Substituted 4-phenylthiazoles: Development of potent and selective A1, A3 and dual A1/A3 adenosine receptor antagonists

Article abstract of DOI:10.1016/j.ejmech.2019.111879

Adenosine acts as a powerful signaling molecule via four distinct G protein-coupled receptors, designated A₁, A_2A, A_2B and A₃ adenosine receptors (ARs). A_2A and A_2B ARs are G_s-coupled, while A₁ and A₃ ARs inhibit cAMP production via G_i proteins. Antagonists for A₁ and A₃ ARs may be useful for the treatment of (neuro)inflammatory diseases including acute kidney injury and kidney failure, pulmonary diseases, and Alzheimer's disease. In the present study, we optimized the versatile 2-amino-4-phenylthiazole scaffold by introducing substituents at N2 and C5 to obtain A₁ and A₃ AR antagonists including dual-target compounds. Selective A₁ antagonists with (sub)nanomolar potency were produced, e.g. 11 and 13. These compounds showed species differences being significantly more potent at the rat as compared to the human A₁ AR, and were characterized as inverse agonists. Several potent and selective A₃ AR antagonists, e.g. 7, 8, 17 and 22 (K_i values of 5–9 nM at the human A₃ AR) were prepared, which were much less potent at the rat orthologue. Moreover, dual A₁/A₃ antagonists (10, 18) were developed showing K_i values between 8 and 42 nM. Docking and molecule dynamic simulation studies using the crystal structure of the A₁ AR and a homology model of the A₃ AR were performed to rationalize the observed structure-activity relationships.

Full text of DOI:10.1016/j.ejmech.2019.111879

Journal Pre-proof
Substituted 4-phenylthiazoles: Development of potent and selective A A and dual
1, 3
A /A adenosine receptor antagonists
1
3
Aliaa Abdelrahman, Swapnil G. Yerande, Vigneshwaran Namasivayam, Tim A.
Klapschinski, Mohamad Wessam Alnouri, Ali El-Tayeb, Christa E. Müller
PII:
S0223-5234(19)31031-1
DOI:
https://doi.org/10.1016/j.ejmech.2019.111879
EJMECH 111879
Reference:
To appear in: European Journal of Medicinal Chemistry
Received Date: 30 July 2019
Revised Date: 10 November 2019
Accepted Date: 11 November 2019
Please cite this article as: A. Abdelrahman, S.G. Yerande, V. Namasivayam, T.A. Klapschinski, M.W.
Alnouri, A. El-Tayeb, C.E. Müller, Substituted 4-phenylthiazoles: Development of potent and selective
A
A and dual A /A adenosine receptor antagonists, European Journal of Medicinal Chemistry
3 1 3
1,
(2019), doi: https://doi.org/10.1016/j.ejmech.2019.111879.
This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition
of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of
record. This version will undergo additional copyediting, typesetting and review before it is published
in its final form, but we are providing this version to give early visibility of the article. Please note that,
during the production process, errors may be discovered which could affect the content, and all legal
disclaimers that apply to the journal pertain.
©
2019 Published by Elsevier Masson SAS.

1
Substituted 4-phenylthiazoles: development of potent
and selective A A and dual A /A adenosine
1,
3
1
3
receptor antagonists
a,b,&
c,d,&
a,b
Aliaa Abdelrahman,
Swapnil G. Yerande,
Vigneshwaran Namasivayam, Tim A.
a,b
a
a,b
a,b,
Klapschinski, Mohamad Wessam Alnouri, Ali El-Tayeb, and Christa E. Müller *
a
Pharmaceutical Institute, Department of Pharmaceutical & Medicinal Chemistry, University
of Bonn, An der Immenburg 4, D-53121 Bonn, Germany.
b
Pharma Center Bonn, University of Bonn, Brühlerstr. 7, D-53121 Bonn, Germany
c
Department of Medicinal Chemistry, B. V. Patel Pharmaceutical Education and Research
Development (PERD) Centre, Ahmedabad 380 054, Gujarat, India.
d
present address: Deepak Nitrite Ltd., Vadodara, Gujarat, India
&
Authors contributed equally
christa.mueller@uni-bonn.de
*
Address correspondence to:
Prof. Dr. Christa Müller
Pharmazeutisches Institut
Pharmazeutische & Medizinische Chemie
An der Immenburg 4, D-53121 Bonn, Germany
Phone: +49-228-73-2301; Fax: +49-228-73-2567

