P2Y1 antagonists: Combining receptor-based modeling and QSAR for a quantitative prediction of the biological activity based on consensus scoring

Article abstract of DOI:10.1021/jm0700971

P2Y₁ is an ADP-activated G protein-coupled receptor (GPCR). Its antagonists impede platelet aggregation in vivo and are potential antithrombotic agents. Combining ligand and structure-based modeling we generated a consensus model (LIST-CM) corr

Full text of DOI:10.1021/jm0700971

J. Med. Chem. 2007, 50, 3229-3241
3229
P2Y1 Antagonists: Combining Receptor-Based Modeling and QSAR for a Quantitative
Prediction of the Biological Activity Based on Consensus Scoring
Stefano Costanzi,*,† Irina G. Tikhonova,† Michihiro Ohno,‡ Eun Joo Roh,‡ Bhalchandra V. Joshi,‡ Anny-Odile Colson,†
Dayle Houston,§ Savitri Maddileti,§ T. Kendall Harden,§ and Kenneth A. Jacobson*,‡
Laboratory of Biological Modeling and Molecular Recognition Section, National Institute of Diabetes and DigestiVe and Kidney Diseases,
NIH, Bethesda, Maryland 20892, and Department of Pharmacology, School of Medicine, UniVersity of North Carolina, Chapel Hill,
North Carolina 27599
ReceiVed January 24, 2007
P2Y1 is an ADP-activated G protein-coupled receptor (GPCR). Its antagonists impede platelet aggregation
in ViVo and are potential antithrombotic agents. Combining ligand and structure-based modeling we generated
a consensus model (LIST-CM) correlating antagonist structures with their potencies. We docked 45 antagonists
into our rhodopsin-based human P2Y1 homology model and calculated docking scores and free binding
energies with the Linear Interaction Energy (LIE) method in continuum-solvent. The resulting alignment
was also used to build QSAR based on CoMFA, CoMSIA, and molecular descriptors. To benefit from the
strength of each technique and compensate for their limitations, we generated our LIST-CM with a PLS
regression based on the predictions of each methodology. A test set featuring untested substituents was
synthesized and assayed in inhibition of 2-MeSADP-stimulated PLC activity and in radioligand binding.
LIST-CM outperformed internal and external predictivity of any individual model to predict accurately the
potency of 75% of the test set.
Introduction
we have extensively investigated the structure-activity relation-
ships (SAR) of A3P5P derivatives.8-15 The compounds which
were synthesized and tested include also acyclic derivatives,
such as 2, in which the ribose is substituted with acyclic
moieties,11,12 and conformationally locked derivatives, such as
3a, 3b (MRS2279), and 3c (MRS2500).12-14 In the latter
derivatives, the ribose is replaced with a bicyclo[3.1.0]hexane
ring, also known as a Northern (N)-methanocarba ring system,
which constrains the pseudosugar in the 2′-exo conformation
which is preferred by the receptor binding site. These derivatives
are competitive in binding and several have been shown to be
useful as tracers in radioactive form.16 Compound 3c was shown
to be effective in reducing thrombus formation in vivo and to
be more resistant to degradation than the related 9-riboside
derivatives.6
P2Y receptors are a family of nucleotide-activated G pro-
tein-coupled receptors (GPCRs). Eight subtypes, divided into
two subgroups, have been identified in human: the P2Y1
subgroup encompasses P2Y1, P2Y2, P2Y4, P2Y6, and P2Y11,
which couple to the stimulation of PLC through Gq, while the
P2Y12 subgroup encompasses P2Y12, P2Y13, and P2Y14, which
couple to the inhibition of adenylyl cyclase through Gi. The
two subgroups differ significantly in overall sequence similarity
and in their proposed mechanisms of ligand recognition. Natural
ligands of the P2Y receptors include ATP, ADP, UTP, UDP,
and UDP-glucose.1,2
The P2Y1 and P2Y12 receptors have both been identified in
platelets, where their activation by adenosine diphosphate (ADP,
1a) mediates a proaggregatory response through different
pathways. Activation of P2Y1 leads to shape change and rapid
aggregation, while activation of P2Y12 causes amplification of
the response and sustained aggregation.3,4 Synergistic activation
of both receptors is required to induce ADP-mediated platelet
aggregation. Thus, antagonizing each subtype separately has
antithrombotic effects. Antagonism of the P2Y12 receptor is a
validated target in antithrombotic therapy, with the widespread
use of the selective antagonist clopidogrel, which acts through
a metabolite that irreversibly binds to the P2Y12 receptor.5
Recently, the P2Y1 receptor has been investigated in the same
therapeutic context, based on pharmacological studies using both
novel ligands and genetic deletion of the P2Y1 receptor in mice.6
While ADP (1a) is the natural agonist, adenosine 3′,5′-
bisphosphate (A3P5P, 1b) was found to act as a competitive
antagonist of the human P2Y1 receptor (Chart 1).7 Since then,
In the present work, we combined receptor-based modeling
and QSAR (quantitative structure-activity relationships) meth-
odologies in the attempt to generate models capable of cor-
relating the chemical structures of P2Y1 antagonists with their
biological activities. Such models would be of great assistance
in the drug discovery process, as the activity of new compounds
could be estimated in a quantitative manner before their
synthesis and testing.
Since a crystal structure is available only for bovine rhodopsin
in the GPCR family,17 we developed rhodopsin-based molecular
models of the P2Y1 receptor.1,11 Subsequently, we studied the
interactions of competitive antagonists with the putative binding
site of the receptor with an iterative approach which included
flexible molecular docking, mutagenesis, and chemical modi-
fication of the ligands.1,11,14,18
Docking scores alone are not a viable route for the prediction
of the affinity of new ligands and cannot be applied to the lead
optimization process.19,20 This is especially true when the
structure of the receptor is inferred by homology modeling rather
than being experimentally determined. However, experimentally
supported docking models reliably provide the bioactive con-
formation of the ligands and their binding modes into the
* Corresponding authors. (S.C.) Email: stefanoc@mail.nih.gov. Phone:
301-451-7353. Fax: 301-480-4586. (K.J.) Email: kajacobs@helix.nih.gov.
Phone: 301-496-9024. Fax: 301-480-8422.
† Laboratory of Biological Modeling, National Institute of Diabetes and
Digestive and Kidney Diseases.
‡ Molecular Recognition Section, National Institute of Diabetes and
Digestive and Kidney Diseases.
§ University of North Carolina, Chapel Hill.
10.1021/jm0700971 CCC: $37.00 © 2007 American Chemical Society
Published on Web 06/12/2007

3230 Journal of Medicinal Chemistry, 2007, Vol. 50, No. 14
Costanzi et al.
Chart 1
receptor binding pockets.19,21 Here we used the structural
alignment and the bioactive conformations resulting from
docking for the construction of QSAR models based either on
molecular fields or on molecular descriptors. Subsequently, in
order to benefit from the strength of each individual technique,
we combined the models based on receptor-ligand interactions
and those based on QSAR to generate a PLS model based on
consensus scoring.
To validate our consensus model, we assayed the antagonistic
potency of 8 analogues of A3P5P (1b), 7 of which were newly
synthesized, in the methanocarba and acyclic series. The
compounds were also tested for their binding affinity. In parallel
with the experimental measurements, we predicted theoretically
the potency of the new compounds. Finally, experimental and
theoretical values were compared to assess the predictive
capabilities of our models.
Throughout this paper (a) the statistical models linking the
biological activities (dependent variables) to the molecular
properties (independent variables) were always built by means
of partial least-square (PLS) regressions, (b) the degree of
correlation of experimental versus predicted activities was
always expressed in terms of the square of the correlation
coefficient (r2), which indicates the fraction of explained
variance, (c) the internal predictivity was always measured in
terms of cross-validated r2 (q2), after cross-validation with the
leave-one-out method, and (d) to facilitate the comparison
among receptors, we used the GPCR residue indexing system
as explained elsewhere.1
by the presence of two monophosphate groups and bear various
substitutions at the purine ring. The compounds are endowed
with potencies ranging from low nM to high µM. The most
potent is the (N)-methanocarba analogue 3c, which shows an
IC50 of 8.4 nM in inhibition of PLC activity and a Ki of 0.78
nM in binding studies (Table 1).14 A table reporting the
structures of all the compounds in the training set, their
potencies, and references to the original publications is supplied
as Supporting Information (Table S1).
Receptor-Based Molecular Modeling. We refined our model
of the P2Y1-3c complex by means of a cycle of Monte Carlo
Multiple Minima (MCMM)22 conformational analysis granting
full flexibility to the ligand and to the residues forming the
binding cavity (Figure 1a). Consistently with our previous
reports, the phosphates are coordinated by three cationic residues
located in TM3, TM6, and TM7, namely R128(3.29), K280-
(6.55), and R310(7.39), which proved fundamental for the
recognition of agonists and antagonists. The amino group at
the 6 position of the purine ring (N6) donates a hydrogen-bond
(H-bond) to Q307(7.36), while the nitrogen at the 1 position
(N1) accepts an H-bond from S314(7.43).
After the optimization of the P2Y1 model around 3c, we
automatically docked into the receptor each of the 45 compounds
of the training set. The docking experiments have been
performed by means of the Glide23 module of the Schro¨dinger
package granting full flexibility to the ligands. The receptor was
treated as a rigid grid with softened van der Waals (vdW)
potentials. To account for the flexibility of the receptor and
induced fit phenomena, we submitted the docking complexes
to an energy minimization cycle by means of the Liaison23
module of the Schro¨dinger package and subsequently recalcu-
lated the docking scores (LiaScore). The highest scoring poses
of the training set compounds aligned very well inside the P2Y1
binding cavity, establishing with the receptor the characteristic
interactions proposed by our experimentally supported model
(Figure 1b).
Results
The Training Set. We collected 45 P2Y1 antagonists
published by our group between 1998 and 2004. The antago-
nistic properties of all the compounds have been tested, in the
course of the years, in the laboratory of T. Kendall Harden for
the inhibition of the phospholipase C (PLC) activity stimulated
by 2-MeSADP.8-11,13,14 The 45 compounds are all A3P5P (2)
analogues and encompass ribose derivatives, anhydrohexitol
derivatives, acyclic derivatives, (N)-methanocarba derivatives,
and bicyclo[2.2.1]heptane derivatives. They are all characterized
Overall, the PLS regression of LiaScore resulted in predicted
potencies which did not significantly deviate from the experi-
mental potencies, as indicated by a root-mean-square error

