. Prediction of protein structure at better than 70% accuracy. J Mol Biol 1993; 232:544–599. 27. Johnson HM, Langford MP, Lakchaura B, Chan TS, Stanton GJ. Neutralization of native human gamma interferon by antibodies to a synthetic peptide encodded by the 5' end of human gamma interferon cDNA. J Immunol 1982; 129:2357–2359. 28. Langford MP, Gray PW, Stanton GJ, Lakchaura B, Chan T-S, Johnson HM. Antibodies to a synthetic peptide corresponding to the N-terminal end of mouse gamma interferon (IFN-α). Biochem Biophys Res Comm 1983; 117:866–871. 29. Russell JK, Hayes MP, Carter MJ, Torres BA, Dunn BM, Russell SW, Johnson HM. Epitope and functional specificity of monoclonal antibodies to mouse interferon gamma: the synthetic peptide approach. J Immunol 1986; 136:3324–3328. 30. Magazine HI, Carter JM, Russell JK, Torres BA, Dunn BM, Johnson HM. Use of synthetic peptides to identify an N-terminal epitope on mouse gamma interferon that may be involved in function. Proc Natl Acad Sci USA 1988; 185:1237–1241. 31. Jarpe MA, Johnson HM. Topology of receptor binding domains of mouse IFN-α. J Immunol 1990; 145:3304–3309. 32. Griggs ND, Jarpe MA, Pace JL, Russell SW, Johnson HM. The N-terminus and C- terminus of interferon gamma are binding domains for cloned soluble interferon gamma receptor. J Immunol 1992; 149:517–520. http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_455.html (1 of 2) [4/9/2004 12:12:22 AM]

Document

33. Szente BE, Johnson HM. Binding of IFN-α and its C-terminal peptide to a cyto-plasmic domain of its receptor that is essential for function. Biochem Biophys Res Comm 1994; 201:215–221. 34. Zavodny PJ, Petro ME, Chiang TR, Narula SK, Leibowitz PJ. Alterations of the amino terminus of murine interferon gamma: expression and biological activity. J Interferon Res 1988; 8:483–494. 35. Arakawa T, Hsu YR, Parker CG, Lai PH. Role of polycationic C-terminal portion in the structure and activity of recombinant human interferon gamma. J Biol Chem 1986; 261:8534–8539. 36. Leinikki PO, Calderon J, Luquette MH, Schreiber RD. Reduced receptor binding by a human interferon gamma fragment lacking 11 carboxyl-terminal amino acids. J Immunol 1987; 139:3360–3366. 37. Wetzel R, Perry LJ, Veilleux C, Chang G. Mutational analysis of the C-terminus of human interferon gamma. Prot Eng 1990; 3:611–623.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_455.html (2 of 2) [4/9/2004 12:12:22 AM]

Document

Page 456

38. Lundell D, Lunn C, Dalgarno D, Fossetta J, Greenberg R, Reim R, Grace M, Narula S. The carboxylterminal region of human interferon-gamma is important for biological activity: mutagenic and NMR analysis. Prot Eng 1991; 4(3):335–341. 39. Zu X, Jay FT, The E1 functional epitope of the human interferon-α is a nuclear targeting signal-like element. J Biol Chem 1991; 266:6023–6026. 40. Bader T, Wietzerbin J. Nuclear accumulation of interferon-gamma. Proc Natl Acad Sci USA 1994;91:11831–11835. 41. Van Volkenburg MA, Griggs ND, Jarpe MA, Pace JL, Russel SW, Johnson HM. Binding site on the murine interferon-gamma receptor for interferon-gamma has been identified using the synthetic peptide approach. J Immunol 1993; 151:6206– 6213. 42. Fernando LP, LeClaire RD, Obici S, Zavodny PJ, Russell SW, Pace JL. Stable expression of a secreted form of the mouse IFN-α receptor by rate cells. J Immunol 1991; 147:541–547. 43. Szente BE, Soos JM, Johnson HM. The C-terminus of IFN-α is sufficient for intracellular function. Biochem Biophys Res Comm 1994; 203:1645–1654. 44. Fidler IJ, Fogler WE, Kleinerman ES, Saiki I. Abrogation of species specificity for activation of tumoricidal properties in macrophages by a recombinant mouse or human interferon gamma encapsulated in liposomes. J Immunol 1985; 135:4289–4296. 45. Sancéau J, Sondermeyers P, Béranger F, Falcoff R, Vaquero C. Intracellular human interferon triggers an antiviral state in transformed murine L cells. Proc Natl Acad Sci USA 1987;84:2906–2910. 46. Smith MR, Muegge K, Keller JR, Kung HF, Young HA, Durum SK. Direct evidence for an intracellular role for interferon-gamma. J Immunol 1990; 144:1777–1782.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_456.html (1 of 2) [4/9/2004 12:12:24 AM]

