00:00:30 Welcome to the 11th American Chemical Society Satellite Television Seminar.

00:00:59 I'm your moderator, Paul Anthony, and we're pleased that you were able to join us today.

00:01:03 Since we began running satellite TV seminars in 1989, we've constantly tried to offer you

00:01:08 something unique in each program.

00:01:11 We've had world-famous scientists talk about their specialties, and ACS presidents host

00:01:15 programs covering the latest developments in technology.

00:01:18 Our previous programs have included primarily a North American perspective.

00:01:23 Today we're retaining these established features, and we're adding a new wrinkle to the program.

00:01:29 A transatlantic outlook.

00:01:31 Two speakers are joining us via telephone from Europe, and several groups are watching

00:01:35 our broadcast in Europe as well.

00:01:37 We hope that you'll enjoy this expanded international character.

00:01:42 The panelists here in the studio are Helen Free, current president of the American Chemical

00:01:46 Society, Jonathan Greer of Abbott Laboratories and the technical coordinator for this program,

00:01:53 John Baldwin, distinguished senior scientist at Merck Laboratories, John Erickson, director

00:01:58 of the structural biochemistry program at the National Cancer Institute, and Michael

00:02:03 Varney, director of medicinal chemistry at Agaran Pharmaceuticals.

00:02:09 Joining us in Europe are Gerard Folkers and Didier Rognon from the Department of Pharmacy

00:02:14 of the Swiss Federal Institute of Technology in Zurich, and Peter Warner, principal medicinal

00:02:20 chemist at Zeneca Pharmaceuticals in England.

00:02:23 Helen, it's still early in your term as ACS president, but would you share your focus

00:02:28 for ACS this year and what this program has to do with it?

00:02:31 I'd be glad to, Paul.

00:02:33 You, the ACS members in the audience, have given me the opportunity to make a personal

00:02:38 impact on one of the largest and most influential scientific organizations in the world, and

00:02:44 I welcome that challenge.

00:02:46 To me, the American Chemical Society is a dream team that's comparable to last summer's

00:02:51 American Olympic basketball team with star players from a variety of different professional

00:02:57 teams.

00:02:58 We have a host of all-star players, the American Chemical Society board, the council, many

00:03:04 committees, 145,000 members, our professional staff, and divisions and local sections.

00:03:10 This diverse group plays a coordinated game in which nobody loses and everybody in the

00:03:15 chemical professions wins.

00:03:18 There are five objectives that I personally want to emphasize this year, and two in particular

00:03:23 apply to today's program.

00:03:25 The five are to improve the public's awareness of chemistry's contributions to the quality

00:03:31 of our daily lives, to encourage children's interest in science at an early age and maintain

00:03:37 that interest, to provide opportunities for active participation by industries, non-PhD

00:03:44 chemists, to interact with scientific groups from other countries towards a world chemistry

00:03:49 celebration in 1999, and to urge more members to be more active in the American Chemical

00:03:56 Society.

00:03:58 The third and fourth points on this list are especially pertinent to this program with

00:04:03 its focus on successes in industrial research and on international communication.

00:04:09 ACS thrives in no small part because the chemical community we serve includes a creative and

00:04:16 profitable chemical and chemical processing industry.

00:04:20 I plan to do everything I can to encourage ACS members who work in industry to play more

00:04:25 active roles in ACS activities.

00:04:29 For example, the theme of the American Chemical Society national meeting in Chicago this fall

00:04:34 will be industry.

00:04:36 Many ACS divisions are already arranging a variety of industry-oriented symposia and

00:04:41 technical sessions.

00:04:43 Some of these are specifically designed for non-PhD chemists.

00:04:47 We need also to look beyond our national borders as we're doing today.

00:04:52 One of the most wonderful things that can be said of scientists is that no matter what

00:04:55 the political climate, they find ways to keep in touch and to participate in a continual

00:05:01 flow of information.

00:05:03 The conduct of science transcends borders, cultures, and ideology.

00:05:07 We're without doubt part of a remarkable establishment.

00:05:12 For several years, the American Chemical Society has been sponsoring a National Chemistry

00:05:17 Week in the United States during November.

00:05:20 It's an effective, local-level outreach program to show the public the benefits of science

00:05:25 and chemistry in particular.

00:05:28 In 1999, we plan to expand this to a World Chemistry Celebration.

00:05:33 I encourage you chemists in our international audience to work with your own national chemical

00:05:39 societies to join with us in this event.

00:05:42 We also plan to continue with the international aspect of these satellite television seminars,

00:05:48 and we'll continue to bring you speakers of the highest caliber discussing cutting-edge

00:05:53 advances.

00:05:55 For those in the audience who are not ACS members, I invite you to join and to become

00:06:00 active contributors to this myriad of programs.

00:06:03 We've even included a membership application form in your seminar notes.

00:06:07 If you're already a member, give your application to a colleague and offer to sponsor his or

00:06:12 her membership.

00:06:14 The way to make the most of 1993 is for us all to make it together.

00:06:19 Thank you, Helen, very much.

00:06:21 Next, Jonathan Greer will present a brief introduction to molecular modeling and its

00:06:25 role in drug discovery and introduce the rest of the program.

00:06:29 Jonathan?

00:06:30 Thank you, Paul.

00:06:32 Today we have what I believe you will find an exciting program, the latest of an interesting

00:06:37 series of ACS satellite television presentations on the use of molecular modeling and the

00:06:42 three-dimensional structure of macromolecules to study biological function and ligand binding

00:06:48 and design.

00:06:49 In 1990, the first program of this series, Molecular Modeling for Biological Systems,

00:06:55 included an introduction to the use of molecular dynamics, host-guest chemistry, distance geometry,

00:07:02 and solution structures from NMR data.

00:07:04 In 1991, a second program, Developing and Using Protein Models, presented the range

00:07:10 of methods that currently exist to determine three-dimensional structures of proteins,

00:07:15 including X-ray crystallography, NMR, simulations, and comparative modeling.

00:07:22 These programs established a background understanding of the methods and techniques used to study

00:07:27 macromolecular structure, both experimental and theoretical, and are still available on

00:07:32 videotape from the ACS.

00:07:34 I've included a page in your handouts about them.

00:07:39 We wanted to devote this ACS satellite presentation to the application of these methods for the

00:07:44 solution of problems of particular interest to chemists.

00:07:48 Knowledge of protein structure has proven valuable in a broad range of cases, such as

00:07:53 to help understand protein function and to rationalize molecular biological manipulation

00:07:59 of the protein structure through site-specific mutagenesis.

00:08:02 However, from the perspective of the chemist, perhaps the most challenging and promising

00:08:07 use of protein structure is in the rational design of novel therapeutic agents using the

00:08:13 three-dimensional structure of the protein target molecule whose function we are attempting

00:08:17 to control and modify.

00:08:20 We are presenting today a series of talks that illustrate how protein structure and

00:08:24 molecular modeling can indeed, under the correct circumstances, be used to aid in

00:08:29 the design of new drug entities.

00:08:33 For many years, the discovery of new drugs has been achieved by taking a lead structure

00:08:38 and iterating cycles of new compound synthesis with biological testing of those compounds

00:08:43 to produce some measure of therapeutic efficacy.

00:08:46 The initial lead structures were either the natural ligand or the result of a random screening

00:08:51 program of compounds using in vitro or even in vivo tests.

00:08:55 Indeed, frequently projects were chosen according to the results of random screening.

00:09:01 Analogs were selected for synthesis during the iteration cycle based upon a combination

00:09:06 of inspired trial and error and medicinal chemistry experience and intuition.

00:09:12 By contrast, rational drug design, widely practiced today in the pharmaceutical industry,

00:09:17 requires selecting a protein target molecule whose natural ligand will be manipulated to

00:09:22 produce an inhibitor, an agonist, or an antagonist based upon the therapeutic need and capitalizing

00:09:30 upon our knowledge of the mechanism of action of the protein ligand complex.

00:09:34 In order to permit the chemists to more fruitfully design modifications of the lead structure,

00:09:39 it's helpful to have a three-dimensional structure for the bioactive confirmation of the ligand

00:09:44 as it binds to the receptor.

00:09:48 Experience has taught us that this confirmation is not the solution structure, nor is it necessarily

00:09:53 the crystal structure of the ligand.

00:09:55 Rather, it's the confirmation of the ligand when it's bound to its receptor or enzyme-active site.

00:10:03 Knowledge of the bioactive confirmation should better permit the chemist to modify analogs

00:10:07 constructively to produce novel structures that are potent and specific.

00:10:12 Given knowledge of the bioactive confirmation of the ligand, it should be useful to understand

00:10:17 the detailed interactions of the ligand with its receptor protein.

00:10:21 This allows the chemist to preserve the critical interactions with the protein while modifying

00:10:26 the analog to interact more precisely with the receptor and to occupy subsidiary sites,

00:10:32 resulting in better potency and specificity.

00:10:35 With the structure of the target protein ligand complex, one can better understand the structure-activity

00:10:41 relationships of existing compounds, suggest new analogs to synthesize in current series,

00:10:47 and develop novel concepts and ideas for completely new ligand moieties.

