00:00:00 The active site cavity itself is divided into two hemispheres.

00:00:25 The upper portion of the cavity, as we'll be viewing it today, is composed mainly of

00:00:30 hydrophilic amino acids, and their residues are shown in red.

00:00:35 Now the bottom of the active site cavity is mainly lipophilic residues, and their side

00:00:41 chains are shown in the blue coloration.

00:00:43 Here we can note a few of the amino acid residues that are going to be important in today's

00:00:47 discussion.

00:00:49 To the right of the zinc is histidine-64.

00:00:54 Slightly below the zinc to the left is phenylalanine-131.

00:00:58 And then situated between those two histidine residues is glutamine-92.

00:01:05 The next view of carbonic anhydrase will be in the Connelly surface format.

00:01:12 Here we see the surface of the active site cavity itself.

00:01:15 It's about 15 angstroms across and descends about 15 angstroms to the zinc atom.

00:01:22 It becomes very narrow at the zinc, it's a cone-shaped cavity.

00:01:26 Again we can note those key residues.

00:01:30 Histidine-64 appears at about 2 o'clock.

00:01:34 The large bulge at about 7 o'clock is phenylalanine-131 with glutamine-92 at about 11 o'clock.

00:01:42 The particular class of inhibitors that we're going to be focusing on today are these heteroaryl

00:01:48 sulfonamides.

00:01:50 Now the sulfonamide group presumably binds in its deprotonated form to the zinc atom

00:01:57 where it's replacing the hydroxyl ion as the fourth ligand.

00:02:01 The heteroaryl group will be this thionothiopyrin that has an alkyl amino group in the 4th position.

00:02:09 And we're going to explore what effect that chirality might have on the ability of these

00:02:14 compounds to interact within the active site cavity.

00:02:18 We're going to explore the size of that alkyl group and the effect it may have.

00:02:23 And then we're going to look at what effect a substituent in the 6th position has on the

00:02:28 relative affinity of this class of compounds.

00:02:31 Both of the two optical isomers have been isolated.

00:02:35 Both have been co-crystallized with the enzyme and we have data that shows how both of these

00:02:41 two optical isomers interact with the enzyme.

00:02:46 Now we evaluate this interaction in two ways.

00:02:50 One which is given in terms of an I50.

00:02:54 This is a functional assay which measures the ability of these compounds to inhibit

00:02:58 the conversion of CO2 to bicarbonate.

00:03:00 It's a pH stat titration type assay.

00:03:04 The second is given in terms of a Ki.

00:03:07 Now this is a competition assay with a fluorescent ligand, the ancylamide, and it measures the

00:03:12 relative affinity of these compounds for the active site.

00:03:16 Both of these assays indicate that the SN antemur, MK417, is 100-fold more potent and

00:03:24 100-fold greater affinity for the enzyme than its RN antemur.

00:03:29 We were curious as to why a chiral center which is so far removed from the catalytic

00:03:36 site would have such a profound role and affinity.

00:03:41 Why there is this 100-fold difference, this 3 kilocalorie per mole difference in binding

00:03:46 energy.

00:03:48 And this is when we examined both of these compounds then within human carbonic anhydrase

00:03:52 2 to try to discern the answer.

00:03:55 Before we move to the results of the X-ray crystallographic studies, I want to remind

00:04:00 you that the thiopyrin ring can exist in two conformational forms.

00:04:07 Now as shown at the top of the slide, the alkyl amino group can be placed pseudo-axial

00:04:14 or at the bottom of the slide, pseudo-equatorial.

00:04:16 These two different conformers are interconvertible simply by a ring flip.

00:04:22 We would assume that the pseudo-equatorial conformer would be the more stable conformer

00:04:27 and this is what we might expect to see within the active site cavity of the enzyme.

00:04:32 Now in order to convince ourselves that pseudo-equatorial form is indeed favored, Mark Mirko, who is

00:04:39 now head of modeling at Vertex Corporation, undertook some ab initio calculations.

00:04:44 These calculations were done in cooperation with a group from the Cray Corporation.

00:04:51 At the highest level of theory at the 631 G star level, indeed the pseudo-equatorial

00:04:56 conformer is preferred over the pseudo-axial by about 1 kilocalorie per mole.

00:05:02 Now let's turn to the results of the X-ray crystallographic study and see whether this

00:05:07 lower energy conformer, the pseudo-equatorial conformer, is indeed bound within the active

00:05:12 site.

00:05:13 Here we view the more active S enantiomer bound within the active site of human carbonic

00:05:18 anhydrase 2.

00:05:20 Notice first that the alkyl amino side chain is not in the preferred pseudo-equatorial

00:05:24 orientation but rather in the higher energy pseudo-axial.

00:05:29 Now in order to be accommodated within the active site cavity, histidine 64 has rotated.

00:05:35 It's rotated 120 degrees.

00:05:38 This now allows room for the isobutyl amino group to fit within the cavity.

00:05:43 Let's turn to the zinc atom, shown here as the red sphere.

00:05:47 It's coordinated to the sulfonamide group, presumably through its deprotonated form.

00:05:52 The sulfonamide nitrogen is 2 angstroms from the zinc.

00:05:55 It's now the fourth ligand to zinc.

00:05:58 It's also hydrogen bound to threonine 199.

00:06:04 The lower oxygen of the SO2 is also bound to threonine 199 through the nitrogen of its

00:06:10 backbone.

00:06:11 Upper oxygen of the SO2 on the left is within 3 angstroms of the zinc atom and can form

00:06:19 a possible fifth coordinate to the metal.

00:06:23 Let's turn to the two oxygens of the sulfone.

00:06:26 The upper oxygen of the sulfone is within hydrogen binding distance of glutamine 92.

00:06:31 The lower oxygen of the sulfone is forming a weakly polar interaction with phenylalanine

00:06:36 131.

00:06:38 Let's contrast this orientation, the S enantiomer, with the less active R enantiomer.

00:06:45 Again we see the alkyl amino group is in the less favored pseudo-axial orientation.

00:06:51 Histidine 64 has again rotated.

00:06:54 But now to be accommodated within the active site cavity, two things have happened.

00:06:59 There's been rotation around the thiophene sulfonamide bond and the alkyl amino group

00:07:06 has folded into this Gauss geometry.

00:07:10 The sulfonamide group itself is coordinated to the zinc exactly as we've seen before.

00:07:15 This will be invariant in all the examples we look at today because of this high interaction

00:07:22 within the catalytic site.

00:07:24 Both oxygens of the SO2 are capable of forming the same polar interactions that we spoke

00:07:29 about with the S enantiomer.

00:07:31 In order to better view the two enantiomers and their relative positions in the active

00:07:37 site cavity, we can now see the S enantiomer shown in white, the R enantiomer shown in

00:07:44 blue, and the rotation that has occurred around the thiophene sulfonamide bond becomes evident

00:07:50 from the relative position of the thiophene rings.

00:07:54 Note that the side chain is transoid for the S enantiomer, it's Gauss for the R enantiomer.

00:08:01 Now this rotation and elevation of the thiopyrin portion of the ring place the alkyl amine

00:08:08 in exactly the same position in both optical isomers.

00:08:12 The terminal methyl groups are essentially identical.

00:08:16 We want to try to explain this difference in 100-fold in the relative affinity of these

00:08:21 two enantiomers for the enzyme by two components, the rotation that has occurred around the

00:08:27 thiophene sulfonamide bond and the Gauss versus trans geometry of the side chain.

00:08:35 Let's first look at the rotation that has occurred around the thiophene ring.

00:08:39 We'll examine this torsional component using a simplified model.

00:08:44 That model is thiophene sulfonamide itself.

00:08:49 The torsional contribution was calculated at the 321 G star level, and the preferred

00:08:54 rotamer places the sulfonamide nitrogen 75 degrees from the plane of the thiophene ring.

00:09:02 So as we see at the bottom of the slide, both the S and the R enantiomer are binding in

00:09:08 higher energy conformation with dihedral angles of 150 and 170 degrees.

00:09:14 Of these, the S enantiomer is preferred, and it's preferred by roughly one kilocalorie per mole.

00:09:22 So just this rotation around the thiophene sulfonamide bond influences the relative affinity

00:09:30 by about one-third of that three kilocalorie energy difference.

00:09:35 Let's turn to the conformational component, that is, the difference between the trans

00:09:40 and the Gauss geometry of the side chain.

00:09:44 And we've looked at this problem three different ways, one using the free energy perturbation

00:09:48 techniques that have been pioneered by Peter Coleman and others.

00:09:53 And this work is currently underway with Professor Mervs at Penn State.

00:09:58 The second approach has been using the Merck-Optimol program, which is an MM2-type program.

00:10:04 And finally, what I'll show you today is some ab initio calculations where we've used the

00:10:09 simplified model isopropyl isobutyl amine.

00:10:13 We've placed each atom of this model in their respective positions as found by X-ray crystallographic studies.

00:10:21 At the 321 G-star level, one finds indeed that the trans geometry is preferred over

00:10:29 the Gauss geometry by about one kilocalorie per mole.

00:10:33 So we can in fact account for the major part of the difference, this 100-fold difference

00:10:38 in binding between these two enantiomers in terms of the torsional component, that

00:10:44 is, the rotation around the thiophene sulfonamide bond, and the conformational component, that

00:10:48 is, the difference between the trans and the Gauss geometry.

