lead Discovery and Optimization of Thrombinand Factor Xa inhibitors, C5a Receptor

Antagonists, and Anti-microbial Agents using theOne-Bead-One-Compound (Selectide) Process

Bruce Seligmann1, Richard Kris, Serene Josiah, Marianne Wildgoose1, Steve Baum,Christine Sizemore, Sydney Wilson, Steve Felder, Jim Ostrem, Charlie Chen, PeterWildgoose1, David Thorpe, Olga Isakova, Farid Abdul-Latif. Fahad Al-Obeidi, Alena

Safarova, Pavel Safar, Viktor Krchnak, Zuzana Flegelova, Ken Wertman, NikolaiSepetov, Michal Lebl, Peter Strop

1Hoechst, Marion, Roussel, 1Cincinnati, OH 45215, U.S.A. and Tucson, AZ 85737, U.S.A.

Graphical Abstract

LEAD DISCOVERY AND OPTIMIZATION OF THROMBIN AND FACTOR XaINHIBITORS, C5a RECEPTOR ANTAGONISTS, AND ANTI-MICROBIAL AGENTSUSING THE ONE-BEAD-ONE-COMPOUND (SELECTIDE) PROCESSBruce Seligmann1, Richard Kris, Serene Josiah, Marianne Wildgoose1, Steve Baum, ChristineSizemore, Sydney Wilson, Steve Felder, Jim Ostrem, Charlie Chen, Peter Wildgoose 1, DavidThorpe,Olga Isakova, Farid Abdul-Latif. Fahad Al-Obeidi, Alena Safarova, Pavel Safar, ViktorKrchnak, Zuzana Flegelova, Ken Wertman, Nikolai Sepetov, Michal Lebl, Peter Strop1Hoechst, Marion, Roussel, 1Cincinnati, OH 45215, U.S.A. and Tucson, AZ 85737, U.S.A.This manuacript describes the application of the one-bead-one-compound strategy for thesynthesis of combinatorial libraries to on-bead and solution screening assays to identify andoptimize leads which inhibit thrombin (1), factor Xa (2), and bacterial growth, and antagonize theaction of C5a. -

(1) phe-Pro-Arg-Pro-Phe-Gly-Tyr-Arg-Val-NH2(2) Tyr-lle-Arg-Leu-Ala-Ala-Phe-Thr-NH2(2) tyr-lle-Arq-Leu-Pro-Nl-le(2) Ac-pAph-Chg-Pal(3)(Me)-Leu-Pro-NH2

NH

J-.et~JQH 0 N.Y-H 0 H Y CONH2b SEL271l(2)

Abstract

The one-bead-one-compound strategy of synthesizing and screening combinatoriai libraries wasused to identify and optimize leads for a number of targets which are of therapeutic interest.Three screening assay formats were used. An on-bead binding assay where soluble targetmolecules bind to and identify beads containing active compounds was used to discover andoptimize thrombin and factor Xa enzyme inhibitors, A single release agar diffusion assay wasused to identify antimicrobial agents in a functional growth inhibition assay. A double releaseradiolabeled membrane binding assay and functional whole cell calcium rise assay was used toidentify antagonists of a seven trans-membrane spanning G-protein coupled receptor, the C5areceptor. These data demonstrate that many classes of targets which are of pharmaceuticalinterest can be pursued using the one-bead-one-compound combinatorial chemistry drugdiscovery strategy.

Introduction

In 1991 Kit Lam published a seminal paper in the field of combinatorial chemistry (1). The solidphase synthesis split and mix method of assuring that every building block coupled completeiyduring the synthesis of mixtures of peptides had already been described by Furka (2). Lamrecognized that by retaining the relationship between each compound and the solid phaseparticle on which it had been synthesized the split and mix process could be used to produceuseful Iibraries in a one-bead-one-compound format from which information on the solid phaseparticle"could be used to identify positive compounds. This obviated the need to physically trackthe synthesis of each compound in order to determine identity, as well as obviating the need toidentify active compounds from mixtures using some method of target-dependent enrichment.This publication by Lam took a simple synthetic method to assure that every compound in amixture was synthesized (split and mix synthesis) to a combinatorial chemistry drug discoverystrategy which has revolutionized the pharmaceuticai industry. Selectide was founded based onthis concept, and the process was identified as the "Selectide Process". The proliferation ofcompanies which utilize solid phase synthetic strategies to produce libraries of compounds in theone-bead-one-compound format attests to the power of this approach.

There are two basic screening approaches using the one-bead-one-compound strategy. In one,the solid phase particles are directly assayed with the compounds remaining attached to thebeads, using soluble targets and a detection system which labels positive beads (Figure 1) (3, 4).The positive beads are then selected, and the identity of the compound on each determineddirectly or from the chemical information on the bead. In the second approach the testcompounds are released from the solid particles in such a manner that the bead of origin foreach released compound is known or can be subsequently determined (Figure 2) (5, 6). Thereleased compounds are tested in traditional solution assays, and the beads corresponding toeach positive compound are recovered and the identity of each compound determined frominformation remaining on the solid phase particle. This latter synthesis and assay strategy was amajor breakthrough in the practical utilization of combinatorial libraries of discrete compounds,since it represented a very simple strategy to synthesize libraries of discrete compounds (one perparticle) in a format which enabled highly efficient screening in traditional assays in order toidentify active compounds in just one or two steps. This approach opened up many classes oftargets to the combinatortal chemistry approach, and redefined the concept of high throughputscreening. Thus, the one-bead-one-compound strategy enables the efficient, simple synthesis ofcomplex libraries containing millions of individually separable compounds, one per solid phaseparticle, and high throughput screening, complementing the synthetic capacity of the process.Technical aspects of this process were recently reviewed (4, 6 to 10). This manuscript will focuson the application of the one-bead-one-compound strategy to combinatorial chemistry driven leadgeneration and optimization.