P2Y1 Antagonists
Journal of Medicinal Chemistry, 2007, Vol. 50, No. 14 3231
Table 1. In Vitro Pharmacological Data at the Human P2Y1 Receptor for Inhibition of Radioligand Binding and for Inhibition of PLC Activity
Stimulated by 2-MeSADP
compound
R
binding, Ki, nMa,c
antagonism, IC50, nMb-d
3ae
3be
3ce
4
H
Cl
I
17.6 ( 2.713
2.5 ( 0.413
0.78 ( 0.0813
91 ( 1213
540 ( 133
650 ( 266
1430 ( 930
3.6 ( 0.713
80 ( 1013
330 ( 7013
95 ( 39
157 ( 6013
52 ( 113
8.4 ( 0.813
221 ( 3013
6900 ( 2500
1750 ( 1100
870 ( 180
48 ( 113
SCH3
CN
CONH2
CO2H
CH3
(CH2)5CH3
trans-CHdCH(CH2)3CH3
CtCH
5f
6f
7f
8
9
452 ( 22113
1870 ( 59013
93 ( 30
10
11e, f
12
13 f
14 f
15 f
16e, f
CtC(CH2)3CH3
C6H5
430 ( 20013
273 ( 26
2400 ( 60013
1940 ( 1000
1560 ( 540
620 ( 40
706 ( 219
76 ( 1015
66 ( 12
Cl
I
790 ( 1600
a Affinities determined by using [3H]3b in a radioligand binding assay. The human P2Y1 receptor was expressed to high levels in Sf9 insect cells with
a recombinant baculovirus. Membranes prepared from these cells were incubated for 30 min at 4 °C in the presence of ∼20 nM [3H]3b. b Antagonist IC50
values represent the concentrations needed to inhibit by 50% the effect elicited by 100 nM 2-MeSADP on the human P2Y1 receptor expressed in 1321N1
astrocytoma cells. c Mean ( SEM given for three separate determinations. d None of the compounds displayed agonist effects. e 3a, MRS2275; 3b, MRS2279;
3c, MRS2500; 11, MRS2611; 16, MRS2609. f Test set compounds.
Figure 1. (a) Details of the P2Y1 receptor binding site complexed with the antagonist 3c as obtained after fully flexible Monte Carlo conformational
search. A schematic representation of the receptor-ligand complex is given in the lower left inset. (b) Docking complexes of the P2Y1 receptor and
the 45 compounds of the training set. In the tube representations the receptor is colored according to residue positions, with a spectrum of colors
that ranges from red (N-terminus) to purple (C-terminus): TM1 is in orange, TM2 in ochre, TM3 in yellow, TM4 in green, TM5 in cyan, TM6 in
blue, TM7 in purple.
(RMSE) of 0.660. Thus, the molecules in the training set were
recognized as binders and their potencies were estimated with
a reasonably low error. However, the LiaScore model clearly
appeared unable to rank the compounds according to their
potency, as indicated by r2 and q2 values close to 0 (Table 2).
We also calculated the free energy of binding based on the linear
interaction energy method (LIE), also known as linear response
method (LRM),24,25 in a surface generalized Born (SGB)
continuum model for solvation as implemented in Liaison.26-28
The calculation, which relied on energy minimizations, yielded
a PLS model showing a lower RMSE and a significant
improvement of correlation and internal predictivity (Table 2).
However, with a q2 of 0.317 the model certainly did not offer
the reliability necessary to estimate the potency of novel
compounds.
Docking-Based QSAR: (a) Molecular Fields. Comparative
molecular fields analysis (CoMFA) and comparative molecular
similarity indices analysis (CoMSIA) are 3D-QSAR methods
intended to correlate the molecular features of a series of com-
pounds with their biological activities. CoMFA and CoMSIA
do not take into account receptor-ligand interactions but rely
only on the calculation of the molecular fields of the ligands
and their subsequent correlation, by PLS regression, to their biolog-
ical activities. In particular, CoMFA is based on the calculation

3232 Journal of Medicinal Chemistry, 2007, Vol. 50, No. 14
Costanzi et al.
Table 2. PLS Statistics for the Predictions of the Antagonistic Potency
of the Training Set Compounds. Plots of Experimental vs Calculated
Potencies for the Training Set Compounds Are Available as Supporting
Information (Figure S1)
In order to provide meaningful models, CoMFA and CoMSIA
are dependent on knowledge of the bioactive conformation of
the ligands and of their relative structural alignment. Here we
derived this information directly from our docking studies, thus
generating docking-based 3D-QSAR models.
method
r2 a
q2 b
RMSE c
LiaScore
SGB-LIE
CoMFA
CoMSIA
MOE 2D
MOE 3D
MOE 2D and 3D
consensus
0.080
0.441
0.904
0.815
0.766
0.690
0.876
0.922
0.017
0.317
0.590
0.513
0.372
0.563
0.688
0.890
0.660
0.514
0.229
0.314
0.573
0.383
0.243
0.192
When compared to docking scores and SBG-LIE calculations,
the PLS analysis of our docking-based CoMFA (r2 ) 0.904; q2
) 0.590) and CoMSIA (r2 ) 0.815; q2 ) 0.513) appeared
statistically more robust and showed a significantly higher
internal predictivity. The optimal number of components was
found to be 5 for CoMFA and 4 for CoMSIA.
a Square of the correlation coefficient. b Cross-validated square of the
correlation coefficient (leave-one-out method). c Root-mean-square error
(RMSE) in the non cross-validated analysis.
Since the QSAR models were built on the basis of docking,
we could dock the 3D molecular fields produced by CoMFA
and CoMSIA into the receptor binding pocket, thus evaluating
the complementarity of the ligand-based QSAR and our
proposed model of the ligand-receptor interactions. The
of steric and electrostatic fields, while CoMSIA considers also
hydrophobic, H-bond acceptor, and H-bond donor fields.
Figure 2. Contour maps of the CoMFA and CoMSIA models (StDev*Coeff) docked into the P2Y1 binding site. Compound 3c is shown as a
representative antagonist. The moieties of the antagonists responsible for the productions of the molecular fields are labeled in yellow. (a) CoMFA
electrostatic fields: positive charges increase the potency in the blue areas and decrease it in the red areas (contour levels 90% and 20%). (b) CoMFA
steric fields: bulky substituents increase the potency in the green areas and decrease it in the yellow areas (contour levels 80% and 20%). (c)
CoMSIA electrostatic fields: positive charges increase the potency in the blue areas and decrease it in the red areas (contour levels 90% and 20%).
(d) CoMSIA steric fields: bulky substituents increase the potency in the green areas and decrease it in the yellow areas (contour levels 90% and
20%). (e) CoMSIA H-bond fields: H-bond donors increase the potency in the cyan areas, while H-bond acceptors increase it in the magenta areas
(contour levels 85%). (f) CoMSIA hydrophobic fields: hydrophobic substituents increase the potency in the orange areas (contour levels 85%).

P2Y1 Antagonists
Journal of Medicinal Chemistry, 2007, Vol. 50, No. 14 3233
Table 3. Description and Relative Weight of the Molecular Descriptors
Used To Build the QSAR Model. All the Descriptors Were Calculated
with the QSAR Applications in MOE 200529,38
electrostatic fields suggest the importance of the aromaticity of
the purine ring, which binds in a region surrounded by aromatic
residues (Figure 2a and 2c, red fields). The N6 group shows
electrostatic (CoMFA, Figure 2a, blue field) or H-bond (CoM-
SIA, Figure 2e, cyan field) interactions with the essential residue
Q307(7.36). The N1 produces a strong H-bond acceptor field
corresponding to S314(7.43) (CoMSIA, Figure 2e, magenta
field). H-bond donor fields are detected by CoMSIA in
proximity of the phosphate groups and are complemented in
the receptor model by the three critical cationic residues R128-
(3.29), K280(6.55), and R310(7.39), and by H132 (3.33) (Figure
2e, magenta field). The latter is characteristic of P2Y1 and may
contribute to the selectivity of A3P5P analogues for this subtype.
In agreement with the arrangement of the cationic residues,
CoMSIA distinguishes between the orientations of the 3′-
phosphate that are beneficial (Figure 2c, red field) or detrimental
(Figure 2c, blue field) for the potency. Since all the studied
ligands are A3P5P analogues, CoMFA does not reveal the
contribution of the phosphates to the potency of the compounds.
If we add to the training set an analogue of 3c without the
phosphates (i.e., 2-iodo-N6-methyl-2′-deoxyadenosine) and we
assume the compound to be inactive, strong electrostatic fields
are produced in the regions corresponding to the phosphates
(Supporting Information, Figure S2c, lower right inset). CoMFA
and CoMSIA suggest that at the 2 position small substituents
are beneficial (Figure 2b, green fields, and Figure 2f, orange
field), while large groups, which cause steric clashes with
residues in TM1 and TM2, are not tolerated (Figure 2b and 2d,
yellow fields). Finally, CoMFA and CoMSIA reveal the positive
effect of the cyclopropane ring of the methanocarba pseudosugar
(Figure 2b and 2d, green fields). This is not due to a favorable
interaction with the binding site but to the locked Northern
conformation induced by the methanocarba system.
weight
namea
descriptiona
2D Physical Properties
0.005353 mr
0.082981 AM1_HF
molecular refractivity
AM1 heat of formation
0.173345 PM3_HOMO PM3 energy of the HOMO
0.005953 SlogP
0.211838 TPSA
0.016101 logS
log of the octanol/water partition coefficient39
polar surface area with group contributions40
log of the aqueous solubility41
2D SubdiVided Approximate Van der Waals Surface Areas38
0.215821 SlogP_VSA0
0.077908 SlogP_VSA1
0.430744 SlogP_VSA2
0.260983 SlogP_VSA3
0.135706 SlogP_VSA4
0.276779 SlogP_VSA5
0
SlogP_VSA6b
0.221864 SlogP_VSA7
0.735487 SlogP_VSA8
0.352781 SlogP_VSA9
2D Kier and Hall connectiVity
0.018255 chi0
atomic connectivity index (order 0)42
3D Surface Area and Shape Descriptors
0.241597 ASA
0.080319 pmiX
0.49167 pmiY
0.205061 pmiZ
water accessible surface area
x component of the principal moment of inertia
y component of the principal moment of inertia
z component of the principal moment of inertia
van der Waals surface area
1
VSA
3D Conformation Dependent Charge Descriptors
0.053425 ASA_H
0.059199 ASA_P
0.04164 dipole X
0.013659 dipole Y
0.012961 dipole Z
water accessible surface area of hydrophobic atoms
water accessible surface area of polar atoms
x component of the dipolar moment
y component of the dipolar moment
z component of the dipolar moment
a More details on the descriptors are available at: http://www.chem-
comp.com/journal/descr.htm. b SlogP_VSA6 was 0 for all the compounds
in the training set and was not used in the PLS analysis.
Docking-Based QSAR: (b) Molecular Descriptors. Alter-
natively, using the QSAR applications of the molecular operating
environment (MOE 2005),29 we have calculated 184 2D
descriptors and 67 3D descriptors for the compounds of the
training set. While 2D descriptors do not depend on the
conformation of the molecules, 3D descriptors are always
conformation-dependent and, in some cases, alignment-depend-
ent. Also in this case, we made use of the putative bioactive
conformations and the 3D alignment of the molecules resulting
from the docking studies.
QSAR with 2D descriptors, MOE QSAR with 3D descriptors,
and MOE QSAR with 2D and 3D descriptors. Remarkably, the
consensus model resulted in significantly more robust statis-
tics (Table 2). Notably, with a q2 of 0.890, this model showed
an internal predictivity superior to that of any individual
methodology.
The Test Set. To measure the external predictivity of our
models, we chose a particularly challenging test set of eight
compounds to be synthesized and tested biologically, each of
which showed at least one substituent not featured in the training
set.
The eight compounds of the test set belong either to the
methanocarba or to the acyclic series (Table 1, compounds 5,
6, 7, 11, 13, 14, 15, 16). In particular, six compounds are
analogues of 3c bearing different substituents at the 2 position
or lacking the 5′-phosphate, while the 2 remaining compounds
belong to the acyclic series and bear phosphonate groups at the
3′ and 5′ positions. To avoid any bias in the construction of the
models, the potencies of the compounds were quantitatively
predicted before the compounds were tested.
The Test Set: Synthesis. The (N)-methanocarba bisphos-
phate analogues 5-7, 11, and 13 were synthesized using a
Mitsunobu reaction to combine adenine and bicyclohexane
moieties.13,15,30 The functional group at the 2 position was
formed by both prefunctionalization of the adenine moiety and
later manipulation following condensation with the pseudoribose
ring. The cyano group was introduced by the reaction of zinc
cyanide on the corresponding 2-iodo intermediate 17 using
palladium chemistry in tetramethylurea (Scheme 1).14 The Pd-
(0) complex was generated in situ from tri-(2-furyl)phosphine
We identified the 16 2D descriptors and 10 3D descriptors
that best correlate with the potency of the training set com-
pounds. In particular, the 2D descriptors included subdivided
surface areas, Kier and Hall connectivity indices, and six
different physical properties, while the 3D descriptors included
conformation-dependent charge descriptors as well as surface
area and shape descriptors (Table 3).
PLS analysis (Table 2) showed that use of the 3D descriptors
alone (q2 ) 0.563) yielded a higher internal predictivity than
2D descriptors alone (q2 ) 0.372). However, the best result
was obtained from the combination of 2D and 3D descriptors
(q2 ) 0.688).
Combining QSAR and Docking Scores in a Ligand and
Structure-Based Consensus Model (LIST-CM). Combining
the scores obtained with different methodologies may result in
a significant improvement of the predictivity over the individual
techniques (see Discussion). Thus, we treated the predictions
obtained with the individual methodologies as independent
descriptors and correlated them to the potency of the compounds
through PLS regression, generating a consensus model (LIST-
CM) based on LiaScore, SGB-LIE, CoMFA, CoMSIA, MOE

