Design, Virtual Screening, and Synthesis of Antagonists of

Design, Virtual Screening, and Synthesis of Antagonists of αIIbβ3 asAntiplatelet AgentsPavel G. Polishchuk,† Georgiy V. Samoylenko,† Tetiana M. Khristova,†,‡ Olga L. Krysko,†

Tatyana A. Kabanova,† Vladimir M. Kabanov,† Alexander Yu. Kornylov,† Olga Klimchuk,‡

Thierry Langer,§ Sergei A. Andronati,† Victor E. Kuz’min,† Andrei A. Krysko,*,† and Alexandre Varnek*,‡

†A.V. Bogatsky Physico-Chemical Institute of National Academy of Sciences of Ukraine, Lustdorfskaya doroga 86, Odessa 65080,Ukraine‡Laboratory of Chemoinformatics (UMR 7140 CNRS/UniStra), University of Strasbourg, 1, rue B. Pascal, Strasbourg 67000, France§Department of Pharmaceutical Chemistry, Faculty of Life Sciences, University of Vienna, Althanstraße 14, 1090 Vienna, Austria

*S Supporting Information

ABSTRACT: This article describes design, virtual screening,synthesis, and biological tests of novel αIIbβ3 antagonists, whichinhibit platelet aggregation. Two types of αIIbβ3 antagonistswere developed: those binding either closed or open form ofthe protein. At the first step, available experimental data wereused to build QSAR models and ligand- and structure-basedpharmacophore models and to select the most appropriate toolfor ligand-to-protein docking. Virtual screening of publiclyavailable databases (BioinfoDB, ZINC, Enamine data sets)with developed models resulted in no hits. Therefore, smallfocused libraries for two types of ligands were prepared on the basis of pharmacophore models. Their screening resulted in fourpotential ligands for open form of αIIbβ3 and four ligands for its closed form followed by their synthesis and in vitro tests.Experimental measurements of affinity for αIIbβ3 and ability to inhibit ADP-induced platelet aggregation (IC50) showed that twodesigned ligands for the open form 4c and 4d (IC50 = 6.2 nM and 25 nM, respectively) and one for the closed form 12b (IC50 =11 nM) were more potent than commercial antithrombotic Tirofiban (IC50 = 32 nM).

■ INTRODUCTION

Thrombus formation is the most important pathologicalmechanism underlying atherothrombotic diseases such asacute coronary syndromes and ischemic stroke/transientischemic attack, which are responsible for elevated mortalityworldwide1,2 and which are a platelet-mediated phenomenon.To start to form clots, platelets should be turned from therested state to the activated one.3 Rupture of atheroscleroticplaques is supposed to be the main cause of arterial thrombusformation.4,5 This exposes such platelet activating proteins astissue factor, von Willebrand factor, collagen, etc. Activatedplatelets are able to excrete other agonists of platelet activationsuch as adenosine diphosphate and thromboxane A2, whichpromote activation of adjacent platelets.6 Activated plateletschange their shape and expose fibrinogen receptors, integrinαIIbβ3, which change their conformation from bent conforma-tion to extended conformation with closed headpiece (Figure1). Then β-subunit moves away from α-subunit and thereceptor goes into high-affinity state with open headpiece inwhich it binds fibrinogen and von Willebrand factor, resultingin clot formation and clot adherence, respectively.7 Thus,inhibition of αIIbβ3 can prevent clot formation regardless of theplatelet activation pathway.8−10

Most antagonists of αIIbβ3 represent peptidomimetics, whichmimic RGD or KGD sequence of fibrinogen and bind to theopen form of the integrin (Figure 1, Ligand B). For a long timeresearchers focused their efforts on design of novel RGD-

Received: April 3, 2015Published: September 14, 2015

Figure 1. Conformational changes in integrin αIIbβ3 upon activationand ligand binding.

Article

pubs.acs.org/jmc

© 2015 American Chemical Society 7681 DOI: 10.1021/acs.jmedchem.5b00865J. Med. Chem. 2015, 58, 7681−7694

pubs.acs.org/jmc

http://dx.doi.org/10.1021/acs.jmedchem.5b00865

peptidomimetic that resulted in two marketed drugsTirofiban11 and Eptifibatide.12 The third marketed drug isAbciximab, a monoclonal antibody specific for an epitope on β3subunit.13 The existed drugs have proven their efficiency inreducing a risk of periprocedural myocardial infarction and inurgent target vessel revascularization during catheterization.14

However, these compounds have some drawbacks like inducingthrombocytopenia,15,16 which is supposed to be an immuno-logical response of an organism on the conformational changesin integrin αIIbβ3 upon binding with RGD-peptidomimetics.17,18

Recently, a novel antagonist of αIIbβ3, RUC-1, has beendiscovered during experimental screening of ∼33 000 smallcompounds.19 According to mutagenesis studies, RUC-1 bindsonly to the αIIb subunit of the integrin. As it is shown in gelfiltration and dynamic light scattering experiments, it does notinduce transformations leading to open headpiece form (Figure1, Ligand A). Later on, this was confirmed by X-ray study of thecomplex of RUC-1 and αIIbβ3.

20 Notice that RUC-1 has arelatively weak inhibition potency of ADP-induced aggregationtested on human platelet rich plasma (IC50 = 13 μM).19 Inorder to explore the RUC-1 binding pocket and to obtainadditional information concerning binding mechanism andinduction of conformational changes in the receptor, a series ofderivatives of RUC-1 have been synthesized.21 One of them,named RUC-2, was found some 100 times more potent ininhibiting ADP-induced platelet aggregation than RUC-1 (IC50= 96 nM).21 At the same time RUC-2 does not induce anyconformation changes in the αIIbβ3 headpiece,21 which mayreduce adverse effects. Recently more potent ligands of theintegrin’s closed form RUC-3 (IC50 = 45 nM) and RUC-4(IC50 = 33 nM) have been reported.22 According to moleculardocking and molecular dynamics simulations they interact withthe same residues as RUC-2.Protein−ligand binding patterns in αIIbβ3 open and closed

forms differ. Thus, Tirofiban binds with Asp224 residue of theαIIb subunit and with Mg2+ ion of metal ion-dependentadhesion site at the β3 subunit in the open form of integrin(Figure 2). However, RUC-2 binds to Asp224 residue of theαIIb subunit and to Glu220 residue of the β3 subunit, and thus,it displaces Mg2+ ion. These differences are key factorsdetermining ligands effects on the conformational state of thereceptor.In this article, we report design of novel antagonists of αIIbβ3,

which bind to either open or to closed forms of the receptor.To achieve our goal we applied various ligand- and structure-based modeling techniques to screen either commercialdatabases or generated in this work virtual combinatoriallibraries. Since antagonists of the open form have been

investigated for a long time, a lot of data have been collectedfrom in-house studies and from literature sources. Thus, for thedesign of the ligands for open form of αIIbβ3, we used bothligand-based and structure-based approaches. Only few ligandsof closed form of αIIbβ3 have been reported in the literature.Therefore, here only structure-based approaches were used todesign new compounds.

■ RESULTS1. Design of Antagonists of the Open Form of αIIbβ3.

Two data sets have been collected from literature sources23 andin-house studies:24−29 (1) the affinity data set comprising 338compounds with reported affinity values for αIIbβ3 and (2) theantiaggregation activity data set comprising 453 compoundstested under similar conditions. These data sets were used forthe development of QSAR and ligand-based pharmacophoremodels. X-ray structures of complexes of different antagonistswith αIIbβ3 have been used for the development of structure-based pharmacophore models and molecular docking.Ensemble of QSAR and pharmacophore models and dockingexperiments formed a virtual screening pipeline, which has beenfurther used for compounds selection and prioritization.

1.1. QSAR Models. Three individual 2D QSAR models bothfor ligands’ affinity and antiaggregation activity have been builtusing Random Forest30 method in combination with threetypes of fragment descriptors: simplexes (SiRMS),31,32 ISIDAfragments (SMF),33 and fuzzy pH-dependent pharmacophorictriplets (FPT).34,35 Consensus QSAR models have beendeveloped by averaging predictions of the correspondingindividual models. Predictive performance of the models wasestimated by 5-fold external cross-validation. Usage ofapplicability domain (AD) for consensus models did notsignificantly improve prediction performance, see Table 1.Relatively large RMSE values of predicted activities could be

explained by several reasons. The first one is related to someheterogeneity of experimental data collected from differentsources: experimental methods of evaluation of biologicalresponses were not always similar. Different experimental

Figure 2. Interaction patterns of Tirofiban and RUC-2 compounds with integrin αIIbβ3 in its open (left) and closed (right) forms.

Table 1. Five-Fold External Cross-Validation Statistics forConsensus 2D QSAR Models of Affinity for αIIbβ3 andAntiaggregation Activity

R2 RMSE RAD2a RMSEAD

aAD

coverage

affinity for αIIbβ3 0.75 0.76 0.76 0.72 0.97antiaggregation activity 0.52 0.77 0.54 0.74 0.99

aStatistical parameters take into account applicability domain.

Journal of Medicinal Chemistry Article

DOI: 10.1021/acs.jmedchem.5b00865J. Med. Chem. 2015, 58, 7681−7694

7682


conditions may significantly affect activity values. For example,usage of different agonists and anticoagulants in antiaggregationexperiments may change measured activity values by up to 1order of magnitude.36 Therefore, such data were not includedin the data set. However, we accepted data measured bydifferent experimental methods in order to collect a sufficientlybig data set of compounds studied under similar conditions.The second reason is related to high interindividual variabilityof αIIbβ3 population in platelets, which can fluctuate in therange 71 700−1020 00 receptors per platelet.37 This can lead tosignificant difference in both dissociation constants offibrinogen (KD = 70−255 nM)38 and affinity values ofantagonists of αIIbβ3; for example, clinically relevant concen-trations of Tirofiban, which cause inhibition of fibrinogenbinding range from 17% to 88%.39 Clearly the above reasonsintroduce some inevitable noise in the data, which affectspredictive performance of obtained QSAR models.1.2. Pharmacophore Models. Three structure-based

pharmacophore models were obtained with LigandScout40

from three available X-ray structures of αIIbβ3 complexes withsmall molecule antagonists L-739,758, Tirofiban, Eptifibatide(PDB codes 2VC2, 2VDM, and 2VDN, respectively).41 Thebinding pockets in these complexes are very similar: all majoramino acid residues interacting with ligands are almost at thesame positions. The models performance has been assessed invirtual screening of a validation set combining the affinity set(338 compounds) with decoys selected from the ChEMBLdatabase (1518 compounds). We found that the modelsautomatically built by LigandScout contained too manyfeatures, and by this reason they were not able to select anyactive in the validation set. Therefore, some features weremanually marked as optional and some of them were removedin order to make models less specific (see an example on Figure3). This significantly improved the models performancereaching precision = 0.76−0.86 and recall = 0.13−0.26. Thejoint application of all three models to the validation set slightlyimproved overall prediction performance (precision = 0.81 andrecall = 0.36), whereas enrichment ratio at 1% and 5% wasequal to 5.56 and 6.78, respectively.Structure-based models have provided important information

about distance (∼16 Å) between positively and negativelycharged centers, which are considered as essential for ligandbinding by many authors.11,42,43 This information has beenused for preselection of conformers used for building of ligand-

based pharmacophore models on the compounds from theaffinity data set.Active compounds (pIC50 ≥ 8) from the affinity data set have

been clustered by LigandScout according to their pharmaco-phoric representation. Then, for each of the seven clusters, oneindividual pharmacophore model was produce (see an exampleon Figure 4 and all other models in Figure S3 in the Supporting

Information). Joint application of all seven obtained modelsresulted in a reasonable predictive performance (precision =0.67 and recall = 0.93) and demonstrated higher enrichmentratios than structure-based models: 9.04 and 7.80 at 1% and5%, respectively.To speed up a virtual screening of large databases, a simple

topological 2D pharmacophore model has been developed onthe basis of structure- based pharmacophores. It consists ofonly two features, centers of positive and negative charges,separated by 13 bonds. As one may see from Figure 5, thisroughly corresponds to the distance of 16 Å separating thesefeatures in 3D pharmacophores.