The issue for the future is whether the one-bead-one-compound strategy will become the methodof choice for the synthesis and screening of non-peptide libraries. Certainly as organic reactionsare reduced to practice on solid phase the process can be applied. However, the requirement foran anaiytical method to directly determine the identity of active compounds, or for orthogonalchemistry by which to synthesize a tag, as well as the requirement for a synthetically inertmethod for releasing compound from the solid phase particles all represent hurdles to theapplication of the one-bead-one-compound strategy to the synthesis of non-peptide libraries.Coding, for example the process of coding via electroferric tags described by Clark Still (13, 14)and used by Pharmacopeia (15, 16), seems to be a promising approach to address this issue.However, it may be that for the interim other combinatorial synthetic and screening strategies (fora review of alternative techniques, see e.g. references 9 to12) will help fulfill the needs of drugdiscovery until the non-peptide organic solid phase reactions become sufficiently facile to permitroutine use of the one-bead-one-compound strategy.

Materials and Methods

Reagents and Materials

Thrombin was purchased from Enzyme Research Laboratories (South Bend, IN), human plasmaderived factor Xa was obtained from Dr. Waiter Kisiel (Univ. of New Mexico) and EnzymeResearch Laboratories, 5-bromo-4-chloro-3-indolyl phosphate (BCIP), biotin and alkalinephosphatase labeled streptavldin {SAP} were purchased from Pierce {Rockford, IL}. BCIP bufferfor dissolving the BCIP substrate included 50 mM Tris base, 50 mM NaCl, 1.2 mM MgCI2' pH8.4. BCIP substrate assay buffer contained 50 mM Tris-HCI (pH 7.8), 200 mM NaCI, 0.5% PEG­8000 (w/v), and 0.01 % NaN3 {w/v}. Phosphate buffered saline (PBS) was made in the laboratory(137 mM NaCI, 2.7 mM KCI, 4.3 mM Na2HP04, 1.4 nM KH2P04, pH 7.2), and fortified withGelatin (0.05% w/v) as needed to make a PBS-Gelatin buffer, or with Tween-20 {O.1 % v/v} tomake a PBS-Tween buffer. High salt binding buffer {HSBB} contained 50 mM Tris-HCI (pH 7.8),0.8 M NaCl, 0.05% PEG-8000 {w/v}, 0.1 mg/ml bovine serum albumin, 0.25% Tween-20 (v/v),and 0.02% NaN3 (w/v). Guanidine hydrochloride {GuHCl, Sigma} was made up as a 6Msolution, pH 1.0. The thrombin substrate S2366 and the factor Xa substrate S2765 werepurchased from Kabl {Sweeden}. The C5a receptor was obtained from Dr. Craig Gerard{Harvard}, sub cloned into pCDNA1-AMP and used to transfect K562 cells line {producing aK562A42 C5aR+ sub clone, expressing -100,000 receptors per cell}. This produced functionalreceptor which bound C5a with high affinity (Kd 7 pM), and a transfected cell phenotype whichexhibited a rise In calcium upon stimulation with C5a. Recombinant human C5a was purchasedfrom Sigma, and 1251_C5a was purchased from New England Nuclear. Agarose LE waspurchased from FMC BioProducts {Rockland, ME}, and bacterial strains {Staphylococcus aureus,ATCC#25923, and Pseudomonas aeruginosa, ATCC#10145} were obtained from the AmericanType Culture Collection.

Synthesis of libraries on solid phase

The synthesis of libraries of compounds for on-bead screening was carried out essentially asdescribed previously {1, 3, 4, 17}, using polyethyiene giycol-grafted polystyrene (PEG-PS) resinbeads including TentaGel (90 and 130 urn, Rapp Poiymere, Tublnpen, Germany) with asubstitution of reactive sites of 0.23 mmol/g and high substitution PEG-PS resin (130 urn) of 0.9mmol/g (500 pmol/bead), as well as 220 urn beads with a substitution of 0.45 mmoi/g (-4500pmol/bead), purchased from Miilipore. The synthesis of libraries of compounds for screening in asingle or double release assay was performed using a double cieavabie iminodiacetic aciddipeptide-based linker which provides three carboxyl groups, two for points of compoundsynthesis and one for attachment to the solid phase resin, essentially as previously described {5,6, 17, 18}. Once the Boc-protected amino group is deprotected this structure is unstable atneutral pH in aqueous buffer, forming a diketopiperazine ring and reieasing the compoundattached to one of the carboxyl groups of the linker via the ester bond. The compoundsynthesized on the remaining carboxyl group can be cleaved by alkaline hydrolysis at basic pH(5,6). In the specific case of the C5a example the linker was modified in such a way thatcleavage left the free carboxyl group. This modification was used because the Gly-NH{CH2)30H"tail" common to the compounds released from the specific linker described in the Kocispublication (17) attenuated the activity of the C5a peptides anywhere from between 50 and 1000­fold (results not shown). Libraries were stored dry until immediately before use, and thensuspended and handled in 25 mM HCI to prevent any cleavage and loss of compound before therelease step in the assay protocol.

On-Bead screening assay

" '

The method used was a modification of that previously published (1, 3, 4). The target molecule(thrombin, Factor Xa) was labeled with biotin and then conjugated in distilled H20 for 30 to 45min with alkaline phosphatase labeled streptavidin (SAP) to make a stock Target/SAP solution20-fold higher concentration then used in the incubation with library beads. Library beads werewashed three times with 1/4th strength HSBB-T, then blocked using a one hour incubation in1/4th strength HSBB-T plus Gelatin (0.5 gn), and then washed again with 114th strength HSBB-T.PBS-Gelatin or 114th strength HSBB-T plus CaCI2 were used to dilute the Target/SAP complexto the final volume necessary, beads added, and the mixture incubated 1 to 2 hr on the bench topwith gentle agitation. The beads were washed three times with PBS-Tween or 1/4th strengthHSBB"T, and then washed three times with BCIP buffer. The washed beads were incubated withBCIP substrate (BCIP was dissolved at 50 mg/ml in DMF, then 33 ul of this solution was addedto 10 ml of BCIP buffer supplemented with 0.5 ml of a 10% v/v Tween-20 solution) for 0.5 to 1 hr.The color development was stopped by washing the beads several times with 1/4th strengthHSBB-T. The beads which were colorized were selected by aspiration into the pipette of amanual micropipettor while observing the beads under a dissecting microscope. In someinstances stained beads were de-stained and then re-stained, The de-staining procedureinvolved sequential washes of the beads with GuHCI followed by a wash in dimethylformamide(DMF).

