Pyrimidine derivatives as potent and selective A3 adenosine receptor antagonists

Article abstract of DOI:10.1021/jm100843z

Two regioisomeric series of diaryl 2- or 4-amidopyrimidines have been synthesized and their adenosine receptor affinities were determined in radioligand binding assays at the four human adenosine receptors (hARs). Some of the ligands prepared herein exhib

Full text of DOI:10.1021/jm100843z

J. Med. Chem. 2011, 54, 457–471 457
DOI: 10.1021/jm100843z
Pyrimidine Derivatives as Potent and Selective A3 Adenosine Receptor Antagonists
§
,§
†,‡
#
Vicente Yaziji,†,‡ David Rodrı
guez, Hugo Gutierrez-de-Teran,* Alberto Coelho, Olga Caamano,
ꢀ ꢀ ~
´
a Isabel Loza,*, Marıa Isabel Cadavid, and Eddy Sotelo*,†,‡,#
ꢀ
Xerardo Garcıa-Mera, Jose Brea, Marı
´ ´ ´
#
†Combinatorial Chemistry Unit (COMBIOMED), Institute of Industrial Pharmacy, ‡Center for Research in Biological Chemistry and Molecular
Materials, University of Santiago de Compostela, §Public Galician Foundation of Genomic Medicine, Santiago de Compostela University Hospital
(CHUS), 15706, Santiago de Compostela, Spain, #Department of Organic Chemistry, and Department of Pharmacology, Faculty of Pharmacy,
University of Santiago de Compostela, Santiago de Compostela, 15782, Spain
Received June 11, 2010
Two regioisomeric series of diaryl 2- or 4-amidopyrimidines have been synthesized and their adenosine
receptor affinities were determined in radioligand binding assays at the four human adenosine receptors
(hARs). Some of the ligands prepared herein exhibit remarkable affinities (Ki<10 nm) and, most
noticeably, the absence of activity at the A1, A2A, and A2B receptors. The structural determinants that
support the affinity and selectivity profiles of the series were highlighted through an integrated
computational approach, combining a 3D-QSAR model built on the second generation of GRid
INdependent Descriptors (GRIND2) with a novel homology model of the hA3 receptor. The robustness
of the computational model was subsequently evaluated by the design of new derivatives exploring the
alkyl substituent of the exocyclic amide group. The synthesis and evaluation of the novel compounds
validated the predictive power of the model, exhibiting excellent agreement between predicted and
experimental activities.
Introduction
therapeutic applications derived from the modulation of this
receptor subtype have been reviewed recently.11-18 In parti-
cular, it is becoming increasingly apparent that antagonists of
A3AR might be therapeutically useful for the acute treatment
of stroke and glaucoma,19 inflammation,20-22 and in the
development of cerebroprotective,23,24 antiasthmatic and
antiallergic drugs.25,26 Furthermore, recent evidence27-31
of high levels of expression of A3ARs in several cell lines has
suggested potential applications for A3AR antagonists in
cancer chemotherapy.
The putative applications of these compounds as drugs, as
well as the growing demand for pharmacological tools to
study the human A3AR roles, has made the identification of
potent and selective small molecule antagonists of this receptor
subtype a topic of great interest.11-18 The search for A3AR
antagonists began with the observation that xanthines ; a
successful structural motif in the search for antagonists for the
other ARs subtypes ; exhibit low binding affinities for the A3
receptor subtype. The pursuit of A3AR antagonists therefore
focused on the exploration of structurally diverse heterocyclic
libraries. Nowadays, the best known class of A3AR ligands
(Figure 1) includes highly diverse families of tri- and bicyclic
heteroaromatic scaffolds and, to a lesser extent, mono-hetero-
cyclic systems. Whereas the systematic structural elaboration
of these prototypes has provided derivatives possessing good
affinity,11-18 the selectivity issue and the relatively poor
bioavailability profiles of drug candidates have remained
elusive until recently.5-7
The ubiquitous nucleoside adenosine is essential for the
proper functioning of every cell in mammalian species. Ade-
nosine is directly linked to energy metabolism through ATP,
ADP, and AMP, while at the extracellular level it regulates a
wide range of biological functions through activation of
specific receptors (adenosine receptors, ARs),1-3 which are
classified as A1, A2A, A2B, and A34 and belong to the super-
family of the G-protein coupled receptors (GPCRsa). The
improved understanding of the physiology, pharmacology,
structure, and molecular biology of adenosine and its recep-
tors has provided solid foundations that support the potential
the development of conceptually unexplored therapeutic stra-
tegies to address serious unmet medical needs. The advances
in the medicinal chemistry of this emerging family of ther-
apeutics have been reviewed recently.5-7
The A3AR subtype is the most recently characterized
member of the family.8 Activation9,10 of this subtype has been
shown to inhibit adenylate cyclase, to increase phosphatidy-
linositol-specific phospholipase C and D activity, to elevate
intracellular Ca2þ and IP3 levels, and to enhance the release of
inflammatory and allergic mediators from mast cells. The
*To whom correspondence should be addressed. (H.G.) tel.: þþ34-
881813873, e-mail: hugo.teran@usc.es. (M.I.L.) tel.: þþ34-881815005,
e-mail: mabel.loza@usc.es. (E.S.) tel.: þþ34-881815221, Fax.: þþ34-
981-528093, e-mail: eddy.sotelo@usc.es.
a Abbreviations: hARs, human adenosine receptors; ADP, adenosine
diphosphate; ATP, adenosine-50-triphosphate; AMP, adenosine mono-
phosphate; GPCRs, G-protein coupled receptors; 3D-QSAR, three-
dimensional quantitative structure-activity relationships, GRIND2,
GRid INdependent Descriptors; IP3, inositol trisphosphate; CHO cells,
Chinese hamster ovary cells; CLACC, consistency large auto and cross
correlation; MIF, molecular interaction fields; PDB, Protein Data
Bank; LOO, leave-one out; rmsd, root mean square deviation.
The pyrimidine core, being part of the heterocyclic moiety
of the endogenous ligand of these receptors (adenosine), is a
recurrent substructural motif within bi- and tricyclic ARs
antagonists.5-7 The well-documented contributions of this
chemotype to the field notwithstanding, relatively few papers
r
2010 American Chemical Society
Published on Web 12/27/2010
pubs.acs.org/jmc

458 Journal of Medicinal Chemistry, 2011, Vol. 54, No. 2
Yaziji et al.
Figure 1. Structures of representative selective A3 adenosine receptor antagonists.
Figure 2. Structures and biological data for representative diphenyl 2- or 4-amidopyrimidines as selective A1 adenosine receptor antagonists.
The substitutions further explored in the present report follow those established in the early A1AR model,39 labeled as L1, L2, and L3,
indicating the lipophilic pockets in the receptor. Note that there is no substitutuent in L1 in the series of aminopyrimidines and that L2=L3 for
all compounds described herein.
have concerned focused programs based on this scaffold,32-36
or its biososters (e.g., triazines).37 Two recent publications
have covered (1) a molecular simplification study from tria-
zoloquinoxalines to pyrimidines38 and (2) the elaboration of a
pharmacophoric model for A1 adenosine receptors, based on
structurally simple regioisomeric diarylpyrimidine scaffolds
(Figure 2).39 The latter work not only enabled the identifica-
tion of potent and selective A1AR antagonists derived from
either the 4,6-diphenyl-2-amidopyrimidine or 2,6-diphenyl-
4-amidopyrimidine templates (Figure 2) but also provided a
valuable structural model that could be exploited for the
design of new series of compounds.
In light of these precedents, and particularly residual
activity toward the hA3AR subtype observed for some pre-
viously reported compounds (Figure 2),39 it was envisioned
that the structural redecoration of the aryl fragments on the
amidopyrimidine templates would modify the adenosine
receptor selectivity profile and provide new selective A3AR
antagonists. We therefore focused on the exploration of
diverse aryl moieties on the heterocyclic scaffold, with parti-
cular attention paid to structural elements that had been
previously identified as contributors in the molecular recogni-
tion ofthe A3ARsubtype(e.g., 4-methoxyphenylgroup).40-43
From methodological and practical points of view, it was
decided first to explore the synthesis and screening of libraries
incorporating identical aryl groups at positions 4,6 and 2,6, as
a proof of concept. Thereafter, depending on the results of this
first series (reported inthe current manuscript) the synthesisof
nonidentical series will be performed. The design of the new
chemical entities was assisted, and interpreted, by developing
an integrated molecular modeling approach that combined
ligand docking and 3D quantitative-structure activity (3D-
QSAR) studies. Although limitations in the homology model-
ing of ARs in the design of new ligands have recently been
recognized,44-46 the recent release of the crystal structure of
human A2AAR in complex with the potent inhibitor 4-(2-[7-
amino-2-(2-furyl)[1,2,4]triazolo[2,3-a][1,3,5]triazin-5-ylamino]-
ethyl)phenol (ZM241385)47 has been a breakthrough in this

Article
Journal of Medicinal Chemistry, 2011, Vol. 54, No. 2 459
Scheme 1a
a Reagents: (a) Pd[(PPh)3]4, DME/H2O, Na2CO3, (b) THF, TEA.
area, as occurred earlier with the release of the structure
of the hβ2 adrenergic receptor.49 In fact, the most recent
models of the A3AR have already taken advantage of this
crystal structure in the description of receptor-antagonist
recognition.48 On the other hand, structure-based approaches
have frequently been combined in G protein-coupled receptor
(GPCR) research with ligand-based techniques, such as phar-
macophore models39 or 3D-QSAR studies.46 In the present
work, a new A3AR model, derived from the recent crystal
structure of A2AAR, is reported and used as a basis for the
automated docking of the series reported here. In a first
iteration, an initial batch of compounds was synthesized,
tested, and computationally investigatedfor the bindingmode
of the series, which guided the design of the rest of the
compounds series. Once the experimental affinities were avail-
able for the compounds here reported, the new ligands were
computationally described and their structure-affinity was
modeled by using the most recent version of the GRid
INdependent Descriptors (GRIND-2),50,51 thus providing a
rational interpretation of the structure-activity and structure-
selectivity relationships. To further challenge the computa-
tional model in terms of robustness and predictive capability,
it was used for the design of novel compounds bearing new
alkyl substitutions on the L1 site (see Figure 2). The synthesis
and evaluation of the novel compounds validated the pre-
dictive power of the model, exhibiting excellent agreement
between predicted and experimental activities.
residues on the functionalized pyrimidine templates, a short
and divergent synthetic strategy was optimized.52 The syn-
thetic pathway developed to access the designed regioisomeric
libraries is presented in Scheme 1, and this relied on the
commercial availability of the 2- or 4-aminodichloropyrimi-
dines 1a-b as precursors. Application of the standard condi-
tions of the highly reliable and well-established Suzuki-
Miyaura cross-coupling reaction to a collection of commer-
cially available boronic acids (2a-q), which representatively
cover both the aryl and heteroaryl series (Scheme 1), enabled
the rapid decoration of the heterocyclic core at positions
2,4- or 4,6- to afford diarylpyrimidinamines 3a-q and 4a-q,
which can be considered bioisosteres of previously described
2-amino-4,6-diaryltriazines.37 Derivatization of the amine
function in the heterocyclic precursors 1a-b by treatment
with acid chlorides 5a-c and subsequent palladium-catalyzed
(hetero)arylation afforded two regiosomeric series of di(hetero)-
aryl 2- or 4-amidopyrimidines (8-13).
In an attempt to validate the robustness and predictive
capability of the herein developed computational model some
computer-generated new ligands, designed to evaluate the
tolerance of A3AR to the introduction of bulky alkyl residues
in the amide moiety (L1) of the pyrimidin-4-amine series, were
prepared. Treatment of two representative amines, incorpo-
rating binding residues that conferred high A3 AR affinity (4d
and 4k), with three additional acid chlorides (Scheme 2)
afforded the new structures 14a-f.
The synthetic program provided a focused library of 142
members, which in turn can be subdivided into two regioiso-
meric sublibraries [34 amines (3-4) and 108 amides (8-14)]
that were structurally characterized. A detailed account of the
experimental procedures and the complete description of the
Chemistry
Given that the feasibility of the proposed aim is heavily
reliant on the exhaustive exploration of diverse (hetero)aryl

