TETHERING: Fragment-Based Drug Discovery

30 Apr 2004 18:12 AR AR214-BB33-10.tex AR214-BB33-10.sgm LaTeX2e(2002/01/18)P1: FHD10.1146/annurev.biophys.33.110502.140409

Annu. Rev. Biophys. Biomol. Struct. 2004. 33:199–223doi: 10.1146/annurev.biophys.33.110502.140409

Copyright c© 2004 by Annual Reviews. All rights reserved

TETHERING: Fragment-Based Drug Discovery

Daniel A. Erlanson, James A. Wells,and Andrew C. BraistedSunesis Pharmaceuticals, Inc., 341 Oyster Point Boulevard, South San Francisco,California 94080; email: [email protected]; [email protected]

Key Words disulfide exchange, small-molecule inhibitors, fragment assembly,structure-based drug design, molecular recognition

■ Abstract The genomics revolution has provided a deluge of new targets for drugdiscovery. To facilitate the drug discovery process, many researchers are turning tofragment-based approaches to find lead molecules more efficiently. One such method,Tethering1, allows for the identification of small-molecule fragments that bind to spe-cific regions of a protein target. These fragments can then be elaborated, combined withother molecules, or combined with one another to provide high-affinity drug leads. Inthis review we describe the background and theory behind Tethering and discuss itsuse in identifying novel inhibitors for protein targets including interleukin-2 (IL-2),thymidylate synthase (TS), protein tyrosine phosphatase 1B (PTP-1B), and caspases.

CONTENTS

INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200BACKGROUND . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200

Fragment-Based Drug Discovery. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200Finding Fragments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201Designing Fragments for Tethering. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204

TETHERING AS A FRAGMENT DISCOVERY TOOL. . . . . . . . . . . . . . . . . . . . . . . 204IL-2: Probing an Adaptive Binding Site. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204IL-2: Combining Medicinal Chemistry with Tethering. . . . . . . . . . . . . . . . . . . . . . 208Thymidylate Synthase: Discovering and Elaboratinga Core Fragment. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209

TETHERING WITH BREAKAWAY EXTENDERS: DISCOVERY OFA NEW PHOSPHOTYROSINE MIMETIC FOR PTP-1B. . . . . . . . . . . . . . . . . . . . 211

TETHERING WITH EXTENDERS: TETHERING AS A FRAGMENTASSEMBLY TOOL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214Discovery of Nonpeptidic Caspase Inhibitors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215Opportunities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218

CONCLUSIONS AND FUTURE DIRECTIONS. . . . . . . . . . . . . . . . . . . . . . . . . . . . 218

1Tethering is a servicemark of Sunesis Pharmaceuticals, Inc. for its fragment-based drugdiscovery.

1056-8700/04/0609-0199$14.00 199

30 Apr 2004 18:12 AR AR214-BB33-10.tex AR214-BB33-10.sgm LaTeX2e(2002/01/18)P1: FHD

200 ERLANSON ¥ WELLS ¥ BRAISTED

INTRODUCTION

Modern molecular biology has overwhelmed drug discovery organizations withmyriad novel drug targets. These targets have not only expanded the familiarfamilies of G protein–coupled receptors (GPCRs) and enzymes that most drugsaddress but have provided new families of enzymes and protein-protein targets.A major challenge for the drug discovery community is validating this wealthof targets with small molecules, and ultimately developing drugs for the morepromising targets.

The problem of finding a good small-molecule starting point is one of search-ing chemical space. The number of possible small (up to 30 nonhydrogen atoms)drug-like molecules is estimated to be greater than 1060 (8a). Fortunately, manysolutions exist to interrogate a site of interest with small molecules. Structural bi-ology, biochemistry, enzymology, biophysics, and molecular modeling can definekey binding interactions and restrict the chemical search dramatically. Moreover,the drug discovery process is not purely random. Medicinal and combinatorialchemistry can improve the binding affinity of a weak starting compound. Nonethe-less, given the vastness of chemical space, finding good starting points remains asignificant bottleneck in the drug discovery process.

One way to make chemical diversity space more manageable is to search forsmaller molecules, or even fragments of molecules, and then merge or elaboratethese fragments. Discovering drugs in pieces can reduce the dimensionality ofthe search and dramatically improve the chances of finding good starting pointsfor the drug discovery process against novel drug targets. This review describes apowerful new technology for doing so.

BACKGROUND

Fragment-Based Drug Discovery

The notion that entities with weak binding affinity can be linked to form higher-affinity molecules was first shown for ligands binding to metal ions, giving rise tothe so-called “chelate effect.” Jencks (38) applied this theory to small molecule–protein interactions. An early test of this theory (51) showed that free energiesfor the binding of fragments of substrates and analogs to beta-hydroxy-beta-methylglutaryl coenzyme A (HMG-CoA) reductase were additive and, when cor-rected for the linking energy, approximated the binding affinity of the parentmolecules. Recent structural work comparing the binding of parent compoundswith that of fragments derived from them in thymidylate synthase neatly shows thefragments binding in a fashion similar to that of moieties in the parent compound(69). The concept of additive binding effects has been widely exploited in bothmedicinal chemistry and protein engineering.

Additive binding allows the search for drugs in pieces, which offers a tremen-dous combinatorial advantage over discovery of drugs intact. For example, to


TETHERING 201

produce a million-compound library (the size of many pharmaceutical compoundcollections), one could link 1000 different fragments asymmetrically in all binarycombinations. Assuming one can screen the fragments separately and then linkthem, the diversity of this huge library could be gleaned from a much smallercollection of compounds, in this case just over 1000. This huge combinatorialadvantage is one of the main drivers for the development of fragment-based drugdiscovery methods, but there are additional advantages to fragment-based drug dis-covery as well. Small fragments can gain access to smaller nooks in the protein thatmay not be easily accessible to a larger, nonoptimized scaffold. Small fragmentsare synthetically more accessible and often more soluble. Small libraries (of smallfragments) are easier to maintain and fully characterize than massive compoundcollections. And “growing” a molecule from a small scaffold can offer a muchsimpler chemical pathway than does remodeling a larger and more complicatedscaffold while maintaining a constant molecular weight.

Finding Fragments

NON-TETHERING METHODS Detecting fragments that bind a target can be chal-lenging. Traditional high-throughput screening (HTS) typically relies on inhibi-tion assays. But high concentrations of compound may be needed to find weakinhibitors. For example, an inhibitor with an IC50 of 1 mM produces less than10% inhibition when screened at 100µM, an effect that may fall within the back-ground error range of the screen. Moreover, HTS that employs high compoundconcentrations is often not practical, either because many molecules are insolubleor because it requires prohibitively large amounts of compound. Even more sig-nificant is the problem of false positives: Causes other than the desired one-to-onebinding interaction can yield assay inhibition. A variety of chemotypes has beenidentified as nonspecifically chemically reactive, and strongly colored compoundscan be problematic in spectrophotometric screens (61). Recently, Shoichet andcolleagues (49) have demonstrated that a number of molecules can act “promis-cuously” against a wide range of targets, probably by forming large aggregates.In response to these pitfalls, many researchers have pursued “positive” selectionprocedures (75). The rationale is that, whereas many factors may interfere withan assay and produce a negative signal that would be misinterpreted as inhibition,fewer artifacts can produce a positive signal.