Single release assay

A variation of a standard agar-based diffusion assay was used to detect anti microbial activity ofcompounds released from library beads (19). Dry library beads were suspended in 500 ftl of 25mM HCI in a 100 mm petri dish. Melled agarose growth medium (5 ml of an 0.8% molten broth­agarose, inoculated with the test organism, Staphylococcus aureus) was poured over the beadswhich were evenly distributed by gentle swirling. Once the agar had solidified (-5 min) the petridishes were incubated under optimal conditions for the growth of the microorganism. The pH ofthe medium was 7.2, which caused cleavage of compound from the beads. Compound diffusedfrom the beads, creating a symmetrically round diffusion gradient zone around the bead of origin.In the case of compounds which inhibited the growth of microorganisms the zone containing thecompound remained clear of growth. The beads in the center of these clear zones of inhibitionwere harvested and the identity of the anti microbial compound was determined. The MinimalInhibitory Concentration (MIG) was carried out in 96 well microtiter plates in liqUid broth, Eachwell was seeded with 50 ftl of llquid broth seeded with the test organism. Two-fold serial dilutionsof test compound (50 ftl volume) were added to consecutive wells. The plates were incubated at370 C for -14 hr and the optical density at 570 nm determined in a microplate absorbancereader. The lowest concentration of compound resulting in no visible growth was determied to bethe MHC.

Double release assay

The method used was a modification of that previously published (5, 6). Libraries were storeddry until use and then suspended in 25 mM HCI to prevent any cleavage and release ofcompound from the library beads. To determine the number of beads to pool for the first releasea preliminary background test was carried out. Beads from the library were dispensed into wellsof a filter plate in increasing numbers (10, 25, 50, 75, 100, 150, ...500), suspended in 25 mMHC1/0.01 % Tween-20. The beads were then washed two times with water (50ftl aliquots), twotimes with 0.1M HEPES to neutralize any remaining HCI, and then washed one time with water toremove residual HEPES before adding 50 ftl of assay buffer, pH 7.2. After 12 hrthe releasatewas filtered into a 96-well round bottom (low binding) microtiter plate. The beads were washedonce with 50 ftl assay buffer, and the wash solution also collected to give a total volume of -100ul of releasate for testing. The effect of releasate on binding was then determined using the C5areceptor K562 membrane binding assay described below. An effect of greater than 20% •

'J c

inhibition was considered too high of a background, and the highest number of beads giving lesstha!),!,his level of inhibition was set as the maximum pODI size allowable for that particular library ofcompounds. Beads were then dispensed into master filter plates at the selected poolinq limit. inthis case 50 beads per well, and the first release performed as described above. Following thecollection of the releasate the beads in the master filter plate were washed twice with water. thenonce with 50 mM HCI. and then suspended in 50 f.ll of 50 mM HCI for storage at 40C in a sealed(hydrated) bag. This procedure prevented any release of the remaining compound on the beads.In experiments where a second release was carried out the beads from the appropriate wells ofthe first master filter plate were recovered by repeated aspiration and dispensing into aneppendorf tube using water as the carrier solution. Once all beads were recovered the collectionwas washed two times by centrifugation in water, and then redistributed. one bead per well. into asecond master filter plate into a solution of 25 mM HCI/O.01 % Tween-20. The same washeswere carried out as for the first release. The second release was then carried out by the additionof 50 f.ll of a 50 mM NaOH (pH 13) solution to each well and 1 hr incubation. or conducted in thedry state by gas phase ammonia vapors. In the case of solution phase NaOH. the solution in thefilter plate was neutralized by the addition of 25 f.ll of an HCI neutralizing solution (stockneutralizing solution consisting of equal volumes of 200 mM HCI and 5-fold strength assaybuffer). and then the releasate was collected into a 96 well round bottom (low binding) microtiterplates, and washed once with 50 ul of assay buffer to give a total volume of approximately 125 ulreleasate for use in assay and compound identification.