00:10:51 I would next like to turn to the contribution of a 6-alkyl substituent.

00:10:57 What effect this substituent might have on the relative preference for the axial versus

00:11:02 equatorial orientation of the alkyl amino group?

00:11:06 And then secondly, what effect this substituent might have in non-bonded hydrophobic interactions

00:11:12 with the lipophilic surface?

00:11:14 We'll first examine then the effect that a 6-methyl substituent has on this axial-equatorial

00:11:20 preference.

00:11:21 Now, as you remember, without the methyl group, the amino group was in the preferred pseudo-equatorial

00:11:28 orientation.

00:11:30 Now, simply by placing a methyl group trans to that alkyl amino substituent, we have removed

00:11:37 that bias for pseudo-equatorial so that now pseudo-axial and pseudo-equatorial are essentially

00:11:44 at the same energy level, and there would no longer be a penalty for binding the pseudo-axial

00:11:49 amino group.

00:11:50 By introducing such a 6-methyl substituent, of course, we generate a second chiral center.

00:11:56 That means we now have four optical isomers.

00:11:59 All four of these isomers have been prepared and have been co-crystallized with human carbonic

00:12:04 anhydrase 2.

00:12:05 We'll now look at the results of placing that methyl group in the 6-position and then try

00:12:10 to explain it based on the geometry of the inhibitors.

00:12:14 The trans-SS enantiomer, the expected most active enantiomer, is about 100-fold more

00:12:21 potent than the cis-RS enantiomer.

00:12:26 That is, it has a I50 of about 0.3 nanomolar versus 29 nanomolar for the cis compound.

00:12:34 This is also true in terms of our Ki measurements.

00:12:37 The more active trans enantiomer has a Ki of about 0.4 nanomolar versus the cis enantiomer,

00:12:44 which has a Ki of around 15 nanomolar.

00:12:49 So we're now going to look at the results of the X-ray crystallographic experiments

00:12:52 and see if we can rationalize some of these differences in relative affinity.

00:12:56 First we'll view the most active trans-SS enantiomer.

00:13:02 Notice that it's situated within the active site cavity, exactly as our more active S

00:13:07 enantiomer previously.

00:13:09 The alkyl amino group is pseudo-axial, but presumably now we're not taking a penalty

00:13:14 for that pseudo-axial orientation.

00:13:16 The methyl group is pseudo-equatorial.

00:13:19 Histidine 64 has again rotated to its new position.

00:13:24 The sulfonamide group is in exactly the same alignment as we've seen previously, and both

00:13:28 oxygens of the SO2 are capable of the same polar interactions.

00:13:33 Next let's turn to the more active of the two cis enantiomers, the cis enantiomer having

00:13:40 the SR absolute configuration.

00:13:43 The alkyl amino and the methyl group both come off pseudo-equatorially.

00:13:48 Histidine 64 is again displaced.

00:13:50 The sulfonamide group is exactly the same as it was previously, and both oxygens of

00:13:55 the SO2 are capable of engaging in the polar interaction with glutamine-92 and phenylalanine-131.

00:14:04 Notice now we're beginning to see some rotation around the thiophene sulfonamide bond.

00:14:09 And next we'll turn to the two least active enantiomers.

00:14:14 First we'll view the trans-RR enantiomer.

00:14:18 Here we see major rotation has occurred around the thiophene sulfonamide bond.

00:14:22 The alkyl amino group is pseudo-axial, reaching down and displacing histidine 64.

00:14:28 The methyl group is pseudo-equatorial.

00:14:31 Again the two oxygens of the sulfone are still capable of engaging in the same polar interactions.

00:14:38 We'll contrast these three with the least active cis-enantiomer, the RS compound.

00:14:44 Here we observe again significant rotation around the thiophene sulfonamide bond.

00:14:50 The alkyl amino group and the methyl group are both pseudo-equatorial.

00:14:53 The alkyl amine reaches over and again causes rotation of histidine 64.

00:14:59 Now the best way to view these relative rotations within the active site cavity is by placing

00:15:04 them all in the same slide, and we'll view that next.

00:15:07 In this view, each compound is in its respective position within the active site cavity.

00:15:13 The two lower compounds, that is the two closest to the lipophilic surface, are the two most potent.

00:15:20 Their alkyl amino groups reach up and their terminal methyl groups are in the same position.

00:15:26 The two least active isomers fit highest in the active site cavity, furthest removed from

00:15:31 the lipophilic surface, and again their side chains are most similar.

00:15:36 The amino group is nearly in the same position in all four optical isomers.

00:15:42 The sulfur of the thiophene ring acts as a hinge, and each compound is a leaf, opening

00:15:47 up within the active site cavity.

00:15:50 Now although the sulfone oxygens differ in somewhat of their relative position, all seem

00:15:55 capable of engaging in the same polar interactions.

00:15:58 Let's examine the four optical isomers again, but now with the dihedral angle shown, that

00:16:04 is the angle between the thiophene and the sulfonamide nitrogen.

00:16:09 Now here we can see that the most active SS enantiomer is the most preferred, having a

00:16:15 dihedral angle of 140 degrees.

00:16:19 Now let's contrast that to the trans-RR enantiomer, with a difference of about 30 degrees.

00:16:27 That difference would translate down to about 1.5 kilocalories per mole.

00:16:32 Now if we look at the KI difference between the two, we see that the trans-SS is a KI

00:16:36 of around 0.4, the trans-RR at about 5.5, and this difference is about tenfold, or about

00:16:43 1.5 kilocalories per mole.

00:16:45 Thus we can account for nearly the entire difference in binding between these two optical

00:16:52 isomers simply based on the rotation around the thiophene-sulfonamide bond.

00:16:57 If we look at the most active again, the trans-SS, and contrast that to the cis-RS, the least

00:17:04 active, we see now that this difference of 50 to 100-fold can be explained in large measure

00:17:11 by the difference again in the rotation around the thiophene-sulfonamide bond.

00:17:15 That is a difference of about 30 degrees.

00:17:17 I next want to turn to the possibility that the 6-alkyl group could engage in non-bonded

00:17:24 hydrophobic interactions with the lipophilic surface.

00:17:29 Now here in red we can see that lipophilic surface, and through modeling we find that

00:17:34 by enlarging the alkyl group coming off the 6-position, we do indeed have the potential

00:17:41 for interacting with that surface.

00:17:43 Okay, we'll analyze the effect of this 6-alkyl substituent in the more active trans-SS series

00:17:50 and in the more active cis-SR series.

00:17:55 Here we place methyl, ethyl, and propyl groups in the 6-position.

00:18:01 Now in the trans-series, we see very little improvement in relative potency as determined

00:18:06 by the KI.

00:18:08 The 6-methyl and the 6-ethyl derivatives are identical.

00:18:12 Only with the 6-propyl do we see a very modest improvement by a factor of 2.

00:18:16 On contrast, in the cis-SR series, we see a step rise improvement in relative potency,

00:18:23 going from 2 nanomolar for the methyl down to 0.2 nanomolar for the propyl, an improvement

00:18:30 of about tenfold, which would translate down into a 1.5 kilocalorie difference in delta-G.

00:18:38 Why do we fail to see an improvement moving from the methyl to the ethyl substituent in

00:18:44 the trans-series?

00:18:45 Well, here in this model, we can see that the terminal methyl group of that ethyl substituent,

00:18:53 when it is pointed down towards the lipophilic canyon, is engaging in a negative 1.5-synpentane

00:19:00 interaction with the lower oxygen of the SO2.

00:19:04 We have explored this in more detail.

00:19:07 At the 321 G-star level of ab initio calculation.

00:19:12 Now in this slide, we see that the methyl group of that ethyl substituent can exist

00:19:19 either in position A, B, or C. In position A, it is pointing down into the lipophilic

00:19:26 canyon, but it's also engaging in this 1.5-synpentane interaction with the lower oxygen of the SO2,

00:19:33 and it is the higher energy conformer, being of higher energy by about 1 to 1.5 kcal over

00:19:41 the possibility when that terminal methyl group is positioned in C or B.

00:19:46 Let's contrast that with the cis-series, and now we're going to observe the effect

00:19:52 of orienting that terminal methyl group of the ethyl substituent down towards the lipophilic

00:19:57 surface.

00:19:58 Now that terminal methyl group is pointed down towards the lipophilic surface in position

00:20:02 D. Now let's contrast that to position F. That terminal methyl group in position F would

00:20:09 now engage in the 1.5-synpentane interaction with the upper oxygen of the SO2.

00:20:16 We now find that pointing down towards the lipophilic canyon is actually the preferred

00:20:20 position for the methyl group.

00:20:23 Thus we think we can explain this difference in enhancement in affinity in the cis-series

00:20:30 by building down into the lipophilic canyon without an energetic penalty.

00:20:35 In contrast, the trans-series, when we build down into the lipophilic canyon, we must overcome

00:20:43 the energetic penalty of the 1.5-synpentane interaction.

00:20:49 Now finally I want to turn to the substituent in the 4 position, the size of the alkyl group

00:20:56 attached to that 4-amino.

00:20:58 Here we see the 4-ethylamino-6-methyl with van der Waals surfaces shown, and very nice

00:21:05 van der Waals contact can be observed with the imidazole, the histidine-64, and its rotated

00:21:11 position.

00:21:13 So we wanted to know what advantage this rotation of histidine-64 might have in the

00:21:18 ability of these compounds to interact within the enzyme surface.