1.3. Docking Studies. At the first step, the most relevantsoftware tool and protein structure have been chosen accordingto self-docking and cross-docking results as well as performanceof virtual screening on the affinity data set. Three ligand−protein complexes (PDB codes 2VC2, 2VDM, and 2VDN)were examined in combination with three different dockingprograms: FlexX,44 MOE,45 and PLANTS.46−48 Since the roleof two water molecules coordinating Mg2+ cation (Figure 2)

Figure 3. Example of initial and manually tuned structure-based pharmacophore models for Eptifibatide complex with αIIbβ3 (2VDN). The followinglabels for pharmacophore features were used: red stars, centers of negative charge; blue stars, centers of positive charge; red arrows − H-bondacceptors; green arrows − H-bond donors; yellow spheres, hydrophobic parts. Exclusion volumes are not shown for clarity.

Figure 4. Example of a ligand-based pharmacophore model. Thefollowing labels for pharmacophore features were used: red star, centerof negative charge; blue star, center of positive charge; red arrow orsphere, H-bond acceptor; green arrow or sphere, H-bond donor;yellow sphere, hydrophobic part. Exclusion volumes are not shown forclarity. Notice that all other ligand-based pharmacophore models aregiven in Figure S3 in the Supporting Material.



7683


was not clear, the docking calculations were performed on thestructure with and without these water molecules. Computa-tional experiments reveal overall performance of MOE appliedto 2VDM binding site including water molecules (see Figure 6and Table S2 in Supporting Information). This setup wasfurther used in the virtual screening.

1.4. Design and Screening of the Focused Virtual Library.Virtual screening of BioinfoDB49 containing about three millionof commercially available compounds with pharmacophore andQSAR models resulted in no hits. This can be explained by alow number of compounds with positively and negativelycharged groups in commercial libraries. Even simple 2Dpharmacophore representation (Figure 5) returns fromBioInfoDB only 210 compounds. Subsequent screening with3D pharmacophore and QSAR models resulted in no hits.Therefore, the focused virtual compound library has beencreated using a fragment-based approach. The main require-ments for new antagonists of αIIbβ3 were derived from thepharmacophore models, docking studies, and some exper-imental observations. They are (i) positively and negativelycharged groups should be separated by at least 16 Å; (ii)lipophilic fragment should be attached to the acidic part of amolecule; and (iii) desirable that the above lipophilic fragmentis linked to a H-bond acceptor able to bind the Arg214 residueof αIIbβ3. According to these rules various Arg- and Asp-mimetic fragments and different linker groups were proposed

(Figure 7). Combinatorial virtual library was generated by in-house computer program. After discarding synthetically

irrelevant structures, the remaining 6930 compounds wereused for the screening.At the first step of virtual screening, pharmacophore and

QSAR models were applied in parallel followed by selection ofcommon hits (Figure 8). Hits selected by QSAR models metspecified threshold values: for affinity pIC50 ≥ 8.0 and forantiaggregation activity pIC50 ≥ 7.0. Common application of

Figure 5. Three-dimensional structure-based pharmacophore (top)and related 2D pharmacophore (bottom) suggested for virtualscreening.

Figure 6. Performance of molecular docking assessed on the affinitydata set with MOE, PLANTS, and FlexX programs applied to 2VDMbinding site.

Figure 7. General scheme of the ligands for open form αIIbβ3 andexamples of building blocks used for generation of virtualcombinatorial library.

Figure 8. Workflow of the virtual screening of the focused librarydesigned for open form αIIbβ3.



7684


pharmacophore and QSAR models resulted in 93 hitsrepresenting 310 individual stereoisomers. Further selectionwith docking lead to 83 compounds (164 stereoisomers). Allthese compounds successfully passed through ADME/Toxfilters using water solubility50,51 and Ames mutagenicity52 in-house models and toxicity assessment with PASS program.53

Two synthetically feasible compounds represented by twoenantiomers each were chosen for further synthesis andbiological evaluation.It is interesting to note that the docking experiments revealed

a little difference between binding poses of differentenantiomers of selected compounds (Figure 9). In the complex,basic nitrogen of tetrahydroisoquinoline group binds to theAsp224, and carboxylic group of ligands binds to Mg2+, whereassulfonamide group forms H-bond with Arg214 residue. Theseinteraction patterns are very similar to those in theTirofiban−αIIbβ3 complex. Thus, it can be expected thatdifferent enantiomers of designed compounds would havesimilar affinity values.1.5. Synthesis and Biological Evaluation. Pure enantiomers

of synthesized compounds were synthesized and tested for theiraffinity for αIIbβ3 receptors and antiaggregation activity. For thepurpose of comparison, experiments were also performed onthe reference commercial compound Tirofiban.DCC/SuOH method has been used for the preparation of

previously reported RGDF mimetics, derivatives of 1,2,3,4-tetrahydroisoquinoline-7-carboxylic acid.27 This method wasused on two stages when the acid 1 or 2 was coupled withsodium salts of appropriate β-alanines. The compound 2described in this article was synthesized from the acid 1 and β-alanine using TSTU as a coupling reagent (Scheme 1). Boc-derivatives 3 have been obtained analogously using PFTU.Acidolytic elimination of Boc-protective groups from com-pounds 3 yielded the target RGDF mimetics 4a−d.Experimental data (Table 2) demonstrate high affinity for

αIIbβ3 and antiaggregation activity of compounds 4a−d.Tirofiban was used as standard inhibitor.2. Design of Antagonists of the Closed Form of αIIbβ3.

In this section we describe design of the analogues of RUC-2ligand displaying high affinity for closed form of αIIbβ3. Sincevery few experimental data on ligands for closed form wereavailable, only structure-based pharmacophore and dockingmethods were used.2.1. Pharmacophore Models. An initial structure-based

pharmacophore model (Figure 10) has been generated withLigandScout using the structure of the RUC-2-αIIbβ3 complex(PDB code 3T3M). This model contains: (i) two positivecenters separated by 15.8 Å, (ii) five H-bond donors associated

with positive centers, two of which directed toward αIIbAsp224amino acid and three to β3Glu220, (iii) three H-bond acceptorsassociated with carbonyl group of the ligand, which binds to

Figure 9. Docking poses of the designed enantiomers 4c (left) and 4d (right).

Scheme 1. Synthesis of Target RGDF Mimetics 4a−da

aReagents: (i) NEt3, TSTU; β-alanine, NaHCO3; H+; (ii) NEt3,PFTU; α-sulfonamido-β-alanines, NaHCO3; H

+; (iii) CH2Cl2, HClgas.

Table 2. Predicted and Experimentally Measured Values ofAffinity for αIIbβ3 and Antiaggregation Activity of theDesigned αIIbβ3 Antagonist of the Open Form of Integrin

aPredicted by consensus 2D QSAR models.



7685


αIIbAsp232 residue via two water molecules (see also Figure 2),(iv) H-bond donor bounded with β3Asn215, and (v) one H-bond acceptor and hydrophobic feature shifted toward one ofthe positive centers.Two-dimensional topological pharmacophore has been

derived from the 3D structure-based pharmacophore similarlyto that previously described in section 1.2. It contains twopositively charged centers separated by, at least, 12 bonds. Thiscorresponds to the distance of 15.8 Å between two positivecenters in the 3D pharmacophore model.2.2. Docking Studies. As previously described for the open

form ligands, a preliminary study has been performed in orderto select the most appropriate docking tool (FlexX or MOE)able to characterize a binding mode of molecules interactingwith the binding pocket of RUC-2-αIIbβ3 (3T3M) complex.Self-docking studies demonstrated the performance of FleX(RMSD = 0.78 Å), which retained all important ligand−proteininteractions, whereas MOE (RMSD = 2.2 Å) failed to bindligand to Glu220 residue, which seems to be crucial for ligand−protein recognition.2.3. Virtual Screening of Publicly Available Databases.

Developed 2D and 3D pharmacophore models were used toscreen several large databases of commercially availablecompounds: (i) advanced and HTS Enamine databases,containing collection of 1.5 million structurally diversecompounds;54 (ii) REAL Enamine database, containing ∼17million synthetically feasible compounds;55 and (iii) ZINCdatabase,56 which ensembles collections of compounds fromdifferent vendors with overall more than 17 millioncompounds. This resulted in 50 compounds, which havebeen docked with FlexX. Only two high score compounds havebeen selected. One of them is the known drug Nafamostat, aserine protease inhibitor.57 This compound was also identifiedby Negri et al. in their structure-based virtual screening.58

Nafamostat has some clear drawbacks: it does not possess highantiaggregation activity (IC50 = 12.5 μM), and it may havesome side effects because of its ability to bind different proteins,such as thrombin, urokinase, trypsin, plasmin, etc.59−61 Itshould, however, be noted that Nafamostat was introduced asan alternative anticoagulant in continuous renal replacementtherapy (CRRT) in 1990, but its usage is mainly limited toJapan.62,63 Another selected compound, N-[(E)-[1-(2-{5-[(1E)-1-(carbamimidamidoimino)ethyl]-4-methyl-1,3-thiazol-2-yl}-4-methyl-1,3-thiazol-5-yl)ethylidene]amino]guanidine,was not available for purchasing at that moment.2.4. Design and Screening of Focused Virtual Library.

Since no compounds were selected from the commercialdatabases, a small focused virtual library of RUC-2 analogueshas been designed. According to 3D structure analysis, a ligandfor the integrin’s closed form should possess: (i) a positively

charged part (preferably pyperazine residue) able to interactwith the Asp224 residue; (ii) a heterocyclic moiety interactingwith the Tyr190 residue; (iii) an acceptor group (preferablycarbonyl) interacting with the Asp232 residue, and (iv)positively charged part (amino group) displacing Mg2+ ionand, in such a way, providing with interactions with Glu220residue of the β3 subunit. Potentially, a molecule combining 6-amino-2-(piperazin-1-yl)-3H-quinazolin-4-one scaffold con-nected to amino-group, as it is shown on Figure 11, may fulfill

these conditions. Notice that substituted quinazolinediones andquinazolinones derivatives are known as platelet aggregationinhibitors and fibrinogen receptor antagonists.64

Based on these considerations, 29 virtual compounds (41stereoisomers) were designed varying a linker separating 6-amino-2-(piperazin-1-yl)-3H-quinazolin-4-one scaffold andamino-group (see Figure 11).Designed compounds were screened against 3D pharmaco-

phore models, followed by docking with FlexX and applicationof ADME/Tox filters described in section 1.4. This resulted in20 hits, three of which (compounds 12a−c in Table 3) wereselected for the synthesis and biological tests. Their bestdocking poses (Figure 12) reveal the binding pattern similar to

Figure 10. Pharmacophore model derived from the RUC-2-αIIbβ3complex. Description of labels is given in the caption for Figure 3.