The three techniques most frequently used for positively detecting fragments arenuclear magnetic resonance (NMR), X-ray crystallography, and mass spectrome-try (MS). Of these, NMR screening has been most extensively utilized for fragmentdiscovery. NMR screening can be performed at high compound concentrations andoften provides a good understanding of binding interactions. Structure activity rela-tions by NMR (SAR by NMR), a technique developed by Fesik and coworkers (45,67), is a powerful and versatile strategy. Other research groups (8, 22, 23, 37) havedeveloped complementary approaches. X-ray crystallography is unmatched in thelevel of detail it provides to deconstruct small molecule–protein interactions, and



several groups (7, 44, 53) have used crystallography to discover and develop frag-ments. Crystallography can also be combined with NMR to yield a hybrid approach(76). Finally, MS can also provide a positive detection method. Although informa-tion available about the site of binding is usually more limited than that obtainedwith NMR or crystallography, MS can be extremely high throughput (41, 70).

TETHERING In general, the preceding three methods of fragment discovery relyon noncovalent interactions between the target protein and small-molecule ligands,requiring high ligand concentrations for detectable occupancy on the protein. Incontrast, Tethering relies on reversible covalent bond formation between the frag-ment and the protein of interest. This in effect amplifies the affinity of the fragmentfor the target molecule, enabling detection at lower concentrations. The methodalso controls the region where selected fragments bind on the protein. Tetheringrequires relatively little protein (10–50 mg to screen more than 10,000 fragments)compared to other fragment-based approaches and is not limited by protein sizeor crystallization properties. The only essential requirements for Tethering are acoarse three-dimensional model of the protein target and the ability to analyze theprotein by MS. Tethering also provides an in situ linking procedure not affordedby other methods (see below).

The basic method of Tethering is shown in Figure 1 (19). If the target proteincontains a cysteine residue within or near the targeted site, it can be used directly.Otherwise, site-directed mutagenesis can introduce a cysteine residue. The cysteineresidue should generally be within 5 to 10A of the site of interest and should berelatively surface exposed to facilitate thiol-disulfide exchange (see below). Oncethe cysteine is in place, the protein is reacted with a library of disulfide-containingfragments under partially reducing conditions. In theory, the cysteine should forma disulfide with each of the fragments, and the exchange should occur rapidly.Assuming that the reactivity of the disulfide in each fragment is equivalent (seebelow), and assuming no noncovalent interaction between the fragments and theprotein, the mixture at equilibrium should consist of the protein disulfide-bondedto each fragment in equal proportion. However, if one of the fragments has inherent

Figure 1 Tethering fragment-based drug discovery.


TETHERING 203

Figure 2 Simplified schematic showing some of the equilibria in Tethering. Fragmentswith inherent affinity for the protein (Square B) have a more stable disulfide bond andpredominate at equilibrium.

affinity for the protein of interest, and if it binds to the protein near the introducedcysteine, then the thiol-disulfide equilibrium will be shifted in favor of the disul-fide for this fragment, and this protein-fragment complex will predominate. Inother words, the fragment will be “selected” by the protein. The overall level ofmodification can be tuned using a reductant, such as 2-mercaptoethanol (2-ME).

Figure 2 illustrates some of the competing reactions that can occur in a simpli-fied system. Multiple equilibria are possible for a reaction consisting of just twodisulfide-containing fragments, the protein, and a reductant. Even in this simpli-fied case, more than a dozen species will be present. However, if only the proteinspecies are considered, only four possibilities exist: protein alone, protein+ re-ductant, protein+ fragment A, and protein+ fragment B. The dominant proteinspecies can be identified with MS. If each fragment in a given pool has a uniquemolecular weight, MS also reveals which fragment binds predominantly to theprotein and, thus, which has the highest intrinsic binding affinity for the protein.We have demonstrated that a single compound can be selected from a pool of asmany as 100 different fragments (19). Because of resolution limits in the massdetermination of larger proteins and the small size and mass redundancy of thefragments (generally between 100 and 250 Da), pools of around 10 compoundsare most convenient for screening. To facilitate identification, the pools are con-structed so that each fragment differs by at least 5 Da from any other fragment inthe same pool.

Tethering is related to dynamic combinatorial chemistry (DCC), a concept inwhich reversible equilibria are exploited to generate receptors or ligands in situfrom fragments (32, 43, 60). The system is generally set up to thermodynamicallyselect molecules that have a desirable property (for example, high affinity to a



receptor) through bond-forming and bond-breaking reactions among a group offragments. A variety of bond-forming reactions have been investigated in the caseof DCC, including metal-ligand interactions, oxime formation, imine formation,boronic acid ester formation, disulfide formation, and others (60). In theory, anyof these chemistries could be used for Tethering. In practice, the thiol-disulfideexchange is ideal. The cysteine residue is naturally occurring and therefore easyto introduce into proteins. The disulfide bond can be formed and broken underextremely mild conditions with no other effects on biological molecules. Theequilibria in a thiol-disulfide exchange system can be tuned to more or less strin-gent conditions merely by altering the ratio of free thiol to disulfide (26). Finally,previous research has shown that the stability of intermolecular (56, 68) and in-tramolecular (14) disulfide bonds can be coupled to the binding energies of theinteracting molecules. A recent example describes using disulfide exchange toselect peptide ligands to the translocase protein Tom20 (55).

Designing Fragments for Tethering

To be useful for drug discovery, fragments should be small (molecular weightpreferably lower than 250 Da), heavily functionalized, and contain no “toxi-cophores” or other malignant functionalities. Several researchers have writtenabout the design of compounds for fragment-based discovery methods (23, 36).

Fragments for Tethering must also contain a thiol or disulfide bond. Thoughsynthetically simple, few disulfide-containing small molecules are commerciallyavailable, so it is necessary to custom build the fragment library. For convenience,each molecule contains a unique fragment on one side of the disulfide and acommon, small, solubilizing group such as aminoethanethiol on the other side.

A critical assumption unique to Tethering is that the disulfide moiety in eachfragment has a similar reduction potential. In reality, the electronic and stericcontext of a disulfide affects its overall stability (31, 42). Because at least a coupleatoms separate the disulfide bond from the variable portion in our fragments, weobserve only small differences in reactivity. Moreover, to minimize the impact ofthese differential reactivities, we group molecules so that all fragments in a givenpool have the same connectivity between the variable portion and the disulfidebond.

TETHERING AS A FRAGMENT DISCOVERY TOOL

IL-2: Probing an Adaptive Binding Site

Protein-protein interfaces are considered one of the most difficult target classesto inhibit with small molecules, yet they also represent a large and importantclass for therapeutic intervention (13, 73). In contrast to enzyme targets, wheresubstrate analogs can provide starting points and where the active site offersboth a binding pocket to exclude solvent and side chain functionality for binding


TETHERING 205

interactions, protein-protein interfaces tend to be large, relatively flat surfaces. Fewwell-characterized small-molecule ligands inhibit protein-protein interactions, andthose that have been described are often high molecular weight or exploit an al-losteric site.

The cytokine interleukin-2 (IL-2) is a critical component in the immune re-sponse (52). Binding of IL-2 to its hetero-trimeric receptor (α, β, andγ chains)induces T cell proliferation, and IL-2 has been implicated in several immune sys-tem disorders. Two antibody therapeutics that target the IL-2α receptor, Zenapax(Roche) and Simulect (Novartis), have been approved for treating graft-host dis-ease associated with renal transplantation. The efficacy of these drugs validatesthe manipulation of the IL-2 signaling pathway in treating diseases of the immunesystem and identifies IL-2 as an important therapeutic target. Unlike these proteintherapeutics, small molecules offer the potential for oral bioavailability and morecontrolled pharmacokinetics.