00:21:23 To answer this question, we've compared three compounds.

00:21:27 The example where we have an unsubstituted amino group in the 4 position, one where we

00:21:32 have a methyl amino group, and one where we have the ethyl amino substituent.

00:21:37 Now our experimentally determined KIs would suggest there's absolutely no advantage in

00:21:41 moving from the free amino to the methyl amino with KIs of 1.5 to 1.8 nanomolar.

00:21:49 But we see a significant improvement on the ethyl amino with a KI of 0.4, it's an improvement

00:21:56 of about 5-fold, and would translate to an improvement in binding energy of about 1 kilocalorie

00:22:02 per mole.

00:22:03 We have co-crystallized each of these examples with human carbonic anhydrase 2, searching

00:22:10 for an explanation for this improvement in binding affinity as we move from methyl amino

00:22:16 to ethyl amino.

00:22:17 This is a result of our co-crystallization experiment with the 4-amino derivative.

00:22:24 Notice that the alignment within the active site cavity is very similar to our most active

00:22:30 examples, but histidine-64 is in its native position, it is not rotated.

00:22:37 Now the red sphere to the right of the imidazole residue is meant to designate an ordered water

00:22:44 molecule.

00:22:45 On the X-ray crystallographic experiments, density was seen within hydrogen bonding distance

00:22:51 of the imidazole.

00:22:52 It was regular in shape, and it was interpreted as a water molecule, which is part of a chain

00:22:57 of water molecules, which reach out towards the surface of the enzyme.

00:23:02 Now we'll turn to the methyl amino derivative.

00:23:06 Again we find nearly identical alignment within the active site cavity.

00:23:11 Histidine-64 remains in its native position, and the ordered water molecule is still present.

00:23:18 Now we'll contrast the ethyl amino derivative with the 4-amino and the 4-methyl amino and

00:23:24 see if we can discern a difference.

00:23:27 The ethyl amino group now reaches up and displaces histidine-64 to its rotated position.

00:23:35 That ordered water molecule has been displaced.

00:23:38 It is no longer found in the chain of water molecules that run out towards the surface.

00:23:43 We believe that the ability of the ethyl amino group to displace histidine-64 and eliminate

00:23:49 one of the ordered water molecules and its resulting increase in entropy can explain

00:23:55 the five-fold improvement in binding that we find when moving from the methyl amino

00:24:00 to the ethyl amino substituent.

00:24:03 This iterative approach, this multiple X-ray crystallographic analysis, has led to five

00:24:10 determinants in the interaction between the inhibitors and carbonic anhydrase 2.

00:24:17 First there is the confirmation that the inhibitor must adopt in order to be accommodated within

00:24:22 the active site cavity.

00:24:24 Secondly is the rotation, the side chain of histidine-64.

00:24:29 Thirdly is the displacement of an ordered water molecule.

00:24:33 Fourth are the polar interactions between the inhibitor and glutamine-92 and phenylalanine-131.

00:24:39 And fifth are the lipophilic interactions between the sixth substituent and the lipophilic

00:24:43 surface of the enzyme.

00:24:46 Now by a careful analysis of these five determinants of binding, we can understand the difference

00:24:52 in relative potency between the various inhibitors.

00:24:55 And by understanding this relative difference, we can be led to new rational approaches for

00:25:01 drug design.

00:25:02 These approaches should be superior to the more empirical, classical approach to drug

00:25:07 discovery.

00:25:11 The next speaker is Dr. Michael D. Varney of Agaran Pharmaceuticals in San Diego, California,

00:25:18 where he serves as Director of Medicinal Chemistry.

00:25:21 His current research concerns the area of molecular design as it applies to the discovery

00:25:27 of novel drugs.

00:25:28 Dr. Varney's topic today is the crystal structure-based design of novel inhibitors of thymidylate

00:25:35 synthase.

00:25:36 Hello.

00:25:37 What I would like to talk to you about today is iterative protein crystal structure-based

00:25:45 design.

00:25:46 Iterative protein crystal structure-based design is a process by which crystallographic

00:25:53 structural information is used as the basis for the design of drugs.

00:25:58 The process itself is made up of a number of disciplines, that being protein crystallography,

00:26:05 computational chemistry, synthetic or medicinal chemistry, biochemistry, and pharmacology.

00:26:13 Shown here is a graphical depiction of this process.

00:26:18 The process itself begins with the structure of a particular enzyme or receptor of therapeutic

00:26:25 interest.

00:26:26 Once that structure has been solved, there is a characterization or computational analysis

00:26:31 phase in which we analyze the structure for information that we can use later on in the

00:26:38 design process.

00:26:41 Once we've characterized the active site, we go ahead and design potential inhibitors

00:26:47 and then we prioritize those inhibitors and synthesize the ones that we believe have the

00:26:53 highest likelihood of success.

00:26:56 Once the compounds are synthesized, we have a biochemical evaluation, which means that

00:27:01 we measure the binding constant of that compound to the enzyme in which it was designed, and

00:27:07 then we solve the structure of that compound bound in the protein again.

00:27:16 This process can be iterated over and over again and eventually can lead to drug or clinical

00:27:26 candidates.

00:27:28 There are two areas in which this type of a process has utility.

00:27:32 The first, shown here, is what we would refer to as de novo design or design into an empty

00:27:38 active site.

00:27:39 This type of a design allows for the design of diverse groups or structurally unique lead

00:27:46 compounds.

00:27:48 That is, that there is more than one solution to inhibiting a particular enzyme and if we

00:27:55 design compounds from scratch, we should be able to design more than one structural type.

00:28:01 The second area in which this type of a process is useful is in analog design.

00:28:07 That is, the design of analogs of a compound for which we have now solved the crystal structure

00:28:14 of it bound in the protein.

00:28:16 This type of an analog design process allows for us to optimize the binding properties

00:28:21 of a lead compound, and that lead compound can come from, for example, the de novo design

00:28:27 process or from the literature or a substrate analog.

00:28:32 In addition, analog design allows for us to manipulate the physical properties of a compound

00:28:38 to optimize not only its binding properties, but to optimize its biological and physical

00:28:45 properties as well.

00:28:47 For example, if the solubility of a particular compound is a problem, we can use the structure

00:28:53 of that compound bound in the protein to design an analog that will be more soluble, but because

00:29:00 we know how the compound binds, we can put that solubilizing functionality in an area

00:29:05 that does not compromise the binding of the molecule to the target enzyme.

00:29:11 Once the structure of a protein has been solved, the analysis phase of the process can begin.

00:29:16 Contained within a particular protein structure is an enormous amount of information that

00:29:20 can be used for design purposes, and we would like to extract as much of that information

00:29:25 as possible.

00:29:28 Shown here is a number of the methods by which we characterize a particular protein

00:29:34 active site.

00:29:36 Initially, what we do is determine the shape and the boundaries of the active site, and

00:29:43 the manner in which we do this is usually by running a water molecule around in the

00:29:47 active site and then displaying the output of that water molecule's contacts with the

00:29:56 protein in a continuous collection of dots called a Connelly surface.

00:30:02 Second, we calculate the electrostatic potentials on a three-dimensional grid in the active

00:30:08 site.

00:30:09 What this allows us to do is to determine the areas of an active site where there are

00:30:14 discrete or partial charges, because during the design process we would like to complement

00:30:20 those charges.

00:30:22 Third, we would like to identify and orient water molecules.

00:30:28 In a protein structure that is highly refined, water molecules that are ordered can be found.

00:30:36 We can orient those water molecules based on their hydrogen bonding needs.

00:30:41 Fourth, we identify the type and the orientation of hydrogen bonding functional groups that

00:30:48 the protein will present to any ligand that will bind.

00:30:52 Fifth, we identify the shape and the size of hydrophobic space.

00:30:58 The reason for this is because later on during the design process, for example, we might

00:31:03 have to decide whether we should put a phenyl group into a particular area or a cyclohexane,

00:31:10 and the shape and the size of a hydrophobic pocket can help us make that decision.

00:31:14 And last, in the characterization phase, we determine the flexibility of side chains

00:31:20 and backbone atoms that make up the active site.

00:31:24 The reason for this is because if we know that an area of a particular protein active

00:31:29 site is flexible, we have more leverage or more leeway in putting functional groups that

00:31:38 might be larger, for example, than was previously known.

00:31:44 Once the analysis phase is completed, we can then start the design phase.

00:31:52 Shown here is a collection of properties or design tactics that we would take in the de

00:32:00 novo design process.

00:32:03 So for example, in an empty active site with an initial molecule, what we would like to

00:32:08 do is to fill up as much space as we possibly can.

00:32:13 That is, to occupy the largest percentage of volume of the active site and, as a result,

00:32:21 increase the total surface area of interaction of our ligand with the protein.

00:32:27 Second, we can now use the periphery of our design skeleton to place hydrogen bonding

00:32:37 functionality that will complement the protein.

00:32:40 Third, we address each water molecule that is found in the active site with the question

00:32:46 in mind of, do we interact with this water molecule directly or do we displace the water

00:32:53 molecule and interact with the protein?

00:32:56 Fourth, we evaluate the conformational energy of potential inhibitors.

00:33:01 We can do this using conformational search techniques.

00:33:05 And the reason why we do this is because we would like to know what the internal energy

00:33:11 penalty a small ligand must pay to attain its binding conformation.