Figure 11. Schematic representation of ligands for closed form ofαIIbβ3 used for generation of virtual focused library.

Table 3. Experimental Values of Affinity for αIIbβ3 andAntiaggregation Activity of the Designed αIIbβ3 Antagonistsof the Closed Form of the Receptor



7686


those observed in X-ray RUC-2-αIIbβ3 complex : ligands’ aminogroup interacts with β3Asn215 and piperazin group withαIIbAsp224 (Figure 11, right). However, among 10 best poseswe also discovered those with opposite orientation: aminogroup interacts with αIIbAsp224 and piperazine group, withβ3Asn215 (Figure 11, left). These observations show that RUC-2 analogues may have different binding modes in the integrin’sbinding pocket.2.5. Synthesis and Biological Evaluation. In this section we

describe a synthesis of compounds 12a−c selected in virtualscreening. The synthetic route to a set of 2-piperazin-1-yl-quinazolines 12a−c is summarized in Scheme 2. Quinazoline-2,4-dione (6) was obtained by heating the isatoic anhydride andurea in DMF.65 As a result of nitration of the compound 6,there was obtained the 6-nitro-3H-quinazoline-2,4-dione (7).There was carried out the chlorination of compound 7 usingPOCl3 to obtain 6-nitro-2,4-dichloro-quinazoline, which wasfurther hydrolyzed to form 6-nitro-2-chloro-3H-quinazolin-4-one (8). The piperazyl ring was then conveniently introducedto the 2-position of the intermediate 8 by the reaction of it withthe 1-Boc-piperazine, which gave the compound 9. Thereduction of the nitro group of compound 9 using H2−Pd(C) gave the amine 10 as a crucial substrate for theconstruction of the target molecules. Condensation of Boc-

acids with amine 10 has been conducted using the HATU. Theelimination of Boc-protective groups yielded the compounds12a−c.Results of in vitro biology testing are summarized in Table 3.

As one may see, all synthesized compounds are characterizedby high affinity for αIIbβ3 and antiaggregation activity values.

■ DISCUSSION

This study demonstrated that virtual screening of largecommercial databases resulted in no (for open form ligands)or very few (for closed form ligands) hits. This is not surprisingtaking into account that studied peptidomimetics or RUC-2analogues have very specific features (charged parts separatedby a certain distance), which do not occur in most ofcommercial compounds. That is why the only solution wasgeneration of focused libraries followed by their screening.The lack of potential αIIbβ3 binders in commercial databases

is confirmed in recent publications by Negri et al.58 and byWanga et al.66 reporting compounds with relatively weakantiaggregation potency (inhibition of ADP induced plateletaggregation IC50 = 12−47 μM58 and IC50 = 20−90 μM66)resulted from structure-based virtual screening of these datasources.

Figure 12. Docking poses of the compound 12b.

Scheme 2. Synthesis of Compounds 12a−ca

aReagents: (i) (H2N)2CO, DMFA, 150 °C, 3 h, 63%; (ii) HNO3, H2SO4, −5 °C, 1 h, 67%; (iii) POCl3, N,N-dimethylaniline, reflux, 6 h; (iv) 1 MNaOH, H2O, room temperature; (v) 2 h, 1 M HCl, 61%; (vi) 1-Boc-piperizine, NEt3, ACN, 50 °C, 2 h, 82%; (vii) H2/Pd (C), MeOH/C6H6, roomtemperature, 3 h, 92%; (viii) NEt3, HATU, Boc-acids, ACN, 50 °C, overnight, 46-58%; (ix) CH2Cl2, HCl gas, room temperature, 1 h, 92−96%.



7687


Important feature of the given study is application ofdifferent chemoinformatics approaches: QSAR, 2D and 3Dpharmacophores, and ligand-to-protein docking. Joint applica-tion of different modeling techniques provided with very highsuccess rate: almost all designed compounds displayed highaffinity for αIIbβ3 and antiaggregation activity, and some of themperform better than commercial drug Tirofiban. Notice that theclosed form ligand 12b (antiaggregation activity IC50 = 11 nM)designed in this study outperforms recently reported RUC-3(IC50 = 45 nM) and RUC-4 (IC50 = 33 nM) molecules.22

■ CONCLUSIONThis work is devoted to design, virtual screening, synthesis, andin vitro tests of novel αIIbβ3 antagonists. Two types of ligandswere designed: those binding to open or to closed form of theprotein. Various theoretical approaches were applied: QSAR,structure- and ligand-based pharmacophore and docking.Consensus virtual screening involving all these techniquesallowed us to select very few molecules for the synthesis and invitro tests. Experimental validation demonstrated very highsuccess rate of our approach: 2 out of 4 hits for open-formligands and 2 out of 4 hits for closed-form ligandsdemonstrated higher affinity and antiaggregation activity thancommercial antithrombotic Tirofiban.

■ EXPERIMENTAL SECTIONDescription of Data Sets of RGD-Peptiomimetics Bound to

the Open Form of αIIbβ3. Two data sets of RGD-peptidomimetics,which possess affinity for αIIbβ3 or antiaggregation activity, have beenprovided by A.V. Bogatsky Physical-Chemical Institute of NationalAcademy of Sciences of Ukraine (PCI). All of those compounds havebeen synthesized and tested for affinity for αIIbβ3 and antiaggregationactivity at the Medicinal Chemistry Department of A.V. BogatskyPhysical-Chemical Institute by earlier described methods. Antiag-gregation activity of compounds was measured by Born’s method onhuman platelet rich plasma.67 Affinity for αIIbβ3 was measured asinhibition of fluorescein isothiocyanate-labeled fibrinogen binding toactivated human platelets by tested compounds.38 These two data setscontained relatively small number of compounds (45 compounds withreported affinity values and 53 with reported antiaggregation activity),and they were significantly imbalanced as they contained mostly activecompounds. Due to this fact these data sets have been extended bydata taken from CHEMBL database (version 7),23 which is a publiclyavailable collection of organic compounds with reported bioactivitydata. The compounds from CHEMBL database have been selectedtaking into account similarity of the used bioassays to those oneswhich were used in PCI tests. This was the crucial step because activityvalues for the same compound obtained in different assays can differmore than order of magnitude. Thus, we expect that in such a wayprepared data sets should be less heterogeneous and more reliable formodeling.Curation of the two data sets has been performed by Chemaxon

Standardizer tool:68 (i) mixtures and inorganics were removed, (ii)salts were cleaned or removed, (iii) normalization of specificchemotypes (aromaticity and nitro groups were checked) wereperformed, (iv) explicit hydrogen atoms were added, and (v)treatment of tautomeric forms were done. The resulted data setscontained achiral and chiral compounds (single stereoisomers andracemic mixtures). During data set curation single stereoisomers wereexcluded if corresponding racemic compounds were present in thedata set. Duplicates in the data sets have been removed usingChemaxon Instant JChem.69 All values of affinity for αIIbβ3 andantiaggregation activity of compounds were converted to pIC50(−lgIC50, IC50 in mol/L units). The extended data sets have becomemore balanced with better distributed and wider range of activityvalues (see Figure S1 in Supporting Information). These data sets areavailable as Supporting Information.

QSAR Modeling of Open Form αIIbβ3 Antagonists. Threedifferent approaches for representation of molecular structure on 2Dlevel have been used: simplex representation31,32,70 and two types ofISIDA descriptors, substructure molecular fragments33 and fuzzy pH-dependent pharmacophore triplets.34,35 As these approaches are welldescribed in the literature we did not provide their detailed descriptionhere.

Random Forest method30 (implemented in the CF software71) wasused for QSAR models development because this method has provedits applicability for solution of various QSAR tasks.72−75 Predictiveperformance of obtained models, expressed as determinationcoefficient (R2) and root mean-squared error (RMSE) (see eqs 1and 2), has been assessed by 5-fold external cross-validation procedure.To perform 5-fold external cross validation all compounds in the dataset were sorted according to their pIC50 values, and each fifthcompound went to the separate bin; thus, five bins were obtained.Then the compounds of four out of the five bins have been combinedtogether, and QSAR model has been developed, compounds from theremaining fifth bin (which is actually an external test set) have beenpredicted by this model. This procedure was repeated five times to useall bins as an external test set only once. Predictions of external setswere combined and cross-validation statistics were calculatedaccording to the eqs 1 and 2. Consensus predictions of affinity valuesfor αIIbβ3 and antiaggregation activity of novel compounds have beenmade by averaging predictions of corresponding individual QSARmodels.

= −∑ −

∑ −

=

=

Ry y

y y1

( )

( )in

i i

in

i

2 1 pred, exp,2

1 exp, training2

(1)

∑=−

−=

RMSEn

y y1

1( )

i

n

i i1

pred, exp,2

(2)

where n is the number of compounds in an external set; yexp,i is theobserved activity value of ith compound in an external set; ypred,i is thepredicted activity value of ith compounds in an external set; ytraining isthe mean activity value for compounds of a training set.

To check whether compounds were inside or outside of AD of theconsensus models we estimated root-mean-square deviation of allpredictions made by individual QSAR models. If a calculated standarddeviation value was less than or equal to 0.5, then this compound wasinside AD; otherwise, it was outside AD.

ADME/Tox Assessments of Novel Compounds. Some ADME/Tox properties of screening compounds, such as mutagenicity andsolubility were estimated using previously reported QSAR models.The classification 2D QSAR model52 based on the Ames mutagenicitydata set76 was applied for prediction of mutagenicity of novelcompounds. This model was developed using Random Forest methodand based on the simplex representation of molecular structure. Theaqueous solubility of screened compounds was predicted by two 2DQSAR models in order to make average consensus prediction. Theformer is based on ISIDA descriptors and developed using MultipleLinear Regression method,50 and the latter was developed usingRandom Forest method in combination with simplex descriptors.51

Possible adverse effects were assessed by PASS53 program, whichpredicts probability of many types of activity and toxic effects.

Pharmacophore Model Development. Structure-based andligand-based approaches implemented in LigandScout have beenused for developing of 3D feature-based pharmacophore models.40 Forgeneration of structure-based pharmacophore models ligands incomplexes were ionized, and their energy was minimized usingMMFF94 force field afterward structure-based pharmacophore modelswere produced.