G5a receptor K562 membrane binding assay

Membranes from K562 cells transfected with the C5a receptor (K562A42 sub clone) and grown ina 1 liter spinner flask were prepared by washing in 20 mllysis buffer (see below) withcentrifugation (5 min. 800g. 4°C). re suspending the pellet in 75 ml of lysis buffer. and after a 5min incubation to permit swelling. disrupting the cells in a Dounce homogenizer (type B glasspestle. 25 strokes) while suspended in a lysing buffer containing 50 mM Tris-Cl, pH 8.0. 1 mMEGTA, 5 mM MgCI2 and a cocktail of protease inhibitors (50 f.l9/ml l.eu-peptin, 5 f.lg/ml PepstatinA. 10 f.lg/ml Aprotenin, 200 f.lg/ml Bacitracin and 1 mM PMSF). Disruption was confirmedmicroscopically. The disrupted preparation was centrifuged for 5 min at 800g. 40C to removeany unbroken cells. The pellet from this centrifugation was re homogenized in 20 mllysis buffer.centrifuged again and the supernatants pooled. The polled supernatant containing the cellmembranes was centrifuged for 30 min at 20,OOOg. 40C to pellet the membranes. which werethen re suspended in 10 ml HEPES-EGTA buffer (20 mM HEPES-KOH. pH 8.0.1 mM EGTA. 5mM MgCI2' 30% Glycerol, plus protease cocktail). The membrane concentration was determinedusing a Bradford assay. the membranes were diluted to produce a stock concentration of 1 f.lg/f.ll.aliquot in volumes of 1. 0.5. and 0.1 mi. quick frozen in EtOH/Dry ice. and stored at -BOoC untiluse. The radio label membrane binding assay was carried DUt in 96 well low binding microtiterplates using these membranes as the source of C5a receptor, Membranes were added at aconcentration of 1ug/well, brought to a volume of 30 ul, and incubated at room temperature for 1hr. Assay buffer or releasate (50 ul) was added to each well and incubated one hour. Thebinding competition was initiated by rapidly adding 20 ul of 1251-C5a (7pM final concentration) toeach well. After a 5 min incubation the membranes were harvested. Glass fiber-C UNI­filterplates (Packard) which had been treated with PEl (25 f.ll of a 0.33% solution) for 10 min. andwashed three times with 50 mM Tris-CI (pH 7.4), were used tD harvest the membranes. Thesolution (membranes pius reagents) in the microtiter plate incubations were harvested onto thefilter plates and washed four times with 50 mM Tris-Cl using a Packard Filtermate 196 cellharvester. All tests of releasate were carried out in duplicate. and each well contained a measureof total binding (no competitor). nonspecific binding (no membranes) and a competitive bindingcontrol (agonist peptide SEL 2289). The filter plates were dried in a hood for 1 hr. the bottom ofeach plate was sealed. 50 ul of Microscint-20 scintillation cocktail was added to each well. rhe top

!i i 1

of the plates were sealed, and the plates were counted on a Packard Topcount microplatecounter, 1 min/well.

G5a mediated K562 Intracellular calcium rise assay

The agonist/antagonist activity of compounds was determined by assessing both their ability toelicit a calcium rise from transfected K562A42 celis and their ability to inhibit the response elicitedby recombinant human G5a. The K562a42 celis were loaded with 2 11M Fura2-AM essentialiy asdescribed previously (20). The Fura2 calcium sensitive fluorescence ratio of 340/380 nmexcitation was used to monitor changes in intraceliular calcium levels foliowing stimulation. Celiswere maintained in a suspension in a curette using a magnetic stir bar. Measurements weremade using a SPEX slewing monichromator spectrofluorometer. 50 pM C5a was used as toproduce an approximately 50% maximal response as a positive control and as a chalienge forevaluating antagonist efficacy.

(,

Results

Lead generation and optimization using an On-Bead assay: discovery and optimization ofthrombin inhibitors

Pentapeptide libraries containing the 19 naturaily occuring (L) amino acids and their D-isomersynthetic analog~ere synthesized in order to identify novel pentapeptide thrombin inhibitors.Peptides which bbund thrombin were selected in the on-bead binding assay as described above,identified by Edman degradation and amino acid sequencing, and then were synthesized (as theamide or C-terminal glycine-amide adduct) and tested for inhibition of thrombin enzyme activity.The results are shown in Table 1. Based on the published literature the sequence fPRP wasselected as the basis for the synthesis of a secondary optimization library of the composition phe­Pro-Arq-Pro-Xxxxx where the X's represent the points of randomization using 19 L-amino acids.Since fPRP itself exhibited a Ki of 20 J.1M, the stringency of the on-bead binding assay wasincreased by using a lower concentration of thrombin. From this assay a 25 nM inhibitor ofthrombin was identified, SEL 1172, resynthesized and tested as SEL 1478 (phe-Pro-Arq-Pro­Phe-Gly-Tyr-Arq-Vat-Nl-lp) without the linker. This compares very favorable with Hirulog (phe­Pro-Arg-Pro-Gly-Gly-Gly-Gly-N-Gly-Asp-Phe-Glu-Glu-ile-Pro-Glu-Glu-Tyr-Leu), a 2 nM inhibitorin clinical trials as an antithrombotic agent. A simple evaluation of structure activity relationshipwas periormed by the synthesis of individual compounds in which each new residj3 was replacedwith alanine, or the parent structure was progressively truncated from the C-termihal end with theresult shown in Table 2.

Lead generation using an On-Bead assay: discovery of a Factor Xa inhibitor

The on-bead binding assay also was used to screen compounds from peptide, peptidomimetic,end non-peptide libraries for inhibition of Factor Xa. Screening octamer peptide librariescontaining the naturaily occuring L-amino acids produced a family of compounds, the most activeof which was SEL 1691, with a Ki of 15 !J.M. This was a novel structure (Tyr-ile-Arg-Leu-Ala-Ala­Phe-Thr-NH2) containing a novel motif not previously associated with serine proteases, Tyr-ile­Arg. Optimization pentamer libraries based on this structure but incorporating unnatural aminoacids produced SEL 2060 (tyr-lle-Arp-Leu-Pro-Nl-ly) with a Ki of 150 nM as the most activeanalog, containing a 0 rather than L-tyrosine. However, the library efforts also demonstratedadditional unnatural amino acid substitutions which permited an integrated medicinal chemistryeffort to produce an orally active inhibitor, SEL 2711 (Figure 3). Thisstructure is acetyl-p-

!Iv fVQI .~idinoyihenY~lanYljtYcloheXYI~IYCYI-3-~ethy(pYridiniumlAlanYI~ucYI;Proline-amide,'I 0 abbreviated as Ac-pAph-Chg-Pal(3)Me-Leu-Pro-NH2' The molecule containing ail L-isomers,

SEL 2711, exhibited a Ki of 3 nM for inhibition of Factor Xa, and greater than 3000-fold selectivitycompared to other enzymes (thrombin, plasmin, protein C, Factor Vila, trypsin, etc.).