00:33:17 The reason why this is important is that that energy penalty is energy that's lost in binding

00:33:24 when the compound actually binds, and compounds that have a higher internal energy penalty

00:33:29 upon binding are, in general, poorer inhibitors.

00:33:34 Fifth, during the de novo design process, we always keep in mind the ease of synthesis.

00:33:40 We wouldn't want to design the perfect molecule and spend two years making it, nor would we

00:33:45 want to make a number of compromises in our design simply to make a compound very quickly.

00:33:52 And last, during this process, we keep in mind classical medicinal chemistry principles

00:33:58 because what we are trying to do is design drugs and not just protein inhibitors.

00:34:04 Before we start the actual design process, I would like to define a couple of terms.

00:34:10 The first is our definition of a lead compound.

00:34:15 Shown here are the criteria that we would apply to a potential ligand before we would

00:34:21 call it a lead compound.

00:34:23 A lead compound, in our hands, must have a measurable Ki of at least 100 micromolar.

00:34:29 The reason for this is because of the physical properties of small molecules and their solubility

00:34:34 limits.

00:34:37 More importantly, this compound must have a solubility that is higher than its inhibition

00:34:43 constant.

00:34:44 And the reason for this is because we want to dissolve that compound in a drop that we

00:34:49 are going to grow protein crystals from and solve the structure.

00:34:53 And if the compound cannot reach solubility that is high enough to form the complexes

00:34:59 in the drop, we cannot grow the crystal.

00:35:01 Third, we must have the structure of this compound solved with our protein of interest.

00:35:08 Fourth, this molecule must have some measure of flexibility.

00:35:13 The flexibility is important because the positions of atoms in an active site of a protein are

00:35:20 not known to any better than about 0.2 to 0.4 angstroms.

00:35:26 And certain molecular interactions fall off very quickly with distance.

00:35:31 So if a molecule were too rigid at the beginning of the process, we might miss certain interactions

00:35:38 that were important to binding and not understand why.

00:35:43 So initially, we allow some flexibility in the molecule to adjust itself and then later

00:35:49 on in the iterative design process, rigidify the molecule to optimize those interactions.

00:35:55 Fifth, we must have positions for elaboration.

00:35:59 What this means is that if we design a compound from scratch, we do not want to substitute

00:36:04 all of the positions on the molecule with functionality and then solve the structure

00:36:09 of that initial compound and have nowhere to go.

00:36:12 The particular enzyme that I would like to talk about today is thymidylate synthase.

00:36:18 Thymidylate synthase is a ubiquitous enzyme and it is the sole de novo source of deoxythymidylate

00:36:24 monophosphate in all living organisms.

00:36:28 Shown here is the thymidylate synthase cycle.

00:36:32 Thymidylate synthase catalyzes the reductive methylation of the 5-position of deoxyuridylate

00:36:38 monophosphate.

00:36:41 As part of that enzymatic conversion, 5,10-methylene tetrahydrofolate, which is the cofactor, is

00:36:47 converted into dihydrofolate.

00:36:50 To complete the cycle, dihydrofolate reductase reduces dihydrofolate to tetrahydrofolate

00:36:56 and serine hydroxymethyltransferase converts tetrahydrofolate to 5,10-methylene tetrahydrofolate.

00:37:04 The important point to remember is that inhibitors of this enzyme should be broad-spectrum antiproliferatives

00:37:13 and that can range from being antibacterial to antifungal to, our particular interest,

00:37:20 antitumor agents.

00:37:22 The important point to remember about thymidylate synthase is that it binds two ligands.

00:37:27 It binds its substrate, deoxyuridylate monophosphate, and it binds its cofactor, 5,10-methylene tetrahydrofolate.

00:37:36 Shown here are the structures of those two ligands.

00:37:40 The enzyme has two binding pockets, one that binds the substrate and the other that binds

00:37:44 the folate, and all of the design work that I will talk about today has been done in the

00:37:49 methylene tetrahydrofolate binding pocket.

00:37:54 Also shown is the first of the very active folate analogs that were synthesized and it

00:38:00 is a compound called 10-propargyl 5,8-didiazepholic acid.

00:38:04 This compound went into clinical trials in the early 1980s and was found to have antitumor

00:38:11 properties in humans, but ultimately never made it to the market because of unrelated

00:38:16 toxicities.

00:38:18 So our initial process began with the structure of CB3717 bound into thymidylate synthase,

00:38:26 and then ultimately for our de novo design, what we did was on the graphics machine, remove

00:38:31 that ligand from the protein and then use what we would refer to as the inhibited form

00:38:37 of the enzyme for our design.

00:38:40 Next I would like to introduce you initially just to the structure of the protein itself.

00:38:47 Shown here is a cartoon of the E. coli thymidylate synthase structure.

00:38:54 It is a homodimer and it has two active sites that are remote from each other.

00:38:59 This is a close-up of the active site itself.

00:39:02 On the bottom of the active site in yellow is the deoxyuridylate monophosphate analog

00:39:09 5-fluorodeoxyuridylate monophosphate, it is covalently bound to the thymidylate synthase

00:39:15 enzyme and it creates the floor of the binding pocket for the folate.

00:39:22 Shown in red is the CB3717 bound in the TS structure.

00:39:28 What I would like to show you now is a close-up of the active site of thymidylate synthase

00:39:34 looking into the opening of the binding pocket from bulk solvent.

00:39:40 Shown here is CB3717 in a stick model made from the protein structure itself bound into

00:39:47 thymidylate synthase.

00:39:49 The black hole in the middle is the opening to the binding pocket and in the back is the

00:39:53 quinazoline portion of CB3717.

00:39:57 Shown here is an electrostatically color-coded Conley surface.

00:40:02 The Conley surface itself is used for two purposes.

00:40:06 One is to visualize the boundaries of a particular active site or the van der Waals surfaces

00:40:12 of the protein and in addition it can be color-coded for different properties of the enzyme.

00:40:18 For example, the flexibility of the enzyme or in this case here, the charge or partial

00:40:24 charge of the atoms that make up the active site.

00:40:27 In this particular case, the spectrum of colors go from red for positively charged areas to

00:40:34 blue for negatively charged areas.

00:40:37 These types of color-coded Conley surfaces are used as guides in the design of ligands

00:40:43 and in particular used to put charged or partially charged functionalities into the areas where

00:40:50 they are most favored.

00:40:52 In the design process, we have used a number of computational tools to both characterize

00:40:58 active sites and to help us design ligands.

00:41:03 One of those programs is a program referred to as GRID that was written by Peter Goodford's

00:41:08 group at Oxford.

00:41:10 What this program does is it runs pieces of molecules or functional groups of molecules

00:41:16 on a three-dimensional grid in the active site of a protein and calculates potential

00:41:21 energies.

00:41:22 The output of a program like this can be displayed in what we would refer to as isoenergetic contour

00:41:27 maps.

00:41:29 Shown here is one of those isoenergetic contour maps on the active site of TS.

00:41:35 What this type of output allows us to see is, is in a graphic sense, what are the areas

00:41:40 of an active site where certain types of functionality would like to bind the most.

00:41:46 This particular slide is the areas where a methyl group or in this case a representative

00:41:51 hydrophobic functionality would be most favored in the TS active site.

00:41:57 Shown here is the output of GRID for an azine nitrogen.

00:42:02 And as you can see, the morphology of these contour maps are different than for the methyl

00:42:07 group.

00:42:08 What we can now do is display the output from different functionalities onto the same active

00:42:14 site in different colors and this can allow us to assign functionality to the periphery

00:42:20 of an initially designed skeleton in the hopes of having the best inhibitor.

00:42:26 What I would like to do now is introduce you to the de novo design process as we have practiced

00:42:32 it.

00:42:33 In this particular series of compounds, we began with a naphthalene ring that was placed

00:42:38 into the active site of TS using the GRID outputs as a guide.

00:42:44 Shown here is that naphthalene ring placed in the E. coli TS from top to bottom.

00:42:51 Once that naphthalene ring had been placed, we went to the periphery of that and used

00:42:56 GRID as a guide to design hydrogen bonding functionality that can complement the active

00:43:01 site.

00:43:03 So for example, we used a lactam NH to donate a hydrogen bond to an aspartic acid and we

00:43:09 used a carbonyl oxygen of that lactam to accept the hydrogen bond from a water molecule on

00:43:16 the lower part of the active site and that is the red X down there.

00:43:20 On the right of this, you can see now that there is empty space.

00:43:24 The Connelly surface has nothing to fill up that area and so now what we want to do is

00:43:30 to use the 6th position of what is now a fused tricycle or a benzindol to fill up that

00:43:37 area with substituents and also to have linker atoms that will allow us to exit the active

00:43:43 site.

00:43:44 What we used was a nitrogen atom.

00:43:47 The reason why we used a nitrogen atom at the 6th position was because we can make substituents

00:43:51 relatively easily and in addition to that, when we substitute that nitrogen, we do not

00:43:56 create chiral centers.

00:43:59 So initially, our substituents are an N-ethyl which you can see fills up the space relatively

00:44:04 nicely and then a CH2 group that will allow us to point out of the active site with an

00:44:12 additional functional group.

00:44:14 Shown here is the way in which we place that naphthalene ring with its initial set of substituents

00:44:20 into the active site of TS using, in this case, a hydrophobic probe from GRID as a guide.