IL-2 is also one of the few protein-protein targets for which a well-validatedsmall-molecule inhibitor has been identified. Hoffman-La Roche identifieda molecule (compound1, Figure 3, IC50 = 3 µM) that binds specifically andreversibly to IL-2 (72). We determined the structure of the complex between1 andIL-2 to pinpoint the binding site’s location and the specific interactions between1 and IL-2 (5). The X-ray structure revealed a striking transformation: The sur-face of unliganded IL-2 is relatively flat and offers no obvious binding pockets,whereas the structure of the complex reveals a series of significant side chain andloop movements that create a hydrophobic binding pocket accommodating thebiaryl-alkyne of1 while the guanidine group forms a bidentate salt bridge withE62 (Figure 4A). Intriguingly, the region where the small-molecule binding siteis located closely overlaps with the residues that have been implicated as criticalfor IL-2Rα binding by alanine-scanning mutagenesis (65, 79). The commonalitybetween the receptor binding site and the ligand binding site suggests that thisregion is a true binding hot spot (17).

The observation that substantial rearrangements occurred in the bound stateestablished that the binding site on IL-2 was mobile and adapted to1 upon bind-ing. Thus, targeting these adaptive regions may identify binding sites that are notobvious from the native protein conformation and that, because of the lability ofthe structure, are not amenable to structure-based design.

We applied Tethering to address this binding site with several goals in mind.First, we wanted to identify fragments that bind to this site to better understand thekinds of chemical functionality preferred in this region. Second, we were interestedin determining whether these fragments could be used to create new lead moleculesor whether the fragments could be used in conjunction with a medicinal chemistryprogram to advance low micromolar hits into high-affinity ligands.

We first prepared a panel of 11 cysteine mutants (Figure 4B) targeting the regionsurrounding the binding site of1 (5). To confirm that the protein was correctlyfolded, each mutant was evaluated for its ability to bind IL-2Rβ. Those muta-tions that were at positions already identified as important for IL-2Rα binding by

30A

pr2004

18:12A

RA

R214-B

B33-10.tex

AR

214-BB

33-10.sgmLaTeX

2e(2002/01/18)P1:FH

D

206E

RLA

NS

ON¥

WE

LLS¥

BR

AIS

TE

D

Figure 3 IL-2 ligands.


TETHERING 207

alanine scanning also affected IL-2Rα binding in our screens. Using Tethering, wescreened all cysteine mutants against a library of approximately 7000 fragments.

The redox conditions used for Tethering were tuned such that hits occurredat a frequency of 0.1% to 1%. Certain cysteine mutants selected many more hitsthan did others; these cysteines mapped to the adaptive binding region, whereasthe hit-poor mutants generally targeted the more fixed regions of IL-2. Many hitsidentified in the adaptive region had structural similarities, suggesting specificSAR. Different cysteine mutants rarely selected the same hits, although similarfragments with different length linkers were occasionally selected by cysteinesthat could access the same regions.

Because the binding of1alters the structure of IL-2, we investigated its influenceon the selection of fragments through Tethering. By running the experiments ineither the presence or absence of1, we determined that some fragments werecompetitive with1and that others were selected synergistically (35). This approachhas the potential to rapidly identify complementary fragments in the absence ofstructural information and thus to overcome a key challenge of fragment assembly:identifying which fragments can be productively linked.

We initially focused our attention on one of the strongest hits, an indoleglyoxylate-derived fragment (compound2, Figure 3) that was selected by Y31Cin the absence of1. A Tethering SAR study of2 was accomplished by synthe-sizing analogs containing a common disulfide linker. By pooling the fragmentsand allowing them to compete for binding and concomitant disulfide bond for-mation, Tethering enabled a rapid rank ordering of the fragment binding strength.Selected compounds were tested as IL-2 inhibitors in an ELISA assay and werefound to inhibit IL2-Rα binding in the 500–600µM range, suggesting that thefragments identified through Tethering can be quickly modified to generate weakbut measurable binding activity (see also Thymidylate Synthase, below).

The crystal structure of2 and IL-2 Y31C was solved, revealing the binding sitefor this fragment (Figure 5A). The overlap between the phenyl ring of the fragment2 and the phenyl ring of1 is close, suggesting that this adaptive region contains aconsensus aromatic binding pocket. Fewer rearrangements occur upon fragment2 binding than upon compound1 binding. Most strikingly, F42 is in the wild-typeconformation compared to the “flipped” conformation observed with the bindingof 1.

The close overlay between1 and 2 suggested a fragment merging strategy:Compound3 (Figure 3), which replaces the phenyl ring of1 with the indoleglyoxylate 2, was synthesized and found to have an IC50 value of 1.5µM, atwofold improvement over1. The crystal structure of3 bound to IL-2 was solved,demonstrating that3bound IL-2 as expected (J.D. Oslob, A.C. Braisted, M. Randal& M.R. Arkin, unpublished data). The orientation of the indole fragment is nota perfect match with the structure of the covalently bound fragment (Figure 5B),which may explain why only a modest improvement in activity was observed.Nonetheless, this result demonstrated the feasibility of merging an establishedlead molecule with fragments discovered by Tethering.



IL-2: Combining Medicinal Chemistry with Tethering

Our interest in IL-2 as a target led us to pursue1 with medicinal chemistry as well.A chemistry effort provided a novel compound series exemplified by compound4(Figure 3), but this series struck an activity plateau in the low micromolar range.We decided to craft new compounds by merging4 with fragments identified usingTethering (9).

Because most of the fragments were selected on the adaptive region of IL-2, thisarea appeared to have significant potential binding energy. Y31C and L72C weretwo of the more hit-rich mutants, and both mutants showed a statistical preferencefor small aromatic carboxylic acids. We suspected that hits selected from Y31C(the mutant that selected2) might also overlap with the lead compounds, so wefocused on L72C. L72C is located on anα-helix, and analysis of similar structuresfrom the PDB indicated three possible vectors to orient the fragment. Two of thesevectors would direct the fragment toward solvent, but the third would bury thefragment into the adaptive region with only minor side chain rearrangements. Wewere unable to obtain fragment-conjugated crystal structures, and the adaptivityof this region makes modeling highly speculative, but the suggestion of a proximalaccessory site indicated how to merge fragments with the soluble lead. On thebasis of the length of the linker and the predicted vector of the disulfide bond, wedesigned and tested a series of 20 compounds that attach an aromatic ring with apredominately acidic functionality off the 4 position of4 using a two-atom linker.

While the activity of4 was only 3µM, 8 of the 20 compounds that merged4with fragments from Tethering had nanomolar activity, representing improvementsbetween 5- and 50-fold (Figure 6). All of the more potent compounds contained acarboxylic acid, as the results from Tethering had indicated. The most potent com-pound,7, inhibits the IL-2:IL2Rα interaction with an IC50 of 60 nM, which makesit the most potent small-molecule inhibitor yet characterized against a cytokinetarget. Subsequent crystallography has revealed the binding mode of7 (70a). Per-haps most important is the efficiency with which this molecule was identified.More traditional efforts to improve this series by modifying the core failed to

Figure 6 IC50 values for IL-2 inhibitors designed by appending fragments from Tetheringonto a known binder.


TETHERING 209

improve affinity. By incorporating the fragments from Tethering and thereby in-troducing larger, more directed functionality, we accessed a previously unknownsubsite. Making such large changes to a lead series without any indication of po-tential benefit is inefficient. Integrating Tethering with medicinal chemistry guidesnot only the direction of growth but also what types of chemical functionality toincorporate.

Thymidylate Synthase: Discovering and Elaboratinga Core Fragment

In addition to identifying novel sites on a protein, Tethering can identify novelfragments in a known ligand binding site. This strategy can be particularly powerfulwhen combined with existing information about the active site, as any new fragmentcan be merged with elements from known inhibitors to produce new, high-affinityinhibitors. The first published example of Tethering did just this (19).