00:10:53 Obtaining the three-dimensional structure of the protein target frequently requires

00:10:57 considerable effort to clone the appropriate DNA, express and purify multi-milligram amounts

00:11:03 of the protein, followed by crystallization and structure solution.

00:11:07 For NMR studies, isotope labeling is crucial to the structure determination.

00:11:13 Not only is the three-dimensional structure of the protein target desired, but especially

00:11:17 complexes with the lead ligand compounds of interest to the chemist.

00:11:22 However, the existence of an experimental structure of the protein ligand complex allows

00:11:27 one to go beyond examining the binding site and using this information to design new analogs.

00:11:33 It allows one to take the designed, synthesized, and assayed compound and return it to the

00:11:39 crystal or NMR tube and redetermine the structure of this complex with this new compound.

00:11:46 Thus one can determine experimentally whether the design concept was structurally correct.

00:11:51 If the molecule was potent, was it potent for the correct reasons built into the design?

00:11:55 If the compound was weaker than expected, how and why did the design concept fail?

00:12:00 First of all, the new structure can be used as the basis for a new round of design, synthesis,

00:12:06 and compound testing.

00:12:08 This process can be iterated with further rounds of protein ligand structure determination,

00:12:13 design, synthesis, testing, and so on to ultimately produce potent and specific compounds.

00:12:20 The design cycle depicted here has proven to be the most powerful implementation of

00:12:25 structurally based drug design and is the subject of our program today.

00:12:30 Indeed, the program being presented today is devoted to illustrating how the above ideas

00:12:36 have been successfully reduced to practice, allowing the design and synthesis of novel,

00:12:41 more potent, and specific compounds than would otherwise have been achieved.

00:12:46 The subjects to be covered are the use of structure and the design of novel HIV protease

00:12:52 inhibitors by John Erickson, the design of carbonic anhydrase inhibitors for the treatment

00:12:58 of glaucoma by John Baldwin, and the design of thymidylate synthase inhibitors as anti-cancer

00:13:04 agents by Michael Varney.

00:13:07 These three presentations were videotaped in advance at each of the speaker's home institutions.

00:13:13 At the beginning of each of the two question sessions, there will be a short live presentation

00:13:18 from our speakers in Europe, the first by Gerard Falkers and Didier Ragnan from the

00:13:23 IHTH in Switzerland, and will describe their work on peptide binding to major histocompatibility

00:13:29 complex protein, and the second by Peter Warner of Zeneca Pharmaceuticals in England, who

00:13:36 will tell about their studies on human leukocyte elastase inhibitors for emphysema.

00:13:42 You should have received printed seminar notes that contain copies of most of the material

00:13:46 that will be shown on the television screen.

00:13:49 We hope that this will help you to follow and take notes on the presentations.

00:13:53 We expect that you will find this exciting program of great interest.

00:13:57 And now, we will hear from John Erickson.

00:13:59 Dr. John W. Erickson is the Director of the Structural Biochemistry Program at the National

00:14:07 Cancer Institute in Frederick, Maryland.

00:14:10 His academic and pharmaceutical industry research experiences span the areas of cellular immunology,

00:14:17 virology, x-ray crystallography, and drug discovery.

00:14:22 Dr. Erickson and his former colleagues at Abbott Laboratories pioneered the design of

00:14:27 structure-based, C2-symmetric HIV protease inhibitors.

00:14:31 Dr. Erickson's topic today is the use of structure in the design of HIV protease inhibitors.

00:14:39 One of the most active areas of drug discovery research in the world today concerns the efforts

00:14:46 to stem the tide of the AIDS pandemic.

00:14:50 With the discovery that HIV, a retrovirus, causes AIDS, we have witnessed an explosion

00:14:56 of information on the molecular biology of this virus.

00:14:59 This has resulted in strategies that target virtually every aspect of the viral life cycle

00:15:05 towards the design of specific antiviral agents.

00:15:08 In today's lecture, I would like to discuss efforts to use structure in the design of

00:15:14 inhibitors for HIV protease.

00:15:18 This is a very active area of research today, and I'm only going to focus on the research

00:15:24 that I've been directly involved with, formerly with my colleagues at Abbott Labs and presently

00:15:29 here at the NCI and Frederick.

00:15:31 To begin with, I'd like to present some structural concepts that underlie the design of a novel

00:15:38 class of pseudo-C2-symmetric diamino alcohols.

00:15:43 I'll then briefly compare the structures of a number of structure-based peptidomimetic

00:15:50 inhibitor complexes with HIV protease, and then I'll come back to discussing a second

00:15:57 series of C2-symmetric compounds, the diamino diols.

00:16:03 We'll look at some binding modes of some of these diol analogs, and finish by looking

00:16:09 at how we're trying to use structure to analyze the problem of drug resistance with HIV protease

00:16:16 inhibitors.

00:16:18 HIV protease replicates via the production of a number of polyproteins, which are encoded

00:16:26 by the viral genome.

00:16:29 These polyproteins have to be processed or cleaved at specific locations to produce the

00:16:34 structural proteins and replicative enzymes that are normally isolated from fully infectious

00:16:41 mature virus particles.

00:16:44 It was discovered some years ago that HIV protease, an enzyme encoded by the viral genome,

00:16:51 was responsible for this processing event.

00:16:54 With the knowledge then that HIV protease is an essential function of the viral life

00:16:58 cycle came the crystallization, in a lot of people's minds, of the idea that HIV protease

00:17:05 was a very good target for antiviral drug design.

00:17:08 The structure of HIV protease was first reported several years ago by scientists at Merck and

00:17:13 several months later by scientists at the NCI here in Frederick.

00:17:18 The structure of the enzyme, as shown here, is composed of two subunits, colored differently

00:17:24 for clarity.

00:17:26 Each subunit contributes an aspartic acid to the active site of the enzyme, which is

00:17:31 formed by the cleft that's shown in the middle of the molecule.

00:17:36 The active site is bordered on the bottom by the core of the molecule and on the top

00:17:40 by the presence of these two beta hairpin-like structures, often referred to as the flaps.

00:17:57 The way in which the subunits are related to each other is critical for our design strategy

00:18:05 and can be best seen by rotating the enzyme about 90 degrees.

00:18:10 You can see that the two subunits are related by a two-fold axis of symmetry and the consequence

00:18:16 of this symmetry is that the left half and the right half of the active sites are chemically

00:18:20 identical.

00:18:22 You can see this two-fold symmetry very clearly by rotating the top subunit about the two

00:18:26 fold axis, which exactly superimposes the top onto the bottom subunit.

00:18:32 When we first began thinking about how to design inhibitors of HIV protease, we didn't

00:18:38 know the structure of HIV protease, nor the structure of any other retroviral protease

00:18:43 for that matter.

00:18:44 However, several retrovirologists had noted over the years that there was a particular

00:18:49 sequence of an aspartic acid followed by a threonine or a serine followed by a glycine,

00:18:55 the Asp3 gly, which is the signature sequence of the active site of aspartyl proteases,

00:19:02 for which we knew quite a bit about.

00:19:04 In fact, because of the importance for hypertension, because of the key role that the enzyme renin

00:19:12 plays in hypertension, there had already been several drug discovery projects in a quite

00:19:18 mature stage that were involved in making inhibitors for renin and had developed what

00:19:24 has now become known as a classical substrate-based approach for peptidomimetic inhibitor development.

00:19:32 As you can see here, this approach is quite simply to replace the cisyl amide bond by

00:19:40 a number of non-cleavable transition state analog or isosteres.

00:19:47 For instance, the reduced amide or the hydroxyethylene isostere, phosphonates, hydroxyethylamines

00:19:54 and so on.

00:19:56 Virtually all of the replacements shown here have been tried for HIV protease inhibitors

00:20:01 as well as for renin inhibitors.

00:20:04 We wanted to try to get away from the classical design and try something a little bit more

00:20:09 daring and for this we relied on the concept of active site symmetry.

00:20:15 And the question that we asked was a simple one and that is, can we design a synthetically

00:20:20 feasible compound that would embody the same symmetry as the active site of HIV protease

00:20:28 and yet still be able to bind in a productive fashion?

00:20:31 This question can be rephrased as the following pair of constraints.

00:20:38 One is that the inhibitor should possess the same symmetry as the enzyme active site, in

00:20:43 this case it would be two-fold symmetry, and the other is that the symmetry operators of

00:20:48 the inhibitor and the enzyme should nearly superimpose in the complex when the inhibitor

00:20:54 is bound.

00:20:55 The design is shown here and consists of placing the two-fold axis through the central carbon

00:21:05 atom of the putative tetrahedral intermediate shown as this gem diol in this dipeptide transition

00:21:13 state intermediate.

00:21:15 We simply throw away everything on the right side of the two-fold axis, operate on the

00:21:19 left half by the two-fold axis to generate the diamino alcohol core structure and these

00:21:26 compounds could be synthesized and shown to be weakly inhibitory.

00:21:31 The real problem was how could we establish that these kinds of compounds would possibly

00:21:39 bind productively?

00:21:42 About the time that we began the synthesis of these compounds, the first structure of

00:21:47 a retroviral protease, that actually coming from Rous sarcoma virus, a chicken retrovirus

00:21:53 to be exact, was reported by Alex Glodower and his colleagues here at the NCI and Frederick.

