Synthesis and 3D QSAR of new pyrazolo[3,4-b]pyridines: potent and selective inhibitors of A1 adenosine receptors

Article abstract of DOI:10.1021/jm050407k

A number of 4-aminopyrazolo[3,4-b]pyridines 5-carboxylic acid esters (2-8) were synthesized and evaluated for their binding affinity at the A₁, A_2A, and A₃ adenosine receptors (AR), in bovine cortical membranes, as well as for their affinity toward human A₁AR (hA ₁AR). Some of the new compounds were characterized by a high affinity and selectivity toward the A₁ receptor subtype, showing a significant improvement in comparison with other pyrazolo-pyridines previously reported in the literature. In particular the methyl ester 2h as well as the isopropyl ester 5h, both of them bearing a p-methoxyphenylethylamino side chain at the position 4, presented K₁ values of 6 and 7 nM, respectively. To rationalize the relationships between structure and affinity of the novel compounds, a 3D QSAR model was also generated starting from compounds belonging to different classes of known A₁AR antagonists. 2005 American Chemical Society.

Full text of DOI:10.1021/jm050407k

7172
J. Med. Chem. 2005, 48, 7172-7185
Synthesis and 3D QSAR of New Pyrazolo[3,4-b]pyridines: Potent and Selective
Inhibitors of A1 Adenosine Receptors
Fabrizio Manetti,† Silvia Schenone,*,‡ Francesco Bondavalli,‡ Chiara Brullo,‡ Olga Bruno,‡ Angelo Ranise,‡
Luisa Mosti,‡ Giulia Menozzi,‡ Paola Fossa,‡ Maria Letizia Trincavelli,§ Claudia Martini,§ Adriano Martinelli,|
Cristina Tintori,† and Maurizio Botta†
Dipartimento Farmaco Chimico Tecnologico, Universita` degli Studi di Siena, Via Aldo Moro, I-53100 Siena, Italy,
Dipartimento di Scienze Farmaceutiche, Universita` degli Studi di Genova, Viale Benedetto XV n. 3, I-16132 Genova, Italy,
Dipartimento di Psichiatria, Neurobiologia, Farmacologia e Biotecnologie, Universita` degli Studi di Pisa,
Via Bonanno, I-56126, Italy, and Dipartimento di Scienze Farmaceutiche, Universita` di Pisa,
Via Bonanno 6, I-56126, Pisa, Italy
Received April 29, 2005
A number of 4-aminopyrazolo[3,4-b]pyridines 5-carboxylic acid esters (2-8) were synthesized
and evaluated for their binding affinity at the A1, A2A, and A3 adenosine receptors (AR), in
bovine cortical membranes, as well as for their affinity toward human A1AR (hA1AR). Some of
the new compounds were characterized by a high affinity and selectivity toward the A1 receptor
subtype, showing a significant improvement in comparison with other pyrazolo-pyridines
previously reported in the literature. In particular the methyl ester 2h as well as the isopropyl
ester 5h, both of them bearing a p-methoxyphenylethylamino side chain at the position 4,
presented Ki values of 6 and 7 nM, respectively. To rationalize the relationships between
structure and affinity of the novel compounds, a 3D QSAR model was also generated starting
from compounds belonging to different classes of known A1AR antagonists.
Introduction
as tracazolate, etazolate, and cartazolate, were origi-
nally identified as A1AR antagonists,13 during a wide
screening on nitrogenated heterocyclic molecules related
to purinic derivatives. Subsequently, a plethora of
additional pyrazolo[3,4-b]pyridines has been also syn-
thesized, among which the most active compound pos-
sess affinity of 0.3 µM for A1AR, but scarce selectivity.14
Finally, many other compounds with similar activity
have been patented.15
In this context, we have recently synthesized a series
of 4-amino-1-(2-chloro-2-phenylethyl)-1H-pyrazolo[3,4-
b]pyridine-5-carboxylic acid ethyl ester derivatives 1
(Chart 1),16,17 possessing an interesting antagonistic
profile of affinity and selectivity toward A1AR. Among
them, the most active and selective compound (1c, Table
3), bearing a phenylethylamino side chain at the posi-
tion 4, was characterized by a 50 nM affinity toward
bovine A1 receptors (bA1AR), while showed only a 4 and
34% inhibition of the specific radioligand binding toward
bA2A and bA3AR, respectively.16 Moreover, to get a
better understanding of the structure-activity relation-
ships (SAR) of the synthesized compounds and to find
more active and A1AR selective antagonists, we have
also built a pseudoreceptor model17 with the aim of
guiding the design of new compounds by predicting their
A1AR affinity. Unfortunately, biological evaluation of
several compounds suggested by the model did not give
the expected results, demonstrating that the model itself
was scarcely able to predict the activity of inhibitors
subjected to modifications at the position 5.18
Adenosine is an endogenous neuromodulator distrib-
uted in a wide variety of tissues, in both the periphery
and the central nervous system.1,2 This nucleoside
exerts physiological response by interacting with four
subtypes of receptors, named A1, A2A, A2B, and A3,3
belonging to the superfamily of G protein-coupled recep-
tors.4 The stimulation of adenosine receptors activates
several effector systems, such as the enzyme adenylyl
cyclase. Activation of A1 and A3AR leads to an inhibition
of adenylyl cyclase activity, while activation of A2A and
A
2BAR causes a stimulation of adenylyl cyclase.5 The
four AR subtypes are also coupled with other second
messenger systems, including calcium or potassium ion
channels, phospholipase A2 or C, and guanylate cyclase.6
The physiological significance and functions of adeno-
sine have been extensively studied. In the last two
decades, a large number of adenosine receptor ligands
(agonists and antagonists) have been developed.5,7
Particularly, many efforts have been invested in the
synthesis of A1AR antagonists8 that can stimulate
cerebral activity by blocking the adenosine central
inhibitory activity. A1AR antagonists have therapeutic
potential in the treatment of various form of dementia,
such as the Alzheimer’s disease,5,9 depression,10 and as
cognition enhancers in geriatric therapy.11 Moreover, A1-
AR antagonists are currently studied as potassium-
saving diuretics and for the treatment of acute renal
failure.12 4-Amino-substituted pyrazolo[3,4-b]pyridines,
* To whom correspondence should be addressed. Tel: +39 010
3538866, fax: +39 010 3538358, e-mail: schensil@unige.it.
† Universita` degli Studi di Siena.
As a consequence, a new and more accurate model
was required to better predict the activity of A1AR
inhibitors belonging to this family. To reach this goal,
we have planned to synthesize compounds 2-8 (Table
1), bearing various substituents as the terminal portion
‡ Universita` degli Studi di Genova.
§ Dipartimento di Psichiatria, Neurobiologia, Farmacologia e Bio-
tecnologie, Universita` degli Studi di Pisa.
| Dipartimento di Scienze Farmaceutiche, Universita` di Pisa.
10.1021/jm050407k CCC: $30.25 © 2005 American Chemical Society
Published on Web 10/15/2005