2
Abstract
Adenosine acts as a powerful signaling molecule via four distinct G protein-coupled
receptors, designated A , A , A and A adenosine receptors (ARs). A and A2B ARs are
1
2A
2B
3
2A
G -coupled, while A and A ARs inhibit cAMP production via G proteins. Antagonists for
s
1
3
i
A and A ARs may be useful for the treatment of (neuro)inflammatory diseases including
1
3
acute kidney injury and kidney failure, pulmonary diseases, and Alzheimer’s disease. In the
present study, we optimized the versatile 2-amino-4-phenylthiazole scaffold by introducing
substituents at N2 and C5 to obtain A and A AR antagonists including dual-target
1
3
compounds. Selective A antagonists with (sub)nanomolar potency were produced, e.g. 11
1
and 13. These compounds showed species differences being significantly more potent at the
rat as compared to the human A AR, and were characterized as inverse agonists. Several
1
potent and selective A AR antagonists, e.g. 7, 8, 17 and 22 (K values of 5-9 nM at the human
3
i
A AR) were prepared, which were much less potent at the rat orthologue. Moreover, dual
3
A /A antagonists (10, 18) were developed showing K values between 8-42 nM. Docking and
1
3
i
molecule dynamic simulation studies using the crystal structure of the A AR and a homology
1
model of the A AR were performed to rationalize the observed structure-activity
3
relationships.
Keywords
Adenosine receptors; A adenosine receptor; A adenosine receptor antagonist; allosteric
1
3
modulator; Alzheimer’s disease; inverse agonist; multi-target drugs; selectivity; species
differences; structure-activity relationships; synthesis; thiazole
Abbreviations
ADA, adenosine deaminase; AR, adenosine receptor(s); BSA, bovine serum albumin; CADO,
3
3
6
2
-chloroadenosine; [ H]CCPA, [ H]2-chloro-N -cyclopentyladenosine; CHO, Chinese

3
6
hamster ovary; CPA, N -cyclopentyladenosine; cyclic adenosine monophosphate (cAMP);
DMSO, dimethylsulfoxide; DPCPX, 8-cyclopentyl-1,3-dipropylxanthine; GTP, guanosine-5'-
3
3
triphosphate; h, human; m, mouse; [ H]MSX-2, [ H](E)-3-(3-hydroxypropyl)-8-(2-(3-
methoxyphenyl)vinyl)-7-methyl-1-prop-2-ynyl-3,7-dihydropurine-2,6-dione, NECA, 5'-(N-
6
ethylcarboxamido)adenosine; R-PIA, (R)-N -phenylisopropyladenosine; PSA, polar surface
3
3
area;
[ H]PSB-11,
[ H]2-phenyl-8-ethyl-4-methyl-(8R)-4,5,7,8-tetrahydro-1H-
3
3
imidazo[2.1-i]purin-5-one;
[ H]PSB-603,
[ H]8-(4-(4-(4-chlorophenyl)piperazine-1-
sulfonyl)phenyl)-1-propylxanthine; r, rat; TRIS, tris(hydroxy-methyl)aminomethane.
Introduction
Extracellular adenosine acts at G protein-coupled cell membrane receptors (GPCRs), the so-
called purine P1 or adenosine receptors (ARs), which are divided into four subtypes: A , A ,
1
2A
A2B and A [1]. A and A ARs are coupled to G proteins which mediate inhibition of
3
1
3
i/o
adenylate cyclase thereby reducing cAMP production, while A2A and A2B ARs stimulate the
production of cAMP via G proteins [1, 2]. A ARs can be additionally coupled to Gq
s
2B
proteins leading to phospholipase C activation and subsequent calcium mobilization [3]. ARs
are widely distributed in the body, and elevated adenosine levels and/or receptor upregulation
have been observed under many pathological conditions [4]. The A AR is found in high
1
density in the brain, in cortex, hippocampus, and cerebellum, mediating neurotransmitter
release from neurons; moreover, it is expressed on astrocytes, oligodendrocytes, and
microglia [5]. In the periphery, A ARs are present in many organs and tissues including heart
1
and kidneys [6]. A ARs have been detected in various organs such as lung and liver [7]; in
3
addition, they are expressed on immune cells and on various cancer cells [7,8].
Both, A and A AR antagonists, may be useful and might even show synergistic effects for
1
3
the treatment of important diseases including (i) acute kidney injury and kidney failure [2],