00:21:59 When this structure became available to us, it allowed us to do a crucial modeling experiment.

00:22:06 The first thing it allowed us to do was to define the extent of structural homology between

00:22:12 the retroviral protease and other aspartyl proteases.

00:22:17 And secondly, it allowed us to do a docking experiment to examine the binding of a peptidomimetic

00:22:23 inhibitor into RSV protease and then to generate, using computer modeling, a two-fold symmetric

00:22:28 inhibitor.

00:22:29 The first part of this experiment is shown here.

00:22:34 In my lab at Abbott at the time, Thalia Bodzapatero and I had been studying the structure of porcine

00:22:40 pepsin, which is shown on the left.

00:22:44 The backbone structure of the Rous sarcoma virus protease is shown on the right.

00:22:49 Now the sequence homology between these two enzymes is very weak and yet you can see by

00:22:56 the regions that are highlighted that fully a third of the backbone atoms of the Rous

00:22:59 sarcoma virus protease can be superimposed onto the backbone structure of pepsin.

00:23:06 In pepsin, those backbone regions which are highlighted actually exhibit a pseudo two-fold

00:23:11 symmetry themselves.

00:23:13 Using this structural homology, we were then able to perform a docking experiment and for

00:23:19 this we used the structure of a peptidomimetic inhibitor that was originally designed for

00:23:25 renin as this substrate-based inhibitor containing the reduced phe dipeptide isostere.

00:23:36 Now this inhibitor fortuitously bound to rhizopus pepsin, an aspartyl protease derived from

00:23:42 a fungus.

00:23:43 And the structure of a complex of this inhibitor with rhizopus pepsin was solved in David Davies'

00:23:49 lab at the NIH.

00:23:51 We utilized the structural homology between aspartyl proteases and retroviral proteases

00:23:57 to dock the reduced peptide inhibitor into the active site and that's shown in the following

00:24:03 way.

00:24:04 We replaced the structure of pepsin on the left by the structure of the rhizopus pepsin

00:24:08 inhibitor complex and superimposed the two enzymes onto each other using the structural

00:24:14 homology of the active site regions.

00:24:18 Then we simply disappear the structure of rhizopus pepsin and we're left with the reduced

00:24:23 peptide docked into the structure of the active site of RSV protease.

00:24:27 Now for the second part of the experiment, we throw away everything C-terminal to the

00:24:33 central tetrahedral carbon atom of the reduced peptide and then operate using a two-fold

00:24:39 axis placed through the carbon atom on the N-terminal half of the molecule to generate

00:24:45 a two-fold symmetric inhibitor shown here that's docked into the active site of the

00:24:50 retroviral protease.

00:24:51 A cursor examination shows that the interactions were quite reasonable and in fact this experiment

00:24:58 demonstrated clearly to us that the idea of a symmetric inhibitor design should work.

00:25:04 I've listed some of the compounds that were synthesized in the diamino alcohol series

00:25:10 here and you can see that the first few compounds at the top were very weakly potent and in

00:25:17 fact we were at the level of the block compounds when we did our modeling experiment which

00:25:27 really gave us the courage to proceed further in this series rather than to fall back onto

00:25:33 the more classical structure-based approaches.

00:25:36 As you can see, we were able to get into the nanomolar regime of potency by putting

00:25:42 a CBZ block valine on either side of the diamino alcohol.

00:25:47 We next wanted to verify that in fact these compounds bound in a symmetric fashion that

00:25:54 we had initially designed.

00:25:56 To do this, Dave Neidhard in our lab grew a single crystal of a complex of HIV protease

00:26:02 with A74704 and the structure was solved by crystallographic methods and revealed a

00:26:10 couple of interesting features.

00:26:13 One of the most striking things that we noticed was that the structure of the enzyme had changed

00:26:17 remarkably in going from the unbound to the bound or inhibited form.

00:26:23 You can see that the flaps have now moved in by about 7 angstroms to clasp the inhibitor

00:26:29 very tightly in the active site.

00:26:32 This same sort of change had been noted earlier in the structure of a complex of HIV protease

00:26:39 with a reduced peptide structure-based inhibitor.

00:26:44 On the active site region of the complex, you can see a network of hydrogen bonds which

00:26:50 is very pronounced.

00:26:54 The top network of hydrogen bonds that are formed with the aspartic acid portion of the

00:26:59 active site is very similar to what was observed for renin inhibitors bound to aspartyl proteases.

00:27:06 This shouldn't come as a great surprise since we've already established that there was a

00:27:10 great deal of structural homology at the backbone level of these two sets of enzymes and in

00:27:15 fact the hydrogen bonds are primarily being formed between the backbone atoms.

00:27:19 On the other side of the inhibitor, you see a unique network of hydrogen bonds that are

00:27:26 made between the inhibitor and the flaps of the enzyme including a buried water molecule

00:27:32 which is tetrahedrally coordinated to two carbonyl oxygens of the inhibitor backbone

00:27:37 and two amide nitrogens on the flaps.

00:27:43 Now I'd like to compare very briefly the structure of the symmetric complex with structures of

00:27:51 other asymmetric structure-based peptidomimetic inhibitors.

00:27:57 There have been over a dozen crystal structures reported in the literature of HIV protease

00:28:03 inhibitor complexes, some of which are shown here, and they range from the reduced peptide

00:28:09 to a variety of hydroxyethylene-containing isosteres including the diol through hydroxyethylamine-containing

00:28:16 isosteres both in a cyclized and in an acyclic molecules as well as with statin-containing analogs.

00:28:24 It's instructive to compare the structures of some of these complexes and we can do this

00:28:29 in a couple of ways.

00:28:30 If we first look at how well the structures of the protein portions of the complexes superimpose,

00:28:37 you can see that they basically all agree quite well with each other to within about

00:28:42 half an angstrom root mean square deviation.

00:28:46 And this is pretty much within the error of refinement and these protein structures have

00:28:51 come from different space groups and been refined in different laboratories.

00:28:55 If we now look at how well the inhibitor portions superimpose after aligning the proteins onto

00:29:01 each other, you can see that for the same series or for inhibitors of the same class,

00:29:08 namely the hydroxyethylene-containing compounds, the backbones of the inhibitors all align

00:29:13 quite well onto each other in spite of the fact that they may contain quite different

00:29:18 side chain groups which emanate off of the backbones and occupied subsites of the enzyme.

00:29:23 Now what happens when you look at inhibitors of different classes?

00:29:26 There you see a quite different picture and for illustration I've only included three

00:29:31 different classes because the picture gets quite complicated.

00:29:34 Here is a symmetric inhibitor, the A7474 structure I showed earlier, superimposed onto

00:29:41 the conformations of a reduced peptide inhibitor and a hydroxyethylene-containing inhibitor.

00:29:47 And you can see that they all align quite nicely along the N-terminal halves of the

00:29:51 molecule on the left, but once they get to the position which is different,

00:29:56 the conformations then are all dramatically different.

00:30:00 We can draw several conclusions from this analysis.

00:30:03 One is that the protease enzyme wants to interact with the peptidomimetic inhibitors

00:30:10 in a single low-energy type of conformation and the conformational flexibility of the

00:30:19 peptidomimetics enables them to search a variety of conformations in order to try to optimize

00:30:26 or maximize the contacts with the enzyme active site.

00:30:34 One of the surprises to me in this analysis was not so much that A7474, a symmetric inhibitor,

00:30:41 can bind to the protease in a symmetric fashion, but that many of the asymmetric structure-based

00:30:48 inhibitors do so as well. In fact, regardless of class, we see the same very symmetric structure

00:30:54 of the enzyme.

00:30:57 One of the questions now that this raises is can we design a non-peptide inhibitor which

00:31:02 will make favorable interactions with the enzyme and will such an inhibitor induce the

00:31:08 same structural transitions that we see in the enzyme with peptidomimetic inhibitors

00:31:13 or will, in fact, we observe a different class of conformation, possibly one that will be

00:31:20 intermediate between the open and closed forms that we've seen thus far.

00:31:24 I'd like to turn now to a second class of symmetric inhibitors, so-called diaminodiols.

00:31:31 Buoyed by our success with the diaminoalcohols, we next designed a second class of inhibitors

00:31:38 in which we moved the C2 axis of rotation off from the central carbon atom to now bisect

00:31:46 the carbon-nitrogen sessile bond, and in this manner generated a series of compounds

00:31:52 which had an additional hydroxymethyl group inserted in the core resulting in the diaminodiol

00:31:59 series.

00:32:00 Now, the analog of A74704 in the diaminodiols, shown here as A75925, was about 10 to 20-fold

00:32:09 more potent than the alcohol and in addition showed a high degree of specificity for HIV

00:32:15 protease.

00:32:16 Unfortunately, this series of compounds had very poor solubility and could not really

00:32:23 be formulated as a drug candidate.

00:32:25 Now, how are we going to use structure to help solve the problem of pharmacology now

00:32:30 that we're already in the sub-nanomolar binding potency?

00:32:33 In this case, we use structure not so much to tell us what to do as to tell us what not

00:32:38 to do.

00:32:39 Now, examination of the surface of the active site region of the enzyme, which binds to

00:32:45 the inhibitor, shows that it's very continuous everywhere except at the ends of the molecule.