Inhibitors of A1 Adenosine Receptors
Journal of Medicinal Chemistry, 2005, Vol. 48, No. 23 7173
Table 1. Affinity of Compounds 2-8 toward A1, A2A, and A3 Adenosine Receptors
Ki (nM) or % inhibitiona
b
c
d
compd
R′
R
A1
251 ( 24
A2A
A3
2a
2b
2c
2d
2e
2f
2g
2h
3a
3b
3c
3d
3e
3f
3g
3h
4a
4b
4c
4d
4e
4f
CH3
NHcyclopropyl
52%
n.d.
n.d.
n.d.
n.d.
n.d.
0%
40%
41%
38%
49%
53%
n.d.
37%
n.d.
n.d.
46%
2%
CH3
NHC3H7
170 ( 12
7671 ( 680
0%
CH3
1-pyrrolidinyl
4241 ( 402
51%
CH3
4-morpholinyl
0%
CH3
NHCH2C6H5
141 ( 12
88 ( 7
7574 ( 632
1232 ( 116
CH3
NHCH2CH2C6H5
NHCH2CH2C6H4-pCH3
NHCH2CH2C6H4-pOCH3
NHCH2CH2C6H4-pCH3
NHCH2CH2C6H4-pOCH3
NHCH2CH2C6H4-oF
NHCH2CH2C6H4-mF
NHCH2CH2C6H4-pF
NHCH2CH2C6H4-oCl
NHCH2CH2C6H4-mCl
NHCH2CH2C6H4-pCl
NHcyclopropyl
CH3
15 ( 2 (148 ( 18)
6 ( 1 (94 ( 7)
22 ( 3
10 ( 1 (25 ( 4)
16 ( 1 (29 ( 3)
33 ( 3
12 ( 1 (16 ( 2)
76 ( 3
178 ( 11
19 ( 1 (153 ( 13)
156 ( 15
145 ( 13
2973 ( 265
2154 ( 201
1374 ( 110
130 ( 10
36 ( 7
26%
CH3
29%
C2H5
C2H5
C2H5
C2H5
C2H5
C2H5
C2H5
C2H5
C3H7
C3H7
C3H7
C3H7
C3H7
C3H7
41%
53%
36%
25%
30%
0%
21%
63%
1562 ( 123
1388 ( 129
1988 ( 176
31%
NHC3H7
n.d.
n.d.
n.d.
n.d.
n.d.
0%
1-pyrrolidinyl
4-morpholinyl
NHCH2C6H5
0%
NHCH2CH2C6H5
NHcyclopropyl
1358 ( 104
7503 ( 420
11%
5a
5b
5c
5d
5e
5f
CH(CH3)2
CH(CH3)2
CH(CH3)2
CH(CH3)2
CH(CH3)2
CH(CH3)2
CH(CH3)2
CH(CH3)2
CH(CH3)2
CH(CH3)2
CH(CH3)2
C4H9
NHC3H7
24 ( 4
0%
1-pyrrolidinyl
1078 ( 90
1822 ( 104
559 ( 32
31 ( 2
30%
16%
n.d.
n.d.
n.d.
0%
4-morpholinyl
NHCH2C6H5
0%
NHCH2CH2C6H5
NHCH2CH2C6H4-pCH3
NHCH2CH2C6H4-pOCH3
NHCH2CH2C6H4-oCl
NHCH2CH2C6H4-mCl
NHCH2CH2C6H4-pCl
NHcyclopropyl
1517 ( 124
26%
5g
5h
5i
48 ( 3
n.d.
52%
n.d.
n.d.
n.d.
0%
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
7 ( 1 (17 ( 1)
44 ( 3
34%
13%
5j
100 ( 7
41 ( 2
14%
5k
6a
6b
6c
6d
6e
7a
7b
7c
8a
8b
8c
27%
895 ( 69
959 ( 74
1893 ( 131
63%
11%
C4H9
NHC3H7
20%
C4H9
1-pyrrolidinyl
45%
C4H9
4-morpholinyl
16%
C4H9
NHCH2C6H5
27%
5%
CH(CH3)C2H5
CH(CH3)C2H5
CH(CH3)C2H5
CH2-cyclopropyl
CH2-cyclopropyl
CH2-cyclopropyl
NHcyclopropyl
197 ( 12
51 ( 5
33%
NHC3H7
40%
NHCH2CH2C6H5
NHC3H7
NHCH2CH2C6H5
NHCH2CH2C6H4-pCH3
62 ( 4
46%
109 ( 7
23%
48 ( 4
54%
63 ( 5
4%
a Ki values are means ( SEM of three separate assays, each performed in triplicate. b Displacement of specific [3H]DPCPX binding in
bovine cortical membranes or percentage of inhibition of specific binding at 10 µM concentration. In parentheses, affinity values toward
human A1 CHO transfected cells are also reported. c Displacement of specific [3H]CGS21680 binding in bovine striatal membranes or
percentage of inhibition of specific binding at 10 µM concentration. d Displacement of specific [125I]AB-MECA binding in human A3 CHO
transfected cell membranes or percentage of inhibition of specific binding at 10 µM.
Chart 1
the side chain at N1, while both the positions 4 and 5
were submitted to a variation of substituents. As a
result, in the present study, we describe the synthesis
and biological evaluation of a series of new 4-amino-
(substituted)-1H-pyrazolo[3,4-b]pyridine-5-carboxylic acid
esters (2-8), bearing the 2-chloro-2-phenylethyl chain
at N1. The new compounds, along with their affinity
data, were used to generate a highly predictive 3D
QSAR model for antagonists of A1AR.
Chemistry. The synthesis of the new compounds 2-8
is reported in Scheme 1. Acid hydrolysis (EtOH/H2SO4
at reflux for 24 h) of the 4-chloro-1-(2-chloro-2-phenyl-
ethyl)-1H-pyrazolo[3,4-b]pyridine-5-carboxylic acid ethyl
ester 9, prepared in six steps from 2-hydrazino-1-
phenylethanol and ethoxymethylene cyanoacetate ac-
cording to our published procedure,17 afforded the acid
of the ester side chain. Variation of the ester function
was envisaged as an interesting field of investigation
to enlarge the knowledge on the SAR concerning such
a substituent. Moreover, SAR data of our first genera-
tion A1AR inhibitors (compounds 1) evidenced the
importance of either the 2-chloro-2-phenylethyl group
at N1 or the cyclic and linear substituents at the
position 4, for the activity toward A1AR. Taking into
account all these suggestions, we have chosen to keep

7174 Journal of Medicinal Chemistry, 2005, Vol. 48, No. 23
Manetti et al.
Scheme 1a
a Reagents: (a) 3 M H2SO4, EtOH, reflux, 24 h; (b) R’OH, concd H2SO4, 80 °C, 18 h; (c) POCl3/DMF, CHCl3, reflux, 8 h; (d) (1) SOCl2,
CHCl3, reflux, 4 h, (2) cyclopropylmethanol, CHCl3, reflux, 6 h; (e) amines, toluene, rt, 2 days.
Table 2. Intrinsic Activity of 2h and 5h toward A1AR,
10, where the chlorine atom at the position 4 was
substituted by a hydroxy group. It is important to note
that a basic hydrolysis caused an undesired dehydroalo-
genation at the N1 side chain, leading to the corre-
sponding styryl derivative. The intermediate 10 was
transformed in good yield into the corresponding ester
derivatives 11a-e, through a Fisher esterification with
methyl, propyl, isopropyl, butyl, and sec-butyl alcohol,
in the presence of concentrated H2SO4 (80 °C, 18 h)
(Method A). On the other hand, the ester intermediate
13 was prepared by treating 10 with thionyl chloride
at reflux, and then, without isolation of the intermediate
acyl chloride, with cyclopropylmethanol in CHCl3 at
reflux for 6 h.
Treatment of 11a-e and 13 with the Vilsmeier
complex (POCl3:DMF 1:1) in CHCl3 at reflux for 8 h
(Method B) afforded the corresponding 4-chloro deriva-
tives 12a-e and 14, respectively, that were in turn
purified by chromatography with Florisil and CHCl3 as
the eluant (55-70% yield). Regioselective substitution
of the C4 chloride substituent of compounds 9, 12 and
14 with an excess of various amines (Method C) led to
the final compounds 2-8, in a very satisfactory yield
(65-90%).
Biology. Compounds were tested for their ability to
displace [3H]-8-cyclopentyl-1,3-dipropylxanthine, [3H]D-
PCPX, from A1AR, and [3H]-2-[[4-(2-carboxyethyl)phen-
ethyl]amino]-5-(N-ethylcarbamoyl)adenosine, [3H]CGS2-
1680, from A2AAR. The ability of the most active com-
pounds to displace [125I]-N-(3-iodo-4-aminobenzyl)-5-N-
methylcarboxamidoadenosine, [125I]AB-MECA, from A3-
AR was also evaluated in CHO cells transfected with
human A3AR.19 Moreover, the most A1AR selective com-
pounds were also tested for their affinity toward human
A1AR CHO transfected cells. Binding affinities toward
A1 (both bovine and human receptors), A2A, and A3AR,
Expressed as GTP Shifta
Ki (A1AR, nM)
compd
-GTP
+ GTP
GTP shift
R-PIA
2h
5h
4.2 ( 0.3
6.3 ( 0.3
7.2 ( 0.5
20 ( 1
7.3 ( 0.4
7.7 ( 0.3
4.7
1.1
1.0
a Displacement of [3H]DPCPX from bovine cortical membranes
in the absence and in the presence of 1 mM GTP. Values were
taken as mean ( SEM from three different experiments.
expressed as affinity constant values (Ki) or % inhibition
of specific radioligand binding, are reported in Table 1.
Moreover, selected compounds (namely, 2h and 5h)
were also profiled in functional assays for their intrinsic
activity at the A1AR (displacement of [3H]DPCPX from
bovine cortical membranes in the absence and in the
presence of 1 mM GTP) and results are reported in
Table 2. No significative GTP shift was evidenced,
suggesting that they elicited an antagonist profile. In
contrast, R-PIA, used as a standard agonist, exhibited
a larger GTP shift value of 4.7.
SAR Analysis. The new pyrazolo-pyridines 2-8
reported in this paper showed a very interesting A1AR
affinity profile that, together with affinity data already
reported for a series of pyrazolo-pyridine ethyl esters,16
allowed better clarification of the role of the amino and
the ester substituents at the position 4 and 5, respec-
tively, in the modulation of the affinity toward A1AR.
With the aim of investigating their influence in
defining affinity toward adenosine receptors, various
substituents (F, Cl, Me, OMe) have been introduced on
the phenyl ring of the phenylethylamino moiety at the
position 4 of the ethyl ester parent compound 1c (Table
3), leading to derivatives 3 (Table 1). With the exception
of compounds 3f and 3g, bearing an ortho and meta
chloro substituent, respectively, and found to be less