00:44:28 Shown here is now looking into the opening of the active site from solvent and you can

00:44:34 see that what we have done is put a phenyl group on that CH2 and that phenyl group now

00:44:38 lays against an isoleucine side chain and it makes an edge-on interaction with a phenylalanine

00:44:44 side chain.

00:44:45 What we have now done is design a compound, essentially atom by atom, that we are now

00:44:52 going to go to the bench and synthesize.

00:44:55 That molecule is shown here and in addition to the phenyl group out on the periphery of

00:45:01 the molecule, we have added an SO2 papirazino group and the reason for this is the papirazino

00:45:07 group allows us to make salts which will be water soluble and because we were not able

00:45:13 to predict the binding constant of this particular compound, we wanted to make sure that the

00:45:19 compound was water soluble so that even if it were not the best inhibitor, we would

00:45:24 still be able to grow protein crystals of it.

00:45:28 This compound, as you can see, turns out to be an inhibitor of thymidylate synthase and

00:45:33 it is about a 20 micromolar inhibitor of the E. coli enzyme and it is about a 1.6 micromolar

00:45:39 inhibitor of the human enzyme.

00:45:41 Once this initial compound had been synthesized and its inhibition constant measured, our

00:45:47 protein crystallography group at Agaron grew crystals of the compound in E. coli thymidylate

00:45:52 synthase and solved the structure.

00:45:55 That structure is shown here.

00:45:57 What we can see is that the compound binds grossly, essentially the way that it was bottled

00:46:02 to bind with a few minor changes.

00:46:05 The most important is that the water molecule that the carbonyl oxygen was intended to interact

00:46:12 with, that is except the hydrogen bond, has now been displaced.

00:46:17 We can tell that because there is no electron density around that water molecule, which

00:46:23 is the yellow X at the bottom of the slide.

00:46:27 What the compound looks like in the stick model made from the protein structure.

00:46:32 As you can see, there is no water molecule down on the bottom of the slide and the Connelly

00:46:37 surface now protrudes into that area.

00:46:41 Once the structure had been solved, we spent some time analyzing it in detail and comparing

00:46:46 it to the models that we had made.

00:46:50 Shown here is a comparison of the binding, the actual binding confirmation of the molecule

00:46:57 versus the model.

00:46:59 The structure on the right is the modeled compound and the structure on the left is

00:47:05 the fit from the protein structure.

00:47:07 The difference is that the compound appears to have moved to the left or wedged itself

00:47:13 a little bit deeper into the active site and as a result, the carbonyl oxygen of the lactam

00:47:19 bangs into the carbonyl oxygen of the penultimate residue, alanine-263.

00:47:25 The result of this unfavorable interaction is a movement of alanine-263 carbonyl oxygen

00:47:32 out of the way and when that happens, the water molecule no longer has a pocket to bind

00:47:38 in and is displaced.

00:47:41 Now what we would like to do is to use this structure that we have solved to design an

00:47:46 analog that should have better binding by fixing some of the problems that the initial

00:47:53 compound had.

00:47:56 And the obvious place to go with this type of a molecule is the lactam area.

00:48:01 What we need to do is to either design a functionality that will interact with the

00:48:07 carbonyl oxygen of alanine-263 in a favorable way so that those two substituents can approach

00:48:14 each other more closely or to design functionality on the ligand that will fill up the space

00:48:21 left behind by the water molecule and interact with the protein directly.

00:48:26 What we decided to do was to replace the lactam oxygen with a nitrogen atom.

00:48:32 Now what this type of a replacement will do is to put two hydrogens on that nitrogen,

00:48:39 which can be potential hydrogen bond donors, but in addition to that, that nitrogen will

00:48:45 make that functionality an amidine functionality, which should have a pKa that would allow it

00:48:51 to be protonated while it was bound.

00:48:55 The reason why we had hoped that this would be the case is because that nitrogen that

00:49:00 is protonated now can interact in an electrostatic fashion with the aspartic acid, but in addition

00:49:07 to that, protonated or positively charged nitrogens are better hydrogen bond donors.

00:49:16 Shown here is a model of how that compound is expected to bind.

00:49:21 The NH2 group is now in the same space where the carbonyl oxygen was, but you can see that

00:49:28 it now has a hydrogen that can point towards the carbonyl oxygen of alanine-263.

00:49:35 With this analysis and design complete, what we end up with is a new molecule that now

00:49:40 has the lactam oxygen replaced with a nitrogen atom, and that is shown on this slide.

00:49:48 The compound, as you can see, is now a 34 nanomolar inhibitor of the human TS, which

00:49:57 means that it is about 50 times more active than the lactam-containing compound.

00:50:03 The crystallography group at Agaron grew crystals of this compound and solved the structure

00:50:09 of it bound in E. coli TS.

00:50:12 Again, what we see is that the compound, in a gross sense, binds similar to the way that

00:50:19 it was modeled, but the most important thing that we get from this structure is that now

00:50:24 there is electron density around the missing water molecule.

00:50:29 That is, that the water molecule has returned, and the reason for that, we believe, is because

00:50:35 the carbonyl oxygen of alanine-263 is now more closely approaching the nitrogen and

00:50:42 makes a hydrogen bond, and as a result of that, the pocket for the water molecule is

00:50:47 now there.

00:50:49 Shown here is a look into the active site from water of this compound bound in E. coli TS.

00:50:57 Now, with this second structure solved, what we wanted to do was to optimize the compound

00:51:03 even more, and the way that we chose to do that was to fill up additional hydrophobic

00:51:08 space that was not filled by the previous compound.

00:51:14 Shown here is one of those types of molecules.

00:51:17 This compound has two additional methyl groups in the previous compound.

00:51:21 One is in the periposition, or the 5 position of the benzindol, and the other is a methyl

00:51:26 group on the nitrogen of the amidine because that hydrogen of that nitrogen was not hydrogen

00:51:32 bonding to any functionality in the active site.

00:51:36 That newly designed molecule looks like this.

00:51:40 As you can see, it is what we would refer to as an optimized structure.

00:51:44 It is a two nanomolar inhibitor of the human enzyme, which puts it a factor of a thousand

00:51:50 better than the first compound in this series, and our protein crystallography group solved

00:51:57 the structure.

00:51:58 What we see is, is we see that the compound binds very similar now to the way that it

00:52:04 was modeled, and that the water molecule is very tightly bound and is hydrogen bonded

00:52:11 to the amidine NH.

00:52:14 We see a very tight association around the periphery of the molecule with the protein

00:52:19 active site, and very good complementary functionality.

00:52:24 And here is the hydrogen bonds that this molecule makes in the active site.

00:52:29 The amidine functionality, which we believe is protonated, is donating a hydrogen bond

00:52:35 to an aspartic acid, and at the bottom, we can see that the NH is making what we believe

00:52:41 to be a bifurcated hydrogen bond that is both to the alanine-263 carbonyl oxygen and the

00:52:49 water molecule that is hydrogen bonded additionally to the alanine-263.

00:52:55 So what we've been able to do now is design a compound essentially from scratch, atom

00:52:59 by atom.

00:53:01 That compound was an inhibitor of E. coli and human thymidylate synthase, and then using

00:53:07 the iterative crystallographic process, we refined that structure.

00:53:12 Shown here is a little bit more data on the activity of this type of compound.

00:53:19 As you can see, the initial compound, which we refer to as AG-117, had a Ki, or an inhibition

00:53:25 constant, of 1.6 micromolar against human TS.

00:53:30 In addition, it had measurable activity in cell culture as an antitumor agent, but it

00:53:37 had no thymidine reversal.

00:53:39 What this means is that even though the compound had some measurable cytotoxicity in cell culture,

00:53:47 inhibition or cytotoxicity was not due to inhibition of TS.

00:53:54 The next compound that I talked about was a 34 nanomolar inhibitor of the human TS.

00:54:01 It's a little bit more active in cell culture now.

00:54:03 As you can see, it's a 0.38 micromolar inhibitor of the L1210 cells, and it has some slight

00:54:10 reversibility in cell culture.

00:54:13 The last two compounds shown here are the optimized structures in this series of compounds.

00:54:19 The compound referred to as AG-331 is one of the most potent TS inhibitors known.

00:54:27 This compound has good activity in the L1210 cell culture, and as you can see also, it

00:54:34 has very good thymidine reversibility, which means that it is a very specific inhibitor

00:54:39 for TS in cells.

00:54:41 This particular compound is one of the first clinical candidates to come from the TS program.

00:54:49 So as you can see, what we have done is we have used the iterative design process to

00:54:55 design a compound from scratch and then optimize this series to the point where we have a clinical

00:55:03 candidate.

00:55:05 What this process can give you then is a more rapid access to active compounds and probably

00:55:12 more importantly though, a diversity of structural types that inhibit the same enzyme.

00:55:21 Thank you, Drs.

00:55:22 Varney and Baldwin.

00:55:23 In a few minutes, we will again be ready to start taking your telephone calls.

00:55:27 The telephone numbers again are 800-368-5781 and 5782 and 202-463-3170.

00:55:36 And the fax number, 202-887-3457.

00:55:39 You may begin calling in now if you care.

00:55:43 Before we begin the final general discussion, Dr. Peter Warner, who is Principal Medicinal

00:55:47 Chemist with Zeneca Pharmaceuticals in Macclesfield, England, is standing by to report on their

00:55:52 work on the use of protein structures in the design of human leukocyte elastase inhibitors.

00:55:57 Welcome Dr. Warner.