The enzyme thymidylate synthase (TS) is an ancient, highly conserved enzymepresent in virtually all organisms. It plays a critical role in the nucleotide syn-thesis pathway and is especially important in actively dividing cells where DNAis rapidly synthesized. As a result, it is a prominent cancer target; the widelyprescribed chemotherapeutic agent 5-fluorouracil targets this enzyme. Mecha-nistically, the enzyme relies on a cysteine located in the active site to facili-tate catalysis. Moreover, the catalytic pocket is large, as the enzyme needs tobe able to bind both deoxyuracil monophosphate (dUMP) as well as the co-factor (6R)-5,10-methylenetetrahydrofolic acid (mTHF). TheEscherichia coliversion of the enzyme has been extensively studied both crystallographicallyand mechanistically (15). These factors made TS an ideal choice for developingTethering.

We initially screened TS against a small library of about 1200 disulfide-containing fragments. Although the TS construct we used contains four cysteines inaddition to the active-site cysteine, examination of the crystal structure and prelimi-nary modification studies demonstrated that only the active-site cysteine is accessi-ble for disulfide-exchange. Of all of the compounds screened,N-arylsulfonamide-proline derivatives were most strongly selected, represented byN-tosyl-proline.Substituents around the phenyl ring showed some variability, but the presence ofthe proline ring itself was essential.

To understand this SAR, we determined the crystal structures ofN-tosyl-D-proline bound to TS, with and without the disulfide linker, and compared its po-sition with the natural substrates (Figure 7). Significantly, the presence of thecovalent linker minimally affected the overall structure (Figure 8) (19). The crys-tallographic structure of N-tosyl-D-proline bound to TS shows a hydrogen bondbetween a sulfonamide oxygen and Asn177 on TS as well as hydrophobic inter-actions between the proline ring itself and Trp80; altogether, about 80% of thesmall molecule is buried. When we mutated the active-site cysteine to a serine andintroduced a new cysteine residue nearby (L143C), we were also able to select the



Figure 9 Inhibition constants (Kis) forN-phenyl-sulfonamide-proline derivatives.

N-tosyl-proline fragment, demonstrating that the same hit can be found from dif-ferently placed cysteine residues. X-ray crystallography showed that the fragmentbound similarly to the noncovalent and C146-linked structures, as shown in Figure8 (19). Thus, the disulfide linkage itself has at most a modest effect on fragmentbinding.

Enzymatic assays demonstrated thatN-tosyl-D-proline (compound8, Figure 9)inhibits TS competitively with respect to dUMP, albeit with a low activity (Ki =1.1 mM), a level that would almost certainly not have been detected in conventionalHTS. Superimposing the crystallographic structure of fragment8 with that of TSbound to its cofactor mTHF (34) (Figure 7) revealed that appending a glutamateresidue at the 4 position of theN-tosyl group would mimic the glutamate residueof mTHF. Synthesis of this compound9 (Figure 9) does in fact result in a 30-fold increase in activity. The SAR around this glutamate residue indicates thattheα-carboxylate is important for binding (compare compound9 with compound11or compound12), whereas theγ -carboxylate is much less important (comparecompound9with compound10). In general, the SAR around the glutamate moietyis similar to that previously observed for mTHF (40) and mTHF mimics (47).

Next, we elaborated the proline side chain to try to further improve the affin-ity. Compound13 (Figure 9) revealed that the negative charge is not necessary,whereas compound14 showed that some substituent is essential for the inhibitorto be competitive with respect to dUMP; this observation is consistent with the


TETHERING 211

structure shown in Figure 7, which reveals that this side chain is the only portionof the molecule that extends into the dUMP binding site. These data suggested thepossibility of trying to mimic the phosphate moiety of dUMP directly, which ledto the synthesis of compound15, which has a Ki of 330 nM.

Introducing an acidic side chain off the proline ring enhanced affinity sig-nificantly (compare compound15 and compound13). Compound16, which isnearly isosteric with compound15, has a much lower affinity, indicating that theacid moiety is essential. Compound17, with a positively charged substituent,is only partially competitive, suggesting that its side chain does not occupy thedUMP binding site. To discover the origins of this SAR, we determined the crystalstructure of compound15 (Figure 10) bound to TS. Compound15 forms a well-ordered complex with TS in which theβ-alanine side chain is involved in multiplehydrogen-bond contacts with residues Tyr209, His207, and Ser167. The first twoof these residues normally interact with the 3′-hydroxyl of the substrate dUMP.

This example demonstrates how Tethering can be used to discover a core frag-ment that binds to the active site of an enzyme. Structure-based design allowed usto rapidly improve the affinity of this millimolar hit by more than three orders ofmagnitude.

TETHERING WITH BREAKAWAY EXTENDERS: DISCOVERYOF A NEW PHOSPHOTYROSINE MIMETIC FOR PTP-1B

In the case of TS, we used Tethering within the active site to discover a novel corefragment that we then elaborated to enhance affinity. This was successful partlybecause TS has a large, open active site. However, for many enzymes the activesite is smaller and more easily disrupted. A case in point is the class of protein ty-rosine phosphatases, or PTPs. These enzymes, which remove the phosphate groupfrom phosphotyrosine (pTyr) residues in peptides and proteins, are increasinglyrecognized as an important but difficult class of drug targets (39, 78). One of theprimary challenges associated with these enzymes is the fact that the active site hasevolved to recognize the highly charged and therefore nondrug-like pTyr residuewith great specificity. Several research groups have sought to find replacementsfor the pTyr residue (3, 11, 45, 54), but few compounds have made the transitioninto clinical trials.

PTPs contain a conserved cysteine residue in the active site that is highly reactivetoward electrophiles (4, 28), and may in fact be regulated in vivo through disulfidebond formation with small endogenous thiols (6). However, the active site is deepand narrow, with room for little other than the pTyr residue itself. Therefore, we feltthat Tethering from the native active-site cysteine would not be useful for discover-ing pTyr replacements, as any identified fragments would likely bind outside of theactive site. At the same time, we were reluctant to introduce mutations too close tothe active site for fear of disrupting the fragile and highly tuned catalytic machinery.



Figure 11 Tethering with breakaway extenders to identify new fragments within anactive site.

To access the active site for Tethering without disrupting it, we developed Teth-ering with breakaway extenders, as illustrated in Figure 11 (21). In this approach,we first introduce a cysteine residue well outside of the active site. We then modifythis residue with a cleavable “breakaway extender,” which, when cleaved, posi-tions a thiol close to the active site. The newly introduced thiol is then accessiblefor Tethering. We chose to develop the approach on the antidiabetic target PTP-1B,which is well characterized structurally, biochemically, and physiologically. Wemutated Arg47 to a Cys (R47C), which places a thiol near the pTyr-1 binding site(62, 64), but places the mutation well outside the active site.

To validate the concept, we first synthesized prototype extender18(Figure 12).This molecule contains a bromoacetamide moiety designed to react with the R47Cresidue as well as an oxalamic acid moiety, a common pTyr mimetic designed tobind in the pTyr binding site. Earlier experiments demonstrated that simple bro-moacetamides react rapidly with the active-site cysteine C215. However, the pTyr

Figure 12 PTP-specific breakaway extenders.


TETHERING 213

mimetic part of the prototype extender binds in the active site, sterically blockingthe active-site cysteine from reacting with the bromoacetamide. Binding of theoxalamic acid moiety in the active site also positions the bromoacetamide to reactwith R47C. The design was confirmed by X-ray crystallography, which demon-strated that the prototype extender18binds as expected. In fact, the oxalamic acidmoiety superimposes closely on the pTyr moiety of a typical peptide substrate (21).

Having demonstrated that we could protect the active-site cysteine by plug-ging the pTyr binding pocket and simultaneously modify R47C, we next designedbreakaway extender19 (Figure 12). Although this extender is roughly the samelength as the prototype extender18, there are several differences, most notablythe presence of the thioester bond in the middle. Nonetheless, we found that thisbreakaway extender cleanly and selectively modified our PTP-1B at R47C, andwe could cleave the thioester efficiently with hydroxylamine.