4
(ii) inflammatory pulmonary disease, asthma and allergy [9], and (iii) Alzheimer’s disease
[
10-12]. Importantly, both receptors target the same, G -mediated pathway.
i
There is recent evidence that A AR antagonists reduce tau aggregation [10], while A AR
1
3
antagonists protect from amyloid generation [11]. A antagonists are known to improve
1
cognition, and both, A and A antagonists were found to exert neuroprotective properties in
1
3
preclinical studies [12].
Selective A and A AR antagonists have been developed, and selective A AR antagonists
1
3
1
were clinically evaluated [13]. A phase III clinical trial of the selective A antagonist
1
rolofylline for the treatment of congestive heart failure had shown excellent tolerability of the
drug, but lack of efficacy [14]. A recent stratified re-evaluation of the data showed that
rolofylline could be harmful in low-risk acute heart failure patients, while it might be
significantly beneficial for higher risk patients [15]. A ARs antagonists may be useful for the
1
treatment of cognitive dysfunction, as observed in Alzheimer’s disease (AD), due to their
CNS stimulatory effects. A recent study indicates that rolofylline is an interesting candidate to
combat the hypometabolism and neuronal dysfunction associated with tau-induced
neurodegenerative diseases [10].
Selective A AR antagonists have not entered into clinical trials so far. One reason may be the
3
large species differences between human and rodent A ARs. The development of antagonists
3
that are equally potent in both species has so far only met with limited success [2,8,16].
It is known that the treatment of complex diseases requires multifaceted approaches instead of
targeting a single pathway [17]. Multi-target drugs are considered safer than drug
combinations since they have lower toxicities and a lower risk of drug-drug interactions [18].
Moreover, the combination of several drugs often reduces patient compliance [18]. For these
reasons, the development of dual- or multi-target drugs can offer significant advantages [19].

5
A dual A /A AR antagonist was designed as an anti-asthmatic agent [20], while dual A
2A
2
B
3
agonists/A AR antagonists have been developed for the treatment of asthma and other
3
inflammatory diseases [21,22]. A /A dual AR antagonists were previously synthesized with
1
2A
the goal to treat neurodegenerative diseases including Parkinson’s [23].
No pharmacological data on dual A /A AR ligands have been published, but a combination
1
3
of an A AR agonist with an A AR antagonist was claimed in a patent for the treatment of
1
3
glaucoma [24]. Recently, we obtained
a
dual A /A
AR antagonist, 5-
1
3
diphenyl[1,2,4]triazolo[1,5-c]quinazoline (K human A AR 51.6 nM, human A AR 11.1
i
1
3
nM), within a series of antagonists targeted towards the A AR [25].
2
A
The naturally occurring xanthine derivatives caffeine (I) and theophylline (II) are weak,
nonselective AR antagonists (Fig.1) [1,26]. Caffeine (I) has been used as a drug for various
indications including central stimulation to increase alertness and wakefulness, in
combination with analgesics to treat pain, and for the treatment of apnea in premature babies
[
2,27]. Modification of the xanthine structure, in particular by substitution at the 8-position,
has led to derivatives provided with high affinity and subtype-selectivity. For example, PSB-
6 (III) is one of the most potent A AR antagonists with high affinity for human, rat and
3
1
mouse ARs, and high selectivity for the A AR against the other AR subtypes [28]. Very
1
recently X-ray structures of the A and the A AR in complex with PSB-36 have been
1
2A
published [29]. The 1,3-dipropyl-8-(3-noradamantyl)xanthine, rolofylline (IV), is another
potent and selective adenosine A AR antagonist [30]. PSB-10 (Va) and PSB-11 (Vb), see
1
Fig. 1, are tricyclic xanthine derivatives, that show high affinity and selectivity for the human
A AR [31]. The pyridine derivative MRS1523 (VI) was the first A AR antagonist that
3
3
displayed considerable potency of the rat A AR [32].
3

6
Fig. 1. Structures and affinities of selected AR antagonists. (h = human, r = rat)

7
Thiazole derivatives, such as 7-imino-2-thioxothiazolo[4,5-d]pyrimidines (e.g. VII), related
thiazolotriazolopyrimidinethiones (e.g., VIII) and 4-benzylideneamino-2,3-dihydro-2-
thioxothiazole-5-carbonitrile derivatives, e.g. compound IX (Fig. 2), have been reported as
potent A -selective AR antagonists [33-35]. The benzothiazole derivative tozadenant (SYN-
2
A
1
15, X) was developed as a potent and selective A AR antagonist, and evaluated in phase III
2A
clinical trial for the treatment of Parkinson’s disease, but has been discontinued due to
toxicity [2,36]. We previously showed that substitution of the 2-amino-4-phenylthiazole
scaffold with a benzoyl moiety at the amino group and an aroyl moiety at the 5-position
shifted the selectivity profile towards the A AR; the most potent compound in this series was
1
XI (Fig. 2) [37]. Recently, furanoylamide derivatives of 4,5-disubstituted 2-aminothiazoles
were synthesized as adenosine receptor antagonists, e.g. compound XII, which showed a high
affinity for ARs in the low nanomolar range but lacked subtype-selectivity (Fig. 2) [38].
Thiazole derivatives were also found to be promising scaffolds for A AR antagonists, for
3
example, compound XIII displayed a subnanomolar affinity for the human A AR combined
3
with high selectivity versus the other AR subtypes (Fig. 2) [39-40].

8
Fig. 2. Structures and affinities of selected thiazole derivatives as AR antagonists and general
structures of the designed compounds.