00:32:53 If we look at a 90-degree view, we're actually looking down the axis of the inhibitor through

00:32:59 what looks like a solid tunnel, which is about 30 angstroms long.

00:33:03 It's only open at the ends.

00:33:05 So what this told us was stay away from the middle of the inhibitors, the guts of the

00:33:10 inhibitor, and tinker with the ends.

00:33:13 And that's effectively what was done.

00:33:17 And this led to a new series of compounds, which had markedly improved pharmacologic

00:33:23 properties.

00:33:24 The key aspect of these compounds, which were designed by Dale Kemp and his colleagues,

00:33:32 was the replacement of the CBZ group at the end by a 2-pyridyl moiety and the replacement

00:33:39 of the carbamate linkage at the end with an N-methylurea type of linkage.

00:33:44 This led to a series of compounds with markedly improved solubility without sacrificing either

00:33:50 the potency for the protease or for the virus.

00:33:53 And, in fact, led to the development of A77003 as a clinical candidate.

00:34:01 Now, the DIAL series raised several interesting issues for us in the design of protease inhibitors.

00:34:08 First of all, all else being equal, replacement of the alcohol core by the DIAL core nearly

00:34:14 always resulted in a 10- to 50-fold increase in potency of these compounds.

00:34:19 The second interesting issue is one of stereochemistry.

00:34:23 There are three possible diastereomeric DIALs, the RR and the SS DIALs, each of which has

00:34:30 an infinite number of exact C2 symmetric conformations, and the RS DIAL, which is equivalent

00:34:37 to the SR in this case.

00:34:39 Now, if, in fact, the second hydroxy group in the DIAL is responsible for the increased

00:34:44 potency that we observe, then the stereochemistry should exert a profound effect on the potency

00:34:49 of these compounds.

00:34:50 In some cases, we do see effects of stereochemistry.

00:34:54 In others, we don't see an effect of stereochemistry.

00:34:58 And, in fact, the ordering of a particular series of diastereomers seems to depend both

00:35:04 in the qualitative as well as in the quantitative sense on the nature of the substituents that

00:35:09 are placed at the P2 and the P3 positions.

00:35:11 My group here at the NCI in Frederick is trying to look at the…is looking at the crystal

00:35:17 structures of a number of different DIAL and DIAL analogs with the aim of trying to understand

00:35:22 some of these activity relationships that have been observed.

00:35:25 As shown here are two different possible modes of binding of the DIALs with HIV protease.

00:35:33 With reference to the alcohol structure, you can see that here's the symmetric binding

00:35:41 mode that has been placed in the hydroxy group between the two aspartic acids of the active

00:35:47 site.

00:35:48 Now, if the DIAL were to bind in a symmetric fashion, the hydroxies would be displaced

00:35:53 from the two-fold axis of the enzyme, which would now pass roughly through the midpoint

00:35:58 of the DIAL bond.

00:36:00 And, in this case, the two hydroxies of the DIAL would be maximizing their interactions

00:36:04 with the aspartic acids.

00:36:06 On the other hand, if the binding of the central hydroxy group for the alcohol compound was

00:36:14 critical, then you might expect that only one of the two alcohols on the DIAL would

00:36:20 be making favorable interactions with the aspartates, and the other hydroxy group would

00:36:25 be displaced.

00:36:26 We first solve the structure of the DIAL A77003, our clinical candidate with HIV protease,

00:36:32 and this is the RS compound.

00:36:35 Examination of the active site structure of this complex indicated quite clearly that

00:36:41 the mode of binding of this compound is asymmetric.

00:36:45 A single hydroxy is interacting with the two aspartic acids.

00:36:48 The first hydroxy and the second hydroxy are hydroxy, and it's actually positioned

00:36:52 very similar to the hydroxy in the alcohol series.

00:36:55 The second hydroxy is making really no strong interactions with any aspect of the enzyme

00:37:02 structure, and that suggests that the second hydroxy group may actually be a liability

00:37:07 for binding rather than giving us an improvement in potency, which seems to pose us with a

00:37:13 bit of a dilemma.

00:37:15 And if you don't make compensating interactions for the loss of dehydration, if you don't

00:37:21 compensate for these interactions in the enzyme, then you might have to pay an energetic penalty.

00:37:27 Now, to test this hypothesis, we synthesized a deshydroxy form of the compound, which was

00:37:34 missing the second hydroxy group, but the rest of the compound was structurally identical.

00:37:41 The compound shown here, referred to as 78791, consists of a single hydroxy group, which

00:37:48 in this case would have the S stereochemistry.

00:37:51 We solved the structure of 78791 complex with HIV protease, and the best way to show the

00:37:59 results is to simply superimpose the inhibitor structure of the deshydroxy diol with the diol.

00:38:07 Both conformations were identical, the only difference being the presence or absence of

00:38:11 a hydroxy group in the secondary position.

00:38:15 This showed quite clearly that the presence of the second hydroxy group really plays no

00:38:20 role on the conformation of this inhibitor.

00:38:24 In fact, it seems to lead to a slightly weaker inhibitor.

00:38:29 In our laboratory here at the NCI, we are trying to simulate these conformational differences

00:38:35 that we observe in the diols, and in addition we're examining the crystal structures of

00:38:40 other diol analogs with the aim of trying to understand those critical interactions

00:38:45 which govern potency between HIV protease and its inhibitors.

00:38:51 An important question to ask at this point is, will HIV be able to mount resistance to

00:38:57 HIV protease inhibitors as well as it has been able to do to reverse transcriptase inhibitors?

00:39:04 Well, if you were the virus and you were presented with a compound like A77003 shown

00:39:11 here, bound in the active site of the enzyme, how would you try to defend yourself against

00:39:16 this drug?

00:39:18 One way would be to try to identify which regions of the inhibitor are interacting most

00:39:24 strongly with the enzyme and identify those regions on the enzyme that are involved and

00:39:31 change them.

00:39:33 Now it's very difficult for us to precisely evaluate the strength of interactions between

00:39:38 various portions of the inhibitor and the portions of the enzyme with which it interacts,

00:39:44 but there are some general considerations that govern the strength of interactions.

00:39:49 For instance, we know that the strength of non-bonded contacts generally diminishes very

00:39:57 rapidly as the distance between the atoms involved in the interaction increases.

00:40:05 One of the things we've been doing here at the NCI is to try to develop some procedures

00:40:10 with which to evaluate these interactions.

00:40:13 And working with T.J. O'Donnell, we have developed a procedure for visualizing the

00:40:21 closeness of approach of inhibitor with enzyme active site atoms.

00:40:27 The procedure involves calculating what we call difference surfaces.

00:40:32 By contouring these different surfaces at different levels, you get a feeling for how

00:40:37 close or far away different atoms in these contact points are.

00:40:43 For example, shown here is a different surface map contoured at a very high level, which

00:40:50 illustrates those atoms on the inhibitor and the enzyme which come into closest contact

00:40:55 with each other.

00:40:56 In the center of the molecule, the active site aspartic acids are involved in one such

00:41:02 contact with the hydroxy group of the diol.

00:41:05 Unfortunately for the enzyme, these aspartic acids are immutable, so that's not a place

00:41:10 where the enzyme can change.

00:41:13 Near the edge of the inhibitor, you can see another very extensive surface or patch between

00:41:20 the aromatic ring of the pyridine inhibitor and the arginine side chain of the enzyme.

00:41:30 And this appears to be a very interesting contact by virtue of the fact that it's not

00:41:35 only a close Van der Waals contact, but is an example of an induced dipole electrostatic

00:41:42 interaction between the positively charged guanidinium group and the pi electrons of

00:41:47 the pyridine ring.

00:41:49 There's yet a third patch near the top of the view, which is between the gamma methyl

00:41:59 groups of valine and the phenylalanine aromatic ring of the inhibitor in the P1 position.

00:42:08 The arginine interaction with pyridine is particularly interesting to us because there

00:42:12 are not that many examples of stacked arginine groups with aromatic rings.

00:42:17 If we examine the electron density map of the complex, you can see that the arginine

00:42:24 has very well-defined electron density associated with it.

00:42:27 This is indicative of a very well-ordered and relatively rigid side chain.

00:42:33 Arginine is very often found in a disordered configuration where you don't have that

00:42:37 strong electron density.

00:42:39 So this is experimental evidence, if you will, that this interaction is probably a fairly

00:42:43 strong one.

00:42:44 We've been collaborating with Dr. David Ho in New York, who has been growing HIV in the

00:42:50 presence of sublethal concentrations of A77003.

00:42:55 And he has been able to isolate several variants of the virus which show a decreased susceptibility

00:43:01 to the drug.

00:43:04 Sequencing analysis has shown that these variants often have a single base substitution, which

00:43:11 leads to a single amino acid change that maps to the region of the protease.

00:43:16 As shown here, two of these substitutions map to arginine 8 and to valine 82, precisely

00:43:23 those regions which we saw from the interaction map were predicted to make relatively strong

00:43:29 contacts between the inhibitor and the enzyme.

00:43:32 In the case of arginine 8, the replacement was to a glutamine, and in the case of position

00:43:37 82, the valine was replaced by an alanine.