3234 Journal of Medicinal Chemistry, 2007, Vol. 50, No. 14
Costanzi et al.
Scheme 1
Scheme 2
Scheme 3
and Pd2(dba)3 (tris(dibenzylideneacetone)dipalladium(0) chlo-
roform adduct). The cyano group was later converted in TFA
to the carboxamide 6 or hydrolyzed to give the carboxylic acid
derivative 7. A 2-iodo intermediate reacted with trimethylsilyl-
acetylene followed by deprotection to provide 11 (Scheme 2).
A 2-phenyladenine intermediate 25 was generated by a Suzuki
coupling reaction followed by introduction of the N6-Me group,
deprotection at the 9 position, condensation with the prephos-
phorylated bicyclic ring system, and final deprotection to yield
13 (Scheme 3). To probe the need for bisphosphorylation, the
5′-hydroxyl analogue 14 was prepared using a selective alcohol
protection approach with phosphorylation and deprotection as
the final steps (Scheme 4). Compound 14 also served as a
precursor for the enzymatic synthesis of a radiolabeled form
of 3c.16
The synthesis of acyclic phosphonate derivative 16 (Scheme
5) started with coupling of 2-iodo-6-chloro purine12,31 with the
alcohol15 36 by a Mitsunobu reaction to afford 37 in 52% yield.

P2Y1 Antagonists
Journal of Medicinal Chemistry, 2007, Vol. 50, No. 14 3235
Scheme 4
Scheme 5
The displacement of the 6-chloro group with methylamine
followed by treatment with excess of iodotrimethylsilane
afforded the requisite 2-iodo phosphonate analogue 16.
The Test Set: Pharmacological Activity. Unlike 2-halo
substituents, the cyano moiety and polar carboxy and carbox-
amido substituents at the 2 position (5-7, Table 1) were poorly
tolerated by the P2Y1 receptor. Interestingly, among these three
new molecules, compound 5, which bears the 2-cyano substitu-
ent, showed the highest binding affinity but the lowest antago-
nistic potency, while the opposite happened for 7, which bears
the 2-carboxy substituent.
The 2-ethynyl-substituted compound 11 proved to have higher
affinity and potency than the corresponding compound 12, which
bears a longer hexynyl chain. This is consistent with the
previously observed tendency of the P2Y1 receptor to prefer
small substituents at the 2 position, as in 8, rather than larger
chains, as in 9, 10 or 12.13 This trend also appears evident upon
analysis of the CoMFA and CoMSIA steric contours (Figure
2b and 2d). Consistently, the 2-phenyl-substituted compound
13 showed lower affinity and potency than the less bulky
compound 11.
not lead to a more potent inhibitor as it did in the (N)-
methanocarba series.
The Test Set: Prediction and Validation. The predicted
antagonistic potencies of the compounds in the test set,
expressed as pIC50, are reported in Table 4. Plots of experimental
versus predicted potencies for the test set compounds, together
with a synoptic bar graph representation, are given as Supporting
Information (Figures S3 and S4).
The root-mean-square error of prediction (RMSEP) proved
to be low for all the methodologies. Thus, all compounds were
correctly recognized as P2Y1 antagonists. However, none of the
models showed good ranking capabilities over the whole test
set, measured in terms of r2.
All the methodologies showed outliers, defined as compounds
with an error of prediction higher than 1 log unit. In particular,
the LiaScore model showed only 1 outlier and proved to be the
methodology endowed with the lowest RMSEP. Thus, the
LiaScore model proved to be the methodology least affected
by the structural diversity.
On the other side of the spectrum, the MOE descriptor-based
QSAR proved to be very dependent on the structures of the
compounds included in the training set. Six test set compounds
were outliers for the model based on 2D and 3D descriptors.
These are all molecules that bear functionalities very distant
from those present in the training set: compound 11 bears an
alkynylic chain significantly shorter than those included in the
In compound 14, the removal of the 5′-phosphate group of
3c led to a great loss in affinity and potency.
In the acyclic phosphonate series, the 2-iodo substituent of
16 was well tolerated, showing affinity and potency similar to
those of the corresponding 2-chloro derivative 15. Thus, it did

3236 Journal of Medicinal Chemistry, 2007, Vol. 50, No. 14
Costanzi et al.
Table 4. Prediction of the Antagonistic Potencies of the Compounds in the Test Set. Errors of Predictions Are Shown in Parentheses
pIC50 in inhibition of 2-MeSADP stimulated PLC activity
experimental
5.16
consensus
LiaScore
SGB-LIE
CoMFA
CoMSIA
MOE 2D
MOE 3D
MOE 2D and 3D
5
5.66
(0.50)
6.02
(0.26)
6.15
(0.09)
6.76
(-0.27)
5.99
(0.28)
6.47
(0.66)
7.54
(1.33)
7.78
6.19
(1.03)
6.32
(0.56)
5.98
(-0.08)
6.25
(-0.78)
6.49
(0.78)
4.94
6.62
(1.46)
6.91
(1.15)
6.95
(0.89)
6.68
(-0.35)
6.56
(0.85)
1.82
(-3.99)
5.66
(-0.55)
6.08
7.15
(1.99)
6.89
(1.13)
7.26
(1.20)
6.96
(-0.07)
6.18
(0.47)
5.52
(-0.29)
6.50
(0.29)
6.69
6.84
(1.68)
5.52
(-0.24)
6.45
(0.39)
6.50
(-0.53)
6.41
(0.70)
7.23
(1.42)
6.50
(0.29)
6.83
10.10
(4.94)
6.66
(0.90)
4.65
(-1.41)
3.65
(-3.38)
6.44
(0.73)
6.81
2.91
(-2.25)
3.87
(-1.89)
3.64
(-2.42)
6.23
(-0.80)
7.11
(1.40)
10.33
(4.52)
7.07
(0.86)
7.52
4.42
(-0.74)
4.19
(-1.57)
3.48
(-2.58)
5.33
(-1.70)
6.01
(0.30)
9.97
(4.16)
7.96
(1.75)
8.55
6
5.76
6.06
7.03
5.71
5.81
6.21
6.10
7
11
13
14
(-0.87)
5.73
(-0.48)
5.74
(1.00)
3.90
(-2.31)
4.98
15
16
(1.68)
0.83
(-0.36)
(-0.02)
(0.59)
0.96
(0.73)
0.90
(-1.12)
2.42
(1.42)
2.24
(2.45)
2.21
RMSEP
r2
0.68
1.63
(0.39)a
0.30
no correlation
1
0.01
3
0.01
3
no correlation
2
no correlation
6
0.07
6
0.01
6
(0.79)a
2
outliers
a Value obtained excluding compounds 15 and 16 (outliers) from the PLS regression.
training set, compounds 6 and 7 bear a carboxamido and a
carboxy substituent at the 2 position, compound 14 bears a 5′-
OH group, and compound 15 and 16 bear phosphonate groups.
Confirming the dependence of the descriptor-based QSAR from
the training set, when either compound 6 or 7 was introduced
in the training set, the error of prediction on the other com-
pound was significantly reduced (Supporting Information, Table
S2).
Our consensus model (LIST-CM) performed significantly
better than any individual methodology also in the evaluation
of the potencies of the test set compounds. With a predictive r2
of 0.30, consensus scoring proved to be the only model able to
rank the potencies of the compounds in the test set. Indeed, the
potency of six out of the eight new compounds was accurately
estimated. Remarkably, if we exclude the two outliers, consensus
scoring shows an r2 of 0.79 and an RMSEP as low as 0.39.
Thus, our consensus model accurately and sharply predicted
the potency of 75% of the test set compounds.
Furthermore, docking programs provide accurate binding
modes and ligand conformations. Here, we used the conforma-
tion and the alignment of the ligands obtained in the docking
studies to build QSAR models. An advantage of docking-based
3D-QSAR is the possibility of plugging the contour maps into
the receptor binding pockets. Thus, we docked the CoMFA and
CoMSIA maps into our P2Y1 model, ascertaining good comple-
mentarity between the ligand-based 3D contour maps and the
molecular features of the receptor.
All of our docking-based QSAR studies yielded very good
correlation coefficients and good internal predictivities, which
were always higher than those obtained on the basis of docking
scores and free binding energy predictions. Notably, the MOE
descriptor-based QSAR yielded the highest internal predictivity,
as measured by the q2 value. However, this model appeared
very sensitive to the nature of the training set, and its good
internal predictivity did not translate into a good external
predictivity.
Thus, we merged receptor-based modeling and QSAR by
generating a PLS model based on consensus (LIST-CM).
Consensus scoring allows combining the strength of the
individual methodologies while mitigating their inherent limits.
The intrinsic coarseness of the receptor-based scores is com-
pensated by the higher accuracy of the QSAR-based scores. At
the same time, the substantial independence of the receptor-
based scores from the chemical nature of the compounds in the
training set compensates for the limits of the QSAR-based
scores. To validate our consensus model, we chose a particularly
challenging test set, in which each compound presents at least
one feature not included in the training set. Consensus scoring
proved to be more robust than any individual model both in
terms of internal and external predictivity. Remarkably, con-
sensus scoring predicted with accuracy the potency of 75% of
the test set compounds.
Discussion and Conclusions
The lack of ranking ability of the PLS regression based purely
on receptor-ligand interactions did not surprise us. Docking
scores and prediction of the free energy of binding provide a
direct estimation of the affinity of the compounds (∆G ) RT
ln Ki). The free energy of binding can also be linked to the
IC50 with the equation ∆G = RT ln (IC50), which however stays
in place only if the potencies are linearly correlated with the
affinities. Although this is expected to generally happen for
closely related compounds, the discrepancies between the
binding and functional data for our test set suggest that it may
not necessarily be the case for the ligands that are the object of
our study. The use of binding data in our QSAR analyses was
not an option due to the nonavailability of such data for most
of the training set. Aside from these considerations, a homology
model does not provide a sufficiently high resolution of the
binding site to accurately infer the difference in the activities
of various compounds.
Experimental Section
Conformational Search of 3c within the P2Y1 Receptor
Binding Pocket. Our previous rhodopsin-based model of the P2Y1
receptor complexed with the antagonist 3c was subjected to an
optimization by fully flexible conformational search with the Monte
Carlo Multiple Minimum (MCMM) method,22 as implemented in
However, docking scores are powerful tools to handle
structurally diverse compounds and to distinguish between
binders and nonbinders. In particular, our LiaScore model
proved to be essentially independent from the training set.