9
In the present study, we optimized the versatile 2-amino-4-phenylthiazole scaffold by various
modifications at the amino group and at the 5-position (for structures of the new compounds
see Fig. 2). Our goal was to obtain antagonists with high affinity for A and/or A ARs,
1
3
preferably acting as dual-target drugs, due to their therapeutic potential for the treatment of
kidney failure, inflammatory pulmonary diseases, and Alzheimer’s disease.
Results and Discussion
Chemistry
Thiazole derivatives were prepared as described in Scheme 1. 4-Methoxybenzoyl chloride
(1b), benzoyl chloride (1a), or furan-2-carbonyl chloride (1c) were reacted with ammonium
thiocyanate in acetone to afford the intermediate isothiocyanates [41], which were treated
with N,N-diethylbenzimidamide in acetone to give 2a-c (Scheme 1) in good yields (65-80%)
[
42]. N,N-Diethylbenzimidamide was obtained by reaction of benzonitrile with diethylamine
in the presence of aluminium chloride as previously described [43] (for details see Supporting
Information (SI)). 2-Chloromethylquinazoline-4-one derivatives used for the synthesis of
thiazoles 6-9, 15-17 and 20-23 were prepared by reaction of anthranilic acid with chloroacetyl
chloride in dichloromethane in the presence of triethylamine yielding 2-(2-
chloroacetylamino)benzoic acid in 64% yield [44]. Activation of 2-(2-
chloroacetylamino)benzoic acid with phosphorus oxychloride in toluene yielding the acid
chlorides followed by cyclization with aniline or its derivatives afforded the desired
2
-chloromethylquinazoline-4-ones [43]. Phenacyl bromide derivatives used for the synthesis
of thiazoles 3, 5, 10-14, and 18, 19 were obtained by reaction of the corresponding
acetophenones with bromine in glacial acetic acid in the presence of a catalytic amount of

1
0
aqueous hydrobromic acid [38]. 2-(Bromomethyl)pyridine employed in the synthesis of
thiazole 4 was obtained from commercial sources. Reaction of 2a-c with the appropriate
phenacyl bromide derivatives, 2-chloromethylquinazoline-4-one derivatives, or
bromopyridine in acetonitrile or methanol, respectively, led to the formation of thiazole
derivatives 3-23 in acceptable to high yields ranging from 28% to 96% (Scheme 1) [38,45].

1
1
a
Scheme 1. Synthesis of thiazole derivatives.
1
2
Compound
R
R
1
a, 2a,
3
-9
F
1
b, 2b,
0-17
Cl
OCH3
CH3
F
O
O
O
O
O
1
10
11
12
13
14
N
N
N
O
N
N
N
O
O
H C
3
H CO
3
1
5
16
17
1
c, 2c,
8-23
1
a
Reagents, conditions: (a) NH NCS, acetone, reflux, 5 min; (b) N,N-diethylbenzimidamide,
4
2
acetone, 0 °C; (c) R -CH Br, acetonitrile or methanol, rt, 1 h.
2
Thiazoles 5-10 and 20-23 had previously been described but they had not been biologically
evaluated [46]. Those compounds were now resynthesized and investigated. All other
synthesized products (3-4 and 11-19) are new and not previously described in the literature.
Thiazole derivatives 3-23 precipitated from the reaction medium as pure products, and
chromatographic purification was therefore not required. The structures of the synthesized

1
2
1
13
compounds were confirmed by H- and C-NMR spectroscopy and high resolution mass
spectrometry (HRMS). In addition, IR spectra were recorded. Purity of the final compounds
was determined by HPLC-ESI-UV/MS analysis and found to be > 95% in all cases (see
Experimental Section and SI).
Biological investigations
Receptor-radioligand binding studies at human and rat A , A , A and A ARs were
1
2A
2B
3
performed as previously described [47-51]. Initially, the 4-phenylthiazoles were tested at a
single concentration of 10 µM, 1 µM or 100 nM, respectively, depending on their solubility.
For compounds that inhibited radioligand binding by more than 50% full concentration-
inhibition curves were recorded and K values were determined. All results are collected in
i
Table 1 together with data for standard compound XI [37] which we tested in the same test
system. For selected compounds, cAMP accumulation assays in Chinese hamster ovary
(
CHO) cells recombinantly expressing the human A AR were performed (see Fig. 6).
3
Receptor functionality at the A AR was determined by GTP shift assays, in which affinities
1
of selected compounds for the A AR were determined in the absence and presence of 100 µM
1
GTP [37] (see Fig. 7 and Table 2). In addition, affinities for the A AR were determined vs.
1
3
3
the antagonist radioligand [ H]DPCPX as well as vs. the agonist radioligand [ H]CCPA (see
Table 3). Agonists will show significantly higher affinity versus an agonist radioligand than
versus an antagonist radioligand, while neutral antagonists will display similar affinities
versus both, agonist and antagonist radioligands [49]. Antagonists with inverse agonistic
activity, on the other hand, will show higher affinity when measured versus an antagonist than
versus an agonist radioligand.