We then screened the modified protein against our library of approximately15,000 disulfide-containing fragments. We discovered a number of hits, virtuallyall of which are negatively charged. The pyrazine fragment20 (Figure 13) was ofparticular interest: It was by far the strongest hit, and it demonstrated remarkablysharp SAR. In particular, changes to the cyclohexyl group or removal of the nega-tive charge completely abolished binding. To determine the origin of this SAR, wefirst grew crystals of the modified PTP-1B with the cleaved extender and soakedthese with the pyrazine disulfide. X-ray crystallography then revealed the preciseinteractions that give rise to the SAR (Figure 13). Particularly noteworthy is thefact that thecis-cyclohexyl group, which was seen to be critical, sits snugly withina hydrophobic pocket created by three protein residues. The acid substituent off thepyrazine is likewise involved in several hydrogen-bond interactions. In addition,we found that PTP-1B adopts the “open” conformation, in which the “WPD-loop”that normally packs down against the pTyr is open. Although this conformation hasbeen observed in a few other pTyr mimetic structures, it is relatively uncommonand suggests an additional mechanism to gain specificity.

Figure 13 Schematic of selected fragment20bound to PTP-1B (bb, backbone).



Inhibition studies revealed that the pyrazine-cyclohexyl moiety20possesses arelatively weak affinity (Ki = 4.1 mM), but this is actually comparable to pTyritself (Km = 4.9 mM) as well as other small pTyr mimetics that have subsequentlybeen optimized for affinity (3, 11, 54). Therefore, this novel fragment may be asuitable starting point for a medicinal chemistry effort.

TETHERING WITH EXTENDERS: TETHERING AS AFRAGMENT ASSEMBLY TOOL

In all of the preceding cases, we used Tethering to identify novel fragments. In thecase of IL-2, we identified an adaptive region of the protein and used some of thenewly identified fragments to enhance a known inhibitor. In the case of TS, weused structure-based design and medicinal chemistry to optimize the affinity of alow-affinity hit from Tethering. In the case of Tethering with breakaway extenderson PTP-1B, we identified a novel pTyr mimetic. But optimizing the binding affinityof a fragment that binds weakly is often not straightforward, particularly in theabsence of structural information. Ideally, we want to use Tethering not only toidentify fragments but also to incorporate them into a larger molecule to create ahigh-affinity inhibitor. To this end, we developed Tethering with extenders (20).Effectively, this transforms Tethering from a fragment discovery tool to a fragmentassembly tool.

Figure 14 shows a schematic of this approach. As in Tethering with break-away extenders, we first introduce an extender into the protein of interest; this isa molecule that (a) has inherent binding affinity for the target protein, (b) reactswith a specific cysteine (or other residue) in the protein of interest, and (c) con-tains a (masked) thiol. Fragments identified from an initial Tethering screen aregood candidates for converting into extenders. Whereas breakaway extenders werecleaved to remove the binding element of the extender, the present technique usesthe binding element as a springboard from which to probe for new fragments: The

Figure 14 Tethering with extenders to identify fragments in the context of a largermolecule.


TETHERING 215

whole assembly is screened against the disulfide library to identify hits that bindto the protein and also form a disulfide bond to the extender. In effect, this forcesthe selection of fragments that bind to the protein of interest in the presence of afirst binding element. Because the linkage length through the disulfide is known, asynthetic molecule containing binding elements from the extender as well as fromthe fragment can be rapidly constructed, in which the disulfide is replaced by amore stable, drug-like moiety. The resulting molecule should have higher affinityfor the target protein than do either of the component fragments alone.

Discovery of Nonpeptidic Caspase Inhibitors

In developing the approach, we first considered proteins that contain a nucleophilicactive-site cysteine. Such proteins often have known small-molecule inhibitors thatinteract with this cysteine and that could serve as starting points for extender design.Cysteine proteases are a large class of enzymes that contain a catalytic cysteineresidue, have known covalent inhibitors, are pharmaceutically relevant, and areideal test cases. The first example to which we applied this approach was theenzyme caspase-3, a member of the cysteine aspartyl proteasefamily. Caspase-3is a central executioner in the apoptosis pathway leading to programmed cell deathand has been investigated as a point of therapeutic intervention in diseases rangingfrom Alzheimer’s and Parkinson’s to myocardial infarction and sepsis (48, 77).

As do all members of the caspase family, caspase-3 has a linear peptide bind-ing site that neatly accommodates four amino acids residues; tetrapeptides havebeen used as starting points for small-molecule inhibitors (71). The most criticalrecognition element for all caspases is an aspartyl group immediately preced-ing the substrate amide bond: The aspartyl residue binds in a highly conservedpocket in the enzyme. Irreversible inhibitors have been designed by starting withaspartyl-containing peptides and replacing the C-terminal amide with a ketonefunctionalized with a good leaving group alpha to the carbonyl (18, 25). We usedthis template to design extender22, as shown in Figure 15. This extender con-tains the essential aspartyl group with a reactive arylacyloxymethylketone and acleavable thioester. We found that this molecule reacted rapidly and cleanly withcaspase-3, and the thioester was deprotected to give a modified protein that wasthen subjected to Tethering. By far the most prominent hit was the salicylic acidshown at the bottom of Figure 15.

Because we knew the molecular connectivity of the salicylic acid fragment tothe extender, we could readily synthesize an initial “diaphore” containing recogni-tion elements from the extender (in this case, the aspartyl group) and the salicylicacid fragment identified from Tethering. The disulfide linkage was replaced witha simple alkane linkage. Instead of the original irreversible arylacyloxymethylke-tone warhead, we used a reversible aldehyde warhead (46). The resulting molecule(compound23, Figure 16) inhibits caspase-3 with a low micromolar affinity. Sim-ple conversion of the flexible 6-amino-hexanoic acid linker to the more rigidpara-aminomethyl benzoic acid linker (compound24) increased the affinity by morethan one order of magnitude, and functionalization of the central aromatic moiety



Figure 15 Tethering with extenders on caspase-3.

with nitrogen picked up a hydrogen bond in the active site of the enzyme (com-pound25), adding another order of magnitude in increased affinity (12, 20).

We also enhanced the affinity of the initial diaphore by a complementary route.We noted that the simple alkane-linked diaphore could position an element intothe S2 binding site, and when we explored this possibility, we discovered thatit was possible to enhance affinity 30-fold simply by adding a phenyl group (2)(compound27, Figure 16).

Although we were able to rapidly generate two series of potent inhibitors froma common fragment, other approaches would likely have been ineffective. The


TETHERING 217

Figure 16 Evolution of a hit derived from Tethering with extenders to nanomolarleads.

simple salicylic acid fragment by itself (i.e., compound26) showed no inhibitionof the enzyme whatsoever, even at concentrations of several millimolar. Therefore,this fragment would not have been detected in a conventional screen and may havebeen missed even in an NMR screen.

To demonstrate that the approach is general, we made a second extender (ex-tender28), as shown in Figure 17. This extender has the same features as the firstextender (extender22), but it contains a phenyl sulfonamide linkage between the

Figure 17 Tethering with extenders on caspase-3 with a dif-ferent extender.



aspartyl group and the thioester and thus is expected to probe a different area of thecaspase. Consistent with these differences, Tethering did not pick up the salicylicacid fragment, but it did pick up a different fragment, a thiophene sulfone. Whenthis fragment was converted into a diaphore, the resulting compound29 inhibitedcaspase-3 with a potency in the high nanomolar range (20).