P2Y1 Antagonists
Journal of Medicinal Chemistry, 2007, Vol. 50, No. 14 3237
the MacroModel 9.1.32,33 The search was performed on the ligand
and the residues in the binding pocket, defined as the residues
located within 6 Å from the ligand. An additional shell of residues
located within 3 Å from the binding pocket were included in the
calculation, but were conformationally frozen. The following
parameters were employed for the conformational search: number
of steps ) 100; number of structures to save for each search )
100; energy window for saving structures ) 1000.0 kJ/mol. The
calculations were conducted with the MMFFs force field. For the
energy minimizations the Polak-Ribiere Conjugate Gradient
was used with a convergence threshold on the gradient of 0.05
kJ/Å/mol.
Automatic Docking. All the compounds of the training and the
test sets were automatically docked into the P2Y1 binding site by
means of Glide 4.0.23 The model of the P2Y1 receptor optimized
with 3c in the binding pocket obtained as described in the previous
paragraph was used for the docking experiments. All the compounds
were sketched from 3c, subjecting any added moiety to energy
minimization with the MMFFs force field until a convergence
threshold on the gradient of 0.05 kJ/Å/mol was reached with the
Polak-Ribiere Conjugate Gradient. In the receptor grid generation,
the vdW radii were scaled by 0.75 for the atoms with a partial
atomic charge less than 0.25e. The docking box was defined as a
cube of 10 Å centered on the centroid of the 3c. The ligands were
docked with the extra precision (XP) mode using default settings
and scaling by 0.80 the vdW radii of the ligands with partial atomic
charge less than 0.15e. The docking scoring function (LiaScore)
and the linear interaction energy (LIE) were calculated with Liaison
in a surface generalized Born continuum model for solvation
(SGB).26 Energy minimization was used as a sampling method, with
the Truncated Newton algorithm and with a residue-base cutoff
for nonbonded interactions of 15 Å. The OPLS 2005 force field
was used. The ligand and the closest residues were treated as fully
flexible; residues further than 4 Å from the ligand were restrained,
while those further than 7 Å were frozen. LiaScore is essentially
the Glide scoring function23,34 calculated during the Liaison
simulation. The LIE-SGB free binding energies were calculated
according to the empirical equation:
Consensus Scoring. Consensus scoring was achieved by sub-
jecting the pIC50 predicted with the individual methodologies to
PLS regression. The calculation was performed with the QuaSAR
application of MOE.29
Chemistry. 1H NMR spectra were obtained with a Varian
Gemini-300 spectrometer (300 MHz) with D2O, CDCl3, CD3OD,
and DMSO-d6 as a solvent. Purity of the nucleosides was checked
using a Hewlett-Packard 1100 HPLC equipped with a Luna 5µ RP-
C18(2) analytical column (250 × 4.6 mm; Phenomenex, Torrance,
CA). System B: linear gradient solvent system: CH3CN/TBAP
from 20/80 to 60/40 in 20 min, flow rate 1.0 mL/min. System C:
linear gradient solvent system: CH3CN/TBAP from 5/95 to 80/20
in 20 min, flow rate 1.0 mL/min. System D: linear gradient solvent
system: H2O/CH3CN from 95/5 to 0/100 in 30 min; the flow rate
was 1 mL/min. System E: linear gradient solvent system: H2O/
CH3CN/AcOH from 90/10/0.05 to 50/50/0.05 in 40 min and 0/100/
0.05 in 60 min, flow rate 0.5 mL/min. Peaks were detected by UV
absorption with a diode array detector. All derivatives tested for
biological activity showed g97% purity in the HPLC systems. Low-
resolution and high-resolution FAB (fast atom bombardment) mass
spectrometry was performed with a JEOL SX102 spectrometer with
6-kV Xe atoms following desorption from a glycerol matrix.
2,2-Dimethyl-propionic Acid 6-Chloro-2-cyanopurin-9-yl-
methyl Ester (18). Pd2(dba)3-CHCl3 (70 mg, 0.135 mmol) and
tri-(2-furyl)phosphine (210 mg, 0.904 mmol) were dissolved in THF
(0.3 mL) under dry N2 atmosphere, and the solvent was removed
under reduced pressure. 2,2-Dimethyl-propionic acid 6-chloro-2-
iodopurin-9-ylmethyl ester13 (333 mg, 0.814 mmol) and Zn(CN)2
(450 mg, 3.83 mmol) in tetramethylurea (2.0 mL) were added, and
the reaction mixture was stirred at 70 °C for 3 d. The crude reaction
mixture was passed through a short silica gel column (eluted by
AcOEt) to remove the salt and palladium, and the solvent was
evaporated. The residue obtained was purified by silica gel column
chromatography (AcOEt/petroleum ether ) 1/2), which furnished
18 (156 mg, 63%). 1H NMR (CDCl3) δ 8.58 (s, 1H), 6.21 (s, 2H),
1.19 (s, 9H). MS (m/e) (positive-FAB) 294, 295 (peak ratio is 3:1)
(M+ + H).
6-Methylamino-9H-purine-2-carbonitrile (19). To a solution
of 2,2-dimethyl-propionic acid 6-chloro-2-cyanopurin-9-ylmethyl
ester (18) (92 mg, 0.313 mmol) in MeCN (0.5 mL) was added 2 N
MeNH2 in THF (2.0 mL). The reaction mixture was stirred at rt
for 2 d, and the solvent was removed under reduced pressure. The
residue obtained was purified by recrystallization (from AcOEt),
which furnished 6-methylamino-9H-purine-2-carbonitrile (19, 21
mg, 39%). The filtrate was purified by silica gel column chroma-
tography (AcOEt/petroleum ether)1/1), which furnished 2,2-
dimethyl-propionic acid 2-cyano-6-methylaminopurin-9-yl-
methyl ester (20, 52.5 mg, 58%).
∆G ) R ( Ubvdw - Ufvdw ) + â ( Ubelec - Ufelec ) +
γ ( Ubcav - U fcav )
where b indicates the bound form of the ligand, f indicates the free
form of the ligand, and Ucav represent an energy term proportional
to the exposed surface area of the ligand.26,27
Prediction of the pIC50 values were obtained by PLS regressions
of the LiaScore and the LIE-SGB descriptors, which were per-
formed with the QuaSAR-Model application of MOE using default
parameters.29 Experimental pIC50 values were used as dependent
variables.
CoMFA and CoMSIA. A 3D cubic lattice with grid spacing of
1 Å and extending 4 Å behind the aligned molecules in all directions
was used. The molecular fields (steric and electrostatic for CoMFA;
steric, electrostatic, H-bond donor, H-bond acceptor, and hydro-
phobic for CoMSIA) were calculated using a probe atom with the
vdW properties of an sp3 carbon and with a charge of +1.0. For
the generation of the CoMFA fields, a distance dependent dielectric
was selected for Coulombic electrostatic energy calculation and a
cutoff of 30 kcal/mol with smooth transition was used. Standard
parameters were used for the generation of the CoMSIA fields.
The PLS regression, using the experimental pIC50 values as
dependent variable, and the leave-one-out cross validation were
performed in Sybyl. For CoMFA and CoMSIA, the CoMFA
standard scaling and a column filtering of 2.0 kcal/mol were used.
Gasteiger-Huckel charges were used for all ligands.
6-Methylamino-9H-purine-2-carbonitrile, 19. 1H NMR (DMSO-
d6) δ 8.33 (s, 1H), 8.23 (bs, 1H), 2.96 (br, 3H). MS (m/e) (positive-
FAB) 175 (M+ + H).
2,2-Dimethyl-propionic Acid 2-cyano-6-methylaminopurin-
1
9-ylmethyl Ester, 20. H NMR (CDCl3) δ 8.13 (s, 1H), 6.11 (s,
2H), 5.65 (bs, 1H), 3.21 (br, 3H), 1.20 (s, 9H). MS (m/e) (positive-
FAB) 289 (M+ + H).
(1′R,2′S,4′R,5′S)-Phosphoric Acid Di-tert-butyl Ester 4-(2-
Cyano-6-methylaminopurin-9-yl)-1-(di-tert-butoxy-phosphoryl-
oxymethyl)-bicyclo[3.1.0]hex-2-yl Ester (22). To a solution of
triphenylphosphine (78 mg, 0.297 mmol) in anhydrous THF (0.50
mL) was added DIAD (0.060 mL, 0.305 mmol) at rt with stirring
for 2 h. Compound 2113,15 (17 mg, 0.032 mmol) and 6-methylamino-
2-cyano-9H-purine (13.5 mg, 0.78 mmol) in THF (0.8 mL) were
added, and the reaction mixture was stirred at rt for 23 h. The
solvent was removed under vacuum, and the residue obtained was
purified by preparative thin-layer chromatography (AcOEt), which
furnished 22 (6.0 mg, 27%). 1H NMR (CDCl3) δ 8.31 (s, 1H), 6.22
(bs, 1H), 5.34 (dd, 1H, J ) 8.4, 14.5 Hz), 5.11 (d, 1H, J ) 6.6
Hz), 4.68 (dd, 1H, J ) 3.0, 9.6 Hz), 3.88 (dd, 1H, J ) 6.6, 9.6
Hz), 3.20 (bs, 3H), 2.32 (dd, 1H, J ) 8.4, 15.3 Hz), 2.11 (dd, 1H,
J ) 7.8, 15.3 Hz), 1.49 (s, 18H), 1.47 (s, 18H), 1.42 (m, 1H), 1.28
Molecular Descriptor-Based QSAR. Calculation of the mo-
lecular descriptors and PLS regressions were performed with the
QuaSAR suite of applications of MOE using default parameters.29
Experimental pIC50 values were used as dependent variables.
Gasteiger (PEOE) charges were used for all ligands.