Ligand-Based Models of RGD-Peptidomimetics, Preparation ofthe Validation Set, and Considered Statistical Parameters.Compounds of the affinity data set with pIC50 ≥ 8 have been chosenfor generation of ligand-based pharmacophore models. The selectedcompounds have been charged with Filter tool of OpenEye77 and atmost 200 conformers within 10 kcal/mol energy gap have been



7688


generated using Omega.78 Conformers having distance less than 16 Åbetween positively and negatively charged atoms (these are two keyfeatures of αIIbβ3 antagonists according to structure-based pharmaco-phore models) were discarded using in-house Python script based onOpenEye OEChem library.79 Compounds have been clustered basedon their pharmacophoric representation and for each of seven clustersof compounds a shared ligand-based pharmacophore has beengenerated with the default LigandScout settings.The validation set was prepared from compounds of the affinity data

set and decoys. Compounds from the affinity data set was split on 234active (pIC50 ≥ 6) and 104 inactive (pIC50 < 6) ones. The set ofinactive compounds has been extended by decoys selected fromCHEMBL data set taking into account compounds similarity to knownαIIbβ3 antagonists. At the first step, compounds with reported values ofantiaggregation activity or affinity for αIIbβ3 have been discarded fromthe whole CHEMBL data set. Then, compounds that had the numberof H-bond donor and acceptors, molecular weight, and topologicalsurface area, predicted logD values80 within the range of correspondingvalues for compounds of the affinity data set have been chosen. On thelast step, only 1518 compounds, which had at least one positively andone negatively charged group, have remained as these are key featuresaccording to the structure-based pharmacophore. Thus, the wholevalidation set contained 234 active compounds and 1622 inactives anddecoys. Because some of active compounds were used in modeling ofligand-based pharmacophores they have been excluded from the set ofactives for validation of ligand-based pharmacophore models.All possible stereoisomers were generated for each compound in the

validation data set having unspecified stereocenters. Afterward at most200 conformers within 10 kcal/mol energy window were producedusing Omega.78 The compound of the validation set was predicted asactive if at least one of its stereoisomers fit at least one pharmacophoremodel. Pharmacophore fit scoring function taking into account onlychemical feature overlap was used for ranking of the screening results.Statistical characteristics that were used for estimation of

pharmacophore models performance are given:

= +

= +

= ++

+ + +

recall TP/(TP FN)

precision TP/(TP FP)

enrichment ratioTP

TP FPTP FN

TP FN TN FP

Ligand-to-Protein Docking. Three programs were used to carryout the docking studies: PLANTS, FlexX, and MOE.PLANTS uses stochastic search algorithm and chemplp scoring

function.46−48 For the purpose of correct identification of rotatablebonds, charges, and protonation states of atoms, SPORES81 (structurerecognition and protonation tool) was used. Centre and radius of thebinding pocket were determined using PyMOL.82 Tuning parameterssuch as number of ants and iteration scaling factor remained bydefault.Protein preparation for docking with FlexX44 consists of several

steps: (i) the definition of the binding site by selection of the residuesflanking the binding pocket; (ii) the check of ionization andtautomeric states, position of polar hydrogen atoms, crystal watermolecules, and metal coordination type; and (iii) the addition ofhydrogen atoms to fill out the remaining open valences of thereceptor. FlexX uses systematic search method and its own empiricalscoring function.Binding pockets were prepared in Molecular Operation Environ-

ment (MOE)45 in several steps: (i) protonation of atoms of theprotein, (ii) optimization of the protein−ligand complex, and (iii)removal of redundant water molecules. Stochastic search method andLondon dG scoring function, which is implemented in MOE, wereused in our docking studies.Quality of self-docking and cross-docking studies was evaluated

using RMSD values−root mean squared distance between correspond-ing atoms in an initial conformer and a docked pose. RMSD valueslower than or equal to 2 Å were considered as satisfactory.

Docking of RGD-Peptidomimetics. Binding pockets of threeselected complexes (PDB codes 2VC2, 2VDM, and 2VDN) ofαIIbβ3 headpiece in the open form with three different ligands (L-739,758, Tirofiban, and Eptifibatide) were very similar in geometry:the root-mean-squared-distances (RMSD) between correspondingheavy atoms of residues included in binding pockets of differentcomplexes were 0.8 Å for 2VDM/2VC2 and 2VC2/2VDN and 0.9 Åfor 2VDM/2VDN. Self- and cross-docking studies have beenperformed in order to choose the most appropriate binding pocketand estimate importance of water molecules coordinated with Mg2+

ion. Results for Tirofiban and L-739,758, which were more similar tothe studied ligands than Eptifibatide indicated that the presence ofthose water molecules was preferable (in the presence of the watermolecules there was a greater number of good poses with RMSD ≤ 2Å). It was found that MOE gives better results (lower RMSD values)for Tirofiban and L-739,758 (1.15−2.14 Å) then FlexX (1.85−3.55 Å)and PLANTS (2.07−8.87 Å) (for details see Table S2 in SupportingInformation). To make a final decision, compounds from the affinitydata set were docked into 2VDM binding site using all three programs.To produce ROC curves to estimate docking performance compoundof the affinity data set has been split on active (pIC50 ≥ 6.5) andinactive (pIC50 < 6.5) ones. MOE demonstrated better performance(AUC = 0.72) against this set of compounds than PLANTS (AUC =0.59) and FlexX (AUC = 0.49) (Figure 6). Therefore, MOE softwareand 2VDM binding site were chosen to perform virtual screening.

Docking of RUC-2 and Its Analogues. Redocking of RUC-2 intothe binding pocket in 3T3M has been performed with FlexX andMOE. Results were revealed that water molecules, which participate informing of H-bond of the ligand with the Asp232 residue, wereimportant for correct ligand orientation and binding. MOE failed toreproduce the binding pose (RMSD = 2.2 Å) and to bind the ligand tothe Glu220 residue of the β3 subunit, whereas FlexX almost perfectlyreproduces ligand pose and all ligand-protein contacts (RMSD = 0.78Å). The latter was chosen to carry out the virtual screening.

Virtual Screening of Publicly Available Data Sets andCreated Focused Libraries. All compounds before screening werestandardized with Chemaxon Standardizer68 and charged with Filtertool of OpenEye.77 The virtual screening workflow was common forboth studies. On the first step 2D pharmacophore model implementedas in-house Python script based on OpenEye OEChem library79 hasbeen applied to reduce the number of compounds to the reasonablelevel (several hundred or thousand). For screening of focused librariesthis step was omitted because they had been created taking intoaccount information from 2D pharmacophore. Then, 3D pharmaco-phores, QSAR models (if available), and docking have been applied.On the last stage some ADME/Tox properties and possible side effectshave been estimated, and synthetically feasible compounds have beenselected for synthesis and experimental evaluation.

Chemistry. 1H NMR spectra were recorded on Bruker AvanceDRX 500 spectrometer with chemical shifts in ppm with the internalTMS as a standard. Electron ionization (EI) and fast-atombombardment (FAB) mass spectra were recorded on a VG AnalyticalVG 70-70EQ instrument. FAB spectra were performed equipped withan argon primary atom beam, and an m-nitrobenzyl alcohol matrix wasutilized. The purity was measured by HPLC conducted on anShumadzu system (System Controller CBM-20A, two pumps LC-8A,and Photodiode Array detector SPD-M20A) using a Hypersil GOLD 3μm (4.6 mm × 150 mm) or Hypersil GOLD aQ 3 μm (4.6 mm × 150mm) column. For preparative HPLC Fraction Collector FRC-10A wasused. The progress of reactions was monitored by TLC (silica gel 60F254, Merck).

The acid 1 has been synthesized using a previously publishedmethod.29 The optically active α-sulfonamido-β-alanines wereprepared from L- or D-asparagine using previously publishedmethods.83,84

3-[(2-Boc-1,2,3,4-tetrahydroisoquinoline-7-carbonyl)amino]-propionic Acid (2). The compound 1 (2.0 g, 0.0072 mol) wasdissolved in anhydrous acetonitrile (30 mL). The solution was cooledto −5 °C, and triethylamine (1.0 mL, 0.0072 mol) and then TSTU(2.168 g, 0.0072 mol) were added. The mixture was stirred for 1 h at



7689


−5 °C, and then a solution of β-alanine (1.283 g, 0.0144 mol) andNaHCO3 (1.21 g, 0.0144 mol) in water (20 mL) was added. Thereaction mixture was mixed 3 h at room temperature. The solvent wasevaporated in vacuo to dryness. Then water (20 mL) was added, andpH of the mixture was brought to 3. The product was extracted fromwater by chloroform (2 × 150 mL). The organic layer was washed by 1N HCl (2 × 50 mL) and water (50 mL), then dried with Na2SO4 andfiltered, and the solvent was evaporated in vacuo. The resulting oilyresidue was dissolved in ether (50 mL), and the precipitate wascollected by filtration and dried. Yield 81%; mp = 122 °C. 1H NMRspectral and FAB-MS characteristics of the same product 2 previouslyobtained.27

General Procedure for a Preparation of Compounds 3. Thecompound 2 (0.5 g, 0.0014 mol) was dissolved in anhydrousacetonitrile (30 mL). The solution was cooled to −5 °C, andtriethylamine (0.2 mL, 0.0014 mol) and then PFTU (0.6 g, 0.0072mol) were added. The mixture was stirred for 1 h at −5 °C. Thesolvent was evaporated in vacuo to dryness. The residue was dissolvedin 100 mL of chloroform. The solution was washed with aqueoussolution of 1 M HCl (40 mL), 5% aqueous solution of NaHCO3 (40mL), and water (40 mL). The organic layer was dried over Na2SO4

and filtered, and the solvent was evaporated in vacuo to dryness. Theresulting oily residue was dissolved in acetonitrile (30 mL), and thensolution of α-substituted-β-alanine (0.0028 mol) and NaHCO3 (0.235g, 0.0028 mol) in water (20 mL) was added. The reaction mixture wasmixed 2 h at room temperature. The solvent was evaporated in vacuoto dryness. Then water (20 mL) was added, and pH of the mixture wasbrought to 3. The product was extracted from water by ethyl acetate (2× 100 mL). The organic layer was washed by 1 N HCl (2 × 20 mL),water (20 mL), then dried with Na2SO4, filtered, and the solvent wasevaporated in vacuo. The resulting oil was dissolved in acetonitrile andwas separated by 100 μL portions by preparative HPLC using theconditions: column, Hypersil GOLD aQ 5 μm (20 mm × 150 mm);flow rate, 20 mL/min; mobile phase, water/acetonitrile; isocratic 60/40; detection, UV 254 nm. The product containing fractions wascombined, and the solvent was evaporated in vacuo to dryness.2-(S)-(n-Butylsulfonyl)-3-{3-[(2-Boc-1,2,3,4-tetrahydroisoquino-

line-7-carbonyl)amino]propionyl}aminopropionic Acid (3a). Pre-pared from acid 2 and 2-(S)-(n-butylsulfonyl)amino-β-alanine.83 Yield43%. A colorless glass substance; 1H NMR δ (500 MHz, d6-DMSO)0.85 (t, J = 7.3 Hz, 3 H), 1.30−1.37 (m, 2 H), 1.42 (c, 9 H), 1.56−1.68(m, 2 H), 2.36 (t, J = 6.7 Hz, 2 H), 2.80 (t, J = 5.4 Hz, 2 H), 2.96 (t, J= 7.4 Hz, 2 H), 3.23−3.27 (m, 1 H), 3.37−3.45 (m, 3 H), 3.55 (t, J =4.9 Hz, 2 H), 3.82−3.88 (m, 1 H), 4.52 (s, 2 H), 7.20−7.23 (m, 2 H),7.62−7.64 (m, 2 H), 8.00 (s, 1 H), 8.41 (t, J = 5.1 Hz, 1 H); MS(FAB) m/z: 555 [M + H]+.2-(R)-(n-Butylsulfonyl)-3-{3-[(2-Boc-1,2,3,4-tetrahydroisoquino-

line-7-carbonyl)amino]propionyl}aminopropionic Acid (3b). Pre-pared from acid 2 and 2-(R)-(n-butylsulfonyl)amino-β-alanine. Yield38%. A colorless oil substance. 1H NMR spectral and FAB-MScharacteristics of the same product 3a.2-(S)-(Phenylsulfonyl)-3-{3-[(2-Boc-1,2,3,4-tetrahydroisoquino-

line-7-carbonyl)amino]propionyl}aminopropionic Acid (3c). Pre-pared from acid 2 and 2-(S)-(phenylsulfonyl)amino-β-alanine.84