1
3
Table 1. Affinities of thiazole derivatives at adenosine receptor subtypes.
2
R
Compd.
K ± SEM (nM) at human (h) or rat (r) adenosine receptors
i
(or % inhibition of radioligand binding at indicated concentration)
A1
A2A
vs. [³H]MSX-2
A2B
A3
3
vs. [ H]CCPA
vs.[³H]PSB- vs.[³H]PSB-11(h)
6
03
or
vs.[³H]NECA (r)
2160 ± 880 (h)
XI [37]
57.4 ± 8.3 (h)
6250 ± 1970 (h)
>1000 (h)
-
4
.83 ± 1.06 (r)
>1,000 (r)
1
Thiazole derivatives 3-9: R = phenyl
a
c
a,f
3
55.3 ± 22.5 (h)
77.7 ± 15.6 (h)
>100
67.8 ± 2.7 (h)
b
d
a,e
2
.79 ± 0.09 (r)
≥1000 (43 %) (r)
(15 %) (h)
a
c
4
5
6
90.8 ± 2.5 (h)
83.6 ± 15.5 (h)
≥100
(33 %) (h)
>100
≥1000
b
d
a,e
a,f
3
.47 ± 0.71 (r)
391 ± 178 (r)
(45 %) (h)
a
c
a,f
24.4 ± 4.1 (h)
39.1 ± 2.4 (h)
428 ± 53 (h)
b
b
a,e
0
.939 ± 0.486 (r)
1670 ± 550 (r)
(0 %) (h)
>100
a
c
a,f
>1000 (0 %) (h)
>1000 (0 %) (h)
111 ± 7 (h)
b
d
a,e
>1000 (1 %) (r)
>1000 (3 %) (r)
(0 %) (h)
a,f
7
>1000
>1000
>100
9.05 ± 1.20 (h)
N
O
a
c
a,e
g,h
(
0 %) (h)
(0 %) (h)
(0) (h)
806 ± 178 (r)
N
>1000
>1000
H C
3
b
d
(
23%) (r)
(17 %) (r)

1
4
a,f
8
9
>1000
>1000
(3 %) (h)
>1000
>100
8.23 ± 0.89 (h)
a
c
a,e
g,h
(
6 %) (h)
(0 %) (h)
1320 ± 600 (r)
>1000
b
d
(
10 %) (r)
>1000
(17 %) (r)
>1000
a,f
>100
83.3 ± 2.5 (h)
a
c
a,e
g,h
(4 %) (h)
(0 %) (h)
(12 %) (h)
>10000 (r)
>1000
>1000
(22 %)
b
d
(16 %) (r)
(7 %) (r)
1
Thiazole derivatives 10-17: R = p-methoxyphenyl
a,f
1
0
1
2
3
4
5
≥1000
(41 %) (h)
≥1000
>100
25.4 ± 5.2 (h)
a
c
a,e
g,h
3
6.7 ± 5.1 (h)
(2 %) (h)
>1,000 (r)
b
1
.47 ± 0.16 (r)
(0 %)
d
(32 %) (r)
a,i
b
1
1
1
1
1
32.3 ± 8.3 (h)
>1000
>100
(0 %) (h)
>100
>1000
c
a,e
a,f
1
.01 ± 0.33 (r)
(0 %) (h)
1000
(3 ± 4) (h)
>
>10000
d
j,e
g,h
(12 %) (r)
(0 %) (r)
>100
(0 %) (r)
a
13.6 ± 4.6 (h)
>1000
>1000
b
c
a,e
a,f
1
.77 ± 0.88 (r)
(0 %) (h)
1000
(0 %) (h)
(37 %) (h)
>
d
c
(
16 %) (r)
a
9.69 ± 4.81 (h)
>1000
>100
(2 %) (h)
>100
≥1000
b
a,e
a,f
0
.529 ± 0.043 (r)
(28 %) (h)
1000
(45 %) (h)
>
>10000
d
j,e
g,h
(15 % ) (r)
(7 %) (r)
>100
(13 %) (r)
>1000
>1000
>1000
a
c
a,e
a,f
(9 %) (h)
(14 %) (h)
>1000
(0 %) (h)
(24 %) (h)
ca. 1000
54 %) (r)
>1000
b
d
(
(16 %) (r)
>1000
a,g
>100
174 ± 19 (h)
a
c
a,e
(
0 %) (h)
(20 %) (h)
>1000
(0 ± 17) (h)
>1000