Inhibitors of A1 Adenosine Receptors
Journal of Medicinal Chemistry, 2005, Vol. 48, No. 23 7175
Table 3. Structures, Experimental, and Estimated/Predicted Affinity of Compounds 1-8, 15-80 toward A1AR
A1AR affinity (-log Ki)b
compda
R
R1
R2
exp
calcd
1a
1b
1c
1d
1ec
1fc
2a
2b
2c
2d
2ec
2f
4-morpholinyl
C2H5
C2H5
C2H5
C2H5
C2H5
C2H5
CH3
6.3
6.9
7.3
7.0
6.9
7.0
6.6
6.8
5.4
5.0
6.9
7.1
7.8
8.2
7.7
8.0
7.8
7.5
7.9
7.1
6.7
7.7
6.8
6.8
5.5
5.7
5.9
6.9
7.5
7.7
6.0
5.7
6.2
7.5
7.3
8.1
7.4
7.0
7.4
6.0
6.0
5.7
5.0
5.0
6.7
7.3
7.2
7.0
7.3
7.2
8.3
6.3
6.0
5.2
7.8
7.5
8.3
5.7
6.5
7.6
7.5
7.0
5.6
6.6
6.8
5.5
5.5
6.3
6.9
7.7
8.0
7.8
8.0
7.5
7.3
7.7
7.0
6.9
7.8
6.9
7.3
5.2
5.6
6.7
7.4
7.0
7.4
5.5
5.7
6.5
7.5
7.8
8.0
7.0
6.9
7.8
5.8
6.1
4.8
4.7
4.6
6.7
7.4
7.6
7.1
7.2
7.4
8.4
6.0
7.1
6.8
7.8
7.2
8.1
NHCH2C6H5
NHCH2CH2C6H5
NHC3H7
NHcyclopropyl
1-pyrrolidinyl
NHcyclopropyl
NHC3H7
CH3
1-pyrrolidinyl
CH3
4-morpholinyl
CH3
NHCH2C6H5
CH3
NHCH2CH2C6H5
NHCH2CH2C6H4-pCH3
NHCH2CH2C6H4-pOCH3
NHCH2CH2C6H4-pCH3
NHCH2CH2C6H4-pOCH3
NHCH2CH2C6H4-oF
NHCH2CH2C6H4-mF
NHCH2CH2C6H4-pF
NHCH2CH2C6H4-oCl
NHCH2CH2C6H4-mCl
NHCH2CH2C6H4-pCl
NHcyclopropyl
NHC3H7
CH3
2gc
2h
3ac
3b
3c
3d
3e
3f
CH3
CH3
C2H5
C2H5
C2H5
C2H5
C2H5
C2H5
C2H5
C2H5
C3H7
C3H7
C3H7
C3H7
C3H7
C3H7
3g
3h
4a
4bc
4cc
4d
4ec
4f
1-pyrrolidinyl
4-morpholinyl
NHCH2C6H5
NHCH2CH2C6H5
NHcyclopropyl
NHC3H7
5ac
5b
5cc
5d
5e
5f
CH(CH3)2
CH(CH3)2
CH(CH3)2
CH(CH3)2
CH(CH3)2
CH(CH3)2
CH(CH3)2
CH(CH3)2
CH(CH3)2
CH(CH3)2
CH(CH3)2
C4H9
1-pyrrolidinyl
4-morpholinyl
NHCH2C6H5
NHCH2CH2C6H5
NHCH2CH2C6H4-pCH3
NHCH2CH2C6H4-pOCH3
NHCH2CH2C6H4-oCl
NHCH2CH2C6H4-mCl
NHCH2CH2C6H4-pCl
NHcyclopropyl
NHC3H7
5g
5h
5i
5j
5k
6a
6b
6cc
6d
6e
7a
7b
7cc
8a
8b
8cc
15d
16
17
18
19
20
21d
C4H9
1-pyrrolidinyl
C4H9
4-morpholinyl
C4H9
NHCH2C6H5
C4H9
NHcyclopropyl
NHC3H7
CH3CHC2H5
CH3CHC2H5
CH3CHC2H5
CH2-cyclopropyl
CH2-cyclopropyl
CH2-cyclopropyl
Ph
NHCH2CH2C6H5
NHC3H7
NHCH2CH2C6H5
NHCH2CH2C6H4-pCH3
OH
CH3
CH3
CH3
CH3
CH3
CH3
CH3
Cl
Ph
OCH3
Ph
NH2
Ph
NHcyclohexyl
OH
Ph
p-FPh

7176 Journal of Medicinal Chemistry, 2005, Vol. 48, No. 23
Table 3. (Continued)
Manetti et al.
A1AR affinity (-log Ki)b
compda
R
R1
R2
exp
calcd
22d
23d
24c,d
25d
26d
27
OH
OH
OH
OH
OH
CH3
NHNH2
OH
OH
Ph
Ph
Ph
Ph
Ph
Br
9.1
9.8
7.6
8.3
8.8
5.9
7.0
9.2
7.1
7.3
6.5
8.2
7.6
5.7
5.1
5.0
5.0
7.3
6.7
5.3
6.5
6.1
7.8
8.0
8.8
9.4
7.7
7.7
7.5
8.4
8.3
8.9
9.0
6.5
7.3
8.3
7.1
7.0
7.3
7.1
7.2
7.7
6.6
5.0
5.7
6.1
7.1
6.5
6.0
5.4
5.0
6.6
6.6
6.3
5.7
6.0
5.7
5.2
5.0
8.5
8.2
7.8
8.2
8.1
5.1
7.1
9.3
7.2
7.2
6.9
8.3
7.9
5.9
5.9
4.9
5.0
6.4
5.8
5.5
7.1
7.4
7.9
7.5
8.5
9.1
7.0
7.4
7.5
8.5
8.9
9.1
8.4
7.0
7.5
8.2
7.9
6.4
8.0
6.9
6.8
7.8
6.2
6.5
6.5
6.2
6.9
7.1
6.3
5.1
5.1
6.3
6.7
6.7
5.9
5.9
5.6
5.6
5.1
Cl
OPh
OEt
OCH3
28c
29d
30d
31
Ph
Ph
Ph
Ph
Ph
p-NH2Ph
m-NH2Ph
CH2Ph
H
CH3
CH3
CH2CH2CH3
CH2CH2CH3
CH3
N(CH3)2
NHcyclohexyl
Br
OEt
OEt
OH
32
OEt
33c,d
34c,d
35d
36d
37d
38d
39d
40c,d
41
CH3
OH
CH3
OH
CH3
OH
CH3
OH
OH
OH
NH2
OH
CH3
OH
NH2
H
42d
43c
44d
45
OH
Ph
NHCOCH3
CH3
H
OPh
OH
Ph
Ph
H
H
46
H
cyclohexyl
cyclopentyl
H
47
H
48
p-CH3
m-CH3
m-F
m-CH3
m-F
m-CH3
m-F
p-OCH3
49c
50
H
H
51
cyclohexyl
cyclohexyl
cyclopentyl
cyclopentyl
H
CH2Ph
CH2CH2Ph
cyclohexyl
COPh
CONHPh
Ph
Ph
Ph
52c
53c
54c
55
56
57
58
p-Cl
59
p-Cl
60c
61
p-Cl
H
62
CH3
63c
64
Ph
H
fur-2-yl
p-Cl-Ph
p-OCH3
fur-2-yl
thien-2-yl
fur-2-yl
fur-2-yl
CH3
65c
66
CH3
CH3
67c
68
CH3
CH3
69
Ph
70
CH2Ph
71
H
CH3
72c
73
CH3
CH2Ph
CH2Ph
p-OH-Ph
m,p-OH
m,p-OCH3
74c
75
H
H
76
H
77
NHC3H7
NHC4H9
NHcyclohexyl
NHCH2CH2Ph
78c
79
80
a Compounds 1, 15-44, 45-60, 61-76, and 77-80 are taken from ref 16, 30, 31, 19, and 18, respectively. b Estimated values (expressed
as -log Ki) are for compounds belonging to the training set, predicted values are for compounds of the test set (expressed as -log Ki).
c Test set compounds. All the remaining molecules belong to the training set. d Such compounds were modeled in their quinoid form (i.e.,
with a quinoid oxygen atom at the position 4 and a protonated nitrogen at the position 1 of the cyclic core), according to ref 30.
active than the corresponding unsubstituted parent
compound (76 and 178 nM, respectively, versus 50 nM),
all the remaining compounds of this subclass showed
improved activity (ranging from 10 to 33 nM). In detail,
the best active compounds of this series (3b and 3e)
showed the para position substituted with a methoxy
or a fluoride group, leading to an activity of 10 and 12
nM, respectively. Moreover, it is also important to note
that the fluoro substitution at the various positions on
the phenyl ring was more profitable for affinity with
respect to the corresponding chloro group (3c versus 3f,
3d versus 3g, and 3e versus 3h).
Affinity values of compounds 2 were found compar-
able to those of both the corresponding ethyl analogues
3 and the derivatives previously published by our
research group.16 Moreover, 2h, with a Ki value of 6 nM,