Yield 36%. A colorless glass substance; 1H NMR δ (500 MHz, d6-DMSO) 1.42 (c, 9 H), 2.27 (ddd, J = 28.5 Hz, J = 14.4 Hz, J = 7.6 Hz,2 H), 2.80 (t, J = 5.5 Hz, 2 H), 3.09−3.15 (m, 1 H), 3.29−3.34 (m,1H), 3.36−3.41 (m, 2 H), 3.55 (t, J = 5.0 Hz, 2 H), 3.90 (dd, J = 14.7Hz, J = 7.1 Hz, 1 H), 4.53 (s, 2 H), 7.22 (d, J = 8.2 Hz, 1H), 7.53−7.63(m, 6 H), 7.77 (d, J = 7.4 Hz, 2 H), 7.99 (t, J = 5.5 Hz, 1 H), 8.36 (t, J= 5.5 Hz, 1 H); MS (FAB) m/z: 575 [M + H]+.2-(R)-(Phenylsulfonyl)-3-{3-[(2-Boc-1,2,3,4-tetrahydroisoquino-

line-7-carbonyl)amino]propionyl}aminopropionic Acid (3d). Pre-pared from acid 2 and 2-(R)-(phenylsulfonyl)amino-β-alanine. Yield51%. A colorless glass substance. 1H NMR spectral and FAB-MScharacteristics of the same product 3c.General Procedure for a Preparation of Compounds 4. The

compounds 3 (1 mmol) were dissolved in anhydrous CH2Cl2 (50mL), and the stream of dry HCl was passed through the solution for

30 min. The solvent was evaporated, and the solid residue was dried invacuo (2 mmHg) for 2 h at 40 °C.

Chloride 2-(S)-(n-Butylsulfonyl)-3-{3-[(1,2,3,4-tetrahydroisoqui-nolinium-7-carbonyl)amino]propionyl}aminopropionic Acid (4a).Yield 98%. A hygroscopic solid; 1H NMR δ (500 MHz, d6-DMSO)0.86 (t, J = 7.1 Hz, 3 H, C4H3 Bu

n), 1.35 (dt, J = 14.3 Hz, J = 6.6 Hz, 2H, C3H2 Bu

n), 1.59−1.69 (m, 2 H, C2H2 Bun), 2.37 (t, J = 6.9 Hz, 2 H,

NHCH2CH2CO), 2.97 (dt, J = 15.4 Hz, J = 8.1 Hz, 2 H, C1H2 Bun),

3.04 (t, J = 5.2 Hz, 2 H, C4H2 THIQ), 3.22−3.27 (m, 1 H,C′H2CH(NHSO2Bu

n)CO2H), 3.35−3.44 (m, 5 H, C″H2CH-(NHSO2Bu

n)CO2H and C3H2 THIQ and NHCH2CH2CO), 3.99(dd, J = 15.3 Hz, J = 6.4 Hz, 1 H, CH2CH(NHSO2Bu

n)CO2H), 4.28(s, 2 H, C1H2 THIQ), 7.29 (d, J = 8.0 Hz, 1 H, C5H THIQ), 7.56 (d, J= 9.0 Hz, 1 H, NHSO2Bu

n), 7.70−7.72 (m, 2 H, C6H and C8HTHIQ,), 8.12 (t, J = 5.1 Hz, 1 H, NHCH2CH(NHSO2Bu

n)CO2H),8.48 (t, J = 5.1 Hz, 1 H, NHCH2CH2), 9.58 (s, 2 H, +NH2 THIQ).HRMS (FAB) calculated for C20H31N4O6S [M + H]+, 455.5573;found, 455.5565.

Chloride 2-(R)-(n-Butylsulfonyl)-3-{3-[(1,2,3,4-tetrahydroisoqui-nolinium-7-carbonyl)amino]propionyl}aminopropionic Acid (4b).Yield 95%. A hygroscopic solid; 1H NMR spectral and HRMS(FAB) characteristics of the same product 4a.

Chloride 2-(S)-(Phenylsulfonyl)-3-{3-[(1,2,3,4-tetrahydroisoquino-linium-7-carbonyl)amino]propionyl}aminopropionic Acid (4c).Yield 98%. A hygroscopic solid; 1H NMR δ (500 MHz, d6-DMSO)2.26 (ddd, J = 31.5 Hz, J = 14.8 Hz, J = 7.3 Hz, 2 H, NHCH2CH2CO),3.05 (t, J = 5.2 Hz, 2 H, C4H2 THIQ), 3.13 (dt, J = 19.2 Hz, J = 6.3Hz, 1 H, C′H2CH(NHSO2Ph)CO2H), 3.29−3.45 (m, 5 H, C″H2CH-(NHSO2Ph)CO2H and C3H2 THIQ and NHCH2CH2CO), 3.88 (dd,J = 14.6 Hz, J = 7.0 Hz, 1 H, CH2CH(NHSO2Ph)CO2H), 4.27 (s, 2H, C1H2 THIQ), 7.29 (d, J = 7.3 Hz, 1 H, C5H THIQ), 7.54−7.62 (m,3 H, m-CH and p-CH Ph), 7.69−7.72 (m, 2 H, C6H and C8H THIQ),7.78 (d, J = 7.4 Hz, 2 H, o-CH Ph), 8.08 (t, J = 5.5 Hz, 1 H,NHCH2CH(NHSO2Ph)CO2H), 8.18 (d, J = 8.8 Hz, 1 H, NHSO2Ph),8.47 (t, J = 5.1 Hz, 1 H, NHCH2CH2), 9.68 (s, 2 H, +NH2 THIQ),11.41 (br s, 1 H, CO2H). HRMS (FAB) calculated for C22H27N4O6S[M + H]+, 475.5477; found, 475.5458.

Chloride 2-(R)-(Phenylsulfonyl)-3-{3-[(1,2,3,4-tetrahydroisoquino-linium-7-carbonyl)amino]propionyl}aminopropionic Acid (4d).Yield 97%. A hygroscopic solid; 1H NMR spectral and HRMS(FAB) characteristics of the same product 4c.

6-Nitroquinazolin-2,4-dione (7). The compound 6 (12.97 g, 0.08mol) was dissolved in concentrated H2SO4 (44 mL), and this solutionwas added to a mixture of concentrated H2SO4 (11 mL) and HNO3(5.6 mL, d = 1.5) at −10 °C. The mixture was stirred for 1 h at −5 °C.The reaction mixture was poured onto crushed ice (200 g). Acrystalline solid was collected and washed with water. The filteredproduct was recrystallized from glacial acetic acid. Yield 67%; mp352−354 °C; MS (EI) m/z 207.

2-Chloro-6-nitro-3H-quinazolin-4-one (8). A mixture of com-pound 7 (3 g, 0.0145 mol), freshly distilled of POCl3 (70 mL) andN,N-dimethylaniline (1.18 mL), was refluxed to completely dissolvethe precipitate, about 6 h. Reaction mixture was allowed to cool toroom temperature, and the solvent was evaporated in vacuo to dryness.To the residue was added ice−water (100 g). Precipitate obtained wasfiltered, washed with distilled water, and dissolved in 100 mL ofchloroform. The solution was washed with water (100 mL). Theorganic layer was dried over Na2SO4, filtered off, and the solvent wasevaporated in vacuo to dryness. The crude product, 2,4-dichloro-6-nitroquinazoline, was used directly in the next step.

2,4-Dichloro-6-nitroquinazoline was suspended in 2 M aqueoussodium hydroxide solution (20 mL), and the mixture was stirred for 3h. Reaction mixture filtered to remove unreacted 2,4-dichloro-6-nitroquinazoline. Filtrate was neutralized with dilute acetic, and theprecipitate was collected by filtration, washed with water, and dried inair. Yield 61%; mp 248−250 °C; MS (EI) m/z 227, 225.

2-(4-Boc-piperazin-1-yl)-6-nitro-3H-quinazolin-4-one (9). Thecompound 8 (2.004 g, 0.0089 mol) was dissolved in acetonitrile (50mL), and to this solution was added 1-Boc-piperizine (1.82 g, 0.0098mol) and NEt3 (1.36 mL, 0.0098 mol). The mixture was stirred for 6 h



7690


at 70 °C. Reaction mixture was allowed to cool to room temperature,and the precipitate of crude product was collected by filtration.Recrystallization from mixture of benzene and ethanol (2:1) to givepure compound 9. Yield 82%; mp 214−216 °C; 1H NMR δ (500MHz, d6-DMSO) 1.42 (s, 9 H), 3.43 (br s, 4 H), 3.72 (br s, 4 H), 7.30(d, J = 8.5 Hz, 1 H), 8.27 (d, J = 7.4 Hz, 1 H), 8.59 (s, 1 H), 11.52 (brs, 1 H); MS (FAB) m/z 376 [M + H]+.2-(4-Boc-piperazin-1-yl)-6-amino-3H-quinazolin-4-one (10). The

compound 9 (0.5 g, 0.0013 mol) dissolved in a 100 mL mixture ofbenzene and methanol (1:1) was subjected to catalytic hydrogenationat room temperature for 2−3 h (the reaction was monitored by TLC)in the presence of 3% palladium on carbon (0.05 g). After the reaction,the stream of helium was passed through the solution for 20 min. Thefiltered solution was then evaporated in vacuo to give compound 10.Yield 94%; mp 241 °C (decomposes); 1H NMR δ (500 MHz, d6-DMSO) 1.41 (s 9 H), 3.39 (br s, 4 H), 3.43 (br s, 4 H), 5.17 (s, 2 H),6.95 (d, J = 7.1 Hz, 1 H), 7.06−7.11 (m, 2 H), 11.25 (br s, 1 H); MS(FAB) m/z 346 [M + H]+.General Procedure for a Preparation of Compounds 11. The 0.01

mol of Boc-acid was dissolved in anhydrous acetonitrile (25 mL). Thesolution was cooled to −5 °C, and triethylamine (1.4 mL, 0.01 mol)and then HATU (3.8 g, 0.01 mol) were added. The mixture wasstirred for 1 h at −5 °C, and then 10 mmol of amine 10 was added.The mixture was stirred for 7 h at 50 °C. The residual amount of theactivated ether (At-ether of starting Boc-acid) was destroyed byaddition of few drops of N,N-dimethylpropane-1,3-diamine, and thesolvent was evaporated in vacuo to dryness. The residue was dissolvedin 100 mL of chloroform. The solution was washed with water (40mL), aqueous solution of 1 M HCl (40 mL), and 5% aqueous solutionof NaHCO3 (40 mL). The organic layer was dried over Na2SO4 andfiltered off, and the solvent was evaporated in vacuo to dryness. Theresulting residue was recrystallization from methanol.3-N-Boc-amino-N-[2-(4-Boc-piperazin-1-yl)-4-oxo-3H-quinazo-