3238 Journal of Medicinal Chemistry, 2007, Vol. 50, No. 14
Costanzi et al.
(m, 1H), 1.02 (m, 1H). MS (m/e) (positive-FAB) 685 (M+ + H),
707 (M+ + Na).
added TFA (0.1 mL), and the reaction mixture was stirred at room
temperature for 3 h. After removal of the solvent, water (5.0 mL)
was added in 40% MeNH2 in water (1.0 mL) and stirred for 2 h at
room temperature. The reaction was monitored by HPLC. The
reaction mixture was subsequently frozen and lyophilized. Purifica-
tion of the residue obtained was performed on an ion-exchange
column packed with Sephadex-DEAE A-25 resin. A linear gradient
(0.01-0.7 M) of ammonium bicarbonate was applied as the mobile
phase, and UV and HPLC were used to monitor the elution, which
furnished 11 (7.1 mg, 36%) 1H NMR (CDCl3) 8.65 (s, 1H), 5.05-
5.11 (m, 1H), 4.96 (d, J ) 6.6 Hz, 1H), 4.40 - 4.52 (m, 1H),
3.52-3.60 (m, 1H), 3.07 (bs, 4H), 2.17 - 2.23 (m, 1H), 1.81-
1.88 (m, 2H), 1.10 - 1.13 (m, 1H), 0.92-0.93 (m, 1H) MS (m/e)
(positive-FAB) 460 (M + H)+.
(1′R,2′S,4′R,5′S)-Phosphoric Acid Mono-[4-(2-cyano-6-methyl-
aminopurin-9-yl)-1-phosphonooxymethyl-bicyclo[3.1.0]hex-2-yl]
Ester (5), (1′R,2′S,4′R,5′S)-Phosphoric Acid Mono-[4-(2-car-
bamoyl-6-methylaminopurin-9-yl)-1-phosphonooxymethyl-bicyclo-
[3.1.0]hex-2-yl] Ester (6) and (1′R,2′S,4′R,5′S)-6-Methylamino-
9-(4-phosphonooxy-5-phosphonooxymethyl-bicyclo[3.1.0]hex-2-
yl)-9H-purine-2-carboxylic Acid (7). To a solution of 22 (6.0 mg,
0.0088 mmol) in CH2Cl2 (2.0 mL) was added TFA (0.100 mL),
and the mixture was stirred at rt for 5.5 h. 2 N Et3NH2CO3 buffer
(2 mL) was added to the reaction mixture, and the crude product
was lyophilized. Purification of the obtained residue was performed
by an ion-exchange column packed with Sephadex-DEAE A-25
resin, a linear gradient (0.01 to 0.5 M) of ammonium bicarbonate
was applied as the mobile phase, and UV and HPLC were used to
monitor the elution, which furnished 5 (1.95 mg, 42%) and the
mixture fraction of 5 and 6 (1:2 mixture, 2.0 mg).
The mixture fraction of 5 and 6 (1:2 mixture, 2.0 mg) was
dissolved in H2O (1.5 mL) and EtOH (1.5 mL). To this solution
was added 3 N NaOH aq (0.5 mL) and 30% H2O2 aq (0.3 mL),
and the reaction mixture was stirred at rt for 21 h and lyophilized.
Purification of the obtained residue was performed by HPLC (a
linear gradient of MeCN/TBAP ) 20/80 to 80/20 in 15 min) and
Sephadex-DEAE A-25 resin (a linear gradient 0.01M to 0.5 M of
ammonium bicarbonate), to give 6 (0.67 mg, 14%) and 7 (1.32
mg, 27%).
2,2-Dimethyl-propionic Acid 6-Chloro-2-phenylpurin-9-yl-
methyl Ester (25). Pd2(dba)3-CHCl3 (9.5 mg, 9.17 µmol) and Ph3P
(19.0 mg, 72.4 µmol) were dissolved in THF (0.3 mL) under dry
N2 atmosphere, and the solvent was removed under reduced
pressure. 2,2-Dimethyl-propionic acid 6-chloro-2-iodopurin-9-yl-
methyl ester13 (126 mg, 0.319 mmol), K2CO3 (71 mg, 0.514 mmol),
PhB(OH)2 (42 mg, 0.345 mmol), and toluene (1.0 mL) were added,
and the reaction mixture was stirred at 100 °C for 7 d. The crude
reaction mixture was passed through a short silica gel column
(eluted by AcOEt) to remove the salt and palladium, and the solvent
was evaporated. The residue obtained was purified by silica gel
column chromatography (AcOEt/petroleum ether ) 1/2), which
1
furnished 25 (101 mg, 92%). H NMR (CDCl3) δ 8.55 (m, 2H),
8.34 (s, 1H), 7.51 (m, 3H), 6.26 (s, 2H), 1.19 (s, 9H). MS (m/e)
(1′R,2′S,4′R,5′S)-Phosphoric Acid Mono-[4-(2-cyano-6-methyl-
(positive-FAB) 345, 347 (peak ratio is 3:1) (M+ + H).
aminopurin-9-yl)-1-phosphonooxymethyl-bicyclo[3.1.0]hex-2-yl]
2,2-Dimethyl-propionic Acid 6-methylamino-2-phenylpurin-
9-ylmethyl Ester (26). To a solution of 2,2-dimethyl-propionic acid
6-chloro-2-phenylpurin-9-ylmethyl ester (25) (90 mg, 0.261 mmol)
in THF (3.0 mL) was added 2 N MeNH2 in THF (1.0 mL). The
reaction mixture was stirred at rt for 25 h, and the solvent was
removed under reduced pressure. The residue obtained was purified
by silica gel column chromatography (AcOEt/petroleum ether )
1/1 then 1/0), which furnished 26 (71 mg, 80%). 1H NMR (CDCl3)
δ 8.53 (m, 2H), 7.98 (s, 1H), 7.51-7.41 (m, 3H), 6.19 (s, 2H),
5.89 (bs, 1H), 3.32 (bd, 3H, J ) 4.8 Hz), 1.18 (s, 9H). MS (m/e)
(positive-FAB) 340 (M+ + H).
1
Ester (5). H NMR (D2O) δ 8.64 (s, 1H), 5.08 (d, 1H, J ) 6.6
Hz), 4.70-4.60 (m, 1H), 4.60-4.50 (m, 1H), 3.70 (m, 1H), 3.12
(s, 3H), 2.30(m, 1H), 2.02 (m, 1H), 1.93 (m, 1H), 1.23 (m, 1H),1.02
(t, 1H, J ) 7.4 Hz). 31P NMR (D2O) δ 0.892 (s), 0.364 (s). MS
(m/e) (negative-FAB) 459 (M+ - H). LRMS (negative-FAB) calcd
for C14H17N6O8P2 459.0583, found 459.058. HPLC 17.2 min (99%,
System B), 2.3 min (99%, System D).
(1′R,2′S,4′R,5′S)-Phosphoric Acid mono-[4-(2-carbamoyl-6-
methylaminopurin-9-yl)-1-phosphonooxymethyl-bicyclo[3.1.0]-
hex-2-yl] Ester (6). 1H NMR (D2O) δ 8.63 (s, 1H), 5.30 (m, 1H),
5.20 (d, 1H J ) 6.6 Hz), 4.58 (m, 1H), 3.46 (m, 1H), 3.20 (s, 3H),
2.28 (m, 1H), 2.03 (m, 1H), 1.87 (m, 1H), 1.34 (m, 1H), 1.21 (m,
1H). 31P NMR (D2O) δ 2.00 (s), 1.29 (s). MS (m/e) (negative-
FAB) 477 (M+ - H), 499 (M+ - 2H + Na). HPLC 18.4 min (99%,
System B), 3.8 min (99%, System D).
(1′R,2′S,4′R,5′S)-6-Methylamino-9-(4-phosphonooxy-5-phospho-
nooxymethyl-bicyclo[3.1.0]hex-2-yl)-9H-purine-2-carboxylic Acid
(7). 1H NMR (D2O) δ 8.69 (s, 1H), 5.27 (br, 1H), 5.19 (d, 1H, J )
6.0 Hz), 4.54 (m, 1H), 3.61 (m, 1H), 3.20 (s, 3H), 2.29(m, 1H),
2.01(m, 1H), 1.86(m, 1H),1.18(m, 1H), 0.96 (t, 1H, J ) 7.4 Hz).
31P NMR (D2O) δ 2.82 (s), 2.14 (s). MS (m/e) (negative-FAB)
478 (M+ - H). HPLC 16.0 min (99%, System B), 1.9 min (99%,
System D).
6-Methylamino-2-phenyl-9H-purine (27). To a solution of 2,2-
dimethyl-propionic acid 6-methylamino-2-phenylpurin-9-ylmethyl
ester (26) (58 mg, 0.172 mmol) in THF (1.0 mL) and MeOH (1.0
mL) was added 3 N NaOH aq (0.175 mL). The reaction mixture
was stirred at rt for 3 h, and the solvent was removed under reduced
pressure. The residue obtained was purified by silica gel column
chromatography (AcOEt then CHCl3/MeOH ) 5/1), which fur-
1
nished 27 (35 mg, 90%). H NMR (DMSO-d6) δ 8.40 (m, 2H),
7.94 (s, 1H), 7.47-7.35 (m, 3H), 3.07 (d, 3H, J ) 4.5 Hz). MS
(m/e) (positive-FAB) 226 (M+ + H).
(1′R,2′S,4′R,5′S)-Phosphoric Acid Di-tert-butyl Ester 1-(Di-
tert-butoxy-phosphoryloxymethyl)-4-(6-methylamino-2-phenyl-
purin-9-yl )-bicyclo[3.1.0]hex-2-yl Ester (28). To a solution of
triphenylphosphine (53 mg, 0.202 mmol) in anhydrous THF (0.50
mL) was added DIAD (0.035 mL, 0.178 mmol) at rt with stirring
for 2 h. (1′R,2′S,4′R,5′S)-Phosphoric acid di-tert-butyl ester 1-(di-
tert-butoxy-phosphoryloxymethyl)-4-hydroxy-bicyclo[3.1.0]hex-2-
yl ester13,15 (21, 25 mg, 0.48 mmol) and 6-methylamino-2-phenyl-
9H-purine (46.5 mg, 0.206 mmol) in THF (1.0 mL) were added,
and the reaction mixture was stirred at rt for 23 h. The solvent was
removed under vacuum, and the residue obtained was purified by
preparative thin-layer chromatography (AcOEt then chloroform/
MeOH ) 5/1), which furnished 28 (18 mg, 51%). 1H NMR (CDCl3)
δ 8.52-8.48 (m, 2H), 8.09 (s, 1H), 7.49-7.40 (m, 3H), 5.88 (m,
1H), 5.41 (m, 1H), 5.27 (d, 1H, J ) 7.2 Hz), 4.66 (dd, 1H, J )
4.5, 11.4 Hz), 3.90 (dd, 1H, J ) 4.5, 11.4 Hz), 3.30 (bd, 3H, J )
5.1 Hz), 2.43 (m, 1H), 2.16 (m, 1H), 1.85 (m, 1H), 1.49 (s, 18H),
1.48 (s, 9H), 1.43 (s, 9H), 1.18 (m, 1H), 1.01(m, 1H). MS (m/e)
(positive-FAB) 736 (M+ + H).
(1′R,2′S,4′S,5′S)-Phosphoric Acid Di-tert-butyl Ester 1-(di-
tert-butoxy-phosphoryloxymethyl)-4-(6-chloro-2-(2′-trimethyl-
silyl)ethynylpurin-9-yl)-bicyclo[3.1.0]hex-2-yl Ester (24). To a
solution of compound 23 (30 mg, 0.038 mmol) in 1.5 equiv of
triethylamine solution were added trimethylsilylacetylene (0.01 mL,
0.0758 mmol), bis(triphenylphosphine)palladium(II) dichloride (2.6
mg, 0.1 equiv), and CuI (1.4 mg, 0.2 equiv) and stirred at room
temperature for 7 h. The solvent was removed under vacuum, and
the residue obtained was purified by preparative thin-layer chro-
1
matography, which furnished product (20.5 mg, 71%). H NMR
(CDCl3) 8.59 (s, 1H), 5.28 (d, J ) 6.9 Hz, 1H), 5.26-5.33 (m,
2H), 4.69 (dd, J ) 4.8, 11.4 Hz, 1H), 3.84 (dd, J ) 6.3, 11.3 Hz,
1H), 2.30 (d, J ) 8.4 Hz, 1H), 2.14-2.17 (m, 1H), 1.81-1.85 (m,
1H), 1.49 (s, 36H), 1.18 (dd, J ) 3.9, 6.3 Hz, 1H), 1.02 (dd, J )
8.7, 6.0 Hz, 1H), 0.19 (s, 9H)
(1′R,2′S,4′S,5′S)-4-(2-Ethynyl-6-methylaminopurin-9-yl)-1-
[(phosphato)-methyl]-2-(phosphato)-bicyclo[3.1.0]hexane (11).
The mixture of 24 (33.0 mg, 0.043 mmol) in CH2Cl2 (3 mL) was
(1′R,2′S,4′R,5′S)-Phosphoric Acid mono-[4-(6-methylamino-
2-phenylpurin-9-yl)-1-phosphonooxymethyl-bicyclo[3.1.0]hex-