460 Journal of Medicinal Chemistry, 2011, Vol. 54, No. 2
Yaziji et al.
Scheme 2a
the 4,6-diaryl-2-amidopyrimidines (8a-q-10a-q, Table 1) and
the 2,6-diaryl-4-amidopyrimidines (11a-q-13a-q, Table 2),
as well as for the isomeric amine series (3a-q and 4a-q,
see Supporting Information). Examination of the binding
data indicates that new potent and highly selective ligands
for the A3 receptor subtype have been identified (Table 2,
compounds 11b, 11d, 12d, 13d, 11f, 12h, 11j, 11k, 12k, 13k,
11m). These results validate the initial hypothesis that the
appropriate decoration of the heterocyclic scaffold with pre-
viously unexplored diversities would lead to remarkable
modifications in the pharmacological activity in comparison
to the published results for analogous compounds. Moreover,
the documented data exemplify how the structural manipula-
tion of these privileged scaffolds is able to modify the bio-
logical profile, not only at the quantitative (affinity) level but
also at the qualitative (selectivity) level.
Bearing in mind the considerable number of compounds
tested, and for the sake of brevity and clarity, the analysis and
interpretation of the data will be carried out at two levels. On
the one hand, the most prominent features of the structure-
activity (SAR) and structure-selectivity (SSR) relationships
for both series will be discussed qualitatively. On the other
hand, a more in-depth and quantitative structure-activity
relationship can be obtained on the basis of an integrated
molecular modeling study. Suchan analysiswas performedon
the set of 64 compounds with experimental Ki values in the A3
receptor, and this represents a novel approach based on the
combination of molecular docking on a homology model for
the A3AR and a 3D-QSAR study.
It can be observed from the biological data the amine series
(3a-q and 4a-q) did not exhibit attractive pharmacological
profiles at any of the ARs (see Supporting Information). The
moderate affinity toward the A1 receptor subtype elicited by
the parent compounds of the regioisomeric series (Ar=Ph,
compounds 3a and 4a) was generally extinguished by the
introductionofgroups at the phenyl ringsor their replacement
by diverse heterocyclic cores. The generally disappointing
binding data are common to both regioisomeric amine subsets
(3 and 4). The most remarkablederivative within the series(4l)
combines a potent A1AR antagonistic effect (Ki=7.99 nM)
and a satisfactory selectivity (>30) versus the human A3AR
subtype.
Inspection of the pharmacological data obtained for the
most populated set of compounds prepared in this work (i.e.,
the diaryl 2- or 4-amidopyrimidines 8a-q-13a-q, Tables 1
and 2) confirms that the systematic modification of the
structural prototypes produced a significant, but differen-
tiated, variation in their biological behavior. A comparative
analysis of these data highlights the different activity profiles
elicited for the two regioisomeric series toward ARs (Tables 1
and 2). Thus, compounds that incorporate the amide moiety
at position 4 of the heterocyclic core afforded the most
interesting derivatives identified during this study, while their
regioisomeric congeners gave a somewhat poor activity pro-
file. Within the 2-amidopyrimidine series only those ligands
bearing tolyl groups at positions 4 and 6 of the heterocyclic
core (compounds 8b, 9b, and 10b) and the N-[2,6-di(benzo-
[d][1,3]dioxol-5-yl)pyrimidin-4-yl]acetamide (8l) elicited mod-
erate A3AR affinity (Table 1).
a Reagents: (a) THF, TEA.
analytical and spectroscopic data for all compounds are
available in the Supporting Information.
Biological Evaluation
The affinities of the obtained compounds atthe four human
adenosine receptor subtypes were determined in vitro using
radioligand binding assays according to experimental proto-
cols described elsewhere.53 Human adenosine receptors ex-
pressed in transfected CHO (A1AR), HeLa (A2AAR and
A3AR), and HEK-293 (A2BAR) cells were employed. (3H)-
1,3-Dipropyl-8-cyclopentylxanthine ([3H]DPCPX) for A1AR
and A2BAR, [3H]4-(2-[7-amino-2-(2-furyl)[1,2,4]triazolo[2,3-
a][1,3,5]triazin-5-ylamino]ethyl)phenol for A2AAR, and
[3H]NECA for A3AR were employed as radioligands in
binding assays. The biological data (Tables 1-3) are ex-
pressed as Ki ( SEM (nM, n=3) or percentage of inhibition
of specific binding at 0.1 μM (n = 2, average) for those
compounds that did not fully displace radioligand binding.
Functional Assay at Adenosine A3 Receptors
Some representative ligands that show affinity toward the
hA3AR subtype were also studied through cAMP experi-
ments (see Figure 3 and Table 3). The functional evaluation
was carried out with intact cells expressing the hA3AR.
The inhibition of forskolin-stimulated cAMP production by
the receptor agonist was used as a read-out. Concentration-
response curves of two representative compounds (compounds
11d and 12d) over 0.1 μM NECA-induced A3AR activation
are shown in Figure 3. cAMP formation was measured by
enzyme immunoassay (GE Healthcare). Antagonistic potency,
measured as KB, was calculated from the formula: KB =
(IC50)/((2 þ ([A]/[A50]n)1/n - 1), where IC50 is the concentra-
tion of the antagonist that inhibits the agonist stimulation
by 50%, [A] is the concentration of the agonist in the assay,
[A50] is the concentration of the agonist that elicits the half-
maximum response, and n is the slope of the concentration
response curve.54
All these derivatives fully reverted the A3AR-elicited in-
hibition of cAMP accumulation, unequivocally validating the
antagonistic behavior of these compounds at the human A3
AR. Moreover, a comparative analysis of the KB values for
these compounds (Table 3) during the cAMP experiments
revealed a clear correspondence with the affinity values
determined during the binding experiments (Ki in Table 4).
In clear contrast to previously discussed results for the
amines (3 and 4) and 4,6-diaryl-2-amidopyrimidines (8-10),
the biological data obtained for the 2,6-diaryl-4-amidopyr-
imidine subset (Table 2, compounds 11-13) unequivocally
show the determinant influence that the varied structural
Structure-Activity Relationship and Molecular Modeling
Affinities in radioligand binding assays at the four human
adenosine receptors (A1, A2A, A2B, and A3) are reported for

Article
Journal of Medicinal Chemistry, 2011, Vol. 54, No. 2 461
Table 1. Structure and Affinity Binding Data for the 4,6-Diaryl-2-amidopyrimidines 8, 9, and 10 at the Human Adenosine Receptors
Ki (nM) or % at 0.1 μM
a
b
c
d
comp
Ar
R
hA1
hA2A
hA2B
hA3
8a
9a
Ph
Me
Et
17%
31.3 ( 2.1
10.3 ( 1.9
3%
11%
554 ( 32
3%
6%
13%
531 ( 31
17%
42.5 ( 3.4
14%
3%
10a
8b
Pr
4-Me-Ph
Me
Et
6%
1%
47.3 ( 4.7
157 ( 22
131 ( 19
10%
9b
1%
3%
10b
8c
Pr
3760 ( 225
2%
2%
2%
1%
10%
2%
4-CF3-Ph
4-MeO-Ph
4-MeS-Ph
4-MeCO-Ph
4-F-Ph
Me
Et
9c
12%
2%
9%
2%
2%
10c
8d
Pr
1%
1%
13%
1%
Me
Et
7%
1%
20%
13%
9d
2%
2%
8%
9%
10d
8e
Pr
24%
1%
15%
14%
Me
Et
1%
2%
374 ( 22
2%
9e
505 ( 36
3%
7%
183 ( 16
7%
8%
4435 ( 85
3%
2%
10e
8f
Pr
Me
Et
14%
1%
9f
1%
7%
12%
14%
2%
17%
22%
10f
8g
Pr
2%
1%
Me
Et
2%
8%
22%
9g
14%
18%
2%
3%
5%
1%
2%
10g
8h
Pr
27%
1%
4-Cl-Ph
Me
Et
2%
1%
2%
9h
1%
1%
11%
3%
14%
17%
10h
8i
Pr
1%
2-F-Ph
Me
Et
8%
12%
424 ( 38
647 ( 101
10%
13%
11%
2%
10%
307 ( 19
6%
15%
9i
48.2 ( 4.6
23.5 ( 1.8
4%
279 ( 24
15%
2%
10i
8j
Pr
2-MeO-Ph
2,4-MeO-Ph
3,4- (CH2-O2)-Ph
Ph-CH=CH-
2-furan
Me
Et
1%
2%
9j
21%
13%
1%
22%
18%
10j
8k
9k
10k
8l
Pr
1%
1%
Me
Et
9%
17%
3%
14%
16%
15%
16%
8%
3%
Pr
14%
1%
15%
1%
1%
Me
Et
101 ( 7
127 ( 4
751 ( 18
4%
9l
535 ( 37
678 ( 42
1%
1%
1%
10l
8m
9m
10m
8n
Pr
Me
Et
15%
1%
1%
1%
1%
1%
142 ( 9
12%
9%
Pr
3%
2%
Me
Et
1%
19%
15%
23%
5%
11%
21%
15%
3%
9n
24%
17%
2%
1903 ( 116
1%
21%
10n
8o
Pr
2-thiophene
3-furan
Me
Et
9o
24%
255 ( 31
1%
25%
201 ( 28
1%
2%
3%
21%
24%
10o
8p
Pr
Me
Et
1%
1%
16%
9p
3%
2%
11%
0%
10p
8q
Pr
11%
2%
17%
2%
21%
1%
3-thiophene
Me
Et
1%
3250 ( 261
12%
9q
10q
367 ( 25
85.7 ( 5.2
1893 ( 174
4568 ( 251
1279 ( 215
2%
Pr
a Displacement of specific [3H]DPCPX binding in human CHO cells expressed as Ki ( SEM in nM (n=3) or percentage displacement of specific
binding at a concentration of 0.1 μM (n = 2). b Displacement of specific [3H]4-(2-[7-amino-2-(2-furyl)[1,2,4]triazolo[2,3-a][1,3,5]triazin-5-ylami-
no]ethyl)phenol binding in human HeLa cells expressed as Ki ( SEM in nM (n=3) or percentage displacement of specific binding at a concentration
of 0.1 μM (n=2). c Displacement of specific [3H]DPCPX binding in human HEK-293 cells expressed as Ki ( SEM in nM (n=3) or percentage
displacement of specific binding at a concentration of 0.1 μM (n=2). d Displacement of specific [3H]NECA binding in human HeLa cells expressed as
Ki ( SEM in nM (n=3) or percentage displacement of specific binding at a concentration of 0.1 μM (n = 2).
parameters have on the antagonistic profile of these series.
The exhaustive exploration of the scaffold enabled the
identification of structurally simple derivatives that exhibit
outstanding affinity and remarkable selectivity for the A3AR