The crystal structures of both extender-modified enzymes complexed to theirrespective disulfide hits reveal that the salicylic acid and the thiophene sulfonebind in the same region of the enzyme, the S4 pocket, even though the link-ers take different trajectories to reach this pocket (20) (Figure 18). Also no-tably, the residue Tyr204 collapses into the S2 pocket in the case of extender22,whereas because of the bulk of extender28 the tyrosine swings back to accommo-date the extender. Residues in the S4 binding pocket also readjust slightly to ac-commodate the different moieties, revealing the adaptability of the protein surface.

Tethering with extenders was also applied to caspase-1, using the same ex-tenders, and a number of different fragments were identified. These were alsoconverted into diaphores, and in one case a single-digit nanomolar inhibitor wasrapidly identified (T. O’Brien, B.T. Fahr, M. Sopko, J.W. Lam, N.D. Waal, B.Raimundo, H. Purkey, P. Pham, & M.J. Romanowski, manuscript submitted).

Opportunities

The strength of Tethering with extenders is that it can be applied to any protein forwhich a small-molecule inhibitor is known. If the small-molecule inhibitor is co-valent, one simply needs to install a masked thiol onto the small-molecule inhibitorto generate an extender. Otherwise, it is straightforward to add a covalent warheadto a noncovalent inhibitor for Tethering and to remove it when diaphores are con-structed. As described here, this approach is well suited for enzymes that utilizeproteins as their substrates and therefore bind their substrates in extended formsusing fragment-sized subsites. These include proteases (27, 29, 30, 57, 59, 63), ki-nases (24, 74), and phosphatases (4, 28, 50), all classes of enzymes that are heavilypursued in the pharmaceutical industry. One of the most exciting uses of Tetheringis to identify a first hit, convert this hit to an extender, and then perform Tetheringwith extenders to identify a second hit, such that both fragments are derived fromTethering. This approach can be generalized to virtually any protein target.

CONCLUSIONS AND FUTURE DIRECTIONS

Fragment-based drug discovery offers great promise for providing new startingpoints for drug discovery as well as for facilitating lead optimization. Given theweak intrinsic affinities of many small fragments for their targets, it is crucial thatany fragment-based method selects for one-to-one binding to a specific site onthe protein. Tethering is well suited for this because the stoichiometry is measureddirectly by MS, and the site is localized to the cysteine used for Tethering. The disul-fide bond does not compromise the utility of the discovered fragments, because


TETHERING 219

the selection is under thermodynamic control and X-ray structures that comparethe disulfide-bonded and free fragments show them to be in close alignment.

At Sunesis we have applied Tethering to over a dozen different drug targets andhave produced a spectrum of useful fragments for each. The chemical informationthat Tethering provides can be applied to existing hits to generate higher-affinitycompounds, as was shown for IL-2. Tethering identifies nucleating fragments thatcan be grown into free-standing hits, as was shown for TS. Tethering can “mine”for novel warheads that engage the active sites of enzymes, as shown by the useof breakaway extenders in PTP-1B. And Tethering with extenders can allow theassembly of fragments into high-affinity molecules, as was shown for the caspases.

Several new avenues promise to further expand the utility of Tethering. BecauseTethering is a binding assay and does not rely on protein function, it is more ver-satile than many functional assays. For example, kinases exist in both inactive andactive forms in the cell, and these may have different conformations (33). Thoughsmall molecules that target one form over the other can achieve greater selectivity(66), functional screens are less able to identify molecules that bind selectively toinactive forms of the enzyme, a difficulty obviated by Tethering. Finally, Tetheringprovides a site-directed basis for discovery so that one can choose the sites to beexplored. This offers the possibility of selectively targeting and even identifyingallosteric (58) sites in proteins that may be missed in a functional screen. Suchallosteric sites may prefer different chemical features that could be advantageousto the highly charged fragments preferred by substrate sites in phosphatases andmany proteases. In aggregate, we believe fragment-based approaches will proveto be useful tools in drug discovery and optimization for novel and challengingprotein targets. Tethering is a powerful addition to the arsenal of fragment-baseddrug discovery approaches.

ACKNOWLEDGMENTS

We thank all our colleagues at Sunesis Pharmaceuticals for their research andsupport in making this work possible, and Monya L. Baker for editorial assistance.

The Annual Review of Biophysics and Biomolecular Structureis online athttp://biophys.annualreviews.org

LITERATURE CITED

1. Deleted in proof2. Allen DA, Pham P, Choong IC, Fahr B,

Burdett MT, et al. 2003. Identificationof potent and novel small-molecule in-hibitors of caspase-3.Bioorg. Med. Chem.Lett.13:3651–55

3. Andersen HS, Olsen OH, Iversen LF,

Sorensen AL, Mortensen SB, et al. 2002.Discovery and SAR of a novel selec-tive and orally bioavailable nonpeptideclassical competitive inhibitor class ofProtein Tyrosine Phosphatase 1B.J. Med.Chem.45:4443–59

4. Arabaci G, Xiao-Chuan G, Beebe KD,



Coggeshall KM, Pei D. 1999.α-haloacetophenone derivatives as photore-versible covalent inhibitors of protein ty-rosine phosphatases.J. Am. Chem. Soc.121:5085–86

5. Arkin MR, Randal M, DeLano WL, HydeJ, Luong TN, et al. 2003. Binding ofsmall molecules to an adaptive protein-protein interface.Proc. Natl. Acad. Sci.USA100:1603–8

6. Barrett WC, DeGnore JP, Konig S, FalesHM, Keng Y-F, et al. 1999. Regulation ofPTP1B via glutathionylation of the activesite cysteine 215.Biochemistry38:6699–705

7. Blundell TL, Jhoti H, Abell C. 2002.High-throughput crystallography for leaddiscovery in drug design.Nat. Rev. DrugDiscov.1:45–54

8. Boehm HJ, Boehringer M, Bur D, Gmuen-der H, Huber W, et al. 2000. Novel in-hibitors of DNA gyrase: 3D structurebased biased needle screening, hit vali-dation by biophysical methods, and 3Dguided optimization. A promising alterna-tive to random screening.J. Med. Chem.43:2664–74

8a. Bohacek RS, McMartin C, Guida WC.1996. The art and practice of structure-based drug design: a molecular modelingperspective.Med. Res. Rev.16:3–50

9. Braisted AC, Oslob JD, Delano WL, HydeJ, McDowell RS, et al. 2003. Discoveryof a potent small molecule IL-2 inhibitorthrough fragment assembly.J. Am. Chem.Soc.125:3714–15

10. Deleted in proof11. Burke TR Jr, Yao ZJ, Liu DG, Voigt J, Gao

Y. 2001. Phosphoryltyrosyl mimetics inthe design of peptide-based signal trans-duction inhibitors.Biopolymers60:32–44

12. Choong IC, Lew W, Lee D, Pham P, Bur-dett MT, et al. 2002. Identification ofpotent and selective small-molecule in-hibitors of caspase-3 through the use of ex-tended tethering and structure-based drugdesign.J. Med. Chem.45:5005–22

13. Cochran AG. 2001. Protein-protein inter-

faces: mimics and inhibitors.Curr. Opin.Chem. Biol.5:654–59

14. Cochran AG, Tong RT, Starovasnik MA,Park EJ, McDowell RS, et al. 2001. Aminimal peptide scaffold for beta-turndisplay: optimizing a strand position indisulfide-cyclized beta-hairpins.J. Am.Chem. Soc.123:625–32

15. Costi MP, Tondi D, Rinaldi M, BarloccoD, Pecorari P, et al. 2002. Structure-basedstudies on species-specific inhibition ofthymidylate synthase.Biochim. Biophys.Acta.1587:206–14

16. DeLano WL. 2003. PyMOL moleculargraphics system on the World Wide Web.http://pymol.sourceforge.net/

17. DeLano WL. 2002. Unraveling hot spotsin binding interfaces: progress and chal-lenges.Curr. Opin. Struct. Biol.12:14–20