Inhibitors of A1 Adenosine Receptors
Journal of Medicinal Chemistry, 2005, Vol. 48, No. 23 7177
was the most active pyrazolo-pyridine derivative re-
ported in this paper.
A further lengthening of the ester alkyl chain to a
propyl and a butyl group led to compounds 4 and 6,
respectively, with affinity values in the micromolar
range (spanning from 130 to 2973 nM), suggesting that
the best linear alkyl group for the ester moiety was the
ethyl or the methyl substituent. Very interestingly,
when the alkyl ester was branched to an isopropyl
group, such as in compounds 5, good affinity values were
found for particular substituents at the position 4 of the
core structure. In particular, small aliphatic or cy-
cloaliphatic groups such as the propylamino (5b) and
cyclopropylamino (5a) moiety gave affinity of 24 and 36
nM, respectively, while larger and cyclic amines, such
as 1-pyrrolidinyl (5c) and 4-morpholinyl (5d), were
associated with very lower affinity (1078 and 1822 nM,
respectively). Similarly, affinity for the benzylamino
derivative 5e was found in the low micromolar range
(559 nM), while the phenylethylamino compound 5f
showed an affinity of 31 nM, comparable or better with
respect to all the other phenylethylamino derivatives
(compare with the affinity of 2f, 88 nM; 1c, 50 nM; 4f,
130 nM; 7c, 62 nM; and 8b, 48 nM). Introduction of
substituents on the phenyl ring of the phenylethylamino
moiety led in general to similar (5g, 5i, and 5k) or
decreased (5j) affinity, with the exception of compound
5h, showing a very high affinity (7 nM), comparable
with that of the corresponding methyl and ethyl ester
derivatives 2h and 3b, respectively.
Finally, increasing the size of the isopropyl chain of
compounds 5 to both a sec-butyl and a cyclopropylmethyl
substituent (compounds 7 and 8, respectively), a drop
in affinity was observed.
In summary, the optimal combination of the amino
substituent and the ester group at the positions 4 and
5, respectively, seems to be a p-methoxyphenylethyl-
amino chain together with a short aliphatic group, such
as a methyl, ethyl, and isopropyl moiety.
Compounds with higher affinity toward bA1AR were
also tested for their affinity toward hA1AR. Affinity of
3b, 3c, 3e, and 5h was found to be comparable with
that toward bA1AR (25 vs 10, 29 vs 16, 16 vs 12, 17 vs
7 nM, respectively, Table 1). On the other hand, the
remaining compounds tested showed affinity values
toward the human receptor from 8-fold (3h) to about
10- (2g) and 16-fold (2h) lower than those found toward
the bovine receptor (153 vs 19 nM, 148 vs 15, and 94 vs
6 nM, respectively).
Figure 1. PCA score plot (X-axis and Y-axis report the first
and the second components, respectively) for all 116 ligands.
The uniform distribution of test set compounds (red dots)
within clusters of training set compounds (black dots) suggests
that selection between training set and test set was performed
appropriately.
well as with molecules collected from the literature21
(Table 3). It is important to note that within the huge
amount of A1AR inhibitors found in the literature, to
ensure the highest homogeneity of biological data with
respect to those of the new pyrazolo-pyridine deriva-
tives, we have focused our attention on such compounds
whose affinity toward A1AR was obtained following the
same experimental protocol applied for the new com-
pounds. Moreover, derivatives found to be inactive (Ki
> 10000 nM, Table 3) were also included in the original
set of compounds, assigning them an arbitrary Ki value
of 10000 nM, corresponding to a -log Ki ) 5.0.22
Compounds of the original set were divided into a
training set and a test set, according to the usual
guidelines. In detail, either the training or the test set
should contain compounds in such a way to maximize
their structural diversity (that is, compounds belonging
to the training set and to the test set should be
representative of the molecular diversity of all the
compounds under study) and uniformly span over the
whole range of activity. As a result, the training set was
constituted by 86 compounds, while an external valida-
tion set (test set) of 30 molecules was selected to test
the predictive power of the model.
Moreover, with the purpose of checking the accuracy
in the choice of training and test set compounds,
following a protocol reported in the literature,23 a PCA
was performed by means of the program GOLPE,24
using GRID interaction fields25 as descriptors (corre-
sponding to the interaction energies between appropri-
ate chemical probes and all the compounds under study,
see below). As a result, test set compounds (red dots,
Figure 1) were uniformly distributed within clusters of
training set compounds (black dots), suggesting that
selection between training set and test set was per-
formed appropriately.
Training set compounds are characterized by affinity
values spanning about 5 orders of magnitude, the
minimum value of 5.0 (expressed as -log Ki) being
associated with compounds 37, 38, 80, 2d, 6d, and 6e
and the maximum value of 9.8 being associated with
compound 23. Similarly, compounds belonging to the
test set showed affinity value ranging from 5.0 (com-
pounds 65 and 72) to 9.0 (compound 54), spanning over
The new compounds showed very low or no affinity
toward both A2A and A3AR (Table 1), thus leading to a
high selectivity toward A1AR. In fact, Ki values toward
A2AAR were in the micromolar range for a few com-
pounds, while for the remaining derivatives, only low
percentages of inhibition of specific CGS21680 binding
at 10 µM concentration, were found. Similarly, all the
compounds tested for their affinity toward A3AR showed
very low percentages of inhibition of specific AB-MECA
binding at 10 µM concentration.
3D QSAR Studies
Dataset Selection. A set of 116 compounds was
assembled grouping the new derivatives together with
additional compounds previously reported by us,20 as

7178 Journal of Medicinal Chemistry, 2005, Vol. 48, No. 23
Manetti et al.
Figure 2. (A) Graphical representation of the alignment of compounds 2h (green), 23 (orange), 63 (yellow), 47 (cyan), and DPCPX
(purple), as obtained by application of the field fit alignment method of Cerius2. Pharmacophoric features are labeled, while
heteroatoms and hydrogens bound to hydrogen bond donor groups are color coded following the atom-type notation. (B) A simplified
view of the alignment where pharmacophoric elements have been labeled and color coded: hydrophobic regions, HY1-3, blue;
hydrogen bond acceptor groups, HBA1-2, red; hydrogen bond donor groups, HBD, pink.
4 orders of magnitude. Training set compounds, along
with their biological activity, were used to generate a
3D QSAR model for A1AR antagonists.
employing a rigid field fitting and using C (a generic
carbon atom) as the probe. A fit similarity score was
then calculated for all conformations and the conformer
of each representative molecule that showed the higher
fit was selected and used as the “active conformation”
for its chemical family. In a second step, the active
conformation of each structural class was used as the
template structure to align all the remaining molecules
of the same class. For these molecules, the conformer
with the highest fit to the “active conformation” of the
target model was selected and used in the next calcula-
tions. As a result, six common features were found
(Figure 2), corresponding to three hydrophobic molec-
ular portions (HY1-3), two hydrogen bond acceptors
(HBA1-2), and a hydrogen bond donor group (HBD). In
further detail, HY1 was filled by the chlorophenyl side
chain of 2h, the chloro substituent and part of the
condensed pyridine ring of 23, the phenyl ring at
position 2 of 47, and the phenyl ring at position 3 of 63.
Moreover, the methyl group of the ester function of 2h,
the C3 portion of 23, and the fused benzene rings of both
63 and 47 were embedded into HY2. Finally, the aryl
side chain at position 4 of 2h, the phenyl ring at position
2 of 23, the phenyl ring at position 10 of 63, and the
cyclopentyl ring of 47 were located within HY3. With
the exception of 63, the remaining three compounds
showed a structural motif constituted by a hydrogen
bond acceptor moiety (HBA1) close to a HBD. In
particular, they were represented by N2 and the amino
group of the C4 side chain of 2h, N8 and N1(H) of 23,
and N3 and the amino group of the C4 side chain of 47,
respectively. Compound 63, showing both N1 and N2
as possible hydrogen bond acceptors, lacked the HBD
feature, thus accounting for its lowest binding affinity
among the four compounds. The carbonyl group present
on each of the four structures represented the remaining
pharmacophoric feature HBA2.
In the first step, on the basis of the fact that the active
conformation of the selected compounds is not known
yet, we resorted to a molecular modeling approach to
build the conformational populations to be used for the
generation of the molecular alignment. A local minimum
energy conformer of each compound was built using the
2D-3D sketcher implemented in the software Catalyst,26
while the conformational space of all antagonists was
then explored by a conformational search by means of
the “best” procedure, keeping all conformations within
5 kcal/mol from the global minimum and specifying 250
as the maximum number of conformers.
Field Fit Alignment. It is well-known that one of
the most important steps of a 3D QSAR study is
represented by the alignment of the molecules under
investigation. The alignment of all the molecules se-
lected for this study was performed applying a protocol
that combines the active analogue approach together
with a field fit alignment method. The active analogue
approach is usually used to search the bioactive con-
formers of the characteristic compounds of each family,27
on the basis of the assumption that ligands able to bind
the same site share the same three-dimensional ar-
rangement of structural features essential for recogni-
tion to the receptor (the so-called pharmacophore pat-
tern). The conformers found through this approach are
in turn used as templates to fit the remaining molecules.
For this purpose, the Align Molecules/Drug Discovery
module in Cerius228 was used to perform the alignment
of all the selected compounds. In detail, the low energy
conformer of 23, the most active compound among all
the molecules under investigation, also characterized by
a low conformational flexibility, was selected as the
template structure. The alignment strategy was con-
stituted by two steps. First, the most active compound
of each chemical series, corresponding to 2h, 47, 63, and
77, was collected. All the conformers of each of the most
active compounds was aligned to the target model
The alignment proposed was in perfect agreement
with a model of the bA1AR recently proposed to ratio-
nalize SAR of a set of aryl-triazino-benzimidazoles.19 In
fact, HY1-3, HBA1-2, and HBD of the present model