462 Journal of Medicinal Chemistry, 2011, Vol. 54, No. 2
Yaziji et al.
Table 2. Structure and Affinity Binding Data for the 2,6-Diaryl-4-amidopyrimidines 11, 12, and 13 at the Human Adenosine Receptors
Ki (nM) or % at 0.1 μM
a
b
c
d
comp
Ar
R
hA1
hA2A
hA2B
hA3
11a
12a
13a
Ph
Me
Et
31.2 ( 4.1
22.3 ( 3.3
19.5 ( 3.2
2%
255.3 ( 13
84.5 ( 5.7
103 ( 8
8%
19%
76.6 ( 6.4
1%
12.1 ( 1.3
45.5 ( 7.4
171 ( 21
4.4 ( 0.3
18.3 ( 1.9
59.0 ( 2.3
126 ( 11
12%
Pr
11b (ISVY133)
12b
4-Me-Ph
Me
Et
2%
2%
36.9 ( 4.1
16%
1%
1%
2%
13b
11c
Pr
3%
1%
4-CF3-Ph
4-MeO-Ph
4-MeS-Ph
4-MeCO-Ph
4-F-Ph
Me
Et
1%
17%
12c
13c
3%
3%
8%
6%
Pr
1%
10%
12%
11d (ISVY130)
12d (ISVY074)
Me
Et
1%
8%
4%
3%
3.6 ( 0.2
3.6 ( 0.40
11.0 ( 1.3
71.3 ( 3.5
43.6 ( 1.7
12%
1%
4%
13d (ISVY071)
11e
Pr
8%
1%
1%
2%
Me
Et
1%
1%
12e
13e
1%
3%
3%
1%
Pr
2%
13%
11f
12f
Me
Et
10%
1%
16%
2%
25.2 ( 0.7
43.9 ( 2.4
133 ( 20
16.7 ( 1.4
12.1 ( 0.6
34.8 ( 3.1
63.3 ( 8.2
25.3 ( 0.5
103 ( 6
18.1 ( 0.7
160 ( 14
135 ( 11
24.1 ( 1.3
23.2 ( 0.8
110 ( 15
5.4 ( 0.1
11.3 ( 1.4
10.2 ( 1.1
3.3 ( 0.3
14.5 ( 1.2
59.0 ( 4.3
15.6 ( 2.1
46.9 ( 5.4
25%
1%
1%
13f
11g
Pr
2%
15%
1%
1%
Me
Et
1334 ( 110
429 ( 18
1829 ( 47
10%
12g
13g
83.9 ( 5.0
82.3 ( 3.4
1%
1%
2%
Pr
11h
12h
4-Cl-Ph
Me
Et
1%
3%
16%
16%
20%
21%
13h
11i
Pr
1%
2-F-Ph
Me
Et
17%
73.8 ( 6.0
103 ( 5
142 ( 7
14%
21%
16%
9%
12i
13i
31.6 ( 4.1
18.7 ( 2.5
1%
Pr
11j
12j
2-MeO-Ph
2,4-MeO-Ph
3,4- (CH2-O2)-Ph
Ph-CH=CH-
2-furan
Me
Et
1%
7%
113 ( 9
41.3 ( 2.6
1%
22%
14%
13J
Pr
3%
2%
11k (ISVY167)
12k (ISVY169)
13k
Me
Et
6%
14%
6%
18%
8%
Pr
12%
14%
2%
11l
12l
Me
Et
17.7 ( 3.1
5.28 ( 0.8
9.7 ( 1.4
1%
3345 ( 127
2541 ( 64
22%
1668 ( 39
16%
1%
13l
Pr
11m
12m
13m
11n
Me
Et
13%
2%
17%
1%
2%
8%
Pr
1%
Me
Et
40.7 ( 5.2
15.5 ( 3.1
7.8 ( 0.9
19%
8.1 ( 1.2
6.4 ( 0.7
5.7 ( 0.4
24.6 ( 2.6
114 ( 7
153 ( 11
74.0 ( 3.2
82.8 ( 6.0
544 ( 11
39.8 ( 4.4
63.5 ( 4.1
23%
12.0 ( 1.1
20.5 ( 2.4
16.4 ( 0.7
23%
17%
8%
3.0 ( 0.4
6.2 ( 0.7
9.9 ( 1.2
8.0 ( 0.4
21.8 ( 2.2
23.0 ( 4.0
10.1 ( 0.9
12.6 ( 1.1
3%
12n
13n
Pr
11o
12o
2-thiophene
3-furan
Me
Et
32.9 ( 1.4
33.3 ( 3.2
1%
13o
11p
Pr
Me
Et
302 ( 67
49.0 ( 5.3
1%
12p
13p
132 ( 10
65.1 ( 4.0
13%
Pr
11q
12q
3-thiophene
Me
Et
16%
164.7 ( 47
21%
20.2 ( 3.2
64.2 ( 5.6
11%
39.4 ( 2.3
65.1 ( 4.7
13q
Pr
a Displacement of specific [3H]DPCPX binding in human CHO cells expressed as Ki ( SEM in nM (n=3) or percentage displacement of specific
binding at a concentration of 0.1 μM (n = 2). b Displacement of specific [3H]4-(2-[7-amino-2-(2-furyl)[1,2,4]triazolo[2,3-a][1,3,5]triazin-5-ylami-
no]ethyl)phenol binding in human HeLa cells expressed as Ki ( SEM in nM (n = 3) or percentage displacement of specific binding at a concentration
of 0.1 μM (n=2). c Displacement of specific [3H]DPCPX binding in human HEK-293 cells expressed as Ki ( SEM in nM (n=3) or percentage
displacement of specific binding at a concentration of 0.1 μM (n=2). d Displacement of specific [3H]NECA binding in human HeLa cells expressed as
Ki ( SEM in nM (n=3) or percentage displacement of specific binding at a concentration of 0.1 μM (n=2).
(see Table 2, compounds 11b, 11d, 12d, 13d, 11k, 12k, 13k). A
comparison of these data with the observed activity for the
parent compounds of the series (Table 2, Ar = Ph, com-
pounds 11a, 12a, 13a, relatively A1AR potent but somewhat

Article
Journal of Medicinal Chemistry, 2011, Vol. 54, No. 2 463
Table 3. Antagonistic Potency (Measured as KB) at Human A3 Recep-
tors of Selected Compoundsa
L2/L3 substitutions. The weak modulator effect exerted by the
alkyl residues of the amide functions (L1) on the activity/
selectivity profile within these series can be observed. It can be
clearly appreciated that the size of L1 is inversely correlated
with the affinity within each subseries, an observation that is
consistent with previous findings.38 Moreover, a detailed
inspection of the pharmacological data reported for these
series (Table 2) shows that A3AR selectivity also increases on
reducing the size of the L1 substituent. Only some combina-
tions of L2/L3 substituents show little sensitivity to the nature
of the L1 substituent, in particular, compounds incorporating
4-methoxyphenyl (11d, 12d, 13d) or 2,4-dimethoxyphenyl
(11k, 12k, 13k) residues (i.e., the substituent present in
compounds eliciting the highest affinity).
compound
KB (nM)
12d
13d
11k
11b
12h
1.40 ( 0.09
1.57 ( 0.13
3.85 ( 0.41
3.56 ( 0.23
3.88 ( 1.17
a Values represent the mean ( SEM of two separate experiments.
Once the initial hypothesis that hA3AR could be more
tolerant to bulky L2/L3 substituents had been validated by
the SAR data, an exhaustive molecular modeling study was
developed to gain new insights into the structure-affinity
relationship for the hA3AR. A homology model of the
hA3AR receptor was built using the recently crystallized
hA2AAR structure as a template. This model served as a basis
for an automated docking exploration of the 64 compounds
for which experimental Ki values at the hA3AR are reported.
The choice of the docking algorithm (GOLD program in
combination with the Chemscore scoring function)55 is the
result of an internal validation of different docking alterna-
tives in order to reproduce the experimental binding pose of
4-(2-[7-amino-2-(2-furyl)[1,2,4]triazolo[2,3-a][1,3,5]triazin-
5-ylamino]ethyl)phenol/hA2AAR (data not shown), a valida-
tion that is in agreement with a recent comparative study of
ligand docking tools in ARs.49 The systematic docking ex-
ploration identified one conserved binding mode for both
regioisomeric diaryl amidopyrimidine series reported here.
Thisbinding mode wasfound in62of the 64 compounds(97%
of the cases), and this mode was the top scored pose by
Chemscore in 66% of the cases. Moreover, in 63% of the
cases, this binding pose was the most populated according to
an rmsd tolerance of 1 A for the clustering. This binding mode
is represented in Figure 5 for compound 11d.
The main anchoring point is a double hydrogen bond of the
exocyclic amino/amido group (donating) and its closest nitro-
gen atom in the pyrimidine ring (N3, accepting) with Asn 6.55
(note the Ballesteros-Weinstein residue numbering56), a to-
tally conserved residue of the adenosine receptor family. At
the same time, the pyrimidine ring is flanked by the side chain
of Phe 5.29, in the second extracellular loop (EL2), and Leu
6.51 in helix 6. This interaction pattern of the aminopyrimi-
dine moiety (π-stacking with Phe 5.29, hydrophobic interac-
tions with Leu 6.51 and hydrogen bonding to Asn 6.55)
resembles the experimentally observed binding mode of
4-(2-[7-amino-2-(2-furyl)[1,2,4]triazolo[2,3-a][1,3,5]triazin-
5-ylamino]ethyl)phenol with the hA2AAR.47 Accordingly, the
important role in ligand binding of residues 5.29 and 6.51 has
recently been validated in a site-directed mutagenesis study of
the A2AAR.57 Interestingly, Phe 5.29 is totally conserved in
the ARs family, while Leu 6.51 is substituted by a smaller
valine in the low-affinity hA2BAR, and the replacement of this
residue by an alanine in hA2BAR completely abolishes ligand
binding.57 As far as position 6.55 is concerned, there is
biochemical evidence that suggests the important role of this
residue in ligand binding for several ARs, including the
A3AR.58-60 The molecular alignment of the 62 molecules,
obtained by ligand-docking, is shown in Figure 6. It can be
appreciatedthatthe volumeofthe L1andL2subsites hasbeen
Figure 3. Effect of 11d (O, dashed fitting) and 12d (b, black fitting)
on 0.1 μM NECA-induced cAMP decrease of 10 μM forskolin-
stimulated human A3 receptors. Points represent the mean ( SEM
(vertical bars) of two separate experiments.
promiscuous AR ligands) allows the rapid evaluation of the
effects caused by the structural modifications. In general,
modification of the aromatic substitution pattern completely
extinguished the affinity for the A1AR, while conferring
notable potency and selectivity toward the A3AR subtype.
Remarkably, such a subtle structural modification is able to
produce a radical variation in the activity profile, being
significant not only for a methoxy group at position 4 but
also for the more highly diverse residues explored (e.g.,
methyl, thiomethyl, acetyl, fluoro and chloro). It is also
remarkable that the vinyl analogues (Table 2, compounds
11m and 12m) of the parent compounds proved to be rela-
tively potent and highly selective A3 ligands, a finding that
reaffirms how bulky substituents at sites L2/L3 favor selec-
tivity toward hA3AR. The consequences of introducing a
group at position 2 of the phenyl ring was also briefly assessed
(Table 2, compounds 11i, 12i, 13i, 11j, 12j, and 13j). As
observed, the introduction of fluoro or methoxy groups at
this position afforded relatively potent derivatives, albeit with
markedly different selectivity profiles. Within this ligand
subset only the 2-methoxyphenyl derivative of the 4-aceta-
mide series (compound 11i) elicited a satisfactory affinity/
selectivity profile. Conversely, the simultaneous introduction
of methoxy groups at positions 2 and 4 of the phenyl ring
afforded highly potent and completely selective ligands (11k,
12k, and 13k) toward the A3AR subtype, regardless of the
alkylic residue present in the amide group at position 4 of
the heterocyclic backbone. Finally, in a clear contrast with the
results described sofar, replacement of the phenyl group inthe
parent compounds by heterocyclic cores proved to be highly
discouraging, generating a series of potent but nonselective
ligands.
An integrated analysis of the data presented in Table 2 for
the 4-amide homologous series (compounds 11-13) is shown
in Figure 4. In this representation the experimental Ki values
at hA3AR are plotted as a function of both the L1 and the