18. Dolle RE, Hoyer D, Prasad CVC,Schmidt SJ, Helaszek CT, et al. 1994. P1aspartate-based peptideα-((2,6-dichloro-benzoyl)oxy)methyl ketones as potenttime-dependent inhibitors of interleukin-1b converting enzyme.J. Med. Chem.37:563–64

19. Erlanson DA, Braisted AC, Raphael DR,Randal M, Stroud RM, et al. 2000. Site-directed ligand discovery.Proc. Natl.Acad. Sci. USA97:9367–72

20. Erlanson DA, Lam JW, Wiesmann C, Lu-ong TN, Simmons RL, et al. 2003. Insitu assembly of enzyme inhibitors us-ing extended tethering.Nat. Biotechnol.21:308–14

21. Erlanson DA, McDowell RS, He MM,Randal M, Simmons RL, et al. 2003. Dis-covery of a new phosphotyrosine mimeticfor PTP1B using breakaway tethering.J.Am. Chem. Soc.125:5602–3

22. Fejzo J, Lepre C, Xie X. 2003. Applica-tion of NMR screening in drug discovery.Curr. Top. Med. Chem.3:81–97

23. Fejzo J, Lepre CA, Peng JW, Bemis GW,Ajay, et al. 1999. The SHAPES strategy:an NMR-based approach for lead genera-tion in drug discovery.Chem. Biol.6:755–69


TETHERING 221

24. Fry DW, Bridges AJ, Denny WA, Do-herty A, Greis KD, et al. 1998. Specific,irreversible inactivation of the epidermalgrowth factor receptor and erbB2, by anew class of tyrosine kinase inhibitor.Proc. Natl. Acad. Sci. USA95:12022–27

25. Garcia-Calvo M, Peterson EP, Leiting B,Ruel R, Nicholson D, Thornberry NA.1998. Inhibition of human caspases bypeptide-based and macromolecular in-hibitors.J. Biol. Chem.273:32608–13

26. Gilbert HF. 1995. Thiol/disulfide ex-change equilibria and disulfide bond sta-bility. Methods Enzymol.251:8–28

27. Graczyk PP. 1999. Caspase inhibitors—a chemist’s perspective.Restor. Neurol.Neurosci.14:1–23

28. Ham SW, Park J, Lee SJ, Yoo JS. 1999.Selective inactivation of protein tyrosinephosphatase PTP1B by sulfone analogueof naphthoquinone.Bioorg. Med. Chem.Lett.9:185–86

29. Han MS, Ryu CH, Chung SJ, Kim DH.2000. A novel strategy for designing ir-reversible inhibitors of metalloproteases:acetals as latent electrophiles that inter-act with catalytic nucleophile at the activesite.Org. Lett.2:3149–52

30. Hernandez AA, Roush WR. 2002. Re-cent advances in the synthesis, design andselection of cysteine protease inhibitors.Curr. Opin. Chem. Biol.6:459–65

31. Houk J, Whitesides GM. 1987. Structure-reactivity relations for thiol-disulfide in-terchange.J. Am. Chem. Soc.109:6825–36

32. Huc I, Lehn JM. 1997. Virtual combi-natorial libraries: dynamic generation ofmolecular and supramolecular diversityby self-assembly.Proc. Natl. Acad. Sci.USA94:2106–10

33. Huse M, Kuriyan J. 2002. The conforma-tional plasticity of protein kinases.Cell109:275–82

34. Hyatt DC, Maley F, Montfort WR. 1997.Use of strain in a stereospecific cat-alytic mechanism: crystal structures ofEs-

cherichia colithymidylate synthase boundto FdUMP and methylenetetrahydrofo-late.Biochemistry36:4585–94

35. Hyde J, Braisted AC, Randal M, ArkinMR. 2003. Discovery and characterizationof cooperative ligand binding in the adap-tive region of interleukin-2.Biochemistry42:6475–83

36. Jacoby E, Davies J, Blommers MJ. 2003.Design of small molecule libraries forNMR screening and other applications indrug discovery.Curr. Top. Med. Chem.3:11–23

37. Jahnke W, Florsheimer A, Blommers MJ,Paris CG, Heim J, et al. 2003. Second-siteNMR screening and linker design.Curr.Top. Med. Chem.3:69–80

38. Jencks WP. 1981. On the attribution andadditivity of binding energies.Proc. Natl.Acad. Sci. USA78:4046–50

39. Johnson TO, Ermolieff J, Jirousek MR.2002. Protein Tyrosine Phosphatase 1Binhibitors for diabetes.Nat. Rev. Drug.Discov.1:696–709

40. Kamb A, Finer-Moore J, Calvert AH,Stroud RM. 1992. Structural basisfor recognition of polyglutamyl folatesby thymidylate synthase.Biochemistry31:9883–90

41. Kaur S, McGuire L, Tang D, DollingerG, Huebner V. 1997. Affinity selectionand mass spectrometry-based strategies toidentify lead compounds in combinato-rial libraries. J. Protein Chem.16:505–11

42. Keire DA, Strauss E, Guo W, Nosz´al B,Rabenstein DL. 1992. Kinetics and equi-libria of thiol/disulfide interchange reac-tions of selected biological thiols and re-lated molecules with oxidized glutathione.J. Org. Chem.57:123–27

43. Lehn JM, Eliseev A. 2001. Dynamic com-binatorial chemistry.Science291:2331–32

44. Lesuisse D, Lange G, Deprez P, BenardD, Schoot B, et al. 2002. SAR and X-ray.A new approach combining fragment-based screening and rational drug design:



application to the discovery of nanomo-lar inhibitors of SRC SH2.J. Med. Chem.45:2379–87

45. Liu G, Szczepankiewicz BG, Pei Z,Janowick DA, Xin Z, et al. 2003. Dis-covery and structure-activity relationshipof oxalylarylaminobenzoic acids as in-hibitors of Protein Tyrosine Phosphatase1B. J. Med. Chem.46:2093–103

46. Margolin N, Raybuck SA, Wilson K,Chen W, Fox T, et al. 1997. Substrateand inhibitor specificity of interleukin-1β-converting enzyme and related caspases.J. Biol. Chem.272:7223–28

47. Marsham PR, Wardleworth JM, BoyleFT, Hennequin LF, Kimbell R, et al.1999. Design and synthesis of potent non-polyglutamatable quinazoline antifolatethymidylate synthase inhibitors.J. Med.Chem.42:3809–20

48. McBride CB, McPhail LT, Steeves JD.1999. Emerging therapeutic targets incaspase-dependent disease.Expert Opin.Ther. Targets3:391–411

49. McGovern SL, Caselli E, Grigorieff N,Shoichet BK. 2002. A common mecha-nism underlying promiscuous inhibitorsfrom virtual and high-throughput screen-ing. J. Med. Chem.45:1712–22

50. Myers JK, Widlanski TS. 1993. Mecha-nism-based inactivation of prostatic acidphosphatase.Science262:1451–53

51. Nakamura CE, Abeles RH. 1985. Modeof interaction of beta-hydroxy-beta-methylglutaryl coenzyme A reductasewith strong binding inhibitors: compactinand related compounds.Biochemistry24:1364–76

52. Nelson BH, Willerford DM. 1998. Biol-ogy of the interleukin-2 receptor.Adv. Im-munol.70:1–81

53. Nienaber VL, Richardson PL, KlighoferV, Bouska JJ, Giranda VL, Greer J. 2000.Discovering novel ligands for macro-molecules using X-ray crystallographicscreening.Nat. Biotechnol.18:1105–8