Document

47. Igarashi K, Garotta G, Ozmen L, Ziemiecki A, Wilks AF, Harpur AG, Larner AC, Finbloom DS. Interferon gamma induces tyrosine phosphorylation of interferon gamma receptor and regulated association of protein tyrosine kinases, Jak1 and Jak2, with its receptor. J Biol Chem 1994; 269:14333–14336. 48. Firmbach-Kraft I, Byers M, Shows T, Dalla-Favera R, Krolewski JJ. Tyk2, prototype of a novel class of non-receptor tyrosine kinase genes. Oncogene 1990; 5:1329–1336. 49. Bernards A. Predicted tyk2 protein contains two tandem protein kinase domains. Oncogene 1991; 6:1185–1187. 50. Wilks AF, Harpur AG, Kurban RR, Ralph SJ, Zurcher G, Ziemiecki A. Two novel protein tyrosine kinases, each with a second phosphotransferase-related catalytic domain, define a new class of protein kinase. Mol Cell Biol 1991; 11:2057–2065. 51. Ihle JN, Witthuhn BA, Quelle FW, Yamamoto K, Silvennoinen O. Signaling through the hematopoietic cytokine receptors. Annu Rev Immunol 1995; 13:369–398. 52. Sakatsume M, Igarashi K-I, Winestock KD, Garotta G, Larner AC, Finbloom DS. The Jak Kinases differentially associate with the a and b (accessory factor) chains of the interferon-g receptor to form a functional receptor unit capable of activating STAT transcription factors. J Biol Chem 1995; 270:17528–17534. 53. Khurana Hershey GK, McCourt DW, Schreiber RD. Ligand-induced phosphorylation of the human interferong receptor. J Biol Chem 1990; 265:17868–17875.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_456.html (2 of 2) [4/9/2004 12:12:24 AM]

Document

Page 457

54. Greenlund AC, Farrar MA, Viviano BL, Schreiber RD. Ligand induced interferon-gamma receptor tyrosine phosphorylation couples the receptor to its signal transduction system (p912). EMBO J 1994; 13:1591–1600. 55. Shuai K, Start GR, Kerr IM, Darnell JE. A single phosphotyrosine residue of Stat91 required for gene activation by interferon gamma. Science 1993; 261:1744–1746. 56. Muller M, Briscoe J, Laxton C, Guschin D, Ziemiecki A, Silvennoinen O, Harpur AG, Barbieri G, Witthuhn BA, Schindler C, Pellegrini S, Wilks AF, Ihle JN, Stark GR, Kerr IM. The protein tyrosine kinase JAK1 complements defects in interferon alpha/beta and gamma signal transduction. Nature 1993; 366:129–135. 57. Barbieri G, Velazquez L, Scrobogna M, Fellous M, Pellegrini S. Activation of the protein kinase tyk2 by interferon α/β. Eur J Biochem 1994;223:427–435. 58. Szente BE, Subramaniam PS, Johnson HM. Identification of IFN-γ receptor binding sites for JAK2 and enhancement of binding by IFN-γ and its' C-terminal peptide IFN-γ(95–133). J Immunol 1995; 155:95–133. 59. Jarpe MA, Johnson HM. Stable conformation of IFN-γ receptor binding peptide in aqueous solution is required for IFN-γ antagonist activity. J Interferon Res 1993; 13:99. 60. Ozmen L, Roman D, Fountoualakis M, Schmid G, Ryffel B, Garotta G. Soluble interferon-gamma receptor: a therapeutically useful drug for systemic lupus erythematosus. J Interferon Res 1994; 14(5):283–284.

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_457.html [4/9/2004 12:12:28 AM]

http://legacy.netlibrary.com/reader/message.asp?message=811&BookID=12640&FileName=Page_458.html

The requested page could not be found. Return to previous page

http://legacy.netlibrary.com/reader/message.asp?message=811&BookID=12640&FileName=Page_458.html [4/9/2004 12:12:31 AM]

Document

Page 459

18 The Design of Anti-Influenza Virus Drugs from the X-ray Molecular Structure of Influenza Virus Neuraminidase Joseph N. Varghese Biomolecular Research Institute, Melbourne, Victoria, Australia I. Introduction Influenza has plagued humankind since the dawn of history and continues to affect a significant proportion of the population irrespective of age or previous infection history. These periodic epidemics that reinfect otherwise healthy people have devastated communities world wide. Some pandemics like the 1917–1919 “Spanish flu” were responsible for the death of tens of millions of people throughout the world. The origins, spread, and severity of influenza epidemics have been a puzzle that has only in the last two decades been adequately addressed. In early times it was thought that the disease was the evil influence (sic) of the stars, and other extraterrestial objects. At present it is generally accepted that the disease is of viral origin, spread by aerosols produced by infected animals, and the continual production of new strains of the virus results in reinfection of the disease (reviewed in Reference 1). There are three types of influenza virus classified on their serological cross-reactivity with viral matrix proteins and soluble nucleoprotein (A, B, and C). Only type A and B are known to cause severe human disease. Type B is only found in humans, while type A has a natural reservoir in birds and some mammals like pigs and horses [2]. Influenza, an orthomyxovirus, is a 100 nm lipid-enveloped virus (Figure 1) containing an eight-segment negative single-stranded genome [3]. Two of the segments code for the surfaces glycoproteins, hemagglutinin (which binds to terminal sialic acid), and neuraminidase (which

