2,4,6-Trisubstituted pyrimidines as a new class of selective adenosine A1 receptor antagonists

Article abstract of DOI:10.1021/jm049448r

Adenosine receptor antagonists usually possess a bi- or tricyclic heteroaromatic structure at their core with varying substitution patterns to achieve selectivity and/or greater affinity. Taking into account molecular modeling results from a series of pot

Full text of DOI:10.1021/jm049448r

Evaluation Only. Created with Aspose.PDF. Copyright 2002-2021 Aspose Pty Ltd.
J. Med. Chem. 2004, 47, 6529-6540
6529
2,4,6-Trisubstituted Pyrimidines as a New Class of Selective Adenosine A1
Receptor Antagonists
Lisa C. W. Chang, Ronald F. Spanjersberg, Jacobien K. von Frijtag Drabbe Ku¨nzel, Thea Mulder-Krieger,
Gijs van den Hout, Margot W. Beukers, Johannes Brussee, and Adriaan P. IJzerman*
Division of Medicinal Chemistry, Leiden/Amsterdam Center for Drug Research, P.O. Box 9502,
2300 RA Leiden, The Netherlands
Received July 13, 2004
Adenosine receptor antagonists usually possess a bi- or tricyclic heteroaromatic structure at
their core with varying substitution patterns to achieve selectivity and/or greater affinity. Taking
into account molecular modeling results from a series of potent adenosine A1 receptor
antagonists, a pharmacophore was derived from which we show that a monocyclic core can be
equally effective. To achieve a compound that may act at the CNS we propose imposing a
restriction related to its polar surface area (PSA). In consequence, we have synthesized two
novel series of pyrimidines, possessing good potency at the adenosine A1 receptor and desirable
PSA values. In particular, compound 30 (LUF 5735) displays excellent A1 affinity (Ki ) 4 nM)
and selectivity (e50% displacement of 1 µM concentrations of the radioligand at the other
three adenosine receptors) and has a PSA value of 53 Å2.
Introduction
Adenosine is an endogenous ligand, ubiquitous
throughout the human body. Its many extracellular
functions are mostly accomplished through G protein
coupled receptors, of which thus far four groups have
been cloned and identified, the A1, A2A, A2B, and A3.1
These adenosine receptors are present in varying levels
of expression in different parts of the body. The wide-
Figure 1. (i) Theophylline. (ii) Caffeine.
the pyrazolopyrimidines),8 3-nitrogen tricyclic systems
(e.g., the imidazoquinolines),9 and 4-nitrogen tricyclic
systems (e.g., triazoloquinoxalines),10 to name but a few
groups. Indeed in two recent reviews11,12 it is stated
quite clearly that the different structural classes for A1
adenosine receptor antagonists are bi-and tricyclic
heterocyclic compounds. Moreover, an investigation in
our group highlighted the general low affinity of mono-
cyclic compounds.13 The exceptions to this were two
5-membered heterocycles, namely, thiazoles and thia-
diazoles.13,14 In the wider field of adenosine receptor
ligands, (dihydro)pyridines have been developed and
explored as A3 receptor antagonists by Jacobson et
al.15-17 and as adenosine receptor agonists by Bayer.18
Although these series show affinity for the A1 receptor,
they have not been specifically investigated for A1
receptor efficacy, and in some cases they have been
developed to show particularly good, although not
selective, affinity for the A2B receptor.19
Adenosine A1 receptors are in abundance in the
mammalian brain, and the role that they play in
important functions, such as in the modulation of
neurotransmitter release, sleep regulation, and cogni-
tion enhancement, has been thoroughly investigated.20-22
For this reason it is essential that a compound targeted
at these therapeutic areas is able to cross the blood-
brain barrier (BBB). Research into the BBB and the
ability of a compound to cross it has become a highly
investigated topic in recent years. A recent review23
highlighted some “rules of thumb” which have emerged
from numerous research articles from the past few
spread purpose and presence of adenosine has led to
substantial research into the individual adenosine
receptors as pharmaceutical targets. Over the years
there have been many attempts to design and develop
adenosine receptor antagonists, and over the past
decade the search for ligands that show selectivity
toward individual receptors has intensified as the role
of the receptors in many therapeutic areas expands.1
Xanthines were the first class of adenosine antagonists
to be investigated and, as such, are also the most well
explored. The best known compounds of this series are
theophylline and caffeine (Figure 1). Research into non-
xanthine ligands has grown tremendously in recent
years, and the past decade has seen a number of new
and interesting compounds showing varying degrees of
selectivity for the adenosine receptors. In the early
1990s a paper by van Galen et al. detailed a model for
the adenosine A1 receptor binding site based on the
superimposition of xanthines with adenosine.2 A num-
ber of different computational-based papers for the A1
receptor followed,3-5 providing yet more models for the
supposed binding site of the A1 receptor. The conse-
quential expansion in the field of adenosine A1 antago-
nism generated a number of bi- and tricyclic hetero-
aromatic systems, featuring 2-nitrogen bicyclic systems
(e.g., the naphthyridines),6 3-nitrogen bicyclic systems
(e.g., deazaadenines),7 4-nitrogen bicyclic systems (e.g.,
* To whom correspondence should be addressed. E-mail:
ijzerman@lacdr.leidenuniv.nl. Tel: +31 (0)71 527 4651. Fax: +31 (0)-
71 527 4565.
10.1021/jm049448r CCC: $27.50 © 2004 American Chemical Society
Published on Web 11/13/2004