1
5
b
b
b
d
c
(
11 %) (r)
(1 %) (r)
>1000
a,f
1
1
6
7
>1000
>100
27.8 ± 8.6 (h)
a
a,e
(
0 %) (h)
(0 %) (h)
>1000
(2 %) (h)
>10000
g,h
≥
1000
(15 %) (r)
d
(25 %) (r)
(15 %) (r)
>1000
a,f
>1000
>100
4.63 ± 1.64 (h)
a
c
a
(
3 %) (h)
1000
17 %) (r)
(10 %) (h)
>1000
(0 %) (h)
>10000 (0 %)
g,h
>
(r)
d
(
(1 %) (r)
1
Thiazole derivatives 18-23: R = 2-Furyl
a
a,f
1
1
8
9
7.57 ± 0.71 (h)
>100
42.2 ± 18.5 (h)
b
d
a,e
0
.348 ± 0.134 (r)
1210 ± 709 (r)
(2 %) (h)
a
c,h
178 ± 14 (h)
132 ± 25 (h)
>100
124 ± 2
b
d
a,e
a,f
6
.69 ± 1.35 (r)
73.3 ± 6.6 (r)
(15 %) (h)
(100 %) (h)
a
c
a,f
2
2
0
1
>1000 (1 %) (h)
>1000 (5 %) (h)
>100
129 ± 26 (h)
b
a,e
≥
1000 (26 %) (r)
>1000 (33 %)
(1 %) (h)
d
(r)
a,f
>1000
>1000
(1 %) (h)
>1000
>100
25.4 ± 5.3 (h)
1840 ± 380 (r)
a
c
a
g,h
(2 %) (h)
(0 %) (h)
≥
1000
b
d
(
25 %) (r)
(6 %) (r)
>1000
a,f
2
2
2
3
>1000
≥100
6.45 ± 5.34 (h)
a
c
a,e
g,h
(
7 %) (h)
(11 %) (h)
(30 %) (h)
973 ± 148 (r)
≥
1000
>1000
b
d
(
25 %) (r)
(0 %) (r)
a,f
g,h
>1000
>1000
>100
48.3 ± 2.3 (h)
a
c
a,e
(
0 %) (h)
(8 %) (h)
>1000
(1 %) (h)
607 ± 48 (r)
≥
1000
b
d
(24 %) (r)
(14 %) (r)
a
human recombinant receptors expressed in CHO
b
c
A : rat brain cortex (for A )
1
1
human recombinant receptor expressed in HEK293 cells
d
A : rat brain striatum (for A )
2
A
2A
e
f
inhibition at 0.1 µM
vs.[³H]PSB-11

1
6
g
h
i
vs.[³H]NECA
rat recombinant receptor expressed in CHO cells
Extrapolated K value (due to limited solubility)
i
Structure-activity relationships
The 2-amino-4-phenylthiazole core structure was modified at the amino group by amide
formation with benzoic acid, p-methylbenzoic acid, or 2-furancarboxylic acid, respectively.
These residues were selected to occupy the lipophilic, aromatic binding pocket. They were
combined with different substituents in the 5-position of the 2-amino-4-phenylthiazole core,
in particular differently substituted benzoyl residues and a 3-arylquinazolin-4-one moiety.
The latter bulky substituent had been found in preliminary studies to confer A AR affinity
3
and selectivity, while the smaller benzoyl substitution appeared to be well tolerated by the A1
AR. The obtained series of 4-phenylthiazole derivatives 3-23 was investigated in radioligand
binding studies at the four human AR subtypes. Our goal was to enhance A and/or A AR
1
3
affinity, and to obtain antagonists with selectivity for one of both subtypes, as well as dual
A /A AR antagonists. Test results are displayed in Table 1.
1
3
In the present series, the introduction of phenyl (see results of compounds 3-9),
p-methoxyphenyl (see results of compounds 10-17) or furanyl moieties (see results of
compounds 18-23) attached to the amide linker in the 2-position of the thiazole core was
generally well tolerated. At the human A AR, higher affinity was obtained with compounds
1
1
bearing relatively small residues attached to the position 5 of the thiazole core (see R in
Scheme 1 and Table 1). At human A ARs several compounds showed high affinity in the low
1
nanomolar concentration range. The 2,4-difluorobenzoyl derivative, compound 3, and the 2-
pyridyl derivative 4 displayed K values of 55.3 and 90.8 nM, respectively. The best results in
i
the present series were obtained with benzoyl- (10, K 36.7 nM), p-chlorobenzoyl- (5, K 24.4
i
i
nM; 11, K 32.3 nM), p-methoxybenzoyl- (12, 13.6 nM), and p-methylbenzoyl-substituted
i
thiazole derivatives (13, K 9.69 nM; 18, K 7.57 nM) (see Table 1), compounds 11, 12 and 13
i
i