Inhibitors of A1 Adenosine Receptors
Journal of Medicinal Chemistry, 2005, Vol. 48, No. 23 7179
corresponded to L2, L3, L1, HB2, HB3, HB1, respec-
tively, of the model reported by Da Settimo and co-
workers. Moreover, the alignment was also fully con-
sistent with a previous pharmacophoric model built by
us on the basis of a different collection of bA1AR
antagonists.17
Probe Selection. The ligands, aligned each other,
were imported into the GRID software. Interaction
energies between selected probes and each molecule
were calculated using a grid spacing of 1 Å, while the
size of the grid was set as large as to accommodate all
the aligned ligands in all directions (along X, Y, and Z
axes). Three probes were chosen to describe all the
molecules. The C3 probe, corresponding to a methyl
group, was used to account for steric contacts. The O
probe (a sp2 carbonylic oxygen) and the N2) probe (sp2
amine dNH2+ cation) allowed to evaluate the ability of
each compound to be a hydrogen bond donor and a
hydrogen bond acceptor, respectively, as well as to
calculate electrostatic potential fields on each ligand
structure.
Variable Selection. Interaction energy values at
each grid point were imported into the GOLPE program
and used as independent variables to build a 3D QSAR
model. However, it is well-known that many of the
variables derived from the GRID analysis do not con-
tribute to the correlation between chemical structure
and biological activity, but they could be considered as
noise, which decreases the quality of the model. As a
consequence, to obtain a robust QSAR model, the no-
significative variables were removed using the GOLPE
program. From the 56376 active variables that were
selected by GOLPE, an initial pretreatment performed
by means of the advanced pretreatment tool (absolute
values lower than 0.02 kcal/mol were set to zero and
variables that exhibited only two values were removed)
reduced the number of variables to 32690. Moreover,
the D optimal preselection procedure was applied to
obtain the most informative variables correlated with
the biological activity. This procedure was repeated two
times, using 50% as the reduction level and five
components. Each selection of variables was preceded
by calculation of a PLS model, on the basis of only the
last selected variables. After D optimal variable prese-
lection, the number of variables was further reduced to
8164.
Figure 3. Experimental activity versus estimated/predicted
activity in the final 3D QSAR model.
conclude that many of the original variables generated
noise, instead of contributing to the robustness of the
model.
Moreover, the predictive power of the model was
further assessed by calculating affinity values for the
test set compounds. As a result, a good correlation was
found (0.63 as standard deviation of error of predictions,
SDEP) between experimental and predicted affinity
values of the external set of compounds (Figure 3 and
Table 3).29
Interpretation of Contour Maps. One important
feature of 3D QSAR analysis is the graphical represen-
tation of the model, usually aimed at making its
interpretation easier. GOLPE provides several options
to display the final model. Among these, PLS pseudo-
coefficients are very useful because they allow the
visualization of favorable and unfavorable interactions
between the probes and the molecules under study. For
this reason, GRID plots of PLS coefficients were ana-
lyzed by means of GOLPE with the purpose of getting
a better understanding of the structure-activity rela-
tionships.
C3 Contour Map. Contour maps for the C3 probe
indicated sterically unfavored and favored regions of
space (cyan and yellow, respectively, Figure 4). In
particular, regions with higher positive values (A, B, and
C, Figure 4, yellow) corresponded to a profitable inter-
action between a molecular substituent and the methyl
probe, contributing to ameliorate the affinity. On the
contrary, the principal regions with negative values (D,
E, F, G, and H, Figure 4, cyan) corresponded to an
unfavorable interaction between a substituent of the
molecule and the methyl probe, contributing to decrease
the affinity. Figure 4a shows compound 2h (character-
ized by the best Ki value, 6 nM) embedded into the
contour map obtained with the C3 probe. The p-
methoxyphenylethylamino group at the position 4, as
well as the methyl group of the ester at the position 5,
and the chloro substituent on the side chain at the
position 1, being closed to the yellow regions of the map
(C, B, and A, respectively), are suggested to have a steric
contribution very profitable for affinity toward A1AR.
On the contrary, cyan volumes represented sterically
not permitted regions. In fact, cyclic amino substituents
such as pyrrolidino and morpholino moieties, or shorter
side chains such as the benzyl group of compounds 2-8
showed steric hindrance, interacting with the region E
A Smart Region Definition (SRD) algorithm was
applied with the aim of selecting and grouping the
regions of variables of highest importance for the model.
The groups were then used in the Fractional Factorial
Design (FFD) procedure replacing the original variables.
FFD selection was applied three times, until no further
improvement in the q2 value was observed. As a result,
the model obtained from the reduced set of variables
(1247) was characterized by a significant enhancement
of the predictive power, with respect to the parent model
generated on the whole set of variables. In fact, the
validation procedure (a cross-validation routine with five
random sets of compounds) showed an improved inter-
nal predictive ability (q2 ) 0.70) for the final model with
five components, in comparison to the model based on
all the variables (q2 ) 0.43). This result allowed us to

7180 Journal of Medicinal Chemistry, 2005, Vol. 48, No. 23
Manetti et al.
Figure 4. PLS coefficient plot obtained with the C3 probe. Compounds 2h (a), 23 (b), 63 (c), and 47 (d) are embedded into
positive (yellow) and negative (cyan) regions. Favorable interactions (i.e., an increase of -log Ki) between a substituent and the
probe occur in regions with positive coefficients (0.0003 kcal/mol level, positive regions, yellow), while unfavorable interactions
(i.e., a decrease of -log Ki) between a substituent and the probe occur in regions with negative coefficients (-0.0010 kcal/mol
level, negative regions, cyan).
of the model. Region G shows that a substitution at the
position 2 of the pyrazolo-pyridine nucleus is not
permitted, while region H indicates that a substitution
at the meta position of the phenylethyl side chain should
result in a lack of affinity. The D region suggests that
bulky substituents on the ester group, as butyl and sec-
butyl chains could give a bad steric interaction. Figure
4b shows 23 characterized by favorable steric interac-
tions with both regions A and C. Compound 63 (Figure
4c) has good contacts with regions A, B, and C, while
47 shows profitable steric contacts with regions A and
C (Figure 4d).
N2) Contour Maps. Coefficient plots generated with
the N2) probe mimic the hydrogen bond donor interac-
tions. It is interesting to note that some positive and
negative maps are positioned in the same portion of
space also found for the C3 probe, indicating that the
major effect of the N2) probe in these regions could be
considered as of steric nature. In other words, in such
regions, contribution of the N2) probe to the hydrogen
bond contacts is low. Compound 2h shows profitable
interactions (region A and C, Figure 5a) involving both
the ester group at the position 5 and the N2 atom as
hydrogen bond acceptor moieties. This result, combined
with that found for the C3 probe, clearly shows that the
ester function of the new pyrazolo-pyridine derivatives
is a crucial structural element in defining affinity of
such compounds toward A1-AR. Compound 23 (Figure
5b) interacted favorably through its carbonyl moiety
with both regions A and B, while the N8 contacted
region C of the N2) map. Moreover, the carbonyl group
at the position 4 of 63 is involved in a profitable
interaction with the region B, while both N1 and N2
make hydrogen bond contacts with the region C of the
N2) map (Figure 5c). In a similar way, Figure 5d shows
good interactions between N3 of 47 and the region C of
the map, while the carbonyl group and N5 have only a
weak interaction with regions B and A, respectively.
O Contour Map. The contribution of the O probe to
the PLS model represents both the ability to accept
hydrogen bond contacts and make steric interactions.
In fact, similarly to that found for the N2) probe, some
portions of the O probe maps occupy similar regions of
space previously identified with the C3 probe, suggest-
ing that also the O probe is involved in steric contacts.
Moreover, two regions (A and B, Figure 6) show profit-
able interactions with the hydrogen bond donor moieties
of the molecules under investigation. In detail, the
region A makes favorable interactions with the amino
group at the position 4 of both 2h and 47 (Figure 6a

Inhibitors of A1 Adenosine Receptors
Journal of Medicinal Chemistry, 2005, Vol. 48, No. 23 7181
Figure 5. PLS coefficient plot obtained with the N2) probe. Compounds 2h (a), 23 (b), 63 (c), and 47 (d) are embedded into
positive (cyan) and negative (yellow) regions. Favorable interactions (i.e., an increase of -log Ki) between a substituent and the
probe occur in regions with negative coefficients (-0.0056 kcal/mol level, positive regions, cyan), while unfavorable interactions
(i.e., a decrease of -log Ki) between a substituent and the probe occur in regions with positive coefficients (0.0042 kcal/mol level,
negative regions, yellow).
and 6d, respectively) while the region B interacts with
the NH moiety of 23 (Figure 6b). No contact was found
for compound 63 (Figure 6c), suggesting that the lower
affinity of such a compound toward A1AR in comparison
to the affinity of 2h, 23, and 47, could derive from the
lack of hydrogen bond donor groups in its structure. As
a consequence, the electrostatic component of the bind-
ing energy due to hydrogen bond donor groups in the
inhibitor structure appears to be an important key in
determining affinity toward A1AR.
To verify such a hypothesis, starting from data
derived from molecular structures (variables), we have
evaluated how each probe was involved in defining the
interactions with the studied compounds. Results high-
lighted that interactions with all the three probes (C3,
N2), O) contributed in the same measure to the activity
(33.9%, 32.8%, 33.3%, respectively), suggesting that
hydrogen bond acceptor groups, hydrogen bond donor
groups and steric interactions were equally important
in defining the affinity of all inhibitors toward the
receptor.
the ester group, a butyl or propyl chain seemed to be
too bulky, giving unfavorable steric interactions, as
showed by the methyl contour map. Differently, methyl,
ethyl and isopropyl groups showed favorable steric
interactions and resulted to be optimal substituents for
this position.
Moreover, the structure of a phenylethylamino side
chain at the position 4 was worthy of further consider-
ation. In fact, while the linear portion of the chain
(corresponding to the ethylamino spacer) was embedded,
without any contact, into a narrow tunnel delimited by
regions D and E of the C3 map (Figure 4), the phenyl
ring was able to have profitable interactions with the
region C. As a result of such interactions, a very
interesting affinity was found for the phenylethyl
derivatives with respect to compounds with shorter
(benzyl) or bulkier (five or six membered heterocyclic
rings) substituents at C4. This finding suggested that
substituents at the position 4 should be characterized
by a small volume close to the C4, while the distal
portion of the side chain interacting with the region C
of the C3 map tolerated larger groups. Accordingly,
cyclic amino substituents such as pyrrolidino and mor-
pholino moieties, or shorter side chains such as the
benzyl group showed steric hindrance, interacting with
the regions D and E of the model (Figure 4). In addition,
the model showed that para substituted phenylethyl-
amino side chains were preferred to the corresponding
In summary, results from calculations showed that
the 3D QSAR model was able to explain, at the
quantitative level, the structure-activity relationships
of the 4-amino-1H-pyrazolo[3,4-b]pyridine-5-carboxylic
acid esters. Substituents on the nitrogen at the position
4 and on the ester group at the position 5 were very
important for affinity toward A1AR. In particular, for