464 Journal of Medicinal Chemistry, 2011, Vol. 54, No. 2
Yaziji et al.
Figure 4. Effect of the nature of the L1 substituent on the human A3 adenosine receptor affinity for the series of 2,6-diaryl-4-amidopyr-
imidines. The figure only represents those compounds that have experimental Ki for the three considered L1 substitutions.
Figure 5. Binding mode of compound 11d, showing the main receptor-ligand interactions. Residues that are specific for the human A3
adenosine receptor are shown in boxes in the 3D panel (A), generated in PyMol (http://www.pymol.org). The double hydrogen bond with Asn
6.55 is indicated by dashed lines. Panel B shows a schematic representation of residue-ligand interactions, calculated with LigX as
implemented in MOE. Residues are labeled according to the Ballesteros & Weinstein numbering.56
well explored, while there is a volume tolerance in the subsite
occupied by L3 (helices 2, 3, 7) that was not completely
explored by our ligand series. Even after one energy mini-
mization cycle, the molecular alignment did not change sub-
stantially and the highest variability is still located on the L3 site.
This molecular alignment was the basis for a 3D-QSAR
study that involved the use of the new generation of Grid-
INdependent Descriptors (GRIND-2).50,51 The first generation
of these molecular interaction field (MIF)-based descriptors
was originally conceived precisely to circumvent the necessity
of obtaining a highly accurate molecular alignment of the
molecules prior to the 3D-QSAR analysis.61 However, the
most recent version of the GRIND methodology includes
a new mathematical transformation applied to the MIF
descriptors that guarantees that a given variable represents
exactly the same information for every compound of the
series.51 This method, called consistency large auto and cross
correlation (CLACC), either generates a molecular alignment
of the molecules, on the basis of the correlation of the
variables, or either it uses an input molecular alignment
provided by the user (e.g., obtained by molecular docking).
The first (default) option is recommended for the exploration
of compounds that have closely related structures, while the
second approach (docking alignment) is a good compromise
for series that present problems with the CLACC alignment.
Anadvantageofthe last optionis thatthe interpretation of the
derived models can be easily expressed in the context of
receptor-ligand interactions, allowing it to retrieve structural
information on the binding site. We explored all of these
different settings for the generation of 3D-QSAR models, the

Article
Journal of Medicinal Chemistry, 2011, Vol. 54, No. 2 465
Figure 6. Molecular alignment of the 62 molecules that had the postulated docking pose. This alignment was used as an input for the 3D-
QSAR study. (A) side view of the receptor, and (B) view from the extracellular side. The following transmembrane helices (TM) are shown:
TM2 (cyan), TM3 (green), TM4 (yellow), TM6 (orange), and TM7 (red). The Connolly surface of the receptor is depicted in gray.
Figure 7. Important variables in the 3D-QSAR model C (represented for compound 11d). Green dots denote TIP fields, red dots O fields, blue
dots N1 fields, and yellow dots DRY fields. (Right) The same representation showing the binding site. The correspondence of the TIP fields
with the limiting pockets of the receptor for the L1 (Glu 7.29, Ile 7.35) and L3 (Ile 5.47, Trp 6.48), the superposition of the O-NI short distance
variable with Asn 6.55, and the DRY field with Phe 5.29 can be appreciated.
results being summarized in Table S1 of the Supporting
Information. The model generated on the basis of a docking
molecular alignment and the CLACC method for encoding
the descriptors (Model C in Table S1, Supporting Informa-
tion) was selected for further interpretation. This 3D-QSAR
model has two latent variables (LV) and presents a satisfac-
tory statistical quality, with a fitting parameter of r2=0.86 and
a predictive ability of q2=0.67, as obtained by the LOO cross-
validation test. The standard error for the correlation and the
prediction was 0.31 and 0.48 pKi log units, respectively.
The interpretation of the model highlights the key structural
features for high A3AR affinity. The most important variables,
that is, those with the highest PLS positive coefficients, were
used for the model interpretation and are depicted in Figure 7.
These variables represent, in an ideal case, structural features
that are present in the active compounds but absent in the
inactive compounds. In this respect, the following features are
important for the model interpretation:
(i) The hydrophobic interactions at the extracellular tip
of the binding site, mainly with residues Ile 6.58 and
Leu 7.35, are identified with the O-TIP (optimum
distance at 5.8 A) and the DRY-TIP (6.6 A) correlo-
grams. These two hydrophobic residues are a probable
source of A3 specificity: position 6.58 is occupied by a
threonine in the other subtypes, while position 7.35 has
already been related to interspecies selectivity in the
A1AR.63 The model suggests that an optimal shape
complementarity is achieved by molecules with smaller
L1 substituents (e.g., acetamides) or, alternatively,
molecules bearing larger L1 substituents but smaller
L2/L3 substituents (e.g., 13n). The combination of the
aforementioned descriptors provides information
about the interdependence of the size of L1 and L2/L3
substituents. This descriptor also accounts for the
lack of affinity observed in the aminopyrimidine series
(see Table 1), since these scaffolds do not bear any alkyl
residues on the exocyclic nitrogen.
(ii) The optimal pharmacophoric distance between the
H-bond acceptor probe, corresponding to the carbo-
nyl of the amide, and the shape of the L2 substituent is
located at 16.6 A in the N1-TIP cross-correlogram.
Whereas the role of the carbonyl group could be
hypothesized as a water-mediated interaction with
Glu 7.29 in the third extracellular loop (EL3), this
descriptor mainly identifies the importance of residue
Ile 5.47, interacting with the L2 substituent. Impor-
tantly, Ile 5.47 is occupied by the less bulky valine in
the other AR subtypes, a fact that could be taken into
account to improve A3 selectivity.
(iii) Finally, the O-N1 and N1-N1 autocorrelograms
account for the differences between the 2-amido- or
4-amidopyrimidine series, since these molecular de-
scriptors identify the distances between the exocy-
clic amide and the N1 in the ring. In the series of