Evaluation Only. Created with Aspose.PDF. Copyright 2002-2021 Aspose Pty Ltd.
6530 Journal of Medicinal Chemistry, 2004, Vol. 47, No. 26
Chang et al.
Figure 2. The nine A1 adenosine receptor antagonists modeled in this study. Of the two areas marked on each molecule, the left
ellipse signifies electron-poor regions corresponding to hydrogen-bond donating regions. The area marked with the right-hand
ellipse on each molecule depicts electronically rich regions corresponding to hydrogen-bond accepting regions: a (Ki (hA1) ) 25
47,48
nM);46 b (Ki (bA1) ) 0.15 nM);6 c (Ki (rA1) ) 7.9 nM);40 d (Ki (rA1) ) 21 nM; Ki (hA1) ) 3.5 nM);
(hA1) ) 3.9 nM);48 g (Ki (rA1) ) 10 nM);9 h (IC50 (rA1) ) 28 nM);10 i (Ki (rA1) ) 7.3 nM.41
e (Ki (rA1) ) 7.3 nM;14 f (Ki
years. Among these was the almost qualitative example
that the sum of the nitrogen and oxygen atoms in a
molecule should be five or less for that molecule to have
a high chance of entering the brain.24 Of the more
quantitative prediction techniques, the rule that for
good brain permeation, as for good intestinal absorption,
the polar surface area (PSA) of a molecule should be
below a certain limit has been very thoroughly investi-
gated.25-29 The PSA is defined as the surface area of a
molecule occupied by nitrogen and oxygen atoms, and
hydrogen atoms that are attached to these atoms.29 It
has become one of the most convenient and reliable
parameters to calculate, and the limits for brain pen-
etration according to research, e.g., by Kelder et al.,25
have been proposed to be in the region of 60-70 Å2.
In this present study, we subjected known ligands to
molecular modeling investigations and derived a phar-
macophore upon which we could design potent adeno-
sine A1 receptor antagonists. We describe the develop-
ment of a series of novel mono-heterocyclic compounds
that show high affinity for the human A1 adenosine
receptor, namely, the 2,6-diphenyl-4-amidopyrimidines.
With the notion of the importance of the BBB, we also
calculated PSA values on this set of compounds and
realized that they were within the limits described by
Kelder et al.25 In consequence, we subsequently syn-
thesized a second series of pyrimidines also with “good”
PSA values. These 4,6-diphenyl-2-amidopyrimidines
displayed equal affinity and a notable increase in
selectivity for the A1 adenosine receptor over the A2A
and A3 receptors.
molecular modeling package and minimized, and only
the lowest energy conformer was taken into consider-
ation. This conformer was then subjected to surface
calculations, from which an electrostatic potential en-
ergy was drawn, and upon which the electron density
was mapped; an example for DPCPX is given in Figure
3. In Figure 3, the shades of red and blue denote
relatively electronegative and electron-poor regions,
respectively. These areas are mapped upon the selection
of compounds modeled. It can clearly be seen that all
molecules have in common an adjacent electron-rich and
-poor area at the “bottom” corresponding to hydrogen-
bond accepting and donating regions, as denoted by the
respective ellipses marked in Figure 2. Superimposition
of the compounds using these regions as a basis resulted
in the illustration seen in Figure 4. A simplified version
is shown in Figure 5, where four of the antagonists of
Figure 2 (b, d, e, f,) with the electrostatic potential
energy surface of just one of the molecular models (d)
were chosen to show the general pattern of the resulting
superimposition. This resulting pharmacophore derived
from the molecular modeling is illustrated with a
schematic diagram in Figure 6. There seems to be a
requirement for neighboring electronically rich and poor
regions at the “bottom” of the molecule (labeled regions
A and B) corresponding to hydrogen-bond accepting and
donating regions, respectively. At the “top” of the figure
(region C) there is a requirement for another hydrogen-
bond acceptor. Furthermore, three lipophilic entities
about the central core, labeled L1, L2, and L3, are
desirable for good binding properties.
This phamacophore is in agreement with a model for
the bovine A1 adenosine receptors as published by Da
Settimo et al. to rationalize the SARs of a set of aryl-
triazino-benzimidazolones.31 This model was further
exploited by the same group to provide a series of
imidazotriazines with good affinity for the bovine A1
Results and Discussion
Molecular Modeling. Molecular modeling work was
conducted on a number of ligands with high affinity and
selectivity for the A1 adenosine receptor as listed in
Figure 2. The molecules were drawn in the SPARTAN30