7182 Journal of Medicinal Chemistry, 2005, Vol. 48, No. 23
Manetti et al.
Figure 6. PLS coefficient plot obtained with the O probe. Compounds 2h (a), 23 (b), 63 (c), and 47 (d) are embedded into positive
(cyan) and negative (yellow) regions. Favorable interactions (i.e., an increase of -log Ki) between a substituent and the probe
occur in regions with negative coefficients (-0.0045 kcal/mol level, positive regions, cyan), while unfavorable interactions (i.e., a
decrease of -log Ki) between a substituent and the probe occur in regions with positive coefficients (0.0036 kcal/mol level, negative
regions, yellow).
ortho and meta analogues (compare 3e versus 3c and
3d; 3h versus 3f and 3g; 5k versus 5i and 5j).
compound 1c. In addition, several derivatives described
in this paper revealed a higher affinity compared to the
ethyl esters previously reported by us.16 Moreover,
selected compounds also showed affinity values in the
nanomolar range toward hA1AR. All the most interest-
ing compounds were very selective for the A1AR, show-
ing low or no affinity toward both A2AR and A3AR.
From these experimental data, it was also evident
that either substituents at the position 4 or the ester
group at the position 5 played a very important role in
defining the affinity properties of the studied com-
pounds.
Finally, the new 3D QSAR model, built also using
compounds 2-8, demonstrated to have the knowledge
for estimating/predicting the affinity of a very large
number of inhibitors bearing to different structural
classes. Such a model showed a better prediction ability
with respect to the pseudoreceptor model previously
published by us17 and will be in turn used to provide
suggestions to further optimize the new pyrazolo-
pyridine derivatives.
It was also interesting to note that the 3D QSAR
model was able to explain the low activity observed for
4-amino substituted 1-(2-chloro-2-phenylethyl)-6-meth-
ylthio-1H-pyrazolo[3,4-d]pyrimidines previously re-
ported by us,18 due to lack of any profitable hydrogen
bond interaction with the region A of the N2) map.
Although the good predictivity for the biological
activity of the test set compounds indicated the quality
of the 3D QSAR study, however, we are confident that
a model derived from affinity data toward bA1AR cannot
be considered as a reliable model also for affinity data
toward the hA1AR. In fact, it is well know that bA1AR
and hA1AR are characterized by different antagonist
binding sites,30 reflecting differences in the activity data
found for the same antagonists toward bA1AR and hA1-
AR. As an example, compounds 2g, 2h, and 3h showed
affinity toward bA1AR significantly different from af-
finity values toward hA1AR.
Conclusions
Experimental Section
Results reported in this paper strongly indicated that
some of the newly synthesized pyrazolo-pyridines were
good A1AR ligands toward both human and bovine
receptors, with high selectivity with respect to other
adenosine receptors. In particular the methyl ester 2h
as well as the isopropyl ester 5h, both of them bearing
a p-methoxyphenylethylamino side chain at the position
4, possessed an affinity toward A1AR about 8-fold and
7-fold higher than that of the corresponding parent
Chemistry. Starting materials were purchased from Ald-
rich-Italia (Milan, Italy). Melting points were determined with
a Bu¨chi 530 apparatus and are uncorrected. IR spectra were
measured in KBr with a Perkin-Elmer 398 spectrophotometer.
1H NMR spectra were recorded in a (CD3)2SO solution on a
Varian Gemini 200 (200 MHz) instrument. Chemical shifts are
reported as δ (ppm) relative to TMS as internal standard, J
are expressed in Hz. 1H patterns are described using the
following abbreviations: s ) singlet, d ) doublet, dd ) double

Inhibitors of A1 Adenosine Receptors
Journal of Medicinal Chemistry, 2005, Vol. 48, No. 23 7183
doublet, t ) triplet, q ) quartet, sx ) sextet, sept ) septet, m
) multiplet, br ) broad. All compounds were tested for purity
by TLC (Merk, Silica gel 60 F254, CHCl3 as the eluant).
Analyses for C, H, N were within (0.3% of the theoretical
value.
40-60 °C) (5 mL), to give 2h (0.32 g, 70%) as a white solid;
mp 102-103°C. 1H NMR: δ 3.06 (t, J ) 7.8, 2H, CH2Ar), 3.83
(s, 3H, OCH3), 3.85-3.96 (m, 5H, OCH3+ CH2NH), 4.72-4.87
and 4.97-5.12 (2m, 2H, CH2N), 5.54-5.66 (m, 1H, CHCl),
6.85-7.54 (m, 9H Ar), 8.09 (s, 1H, H-3), 8.84 (s, 1H, H-6), 9.25
(br s, 1H, NH, disappears with D2O). IR cm-1: 3268 (NH), 1680
(CO). Anal. (C25H25N4O3Cl) C, H, N.
1-(2-Chloro-2-phenylethyl)-4-hydroxy-1H-pyrazolo[3,4-
b]pyridine-5-carboxylic Acid (10). To a solution of 4-chloro-
1-(2-chloro-2-phenylethyl)-1H-pyrazolo[3,4-b]pyridine-5-car-
boxylic acid ethyl ester 9 (5 g, 13.7 mmol) in 96% ethanol (50
mL) was added 3 M H2SO4 (20 mL). The solution was refluxed
24 h, and the crude precipitate was filtered and recrystallized
from absolute ethanol, to give 10 (3.26 g, 75%) as a white solid;
Biological Methods. [3H]DPCPX, [3H]CGS21680 and [125I]-
AB-MECA were obtained from DuPont-NEN (Boston, MA).
Adenosine deaminase was from Sigma Chemical Co. (St. Louis,
MO). All other reagents were from standard commercial
sources and of the highest commercially available grade.
A1 and A2AAR Binding Assay. Affinity of the new com-
pounds toward A1 and A2AAR was evaluated by competition
experiments assessing their ability to displace [3H]DPCPX and
[3H]CGS21680 binding from bovine cortical and striatal mem-
branes, respectively. Binding assays were carried out as
previously described.31,32 Pharmacological profile of the most
active compounds toward A1 AR were evaluated by GTP shift
assay.19 Moreover, affinity of the most A1AR selective com-
pounds was also evaluated in human A1AR CHO transfected
cells (kindly supplied by K. N. Klotz from Wurzburg Univer-
sity, Germany).33
1
mp 228-229 °C. H NMR: δ 4.79-4.97 and 5.11-5.29 (2 m,
2H, CH2N), 5.58-5.74 (m, 1H, CHCl), 7.22-7.78 (m, 5H Ar),
8.32 (s, 1H, H-3), 8.76 (s, 1H, H-6). IR cm-1: 3100-2700 (OH),
1641 (CO). Anal. (C15H12N3O3Cl) C, H, N.
Method A. Example. Methyl 1-(2-Chloro-2-phenylethyl)-
4-hydroxy-1H-pyrazolo[3,4-b]pyridine-5-carboxylate (11a).
To a suspension of 10 (1.59 g, 5 mmol) in CH3OH (30 mL) was
added 98% H2SO4 (4 mL) dropwise. The mixture was refluxed
for 18 h, then concentrated under reduced pressure. The crude
was dissolved in CHCl3 (50 mL), washed with a 5% NaHCO3
solution (2 × 20 mL), then with water (20 mL), dried (MgSO4),
and evaporated under reduced pressure. The residue was
crystallized from absolute ethanol to give 11a (1.2 g, 72%) as
A3AR Binding Assay. [125I]AB-MECA binding to A3AR in
bovine cortical membranes was performed in 50 mM Tris, 10
mM MgCl2, and 1 mM EDTA buffer (pH 7.4) containing 0.2
mg of proteins, 2 U/mL adenosine deaminase, and 20 nM
DPCPX. Incubations were carried out in duplicate for 90 min
at 25 °C. Nonspecific binding was determined in the presence
of 50 µM R-PIA and represented approximately 30% of the
total binding. The binding reaction was terminated by filtra-
tion through a Whatman GF/C filter, washing three times with
5 mL of ice-cold buffer. All compounds were routinely dissolved
in dimethyl sulfoxide (DMSO) and diluted with assay buffer
to the final concentration (the amount of DMSO never ex-
ceeded 2%). At least six different concentrations spanning 3
orders of magnitude, adjusted appropriately for the IC50 of each
compound, were used. IC50 values, computer-generated using
a nonlinear regression formula on a computer program (Graph-
Pad, San Diego, CA), were converted to Ki values, knowing
the Kd values of radioligands in the different tissues and using
the Cheng and Prusoff equation.34
Computational Chemistry. All calculations were carried
out on SGI workstations and a SGI Origin300 server. Due to
the fact that biological evaluation of all the new chiral
compounds reported here was carried out using racemic
mixtures, it was arbitrarily decided to model them with
undefined chirality by means of the software Catalyst, thus
allowing the software, during the alignment procedure, to
choose which configuration of the asymmetric carbon atom
common to compounds 1-8 and 77-80 was most appropriate.
This decision can be justified on the basis of the fact that a
number of chiral compounds was included in the training set
and 2h was used as one of the four template structure. The
S-2h enantiomer has been chosen by the program to be aligned
to compound 23 (see below). However, to the best of our
knowledge and our results, there are no experimental data
supporting the hypothesis that S enantiomers are more active
than R enantiomers in the series of pyrazolo[3,4-b]pyridines
bearing a 2-chloro-2-phenylethyl side chain at N1.
1
a white solid; mp 138-139 °C. H NMR: δ 4.01 (s, 3H, CH3),
4.80-4.95 and 5.01-5.18 (2 dd, 2H, CH2N), 5.51-5.65 (m, 1H,
CHCl), 7.28-7.53 (m, 5H Ar), 8.16 (s, 1H, H-3), 8.85 (s, 1H,
H-6). IR cm-1: 3100 (OH), 1674 (CO). Anal. (C16H14N3O3Cl)
C, H, N.
Method B. Example. Methyl 4-Chloro-1-(2-chloro-2-
phenylethyl)-1H-pyrazolo[3,4-b]pyridine-5-carboxylate
(12a). The Vilsmeier complex, previously prepared from POCl3
(6.13 g, 40 mmol) and anhydrous dimethylformamide (DMF)
(2.92 g, 40 mmol), was added to a suspension of 11a (3.31 g,
10 mmol) in CHCl3 (20 mL). The mixture was refluxed for 8
h. The solution was washed with H2O (2 × 20 mL), dried
(MgSO4), and evaporated under reduced pressure, and the
residue oil was crystallized by adding absolute ethanol (10 mL)
to give 12a (2.38 g, 68%) as a white solid; mp 113-114 °C. 1H
NMR: δ 4.00 (s, 3H, CH3), 4.85-4.97 and 5.06-5.20 (2 dd,
2H, CH2N), 5.53-5.64 (m, 1H, CHCl), 7.26-7.52 (m, 5H Ar),
8.23 (s, 1H, H-3), 9.03 (s, 1H, H-6). IR cm-1: 1729 (CO). Anal.
(C16H13N3O2Cl2) C, H, N.
Cyclopropylmethyl 1-(2-Chloro-2-phenylethyl)-4-hy-
droxy-1H-pyrazolo[3,4-b]pyridine-5-carboxylate (13). To
a suspension of 10 (0.5 g, 1.57 mmol) in CHCl3 (10 mL) was
added thionyl chloride (1.5 g, 12.6 mmol). The mixture was
refluxed for 4 h, then concentrated under reduced pressure.
To the oil were added CHCl3 (10 mL) and cyclopropylmethanol
(2.86 g, 40 mmol), and the solution was refluxed for 6 h. The
mixture was then concentrated under reduced pressure, to give
a crude oil, that was dissolved in CHCl3 (10 mL) and washed
with a 5% NaHCO3 solution (2 × 5 mL) and with water (10
mL). The organic solution was dried (MgSO4), filtered, and
evaporated under reduced pressure, and the residue oil was
crystallized by adding a mixture 1:1 of diethylic ether/
petroleum ether (bp 40-60°C), to give 13 (0.29 g, 50%) as a
white solid; mp 101-102°C. 1H NMR: δ 0.35-0.50 and 0.61-
0.78 (2m, 4H, 2CH2 cycloprop), 1.20-1.40 (m, 1H, CH cyclo-
prop), 4.26 (d, J ) 7.2, 2H, CH2O), 4.76-4.83 and 5.05-5.20
(2m, 2H, CH2N), 5.52-5.68 (m, 1H, CHCl), 7.22-7.59 (m, 5H
Ar), 8.19 (s, 1H, H-3), 8.94 (s, 1H, H-6), 12.20-12.30 (br s,
1H, OH, disappears with D2O). IR cm-1: 3200-3000 (OH),
1671 (CO). Anal. (C19H18N3O3Cl) C, H, N.
Catalyst 4.9 was used either to sketch structures with
undefined chirality and to perform a conformational search.
The best searching procedure was applied to select representa-
tive conformers within 5 kcal/mol from the global minimum.
Cerius2 software was applied to align molecules, by ap-
plication of the Field Fit alignment method (using CFF force
field). In detail, Target Model was used as the Align Strategy,
Field as the Align Method, and Rigid as the Align Type. A
high activity and structural rigidity were the parameters
applied to choose the Target Model.
Method C. Example. Methyl 1-(2-chloro-2-phenylethyl)-
4-{[2-(4-methoxyphenyl)ethyl]amino} -1H-pyrazolo[3,4-
b]pyridine-5-carboxylate (2h). To a solution of 12a (0.35
g, 1 mmol) in anhydrous toluene (5 mL) was added 4-meth-
oxyphenethylamine (0.6 g, 4 mmol), and the reaction mixture
was stirred at room temperature for 48 h. The mixture was
washed with H2O (2 × 10 mL), the organic phase was dried
(MgSO4), filtered, and evaporated under reduced pressure, and
the residue oil was crystallized by adding petroleum ether (bp
The descriptive steric, electrostatic, and hydrogen bond
interactions, represented by the Lennard-Jones energy, the
Coulombic energy, and a hydrogen bonding term, respectively,
were calculated using GRID, version 21. Analysis was per-