466 Journal of Medicinal Chemistry, 2011, Vol. 54, No. 2
Yaziji et al.
Figure 8. Multiple pseudosequence alignment of human ARs, taking hA3AR residues within 4.5 A of all docked compounds into account.
Residue positions are denoted by the Ballesteros & Weinstein numbering,56 and shaded in gray according to their distance toward the docked
compounds.
4-amidopyrimidines, optimum distances of 9 and
10.6 A for O-N1 and N1-N1 autocorrelograms,
respectively, are observed. Conversely, in the series of
2-amidopyrimidines, a descriptor in the N1-N1 cor-
relogram, which is negatively related with affinity in
the model, identifies the particular location of the N1
of this scaffold closer to the carbonyl of the amide
(distance 4.6 A).
alignment of this pseudosequence for the human members of
the ARs family is shown in Figure 8, in which the variable
positions are clearly identified.
In this respect, it is remarkable that positions 5.42 and 6.52,
at the bottom edge of the binding site, are both occupied by a
serine in the hA3AR (see Figure 5); in the other three human
ARs these positions are occupied by Asn 5.42 and His 6.52,
respectively. The less voluminous sidechainofa serinein these
positions would allow the accommodation of bulkier L2
substituents in this subsite at the hA3AR, thus offering a
rationale for the observed receptor selectivity. There are also
remarkable differences within the ARs family regarding
residues at the top of the binding site: Ile 6.58, interacting
with L1, is replaced by a smaller valine in the other ARs.
hA3AR presents a valine at position 5.30 (in the tip of EL2),
which replaces a Glu that is conserved in the other three hAR
subtypes. Accordingtothe A2AARcrystallographic structure,
Glu 5.30 hydrogen bonds with the side chain of a His 7.29
in EL3, thus closing the top of the binding site while accept-
ing an additional hydrogen bond from the amino group of
4-(2-[7-amino-2-(2-furyl)[1,2,4]triazolo[2,3-a][1,3,5]triazin-5-
ylamino]ethyl)phenol.47 The amino derivatives would bene-
fit from this interaction, thus explaining the low selectivity
profile displayed by this group of compounds (Table 1).
Interestingly, a recent study of a new series of 2-phenylpyraz-
olo[4,3-d]pyrimidin-7-ones already indicated this difference in
the flap regions as being responsible for the A3AR selec-
tivitiy.49 Finally, Leu 7.35 provides a specific hydrophobic
subsite for the L1 substituent compared to the more volumi-
nous methionine present at this position in the A2AAR and
A2BAR or the polar Thr 7.35 in A1AR. Importantly, site directed
mutagenesis studies have identified this position as being
responsible for the interspecies differences in ligand affinities
in the A1ARs.62
To further challenge the computational model in terms of
robustnessand predictive capability, the tolerance of A3ARto
steric factors imposed by the alkyl residue of the amide
function(L1) was explored. Accordingly, six compounds were
designed which combined three new bulky residues [e.g.,
CH(Me)2, CH(Et)2, and Cy] on the exocyclic amide group
with the scaffold of the amines 4d and 4k. The compounds
were docked on the A3AR and queried to the QSAR model,
which predicted a good affinity profile for the A3AR (see
Table 3). Figure S2 in the Supporting Information shows how
the most bulky compounds (14c and 14f) optimally accom-
modate the cyclohexyl substituent in the hA3AR pocket. On
the other hand, a superposition with the crystallographic
structure of the hA2AAR shows that steric clashes with the
L1 substituent of this receptor might occur, as anticipated by the
pseudosequence analysis shown in Figure 8. The compounds
This last point is intriguing, since the docking model does
not identify any polar interaction for either the oxygen of the
amide group or the nitrogen at position N1 of the pyrimidine
(i.e., the nitrogen that varies in position between 2- and
4-amidopyrimidines). However, if we compare the binding
mode of the molecular series here reported on the A3AR with
the experimental binding mode of 4-(2-[7-amino-2-(2-furyl)-
[1,2,4]triazolo[2,3-a][1,3,5]triazin-5-ylamino]ethyl)phenol on
the A2AAR, it appears that the N1 in the 4-amidopyrimidines
series overlies with a nitrogen in the heterocycle of the
standard (N19 according to PDB nomenclature in entry
3EML).47 Recently, the group of Jacobson63 noted the im-
portance of a polar interaction of this N19 with crystal-
lographic water molecules. In order to check if similar
interactions could be achieved in our 4-amidopyrimidines/
hA3AR complexes, we performed a computational explora-
tion of structural water molecules in the binding site of the
hA3AR model, as detailed in the methods section. The results
(Figure S2, Supporting Information) show an energetically
favorable area for a water molecule that overlaps with the
N1-N1 descriptor, close to the position of the varying nitro-
gen in the 4-amidopytimidines (blue dots at the bottom of
Figure 7). A water-mediated interaction between N1 and Thr
7.42, which somehow resembles the H-bond network ZM24-
1386(N19)-HOH559-HOH550-His7.43(Nε) in the A2AAR, is
thus proposed as an specific polar contact for the 4-amidopyr-
imidine series, lacking in the 2-amidopyrimidines. The experi-
mental validation of such a different interaction behavior is
currently being investigated in our laboratory by synthesizing
additional series of unexplored heterocyclic scaffolds.
As stated above, an advantage of the 3D-QSAR methodology
employed in this study is to place the relevant descriptors in
the context of the binding site. This procedure can not only
deal with the structural requirements of the hA3AR for high
affinity but also enables a comparison of the hot-spots with
the relative positions in other ARs in order to explore the
reasons for selectivity. In an effort to identify these hot-spots,
we built a so-called “pseudosequence” based on the docking
results of this study. This pseudosequence is defined by all of
theresidues of thereceptorlocated at a amaximumdistanceof
4.5 A from the most exposed atom of the group of ligands
docked in the hA3AR (as superimposed in Figure 6). An

Article
Journal of Medicinal Chemistry, 2011, Vol. 54, No. 2 467
Table 4. Structure and Affinity Binding Data for 2,6-Diaryl-4-amidopyrimidines 14 at the Human Adenosine Receptors
Ki (nM) or % at 0.1 μM
pKi @ hA3
a
b
c
d
comp
Ar
R
hA1
hA2A
hA2B
hA3
exp
pred
14a
14b
14c
14d
14e
14f
4-OMe-Ph
CH(Me)2
CH(Et)2
Cy
5%
3%
5%
18%
1%
1%
1%
6%
1%
1%
2%
1%
3%
8%
1%
1%
2%
4%
12.2 ( 0.9
58.1 ( 4.7
32.1 ( 2.4
15.9 ( 1.3
52.4 ( 6.2
56.3 ( 6.1
7.91
7.24
7.49
7.8
8.16
7.82
7.15
7.58
7.57
6.85
2,4-OMe-Ph
CH(Me)2
CH(Et)2
Cy
7.28
6.59
a Displacement of specific [3H]DPCPX binding in human CHO cells expressed as Ki ( SEM in nM (n=3) or percentage displacement of specific
binding at a concentration of 0.1 μM (n=2). b Displacement of specific [3H]4-(2-[7-amino-2-(2-furyl)[1,2,4]triazolo[2,3-a][1,3,5]triazin-5-ylamino]-
ethyl)phenol binding in human HeLa cells expressed as Ki ( SEM in nM (n=3) or percentage displacement of specific binding at a concentration of
0.1 μM (n=2). c Displacement of specific [3H]DPCPX binding in human HEK-293 cells expressed as Ki ( SEM in nM (n = 3) or percentage displacement
of specific binding at a concentration of 0.1 μM (n=2). d Displacement of specific [3H]NECA binding in human HeLa cells expressed as Ki ( SEM in nM
(n=3) or percentage displacement of specific binding at a concentration of 0.1 μM (n=2).
were then prepared and tested at the four human adenosine
receptor subtypes (Table 4). As predicted by the computa-
tional exploration described above, the newer derivatives
exhibit potent and selective activity profiles, which unequi-
vocally confirms the tolerance of hA3AR to the size of the L1
substituent. Excellent agreement is found between predicted
and experimental affinity values for the six compounds
designed and tested in this part of the study [with an impressive
low standard error of the prediction (SDEP=0.37 log pKi
units)], which further confirms the predictive power of the
integrated computational model reported in this work, that is,
combining a molecular alignment from automated docking
with the prediction of activities on the basis of the 3D-QSAR
model (see Supporting Information, Figure S3C) was
observed. It is worth noting that on the solely basis of the avai-
lable literature39 data or the herein established SAR (Tables 1
and 2) the synthesis of compounds 14 would not have been
advisible. Thus, the modeling exploration enabled us to anti-
cipate attractive activity/selectivity profiles for compounds
incorporating hindered fragments at the amide chain (as
consequence of the higher tolerability for the L1 subsite of
the hA3AR).
SARs identified were substantiated by an exhaustive molec-
ular modeling study that combined a receptor-driven docking
model, which was constructed on the basis of the recently
published crystal structure of the hA2AAR, and a ligand-
based 3D-QSAR model, highlighting the key structural fea-
tures required for the optimal interaction with the hA3
receptor subtype in these compounds. The robustness and
predictive capabilities of the model were validated by design-
ing novel series of compounds that explore new alkyl residues
at the L1 subsite, which show high affinity and selectivity
profiles for the hA3AR. We must note that these compounds
would not have been synthesized solely on the basis of the
available SAR data on the literature39 or the qualitative SAR
established on this work (Tables 1-3). On the contrary, the
interest of these compounds was envisioned by the computa-
tional modeling exploration, suggesting that the hA3AR
shows higher tolerability for the L1 subsite. Further experi-
ments are currently in progress in our laboratories to prepare
new libaries incorporating nonidentical aryl groups at posi-
tions 2,6 and 4,6 obtained by adaptation of the herein
documented synthetic strategy according to recently pub-
lished methodologies.64-66 The biological profile of these
new derivatives will be published in due course.
Conclusions
Experimental Section
A new series of structurally simple and highly potent
ligands that exhibit remarkable selectivity profiles toward
the A3AR has been identified. A previous series of potent
and selective A1AR antagonists was selected, and the subse-
quent stepwise structural diversification of these model sub-
strates was carried out in order to radically modify the
activity/selectivity profiles while simultaneously providing
valuable structural information on the requirements for its
binding at the hA3 receptor subtype. Excellent affinity toward
the hA3AR (Kie6 nM) and optimal selectivity profiles (e10%
displacement of 0.1 μM concentrations at the other ARs) were
observed for compounds ISVY133, ISVY130, ISVY074, and
ISVY167, which incorporate 4-tolyl, 4-methoxyphenyl, and
2,4-dimethoxyphenyl moieties at the 2,6-positions of the
heterocyclic backbone. The antagonistic behavior of five
representative derivatives of these series was unequivocally
validated through functional cAMP experiments. The main
Chemistry. Commercially available starting materials, re-
agents, and solvents were purchased (Sigma-Aldrich) and used
without further purification. When necessary, solvents were
dried by standard techniques and distilled. After being extracted
from aqueous phases, the organic solvents were dried over
anhydrous sodium sulfate. The reactions were monitored by
thin-layer chromatography (TLC) with 2.5 mm Merck silica gel
GF 254 strips, and the purified compounds each showed a single
spot; unless stated otherwise, UV light and/or iodine vapor were
used for detection of compounds. The Suzuki cross-coupling
reactions were performed in coated Kimble vials on a PLS (6 ꢀ 4)
Organic Synthesizer with orbital stirring. Filtration and washing
protocols for supported reagents were performed in a 12-channel
vacuum manifold from Aldrich. Purity and identity of all tested
compounds were established by a combination of HPLC, elemental
analysis, mass spectrometry, and NMR spectra as described below.
Purification of isolated products was carried out by column
chromatography (Kieselgel 0.040-0.063 mm, E. Merck) or