1
7
displaying selectivity versus the other AR subtypes (Table 1 and Fig. 3). All tested
compounds were weaker at the human A AR as compared to the rat A AR (e.g. 11: 32-fold,
1
1
1
2: 8-fold, 13: 18-fold difference).
Fig. 3. Concentration-inhibition curves of selected compounds at (A) the human A AR
1
recombinantly expressed in CHO cells, and (B) at rat cortical membrane preparations;
3
[
H]CCPA ( 0.5 nM ) was used as a radioligand. Data points are means ± SEM of three
separate experiments.
In contrast to the human A AR, higher potency for the human A ARs was obtained with
1
3
compounds bearing bulky residues attached to the 5-position of the thiazole core (see for
example compounds 7, 8, 17, 21, and 22, see Table 1). These compounds displayed K values
i
for the human A AR in the low nanomolar range and were highly selective versus the other
3
AR subtypes. The most potent A AR antagonist at the human receptor was 17 (K 4.63 nM)
3
i
The antagonists exhibited significant species differences between human and rat A ARs. For
3
example, compounds 7, 21 and 22 showed 72 to 150-fold higher affinities for the human as
compared to the rat A AR (Table 1). The 3-(p-substituted phenyl)-2-(thiazol-5-yl)quinazolin-
3
4
(3H)-one moiety appeared to be well tolerated by the rat A AR, e.g. compounds 7 (p-
3
methyl), 22 (p-methoxy) and 23 (p-chloro) showing submicromolar K values at the rat A3
i
AR. Compound 23 was the most potent antagonist of the present series at the rat A AR (Ki
3
6
07 nM), but it was about 13-fold more potent at the human A AR (48.3 nM, see Table 1 and
3

1
8
Fig. 4). In contrast, compounds 10, 16, and 17 displayed high affinity for the human A AR in
3
the nanomolar range but no affinity was observed for the rat A AR even at high
3
concentrations.
Fig. 4. Concentration-inhibition curves of selected compounds at human A ARs
3
3
recombinantly expressed in CHO cells. [ H]PSB-11 (0.5 nM) was used as a radioligand. Data
points are means ± SEM of three separate experiments.
A combination of 4-methylbenzoyl in position 5 with a furan-2-carboxamide in position 2 of
the thiazole core resulting in compound 18 showed the highest dual affinity for the human A1
and A ARs (K A 7.57 nM, A 42.2 nM). N-(5-Benzoyl-4-phenylthiazol-2-yl)-4-
3
i
1
3
methoxybenzamide (10) displayed almost identical affinity for the human A AR and the
1
human A AR with K values of 36.7 nM and 25.4 nM, respectively. This compound showed
3
i
no affinity for the other AR subtypes. Both compounds, 10 and 18, are potent dual A /A AR
1
3
antagonists (see Fig. 5 and Table 1).

1
9
Fig. 5. Concentration-inhibition curves of compounds 10 and 18 tested at (A) the human A1
3
AR recombinantly expressed in CHO cells vs. [ H]CCPA (0.5 nM), and (B) the human A AR
3
3
recombinantly expressed in CHO cells vs. [ H]PSB-11 (0.5 nM). Data points represent means
±
SEM from three separate experiments.
In
the
studied
series
only
N-(5-(2,4-difluorobenzoyl)-4-phenylthiazol-2-yl)-4-
methoxybenzamide (14) was inactive at all four AR subtypes. Out of the 21 tested compounds
four showed affinity for the human A AR with K values ranging between 39.1 and 132 nM
2
A
i
(compounds 3, 4, 5 and 19). All other tested compounds were inactive at concentrations up to
1
µM. At the rat A ARs, the compounds showed even lower potency. Only compound 4
2
A
exhibited a measurable K value of 391 nM at the rat A AR (Table 1). All tested compounds
i
2A
were inactive at the human A AR. Selected compounds 11 and 13 tested at the rat A AR
2
B
2B
also showed no affinity at that receptor.
cAMP accumulation assays
In order to confirm that the compounds are antagonists, we investigated compound 23 in
cAMP accumulation assays at A ARs. A concentration-dependent inhibition curve for the
3
agonist NECA was determined in the absence and in the presence of 23 at rat A ARs
3
expressed in CHO cells. Compound 23 (3 µM and 10 µM) shifted of the concentration-