7184 Journal of Medicinal Chemistry, 2005, Vol. 48, No. 23
Manetti et al.
(10) Mu¨ller, C. E. A1 adenosine receptors and their ligands: overview
and recent developments. Farmaco 2001, 56, 77-80.
(11) Chebib, M.; McKeveney, D.; Quinn, R. J. 1-Phenylpyrazolo[3,
4-d]pyrimidines; structure-activity relationships for C6 substit-
uents at A1 and A2A adenosine receptors. Bioorg. Med. Chem.
2000, 8, 2581-2590.
(12) Terai, T.; Kusonoki, T.; Kita, Y.; Akahane, A.; Shiokawa, Y.;
Kohno, Y.; Horiai, H.; Uehara, Y.; Yoshida, K. FK453: A novel
nonxanthine adenosine A1 receptor antagonist as diuretic.
Cardiovasc. Drug Rev. 1997, 115, 44-58.
(13) Daly, J. W.; Hong, O.; Padgett, W. L.; Shamim, M. T.; Jacobson,
K. A.; Ukena, D. Non-xanthine heterocycles: activity as antago-
nists of A1- and A2-adenosine receptors. Biochem. Pharmacol.
1988, 37, 655-664.
(14) Daly, J. W.; Hutchinson, K. D.; Secunda, S. I.; Shy, D.; Padgett,
W. L.; Shamin, M. T. 1-Methyl-4-substituted-1H-pyrazolo[3,4-
b]pyridine-5-carboxylic Acid Derivatives: Effect of structural
alterations on activity at A1 and A2 adenosine receptors. Med.
Chem. Res. 1994, 4, 293-306.
(15) (a) Akane, A.; Kuroda, S.; Itani, H.; Shimizu, Y. Patent NO
WO9803507, 1998; Chem. Abstr. 1998, 128, 154090. (b) Akane,
A.; Nishimura, S.; Kuroda, S.; Itani, H. Patent NO JP10182643,
1998; Chem. Abstr. 1998, 129, 144876.
(16) Schenone, S.; Bruno, O.; Fossa, P.; Ranise, A.; Menozzi, G.; Mosti,
L.; Bondavalli, F.; Martini, C.; Trincavelli, M. L. Synthesis and
biological data of 4-amino-1-(2-chloro-2-phenylethyl)-1H-pyra-
zolo[3,4-b]pyridine-5-carboxylic acid ethyl esters, a new series
of A1-adenosine receptor (A1AR) ligands. Bioorg. Med. Chem.
Lett. 2001, 11, 2529-2531.
(17) 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
selective A1 receptor antagonists. J. Med. Chem. 2002, 45, 4875-
4887.
formed using a grid spacing of 1 Å. The grid dimensions were
(Å): Xmin/Xmax, -13.0/15.0; Ymin/Ymax, -13.0/13.0; Zmin/Zmax
,
-11.0/12.0.
PLS models were calculated using GOLPE, version 4.5.12.
The software automatically rejects variables having a total
sum of square (SS) less than 10-7. Afterward, an advanced
pretreatment was performed that omitted from the analysis
grid points (variables) with too low standard deviation values
(<0.02). Variables which exhibited only two values were also
removed.
Since groups of variables may represent the same structural
information as a single variable, after application of the D
optimal design criterion, the variables were grouped together
according to the Smart Region Definition (SRD) grouping
algorithm, selecting a number of 1879 seeds in the PLS weight
space with a critical cutoff distance of 1.0 Å and a collapsing
cutoff distance of 2.0 Å.
The obtained groups of variables (667 regions) were used
in the Fractional Factorial Design (FFD) variable selection
procedure, replacing the original variables. Following this
protocol, groups of variables instead of single variables were
removed from the data file. To evaluate the effect of the
grouped variables on the predictivity, a number of reduced
models was built by removing the variables according to the
FFD design and using 20% of dummies, on the basis of the
random group cross-validation approach. In detail, the ligands
were randomly assigned to five groups, each one containing
an equal number of ligands. Models were built keeping one of
these groups out of the analysis (the leave-one-out cross-
validation methodology) until all the ligands had been kept
out once. The formation of the groups and the validation was
repeated 20 times, using a maximum dimensionality of five
principal components. After application of the FFD variable
selection, 1247 active variables were kept.
(18) Schenone, S.; Bruno, O.; Bondavalli, F.; Ranise, A.; Mosti, L.;
Menozzi, G.; Fossa, P.; Manetti, F.; Morbidelli, L.; Trincavelli,
L.; Martini, C.; Lucacchini, A. Synthesis of 1-(2-chloro-2-phen-
ylethyl)-6-methylthio-1H-pyrazolo[3,4-d]pyrimidines 4-amino sub-
stituted and their biological evaluation. Eur. J. Med. Chem.
2004, 39, 153-160.
(19) Da Settimo, F.; Primofiore, G.; Taliani, S.; Marini, A. M.; La
Motta, C.; Novellino, E.; Greco, G.; Lavecchia, A.; Trincavelli,
L.; Martini, C. 3-Aryl[1,2,4]triazino[4,3-a]benzimidazol-4(10H)-
ones: a new class of selective A1 adenosine receptor antagonists.
J. Med. Chem. 2001, 44, 316-327.
(20) 4-Amino-1-(2-chloro-2-phenylethyl)-1H-pyrazolo[3,4-b]pyridine-
5-carboxylic acid ethyl ester derivatives as reported in ref 16
(1a-1f) and 1-(2-chloro-2-phenylethyl)-6-methylthio-1H-pyra-
zolo[3,4-d]pyrimidines-4-amino substituted as reported in ref 18
(77-80).
(21) 1,8-Naphthyridine as reported in ref 30 (15-44); 1,2,4-triazolo-
[4,3-a]quinoxalin-1-ones as reported in ref 31 (45-60); 3-aryl-
[1,2,4]triazino[4,3-a]benzimidazol-4(10H)-ones as reported in ref
19 (61-76).
(22) When a Ki value is impossible to be determined for a compound,
there are two alternative ways for reporting its activity. The first
one is to show the percentage of inhibition of specific radioligand
binding at the maximum dose of the compound tested (10 µM).
Alternatively, for compounds 2d, 6d, and 6e, a Ki > 10 µM
couldbe reported, similarly to that described for compounds 37
and 38 (ref 30). However, both cases describe inactive com-
pounds. As a consequence, on the basis of the fact that also
inactive compounds could be a source of important structural
information for qualitative and quantitative structure-activity
relashionship analysis, we (i. Corelli, F.; Manetti, F.; Tafi, A.;
Campiani, G.; Nacci, V.; Botta, M. Diltiazem-like calcium entry
blockers: a hypothesis of the receptor-binding site based on a
comparative molecular field analysis model. J. Med. Chem. 1997,
40, 125-131; ii.) Tafi, A.; Anastassopoulou, J.; Theophanides,
T.; Botta, M.; Corelli, F.; Massa, S.; Artico, M.; Costi, R.; Di
Santo, R.; Ragno, R. Molecular modeling of azole antifungal
agents active against Candida albicans. 1. A comparative
molecular field analysis study. J. Med. Chem. 1996, 39, 1227-
1235) and others (i. Ragno, R.; Artico, M.; De Martino, G.; La
Regina, G.; Coluccia, A.; Di Pasquali, A.; Silvestri, R. Docking
and 3-D QSAR studies on indolyl aryl sulfones. Binding mode
exploration at the HIV-1 reverse transcriptase nonnucleoside
binding site and design of highly active N-(2-hydroxyethyl)-
carboxamide and N-(2-hydroxyethyl)carbohydrazide derivatives.
J. Med. Chem. 2005, 48, 213-223; ii. Pauwels, R.; Andries, K.;
Debyser, Z.; Kukla, M. J.; Schols, D.; Breslin, H. J.; Woesten-
borghs, R.; Desmyter, J.; Janssen, M. A. C.; De Clercq, D.;
Janssen, P. A. J. New tetrahydroimidazo[4,5,1-jk][1,4]-benzodi-
azepin-2(1H)-one and -thione derivatives are potent inhibitors
of human immunodeficiency virus type 1 replication and are
synergistic with 2′,3′-dideoxynucleoside analogues. Antimicrob.
Agents Chemother. 1994, 38, 2863-2870) usually chosed to not
Acknowledgment. Financial support from Italian
MIUR (PRIN 2004037521_002) is gratefully acknowl-
edged. M.B. thanks the Merk Research Laboratories
(2004 Academic Development Program Chemistry
Award). F.M. thanks the Divisione di Chimica Farma-
ceutica della Societa` Chimica Italiana and Farmindus-
tria for the “Premio Farmindustria 2004” award. We are
indebted with Molecular Discovery for the GRID code
and we would like to thank Prof. Gabriele Cruciani
(University of Perugia) for the use of the program
GOLPE in his lab.
Supporting Information Available: Details of the syn-
thesis of compounds 2-14 and their elemental analysis data.
This material is available free of charge via the Internet at
http://pubs.acs.org.
References
(1) Jacobson, K. A.; van Galen, P. J.; Williams, M. Adenosine
receptors: pharmacology, structure-activity relationships, and
therapeutic potential. J. Med. Chem. 1992, 35, 407-422.
(2) Daly, J. W. Adenosine receptors: targets for future drugs. J.
Med. Chem. 1982, 25, 197-207.
(3) Collis, M. G.; Hourani, S. M. Adenosine receptor subtypes.
Trends Pharmacol. Sci. 1993, 14, 360-363.
(4) Ralevic, V.; Burnstock, G. Receptors for purines and pyrimidines.
Pharmacol. Rev. 1998, 50, 413-492.
(5) Poulsen, S. A.; Quinn, R. J. Adenosine receptors: new opportuni-
ties for future drugs. Bioorg. Med. Chem. 1998, 6, 619-641.
(6) Mu¨ller, C. E.; Scior, T. Adenosine receptors and their modulators.
Pharm. Acta Helv. 1993, 68, 77-111.
(7) Mu¨ller, C. E.; Stein, B. Adenosine receptor antagonists: struc-
tures and potential therapeutic applications. Curr. Pharm. Des.
1996, 2, 501-530.
(8) Dhalla, A. K.; Shryock, J. C.; Shreeniwas, R.; Belardinelli, L.
Pharmacology and therapeutic applications of A1 adenosine
receptor ligands. Curr. Top. Med. Chem. 2003, 3, 369-385.
(9) Ribeiro, J. A.; Sebastia˜o, A. M.; de Mendonc¸a, A. Adenosine
receptors in the nervous system: pathophysiological implica-
tions. Prog. Neurobiol. 2002, 68, 377-392.