Evaluation Only. Created with Aspose.PDF. Copyright 2002-2021 Aspose Pty Ltd.
Selective Adenosine A1 Receptor Antagonists
Journal of Medicinal Chemistry, 2004, Vol. 47, No. 26 6531
Figure 5. A simplified picture showing only four of the
antagonists (b, d, e, f) superimposed upon each other, includ-
ing the electrostatic potential energy surface of the antagonist
d.
Figure 3. (a) DPCPX in its lowest energy conformer as
calculated with Spartan. (b) DPCPX with the electrostatic
potential energy surface mapped as a transparent cloud
around the “stick” model. (c) Molecular electrostatic potential
energy surface of DPCPX, where the color coding is in the
Figure 6. Schematic diagram of the resulting pharmacophore
attained from the modeling study. A and B respectively
represent H-bond accepting and donating regions at the
“bottom” of the molecule. An H-bond acceptor is denoted by
C. Three lipophilic domains are represented by L1, L2, and
L3.
range from -60 (deepest red) to 60 (deepest blue) kcal‚mol-1
.
Da Settimo. In our case, the model proposed only
suggests an aromatic core with the given characteristics,
to be substituted accordingly to fill the lipophilic pock-
ets. Thus, it seemed that to fulfill the requirements of
this model, a mono-heterocycle as the central core would
be sufficient. Examining the model in more detail, a
single aromatic group containing a nitrogen atom would
fit the hydrogen-bond accepting region. An amido group
in an adjacent position would fulfill hydrogen-bond
donating and accepting regions B and C. The two
lipophilic pockets L2 and L3 could be “filled” with phenyl
groups to give an almost symmetric core, leaving the
pocket L1 to be explored thoroughly. Thus the 2,6-
diphenyl-4-amidopyrimidines were conceived.
Polar Surface Area Calculations. The polar sur-
face areas (PSA) of the molecules were calculated using
Spartan 5.0 for SGI,30 in combination with an in-house
developed application called PolSurf 1.0. [A copy of
PolSurf and its (C) source code can be obtained via the
author.] First Spartan was used to build the molecule,
optimize its 3D structure, and calculate its property
data; subsequently PolSurf was applied to convert the
raw data from the Spartan “input” and “proparc” files
into the polar surface area of the molecule. The resulting
PSA values were compared with data already available
in the literature for calculations based on a single
conformer and based upon van der Waals surface areas
Figure 4. All nine molecules superimposed upon each other
upon the hydrogen-bond donating and accepting regions, as
seen marked in Figure 2.
receptor.32 In 2002, Bondavalli and co-workers33 carried
out some computational work on a collection of eleven
different A1 antagonists using different methods to
achieve a pharmacophore similar to that of Da Settimo
et al. These models, however, were all based on bi- and
tricyclic heteroaromatic cores, fueling new antagonists
which were again bicyclicsthe pyrazolopyridines of
Bondavallisor tricyclicsthe triazinobenzimidazoles by