lin-6-yl]propionamide (11a). Yield 58%; mp 244−245 °C; 1H NMRδ (500 MHz, d6-DMSO) 1.37 (s, 9 H), 1.42 (s, 9 H), 2.51−2.55 (m, 2H), 3.22 (s, 2 H), 3.40 (br s, 4 H), 3.55 (br s, 4 H), 6.86 (s, 1 H), 7.24(d, J = 6.9 Hz, 1 H), 7.70 (d, J = 6.7 Hz, 1 H), 8.29 (s, 1 H), 10.01 (s,1 H), 11.41 (s, 1 H); MS (FAB) m/z 517 [M + H]+.4-N-Boc-amino-N-[2-(4-Boc-piperazin-1-yl)-4-oxo-3H-quinazo-

lin-6-yl]butyramide (11b). Yield 50%; mp 237−238 °C; 1H NMR δ(400 MHz, d6-DMSO) 1.37 (s, 9 H), 1.42 (s, 9 H), 1.70 (s, 2 H), 2.30(s, 2 H), 2.97 (s, 2 H), 3.40 (br s, 4 H), 3.55 (br s, 4 H), 6.83 (s, 1 H),7.25 (d, J = 6.1 Hz, 1 H), 7.72 (d, J = 5.5 Hz, 1 H), 8.28 (s, 1 H), 9.97(s, 1 H), 11.41 (1 H); MS (FAB) m/z 531 [M + H]+.N-[2-(4-Boc-piperazin-1-yl)-4-oxo-3H-quinazolin-6-yl]-1-Boc-pi-

peridine-4-carboxamide (11c). Yield 46%; mp 241 °C; 1H NMR δ(500 MHz, d6-DMSO) 1.30−1.52 (m, 21 H), 1.77 (d, J = 11.0 Hz, 2H), 2.78 (br s, 2 H), 3.40 (br s, 4 H), 3.55 (br s, 4 H), 4.00 (br s, 2 H),7.24 (d, J = 6.0 Hz, 1 H), 7.73 (d, J = 6.3 Hz, 1 H), 8.29 (s, 1 H), 10.01(s, 1 H), 11.42 (s, 1 H); MS (FAB) m/z 557 [M + H]+.General Procedure for a Preparation of Compounds 12. The

compound 11 (10 mmol) was dissolved in a 100 mL mixture CH2Cl2and methanol (10:1), and the stream of dry HCl was passed throughthe solution for 1 h. The solvent was evaporated, and the solid residuewas dried in vacuo (2 mmHg) for 2 h at 40 °C.Dichloride 3-Ammonium-N-[2-(4-piperazinium-1-yl)-4-oxo-3H-

quinazolin-6-yl]propionamide (12a). Yield 92%; mp >300 °C-(decomposes); 1H NMR δ (500 MHz, d6-DMSO) 2.83 (t, J = 6.3Hz, 2 H, C2H2 propionamide), 3.08 (dd, J = 11.2, 6.0 Hz, 2 H, C3H2propionamide), 3.26 (br s, 4 H, CH2 piperazinium), 4.06 (br s, 4 H,CH2 piperazinium), 4.98 (br s, 1 H, NH quinazoline), 7.86−7.93 (m, 2H, C7H and C8H quinazoline), 8.15 (s, 3 H, +NH3), 8.46 (s, 1 H, C

5Hquinazoline), 9.86 (s, 2 H, +NH2 piperazinium), 10.83 (s, 1 H, NHpropionamide). HRMS (FAB) calculated for C15H21N6O2 [M + H]+,317.3736; found, 317.3751.Dichloride 4-Ammonium-N-[2-(4-piperazinium-1-yl)-4-oxo-3H-

quinazolin-6-yl]butyramide (12b). Yield 94%. Hygroscopic sub-stance; 1H NMR δ (500 MHz, d6-DMSO) 1.91 (dt, J = 14.3, 7.1Hz, 2 H, C3H2 butyramide), 2.34 (t, J = 7.3 Hz, 2 H, C2H2butyramide), 2.81−2.86 (m, 2 H, C4H2 butyramide), 3.27 (br s, 4

H, CH2 piperazinium), 4.09 (br s, 4 H, CH2 piperazinium), 5.25 (br s,1 H, NH quinazoline), 7.91−7.96 (m, 2 H, C7H amd C8Hquinazoline), 8.17 (s, 3 H, +NH3), 8.46 (s, 1 H, C5H quinazoline),9.98 (s, 2 H, +NH2 piperazinium), 11.73 (s, 1 H, NH butyramide).HRMS (FAB) calculated for C16H23N6O2 [M + H]+, 331.4007; found,331.4014.

Dichloride N-[2-(4-Piperazinium-1-yl)-4-oxo-3H-quinazolin-6-yl]-piperidinium-4-carboxamide (12c). Yield 96%. mp >300 °C-(decomposes); 1H NMR δ (500 MHz, d6-DMSO) 1.83−1.99 (m, 2H, C3H′ piperidium), 1.97−2.01 (m, 2 H, C3H″ piperidium), 2.69−2.75 (m, 2 H, C2H′ piperidium), 2.88−2.93 (m, 2 H, C2H″piperidium), 3.23 (br s, 4 H, CH2 piperazinium), 3.53 (dd, J = 27.2,10.1 Hz, 1 H, C4H piperidium), 3.98 (br s, 4 H, CH2 piperazinium),7.70 (s, 1 H, C8H quinazoline), 7.89 (d, J = 6.5 Hz, 1 H, C7Hquinazoline), 8.43 (s, 1 H, C5H quinazoline), 8.83 (s, 1 H, +NH′piperidium), 9.19 (s, 1 H, +NH” piperidium), 9.66 (s, 2 H, N+H2piperazinium), 10.60 (s, 1 H, NH carboxamide). HRMS (FAB)calculated for C18H25N6O2 [M + H]+, 357.4390; found, 357.4379.

In Vitro Biology. Functional activity was determined by measuringthe inhibition of ADP induced platelet aggregation in human platelet-rich plasma (PRP) by Born’s method.67 Mode of action for somecompounds was subsequently revealed in vitro by measuring the abilityof compounds to inhibit the binding of fluoresceinisothiocyanate-labeled fibrinogen (FITC-Fg)85 to αIIbβ3 in a suspension of humanwashed platelets.38

■ ASSOCIATED CONTENT*S Supporting InformationThe Supporting Information is available free of charge on theACS Publications website at DOI: 10.1021/acs.jmed-chem.5b00865.

Additional figures and tables illustrating the data sets,pharmacophores and molecular docking results (PDF)Synthesized compounds and corresponding data (CSV)Ligands of open form of αIIbβ3 with known affinity forαIIbβ3 which were used for QSAR, pharmacophoremodeling and molecular docking studies (CSV)Ligands of open form of αIIbβ3 with known anti-aggregation activity which were used for QSAR modeling(CSV)Hits retrieved in virtual screening for the ligands of theopen form of αIIbβ3 (CSV)Hits retrieved in virtual screening for the ligands of theclosed form of αIIbβ3 (CSV)

■ AUTHOR INFORMATIONCorresponding Authors*E-mail: [email protected].*E-mail: [email protected] authors declare no competing financial interest.

■ ACKNOWLEDGMENTSThe financial support from bilateral Moldova-Ukraine project14.820.18.04.05/U and No M/161-2014) is acknowledged. P.P.thanks French Embassy in Ukraine for scholarship for youngscientists. T.K. thanks French Embassy in Ukraine for Ph.D.scholarship. The authors are grateful to Dr. V.V. Polovinko(Enamine LTD, Ukraine) for the 1H NMR spectra and Dr. I.M.Rakipov (A.V. Bogatsky Physico-Chemical Institute of theNational Academy of Sciences of Ukraine) for the FAB-MSspectra. The authors also thank Dr. D. Horvath and Dr. G.Marcou from the University of Strasbourg for valuablecomments and fruitful discussion. We acknowledge Inte:Ligand



7691

http://pubs.acs.org

http://pubs.acs.org/doi/abs/10.1021/acs.jmedchem.5b00865

http://pubs.acs.org/doi/abs/10.1021/acs.jmedchem.5b00865

http://pubs.acs.org/doi/suppl/10.1021/acs.jmedchem.5b00865/suppl_file/jm5b00865_si_001.pdf

http://pubs.acs.org/doi/suppl/10.1021/acs.jmedchem.5b00865/suppl_file/jm5b00865_si_002.csv





mailto:[email protected]

mailto:[email protected]


company for providing us with LigandScout software. Authorsalso thank reviewers whose comments and remarks helped usto significantly improve the manuscript quality and consistence.

■ REFERENCES(1) Libby, P.; Theroux, P. Pathophysiology of Coronary ArteryDisease. Circulation 2005, 111, 3481−3488.(2) Davì, G.; Patrono, C. Platelet Activation and Atherothrombosis.N. Engl. J. Med. 2007, 357, 2482−2494.(3) Ruf, A.; Patscheke, H. Flow Cytometric Detection of ActivatedPlatelets: Comparison of Determining Shape Change, FibrinogenBinding, and P-Selectin Expression. Semin. Thromb. Hemostasis 1995,21, 146−151.(4) Massberg, S.; Brand, K.; Gruner, S.; Page, S.; Muller, E.; Muller,I.; Bergmeier, W.; Richter, T.; Lorenz, M.; Konrad, I.; Nieswandt, B.;Gawaz, M. A Critical Role of Platelet Adhesion in the Initiation ofAtherosclerotic Lesion Formation. J. Exp. Med. 2002, 196, 887−896.(5) Furie, B.; Furie, B. C. Mechanisms of Thrombus Formation. N.Engl. J. Med. 2008, 359, 938−949.(6) Offermanns, S. Activation of Platelet Function Through GProtein−Coupled Receptors. Circ. Res. 2006, 99, 1293−1304.(7) Wang, X.; Dorsam, R. T.; Lauver, A.; Wang, H.; Barbera, F. A.;Gibbs, S.; Varon, D.; Savion, N.; Friedman, S. M.; Feuerstein, G. Z.Comparative Analysis of Various Platelet Glycoprotein IIb/IIIaAntagonists on Shear-Induced Platelet Activation and Adhesion. J.Pharmacol. Exp. Ther. 2002, 303, 1114−1120.(8) Clappers, N.; Brouwer, M. A.; Verheugt, F. W. A. Antiplatelettreatment for coronary heart disease. Heart 2007, 93, 258−265.(9) Freson, K.; Thys, C.; Wittevrongel, C.; Geet, C. V. Mechanismsof Action and Targets for Actual and Future Antiplatelet Drugs. Mini-Rev. Med. Chem. 2006, 6, 719−726.(10) Andronati, S. A.; Karaseva, T. L.; Krysko, A. A. Peptidomimetics- Antagonists of the Fibrinogen Receptors: Molecular Design,Structures, Properties and Therapeutic Applications. Curr. Med.Chem. 2004, 11, 1183−1211.(11) Hartman, G. D.; Egbertson, M. S.; Halczenko, W.; Laswell, W.L.; Duggan, M. E.; Smith, R. L.; Naylor, A. M.; Manno, P. D.; Lynch,R. J. Non-peptide fibrinogen receptor antagonists. 1. Discovery anddesign of exosite inhibitors. J. Med. Chem. 1992, 35, 4640−4642.(12) Scarborough, R. M.; Naughton, M. A.; Teng, W.; Rose, J. W.;Phillips, D. R.; Nannizzi, L.; Arfsten, A.; Campbell, A. M.; Charo, I. F.Design of potent and specific integrin antagonists. Peptide antagonistswith high specificity for glycoprotein IIb-IIIa. J. Biol. Chem. 1993, 268,1066−1073.(13) Coller, B. S. A new murine monoclonal antibody reports anactivation-dependent change in the conformation and/or micro-environment of the platelet glycoprotein IIb/IIIa complex. J. Clin.Invest. 1985, 76, 101−108.(14) Antoniucci, D. Differences among GP IIb/IIIa inhibitors:different clinical benefits in non-ST-segment elevation acute coronarysyndrome percutaneous coronary intervention patients. Eur. Heart J.Suppl. 2007, 9, A32−A36.(15) Berkowitz, S. D.; Sane, D. C.; Sigmon, K. N.; Shavender, J. H.;Harrington, R. A.; Tcheng, J. E.; Topol, E. J.; Califf, R. M. Occurrenceand clinical significance of thrombocytopenia in a populationundergoing high-risk percutaneous coronary revascularization. J. Am.Coll. Cardiol. 1998, 32, 311−319.(16) McClure, M. W.; Berkowitz, S. D.; Sparapani, R.; Tuttle, R.;Kleiman, N. S.; Berdan, L. G.; Lincoff, A. M.; Deckers, J.; Diaz, R.;Karsch, K. R.; Gretler, D.; Kitt, M.; Simoons, M.; Topol, E. J.; Califf, R.M.; Harrington, R. A. Clinical Significance of ThrombocytopeniaDuring a Non−ST-Elevation Acute Coronary Syndrome: The PlateletGlycoprotein IIb/IIIa in Unstable Angina: Receptor SuppressionUsing Integrilin Therapy (PURSUIT) Trial Experience. Circulation1999, 99, 2892−2900.(17) Bougie, D. W.; Wilker, P. R.; Wuitschick, E. D.; Curtis, B. R.;Malik, M.; Levine, S.; Lind, R. N.; Pereira, J.; Aster, R. H. Acutethrombocytopenia after treatment with tirofiban or eptifibatide is