00:43:40 We have also been collaborating with Dr. Edmund Chang and his colleagues at DuPont Merck,

00:43:45 who have isolated the same valine 82 to alanine replacement using a C2 symmetric inhibitor

00:43:55 that's very closely related to the A77003 compound.

00:44:00 How do these replacements affect the binding potency?

00:44:05 Well, we don't know for sure, but we have made attempts to model these changes into

00:44:10 the active site structure of 77003.

00:44:13 These effects can best be seen by examining CPK or space-filling models of the inhibitor

00:44:21 bound to the active site of the enzyme.

00:44:23 In these structures, the wild-type enzyme is shown on the left side of the image, and

00:44:30 a portion of the inhibitor is drawn in yellow.

00:44:33 You can see the pyridine ring and the phenylalanine rings, which are orthogonally stacked on the

00:44:38 left side of the molecule.

00:44:40 The atoms of the protein are color-coded by atom type, except that the gamma methyl groups

00:44:45 of valine 82 are shaded in gray for clarity.

00:44:50 Now, on the right side is the same molecule, but with the valine at position 82 mutated

00:44:56 to alanine, and the effect of this mutation is to create a hole where the gamma methyl

00:45:02 group was that interacted with the phenylalanine ring.

00:45:06 This hole would be an unfavorable situation, and quite possibly may be leading to some

00:45:13 of the reduced potency that this compound shows with this mutant.

00:45:17 Similarly, if we were to look at the arginine position, which is shown on the bottom left

00:45:22 hand portion of this slide, stacked against the pyridine group, you can see that if you

00:45:31 change that to a glutamine side chain, you would reduce substantially the extent of contact

00:45:38 between these two groups, and more importantly, you would also abrogate the strength of the

00:45:44 electrostatic-induced dipole interaction, since you've now gone from a positively charged

00:45:50 group to a neutral group.

00:45:52 So here's a case where a structure can be used to attempt to rationalize drug resistance.

00:45:58 The task that lies ahead is to design an inhibitor that will now bind to these drug-resistant

00:46:05 variants of HIV protease, or that will be finally tuned to interact with portions of

00:46:11 the protease which are functionally constrained against mutating.

00:46:16 It's important to remember that a mutation at the genetic level in HIV leads to a double

00:46:22 mutation in the protein level, and this is because the protein is a homodimer.

00:46:28 And this may be one of the few advantages that we have in our fight to battle HIV protease.

00:46:35 Well, I've tried to illustrate some of the ways in which we've used structural information

00:46:40 to design and analyze inhibitors to HIV protease.

00:46:45 In this particular case, I feel that we were very fortunate in that this approach led to

00:46:50 the design of an agent which actually has made it into clinical trials and is currently

00:46:55 being tested with AIDS patients.

00:46:59 I hope that I've been able to convey some of the excitement of this research area to

00:47:03 you, and I'd like to end with an analogy, if you'll permit me, that in order to design

00:47:09 compounds that exhibit the tight interactions with their target that you would like to see

00:47:15 requires an environment that promotes strong interactions between interdisciplinary scientists

00:47:21 such as synthetic chemists, structural biologists, and computational chemists.

00:47:27 I've been fortunate in my career to have been associated with some very talented individuals

00:47:31 both at Abbott and now here at the NCI, and I'd like to thank them all for their collaborative

00:47:37 efforts, and I'd like to thank you for listening.

00:47:49 Modeling enhances the understanding of protein structure and interaction.

00:47:54 A model of the solvent-accessible surface of beta-trypsin reveals its catalytic center.

00:48:02 Modeling the alpha-carbon backbone shows this protein's topology.

00:48:08 A slicing plane through the active site reveals various affinities between enzyme and two

00:48:14 para-amidino-phenylpyruvate.

00:48:17 Simulated annealing docking finds a low-energy interaction for this inhibitor.

00:48:22 Understanding protein structure and activity is one of the challenges that Dr. Art Olson

00:48:27 and his colleagues are studying using solutions from Digital Equipment Corporation.

00:48:33 Digital is here, providing the leading computing environment.

00:48:39 Unlocking the doors of discovery is Digital's Alpha AXP full 64-bit RISC solutions.

00:48:47 This new generation of cost-effective systems brings you and your lab into the 21st century

00:48:53 with exceptional performance.

00:48:57 Digital's Alpha AXP solutions have outstanding performance across the full range of computing

00:49:03 systems, and they expand the scope of your problem-solving capabilities.

00:49:10 Alpha AXP provides you with a choice of operating systems.

00:49:14 OSF-1, the standard for Unix, OpenVMS, or Microsoft's Windows NT.

00:49:22 Multiply Alpha AXP performance by networking workstations into farms, and you will have

00:49:27 a cost-effective entry into the super-fast world of parallel computing.

00:49:33 Ask for your favorite applications on Alpha AXP platforms.

00:49:39 Many are already taking advantage of Digital's leadership technology.

00:49:48 For more information on Digital's Alpha AXP products and services, call 1-800-DEC-INFO,

00:49:56 or from your computer, FTP the information from the Digital Gatekeeper on the Internet.

00:50:03 Look for information at the Gatekeeper to subscribe to DEC News, our electronic newsletter

00:50:09 for education and science customers.

00:50:12 If you prefer, you can have the information faxed to you.

00:50:18 Digital's Alpha AXP technology.

00:50:22 The tool to bring molecular sciences research into the 21st century, now.

00:50:29 In a few minutes, we'll be ready to start taking your telephone calls.

00:50:49 At this time, you should see the telephone numbers displayed on the screen.

00:50:53 The numbers are 800-368-5781 and 5782, and for our international callers, 202-463-3170.

00:51:04 You may also fax in your questions to 202-887-3457.

00:51:11 You may begin calling in now, and faxing, for that matter.

00:51:14 While we're getting ready to start the general discussion, Dr. Gerard Folkers,

00:51:18 who is professor in the Department of Pharmacy at the ETH in Zurich, Switzerland,

00:51:22 and his colleague, Dr. Didier Rognon, are waiting on the line to report on their work

00:51:27 on the prediction of an influenza virus-derived nonapeptide-MHC interaction

00:51:32 that was confirmed by crystal structure.

00:51:34 Dr. Folkers and Dr. Rognon, welcome to the program.

00:51:37 Hello?

00:51:38 Yes, go ahead, Dr. Folkers.

00:51:40 Good afternoon, ladies and gentlemen.

00:51:42 First of all, I'd like to thank the organizing committee for inviting us to contribute

00:51:47 to this exciting seminar, a special Jonathan Greer.

00:51:51 And I will talk about cellular immunity.

00:51:54 Cellular immunity is mediated by unique ternary complexes made up of T lymphocytes,

00:52:01 antigenic peptides, and major histocompatibility complex, MHC-encoded proteins.

00:52:09 The function of MHC proteins is to bind to intercellular antigenic peptides

00:52:15 and present them at the surface of infected cells to a T cell receptor.

00:52:21 When we started this study, the crystal structures of the three homologous human

00:52:26 class I MHC proteins were known.

00:52:29 However, previous resolutions were unable to describe the exact conformation

00:52:34 of the co-crystallized bound antigens.

00:52:37 In order to circumvent the relative inaccuracy of the experimental methods,

00:52:43 we used molecular dynamics to simulate single complexes between the X-rayed HLA-A2 MHC protein

00:52:51 and one antigenic non-peptide from the influenza virus matrix protein.

00:52:58 After 40 picoseconds, the molecular dynamic simulation was equilibrated.

00:53:03 RMS positional deviations of the C-alpha and backbone atoms from their crystal

00:53:10 coordinates reach a plateau at 0.9 and 1.9 angstrom, respectively.

00:53:17 The HLA-A2 crystal structure in cyan could be accurately reproduced with two advantages.

00:53:26 The explicit inclusion of a viral peptide not seen in the X-ray studies

00:53:32 and the presence of a water environment.

00:53:36 When RMS atomic fluctuations are compared,

00:53:39 the properties of the experimental and simulated structures agree.

00:53:43 This implies that docking an antigenic peptide in the MHC binding pocket

00:53:49 has not altered the general topology and motional behavior of the protein.

00:53:54 The viral peptide adopts an extended conformation

00:53:58 and occupies the whole length of the protein's binding groove.

00:54:02 The antigen HLA interactions have been identified for time-average structures

00:54:09 and are divided into conserved electrostatic interaction

00:54:13 and specific hydrophobic contacts.

00:54:17 Three peptide side chains at positions 2, 3, and 9

00:54:23 are in close contact with three hydrophobic pockets, B, D, and F

00:54:31 of the protein made up of hypervariable residues.

00:54:37 The polar atoms of the peptide backbone are bound to protein amino acids

00:54:43 that are conserved for all MHC proteins.

00:54:47 The C-terminal end of the peptides are much more tightly anchored to the protein

00:54:54 than the N-terminal ends that developed water-mediated hydrogen bonds to the MHC.

00:55:00 The water inchalation effect was notably mediated

00:55:06 by two of the three water molecules that were treated in the simulation.

00:55:11 Twelve hydrogen bonds between HLA-A2 and its viral peptide could be described.

00:55:18 None of them involved the peptide side chain.