00:55:58 Hello.

00:55:59 Human leukocyte elastase, which I shall refer to today as HLE, is an enzyme produced by

00:56:06 neutrophils to destroy invading organisms, particularly in the lung.

00:56:12 An imbalance between free HLE activity and its endogenous inhibitor has been implicated

00:56:18 in a variety of pulmonary diseases, including emphysema and cystic fibrosis.

00:56:25 It is likely that an HLE inhibitor may have therapeutic value in these diseases.

00:56:31 This slide shows a cartoon representation of the X-ray crystal structure of HLE non-covalently

00:56:37 bound to a protein inhibitor called TOMI.

00:56:41 The structure was sold by Wolfram Bode.

00:56:44 HLE belongs to the styrene protease family of enzymes, which are characterized by a catalytic

00:56:51 triad of amino acids comprising styrene-195, histidine-57, and aspartate-102.

00:57:02 Heptidic trifluoromethyl ketone compounds, such as ICI-200-880, shown on this slide,

00:57:09 can trap the styrene-195 hydroxyl group by readily forming a hemiacetal, and therefore

00:57:15 function as potent, selective, and reversible HLE inhibitors.

00:57:21 Although ICI-200-880 is very effective when dosed directly into the lung, its oral activity

00:57:28 is somewhat less.

00:57:29 Therefore, we began a program to discover orally active trifluoromethyl ketone HLE inhibitors

00:57:37 by attempting to design suitable non-peptidic replacements for the tripeptide backbone using

00:57:43 protein structural information.

00:57:46 To date, no crystal structure of HLE bound to a trifluoromethyl ketone inhibitor exists.

00:57:53 By considering the protein X-ray structure of HLE bound to TOMI and an analogous porcine

00:58:00 elastase structure, we created an active site model of HLE bound to a tripeptidic trifluoromethyl

00:58:07 ketone using the ICI Molecular Modeling System Enigma.

00:58:13 We anticipated that this model would be useful in the design of putative non-peptidic inhibitors.

00:58:21 The active site model revealed several potential interactions between the inhibitor and the

00:58:26 enzyme, which are shown on this slide.

00:58:30 These potential key interactions include covalent attachment of the styrene-195 hydroxyl group

00:58:37 to the trifluoromethyl ketone to form a hemiketal shown in red, hydrogen bonds to the oxyanion

00:58:45 represented by arrows, lipophilic binding of the P1 isopropyl group into the S1 pocket,

00:58:55 and hydrogen bonds between the enzyme and the P1NH, P3 carbonyl, and P3NH group.

00:59:03 A number of possible replacements for the P2 proline and P3 valine residues of ICI-20880

00:59:11 were considered, and it was conceived that a pyridone system shown in blue might fit

00:59:17 into the active site and still maintain the key interactions with the enzyme if substituted

00:59:23 correctly.

00:59:25 It was therefore chosen as an attractive potential dipeptide replacement.

00:59:30 The appropriately functionalized pyridone ring system was arrived at after a considerable

00:59:35 amount of analysis and modeling work, which cannot be described in this brief presentation.

00:59:42 Before synthesizing the putative inhibitor one, we checked that the confirmation of compound

00:59:48 one required to bind into the active site model would be a relatively low energy one.

00:59:54 On the basis of the modeling, we decided to go ahead and make and test compound one,

00:59:59 and we were very encouraged to find that it binds to the enzyme with a Ki of 2.8 micromolar.

01:00:08 Further modeling studies suggested that a suitable lipophilic capping group on the pyridone

01:00:13 amine might be advantageous.

01:00:17 We chose to make the analog two with a benzyloxycarbonyl capping group, which is shown in pink on the

01:00:23 slide.

01:00:26 Compound two was found to be 10 times more potent than compound one.

01:00:31 We postulated that further improvements in potency might be obtained by placing an appropriate

01:00:37 functional group into the S2 pocket of the enzyme.

01:00:42 Modeling of compound one in the active site model indicated that a benzyl group attached

01:00:48 to the five position of the pyridone ring might be suitable.

01:00:53 We therefore synthesized compound three with the pyridone ring bearing a benzyl substituent,

01:01:00 which is shown in pink at the five position, and were delighted to find that it is a potent

01:01:05 inhibitor of HLE with a Ki of 40 nanomolar.

01:01:11 An enzyme kinetic analysis of the compound showed it to be reversible and competitive.

01:01:16 It is also very selective for HLE, even against related enzymes.

01:01:23 The final slide shows compound three, colored blue, modeled into a simplified form of the

01:01:28 active site of HLE.

01:01:32 In conclusion, compound three was discovered with the assistance of protein crystal structural

01:01:38 data and molecular modeling techniques, and fulfills our objective of finding a potent,

01:01:44 reversible, and selective inhibitor of HLE that is substantially non-peptidic in structure.

01:01:51 It is also less stereochemically complex, which was another objective at the outset.

01:01:57 The pyridone trifluoromethyl ketones are a new series of HLE inhibitors with a potential

01:02:04 of possessing oral activity, which may be useful agents to treat diseases such as emphysema

01:02:10 or cystic fibrosis, in which an excess of HLE activity is involved or implicated.

01:02:17 Thank you very much, Dr. Warner.

01:02:21 Although this program has been packed with information, we know that you want to keep

01:02:24 up with continuing advances in modeling techniques and equipment.

01:02:27 Your seminar notes include the names of several useful resources.

01:02:31 Another resource is the Digital Equipment Corporation information line, 800-DEC-INFO.

01:02:38 Please add this number to your list.

01:02:39 That's 800-DEC-INFO.

01:02:42 If you complete and return your post-course questionnaire, we will mail you a printed

01:02:46 list of more resources.

01:02:48 Please let us hear from you during this final discussion period.

01:02:50 Again, the telephone numbers are 800-368-5781 or 5782 and 202-463-3170 if you're calling

01:03:01 either in Washington, D.C. or internationally.

01:03:03 If you get a busy signal, please hang up and try again.

01:03:06 We're probably ordering pizzas or something, right?

01:03:09 Okay.

01:03:10 We go first to Columbus, Ohio, for our first question in this go-round.

01:03:12 Hello.

01:03:13 My name is Albert Soloway.

01:03:14 I'm from the College of Pharmacy at Ohio State.

01:03:19 And my question might be directed to Dr. Varney, but also to others.

01:03:25 Most of the work that's presented relates to enzyme inhibitors.

01:03:29 My own interest in neutron capture therapy, I was interested in putting compounds into

01:03:34 DNA of tumor cells.

01:03:38 And I wondered about the possibility whether there's enough knowledge with respect to kinases

01:03:43 and polymerases to be able to design compounds that could be able to use the enzymatic systems

01:03:50 for incorporating directly into DNA or RNA.

01:03:54 Michael?

01:03:55 I'm not quite sure I understand exactly what you're asking.

01:04:00 Is it that there is enough, is there enough structural information about DNA to design

01:04:04 compounds that bind to it?

01:04:06 Or is there enough structural information about enzymes that bind to DNA?

01:04:10 I'm not, I'm interested not just in binding to DNA, but to actually become incorporated

01:04:17 into the DNA molecule.

01:04:19 In other words, could I design materials that would become incorporated and by that incorporation

01:04:26 be a site for destruction of tumor cells?

01:04:31 I presume that you could do that.

01:04:32 I myself don't have any experience in that.

01:04:35 But I think that there is enough known about the structure of DNA or molecules that interact

01:04:39 that you could, you could do something of that type.

01:04:43 Anyone else have a suggestion on that?

01:04:47 No?

01:04:48 Okay.

01:04:49 Next we go to Cincinnati, Ohio for our next question.

01:04:54 Hello, Cincinnati.

01:04:55 Hello.

01:04:56 This is Herschel Weintraub at Mary and Meryl Dow.

01:05:00 I have a question first for Michael Varney.

01:05:04 In your de novo methods, how often do you hit an active compound such as what you presented

01:05:12 on the first or second try and how often do you have to keep trying and plugging away

01:05:15 at different compounds?

01:05:19 From our experience, what we've found that in designing compounds from scratch, we usually

01:05:24 will make somewhere between two and five examples of that structural type initially.

01:05:30 Somewhere between 30-50% of those compounds will have binding constants in the 10-50 micromolar

01:05:37 range against our target enzyme.

01:05:41 We have a second question from Mary and Meryl Dow in Cincinnati.

01:05:44 Go ahead, please.

01:05:45 Yes.

01:05:46 This is addressed to John Erickson.

01:05:49 You noted that on binding of the inhibitors in the HIV protease, the flap moves.

01:05:54 In all of the structures which have been reported to date, is the flap pretty much in the same

01:05:59 location and is it safe then to do modeling essentially with the fixed flap placed down

01:06:06 into the bound orientation or do you have to always take into account the potential

01:06:10 movement?

01:06:14 All the structures that have been reported to date that are out in the literature show

01:06:21 a remarkable structural similarity in the structures of the flaps to within half an

01:06:30 angstrom RMS deviation or so with a variety of different kinds of peptidomimetic inhibitors,

01:06:37 symmetric inhibitors and so on.

01:06:40 So I think in most cases where you have the resolution you can see and you don't have

01:06:45 the problem of disorder or disordered orientations of the inhibitors.

01:06:50 You can also see the buried water molecule that's conserved interacts between the backbone

01:06:57 of the inhibitor and the backbone of the flap.