associated with antibodies specific for ligand-occupied GPIIb/IIIa.Blood 2002, 100, 2071−2076.(18) Bougie, D. W.; Rasmussen, M.; Zhu, J.; Aster, R. H. Antibodiescausing thrombocytopenia in patients treated with RGD-mimeticplatelet inhibitors recognize ligand-specific conformers of αIIb/β3integrin. Blood 2012, 119, 6317−6325.(19) Blue, R.; Murcia, M.; Karan, C.; Jirouskova, M.; Coller, B. S.Application of high-throughput screening to identify a novel αIIb-specific small- molecule inhibitor of αIIbβ3-mediated plateletinteraction with fibrinogen. Blood 2008, 111, 1248−1256.(20) Zhu, J.; Zhu, J.; Negri, A.; Provasi, D.; Filizola, M.; Coller, B. S.;Springer, T. A. Closed headpiece of integrin αIIbβ3 and its complexwith an αIIbβ3-specific antagonist that does not induce opening. Blood2010, 116, 5050−5059.(21) Zhu, J.; Choi, W.-S.; McCoy, J. G.; Negri, A.; Zhu, J.; Naini, S.;Li, J.; Shen, M.; Huang, W.; Bougie, D.; Rasmussen, M.; Aster, R.;Thomas, C. J.; Filizola, M.; Springer, T. A.; Coller, B. S. Structure-Guided Design of a High-Affinity Platelet Integrin αIIbβ3 ReceptorAntagonist That Disrupts Mg2+ Binding to the MIDAS. Sci. Transl.Med. 2012, 4, 125ra32−125ra32.(22) Jiang, J.-k.; McCoy, J. G.; Shen, M.; LeClair, C. A.; Huang, W.;Negri, A.; Li, J.; Blue, R.; Harrington, A. W.; Naini, S.; David Iii, G.;Choi, W.-S.; Volpi, E.; Fernandez, J.; Babayeva, M.; Nedelman, M. A.;Filizola, M.; Coller, B. S.; Thomas, C. J. A novel class of iondisplacement ligands as antagonists of the αIIbβ3 receptor that limitconformational reorganization of the receptor. Bioorg. Med. Chem. Lett.2014, 24, 1148−1153.(23) Gaulton, A.; Bellis, L. J.; Bento, A. P.; Chambers, J.; Davies, M.;Hersey, A.; Light, Y.; McGlinchey, S.; Michalovich, D.; Al-Lazikani, B.;Overington, J. P. ChEMBL: a large-scale bioactivity database for drugdiscovery. Nucleic Acids Res. 2012, 40, D1100−D1107.(24) Krysko, A. A.; Chugunov, B. M.; Malovichko, O. L.; Andronati,S. A.; Kabanova, T. A.; Karaseva, T. L.; Kiriyak, A. V. Novel fibrinogenreceptor antagonists. RGDF mimetics, derivatives of 4-(isoindoline-5-yl)amino-4-oxobutyric acid. Bioorg. Med. Chem. Lett. 2004, 14, 5533−5535.(25) Andronati, S. A.; Krys’ko, A. A.; Chugunov, B. M.; Kabanova, T.A.; Artemenko, A. G. 4-(Isoindolin-5-yl)amino-4-oxobutyryl-β-alkox-yphenyl-β-alanines, new RGDF mimetics. Synthesis, affinity tofibrinogen receptors, antiaggregatory properties, and their correlationwith the hydrophobicity. Russ. J. Org. Chem. 2006, 42, 1174−1182.(26) Malovichko, O. L.; Petrus, A. S.; Krysko, A. A.; Kabanova, T. A.;Andronati, S. A.; Karaseva, T. L.; Kiriyak, A. V. Derivatives of 7-amino-1,2,3,4-tetrahydroisoquinoline and isophthalic acids as novel fibri-nogen receptor antagonists. Bioorg. Med. Chem. Lett. 2006, 16, 5294−5297.(27) Malovichko, O. L.; Krysko, A. A.; Kabanova, T. A.; Andronati, S.A.; Grishkovets, V. I.; Kachala, V. V.; Panov, D. A. Derivatives of1,2,3,4-tetrahydroisoquinoline-7-carboxylic Acid as Novel FibrinogenReceptor Antagonists. Med. Chem. 2009, 5, 158−164.(28) Krysko, A. A.; Krysko, O. L.; Kabanova, T. A.; Andronati, S. A.;Kabanov, V. M. Derivatives of tetrahydroisoquinoline: Synthesis andinitial evaluation of novel non-peptide antagonists of the αIIbβ3-integrin. Bioorg. Med. Chem. Lett. 2010, 20, 4444−4446.(29) Krysko, A. A.; Samoylenko, G. V.; Polishchuk, P. G.; Fonari, M.S.; Kravtsov, V. C.; Andronati, S. A.; Kabanova, T. A.; Lipkowski, J.;Khristova, T. M.; Kuz’min, V. E.; Kabanov, V. M.; Krysko, O. L.;Varnek, A. A. Synthesis, biological evaluation, X-ray molecularstructure and molecular docking studies of RGD mimetics containing6-amino-2,3-dihydroisoindolin-1-one fragment as ligands of integrinαIIbβ3. Bioorg. Med. Chem. 2013, 21, 4646−4661.(30) Breiman, L. Random forests. Mach. Learning 2001, 45, 5−32.(31) Kuz’min, V. E.; Artemenko, A. G.; Polischuk, P. G.; Muratov, E.N.; Khromov, A. I.; Liahovskiy, A. V.; Andronati, S. A.; Makan, S. Y.Hierarchic System of QSAR Models (1D-4D) on the Base of SimplexRepresentation of Molecular Structure. J. Mol. Model. 2005, 11, 457−467.



7692


(32) Kuz’min, V. E.; Artemenko, A. G.; Muratov, E. N. HierarchicalQSAR technology based on the Simplex representation of molecularstructure. J. Comput.-Aided Mol. Des. 2008, 22, 403−421.(33) Laboratoire de Chemoinformatique UMR 7140 CNRS,Universite de Strasbourg, 1, rue B. Pascal, Strasbourg 67000, France.http://infochim.u-strasbg.fr/.(34) Bonachera, F.; Horvath, D. Fuzzy Tricentric PharmacophoreFingerprints. 2. Application of Topological Fuzzy PharmacophoreTriplets in Quantitative Structure−Activity Relationships. J. Chem. Inf.Model. 2008, 48, 409−425.(35) Bonachera, F.; Parent, B.; Barbosa, F.; Froloff, N.; Horvath, D.Fuzzy Tricentric Pharmacophore Fingerprints. 1. Topological FuzzyPharmacophore Triplets and Adapted Molecular Similarity ScoringSchemes. J. Chem. Inf. Model. 2006, 46, 2457−2477.(36) Mehrotra, M. M.; Heath, J. A.; Smyth, M. S.; Pandey, A.; Rose, J.W.; Seroogy, J. M.; Volkots, D. L.; Nannizzi-Alaimo, L.; Park, G. L.;Lambing, J. L.; Hollenbach, S. J.; Scarborough, R. M. Discovery ofNovel 2,8-Diazaspiro[4.5]decanes as Orally Active Glycoprotein IIb-IIIa Antagonists†. J. Med. Chem. 2004, 47, 2037−2061.(37) Wagner, C. L.; Mascelli, M. A.; Neblock, D. S.; Weisman, H. F.;Coller, B. S.; Jordan, R. E. Analysis of GPIIb/IIIa receptor number byquantification of 7E3 binding to human platelets. Blood 1996, 88,907−914.(38) Xia, Z.; Wong, T.; Liu, Q.; Kasirer-Friede, A.; Brown, E.;Frojmovic, M. M. Optimally functional fluorescein isothiocyanate-labelled fibrinogen for quantitative studies of binding to activatedplatelets and platelet aggregation. Br. J. Haematol. 1996, 93, 204−214.(39) Holmes, M. B.; Sobel, B. E.; Schneider, D. J. Variable responsesto inhibition of fibrinogen binding induced by tirofiban andeptifibatide in blood from healthy subjects. Am. J. Cardiol. 1999, 84,203−207.(40) Wolber, G.; Langer, T. LigandScout: 3-D PharmacophoresDerived from Protein-Bound Ligands and Their Use as VirtualScreening Filters. J. Chem. Inf. Model. 2005, 45, 160−169.(41) Springer, T. A.; Zhu, J.; Xiao, T. Structural basis for distinctiverecognition of fibrinogen γC peptide by the platelet integrin αIIbβ3. J.Cell Biol. 2008, 182, 791−800.(42) Cheng, S.; Craig, W. S.; Mullen, D.; Tschopp, J. F.; Dixon, D.;Pierschbacher, M. D. Design and synthesis of novel cyclic RGD-containing peptides as highly potent and selective integrin.alpha.IIb.-beta.3 antagonists. J. Med. Chem. 1994, 37, 1−8.(43) Ojima, I.; Dong, Q.; Chakravarty, S.; Peerschke, E.; Hwang, S.M.; Wong, A. S. Design, synthesis and SAR of RGD peptide hybrids ashighly efficient inhibitors of platelet aggregation. Bioorg. Med. Chem.Lett. 1995, 5, 1941−1946.(44) FlexX, 2.1.2; BioSolveIT GmbH: Sankt Augustin, Germany.(45) Molecular Operating Environment (MOE), 2008.10; ChemicalComputing Group Inc.: Montreal, QC, Canada.(46) Korb, O.; Stutzle, T.; Exner, T. In PLANTS: Application of AntColony Optimization to Structure-Based Drug Design. Ant ColonyOptimization and Swarm Intelligence; Dorigo, M., Gambardella, L.,Birattari, M., Martinoli, A., Poli, R., Stutzle, T., Eds.; Springer: Berlin/Heidelberg, 2006; Vol. 4150, pp 247−258.(47) Korb, O.; Stutzle, T.; Exner, T. E. Empirical Scoring Functionsfor Advanced Protein−Ligand Docking with PLANTS. J. Chem. Inf.Model. 2009, 49, 84−96.(48) Korb, O.; Stutzle, T.; Exner, T. An ant colony optimizationapproach to flexible protein−ligand docking. Swarm Intelligence 2007,1, 115−134.(49) Rognan, D. BioinfoDB: un inventaire de mole culescommercialement disponibles a des fins de criblage biologique. LaGazette du CINES 2005, 1−4.(50) ISIDA Predictor. http://infochim.u-strasbg.fr/cgi-bin/predictor.cgi?GeneralPropSelect=PhysProp.(51) Muratov, E. N.; Kuz’min, V. E.; Artemenko, A. G.; Kovdienko,N. A.; Gorb, L.; Hill, F.; Leszczynski, J. New QSPR equations forprediction of aqueous solubility for military compounds. Chemosphere2010, 79, 887−890.