Lead generation using a Single Release solution assay: assessment of anti microbial activity

Because of the novel aspect of this assay using pH-dependent release of compound in growthmedium, it was validated using a series of peptide analogs of an anti-microbial peptide describedby Houghten, Ac-Arq-Arq-Trp-Trp-Cys-Arq-Nl-l-, (21,22). These compounds were synthesizedon TentaGel 130 J.1m beads on the double cleavable iminodiacetic acid linker with the Gly­NH(CH2)30H adduct (100 pmol/release) and used in the agar diffusion screening assay testedagainst Staphylococcus aureus. Zones of inhibition after 12 hr incubation at 370 C werecorrelated with the activity of the same compounds released and tested in a solution phasegrowth inhibition assay to determine the liquid Minimai Inhibition Concentration (MIC). Theresults are summarized in Table 3, demonstrating that compounds with MHC values of up to 320J.1g/ml were detected in the agar diffusion assay. A peptapeptide library was prepared using 0

amino acids and unnatural amino acids. From this library additional analogs of the Houghtencompound were identified. as summarized in Table 4.

Lead optimization using a Doubie Release solution assay: identification of G5a receptorantagonists using membrane binding and calcium rise functional screens

Scientists at Abbott pursued the C-terminal residues of CSa in an attempt to identify anantagonist. Their efforts resulted in identification of several potent C-terminal peptide analogaqonists containing unnatural amino acids. Phe-Lys-Ala-cha-Cha-arg-OH (lCso binding 70 nM)(23). and the (Me)PheoLys-Pro-cha-Cha-arg-OH (ICSO binding 10 nM) (24. 2S). A secondaryoptimization Iibrarypy'flexapeptides was based on the originally published agonist modified withproline at the third/position. This library was based on the concept of reacting 20% of the resin ateach posltion.wlth the amino acid of the parent, and splitting the remaining 80% of the resin intoaliquots to.react individually with a set of natural and unnatural amino acid analogs. For example.at the fir,stposition 20% of the resin was coupled to Phe, and the remaining 80 % split andreacted in individual aliquots with 12 natural and unnatural amino acid analogs. The resin wasthoJ6ughly mixed after each step of synthesis so that the 20% aliquot and each of the sub­aJiquots of the remaining 80% of the resin contained all compounds synthesized up to that point.

his library synthesis scheme was followed at each step of the synthesis. The total number ofcompounds contained in the library was 1.2 million. This library was screened to identify onlycompounds with an ICSO better than 50 nM. Screening this library produced a family ofcompounds which inhibited binding and were either agonists (produced a rise in calcium) or wereputative antagonists (did not induce a rise in calcium while inhibiting the CSa stimulated rise incalcium. data not shown). Thirty seven compounds were identified. the most active of which was(Me)Phe-Lys-Pro-cha-Cha-arg-OH (ICSO 16 nM). Several compounds distinct from the parentligand (Phe-Lys-Pro-cha-Cha-arq-Ol-i) were identified. The results for Phe-Lys-Pro-cha-Cha-arq- ~

OH analogs are summarized in Tabie 5. ~~I

U

f'l '

Discussion

Data are presented for four different targets using the one-bead-one-compound strategy in threedifferent synthesis/assay procedures. While each target differs in some specific respect, ail areof therapeutic interest, and two have proven difficult for the pharmaceuticai industry.

Table 1 demonstrates that a iarge number of active anaiogs can be easily identified using the on­bead assay with an enzyme system, thrombin. The on-bead assay is a binding assay which inthe example shown detects compounds which bind in the active site of thrombin. However, thefidelity of this assay was very good in that over 90% of the thrombin binding compounds whichwere identified, resynthesized, and tested for inhibition of enzyme activity were thrombininhibitors. Furthermore, the data indicate that this on-bead binding assay is very sensitive,enabling compounds with as weak as 700 ~M Ki values to be detected. The optimization data isparticuiarlyexciting. Thrombin is an enzyme for which there exists both the catalytic active siteand an exosite which is in addition utilized for binding by high molecular weigl)tt,¥omtJin .' .inhibitors, as well as by the peptide hirulog (phe-Pro-Arg-Pro-Gly-Gly-Gly-GI~-N-Gly-Asp-Phe­

Glu-Glu-lle-Pro-Glu-Glu-Tyr-Leu). The phe-Pro-Arq-Pro element of hiruolg which binds in theactive site of thrombin exhibits a KI of only 20 ~M, and the C-terminal peptide of hirulog whichbinds in the exosite of thrombin exhibits a Ki of only 2 ~M. Yet hirulog exhibits a Ki of 2 nM.Consequently we designed a library to probe for binding outside of, but immediately adjacent to,the fPRP active site pocket, an phe-Pro-Arg-Pro-XXXXX library. This library resulted in thediscovery of a peptide of only nine residues (SEL 1172) which upon resynthesis without linker asSEL 1478 (phe-Pro-Arg-Pro-Phe-Gly-Tyr-Arg-Val-NH2) exhibited a Ki of 25 nM. There was awell defined SAR for this compound, explored through the synthesis of individual compounds, asdetailed in Table 2. This SAR indicates each residue, except possibly tyrosine (residue 7),contributes to the increased potency of this molecule. Because these analogs were synthesizedon solid phase as a batch of individual compounds (multiple synthesis) the analog SAR wasobtained rapidly following on the heels of the initial discovery. Thus the thrombin results not onlydemonstrate how robust the on-bead binding assay is, but also demonstrate the success whichcan be obtained in optimization of a lead to explore additional binding domains and potentially"walk" into another series of enzyme inhibitors.