00:55:23 When a second viral peptide, HIV-Li1 reverse transcriptase,

00:55:29 was docked in the HLA-A2 binding groove and simulated using the same conditions,

00:55:36 similar observations were made suggesting that antigenic peptides

00:55:40 may bind to a defined MHC protein with similar bond conformations.

00:55:47 The hypothetical model for HLA-A2 was confirmed

00:55:51 by the recent crystal structure of allelic proteins HLA-B27 and HLA-AW68

00:55:58 for which the peptide's bond conformation could be resolved.

00:56:02 Ninety percent of the MHC peptide hydrogen bonds enclosed by boxes

00:56:09 effectively occurred in the novel crystal structures.

00:56:13 As predicted by our molecular dynamics study,

00:56:16 some water residues play an important role by binding the interaction,

00:56:22 by bridging the interaction of the peptide's N-terminus

00:56:26 with its MHC complementary binding region.

00:56:30 The middle part of the peptide sequence from residues 4 to 7

00:56:35 in both crystal structures does not participate in peptide MHC interaction.

00:56:42 The result is that the peptide MHC hydrogen bonds

00:56:46 as previously suggested by our molecular dynamics study.

00:56:50 A similar conserved pattern is found for the interaction of the peptide side chains

00:56:54 with the MHC polymorphic proteins.

00:56:57 Pocket B, D, and F are in close contact with the peptide side chains 2, 3, and 9, respectively.

00:57:07 The only difference in the three structures lies in the nature of the interactions,

00:57:12 which in most cases are hydrophobic

00:57:15 and dependent on the polarity of the involved MHC pockets.

00:57:21 This interactive model explains why so few MHC proteins

00:57:26 can recognize the huge amount of potential antigens.

00:57:30 The antigen backbone interacts with the MHC protein amino acid

00:57:35 so that the global recognition is roughly similar for all the peptides.

00:57:41 The length of the antigenic sequence is dictated by the geometry of the protein's binding site.

00:57:48 Since it remains constant for all MHC proteins,

00:57:52 only octa- or mona-peptides are preferentially recognized.

00:57:57 The specificity of the immune recognition is given by the location

00:58:01 and the nature of pockets that can accommodate a great variety of side chains

00:58:06 present at different positions of the peptide sequence

00:58:09 so that many antigenic structures can be recognized by fewer MHC proteins.

00:58:16 The agreement between experimental and theoretical structures has been made possible in part

00:58:21 because of the highly homologous natures of the class I MHC proteins.

00:58:26 Molecular dynamics seems to be an accurate method

00:58:29 for predicting the three-dimensional structure of MHC peptides' antigenic complexes.

00:58:35 The knowledge of the molecular interactions between MHC proteins and viral peptides

00:58:41 provides an invaluable help for modifying the sequence of natural peptides

00:58:46 in order to design new synthetic molecules that could successfully compete with viral antigens.

00:58:53 This approach is currently used in our group for designing new antigens

00:58:57 that could be used in vaccine preparations

00:59:00 for the treatment of viral infections, graft reactions, and autoimmune diseases.

00:59:06 Thank you for your attention.

00:59:11 Thank you, Dr. Volkers and Dr. Ronjan, who is there also.

00:59:14 We haven't heard from him at the moment.

00:59:16 We can begin the general discussion period now.

00:59:18 Again, the telephone numbers are 800-368-5781 and 5782.

00:59:23 And for international callers and those in Washington, D.C.

00:59:26 because this is an international city, isn't it?

00:59:28 202-463-3170.

00:59:31 And when you call, a volunteer will answer and will ask you for your site and location.

00:59:36 And you will be put on hold, and you will hear the program audio on the telephone.

00:59:39 And I will call on you by the name of your location.

00:59:42 So if I say Des Moines, Iowa, that means it's time for you to talk.

00:59:44 When you hear your location called, talk right into your telephone handset

00:59:47 and tell us your name, who your question is for,

00:59:50 if it can be directed to a specific person, and, of course, your question.

00:59:54 We have several telephone lines available now.

00:59:56 But if you should get a busy signal, please hang up and try again.

01:00:00 And our first call is from where?

01:00:02 Montana State.

01:00:04 Let's hear our first call from Montana State.

01:00:06 Go ahead, please.

01:00:07 Hello, my name is Edward Dratz, and I have a question primarily for Dr. Erickson.

01:00:13 And I was interested in if you considered the dynamics or disorder of the inhibitor

01:00:23 in the binding site of the HIV protease, as you'd observe by X-ray thermal ellipsoids.

01:00:29 And I wonder if you need flexibility in parts of good inhibitors,

01:00:36 or would larger ellipsoids show where you could fit in larger substituents?

01:00:43 That's an interesting question.

01:00:49 In the crystal structures that I've described,

01:00:52 we really don't see any great degree of conformational disorder of these inhibitors as they're bound.

01:01:02 The resolution of crystal structures with proteins typically isn't high enough

01:01:08 to be able to model the thermal ellipsoid as you can do with small molecule crystallography.

01:01:15 However, the question that you asked regarding the importance of flexibility of the inhibitor

01:01:21 or portions of it for binding to the HIV protease, I think, is a very profound one

01:01:27 and one that we don't fully understand the answer to.

01:01:30 It's clear that the kinds of inhibitors that I've described

01:01:33 and that many groups are working on the peptidomimetics

01:01:36 are very flexible compounds intrinsically in solution.

01:01:40 And one of the key factors probably governing their binding affinities

01:01:46 are the entropy changes that you can obtain upon binding these flexible inhibitors

01:01:53 to the active site of the enzyme.

01:01:58 Our next question is from Syracuse, New York State University.

01:02:02 There must be under 400 feet of snow over there.

01:02:04 They got 43 inches of snow in Syracuse yesterday.

01:02:07 Are you still able to talk?

01:02:08 Have you been snowed in?

01:02:09 Go ahead, Syracuse.

01:02:11 No, we can still talk.

01:02:13 We got about 450 feet of snow.

01:02:16 This is Darrell Dykes from the SUNY Health Science Center.

01:02:19 My question is for Dr. Erickson.

01:02:21 Several groups are now using computer-based molecular modeling

01:02:25 and dynamics techniques to characterize the functionally important

01:02:28 structural features of these peptides,

01:02:31 and the HIV proteases that you discussed seem to be prime candidates

01:02:35 for this kind of characterization.

01:02:37 Have you considered anything along this line?

01:02:42 Well, I'm not sure I understood the question fully,

01:02:45 but we are actively pursuing the design of HIV protease inhibitors

01:02:52 through the use of X-ray crystallography.

01:02:55 I think HIV protease affords the field of computational modeling

01:03:02 and structure-based design a unique opportunity because of the fact that

01:03:05 there are so many groups working on this area,

01:03:08 and there are more structures coming out in HIV protease complexes right now

01:03:14 than I think we've ever seen in the field of crystallography.

01:03:18 So it's making HIV protease a terrific lab, in a sense,

01:03:21 to study various approaches to structure-based design.

01:03:27 Okay. Next we go to East Hanover, New Jersey, for our next questioner.

01:03:31 Go ahead, please.

01:03:33 My name is Herb Schuster, and I'd like to know from Dr. Erickson,

01:03:37 if in drug-resistant mutations, which provide a means of survival to the virus,

01:03:43 how does the natural substrate bind to the new variants?

01:03:49 Well, that's a question that is obviously very important.

01:03:54 We and other groups have been addressing this by, first of all,

01:03:59 trying to produce the mutated form of the enzyme.

01:04:04 These mutations that I described have been characterized from isolates of virus

01:04:11 that's been grown in the presence of the drug.

01:04:14 And the proteases that have these mutations have to be cloned, expressed,

01:04:20 and then characterized biochemically.

01:04:22 We have been able to do that, in one case, with the ARGTGLN mute that I described,

01:04:28 and we have also been provided several other mutations,

01:04:31 several other mutant proteases from other groups.

01:04:35 In the two cases that we have looked at, the ARGTGLN and the VALE-ALA82,

01:04:43 with at least one of the normal substrates that we use, and it's a fluorogenic substrate,

01:04:48 we haven't noticed any major changes in the KM for that particular substrate.

01:04:54 Growth curves have been done with some of the virus isolates that are resistant,

01:05:02 and in the cases that I've seen, the growth curves are very similar

01:05:08 between the mutant virus and the wild-type virus.

01:05:11 So those particular examples seem to show no deleterious effects on the virus or on the enzyme itself.

01:05:22 All right, that's it? Okay.

01:05:24 Next we go to Burlington, Massachusetts in the Pharmaco Labs.

01:05:27 Go ahead, please.

01:05:28 Hi, this is Mel Belot calling.

01:05:32 I have a question about the affinity for binding,

01:05:38 and how does that correlate with the ability to induce drug resistance?

01:05:48 Well, we don't know the answer to that either,

01:05:52 because I think that we just don't have enough data points on that issue.

01:06:00 I suspect that the tighter that your inhibitor binds to the enzyme,

01:06:04 the easier it is for the enzyme to make a mutant that will allow you to detect a change in terms of the strength of affinity.

01:06:14 But as far as the ability to mount resistance to a particular drug,

01:06:20 I think this is something that we don't know the answer to for the HIV protease inhibitors thus far.