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_459.html [4/9/2004 12:12:33 AM]

Document

Page 460

Figure 1 A schematic diagram of an influenza virus particle that illustrates its constituent components and morphology. The surface antigens hemagglutinin and neuraminidase are attached to the lipid and matrix protein shell that encapsulates the eight negative-stranded RNA genes of the virus and associated nucleoprotein and polymerase.

cleaves terminal sialic acid) and which appear as spikes protruding out of the viral envelope. The viral target in humans is the upper respiratory tract epithelial cells. Replication (see Figure 2) begins with penetration of the virion through the mucin layer covering the epithelial surface, followed by attachment to the viral receptor by the hemagglutinin. Penetration of the cell is achieved by endocytosis and the virion core is released after the fusion of the virion and vesicle membrane mediated by the hemagglutinin. Fusion is enabled by a conformational change in the hemagglutinin made possible by lowering the pH of the endosome by the M2 ion channel protein. Following replication, the progeny virions are released by budding off the cell membrane [4,5].

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_460.html (1 of 2) [4/9/2004 12:12:42 AM]

Document

Page 461

Figure 2 A simplified schematic of the replication cycle of an influenza virion in the host respiratory epithelia. Details of viral transport through mucin and pathways of viral spread on budding from the epithelial cell membrane during replication is not understood. Neuraminidase activity is important for release of budding progeny virions, desialyation of viral glycoproteins, and probably facilitates transport through sialic-acidrich mucin.

Release of virions occur 8 hours post infection and the onset of infection is sudden, resulting in pyroxia, muscular and joint pain, and a dry cough [6]. Virus shedding continues for up to a week, when a rise in virus-specific antibody clears the virus from the host. The vulnerability of the host succumbing to viremea during this week of rising viral titer is mediated by interferon induction [7] 48 hours post infection, which attenuates viral replication until the cell-mediated immune response begins to clear the virus. The severity of the illness is thought to depend on the level of cross protection arising from antibodies raised from previous influenza infections [19]. The course of the illness can be debilitating, and no effective treatment is available at present to halt the progression of the disease. Death can result for susceptible populations (neonate and elderly) primarily as a result of secondary infections [8]. This chapter shall examine a structural basis for the continual emergence of new influenza strains, and the reasons current vaccines against influenza fail to protect against all

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_461.html (1 of 2) [4/9/2004 12:13:07 AM]

Document

Page 462

strains of influenza. The discovery of the active site of influenza neuraminidase and the exploitation of its structural conservation shall be discussed in terms of the design of potent neuraminidase inhibitors. The potential therapeutic use of these inhibitors as antiviral drugs against influenza virus infections shall be examined. A. Antigenic Variation The plethora of different strains of virus that are responsible for the continued reinfection of virus in humans is primarily related to mutations in the viral genes of two surface glycoproteins, hemagglutinin and neuraminidase [9]. The current paradigm for this genetic variation [10,11] is that these mutations arise primarily from incremental changes in the amino acid sequences of these glycoproteins by selection pressure of the immune system of the infected host. This mechanism termed “Antigenic Drift” accounts for most of the strain variation within a particular subtype of influenza. However, infrequently a mutation arises by genetic reassortment of viruses from different animal hosts (“Antigenic Shift”) whereby an entirely new gene for one of the surface glycoproteins is generated that is significantly different (~50%) in amino acid sequence from the parent virus. This is the mechanism by which new subtypes of influenza arise and are primarily responsible for the major pandemics that occur. Strains of influenza virus are classified by type (A, B, or C), geographic location, date of original isolation, and the subtype of the hemagglutinin and neuraminidase antigens. There exist 9 known subtypes (N1 to N9) of neuraminidase and 13 known subtypes (H1 to H13) of hemagglutinin for influenza A in all animal populations. Two neuraminidase (N1 and N2) and three hemagglutinin (H1, H2, and H3) subtypes of influenza A have occurred in strains that have infected humans since 1933 when isolates were first characterized [12]. Prior to 1933 there is indirect evidence of antigenic shift occurring in human populations [13]. The N1 subtype was associated with virus isolated between 1933 and 1957, after which time the N2 subtype appeared in the Asian influenza. No major change in the structure of neuraminidase has occurred since, although the hemagglutinin subtype has changed from H2 to H3 in 1968 in the Hong Kong pandemic, and H1N1 reappeared in 1978 as the Russian pandemic. Influenza B, which infects only human hosts, has only one subtype, but like type A undergoes continual antigenic drift. B. Current Therapeutics Amantidine and Rimantidine are the only class of drugs that have been approved for therapy. At high concentration (>50 mg/mL) Amantidine is thought to buffer the pH of the endosome and prevent the conformational

http://legacy.netlibrary.com/nlreader/nlReader.dll?bookid=12640&filename=Page_462.html [4/9/2004 12:13:12 AM]

Document

Page 463

change of the hemagglutinin necessary for fusion. Drug-resistant mutants arise where the hemagglutinin trimers are thought to be less stable than the wild type [14]. At low concentrations (