P2Y1 Antagonists
Journal of Medicinal Chemistry, 2007, Vol. 50, No. 14 3239
2-yl] Ester (13). To a solution of 28 (10.0 mg, 0.0136 mmol) in
CH2Cl2 (2.0 mL) was added TFA (0.100 mL), and the mixture was
stirred at rt for 2 d. 2 N Et3NH2CO3 buffer (2 mL) was added to
the reaction mixture, and the crude product was lyophilized.
Purification of the obtained residue was performed by an ion-
exchange column packed with Sephadex-DEAE A-25 resin, a linear
gradient (0.01 to 0.5 M) of 0.5 M ammonium bicarbonate was
applied as the mobile phase, and UV and HPLC were used to
bicyclo[3.1.0]hex-2-yl Ester Di-tert-butyl Ester (34). Partial
phosphorylation of the (N)-methanocarba ring was already re-
ported.12 To a stirred solution of 33 (18 mg, 0.026 mmol) and 1H-
tetrazole (10 mg, 0.14 mmol) in anhydrous THF (0.25 mL) was
added di-tert-butyl diethylphosphoramidite (0.035 mL, 0.13 mmol),
and the mixture was stirred for 3.5 h at rt. The reaction mixture
was cooled to -78 °C and treated with a solution of m-CPBA (70%
max, 140 mg) in CH2Cl2 (2.0 mL). The resulting mixture was
warmed up to rt and stirred for 30 min. 5% NaHSO3 (2.0 mL) was
added to the reaction mixture and stirred another 10 min at rt. The
reaction mixture was extracted with AcOEt, and the organic phase
was subsequently washed with sat. aq NaHCO3 and brine, dried
over Na2SO4, and filtered. The solvent was removed under reduced
pressure. The obtained residue was purified by silica gel column
chromatography (AcOEt/petroleum ether ) 1/1 to 1/0 then CHCl3/
1
monitor the elution, which furnished 13 (3.6 mg, 46%). H NMR
(D2O) δ 8.56 (bs, 1H), 8.04 (m, 2H), 7.61-7.50 (m, 3H), 5.25 (m,
1H), 5.09 (d, 1H, J ) 6.0 Hz), 4.58 (dd, 1H, J ) 4.2, 10.8 Hz),
3.72 (dd, 1H, J ) 4.2, 10.8 Hz), 3.27 (bs, 3H), 2.33 (m, 1H), 2.05
(m, 1H), 1.85 (m, 1H), 1.23 (m, 1H), 1.03 (m, 1H). 31P NMR (D2O)
δ 0.44 (bs). MS (m/e) (negative-FAB) 510 (M+ - H). HPLC 19.5
min (99%, System B), 16.2 min (99%, System C).
1
MeO ) 5/1), to give 34 (4.7 mg, 21%). H NMR (CDCl3) δ 8.08
(1′R,2′S,4′S,5′S)-Acetic Acid 1-Acetoxymethyl-4-(6-chloro-2-
iodopurin-9-yl)-bicyclo[3.1.0]hex-2-yl Ester (31). To a solution
of triphenylphosphine (250 mg, 0.953 mmol) in anhydrous THF
(1.00 mL) was added DEAD (0.150 mL, 0.953 mmol) at -78 °C,
and the reaction mixture was stirred at rt for 0.5 h. Acetic acid
2-acetoxy-4-hydroxy-bicyclo[3.1.0]hex-1-ylmethyl ester30 (29, 73
mg, 0.32 mmol) and 2-amino-6-chloropurine (30, 210 mg, 1.24
mmol) in THF (2.0 mL) were added, and the reaction mixture was
stirred at rt for 4 d. The solvent was removed under vacuum, and
the residue obtained was purified by silica gel column chromatog-
raphy (chloroform then AcOEt/petroleum ether ) 2/1). The residue
obtained was dissolved in MeCN (1.00 mL), and diiodomethane
(2.00 mL) and tert-butyl nitrite (0.30 mL, 3.3 mmol) were added.
The tube was sealed after the nitrogen bubbling, and the reaction
mixture was stirred at 80 °C for 3 h. The solvent was removed
under reduced pressure. The residue obtained was purified by silica
gel column chromatography (AcOEt/petroleum ether ) 2/1), which
(s, 1H), 7.41-7.37 (m, 2H), 7.36-7.16 (m, 7H), 6.79 (d, 4H, J )
8.1 Hz), 5.91 (bs, 1H), 5.51 (m, 1H), 5.05 (d, 1H, J ) 6.9 Hz),
3.95 (d, 1H, J ) 9.9 Hz), 3.78 (s, 6H), 3.61 (bs, 3H), 2.98 (d, 1H,
J ) 9.9 Hz), 2.36-2.27 (m, 1H), 2.10-1.95 (m, 1H), 1.47 (m,
1H), 1.42 (s, 9H), 1.39 (s, 9H), 0.94 (m, 1H), 0.80 (m, 1H). MS
(m/e) (positive-FAB) 896 (M+ + H).
(1′R,2′S,4′S,5′S)-Phosphoric Acid mono-[1-hydroxymethyl-
4-(2-iodo-6-methylaminopurin-9-yl)-bicyclo[3.1.0]hex-2-yl] Ester
(14). To a solution of 34 (4.7 mg, 0.0052 mmol) in CH2Cl2 (2.0
mL) was added TFA (0.100 mL), which was stirred at rt for 3 h.
2 N Et3NH2CO3 buffer (2 mL) was added to the reaction mixture,
and the crude product was lyophilized. Purification of the obtained
residue was performed by an ion-exchange column packed with
Sephadex-DEAE A-25 resin, a linear gradient (0.01 to 0.5 M) of
ammonium bicarbonate was applied as the mobile phase, and UV
and HPLC were used to monitor the elution, which furnished 14
(1.70 mg, 63%). 1H NMR (D2O) δ 8.23 (s, 1H), 5.21 (dd, 1H, J )
8.1, 15.6 Hz), 4.95 (d, 1H, J ) 6.9 Hz), 4.12 (d, 1H, J ) 12.6 Hz),
3.58 (d, 1H, J ) 12.6 Hz), 3.08 (bs, 3H), 2.32 (dd, 1H, J ) 8.1,
15.0 Hz), 2.05 (dd, 1H, J ) 7.8, 15.0 Hz), 1.82 (m, 1H), 1.17 (m,
1H), 0.96 (m, 1H). 31P NMR (D2O) δ 0.465 (s). MS (m/e) (negative-
FAB) 480 (M+ - H). HPLC 13.7 min (>99%, System B).
[3-(2-Iodo-6-chloropurin-9-yl)-2-(diethoxyphosphorylmethyl)-
propyl]-phosphonic Acid Diethyl Ester (37). The starting alcohol
tetraethyl 2-hydroxymethyl-1,3-propanebisphosphonate (36) was
dried by coevaporation from anhydrous THF (10 mL) two times.
Subsequently, a solution of 36 (0.064 g, 0.2 mmol) in anhydrous
THF (10 mL) was combined with 2-iodo-6-chloropurine 35 (0.112
g, 0.4 mmol), triphenylphosphine (0.104 g, 0.4 mmol), and diethyl
azodicarboxylate (0.068 g, 0.4 mmol) and stirred at ambient
temperature overnight. The solution was concentrated, and the
resulting residue was purified by purified by preparative thin layer
chromatography by using chloroform/methanol (9:1) as solvent to
provide 37 (0.062 g, 52%). 1H NMR (400 MHz, CDCl3) δ 1.34 (t,
J ) 6.1 Hz, 12H), 1.80-2.08 (m, 4H), 2.95 (m, 1H), 4.05-4.21
(m, 8H), 4.60 (d, J ) 6.0 Hz, 2H), 8.38 (s, 1H); 31P NMR (162
MHz, CDCl3) δ 31.1; TOFMS m/z 609.2 (M + H+).
[3-(2-Iodo-6-methylaminopurin-9-yl)-2-(diethoxyphosphoryl-
methyl)-propyl ]-phosphonic Acid Diethyl Ester (38). A THF
(2 mL) solution of 37 (0.061 g, 0.1 mmol) and 40% methylamine
(0.12 mL, 1.7 mmol) was stirred at ambient temperature for 1 h.
The reaction solution was concentrated, and the residue was purified
directly by preparative thin layer chromatography by using chlo-
roform/methanol (9:1) as solvent to furnish 38 (0.044 g, 74%). 1H
NMR (400 MHz, CDCl3) δ 1.34 (t, 12H, J ) 7.2 Hz), 1.81-2.05
(m, 4H), 2.92 (m, 1H), 3.41 (s, 3H), 4.12 (m, 8H), 4.43 (d, J ) 4.9
Hz, 2H), 8.04 (s, 1H); 31P NMR (162 MHz, CDCl3) δ 29.2; TOFMS
m/z 604.2 (M + H+).
[3-(2-Iodo-6-methylaminopurin-9-yl)-2-phosphonomethylpro-
pyl]-phosphonic Acid Tetraammonium Salt (16). Compound 38
(0.120 g, 0.2 mmol) was dissolved into anhydrous acetonitrile
(4 mL). Iodotrimethylsilane (0.2 mL, 1.35 mmol) was added. The
reaction was stirred at ambient temperature for 4.5 h. Then reaction
was concentrated to dryness. The residue was added with triethy-
lammonium bicarbonate buffer (0.2 mL) followed by water
1
furnished desired product (20.0 mg, 32%). H NMR (CDCl3) δ
8.40 (s, 1H), 5.68 (t, 1H, J ) 8.5 Hz), 5.22 (d, 1H, J ) 7.2 Hz),
4.68 (d, 1H, J ) 12.0 Hz), 3.92 (d, 1H, J ) 12.0 Hz), 2.42 (dd,
1H, J ) 8.5, 15.6 Hz), 2.17 (s, 3H), 2.09 (s, 3H), 1.91 (m, 1H),
1.75 (dd, 1H, J ) 4.2, 8.7 Hz), 1.18 (dd, 1H, J ) 4.2, 8.7 Hz),
1.04 (m, 1H). MS (m/e) (positive-FAB) 491 (M+ + H).
(1′R,2′S,4′S,5′S)-1-Hydroxymethyl-4-(2-iodo-6-methylami-
nopurin-9-yl)-bicyclo[3.1.0]hexan-2-ol (32). To a solution of acetic
acid 1-acetoxymethyl-4-(6-chloro-2-iodopurin-9-yl)-bicyclo[3.1.0]-
hex-2-yl ester (31, 20 mg, 0.041 mmol) in THF (0.030 mL) was
added 2 N MeNH2 in THF (0.30 mL), and the reaction mixture
was stirred at rt for 6.5 h. MeOH (0.10 mL) and 3 N NaOH aq
(0.040 mL) were added to the reaction mixture, which was stirred
overnight at rt. The solvent was removed under vacuum, and the
residue obtained was purified by silica gel column chromatography
(AcOEt then CHCl3/MeOH) 5/1), which furnished 32 (15 mg,
1
94%). H NMR (CD3OD) δ 8.36 (s, 1H), 4.96 (d, 1H, J ) 6.6
Hz), 4.86 (t, 1H, J ) 8.7 Hz), 4.25 (d, 1H, J ) 11.7 Hz), 3.37 (d,
1H, J ) 11.7 Hz), 3.05 (bs, 3H), 2.01 (m, 1H), 1.81 (m, 1H), 1.63
(dd, 1H, J ) 3.6, 8.4 Hz), 1.01 (dd, 1H, J ) 3.9, 5.7 Hz), 0.77 (dd,
1H, J ) 6.0, 8.7 Hz). MS (m/e) (CI) 402 (M+ + H). HPLC 11.4
min (99%, System C), 31.5 min (99%, System E).
(1′R,2′S,4′S,5′S)-1-[Bis-(4-methoxy-phenyl)-phenyl-methoxy-
methyl]-4-(2-iodo-6-methylaminopurin-9-yl)-bicyclo[3.1.0]hexan-2-ol
(33). To a solution of 1-hydroxymethyl-4-(2-iodo-6-methyl-
aminopurin-9-yl)-bicyclo[3.1.0]hexan-2-ol (32, 12 mg, 0.030 mmol)
in DMF (0.50 mL) were added 4,4′-dimethoxytrityl chloride (47
mg, 0.139 mmol) and 4-dimethylaminopyridine (33 mg, 0.270
mmol), and the reaction mixture was stirred at rt for 24 h. The
solvent was removed under vacuum, and the residue obtained was
purified by silica gel column chromatography (AcOEt), which
furnished 33 (19 mg, 92%). 1H NMR (CDCl3) δ 7.92 (s, 1H), 7.46-
7.42 (m, 2H), 7.35-7.15 (m, 7H), 6.84 (d, 4H, J ) 9.0 Hz), 5.76
(bs, 1H), 5.02 (d, 1H, J ) 6.9 Hz), 4.94 (m, 1H), 3.53 (d, 1H, J )
9.9 Hz), 3.33 (d, 1H, J ) 9.9 Hz), 3.17 (bs, 3H), 2.18-2.07 (m,
1H), 1.89-1.77 (m, 1H), 1.47 (m, 1H), 0.99 (m, 1H), 0.66 (m,
1H). MS (m/e) (positive-FAB) 704 (M+ + H).
(1′R,2′S,4′S,5′S)-Phosphoric Acid 1-[Bis-(4-methoxy-phenyl)-
phenyl-methoxymethyl]-4-(2-iodo-6-methylaminopurin-9-yl)-