01:06:26 But it's obviously a very important issue that I'm sure will be explored in the future

01:06:31 as more groups get involved in testing their inhibitors for the ability to develop resistance.

01:06:37 All right, next we go to Maryland, England, to Middlesex, England, to the Glaxo Group Research Company there.

01:06:43 Go ahead with your question, please.

01:06:45 Hello, I have a question for John Erickson.

01:06:48 This is Mike Hamm here at the Glaxo site.

01:06:51 It's a question on behalf of John Saunders,

01:06:53 and the question is, with all the detailed structural information now available for HIV protease,

01:07:00 why has it not been possible to design low molecular weight inhibitors, by which I mean about less than 400?

01:07:07 And should we assume this will never be possible for the aspartyl protease family of enzymes in general?

01:07:14 Mike, that's a good question.

01:07:18 I guess I probably shouldn't say that it's not been possible to design inhibitors of low molecular weight

01:07:27 because I don't want to suggest that there may be groups out there who have done so

01:07:34 but haven't come forward yet with their results.

01:07:38 Certainly, I think the group at SmithKline has sort of delineated in a careful SAR study,

01:07:46 for peptidomimetics at any rate, that the minimal size inhibitor is something that should span the P2 to P2 prime sites.

01:07:57 And that it would seem then that to maximize your binding affinity,

01:08:02 if you want to extrapolate from that study, that you need to fill up those same sites with a non-peptidic inhibitor.

01:08:15 I think that time will tell if we can really improve the affinity of binding enough,

01:08:22 or our predictive ability to improve affinities enough to be able to really apply this to very small molecules

01:08:29 to bind to something like HIV protease, which has a very wide active site.

01:08:35 Okay, John. Let me remind all of our viewers that we still have on the line doctors of Fokler's and Rognon

01:08:43 who are willing to answer any questions you may have of them if you so desire.

01:08:47 Just a reminder that they are still listening in.

01:08:50 Our next site is the University of Missouri at St. Louis. Go ahead, please.

01:08:54 Hello, I'm Fritz with the University of Missouri near the Gateway Arch in Mississippi River.

01:08:59 My question about the inhibitor is, well, on the pyridine there are three possible positions for substitution,

01:09:08 each of which would give a somewhat different pi-density map.

01:09:12 Have these other positions been exploited and what are the results in terms of interaction with the protease?

01:09:21 Well, we've actually have looked at that.

01:09:23 And we've also looked at, in fact, the first compound I described from the Monowall series had a benzene ring there.

01:09:32 And we haven't looked at the structures of the other pyridines,

01:09:38 but we have looked at obviously the CBZ group and the 2-pyrido group.

01:09:42 And they both, we've observed and others have observed, make similar interactions with that arginine side chain.

01:09:49 The position of the nitrogen on the ring in this particular series of compounds doesn't seem to influence the binding affinity all that much.

01:10:01 All right. We go abroad again to Uppsala, Sweden and the Uppsala Biomedical Center. Go ahead, please.

01:10:09 Yes.

01:10:10 My name is Professor Bror Strandberg from Institute of Molecular Biology in Sweden.

01:10:16 I wonder, Jan Eriksson, has any drug been designed against HIV protease,

01:10:25 which works by breaking the contact between the dimers of the enzyme?

01:10:32 Yes. Hello, Bror. It's a pleasure to hear from you again.

01:10:37 There have been several reports in the literature of peptides that have been designed to mimic either the N or C termini of the HIV protease monomer.

01:10:53 And these termini are involved in an anti-parallel beta sheet, two-fold beta sheet, that helps to hold the dimer together.

01:11:04 And these groups have had limited success in the sense that they have been able to demonstrate that they do,

01:11:12 that these peptides do have, inhibit the function of HIV protease and seem to do so by a dissociative mechanism.

01:11:22 The affinities of these peptides is, you know, greater than, generally greater than 10 micromolar or so.

01:11:31 So whether this will be a strategy that will be amenable to, useful for the design of new novel kinds of inhibitors for HIV protease has yet to be shown.

01:11:44 Because in the cases that I know about, these have all been straight peptides.

01:11:49 But it is an interesting approach.

01:11:51 All right, we have a faxed question for Dr. Volkers in Zurich.

01:11:57 And it is from Richard Lewis, who is with Glaxo in Greenford, United Kingdom.

01:12:02 It says, how does MHC peptide binding compare with Src peptide binding motifs?

01:12:11 Could you repeat the question?

01:12:12 Certainly.

01:12:13 Compared to what?

01:12:15 How does MHC peptide binding compare with Src peptide binding motifs?

01:12:27 I do not know.

01:12:29 I cannot.

01:12:30 Okay.

01:12:32 Anybody know here?

01:12:33 No?

01:12:34 Okay.

01:12:37 You don't get a trip to America now.

01:12:39 That was our prize.

01:12:41 We go now to Columbus, Ohio, and the Chemistry Abstracts Service there.

01:12:45 Go ahead, please.

01:12:46 Chemical Abstracts Service.

01:12:47 Yeah, this is Don Widiak at The Ohio State University.

01:12:51 We're meeting in the Chemical Abstracts building, and we have a number of students here.

01:12:56 I'd like to ask the panel and or those in Europe to give some thought to the following.

01:13:05 Molecular modeling is really excellent and very interesting, but most of the work requires a purified protein.

01:13:15 I wonder what the relative percentage of activity in the drug industry is relative to drug design using this mode,

01:13:25 and what the projection will be for 10 years from now.

01:13:29 Okay.

01:13:30 Who would like to take that?

01:13:31 Gemma?

01:13:32 Let's not fight over it now.

01:13:37 Sure, I'll start.

01:13:38 Go ahead.

01:13:39 Jack Baldwin.

01:13:40 I think that the method now, as you would anticipate, is relatively limited for the very reasons that you anticipated.

01:13:49 Many of the problems that are being addressed today are receptor-based problems, channel-based problems,

01:13:56 but I think the whole concept of using protein macromolecular ligand complexes will gain influence

01:14:04 and targets will be selected and more targets will become available.

01:14:08 I'll turn it over to Jonathan for his.

01:14:11 I agree entirely.

01:14:14 We certainly are working to select targets where we can use these types of strategies that we're talking about today.

01:14:22 Certainly in the past, as Dr. Baldwin has said, we've been working with receptor structures that are membrane-bound.

01:14:29 This has been a problem for us, but even there, using comparative homology modeling methods,

01:14:34 we're beginning to develop models to treat those problems as well.

01:14:39 It's also true that one can deal with these problems just by looking at the ligands,

01:14:45 and so I think the methodology is going to expand more and more,

01:14:49 and the success stories of the type that you're hearing today are going to fuel that.

01:14:53 Any other comments before we move on?

01:14:55 All righty.

01:14:56 We go now to Houston, Texas and the Baylor College of Medicine.

01:14:59 Go ahead, please.

01:15:01 Hi, this is Zhang Yichao from Baylor College of Medicine.

01:15:05 The question is for Dr. Erickson.

01:15:08 There are a few aspartic acid proteins in the human body, which shows similarity with the HIV proteinases.

01:15:17 I'm wondering whether you ever tested the binding affinities of those inhibitors to those aspartic acid proteinases.

01:15:25 Yes, that's a good question, and we routinely tested the protease inhibitors for the ability to inhibit some of the important

01:15:41 human aspartic acid proteinases, such as renin, and they were also tested against pepsin and cathepsin D.

01:15:52 Okay.

01:15:55 Okay, we're going to England again, Middlesex, England, to GlaxoResearch.

01:15:58 Go ahead, please.

01:16:00 Hello, it's Mike Hannigan here.

01:16:02 Another question for John Erickson.

01:16:06 The question is on behalf of Kevin Parks from Roche, who are with us here as well.

01:16:11 The question is, do you know any cases of C2 symmetric inhibitors derived from the C-terminus, which show good activity?

01:16:20 If not, have you any explanation for this lack of activity?

01:16:25 I believe Merck has published last year on a C2 or a pseudo-C2 symmetric inhibitor that was derived from the C-terminus.

01:16:39 It was using a, you know, a non-standard amino acid in the P2 and P2 prime positions, I believe.

01:16:47 I think that's correct.

01:16:48 Yeah, but that was quite potent.

01:16:50 That was a quite potent compound, sub-nanomolar.

01:16:54 All right, that's it.

01:16:55 Any follow-up on that?

01:16:56 No.

01:16:57 Okay.

01:16:58 We go to a faxed question.

01:17:01 This is from Greenford, UK, and this is for you, John, again, from Tadeusz Skarzinski.

01:17:07 How about that?

01:17:08 Pretty good.

01:17:09 Now, the asymmetric compound A78791 binds tighter than the symmetric A77003.

01:17:20 Both compounds bind in an asymmetric fashion.

01:17:24 Does not this mean that the protein, although symmetric, evolved to bind asymmetric ligands?

01:17:32 Here's the question you can look at.

01:17:33 Yes, I heard the question.

01:17:34 I was thinking of it.

01:17:35 Well, I think that we can infer maybe that the protein evolved to bind asymmetric ligands

01:17:44 simply from the fact that it does bind and cleave normal peptide or protein substrates.