(52) Sushko, I.; Novotarskyi, S.; Korner, R.; Pandey, A. K.;Cherkasov, A.; Li, J.; Gramatica, P.; Hansen, K.; Schroeter, T.;Muller, K.; Xi, L.; Liu, H.; Yao, X.; Oberg, T.; Hormozdiari, F.; Dao,P.; Sahinalp, C.; Todeschini, R.; Polishchuk, P.; Artemenko, A.;Kuz’min, V.; Martin, T. M.; Young, D. M.; Fourches, D.; Muratov, E.;Tropsha, A.; Baskin, I.; Horvath, D.; Marcou, G.; Muller, C.; Varnek,A.; Prokopenko, V. V.; Tetko, I. V. Applicability Domains forClassification Problems: Benchmarking of Distance to Models forAmes Mutagenicity Set. J. Chem. Inf. Model. 2010, 50, 2094−2111.(53) Poroikov, V.; Filimonov, D.; Lagunin, A.; Gloriozova, T.;Zakharov, A. PASS: identification of probable targets and mechanismsof toxicity. SAR QSAR Environ. Res. 2007, 18, 101−10.(54) Advanced and HTS Enamine Databases. http://www.enamine.net/index.php?option=com_content&task=view&id=22 (accessed 13June 2012).(55) REAL Enamine Database. http://www.enamine.net/index.php?option=com_content&task=view&id=8 (accessed 13 June 2012).(56) Irwin, J. J.; Shoichet, B. K. ZINC − A Free Database ofCommercially Available Compounds for Virtual Screening. J. Chem.Inf. Model. 2005, 45, 177−182.(57) Aoyama, T.; Ino, Y.; Ozeki, M.; Oda, M.; Sato, T.; Koshiyama,Y.; Suzuki, S.; Fujita, M. Pharmacological Studies of Fut-175,Nafamostat Mesilate 0.1. Inhibition of Protease Activity in Invitroand Invivo Experiments. Jpn. J. Pharmacol. 1984, 35, 203−227.(58) Negri, A.; Li, J.; Naini, S.; Coller, B. S.; Filizola, M. Structure-based virtual screening of small-molecule antagonists of plateletintegrin alphaIIbbeta3 that do not prime the receptor to bind ligand. J.Comput.-Aided Mol. Des. 2012, 26, 1005−15.(59) Paques, E. P.; Romisch, J. Comparative study on the in vitroeffectiveness of antithrombotic agents. Thromb. Res. 1991, 64, 11−21.(60) Tsuda, Y.; Nakahara, T.; Ueda, K.; Mori, A.; Sakamoto, K.; Ishii,K. Effect of Nafamostat on N-Methyl-D-aspartate-Induced RetinalNeuronal and Capillary Degeneration in Rats. Biol. Pharm. Bull. 2012,35, 2209−2213.(61) Inman, R.; Chiu, B. Nafamostat mesylate, a serine proteaseinhibitor, demonstrates novel antimicrobial properties and effective-ness in Chlamydia-induced arthritis. Arthritis Res. Ther. 2012, 14,R150.(62) Ohtake, Y.; Hirasawa, H.; Sugai, T.; Oda, S.; Shiga, H.; Matsuda,K.; Kitamura, N. Nafamostat mesylate as anticoagulant in continuoushemofiltration and continuous hemodiafiltration. Contrib. Nephrol.1991, 93, 215−7.(63) Nakae, H.; Tajimi, K. Pharmacokinetics of nafamostat mesilateduring continuous hemodiafiltration with a polyacrylonitrile mem-brane. Ther. Apheresis Dial. 2003, 7, 483−5.(64) Liverton, N. J.; Armstrong, D. J.; Claremon, D. A.; Remy, D. C.;Baldvin, J. J.; Lynch, R. J.; Zhang, G.; Gould, R. J. Nonpeptideglycoprotein IIb/IIIa inhibitors: Substituted quinazolinediones andquinazolinones as potent fibrinogen receptor antagonists. Bioorg. Med.Chem. Lett. 1998, 8, 483−486.(65) Westwood, R.; Tully, W. R.; Murdoch, R. Antihistaminic andbronchospasmolytic triazoloquinazolinones. US4350695 A, 21 Sep-tember, 1982.(66) Wang, Y.; Zhao, Y.; Sun, R.; Kong, W.; Wang, B.; Yang, G.; Li,Y. Discovery of novel antagonists of glycoprotein IIb/IIIa-mediatedplatelet aggregation through virtual screening. Bioorg. Med. Chem. Lett.2015, 25, 1249−1253.(67) Born, G. V. R. Aggregation of blood platelets by adenosinediphosphate and its reversal. Nature 1962, 194, 3.(68) Standartizer, 5.4; Chemaxon: Budapest, Hungary.(69) Instant JChem, 5.4; Chemaxon: Budapest, Hungary.(70) Muratov, E. N.; Artemenko, A. G.; Varlamova, E. V.; Polischuk,P. G.; Lozitsky, V. P.; Fedchuk, A. S.; Lozitska, R. L.; Gridina, T. L.;Koroleva, L. S.; Sil’nikov, V. N.; Galabov, A. S.; Makarov, V. A.;Riabova, O. B.; Wutzler, P.; Schmidtke, M.; Kuz’min, V. E. Per asperaad astra: application of Simplex QSAR approach in antiviral research.Future Med. Chem. 2010, 2, 1205−1226.(71) Polishchuk, P. G. Random Forest software (CF), version 2.13;2010.



7693

http://infochim.u-strasbg.fr/

http://infochim.u-strasbg.fr/cgi-bin/predictor.cgi?GeneralPropSelect=PhysProp

http://infochim.u-strasbg.fr/cgi-bin/predictor.cgi?GeneralPropSelect=PhysProp

http://www.enamine.net/index.php?option=com_content&task=view&id=22





(72) Svetnik, V.; Liaw, A.; Tong, C.; Culberson, J. C.; Sheridan, R. P.;Feuston, B. P. Random forest: A classification and regression tool forcompound classification and QSAR modeling. J. Chem. Inf. Model.2003, 43, 1947−1958.(73) Polishchuk, P. G.; Muratov, E. N.; Artemenko, A. G.; Kolumbin,O. G.; Muratov, N. N.; Kuz’min, V. E. Application of Random ForestApproach to QSAR Prediction of Aquatic Toxicity. J. Chem. Inf. Model.2009, 49, 2481−2488.(74) Kovdienko, N. A.; Polishchuk, P. G.; Muratov, E. N.;Artemenko, A. G.; Kuz’min, V. E.; Gorb, L.; Hill, F.; Leszczynski, J.Application of Random Forest and Multiple Linear RegressionTechniques to QSPR Prediction of an Aqueous Solubility for MilitaryCompounds. Mol. Inf. 2010, 29, 394−406.(75) Ognichenko, L. N.; Kuz’min, V. E.; Gorb, L.; Hill, F. C.;Artemenko, A. G.; Polischuk, P. G.; Leszczynski, J. QSPR Prediction ofLipophilicity for Organic Compounds Using Random ForestTechnique on the Basis of Simplex Representation of MolecularStructure. Mol. Inf. 2012, 31, 273−280.(76) Hansen, K.; Mika, S.; Schroeter, T.; Sutter, A.; ter Laak, A.;Steger-Hartmann, T.; Heinrich, N.; Muller, K.-R. Benchmark Data Setfor in Silico Prediction of Ames Mutagenicity. J. Chem. Inf. Model.2009, 49, 2077−2081.(77) Filter, 2.0.2; OpenEye Scientific Software, Inc.: Santa Fe, NM,2010.(78) Omega, 2.4.3; OpenEye Scientific Software, Inc.: Santa Fe, NM,2010.(79) OEChem, 1.7.4; OpenEye Scientific Software, Inc.: Santa Fe,NM, 2010.(80) cxcalc, 5.4; Chemaxon: Budapest, Hungary.(81) ten Brink, T.; Exner, T. pKa based protonation states andmicrospecies for protein−ligand docking. J. Comput.-Aided Mol. Des.2010, 24, 935−942.(82) Schrodinger, L. L. C. The PyMOL Molecular Graphics System,version 1.3r1; 2010.(83) Askew, B. C.; Bednar, R. A.; Bednar, B.; Claremon, D. A.; Cook,J. J.; McIntyre, C. J.; Hunt, C. A.; Gould, R. J.; Lynch, R. J.; Lynch, J. J.;Gaul, S. L.; Stranieri, M. T.; Sitko, G. R.; Holahan, M. A.; Glass, J. D.;Hamill, T.; Gorham, L. M.; Prueksaritanont, T.; Baldwin, J. J.;Hartman, G. D. Non-Peptide Glycoprotein IIb/IIIa Inhibitors. 17.Design and Synthesis of Orally Active, Long-Acting Non-PeptideFibrinogen Receptor Antagonists. J. Med. Chem. 1997, 40, 1779−1788.(84) Duggan, M. E.; Duong, L. T.; Fisher, J. E.; Hamill, T. G.;Hoffman, W. F.; Huff, J. R.; Ihle, N. C.; Leu, C.-T.; Nagy, R. M.;Perkins, J. J.; Rodan, S. B.; Wesolowski, G.; Whitman, D. B.; Zartman,A. E.; Rodan, G. A.; Hartman, G. D. Nonpeptide αvβ3 Antagonists. 1.Transformation of a Potent, Integrin-Selective αIIbβ3 Antagonist intoa Potent αvβ3 Antagonist. J. Med. Chem. 2000, 43, 3736−3745.(85) Hantgan, R. R. An investigation of fibrin-platelet adhesiveinteractions by microfluorimetry. Biochim. Biophys. Acta, Mol. Cell Res.1987, 927, 55−64.



7694


Documents

Design, Virtual Screening, and Synthesis of Antagonists of