Thrombin, however, is a target which has proven easy for medicinal chemists. In contrast, factorXa is an enzyme with which the pharmaceutical industry has not been very successful inidentifying small molecular weight specific inhibitors, with only one (recent) example of successdescribed to-date (26). The results which are summarized demonstrate the power of thecombinatorial approach to not only identify leads for difficult targets of great therapeuticsignificance, but again point to the success which can result from an optimization effort, in thiscase optimizing the activity from 20 ~M (SEL 1691) to 150 nM (SEL 2060). In this projecttraditional medicinal chemistry efforts fed by the combinatorial results produced an orally activeand highly specific antithrombotic agent (SEL 2711). This success was with peptide libraries anda peptide series of compounds. What is of particular note was the ease with which an orallyavailable inhibitor (SEL 2711) was identified. Traditionally it has been difficult to convert peptidesinto orally available drugs. However, the factof Xa result suggests that combinatorial chemistrymay be generally successful in overcoming the pharmacological property hurdle in peptide leadoptimization.

While the on-bead assays are powerful and can be easily applied to soluble enzyme targets, thepower of the one-bead-one-compound strategy lies in the other screening formats which areavailable to the scientist. The application to a functional assay is presented by the agar diffusion­based anti microbial assay resuits. Table 3 validates this assay, demonstrating that thesensitivity of the agar diffusion assay was comparable to the liquid MHC assay. It is important topoint out that the quantitative size of the zone of inhibition varies from assay-to-assay, is •

2

dependent on the organism, and can also be affected by compound solubility and diffusability andthus should not be the ultimate indication of activity for determining structure activity relationshipsbetween compounds. This is a powerful assay, permitting assessment of functional activity ofmany compounds (20,000 per plate) in a single release format, which simplifies the syntheticchemistry. Compounds can be released from beads in the dry state, and then diffuse out of thebeads when suspended in the agar, broadening the release chemistry which can be used. Table4 se rves to demonstrate that a library approach can produce many analogs of the same series,which may not have been discovered from traditional analoging efforts without considerablelabor. This point is also made with the case of factor Xa.

The C5a receptor is one of a class of very interesting membrane bound receptors which containseven trans membrane spanning regions and are coupled to G-proteins as signal generatingcomplexes. Screening for such a target represents many challenges, because the assay mustbe conducted either as a membrane or whole cell assay. The final example was selectedbecause It demonstrates the application of the one-bead-one-compound strategy to an interestingreceptor which had proven difficult because of the tendency to discover agonists rather thanantagonists. The primary assay was a radio label membrane binding assay, and the secondaryassay was a whole cell functional assay to distinguish antagonists from agonists. Table 5presents the actual screening results of a library synthesized based on an agonist lead. It isparticularly noteworthy that from this library compounds were discovered which were closeanalogs of an antagonist discovered through a traditional medicinal chemistry effort by one group(Konteatis et ai, Merck, ref 27), and which had eluded another research team (Kawai et ai,Abbott, ref 17).

The assays described herein and results demonstrate that most types of targets known to thepharmaceutical industry can be pursued using the one-bead-one-compound strategy. They alsodemonstrate that the process not only produces leads, but enables leads to be efficientlyoptimized. Finally, the results with Factor Xa and C5a demonstrate that difficult targets can besuccessfully pursued with this approach. It is never possible to directly compare efforts betweendifferent laboratories. However, we suggest that the one-bead-one-compound combinatorialchemistry driven drug discovery strategy can be successful where traditional medicinal chemistryhas failed or met with considerable difficulty. Furthermore, even where traditional medicinalchemistry has been successful, the concomitant use of the one-bead-one-compound strategycan accelerate the drug discovery and optimization process and improve the quality of the result.

Acknowledgment

The authors wish to thank the staff at Selectide who carried out many experiments to produce theresults described in this manuscript: Martha Ackerman-Berrier, Dagmar Cabel, Charlie Chen,Ron Ferguson, Kevin Haney, Padrnaja Kasireddy, Zuzana Leblova, Jeff Locascio, Nina Ma,Victor Nikolaev, Anna Maria Robinson, Jim Spoonamore, Magdalena Stankova, Steve Stringer,Shelly Wade, Aleksandra Weichsel, Patti Willson, Luwei Zhao

Figure 1.

Schematic diagram of the on-bead binding assay for identifying positive compounds from a one­bead-one-compound library. Library beads containlnp the test compound (lIB Cpd), and whennecessary a sequencable tagging molecule (t::l Tag), were mixed with the tarqet-blotln/alkalinephosphatase (AP)-Iabeled streptavidin complex and incubated to allow the complex to bind to anyposltlve compounds contained in the library. The mixture was then washed to remove free targetcomplex, and the beads were subsequently incubated with the alkaline phosphatase BCIPsubstrate. A blue percipliate formed on those beads to which the target complex was bound, andafter washing the beads to StDP the reaction, the coiored beads were selected for analysis. Theidentity of the positive compounds was determined either directly from analysis of the compoundon the beads, or in the case where the compounds could not be directly identified, from analysis(such as sequencing) of a coding tag which had been synthesized on the beads.

Figure 2.

Schematic diagram of the double release assay for identifying positive compounds from a one­bead-one-compound library using solution based assays. Step 1: Library beads containing two

copies of the test compound on a double cleavabie linker r<:) and either a third copy of the testcompound or when necessary a sequencable tagging molecule on a non-cleavable linker (t::lCpdlTag), were suspended in HCI and distributed into the wells of a 96-well filterplate at somefixed number per well (e.g. 50 beads/well). The beads were then washed and the acid replacedwith aqueous buffer, pH 7.2, to release one copy of test compound from the double cleavablelinker, creating a mixture of (50) different compounds free in solution in each well. All themixtures were then filtered into a replicate plate and alliquotes of the mixtures of soluble testcompounds taken for testing in the screening bioassay. Step 2: The beads retained in thefilterplate well corresponding to each positive test well were resuspended in acid and redistributedinto a second master filterplate at a density of one bead per well. These beads were washed andthen exposed to aqueous buffer at basic pH to release the second copy of each test compoundfrom the double cleavable linker. The released material from each bead was filtered into areplicate plate and aliquotes of each well, now containing a single test compound free in solution,were tested both in the screening assay and when appropriate Dr desirable in a functional assayto differentiate agonist and antagonist activity. Step 3: The positive compounds identified in thismanner were then submitted to quality assessment to determine purity, concentration, and massspectrum. In some cases the compoudn could be identified from its mass spectrum. In additionthe parent bead was recovered from the master filterplate well and the compound remainingattached through the non-cleavable linker used to determine identity, or in circumstances wheninstead a tagging molecule had been synthesized onto the bead, this was used to identify the testcompound.