3240 Journal of Medicinal Chemistry, 2007, Vol. 50, No. 14
Costanzi et al.
(8) Camaioni, E.; Boyer, J. L.; Mohanram, A.; Harden, T. K.; Jacobson,
K. A. Deoxyadenosine bisphosphate derivatives as potent antagonists
at P2Y1 receptors. J. Med. Chem. 1998, 41, 183-190.
(9) Nandanan, E.; Camaioni, E.; Jang, S. Y.; Kim, Y. C.; Cristalli, G.;
Herdewijn, P.; Secrist, J. A., III; Tiwari, K. N.; Mohanram, A.;
Harden, T. K.; Boyer, J. L.; Jacobson, K. A. Structure-activity
relationships of bisphosphate nucleotide derivatives as P2Y1 receptor
antagonists and partial agonists. J. Med. Chem. 1999, 42, 1625-
1638.
(10) Nandanan, E.; Jang, S. Y.; Moro, S.; Kim, H. O.; Siddiqui, M. A.;
Russ, P.; Marquez, V. E.; Busson, R.; Herdewijn, P.; Harden, T. K.;
Boyer, J. L.; Jacobson, K. A. Synthesis, biological activity, and
molecular modeling of ribose-modified deoxyadenosine bisphosphate
analogues as P2Y1 receptor ligands. J. Med. Chem. 2000, 43, 829-
842.
(11) Kim, H. S.; Barak, D.; Harden, T. K.; Boyer, J. L.; Jacobson, K. A.
Acyclic and cyclopropyl analogues of adenosine bisphosphate
antagonists of the P2Y1 receptor: structure-activity relationships
and receptor docking. J. Med. Chem. 2001, 44, 3092-3108.
(12) Xu, B.; Stephens, A.; Kirschenheuter, G.; Greslin, A. F.; Cheng, X.;
Sennelo, J.; Cattaneo, M.; Zighetti, M. L.; Chen, A.; Kim, S. A.;
Kim, H. S.; Bischofberger, N.; Cook, G.; Jacobson, K. A. Acyclic
analogues of adenosine bisphosphates as P2Y receptor antagonists:
phosphate substitution leads to multiple pathways of inhibition of
platelet aggregation. J. Med. Chem. 2002, 45, 5694-5709.
(13) Kim, H. S.; Ohno, M.; Xu, B.; Kim, H. O.; Choi, Y.; Ji, X. D.;
Maddileti, S.; Marquez, V. E.; Harden, T. K.; Jacobson, K. A.
2-Substitution of adenine nucleotide analogues containing a bicyclo-
[3.1.0]hexane ring system locked in a northern conformation:
enhanced potency as P2Y1 receptor antagonists. J. Med. Chem. 2003,
46, 4974-4987.
(10 mL), and the solution was extracted with diethyl ether (3 mL
× 5). The aqueous layer was concentrated under reduced pressure.
The residue was purified by ion-exchange column chromatography
using Sephadex-DEAE A-25 resin with a linear gradient (0.01-
0.5 M) of ammonium bicarbonate as mobile phase to give 46 mg
1
of 16 as a pale yellow solid (42%). H NMR (400 MHz, D2O) δ
1.61-1.78 (m, 2H), 1.84-1.98 (m, 2H), 2.67 (m, 1H), 3.06 (s,
3H), 4.47 (d, J ) 5.4 Hz, 2H), 8.31 (d, 1H, J ) 6.0 Hz); 31P NMR
(162 MHz, D2O) δ 23.4; TOFMS m/z 490.1 (M- H+).
Pharmacology. 2-MeSADP was purchased from Sigma (St.
Louis, MO). Myo-[3H]inositol (20 Ci/mmol) was obtained from
American Radiolabeled Chemicals (St. Louis, MO).
P2Y receptor-promoted stimulation of inositol phosphate forma-
tion was measured at the human P2Y1 receptor stably expressed in
1321N1 human astrocytoma cells as previously described.35,36 The
IC50 values were averaged from three independently determined
concentration-effect curves for each compound. Briefly, cells plated
in 24-well dishes were labeled in inositol-free medium (DMEM;
Gibco, Gaithersburg, MD) containing 1.0 µCi of 2-[3H]myo-inositol
for 18-24 h in a humidified atmosphere of 95% air/5% CO2 at 37
°C. PLC activity was measured the following day by quantitating
[3H]inositol phosphate accumulation after a 10 min incubation at
37 °C in the presence of 10 mM LiCl. Total [3H]inositol phosphates
were quantified by anion exchange chromatography as previously
described.35,36 The affinities of bisphosphate analogues for the
human P2Y1 receptor were directly determined by using [3H]3b in
a radioligand binding assay, as we recently described in detail.37
Binding and functional parameters were estimated using GraphPAD
Prism software (GraphPAD, San Diego, CA).
(14) Ohno, M.; Costanzi, S.; Kim, H. S.; Kempeneers, V.; Vastmans, K.;
Herdewijn, P.; Maddileti, S.; Gao, Z. G.; Harden, T. K.; Jacobson,
K. A. Nucleotide analogues containing 2-oxa-bicyclo[2.2.1]heptane
and l-alpha-threofuranosyl ring systems: interactions with P2Y
receptors. Bioorg. Med. Chem. 2004, 12, 5619-5630.
Acknowledgment. This research was supported in part by
the Intramural Research Program of the NIH, NIDDK.
(15) Cattaneo, M.; Lecchi, A.; Ohno, M.; Joshi, B. V.; Besada, P.;
Tchilibon, S.; Lombardi, R.; Bischofberger, N.; Harden, T. K.;
Jacobson, K. A. Antiaggregatory activity in human platelets of potent
antagonists of the P2Y1 receptor. Biochem. Pharmacol. 2004, 68,
1995-2002.
(16) Houston, D.; Ohno, M.; Nicholas, R. A.; Jacobson, K. A.; Harden,
T. K. [32P]2-iodo-N6-methyl-(N)-methanocarba-2′-deoxyadenosine-
3′,5′-bisphosphate ([32P]MRS2500), a novel radioligand for quanti-
fication of native P2Y1 receptors. Br. J. Pharmacol. 2006, 147, 459-
467.
(17) Palczewski, K.; Kumasaka, T.; Hori, T.; Behnke, C. A.; Motoshima,
H.; Fox, B. A.; Le, T., I; Teller, D. C.; Okada, T.; Stenkamp, R. E.;
Yamamoto, M.; Miyano, M. Crystal structure of rhodopsin: A G
protein-coupled receptor. Science 2000, 289, 739-745.
(18) Moro, S.; Guo, D.; Camaioni, E.; Boyer, J. L.; Harden, T. K.;
Jacobson, K. A. Human P2Y1 receptor: molecular modeling and site-
directed mutagenesis as tools to identify agonist and antagonist
recognition sites. J. Med. Chem. 1998, 41, 1456-1466.
(19) Ferrara, P.; Gohlke, H.; Price, D. J.; Klebe, G.; Brooks, C. L., III.
Assessing scoring functions for protein-ligand interactions. J. Med.
Chem. 2004, 47, 3032-3047.
(20) Warren, G. L.; Andrews, C. W.; Capelli, A. M.; Clarke, B.; LaLonde,
J.; Lambert, M. H.; Lindvall, M.; Nevins, N.; Semus, S. F.; Senger,
S.; Tedesco, G.; Wall, I. D.; Woolven, J. M.; Peishoff, C. E.; Head,
M. S. A Critical Assessment of Docking Programs and Scoring
Functions. J. Med. Chem. 2006, 49, 5912-5931.
(21) Costanzi, S.; Joshi, B. V.; Maddileti, S.; Mamedova, L.; Gonzalez-
Moa, M. J.; Marquez, V. E.; Harden, T. K.; Jacobson, K. A. Human
P2Y6 receptor: molecular modeling leads to the rational design of a
novel agonist based on a unique conformational preference. J. Med.
Chem. 2005, 48, 8108-8111.
(22) Chang, G.; Guida, W. C.; Still, W. C. An Internal Coordinate Monte
Carlo Method for Searching Conformational Space. J. Am. Chem.
Soc. 1989, 111, 4379-4386.
(23) Glide. [4.0]. 2005. Schro¨dinger, LLC. ref Type: Computer Program.
(24) Aqvist, J.; Medina, C.; Samuelsson, J. E. A new method for predicting
binding affinity in computer-aided drug design. Protein Eng 1994,
7, 385-391.
Supporting Information Available: A table reporting the
structures of all the compounds in the training set, their potencies,
and references to the original publications (Table S1); a table
reporting the MOE QSAR prediction of the potency of the test set
compounds after addition of compound 6 or compound 7 to the
training set (Table S2); plots of experimental vs calculated potencies
for the 45 compounds of the training set (Figure S1); enlarged
versions of all the panels of Figure 1 and Figure 2 (Figure S2);
plots of experimental vs calculated potencies for the eight com-
pounds of the test set (Figure S3); synoptic view of experimental
vs calculated potencies for the eight compounds of the test set
(Figure S4); a table reporting the purity of the target compounds
as measured by HPLC. This material is available free of charge
via the Internet at http://pubs.acs.org.
References
(1) Costanzi, S.; Mamedova, L.; Gao, Z. G.; Jacobson, K. A. Architecture
of P2Y nucleotide receptors: structural comparison based on
sequence analysis, mutagenesis, and homology modeling. J. Med.
Chem. 2004, 47, 5393-5404.
(2) Abbracchio, M. P.; Burnstock, G.; Boeynaems, J. M.; Barnard, E.
A.; Boyer, J. L.; Kennedy, C.; Knight, G. E.; Fumagalli, M.; Gachet,
C.; Jacobson, K. A.; Weisman, G. A. International Union of
Pharmacology LVIII: update on the P2Y G protein-coupled nucle-
otide receptors: from molecular mechanisms and pathophysiology
to therapy. Pharmacol. ReV. 2006, 58, 281-341.
(3) Gachet, C.; Hechler, B. The platelet P2 receptors in thrombosis.
Semin. Thromb. Hemostasis 2005, 31, 162-167.
(4) Jin, J.; Kunapuli, S. P. Coactivation of two different G protein-coupled
receptors is essential for ADP-induced platelet aggregation. Proc.
Natl. Acad. Sci. U.S.A. 1998, 95, 8070-8074.
(5) Savi, P.; Herbert, J. M. Clopidogrel and ticlopidine: P2Y12 adenosine
diphosphate-receptor antagonists for the prevention of atherothrom-
bosis. Semin. Thromb. Hemostasis 2005, 31, 174-183.
(6) Hechler, B.; Nonne, C.; Roh, E. J.; Cattaneo, M.; Cazenave, J. P.;
Lanza, F.; Jacobson, K. A.; Gachet, C. MRS2500 [2-iodo-N6-methyl-
(N)-methanocarba-2′-deoxyadenosine-3′,5′-bisphosphate], a potent,
selective, and stable antagonist of the platelet P2Y1 receptor with
strong antithrombotic activity in mice. J. Pharmacol. Exp. Ther. 2006,
316, 556-563.
(25) Jones-Hertzog, D. K.; Jorgensen, W. L. Binding affinities for
sulfonamide inhibitors with human thrombin using Monte Carlo
simulations with a linear response method. J. Med. Chem. 1997, 40,
1539-1549.
(26) Liaison. [4.0]. 2005. Schro¨dinger, LLC. Ref Type: Computer
Program.
(7) Boyer, J. L.; Romero-Avila, T.; Schachter, J. B.; Harden, T. K.
Identification of competitive antagonists of the P2Y1 receptor. Mol.
Pharmacol. 1996, 50, 1323-1329.