468 Journal of Medicinal Chemistry, 2011, Vol. 54, No. 2
Yaziji et al.
Table 5. Conditions Used for Radioligand Binding Assays Using A1, A2A, A2B, and A3 Human Adenosine Receptors
A1
A2A
2A2B
A3
Buffer A
Buffer B
20 mM Hepes, 100 mM NaCl,
10 mM MgCl2, 2 units/mL
adenosine deaminase
(pH=7.4)
50 mM Tris-HCl,
50 mM Tris-HCl, 1 mM EDTA,
10 mM MgCl2, 0.1 mM
benzamidine, 2 units/mL
50 mM Tris-HCl, 1 mM
EDTA, 5 mM MgCl2,
2 units/mL adenosine
deaminase (pH=7.4)
50 mM Tris-HCl
(pH=7.4)
1 mM EDTA, 10 mM MgCl2,
2 units/mL adenosine
deaminase (pH=7.4)
50 mM Tris-HCl, 1 mM EDTA,
10 mM MgCl2 (pH=7.4)
GF/C
adenosine deaminase (pH=6.5)
50 mM Tris-HCl (pH= 6.5)
20 mM Hepes, 100 mM NaCl,
10 mM MgCl2, (pH=7.4)
GF/C
[3H]DPCPX nM
plate
GF/B
[3H]DPCPX35 nM
400 μM NECA
GF/B
[3H]NECA 30 nM
100 μM (R)-PIA
radioligand
nonspecific
binding
[3H]ZM2413853 nM
50 μM NECA
10 μM (R)-PIA
incubation
25 °C/60 min
25 °C/30 min
25 °C/30 min
25 °C/180 min
medium pressure liquid chromatography (MPLC) on a Combi-
Flash Companion (Teledyne ISCO) with RediSep prepacked
normal-phase silica gel (35-60 μm) columns followed by re-
crystallization. Melting points were determined on a Gallenkamp
melting point apparatus and are uncorrected. The NMR spectra
were recorded on Bruker AM300 and XM500 spectrometers.
Chemical shifts are given as δ values against tetramethylsilane
as internal standard and J values are given in Hz. Mass
spectra were obtained on a Varian MAT-711 instrument.
High resolution mass spectra were obtained on an Autospec
Micromass spectrometer. Analytical HPLC was performed
on a Agilent 1100 system using an Agilent Zorbax SB-Phenyl,
2.1 mm ꢀ 150 mm, 5 μm column with gradient elution using
the mobile phases (A) H2O containing 0.1% CF3COOH and
(B) MeCN and a flow rate of 1 mL/min. Elemental analyses were
performed on a Perkin-Elmer 240B apparatus at the Micro-
analysis Service of the University of Santiago de Compostela,
the elemental composition of the new compounds agreed to
within (0.4% of the calculated value. The purity of all tested
compounds was determined to be >95%. A detailed description
of synthetic methodologies as well as analytical and spectros-
copic data for all described compounds is included in the Sup-
porting Information.
The percentage of displacement of specific binding was calcu-
lated by the expression: % of displacement=((BT - dpm)*100)/
(BT - NSB); where BT is the total binding of the radioligand in
the assay, NSB is the nonspecific binding of the radioligand in
the assay, and dpm are the radioactive measurements obtained
by competing the radioligand binding with a given concentra-
tion of the test compound. Unless otherwise specified, results
shown in the text and tables are expressed as means ( SEM.
Significant differences between two means (p<0.05 or p<0.01)
were determined by one-way analysis of variance (ANOVA) and/or
by Student0s t test for nonpaired data.
Molecular Modeling
Model Building. A homology model of the hA3AR was
built using the recently crystallized hA2AAR as a template.
The modeling protocol is adapted from our participation in
the GPCR Dock 2008 competition.44 Briefly, a sequence
alignment between the two receptors with Clustal (PAM250
substitution matrix, with open and elongation gap penalties
of 10 and 0.05)67 was provided to Modeller v9.4.68 Fifteen
initial models were obtained using standard parameters. In a
first stage, the best five models were selected on the basis of
Procheck69 geometrical quality and DOPE scoring, and
these were subjected to geometrical improvement by the
Molprobity server.70 In a second stage, the best model from
the previous step was subjected to loop optimization with the
LoopModel routine in Modeler,71 again generating 15 re-
fined models. We selected the best model on the basis of a
compromise between Procheck stereochemical quality and
the DOPE energetical ranking. The geometry of the loops in
the selected model was refined by partial energy minimiza-
tion (i.e., nonloop residues were frozen) using the Polak-
Pharmacology. Radioligand binding competition assays were
performed in vitro as previously described53 using A1, A2A, A2B
,
and A3 human adenosine receptors expressed in transfected
CHO (A1AR), HeLa (A2AAR and A3AR), and HEK-293
(A2BAR) cells. The experimental conditions used are summar-
ized in Table 5. In each instance, aliquots of membranes (15 μg
for A1, 10 μg for A2AAR, 18 μg for A2BAR, and 90 μg for A3AR)
in buffer A (see Table 5) were incubated for the specified period
at 25 °C with the radioligand (2-35 nM) and six different
concentrations (ranging from 0.1 nM to 1 μM) of the test
molecule in a final volume of 200 μL. The binding reaction
was stopped by rapid filtration in a multiscreen manifold system
Ribiere algorithm (convergence criteria 0.05 kcal/mol A2)
ꢀ
(Milipore Iberica, Madrid, Spain). Unbound radioligand was
3
removed by washing four times with 250 μL of ice-cold buffer B
for A1 and A2A receptors, and six times with 250 μL of ice-cold
buffer B for A2BAR and A3AR (see Table 5). Nonspecific
binding was determined using a 50 or 400 μM NECA solution
for A2AAR and A2BAR and 10 or 100 μM R-PIA solution for
A1AR and A3AR, respectively. Radioactivity retained on filters
was determined by liquid scintillation counting using Universol
(ICN Biochemicals, Inc.). The binding affinities were deter-
mined using [3H]-DPCPX (130 Ci/mmol; GE-Healthcare,
Barcelona, Spain) as the radioligand for A1AR and A2BAR,
[3H]4-(2-[7-amino-2-(2-furyl)[1,2,4]triazolo[2,3-a][1,3,5]triazin-
5-ylamino]ethyl)phenol (21 Ci/mmol; Tocris, Madrid, Spain)
for A2AAR and [3H]-NECA (15.3 Ci/mmol; NEN-Perkin-Elmer
Life Sciences, Madrid, Spain) for A3AR.
and the OPLS-AA force-field as implemented in Macromodel.72
The general numbering scheme for GPCRs proposed by
Ballesteros and Weinstein56 was adopted through this work.
In essence, every residue is numbered as X.YY, where X
corresponds to the transmembrane helix (X=[1,7]) and YY
is a correlative number in the protein sequence, but taking as
a reference position (YY=50) the most conserved residue in
the given helix.
Protein-Ligand Docking. Automated docking explora-
tion was performed with GOLD version 3.2.73 Each ligand
was docked 20 times with default (high accuracy) genetic
algorithm (GA) search parameters, using the scoring func-
tion Chemscore as implemented in GOLD55 and allowing
full flexibility for the ligand, including flipping of amide
bonds. The search sphere was centered on the side chain
(CD1) of Ile 7.39, and expanded with a radius of 15 A, thus
ensuring a generous enough search space comprising the
The inhibition constant (Ki) of each compound was calcu-
lated by the expression: Ki=IC50 /(1 þ (C/KD)); where IC50 is the
concentration of compound that displaces the binding of radio-
ligand by 50%, C is the free concentration of radioligand, and
KD is the apparent dissociation constant of each radioligand.

Article
Journal of Medicinal Chemistry, 2011, Vol. 54, No. 2 469
antagonist binding site experimentally determined for ade-
nosine receptors.46 The criterion for the selection of docking
poses was based on Chemscore ranking and the population
of the solutions (according to a clustering criteria of 1 A).
Geometrical Optimization. Each docking pose was refined
by partial energy minimization of the binding site with
MOE.74 The site was selected as any atom within a distance
of 4.5 A around the ligand and OPLS-AA parameters were
used in combination with GBSA model for continuum
solvent representation. The convergence criterion for the
steepest descendent algorithm was set to 0.01 rmsd.
Pondal program (Xunta de Galicia, Spain). J.B. is researcher
of the Isabel Barreto program (Xunta de Galicia, Spain). D.R.
is recipient of a PFIS grant from the Instituto de Salud Carlos III
ꢀ
(Ministerio de Ciencia e Innovacion, Spain). We are grateful
to Dr. Manuel Pastor for assistance in the use of the Pentacle
software and helpful discussions.
Supporting Information Available: Experimental details for
the synthesis of compounds described, the spectroscopic, spec-
trometric, and elemental analysis data of all compounds pre-
pared as well as additional molecular modeling details. This
material is available free of charge via the Internet at http://
pubs.acs.org.
3D-QSAR. The conformations of the compounds ob-
tained in the molecular docking step were used to generate
a 3D-QSAR model with the software Pentacle v1.1.75 This
software allows the computation of the second generation of
GRid Independent Descriptors (GRIND-2). This family of
molecular descriptors, which are widely used in QSAR
studies, are generated in a three-step fashion: (i) computation
of molecular interaction fields (MIF) with different Grid
probes,76 (ii) selection of the most relevant MIF nodes, and
(iii) encoding of the descriptors as alignment-independent
vectors of node pairs, obtaining the so-called correlograms.61
The advantages of GRIND-2 include the use of AMANDA
as a new discretization algorithm for the identification of
“hot spots” (most relevant MIF nodes)50 and a new method
for encoding descriptors into alignment-free vectors called
CLACC.51 This method detects consistency in the computed
variables, ensuring that a given vector on the correlogram
corresponds to the description of same pharmacophoric
property within the series. In this work, the MIF were
computed using default values (i.e., GRID probes: DRY,
O, N1, TIP; 0.5 A grid step; dynamic parametrization),
discretization was carried out with default AMANDA para-
meters and the CLACC encoding of the variables was generated
on the basis of the docking alignment (“use CLACC for
alignment” = false) with strict options, meaning that any
variable that is not consistent in the series is removed
(“Remove non-consistent couples”=true). Two rounds of
fractional factorial design (FFD) were applied for the selec-
tion of the most relevant variables in the model. The model
generated was “saved for predictions” through the corre-
sponding menu option in Pentacle. Thereafter, the designed
molecules were docked in the A3 receptor as explained above,
and further imported in the Pentacle software, but using the
generated model as a template for the prediction of activities
on the A3 receptor.
References
(1) Fredholm, B. B.; Arslan, G.; Halldner, L.; Kull, B.; Shulte, G.;
Wasserman, W. Structure and function of adenosine receptors and
their genes. Naunyn-Schmiedeberg’s Arch. Pharmacol. 2000, 362,
364–374.
(2) Murphree, L. J.; Linden, J. Adenosine receptors. Encyclopaedia
Biol. Chem. 2004, 1, 34–39.
(3) Jacobson, K. A.; Knutsen, L. J. S. P1 and P2 purine and pyrimidine
receptor ligands. In Purinergic and Pyrimidine Signalling
(Handbook of Experimental Pharmacology); Abbrachio, M., Williams,
M., Eds.; Springer: Berlin, 2001; Vol. 151/1, pp 129-175.
(4) Fredholm, B. B.; Ijzerman, A. P.; Jacobson, K. A.; Koltz, K. N.;
Linden, J. International union of pharmacology XXV. Nomencla-
ture and classification of adenosine receptors. Pharmacol. Rev.
2001, 53, 527–552.
(5) Baraldi, P. G.; Aghadeh, M.; Tabrizi, M. A.; Gessi, S.; Borea, P. A.
Adenosine receptor antagonists: translating medicinal chemistry
and pharmacology into clinical utility. Chem. Rev. 2008, 108, 238–
263.
(6) Jacobson, K. A.; Gao, Z. Adenosine receptors as therapeutic
targets. Nat. Rev. Drug Discovery 2006, 5, 247–264.
(7) Moro, S.; Gao, Z. G.; Jacobson, K. A.; Spalluto, G. Progress in the
pursuit of therapeutic adenosine receptor antagonists. Med. Res.
Rev. 2006, 26, 131–159.
(8) Zhou, G. L.; Olah, M. E.; Johnson, R. A.; Stiles, G. L.; Clivelli, O.
Molecular cloning and characterization of an adenosine receptor:
the A3 adenosine receptor. Proc. Natl. Acad. Sci. U. S. A. 1992, 89,
7432–7436.
(9) Van Schaick, E. A.; Jacobson, K. A.; Kim, H. O.; IJzerman, A. P.;
Danhof, M. Hemodynamic effects and histamine release elicited by
the selective adenosine A3 receptor agonist Cl-IB-MECA in con-
scious rats. Eur. J. Pharmacol. 1996, 308, 311–314.
(10) Hannon, J. P.; Pfannkuche, H. J.; Fozard, J. R. A role for mast cells
in adenosine A3 receptor-mediated hypotension in the rat. Br. J.
Pharmacol. 1995, 115, 945–952.
(11) Fishman, P.; Bar-Yehuda, S. Pharmacology and therapeutic ap-
plications of A3 receptor subtypes. Curr. Top. Med. Chem. 2003, 3,
463–469.
(12) Muller, C. E. A3 Adenosine receptor antagonists. Mini Rev. Med
Chem. 2001, 1, 417–427.
(13) Baraldi, P. G.; Tabrizi, M. A.; Fruttarolo, F.; Borevo, A.; Avitabile,
B.; Preti, D.; Romagnoli, R.; Merighi, S.; Gessi, S.; Varani, K.;
Borea, P. A. Recent developments in the field of A3 adenosine
receptor antagonists. Drug Dev. Res. 2003, 58, 315–329.
(14) Baraldi, P. G.; Fruttarolo, F.; Tabrizi, M. A.; Romagnoli, R.; Preti,
D.; Carrion, M.; Iaconinoto, A.; Borea, P. A. Recent improve-
ments in the field of A3 adenosine receptor ligands. Exp. Opin.
Ther. Pat. 2005, 15, 1507–1519.
(15) Muller, C. E. Medicinal chemistry of adenosine A3 receptor
ligands. Curr. Top. Med. Chem. 2003, 3, 445–462.
(16) Moro, S.; Deflorian, F. Novel strategies for the design of new
potent A3 antagonists: an update. Curr. Med. Chem. 2006, 13, 639–
645.
(17) Baraldi, P. G. A3 ligands, history and perspectives. Med. Res. Rev.
2000, 20, 103–128.
(18) Van Muijlwijk-Koezen, J. E.; Timmerman, H.; Ijzerman, A. P. The
adenosine A3 receptor and its ligands. Prog. Med. Chem. 2001, 38,
61–113.
(19) Stone, T. W. Purines and neuroprotection. Adv. Exp. Med. Biol.
2002, 513, 249–280.
(20) Beaven, M. A.; Ramkumar, V.; Hydar, A. Adenosine A3 receptors
in mast cells. Trends Pharmacol. Sci. 1994, 15, 13–14.
Water Analysis. The prediction of energetically favorable
regions for structural water molecules in the binding site of
the hA3AR model was carried out with the program Grid,76
using the following parameters: OH2 probe, all program
directives on their default values except for LEAU=2 and
NPLA=2. Further refinement of the position of predicted
water molecules was done with the module FilMap, as
implemented in the Grid software.
Acknowledgment. This work has been developed in the
ꢀ
frame of the Red Gallega de Investigacion y Desarrollo de
Medicamentos (REGID) and was financially supported by
the Fondo Europeo de Desarrollo Social (FEDER) and the
Galician Government (Projects 09CSA016234PR and PS09/63).
E.S. is the recipient of a Consolidation Group Research
ꢀ
Grant from the Conselleria de Educacion (Xunta de Galicia).
E.S., H.G.d.T., and A.C. are researchers of the Isidro Parga