Figure 3.

Structure of SEL 2711, f\C~tYI:P}l]JidihORhenYIAIa:nYI:tpycl?heXY'Ci'IYCYI,3-(MethYIPyripinium)AlanYI-LeucYI-Prolin~-amide, otherwise a~Qj:eliiated as tC'PAPh-Ctlg.pal(3)MOCLeu-proCamidj3.

Table 1

Thrombin On-Bead Screening Results from Pentapeptide Libraries

_XXRXX Format/ LDLDL*

Asp-phe-Ser-arg-ArgThr-arq-Phe-phe-ProLeu-ile-Phe-arg-ProMet-met-Phe-arg-IleGly-phe-Phe-phe-LysLeu-ile-Arg-met-PhePhe-pro-Arg-Gly-Gly

Ki (I-lM)**

>1671172016723

>167>167

00000*

phe-phe-arg-tyr-glnphe-phe-arq-thr-lys

phe-phe-arg-phe-hisphe-phe-arg-pro-thrphe-phe-arg-nle-nlephe-Ieu-arg-glu-tyrphe-pro-arg-ser-glytrp-phe-pro-arg-glyarg-Ieu-met-his-tyr

arg-Ieu-met-arg-phearg-Ieu-nle-arg-phe

67117234353127

>16711711783150

* Structures shown were tested as the G-terminal Gly-amide construct, representing the terminalresidue from the linker used to synthesize library compounds onto beads for screening.** The Ki was calculated from the measured IG50 values.

Table 2

Evaluation of the Structure Activity Relationships for the Thrombin Inhibitor fPRPFGYRV

Selectide Number

SEL 972SEL 1478SEL 1523SEL 1480SEL 1481SEL 1482

SEL 1485SEL 1487SEL 1530SEL 1484SEL 1483

Structure"

phe-Pro-Arg-Pro-Glyphe-Pro-Arg-Pro-Phe-Gly-Tyr-Arg-Valphe-P ro-Arg-Pro-Phe-Gly-Tyr-Argphe-Pro-Arg-Pro-Phe-Gly-Tyrphe-Pro-Arg-Pro-Phe-Glyphe-Pro-Arg-Pro-Phe

phe-Pro-Arg-Pro-Phe-Gly-Tyr-Arg-Ala-B**phe-Pro-Arg-Pro-Phe-Gly-Tyr-Ala-Val-Bphe-Pro-Arg-Pro-Phe-Gly-Ala-Arg-Val-Bphe-Pro-Arq-Pro-Phe-Ala-Tyr-Arg-Val-Bphe-Pro-Arg-Pro-Ala-Gly-Tyr-Arg-Val-B

Ki (nM)

20,0002562167180369

2510946139

2,196

* Synthesized as the C-termlnal amides.** Synthesized with a C-terminal residue (p-alanlne) analogous to the terminal residue of thelinker used to synthesize the original library compounds onto beads for screening.

,

Table 3

Activity of arg-arg-trp-trp-cys-arg-cys Analogs in the Agar Diffusion Assay and Liquid MiC Assay

Compound* Zone of Inhibition (mm) MIC50 (fig/ml)

arq-arq-trp-trp-cys-arq-cys (parent)arq-arq-trp-trp-cys-ala-cysarq-arq-trp-trp-ala-arq-cysala-arq-trp-trp-oys-arq-cys

arp-arq-ala-trp-cys-arq-cysarq-arq-trp-ala-cys-arq-cys

trp-trp-cys-arq-cysarq-arq-trp-trp

3.02.52.52.5

2.02.02.50.8

35203532

320320110

>320

* All compounds tested were the Gly-NH(CH2)30H derivatives, the form released in the agardiffusion assay.

Table 4

D-Amino Acid Peptides with Antimicrobial Activity Identified from Libraries

Compound*

Active in S. aureus AssayAc-trp-trp-arq-arqarq-cys-nal-cys-serarq-cys-nal-cys-alaarq-cys-nal-cys-arqarq-cys-nal-pen-rnet

Active in P. aeruginosa Assayarg-arg-nal-pen-argarg-arg-nal-pen-omarg-orn-nal-pen-arg

Zone of inhibition (mm)

2.52.52.54

3.5

1.74

1.5

*All Crpounds were syntesized as the free acid

• ,-

Table 5

C5a Ligands Identified from a Secondary Optimization Library

Sequence

Phe-Lys-Tic-cha-Cha-arg-OHPhe-His-Tyr-t1e-Chg-arg-OH(NMe)i;'he-Lys-Val-cha-Cha-arg-OHPhe-Lys-Pro-Leu-Cha-arg-OH(NMe) Phe-Lys-Tyr-cha-Cha-arg-OHPhe-Lys-Tyr-cha-Cha-arg-OHPhe-Dab-Pro-cha-Cha-arq-Ol-l

StandardDeviation

3333663

References

1. Lam K.S., Salmon, S.E, Hersh, E.M., Hruby, V.J., Kazmierski, W.M., and Knall, R.J. Nature,1991,245,82

2. Furka, A., Sebestyen, F., Asgedom, M., and Dibo, G. Int. J. Peptide Protein Res., 1991, 37,487

3 Lam,K.S., and Lebl, M. ImmunoMethods, 1992, 1, 11

4. Lam, K.S., and Lebl, M., Methods: A Companion to Methods in Enzymology, 1994, 6, 372.

5. Salmon, S.E., Lam, K.S., Lebl, M., Kandola, A, Khattri, P.S., Wade, S., Patel, M., Kocis, P.,Krchnak, V., Thorpe, D., and Felder, S. Proc. Nat!. Acad. Sci. USA 1993, 90, 11708

6. Lebl, M., Krchnak, V., Salmon, S.E, and Lam, K.S. Methods: A Companion to Methods inEnzymology, 1994,6,381

7. Lebl, M., Krchnak, K., Sepetov, N.F.,Seligmann, B., Strop, P., Felder, S., Lam, K.S.,Biopolymers (Peptide Science), 1995,37,177.