P2Y1 Antagonists
Journal of Medicinal Chemistry, 2007, Vol. 50, No. 14 3241
(27) Zhou, R.; Friesner, R. A.; Ghosh, A.; Rizzo, R. C.; Jorgensen, W.
L.; Levy, R. M. New Linear Interaction Method for Binding Affinity
Calculations Using a Continuum Solvent Model. J. Phys. Chem. B
2001, 105, 10388-10397.
(35) Harden, T. K.; Hawkins, P. T.; Stephens, L.; Boyer, J. L.; Downes,
C. P. hosphoinositide hydrolysis by guanosine 5′-[gamma-thio]-
triphosphate-activated phospholipase C of turkey erythrocyte mem-
branes. Biochem. J. 1988, 252, 583-593.
(36) Boyer, J. L.; Downes, C. P.; Harden, T. K. Kinetics of activation of
phospholipase C by P2Y purinergic receptor agonists and guanine
nucleotides. J. Biol. Chem. 1989, 264, 884-890.
(37) Waldo, G. L.; Corbitt, J.; Boyer, J. L.; Ravi, G.; Kim, H. S.; Ji, X.
D.; Lacy, J.; Jacobson, K. A.; Harden, T. K. Quantitation of the P2Y1
receptor with a high affinity radiolabeled antagonist. Mol. Pharmacol.
2002, 62, 1249-1257.
(38) Labute, P. A widely applicable set of descriptors. Journal of
Molecular Graphics Modeling 2000, 18, 464-477.
(39) Wildman, S. A.; Crippen, G. M. Prediction of physicochemical
parameters by atomic contributions. J. Chem. Inf. Comput. Sci. 1999,
39, 868-873.
(40) Ertl, P.; Rohde, B.; Selzer, P. Fast calculation of molecular polar
surface area as a sum of fragment-based contributions and its
application to the prediction of drug transport properties. J. Med.
Chem. 2000, 43, 3714-3717.
(41) Hou, T. J.; Xia, K.; Zhang, W.; Xu, X. J. ADME evaluation in drug
discovery. 4. Prediction of aqueous solubility based on atom
contribution approach. J. Chem. Inf. Comput. Sci. 2004, 44, 266-
275.
(42) Kier, L. B.; Hall, L. H. Nature of Structure-activity-Relationships
and Their Relation to Molecular Connectivity. Eur. J. Med. Chem.
1977, 12, 307-312.
(28) Singh, P.; Mhaka, A. M.; Christensen, S. B.; Gray, J. J.; Denmeade,
S. R.; Isaacs, J. T. Applying linear interaction energy method for
rational design of noncompetitive allosteric inhibitors of the sarco-
and endoplasmic reticulum calcium-ATPase. J. Med. Chem. 2005,
48, 3005-3014.
(29) MOE. [2005.06]. 2005. Chemical Computing Group Inc. Ref Type:
Computer Program.
(30) Yoshimura, Y.; Moon, H. R.; Choi, Y.; Marquez, V. E. Enantiose-
lective synthesis of bicyclo[3.1.0]hexane carbocyclic nucleosides via
a lipase-catalyzed asymmetric acetylation. Characterization of an
unusual acetal byproduct. J. Org. Chem. 2002, 67, 5938-5945.
(31) Tchilibon, S.; Joshi, B. V.; Kim, S. K.; Duong, H. T.; Gao, Z. G.;
Jacobson, K. A. (N)-methanocarba 2,N6-disubstituted adenine nucleo-
sides as highly potent and selective A3 adenosine receptor agonists.
J. Med. Chem. 2005, 48, 1745-1758.
(32) MacroModel. [9.1]. 2005. Schro¨dinger, LLC. Ref Type: Computer
Program.
(33) Mohamadi, F.; Richards, N. G. J.; Guida, W. C.; Liskamp, R.; Lipton,
M.; Caufield, C.; Chang, G.; Hendrickson, T.; Still, W. C. Macro-
model - An Integrated Software System for Modeling Organic and
Bioorganic Molecules Using Molecular Mechanics. J. Comput. Chem.
1990, 11, 440-467.
(34) Friesner, R. A.; Murphy, R. B.; Repasky, M. P.; Frye, L. L.;
Greenwood, J. R.; Halgren, T. A.; Sanschagrin, P. C.; Mainz, D. T.
Extra precision glide: docking and scoring incorporating a model
of hydrophobic enclosure for protein-ligand complexes. J. Med.
Chem. 2006, 49, 6177-6196.
JM0700971