470 Journal of Medicinal Chemistry, 2011, Vol. 54, No. 2
Yaziji et al.
(21) Ramkumar, V.; Stiles, G. L.; Beaven, M. A.; Ali., H. The A3
adenosine receptor is the unique adenosine receptor which facil-
itates release of allergic mediators in mast cells. J. Biol. Chem. 1993,
268, 16887–16890.
(22) Akkari, R.; Burbiel, J. C.; Hockemeyer, J.; Mueller, C. E. Recent
progress in the development of adenosine receptor ligands as
antiinflammatory drugs. Curr. Top. Med. Chem. 2006, 6, 1375–
1399.
(23) Von Lubitz, D. K. J. E.; Carter, M. F.; Deutsch, S. I.; Lin, R. C. S.;
Mastropaola, J.; Meshulam, Y.; Jacobson, K. A. The effects of
adesnosine receptor stimulation on seizures in mice. Eur. J. Phar-
macol. 1995, 275, 23–29.
selective adenosine A1 receptor antagonists. J. Med. Chem. 2004,
47, 6529–6540.
(40) Cheong, S. L.; Dolzhenko, A.; Kachler, S.; Paoletta, S.; Federico, S.;
Cacciari, B.; Dolzhenko, A.; Koltz, K. N.; Moro, S.; Spalluto, G.;
Pastorin, G. The significance of 2-furyl ring substitution with a
2-(para-substituted) aryl group in a new series of pyrazolo-triazolo-
pyrimidines as potent and highly selective hA3 adenosine receptors
antagonists: new insights into structure-affinity relationship
and receptor-antagonists recognition. J. Med. Chem. 2010, 53,
3361–3375.
ꢀ
(41) Priego, E. M.; Perez-Perez, M. J.; von Frijtag, J. K.; de Vries, H.;
Ijzerman, A. P.; Camarasa, M. J.; Martın-Santamarıa, S. Selective
ꢀ
´
´
human adenosine A3 antagonists based on pyrido[2,1-f]purine-2,
4-diones: novel features of hA3 antagonist binding. ChemMedChem
2008, 3, 111–119.
(24) Nieber, K.; Lewerenz, A.; Hentschel, S.; Vissiennon, Z. Adenosine
A1 and A3 receptors: Neuroprotective targets with perspectives.
Bioforum 2002, 25, 237–240.
(42) Baraldi, P. G.; Tabrizi, M. A.; Preti, D.; Bovero, A.; Fruttarolo, F.;
Romagnoli, R.; Zaid, N. Z.; Moorman, A. R.; Varani, K.; Borea,
P. A. New 2-arylpyrazolo[4,3-c]quinoline derivatives as potent and
selective human A3 adenosine receptor antagonists. J. Med. Chem.
2005, 48, 5001–5008.
(25) Forsythe, P.; Ennis, M. Adenosine, mast cells and asthma. Inflam-
mation Res. 1999, 48, 301–307.
(26) Brown, R. A.; Spina, D.; Page, C. P. Adenosine receptors and
asthma. Br. J. Pharmacol. 2008, 153, S446–S456.
(27) Brambilla, R.; Cattabeni, F.; Ceruti, S.; Barbieri, D. Naunyn-
(43) Van Muijlwijk-Kozen, J. E.; Timmerman, H.; Vollinga, R. C.;
von Drabbe, J. F.; de Groote, M.; Visser, S.; Ijzerman, A. P.
Thiazole and thiadiazole analogues as a novel class of adenosine
receptor antagonists. J. Med. Chem. 2001, 44, 749–762.
(44) Michino, M.; Abola, E.; Brooks, C. L.; Dixon, J. S.; Moult, J.;
Stevens, R. C. Community-wide assessment of GPCR structure
modelling and ligand docking: GPCR Dock 2008. Nat. Rev. Drug
Discov. 2009, 8, 455–463.
Schmiedeberg’s Arch. Pharmacol. 2000, 361, 225–234.
(28) Jacobson, K. A.; Moro, S.; Kim, Y. C.; Li, A. A3 Adenosine
receptors: protective vs damaging effects identified using novel
agonists and antagonists. Drug Dev. Res. 1998, 45, 113–124.
(29) Abbracchio, M. P.; Ceruti, S.; Brambilla, R.; Barbiri, D.; Camurri,
A.; Franceschi, C.; Giammaroli, A. M.; Jacobson, K. A.; Cattabeni,
F.; Malorni, W. Adenosine A3 receptors and viability of astrocytes.
Drug Dev. Res. 1998, 45, 379–386.
(45) Bondavalli, F.; Botta, M.; Bruno, O.; Ciacci, A.; Corelli, F.; Fossa,
P.; Lucacchini, A.; Manetti, F.; Martini, C.; Menozzi, G.; Mosti,
L.; Ranise, A.; Schenone, S.; Tafi, A.; Trincavelli, M. L. Synthesis,
molecular modeling studies, and pharmacological activity of selec-
tive A1 receptor antagonists. J. Med. Chem. 2002, 45, 4875–4887.
(46) Moro, S.; Deflorian, F.; Spalluto, G.; Pastorin, G.; Cacciari,
B.; Kim, S.; Jacobson, K. A. Demystifying the three dimensional
structure of G protein-coupled receptors (GPCRs) with the aid of
molecular modelling. Chem. Commun. 2003, 24, 2949–2956.
(47) Jaakola, V.; Griffith, M. T.; Hanson, M. A.; Cherezov, V.;
Chien, E. Y. T.; Lane, J. R.; Ijzerman, A. P.; y Stevens, R. C.
The 2.6 angstrom crystal structure of a human A2A adenosine
receptor bound to an antagonist. Science 2008, 322, 1211–1217.
(48) Kolb, P.; Rosenbaum, D. M.; Irwin, J. J.; Fung, J. J.; Kobilka,
B. K.; Shoichet, B. K. Structure-based discovery of beta 2-adre-
nergic receptor ligands. Proc. Natl. Acad. Sci. U. S. A. 2009, 106,
6843–6848.
(30) Baraldi, P. G.; Tabrizi, M. A.; Romagnoli, R.; Fruttarolo, F.;
Merighi, S.; Varani, K.; Gessi, S.; Borea, P. A. Pyrazolo[4,3-e]1,2,4-
triazolo[1,5-c]pyrimidine ligands, new tools to characterize A3
adenosine receptors in human tumor cell lines. Curr. Med. Chem.
2005, 12, 1319–1329.
(31) Merighi, S.; Mirandola, P.; Varani, K.; Gessi, S.; Capitani, S.;
Leung, E.; Baraldi, P. G.; Tabrizi, M. A.; Borea, P. A. Pyrazolo-
triazolopyrimidine derivatives sensitize melanoma cells to the
chemotherapeutic drugs: taxol and vindesine. Biochem. Pharmacol.
2003, 66, 739–748.
(32) Cosimelli, B.; Greco, G.; Ehlardo, M.; Novellino, E.; Da Settimo,
F.; Taliani, S.; La Motta, C.; Bellandi, M.; Tuccinardi, T.; Martinelli,
A.; Ciampi, O. Derivatives of 4-amino-6-hydroxy-2-mercaptopyr-
imidine as novel, potent and selective A3 adenosine antagonists.
J. Med. Chem. 2008, 51, 1764–1770.
(33) Van Veldhoven, J. P. D.; Chang, L. C. W.; von Frijtag, J. K.;
Mulder-Krieger, T.; Struensee-Link, R.; Beukers, M. W.; Brussee,
J.; Ijzerman, A. P. A new generation of adenosine receptor antago-
nists: From di- to trisubstituted aminopyrimidines. Bioorg. Med.
Chem. 2008, 16, 2741–2752.
(34) Gillespie, R. J.; Bamford, S. J.; Clay, A.; Gaur, S.; Haymes, T.;
Jackson, P. S.; Jordan, A. M.; Klenke, B.; Leonardi, S.; Liu, J.;
Mansell, H. L.; Ng, S.; Saadi, M.; Simmonite, H.; Stratton, G. C.;
Todd, R. S.; Williamson, D. S.; Yule, I. A. Antagonists of the
human A2A receptor. Part 6: Further optimization of pyrimidine-
4-carboxamides. Bioorg. Med. Chem. 2009, 17, 6590–6605.
(35) Zhang, X.; Tellew, J. E.; Luo, Z.; Moorjani, M.; Lin, E.; Lanier,
M. C.; Chen, Y.; Williams, J. P.; Saunders, J.; Lechner, S. M.;
Markison, S.; Joswig, T.; Petroski, R.; Piercey, J.; Kargo, W.;
Malany, S.; Santos, M.; Gross, R. S.; Wen, J.; Jalali, K.; O0Brien,
Z.; Stotz, C. E.; Crespo, M. I.; Diaz, J. L.; Slee, D. H. Lead
optimization of 4-acetylamino-2-(3,5-dimethylpyrazol-1-yl)-6-pyr-
idylpyrimidines as A2A adenosine receptor antagonists for the
treatment of parkinson0s disease. J. Med. Chem. 2008, 51, 7099–
7110.
(49) Lenzi, O.; Colotta, V.; Catarzi, D.; Varano, F.; Poli, D.; Filacchioni,
G.; Varani, K.; Vincenzi, F.; Borea, P. A.; Paoletta, S.; Morizzo, E.;
Moro, S. 2-Phenylpyrazolo[4,3-d]pyrimidin-7-one as a new scaffold to
obtain potent and selective human A3 adenosine receptor antagonists:
new insights into the receptor-antagonist recognition. J. Med.
Chem. 2009, 52, 7640–7652.
ꢀ
(50) Duran, A.; Martınez, G. C.; Pastor, M. Development and valida-
´
tion of AMANDA, a new algorithm for selecting highly relevant
regions in Molecular Interaction Fields. J. Chem. Inf. Model. 2008,
48, 1813–1823.
(51) Duran, A. Ph.D. Thesis, University Pompeu Fabra, Barcelona,
2010. http://www.tesisenxarxa.net/TDX-0422110-094351
(52) Yaziji, V.; Coelho, A.; El Maatougui, A.; Brea, J.; Loza, M. I.;
Garcıa-Mera, X.; Sotelo, E. Divergent solution-phase diarylpyr-
´
imidine libraries as selective A3 adenosine receptor antagonists.
J. Comb. Chem. 2009, 11, 519–522 and references cited there.
(53) Bosch, M. P.; Campos, F.; Niubo, I.; Rosell, G.; Diaz, J. L.;
Brea, J.; Loza, M. I.; Guerrero, A. Synthesis and biological activity
of new potential agonists for the human adenosine A2A receptor.
J. Med. Chem. 2004, 47, 4041–4053.
(36) Vidal, B.; Nueda, A.; Esteve, C.; Domenech, T.; Benito, S.;
ꢀ
Reinoso, R. F.; Pont, M.; Calbet, M.; Lopez, R.; Cadavid, M. I.;
(54) Leff, P.; Dougall, I. G. Further concerns over Cheng-Prusoff
ꢀ
Loza, M. I.; Cardenas, A.; Godessart, N.; Beleta, J.; Warrellow, G.;
Ryder, H. Discovery and characterization of 40-2(2-furyl)-N-pyridin-
3-yl-4,50-bipyrimidin-20-amine (LAS38096), a potent, selective, and
efficacious A2B adenosine receptor antagonist. J. Med. Chem. 2007,
50, 2732–2736.
analysis. Trends Pharmacol. Sci. 1993, 14, 110–112.
(55) Verdonk, M. L.; Cole, J. C.; Hartshorn, M. J.; Murray, C. W.;
Taylor, R. D. Improved protein-ligand docking using GOLD.
Proteins 2003, 52, 609–623.
(56) Ballesteros, J. A.; Weinstein, H. Integrated methods for the con-
struction of three dimensional models and computational probing
of structure-function relations in G-protein coupled receptors.
Methods Neurosci. 1995, 25, 366–428.
(57) Jaakola, V. P.; Lane, J. R.; Lin, J. Y.; Katritch, V; Ijzerman, A. P.;
Stevens, R. C. Ligand binding and subtype selectivity of the
human A2A adenosine receptor: identification and characteriza-
tion of essential amino acid residues. J. Biol. Chem. 2010, 285,
13032–13044.
(37) Kuefner-Muehl, U., Scheuplein, S. W., Pohl, G., Gaida, W., Lehr,
E., Mierau, J., Meade, C. J. M. Triazines with an adenosine
antagonistic effect. PCT Int. Appl. WO 1999011633.
(38) Morizzo, E.; Capelli, F.; Lenzi, O.; Catarzi, D.; Varano, C.;
Filachioni, G.; Vicenzi, F.; Varani, K.; Borea, P. A.; Colotta, V.;
Moro, S. Scouting human A3 adenosine receptor antagonist bind-
ing mode using a molecular simplification approach: from triazo-
loquinoxaline to a pyrimidine skeleton as a key study. J. Med.
Chem. 2007, 50, 6596–6606.
(58) Gao, Z. G.; Chen, A.; Barak, D.; Kim, S. K.; Muller, C. E.;
Jacobson, K. A. Identification by site-directed mutagenesis
of residues involved in ligand recognition and activation of the
(39) Chang, L. C. W.; Spanjersberg, R. F.; von Frijtag, J. K.;
Mulder-Krieger, T.; van den Hout, G.; Beukers, M. W.; Brussee,
J.; Ijzerman, A. P. 2,4,6-Trisubstituted pyrimidines as a new class of