8. Chen, C., Strop, P., Lebl, M., Lam, K.S., Methods of Enzymology, In Press.

9. Gallop, M.A., Barrett, RW., Dower, W.J., Fodor, S.PA, and Gordon, E.M., J. Med. Chem.,1994,37,1235.

10. Gordon, EM., Barrett, RW., Dower, W.J., Fodor, S.PA, and Gallop, MA, J. Med. Chem.,1994, 37, 1385.

11. Desai, M.C., Zuckermann, R.N., and Moos, W.H. Drug Develop. Res., 1994,33,174.

12. Han, H., Wolfe, M.M., Brenner, S., and Janda, K.D. Proc. Nat!. Acad. Sci. USA, 1995, 92,6419.

13. Ohlmeyer, M.H., Swanson, R.N., Dillard, L.W., Reader, J.C., Asouline, G., Kobayashi, R.,Wigler, M., and Still, W.C., Proc. Natl. Acad. Sci. USA, 1993, 90,10922.

14. Nestler, H.P., Bartlett, PA, and Stili, W.C., J. Org. Chem., 1994, 59, 4723.

15. Baldwin, J.J., Burbaurn, J.J., Henderson, I., and Ohlmeyer, M.H.J., J. Am. Chem. Soc., 1995,117,5588.

16. Burbaurn, J.J., Ohlmeyer, M.H.J., Reader, J.C., Henderson, I., Dillard, L.W., Li, G., Randle,T.L., Sigal, N.H., Chelsky, D., and Baldwin, J.J., Proc. Nat!. Acad. Sci. USA.., 1995, 92, 6027.

17. Kocis, P., Krchnak, V., & Lebl, M. Tetrahed. Lett. 1993, 34, 7251

18. Nikolaiev, V., Stierandova, A., Krchnak, V., Seligmann, B., Lam K.S., Salmon, S.E., and Lebl,M. Pept. Res. 1993,6,161

19. Frost, JA, Threlfall, E.J.., and Willshaw, GA Antibiotics, Assessment of AntimicrobialActivity and Resistance, A.P. Russel and L.B. Quesnel, edts., Academic Press, NYNY 1983, 265

•

20. Hediske, J.J.• Quintavalia, J.C.• Haston, W.• Morrissey, M.M., and Seligmann. B. J.Leukocyte Bioi. 1992. 5, 484

21. Houghten, R.A., Piniila, C.• Blondelle, S.E.• Appel, J.R., Dooley. C.T., and Cuervo, A.H.Nature. 1991. 354. 84.

22. Houghten, R.A., Appel, J.R., Blondelle, S.E.. Cuervo. J.H., Dooley, C.T.• and Pinilla, C.Biotechniques, 1992. 13, 412

23. Drapeau, G., Brochu. S., Godin. D., Levesque, L., Rioux, R.. and Marceau, F. Biochem.Pharm. 1993,45:1289

24. l.uly, J.R.. Megumi, K.• and Wiedeman. P.E. 1994. March, US Patent 005190922A

25. Kawai. M.• Wiedeman, P.E.• Luly, J.R.. and Or, Y.S. World Intellectual Property Organization,1992. Publication W092/12168-A1

26. Yokoyama, T., Keliy, A.B.• Marzee, U.M.• Hansen. S.R.. Kunilada, S., and Harker. L.A.Girculation, 1995, 92. 485.

27. Konleatis, Z.D.• Siciliano, S.J.• Van Riper. G.• Molineaux. C.J., Pandya. S.• Fischer. P.•Rosen, H., Mumford. R.A.. and Springer, M.S. J. Immunol. 1994. 153, 4200•

Tag

~ Cpd plus Target-Biotin/Streptavidin-AP

•Wash

Tag •

~ CpdlTarget-Biotin/Streptavidin-AP

Tag • .JI Add AP BCIP Substrate

~ CpdlTarget-Biotin/Streptavidin-AP

•Wash

•Select Positive Colored (e) Beads

De-stain Positive Beads, Confirm Specificity WithActive-Site Inactivated Target Complex

Identify Active-Site Specific Compounds(sequence Cpd or Tag)

.' ".,..., ..

, , I •

}) ""'.' N

Ste

p2

-Pla

tep

are

nt

be

ad

s,1

/we

ll,re

lea

se2

nd

arm

-Filt

er

rele

asa

te

-Ass

ess

fun

ctio

na

la

ctiv

ity

-Id

en

tlty

po

siti

vew

ell

~%%~~=

i

Cp

dfT

ag

..f7

-C

pd~Cpd

-7 '-

Re

cove

rp

are

nt

be

ad

s,

~?;/

h~CZ

28,{

~cgy

;;;;

J" , r

Ste

p3

r-~

Ste

p1

-Pla

te-5

0b

ea

ds/

we

ll,re

lea

se1s

ta

rm..

-Fllt

er

rele

asa

te

-Id

en

tlfy

po

siti

vew

ell(

s)

-Re

cove

rp

are

nt

be

ad

I!.....

v

I

-Cp

dQ

C:

MS

,pu

rity

,I

con

cen

tra

tio

n.f

!v

-ld

en

tlfy

Cpd

fro

m~

be

ad

info

rma

tio

n

~ .. to,

.. .~.'

~0N'\o -yy CONH2

- . ... ....•..~