01:17:54 If you actually look at the confirmation of either A77003 or A78791 themselves outside of the active site,

01:18:04 you find that outside of the actual core region of the diol,

01:18:09 the two halves of the molecule have almost the identical confirmations if you look at the torsion angles,

01:18:15 and they're related by a two-fold axis as well.

01:18:19 In the case of these two compounds, the two-fold axis of the inhibitor

01:18:22 and the two-fold axis of the enzyme are shifted by about half a bond,

01:18:27 which gives one the impression that they're binding very asymmetrically.

01:18:31 In fact, the asymmetry of binding that I mentioned is really primarily confined to that region about the core.

01:18:37 We have looked at the structures of a couple of the other diastereomers of these diols,

01:18:43 the RR and the SS, which I didn't have time to discuss.

01:18:46 In the case of the SS in particular, the binding is extremely symmetric.

01:18:52 In fact, you can't really distinguish the two halves of the inhibitors in the active site.

01:18:57 The two-fold axis of the inhibitor and the two-fold axis of the enzyme nearly superimpose.

01:19:02 I think that's a consequence of the fact that the active site structure of the enzyme is reasonably not exact,

01:19:11 but reasonably symmetric.

01:19:13 All right.

01:19:14 And we go again to Farm Ecolabs in Burlington, Massachusetts.

01:19:18 Go ahead, please.

01:19:19 This is Dr. Martha Teeter from Boston College, calling from Burlington.

01:19:25 I had a question about the parts of the last case and the chicken-roused sarcoma molecule that were similar.

01:19:36 In homology modeling, it's important to know how many sequences are consistent with one particular structure.

01:19:45 And so in order to evaluate that, what was the RMS deviation between the structurally homologous regions you defined in those two proteins?

01:19:56 And also, what is the percent homology between those two regions?

01:20:03 Well, now I'm going to go back into past history.

01:20:06 I think, Martha, that the RMS deviations between those two regions that I had highlighted in white was about one and a half angstrom RMS,

01:20:16 between one and one and a half angstrom RMS.

01:20:18 I think it's safe to say that.

01:20:22 The actual level of sequence homology was not any higher than what you'd expect based upon chance alone.

01:20:32 In fact, the sequence homology between retroviral proteases and the eukaryotic enzymes is buried in the noise.

01:20:42 The only reason that people thought that retroviral proteases may actually be aspartyl proteases

01:20:48 was because they saw that signature sequence, Asp3Gly, which by itself wouldn't give you the statistics to think that these two enzymes were homologous.

01:20:58 But because that's a signature sequence of aspartyl proteases, that made people think about this.

01:21:03 So there's a case where the actual sequence homology was extremely limited.

01:21:10 Even going from that, Pearl and Taylor, before the structure of RSV protease was solved,

01:21:16 came up with a model for aspartyl proteases, which was, I guess, within about two angstroms RMS overall for much of the core portion of the retroviral proteases.

01:21:26 So they were able, by homology modeling, even with that limited sequence homology, to do a fairly accurate job at it.

01:21:35 We have to conclude the discussion for now. Thank you, John Erickson, for your answers.

01:21:39 And we would also like to thank Drs. Folkers and Ronjan, who are leaving us at this time, but will be back later on in the program.

01:21:46 We still have one more question and answer segment at the end of the program, so please hold your questions until then.

01:21:51 We still have a couple of fax questions here as well.

01:21:53 It's time now for a stretch break.

01:21:56 Before the break, we want to especially recognize two groups, the local site coordinators at each of the participating locations,

01:22:02 and Digital Equipment Corporation, who have supported the participation of 17 ACS local sections.

01:22:09 These programs would not succeed without you.

01:22:12 Whether you are at a company, a university, or an ACS local section, we know that you have gone out of your way to make today's program a success for your group.

01:22:20 Please join me in applauding your local site coordinators for arranging for your group to receive this program all around the country, and the world for that matter.

01:22:28 We'll be back in about five minutes with more good information. Thank you.

01:22:50 ♪♪

01:23:00 ♪♪

01:23:10 ♪♪

01:23:20 ♪♪

01:23:30 ♪♪

01:23:40 ♪♪

01:23:50 ♪♪

01:24:00 ♪♪

01:24:10 ♪♪

01:24:20 ♪♪

01:24:30 ♪♪

01:24:40 ♪♪

01:24:50 ♪♪

01:25:00 ♪♪

01:25:10 ♪♪

01:25:20 ♪♪

01:25:30 ♪♪

01:25:40 ♪♪

01:25:50 ♪♪

01:26:00 ♪♪

01:26:10 ♪♪

01:26:20 ♪♪

01:26:30 ♪♪

01:26:40 ♪♪

01:26:50 ♪♪

01:27:00 ♪♪

01:27:10 ♪♪

01:27:20 ♪♪

01:27:30 ♪♪

01:27:40 Welcome back. A short but sweet break.

01:27:42 We've got a lot of material to go yet.

01:27:44 We're ready to begin the second part of the program.

01:27:46 And in this portion of the program,

01:27:48 we'll hear presentations from John Baldwin and Michael Varney

01:27:51 and wrap up with a brief report from Peter Warner

01:27:54 from the final discussion period

01:27:56 before we go back to your questions again.

01:27:58 Here we are.

01:28:00 The next speaker is Dr. John J. Baldwin.

01:28:03 He is Distinguished Senior Scientist

01:28:06 at Merck Research Laboratories in West Point, Pennsylvania.

01:28:10 He has led a wide range of research efforts while at Merck,

01:28:14 publishing extensively in the drug design of cardiovascular,

01:28:18 ophthalmic, gastrointestinal, and CNS agents.

01:28:22 Dr. Baldwin's topic today is X-ray crystallography and drug design.

01:28:29 For today's symposium on computer-assisted drug design,

01:28:33 I would like to focus on the combination

01:28:35 between macromolecular X-ray crystallography

01:28:38 and computer-assisted modeling.

01:28:41 These are proven to be powerful tools for the medicinal chemist,

01:28:45 for he can view the ligand-macromolecule interaction

01:28:49 at the molecular level.

01:28:51 Now, at this level, one can then begin to rationalize

01:28:54 the differences in relative affinities found experimentally

01:28:58 and base these differences on changes in the structure of the ligand.

01:29:03 Now, these changes can often be subtle,

01:29:05 changes in chirality, changes in substitution patterns.

01:29:09 Now, by understanding this difference in relative affinity,

01:29:13 one can be led into optimization strategies,

01:29:16 and these are the first steps towards rational drug design.

01:29:20 Now, the particular model that we have used

01:29:23 to study this small-molecule-macromolecule interaction

01:29:27 is the enzyme carbonic anhydrase.

01:29:30 Now, inhibitors of carbonic anhydrase

01:29:33 have the potential to be developed into useful therapy

01:29:37 for the treatment of the ocular disease glaucoma.

01:29:40 Now, carbonic anhydrase really defines a whole family of enzymes,

01:29:44 all of which are capable of converting CO2 to bicarbonate.

01:29:48 There are zinc metalloenzymes

01:29:50 in the molecular weight range of 30,000 to 60,000.

01:29:53 Above the isozymes, carbonic anhydrase 2

01:29:56 is the example that's found within these secretory cells

01:29:59 and is involved in the active secretion of aqueous humor,

01:30:03 that fluid which fills the anterior and posterior chamber of the eye.

01:30:07 Recently, evidence has been accumulating

01:30:09 that a membrane-bound isozyme, carbonic anhydrase 4,

01:30:12 may also be involved in this secretory process.

01:30:15 Both of these isozymes are high-activity enzymes.

01:30:18 They're among the most efficient enzymes in the body

01:30:21 for centrally diffusion control.

01:30:23 Of the two, carbonic anhydrase 2 is the best characterized,

01:30:27 and that will be the model that we use today

01:30:30 as we view our inhibitors within its active site.

01:30:33 Now, human carbonic anhydrase 2

01:30:35 was isolated in 1934 by Meldrin.

01:30:39 It's a globular enzyme

01:30:42 containing about 260 amino acid residues

01:30:45 with one zinc per molecule.

01:30:47 The active site cavity itself

01:30:49 is composed of twisted beta sheets

01:30:52 that run through the very center of the molecule.

01:30:54 There are helical regions,

01:30:56 and these are located towards the surface.

01:30:58 In the next few slides,

01:31:00 we'll view carbonic anhydrase 2,

01:31:03 whose structure was determined by X-ray crystallography

01:31:06 at below the 2 Angstrom level of resolution.

01:31:10 Now, the first representation

01:31:12 will be the ribbon representation.

01:31:14 We're looking down into the active site cavity.

01:31:17 The sphere in the center is the zinc atom.

01:31:20 Notice how the active site cavity

01:31:22 is composed of these twisted beta sheets

01:31:25 with a few helical regions on the surface of the enzyme.

01:31:28 We're going to focus more closely

01:31:31 on the active site cavity itself.

01:31:35 Here in the center again is the zinc atom.

01:31:38 It's coordinated to three histidine residues.

01:31:41 The fourth coordinate to zinc is hydroxyl ion,

01:31:44 and this then becomes the attacking species

01:31:47 on CO2 to give bicarbonate.

01:31:55 Thank you.