01:06:59 So I think it's fairly safe to assume that this confirmation of the flap is a preferred

01:07:06 confirmation of the flap.

01:07:07 It may not be the only one that's possible but that it's a safe bet to design against.

01:07:12 Okay.

01:07:13 And we have yet a third question from Marion Merrell Dowell in Cincinnati.

01:07:17 Go ahead please.

01:07:18 Hi.

01:07:19 Thank you.

01:07:20 This is Bob Farr from Marion Merrell Dowell.

01:07:22 The molecular modeling of the Beringer-Ingelheim alfalfa-difluorostatone in which an endanyl

01:07:30 glycine replaces the PT-proline has the endanyl residue unexpectedly sticking into solution.

01:07:38 Dr. Warner, have you examined this compound under your molecular modeling versus the benzyl

01:07:48 pyridone?

01:07:49 And if so, what interactions have you observed between the enzyme and the aromatic rings

01:07:55 of these compounds?

01:07:57 Are they indeed sticking out in the solution or are there some hydrophobic interactions

01:08:02 with the enzyme?

01:08:04 I think what you're referring to is some N-alkylglycine derivatives.

01:08:10 And in fact, we had at ICI a number of years ago made some N-alkylglycine derivatives of

01:08:17 our lead compound, 20880, and discovered them to be quite potent inhibitors.

01:08:23 And we did in fact use that information in the design process.

01:08:27 I wasn't able to go into that in the short presentation today.

01:08:31 Our modeling suggests that it is conceivable that a substituent on the nitrogen could bind

01:08:39 into the P2 pocket, but equally well it could also bind out of that pocket.

01:08:44 We do not have an x-ray crystal structure of that sort of compound as yet from our own

01:08:48 work.

01:08:49 All right.

01:08:50 Thank you, Dr. Warner.

01:08:51 We're going now to Philadelphia to LaSalle University.

01:08:55 Go ahead, please.

01:08:56 Hi.

01:08:57 This is Mordy Chikowicz calling on behalf of Dr. Donald Upson.

01:09:04 The first question is for Dr. Baldwin.

01:09:07 When considering inhibition of zinc metalloenzymes, what is the relative importance of the strength

01:09:14 of metal ligation in comparison to maximizing non-metal interactions?

01:09:24 The question can be addressed perhaps by looking at the simplest unsubstituted arylsulfonamide

01:09:34 like thiophene sulfonamide itself.

01:09:38 That would have a Ki value of in the micromolar range.

01:09:44 So we're getting interaction with the zinc presumably through the sulfonamide.

01:09:49 Now, as we increase the acidity of that sulfonamide group through substitutions and add groups

01:09:56 which are capable of interacting with the macromolecule, we reduce that Ki value from

01:10:02 micromolar level down to subnanomolar level.

01:10:05 So I think the zinc is a critical binding component, but there's other factors in the

01:10:12 overall binding features as well.

01:10:14 Okay.

01:10:15 Did you have a second question from Philadelphia?

01:10:18 This is for Dr. Varney.

01:10:21 What benefits, if any, are there to starting with leads considerably more potent than 50

01:10:27 to 100 micromolar?

01:10:31 The benefits of course would be that it wouldn't take very long to get to a nanomolar range.

01:10:36 I think it depends on where you get that lead, but they're certainly using this approach

01:10:41 that there are advantages to having a more active compound because you can use all of

01:10:47 the things that the three of us talked about in our programs to design derivatives both

01:10:53 for the purposes of increasing binding and changing or modulating pharmacological properties.

01:10:58 So it would be an advantage, sure.

01:11:00 Okay.

01:11:01 You go to the ACS Chicago section in Chicago, Illinois.

01:11:07 Go ahead, please.

01:11:08 Good afternoon from the Windy City.

01:11:12 I have a question on technique.

01:11:16 I have an interest in QSARs.

01:11:20 In your molecular design, either Dr. Varney or Dr. Baldwin, did you use QSARs and how

01:11:28 did you?

01:11:30 And if not, do you see a role for this technique in rational drug design?

01:11:38 I guess I could start.

01:11:41 Currently there's been an awful lot of QSAR analyses done with carbonic anhydrase.

01:11:48 And parameters have been developed and one I just mentioned a minute ago is the acidity

01:11:52 of the sulfonamide group.

01:11:54 And that correlates quite well along with some lipophilicity factors in terms of the

01:12:01 I50 measurements.

01:12:04 So I think QSAR has a role in analysis, but in a lot of ways the ability to view these

01:12:11 inhibitors within the macromolecule is really a very powerful tool which can in many ways

01:12:19 complement, be complemented by QSAR.

01:12:23 Okay.

01:12:25 We have a question for Dr. Baldwin and my interpreter here is going to take care of

01:12:28 it for me.

01:12:29 Okay.

01:12:30 I will try.

01:12:31 Since the N-ethyl group in N-ethylthienothiopyrin is subject to conformational heterogeneity

01:12:39 due to torsion, have the N-isopropyl or N-tertiary butyl analogs been examined?

01:12:46 What are the results?

01:12:48 What about N-cyclopropylcarbinol, N-allyl, and N-benzyl analogs?

01:12:54 Let me tell you where that is from.

01:12:55 I can handle this.

01:12:56 Okay.

01:12:57 Burroughs Welcome in Research Triangle Park, North Carolina.

01:13:00 Okay.

01:13:01 Obviously, we explored a wide range of substitution on the 4-amino substituent.

01:13:14 What we found was, as I've indicated, a strong component for chirality.

01:13:22 As that substituent got larger than isobutyl, potency fell off.

01:13:29 We did not x-rayed any of those particular examples, but presumably, we were not fitting

01:13:33 well then within the active site cavity.

01:13:36 But the N-ethyl derivative and isobutyl derivatives are roughly in the same range.

01:13:42 It doesn't look like we're picking up additional binding energy through isobutyl.

01:13:47 But if you look at the analysis of the x-ray structures very closely, there's a little

01:13:52 bit more perturbation in the isobutyl group.

01:13:55 The ethyl amino group, of course, is situated in this very nice transoid arrangement.

01:14:01 Okay.

01:14:02 This question, which was faxed to us, is for anybody.

01:14:05 I can handle this one.

01:14:06 What is the minimal resolution of a protein structure needed to initiate a rational drug

01:14:10 design strategy?

01:14:12 This comes from the MRI Teleconference Center, Bowman's Gray School of Medicine in Winston-Salem,

01:14:16 North Carolina.

01:14:17 Anybody want to handle that?

01:14:19 What is the minimal resolution of a protein structure needed to initiate a rational drug

01:14:23 design strategy?

01:14:25 I could start with using the carbonic anhydrase analogy.

01:14:30 The early structure of carbonic anhydrase that was developed by Professor Liljes at

01:14:36 Lund University in Sweden was probably not at a range that would be terribly useful.

01:14:43 However, in the mid-1980s, he refined that structure to about the 2 Angstrom level and

01:14:50 was kind enough to provide us with a preprint of his publication.

01:14:55 And it was at that level, then, which served us in good stead in really studying, then,

01:15:02 the ligand macromolecule complexes.

01:15:05 So I would say about 2 Angstrom, and I'll turn it over to the other people.

01:15:08 Yes.

01:15:09 I would actually give a slightly more liberal answer to that question, Jack.

01:15:14 The short answer that I would give would be atomic resolution.

01:15:17 Whatever resolution is required to give you electron density maps with which you can interpret

01:15:23 the atomic structure of the protein in the side chains.

01:15:29 Obviously, any information you can get at the atomic level takes you from nothing to

01:15:36 something.

01:15:37 And then you have at least a framework within which to begin thinking about structure-based

01:15:42 design.

01:15:43 The higher the resolution that you can obtain from that point on, the more sure you can

01:15:49 be about some of the actual conformations and atomic interactions that you see.

01:15:54 Okay.

01:15:55 John?

01:15:56 John, you described some rather subtle differences in how the hydroxyls were interacting with

01:16:00 the aspartase.

01:16:01 Give us a sense of what resolution you might need to really reliably refine structures

01:16:07 and distinguish those structures.

01:16:09 Those structures were actually determined at the 1.8 to 2 Angstrom resolution level

01:16:18 at which the actual electron density maps could clearly distinguish between R and S

01:16:26 stereoisomers, for instance, which was important to make sure that we had the right assignments

01:16:32 on the molecules that we were looking at.

01:16:37 The structure of the monowall compound, A74704, on the other hand, which was the first crystal

01:16:42 structure I presented, was done at about 2.8 Angstrom resolution.

01:16:49 And even in that structure, we could clearly see the position of the hydroxy group.

01:16:54 So to some extent, it depends upon how good your data is.

01:16:58 Not all three Angstrom structures are equivalent.

01:17:02 Not all two Angstrom structures are equivalent.

01:17:04 Okay.

01:17:05 Yes, Michael?

01:17:06 I would like to add one thing to that, and I think the criteria that we use is if you

01:17:10 can find crystallographically bound water molecules, then in general, that structure

01:17:15 is at a resolution that's useful for design.

01:17:17 If you cannot find the water molecules, then in general, we don't pursue the structure.

01:17:23 We have found that about a 2.5 Angstrom cutoff usually exists in a highly refined structure

01:17:29 to find the water molecules.

01:17:30 Okay.

01:17:31 We're off to England again in the Glaxo Research Group in Middlesex.

01:17:34 Go ahead, please.