Article
Journal of Medicinal Chemistry, 2011, Vol. 54, No. 2 471
human A3 adenosine receptor. J. Biol. Chem. 2002, 277, 19056–
19063.
(59) Kim, J.; Wess, J.; van Rhee, A. M.; Schoneberg, T.; Jacobson, K. A.
Mutagenesis identifies residues involved in ligand recognition in
the human A2A adenosine receptor. J. Biol. Chem. 1995, 270,
13987–13997.
(60) Kim, S. K.; Gao, Z. G.; Rompaey, P. V.; Gross, A. S.; Chen, A.;
Calenbergh, S. V.; Jacobson, K. A. Comparison of the binding
domains of A2A agonists and antagonists. J. Med. Chem. 2003, 46,
4847–4859.
(61) Pastor, M.; Cruciani, G.; McLay, I.; Pickett, S.; Clementi, S. GRid-
INdependent descriptors (GRIND): a novel class of alignment-
independent three-dimensional molecular descriptors. J. Med.
Chem. 2000, 43, 3233–3243.
(62) Tucker, A. L.; Robeva, A. S.; Taylor, H. E.; Holeton, D.; Bockner,
M.; Lynch, K. R.; Linden, J. A1 adenosine receptors. Two amino
acids are responsible for species differences in ligand recognition.
J. Biol. Chem. 1994, 269, 27900–27906.
(63) Ivanov, A. A.; Barak, D.; Jacobson, K. A. Evaluation of homology
modeling of G-protein-coupled receptors in light of the A(2A)
adenosine receptor crystallographic structure. J. Med. Chem. 2009,
52, 3284–3292.
I. P.; Clayton, N. M.; Collins, S. D.; Corfield, J. A.; Hartley, C. D.;
Kleanthous, S.; Lambeth, P. F.; Lucas, F. S.; Mathews, N.; Naylor,
A.; Page, L. W.; Payne, J. J.; Pegg, N. A.; Price, H. S.; Skidmore, J.;
Stevens, A. J.; Stocker, R.; Stratton, S. C.; Stuart, A. J.; Wiseman, J. O.
Identification of [4-[4-(methylsulfonyl)phenyl]-6-(trifluoromethyl)-
2-pyrimidinyl] amines and ethers as potent and selective cyclo-
oxygenase-2 inhibitors. Bioorg. Med. Chem. Lett. 2009, 19, 4504–4508.
(67) Thompson, J. D.; Higgins, D. G.; Gibson, T. J. CLUSTAL W:
improving the sensitivity of progressive multiple sequence align-
ment through sequence weighting, position-specific gap penalties
and weight matrix choice. Nucleic Acid Res. 1994, 22, 4673–4680.
(68) Sali, A.; Blundell, T. L. Comparative protein modelling by satis-
faction of spatial restraints. J. Mol. Biol. 1993, 234, 779–815.
(69) Laskowski, R. A.; MacArthur, M. W.; Moss, D. S.; y Thornton,
J. M. PROCHECK: a program to check the stereochemical quality
of protein structures. J. Appl. Crystallogr. 1993, 26, 283–291.
(70) Davis, I. W.; Leaver-Fay, A.; Chen, V. B.; Block, J. N.; Kapral,
G. J.; Wang, X.; Murray, L. W.; Arendall, W. B.; Snoeyink, J.;
Richardson, J. S.; Richardson, D. C. MolProbity: all-atom con-
tacts and structure validation for proteins and nucleic acids.
Nucleic Acids Res. 2007, 35, W375–383.
€
(71) Fiser, A.; Do, R. K.; Sali, A. Modeling of loops in protein
(64) Benderitter, P.; Xavier de Araujo, J.; Schmitta, M.; Bourguignon, J. J.
2-Amino-6-iodo-4-tosyloxypyrimidine: a versatile key intermedi-
ate for regioselective functionalization of 2-aminopyrimidines in
4- and 6-positions. Tetrahedron 2007, 63, 12465–12470.
(65) Gong, B.; Hong, F.; Kohm, C.; Jenkins, S.; Tulinsky, J.; Bhatt, R.;
Vries, P.; Singer, J. W.; Klein, P. Synthesis, SAR, and antitumor
properties of diamino-C,Ndiarylpyrimidine positional isomers:
inhibitors of lysophosphatidic acid acyltransferase-β. Bioorg.
Med. Chem. Lett. 2004, 14, 2303–2308.
structures. Protein Sci. 2000, 9, 1753–1773.
€
(72) Macromodel, version 9.7; Schrodinger, LLC: New York, NY, 2009.
(73) Jones, G.; Willett, P.; Glen, R. C.; Leach, A. R.; Taylor, R.
Development and validation of a genetic algorithm for flexible
docking. J. Mol. Biol. 1997, 267, 727–748.
(74) MOE (Molecular Operative Environment), version 2009.10; Chemi-
cal Computing Group Inc. Montreal, Canada; 2009.
(75) Pentacle, version 1.1; Molecular Discovery, Ltd.: Perugia, Italy, 2009.
(76) Goodford, P. J. A computational procedure for determining
energetically favorable binding sites of biological important
macromolecules. J. Med. Chem. 1985, 28, 849–857.
(66) Swarbrick, M. E.; Beswick, P. J.; Gleave, R. J.; Green, R. H.;
Bingham, S.; Bountra, C.; Carter, M. C.; Chambers, L. J.; Chessell,