54. Niimi T, Orita-Igarashi M, Okazawa M,Sakashita H, Kikuchi K, et al. 2001. De-

sign and synthesis of non-peptidic in-hibitors for the syk C-terminal SH2 do-main based on structure-based in-silicoscreening.J. Med. Chem.44:4737–40

55. Obita T, Muto T, Endo T, Kohda D. 2003.Peptide library approach with a disulfidetether to refine the Tom20 recognition mo-tif in mitochondrial presequences.J. Mol.Biol. 328:495–504

56. O’Shea EK, Rutkowski R, Stafford WFIII, Kim PS. 1989. Preferential hetero-dimer formation by isolated leucine zip-pers from fos and jun.Science245:646–48

57. Otto HH, Schirmeister T. 1997. Cysteineproteases and their inhibitors.Chem. Rev.97:133–72

58. Pargellis C, Tong L, Churchill L, Cir-illo PF, Gilmore T, et al. 2002. Inhibitionof p38 MAP kinase by utilizing a novelallosteric binding site.Nat. Struct. Biol.9:268–72

59. Powers JC, Asgian JL, Ekici OD, JamesKE. 2002. Irreversible inhibitors of serine,cysteine, and threonine proteases.Chem.Rev.102:4639–750

60. Ramstrom O, Lehn JM. 2002. Drugdiscovery by dynamic combinatorial li-braries.Nat. Rev. Drug Discov.1:26–32

61. Rishton GM. 2003. Nonleadlikeness andleadlikeness in biochemical screening.Drug Discov. Today8:86–96

62. Salmeen A, Andersen JN, Myers MP,Tonks NK, Barford D. 2000. Molecularbasis for the dephosphorylation of the ac-tivation segment of the insulin receptorby Protein Tyrosine Phosphatase 1B.Mol.Cell 6:1401–12

63. Salto R, Babe LM, Li J, Rose JR, Yu Z,et al. 1994. In vitro characterization ofnonpeptide irreversible inhibitors of HIVproteases.J. Biol. Chem.269:10691–98

64. Sarmiento M, Puius YA, Vetter SW, KengYF, Wu L, et al. 2000. Structural basis ofplasticity in Protein Tyrosine Phosphatase1B substrate recognition.Biochemistry39:8171–79

65. Sauve K, Nachman M, Spence C, Bailon


TETHERING 223

P, Campbell E, et al. 1991. Localization inhuman interleukin 2 of the binding site tothe alpha chain (p55) of the interleukin2 receptor.Proc. Natl. Acad. Sci. USA88:4636–40

66. Schindler T, Bornmann W, Pellicena P,Miller WT, Clarkson B, Kuriyan J. 2000.Structural mechanism for STI-571 inhibi-tion of abelson tyrosine kinase.Science289:1938–42

67. Shuker SB, Hajduk PJ, Meadows RP, Fe-sik SW. 1996. Discovering high-affinityligands for proteins: SAR by NMR.Sci-ence274:1531–34

68. Stanojevic D, Verdine GL. 1995. De-construction of GCN4/GCRE into amonomeric peptide-DNA complex.Nat.Struct. Biol.2:450–57

69. Stout TJ, Sage CR, Stroud RM. 1998.The additivity of substrate fragments inenzyme-ligand binding.Structure6:839–48

70. Swayze EE, Jefferson EA, Sannes-Lowery KA, Blyn LB, Risen LM, et al.2002. SAR by MS: a ligand based tech-nique for drug lead discovery againststructured RNA targets.J. Med. Chem.45:3816–19

70a. Thanos CD, Randal M, Wells JA. 2003.Potent small-molecule binding to a dy-namic hot spot on IL-2.J. Am. Chem. Soc.125:15280–81

71. Thornberry NA, Rano TA, Peterson EP,Rasper DM, Timkey T, et al. 1997. Acombinatorial approach defines specifici-ties of members of the caspase family andgranzyme B.J. Biol. Chem.272:17907–11

72. Tilley JW, Chen L, Fry DC, Emerson SD,Powers GD, et al. 1997. Identification ofa small molecule inhibitor of the IL-2/IL-

2Rα receptor interaction which binds toIL-2. J. Am. Chem. Soc.119:7589–90

73. Toogood PL. 2002. Inhibition of protein-protein association by small molecules:approaches and progress.J. Med. Chem.45:1543–58

74. Tsou HR, Mamuya N, Johnson BD, ReichMF, Gruber BC, et al. 2001. 6-substituted-4-(3-bromophenylamino)quinazolines asputative irreversible inhibitors of the epi-dermal growth factor receptor (EGFR)and human epidermal growth factor re-ceptor (HER-2) tyrosine kinases with en-hanced antitumor activity.J. Med. Chem.44:2719–34

75. van Dongen M, Weigelt J, Uppenberg J,Schultz J, Wikstrom M. 2002. Structure-based screening and design in drug dis-covery.Drug Discov. Today7:471–78

76. van Dongen MJ, Uppenberg J, SvenssonS, Lundback T, Akerud T, et al. 2002.Structure-based screening as applied tohuman FABP4: a highly efficient alter-native to HTS for hit generation.J. Am.Chem. Soc.124:11874–80

77. Yue T-L, Ohlstein E, Ruffolo R. 1999.Apoptosis: a potential target for discov-ering novel therapies for cardiovasculardiseases.Curr. Opin. Chem. Biol.3:474–80

78. Zhang ZY. 2002. Protein tyrosine phos-phatases: structure and function, substratespecificity, and inhibitor development.Annu. Rev. Pharmacol. Toxicol.42:209–34

79. Zurawski SM, Vega F Jr, Doyle EL,Huyghe B, Flaherty K, et al. 1993. Def-inition and spatial location of mouseinterleukin-2 residues that interact withits heterotrimeric receptor.EMBO J.12:5113–19

TETHERING C-1

Figure 4 (A) X-ray structure of the complex between IL-2 and 1. Residues estab-lished as important for IL-2Rα binding are labeled and colored orange. E62 forms acritical bidentate salt bridge and the adaptive hydrophobic pocket is created aroundresidues F42, L72, K35, R38, and K76. (B) Ribbon diagram of IL-2 that shows thesites chosen for cysteine mutations. Individual mutants were made targeting the bind-ing site of 1. This and all color figures were made using the program PyMOL (16).

Erlanson.qxd 4/30/2004 7:55 PM Page 1

C-2 ERLANSON ■ WELLS ■ BRAISTED

Figure 5 (A) X-ray structure of the disulfide adduct between IL-2 Y31C and 2. Theside chain of F42 is in a conformation similar to that seen in wild-type IL-2; this con-formation would block binding of 1. (B) Overlay of three crystal structures, 1 (yel-low), 2 (green) and 3 (blue), demonstrating the overlap between fragment 2 and com-pound 1 compared with the bound conformation of the compound designed throughfragment merging, 3.


TETHERING C-3

Figure 7 Stereoview showing the crystallographically determined structure of TS boundto N-tosyl-D-proline without the disulfide linker (compound 8) (blue) superimposed withthe cofactor mTHF (yellow) (34).

Figure 8 Overlay of three structures: N-tosyl-D-proline bound to TS noncovalently(green), covalently through C146 (red) or covalently through L143C (blue).


C-4 ERLANSON ■ WELLS ■ BRAISTED

Figure 10 Superposition of two crystallographically determined structures:Compound 15 (blue) and mTHF (yellow)/dUMP (yellow) bound to TS (green).Hydrogen bonds are shown as yellow dashed lines with distances in Ångstroms.

Figure 18 Overlay of caspase-3 modified with either extender 22 (gray) or exten-der 28 (blue) and their selected disulfide fragments. The thick cylinders are theextenders and selected fragments, the thin cylinders are protein residues. Reprintedwith permission from Nature Biotechnology (20).


Documents

TETHERING: Fragment-Based Drug Discovery