01:17:35 Yes.

01:17:36 I have a question for John Baldwin here.

01:17:39 It's about the pseudo-axial and pseudo-equatorial conformational flexibility.

01:17:44 If I interpret your figure 9 correctly, you have the lone pair on the amino substituent

01:17:51 on the same face as the sulfoxide oxygen.

01:17:55 Obviously, in your in vacuo ab initio calculations, there'll be a strong electrostatic propulsion

01:18:01 across the top face of the molecule.

01:18:03 Did you consider, which that, of course, would disfavor the pseudo-axial conformation.

01:18:12 Did you consider different conformations in which either a proton was on the top face

01:18:17 or, in fact, that a minor group was protonated as it may well be?

01:18:22 Okay.

01:18:23 Let me address that question starting at the end.

01:18:29 That amine group actually is a very interesting one in the fact that it is relatively non-basic.

01:18:35 It has a pKa of about 5.8, probably due to components of the SO2 group.

01:18:43 Now, we found that if the basicity present within the molecule is higher than that higher pKa,

01:18:54 then we get a lot of non-specific binding to melanin pigment in the eye.

01:19:00 And compounds with basic nitrogens are not very active in animal models.

01:19:06 So we were pleased with the pKa that was observed.

01:19:09 And to answer your question about the various rotamers, this work was done by Mark Merko and by Graham Smith.

01:19:18 And if one does not consider all the rotamers, obviously, one gets a different answer.

01:19:22 And, yes, indeed, they were all considered.

01:19:24 Okay.

01:19:25 Let's move on because we have many callers that are still waiting and we have a relatively short time to go.

01:19:29 Memphis State University, Memphis, Tennessee.

01:19:31 Go ahead, please.

01:19:32 Yes, sir.

01:19:33 This is Scott Weston of the University of Mississippi calling from Memphis State University in sunny Memphis, Tennessee today.

01:19:38 Dr. Varney touched on a couple of methods in his talk of defining the characteristics of a protein cavity of interest.

01:19:46 I wonder if the other members, including Dr. Varney of your panel there, would discuss the relative advantages

01:19:52 and disadvantages of various methods to define the characteristics of such protein cavities of interest

01:19:59 and also the relative advantages and disadvantages of fitting pieces or entire structures such as caveat, doc, et cetera,

01:20:07 into those protein cavities once they're defined, please.

01:20:11 Go ahead.

01:20:12 Well, I guess I'll start and I'll talk primarily about one tool that I used, and that was GRID.

01:20:19 We have found in general in using GRID that you have to use it in a very qualitative sense.

01:20:27 One can, of course, calculate numbers, potential energies are calculated with the functional groups in GRID,

01:20:33 but that because these are not normalized in any way, areas that are essentially peninsulas

01:20:40 or would surround a functional group on all three sides will in general be deep wells for both amino functionalities

01:20:46 and methyl functionalities.

01:20:49 So in using GRID, we have found that these areas are very difficult to distinguish what functional groups to place in there,

01:20:57 and at that point in time, you have to invoke desolvation arguments or hydrogen bonding complement to decide which.

01:21:04 So in the use of that particular program, I would say use it in a qualitative sense,

01:21:09 essentially to focus your attention on certain areas of an active site.

01:21:13 Okay. Any other comments?

01:21:14 We'll move on to Lehigh University in Bethlehem, Pennsylvania.

01:21:18 Go ahead, please.

01:21:20 I'm Richard Moorbacher, and my question is directed to Drs. Farney and Baldwin.

01:21:27 Could you comment on the relative value in terms of meaningful structural information

01:21:32 that can be obtained from co-crystallized drug or ligand receptor complexes

01:21:38 versus the infusion of a solution of a ligand or drug into a crystalline receptor to get a complex?

01:21:44 Jack Baldwin?

01:21:45 I'll start at least with our experience.

01:21:50 We tried, of course, to diffuse the ligand into carbonic anhydrase, and invariably the crystal cracked,

01:21:59 and I understand that this can be indicative of movement that's occurring in the macromolecule,

01:22:05 the type of movement that we've seen in histidine-64.

01:22:09 So in our case, we couldn't use the diffusion technique.

01:22:12 We had to depend on co-crystallization experiments.

01:22:18 Okay. We're going to handle another facts question here, and if you'll bear with me here so I can read the writing.

01:22:23 Dr. Baldwin described a case where loss of a buried H2O molecule, an entropy-favored event, improved the affinity of binding.

01:22:33 On the other hand, Dr. Varney told us of an example where preserving an internal H2O molecule resulted in a 50x-increased affinity.

01:22:42 In general, what is the role of internal, tightly-bound water molecules in promoting receptor-ligand interactions?

01:22:56 Well, I guess I'll answer that first, to say that the approach that we have taken is to address every water molecule as an individual.

01:23:06 I think that the arguments in principle are that if you can displace that

01:23:11 and essentially isosterically replace that water molecule in terms of the interactions of your functional group with the protein,

01:23:19 that in general, you should get an increase in binding.

01:23:22 But if you cannot do that, and it often is very difficult to do,

01:23:27 that the approach that I would take is to initially just begin by trying to order the water molecule,

01:23:32 to interact with it, and essentially make it a more highly-ordered or a more, I guess, happier water molecule in terms of the number of hydrogen bonds.

01:23:43 So I could address one point as far as the ordered water molecule displacement that we've seen.

01:23:50 That water molecule actually is part of a proton relay network,

01:23:54 which carries the proton from the activation of the water molecule at the zinc and removes it out to bulk solvent.

01:24:02 And there's indications that histidine-64 actually exists in the native enzyme as one major conformer.

01:24:10 But recent studies at the 1.5, 1.6 Angstrom level of resolution by Professor Lilges

01:24:19 suggest that indeed there is a minor rotamer of the side chain of histidine-64.

01:24:25 So that ordered water molecule then disappears in this minor rotamer.

01:24:30 So I think that the water molecule that we see is actually one which is vulnerable to being lost and then replaced as one goes through this shuttling process.

01:24:43 Okay. I'm sorry that we have some of you still on the line, and we were not able to get to you.

01:24:47 Unfortunately, our time period for questions is over right now.

01:24:50 We would like to thank very graciously Dr. Warner, who was with us in Macclesfield, England.

01:24:57 And we appreciate your being with us today, sir, there, and appreciated your input as well.

01:25:01 Before we leave, we have some final comments from Jonathan Greer and Helen Free.

01:25:05 First, Jonathan.

01:25:06 Thank you.

01:25:08 We presented today a series of examples where the three-dimensional structure of the protein target molecule

01:25:14 in conjunction with X-ray crystallographic studies on the protein-ligand complex has been used to aid in the design of novel drugs.

01:25:23 The power of the structural information was used in each case to achieve the particular goals required by the respective project,

01:25:30 high potency, high specificity, and structural novelty.

01:25:35 In all three instances, the efforts have contributed to the discovery of a compound that is currently in clinical trials.

01:25:44 An important and rapidly developing part of the structure-based design cycle depicted here is structure analysis and compound design.

01:25:54 Each of the speakers today described different strategies used to perform this crucial part of the cycle.

01:26:00 There are a broad range of methodologies that have been and are being developed to take advantage of structural information

01:26:06 for the design of analogs that have not as yet been described or illustrated in this series of programs,

01:26:13 but are of increasing use and interest to the chemistry community.

01:26:17 These techniques include de novo lead structure generation with such programs as ALADDIN, CAVIAT, DOC, GRID, LUDI,

01:26:27 determination of pharmacophore groups and their three-dimensional structure using programs such as APEX, CATALYST,

01:26:34 and quantitative potency analysis and prediction of ligand potency using structure-based methods such as COMFA or free energy perturbation.

01:26:43 Many of these techniques can be used even when knowledge of the three-dimensional structure of the protein is not available, only for the ligands.

01:26:52 These new methods are beginning to produce exciting results.

01:26:56 We plan to devote the next ACS satellite television program of this series to present some of these new methods

01:27:03 and examples of their application in the field of structure-based drug design.

01:27:08 Thank you for your support of this series.

01:27:10 We hope that you have found them as stimulating to listen to as we have to produce them.

01:27:16 And thank you, Jonathan, for your support of this program and continuing interest in it, and to all the speakers for their presentations.

01:27:24 As I said earlier, ACS is what our members make it.

01:27:28 I hope that each and every one of you watching today will actively join our ACS Dream Team,

01:27:34 so that we can continue to enrich the professional lives of you and your fellow scientists,

01:27:39 and continue to reach out to the public with the message of how important science, particularly chemistry, is in enhancing their lives.

01:27:48 Well, we've come to the end of today's program, the 11th American Chemical Society Satellite Television Seminar.

01:27:53 Thanks to our speakers, John Baldwin, John Erickson, Gerd Volkers, Didier Rognon, Michael Varney, and Peter Warner,

01:28:00 for making this an interesting and informative program.

01:28:02 And a special thanks to Jonathan Greer for organizing the program, and to Helen Free for hosting it, indeed.

01:28:08 On behalf of the American Chemical Society Continuing Education Department, I want to thank all of you for joining us today.

01:28:14 And please don't forget to complete and return your seminar evaluation forms.

01:28:17 We need your comments and suggestions.

01:28:20 For all of us here, and for all of you out there, thank you very much for joining us today.

01:28:24 Join us next June for our next teleconference.

01:28:50 Thank you.