Evaluation Only. Created with Aspose.PDF. Copyright 2002-2021 Aspose Pty Ltd.
6532 Journal of Medicinal Chemistry, 2004, Vol. 47, No. 26
Chang et al.
Table 1. Affinities and PSA Values of the 2,6-Diphenyl-4-amidopyrimidines 4-20 in Radioligand Binding Assays at the Human
Adenosine Receptors
Ki (nM) or % displacementa
calcd PSA
b
c
d
e
compd
R
values (Å2)
hA1
hA2A
hA2B
hA3
4
Ph
50
50
61
52
51
51
51
51
51
50
51
49
51
52
51
51
51
671 ( 110
35%
17%
0%
0%
489 ( 140
81.9 ( 14
124 ( 17
48%
1300 ( 300
33%
5
4-Cl-Ph
4-MeO-Ph
Me
6
30%
1170 ( 150
6.88 ( 1.9
22.4 ( 8.9
167 ( 51
392 ( 65
267 ( 39
447 ( 62
57%
147 ( 22
49%
39%
25.0 ( 2.4
178 ( 57
170 ( 60
32%
7
37.5 ( 8.1
9.50 ( 4.6
17.6 ( 5.3
109 ( 15
11.1 ( 6.2
14.8 ( 2.7
6.35 ( 0.4
2.22 ( 1.1
27.7 ( 6.2
8.75 ( 4.1
7.58 ( 1.3
6.49 ( 2.2
2.14 ( 0.1
15.5 ( 8.4
8
Et
9
Pr
10
11
12
13
Bu
i-Pr
376 ( 76
157 ( 42
381 ( 33
899 ( 130
18%
CH2CHMe2
CHEt2
CH(Me)(Et)
t-Bu
72%
44%
14 (LUF 5767)
15
16 (LUF 5764)
CH2CMe3
c-Pr
c-Bu
c-Pent
c-Hex
32%
4%
17
18
189 ( 44
179 ( 58
196 ( 66
208 ( 77
74%
52%
19 (LUF 5740)
20
a Ki ( SEM (n ) 3), % displacement (n ) 2). b Displacement of specific [3H]DPCPX binding in CHO cell membranes expressing human
adenosine A1 receptors or % displacement of specific binding at 1 µM concentrations. c Displacement of specific [3H]ZM241385 binding in
d
HEK 293 cell membranes expressing human adenosine A2A receptors or % displacement of specific binding at 1 µM concentrations.
%
Displacement of specific [3H]MRS1754 binding in CHO cell membranes stably transfected with the human adenosine A2B receptor at 1
µM concentrations. e Displacement of specific [125I]AB-MECA binding in HEK 293 cell membranes expressing human adenosine A3 receptors
or % displacement of specific binding at 1 µM concentrations.
as described by Clark.27,28 We observed a highly signifi-
cant correlation for the 75 compounds investigated (R2
) 0.98, see Supporting Information for more details).
Clark compared his calculated PSA values to experi-
mental percentage fractional absorption data27,28 and
the “rule-of-five”.34 In consequence, he concluded that
poor passive intestinal absorption occurred when a
molecule has a PSA value of g140 Å2. Likewise, Kelder
et al.25 published a similar paper stating that orally
active drugs, when transported passively, should not
exceed a PSA of 120 Å2. Furthermore, the Kelder
investigation also studied BBB penetration and con-
cluded that, by decreasing the PSA to 60-70 Å2, brain
penetration could be achieved. As in our calculations,
both parties use van der Waal surface areas, and only
consider the heteroatoms N and O and any hydrogens
that may be attached to them. We therefore propose the
significant value of 60-70 Å2 as the limit for PSA for
brain penetration as suggested by Kelder et al.25 as also
relevant in this paper.
The PSA values of our initial series (the 2,6-diphenyl-
4-amidopyrimidines) were calculated to be under the set
limit, with a range of 49 (R ) tBu) to 61 (R ) 4-MeOPh)
Å2. With reference to this series, a second set of ligands
(4,6-diphenyl-2-amidopyrimidines) was designed with
the pharmacophore and PSA limits in mind. Again, the
hydrogen-bond accepting and donating regions are
fulfilled by the nitrogen in the pyrimidine ring and the
amido hydrogen, and again the carbonyl oxygen ac-
counts for the hydrogen-bond accepting region at the
“top” of the space. With only three nitrogen atoms, one
oxygen atom, and one polar hydrogen atom, the PSA
values (Table 2) were calculated to be within the
proposed 60-70 Å2 limit. The two lipophilic areas
labeled L2 and L3 were “filled” with aromatic phenyl
groups and the third, labeled L1, was again left to be
explored.
Chemistry. The synthetic route to the 2,6-diphenyl-
4-amidopyrimidines (4-20) is depicted in Scheme 1. The
pyrimidinone ring was created by the reaction of the
commercially available â-ketoester, ethyl benzoate, with
benzamidine hydrochloride in the presence of a base to
give the pyrimidinone ring (1). The first attempts to
achieve the condensation using sodium ethoxide ren-
dered only poor yields; changing the base to sodium
hydroxide in accordance to a script by de Valk and van
der Plas35 improved the reaction significantly resulting,
in a yield of 59%. Displacement of the hydroxide
function was achieved with an excess of phosphoryl
chloride containing phosphorous pentachloride, accord-
ing to a preparation by Brown et al.36 Subsequent
substitution with ammonia in a sealed vessel gave the
key 4-amino-2,6-diphenyl pyrimidinyl intermediate (3)
in very good yields. The final step to give compounds
4-20 involved the formation of the amide bond. At-
tempts using the carboxylic acid with standard coupling
agents, for example 1-(3-dimethylaminopropyl)-3-eth-
ylcarbodiimide hydrocloride (EDC) and 1-hydroxyben-
zotriazole (HOBt), gave poor or no yields. As an alter-
native the amine was reacted with acid chlorides in the
presence of triethylamine to give, in general, good to
excellent yields. The pyrimidines 22-40 were synthe-
sized according to the route depicted in Scheme 2.
Guanidine was freed from its hydrochloride salt form
with NaOH(aq) and reacted with benzylideneacetophe-
none according to a method described by Al-Hajjar and