Inhibitors of A1 Adenosine Receptors
Journal of Medicinal Chemistry, 2005, Vol. 48, No. 23 7185
omit them from a statistical treatment such as a QSAR study,
and assumed that their activity could be approximated to that
of the least active compound or to the maximum concentration
of the ligand used to perform biological tests. Accordingly, a Ki
> 10000 nM (a -log Ki ) 5.0) was chosen for compounds 2d,
6d, 6e, 37 and 38, very similar to that of the least active
compound 2e (-log Ki ) 5.4) and to that of the structurally
related compund 80 (-log Ki ) 5.0), and also corresponding to
the maximum concentration of each ligand tested (10 µM).
(23) Nielsen, E. O.; Audouze, K.; Peters, D. New series of morpholine
and 1,4-oxazepane derivatives as dopamine D4 receptor lig-
ands: synthesis and 3D-QSAR model. J. Med. Chem. 2004, 47,
3089-3104.
Affinity of DPCPX (expressed as -log Ki) was estimated to be
9.6, in good agreement with the experimental value of 10.2.
Alignment of such a compound in the current pharmacophoric
model (Figure 2) shows HY1-3 filled by the propyl chain at
positions 1 and 3 and by the cyclopentyl ring, respectively.
Moreover, the 6- and 2-carbonyl groups are HBA1 and HBA2,
respectively, while HBD is the 1-NH group.
(30) Ferrarini, P. L.; Betti, L.; Cavallini, T.; Gianaccini, G.; Lucac-
chini, A.; Manera, C.; Martinelli, A.; Ortore, G.; Saccomanni, G.;
Tuccinardi, T. Study on affinity profile toward native human
and bovine adenosine receptors of a series of 1,8-naphthyridine
derivatives. J. Med. Chem. 2004, 47, 3019-3031.
(31) Colotta, V.; Catarzi, D.; Varano, F.; Cecchi, L.; Filacchioni, G.;
Martini, C.; Trincavelli, L.; Lucacchini, A. Synthesis and struc-
ture-activity relationships of a new set of 2-arylpyrazolo[3,4-
c]quinoline derivatives as adenosine receptor antagonists. J.
Med. Chem. 2000, 43, 3118-3124.
(24) GOLPE 4.5.12; Multivariate Infometric Analyses: Viale dei
Castagni, 16 Perugia, Italy, 1999.
(25) GRID 21; Molecular Discovery Ltd, 215 Marsh Road, Pinner,
Middlesex, UK.
(26) Catalyst 4.9, Accelrys, Inc., San Diego, CA, 2003.
(27) Marshall G. R., Barry, C. D.; Bosshard, H. E.; Dammkoehler,
R. A.; Dunn, D. A. The conformational parameter in drug
design: the active analog approach. In Computer-Assisted Drug
Design, Olson, E. C., Christoffersen, R. E., Eds.; American
Chemical Society: Washington, DC, 1979; Vol. 112, pp 205-
226.
(32) Pirovano, I. M.; IJzerman, A. P.; van Galen, P. J. M.; Soudijn,
W. The influence of molecular structure of N6-(aminoalkyl)-
adenosines on adenosine receptor affinity and intrinsic activity.
Eur. J. Pharmacol. 1989, 172, 185-193.
(33) Catarzi, D.; Colotta, V.; Varano, F.; Calabri, F. R.; Lenzi, O.;
Filacchioni, G.; Trincavelli, M. L.; Martini, C.; Tralli, A.; Mon-
topoli, C.; Moro, S. 2-Aryl-8-chloro-1,2,4-triazolo[1,5-a]quinoxa-
lin-4-amines as highly potent A1 and A3 adenosine receptor
antagonists. Bioorg. Med. Chem. 2005, 13, 705-15.
(34) Cheng, Y. C.; Prusoff, W. H. Relation between the inhibition
constant Ki and the concentration of inhibitor which causes fifty
percent inhibition (IC50) of an enzyme reaction. Biochem. Phar-
macol. 1973, 22, 3099-3108.
(28) Cerius2, Version 4.8; Accelrys Inc., San Diego, CA, 2003.
(29) An additional model was also generated by application of the
same computational protocol and including DPCPX in the
training set. Both statistical parameters (q2 ) 0.7 on the training
set in a PLS model with five components and a SDEP ) 0.57 on
the test set) and PLS pseudo-coefficient contour maps were very
similar to those of the original 3D QSAR model, suggesting that
CPCPX did not bias significantly the generation of the model.
JM050407K