2
0
response curve of the agonist NECA to the right. Thus, 23 can be considered as a competitive
A AR antagonist.
3
The EC values for the agonist NECA were 5.28 nM without agonist, and 56.9 nM and 1720
5
0
nM, respectively, in the presence of 3 µM and 10 µM of 23 (Fig. 6). The calculated K value
B
for compound 23 was 314 nM and thus similar to its K value determined in binding studies.
i
NECA effect
EC (nM)
% decrease in forskolin-
stimulated cAMP
accumulation
5
0
without inhibitor
5.28 ± 3.43
56.9 ± 13.3
1720 ± 1070
66 ± 1
3
µM 23
75 ± 17
95 ± 6
1
0 µM 23
Fig. 6. Inhibition of forskolin (10 µM) induced cAMP production by NECA in CHO cells
stably transfected with the rat A AR, measured in the absence and presence of different
3
concentrations (3 µM and 10 µM) of compound 23. Data were normalized to forskolin
(
100%) and buffer (0%). NECA without inhibitor showed an EC value of 5.28 ± 3.43 with
50
about 66% maximal inhibition of cAMP accumulation. NECA showed an EC values of 56.9
50
±
13.3 and 1723 ± 1067 with maximal inhibition of cAMP accumulation of 75% and 95%
upon addition of 3 µM and 10 µM of 23. The calculated K value for compound 23 was 314 ±
B

2
1
9
2 nM (n=3). At the highest tested concentration of 10 µM of 23, cAMP production was more
strongly inhibited than by NECA alone possibly due to off-target effects.
GTP shift experiments
GTP shift experiments were performed for compounds 3, 5, 13 and 18 in order to determine
whether they behave as agonists or antagonists at the A AR. In general, GTP causes an
1
uncoupling of the A AR from the G protein leading to a shift of the receptor from the high-
1
i
6
to the low-affinity state for agonists such as N -cyclopentyladenosine (CPA) [37]. In this
experiment, the affinities of compounds were determined against the antagonist radioligand
3
[
H]DPCPX in the absence and in the presence of 100 µM GTP using rat brain cortical
membrane preparations as previously described [37]. For the studied compounds 3, 5, 13 and
8 no significant shift of the binding curves in the absence or the presence of GTP was
1
observed (Table 2). In contrast, the full agonist CPA showed a significant rightward shift of
the binding curve of 8.3-fold in the presence of GTP (Fig. 7). Therefore, compounds 3, 5, 13
and 18 are clearly characterized as antagonists at A ARs. Due to the structural similarity of
1
all compounds in this series, we suppose that they all act as antagonists.
Table 2. Affinities and GTP shifts of selected compounds in comparison with the full agonist
6
N -cyclopentyladenosine (CPA).
a
Compound
IC ± SEM (nM) (n = 3)
5
0
b
-
GTP
+ 100 µM GTP
GTP shift
8.3 ± 3.6
0.9 ± 0.1
1.0 ± 0.2
1.0 ± 0.1
1.3 ± 0.3
CPA
64.7 ± 16.2
3.15 ± 2.48
537.6 ± 81.5
3
5
3.14 ± 2.59
0.0981 ±0.036
0.0763 ± 0.006
0.084 ± 0.028
0.0929 ± 0.019
0.0756 ± 0.009
0.067 ± 0.024
1
3
8
1
a
3
Displacement of [ H]DPCPX from rat cortical membranes.

2
2
b
The affinities for the A AR were determined in the absence and presence of 100 µM GTP.
1
The GTP shifts were calculated by dividing the IC values determined in the presence of
5
0
GTP by those measured in the absence of GTP.
Fig. 7. Concentration–inhibition curves of the agonist CPA and compound 13 in the absence
and presence of 100 µM GTP determined at rat A ARs in brain cortical membrane
1
3
preparations. [ H]DPCPX (0.4 nM) was used as a radioligand.
The K value determined for compound 3 at the rat A AR vs. the antagonist radioligand
i
1
3
3
[
H]DPCPX as compared to the agonist radioligand [ H]CCPA was almost identical (2.93 nM
and 2.79 nM, see Table 3). However, antagonists 5, 13 and 18 showed about 10-fold higher
3
affinity for A ARs using the antagonist [ H]DPCPX as a radioligand than vs. the agonist
1
3
radioligand [ H]CCPA (Table 3). DPCPX is an antagonist with inverse agonistic activity [52-
5
4] which means that it prefers and stabilizes an inactive conformation of the A AR, while
1
agonists like CCPA show higher affinity for an activated receptor conformation and stabilize
it. Our results indicate that the investigated A AR antagonists 5, 13, and 18 behave as highly
1
3
efficacious inverse agonists since their affinity for the [ H]DPCPX-labeled receptor
3
conformation is significantly higher than for the [ H]CCPA-labeled, active conformation.

2
3
Table 3. Comparison of K values obtained for selected compounds at rat A ARs using the
i
1
3
3
agonist radioligand [ H]CCPA and the antagonist radioligand [ H]DPCPX
Compounds
K ± SEM (nM)
i
at rat A adenosine receptors
1
3
3
[
H]CCPA
[ H]DPCPX
3
5
2.79 ± 0.09
0.939 ± 0.486
0.529 ± 0.043
0.348 ± 0.134
2.93 ± 2.31
0.0867 ± 0.0189
0.0709 ± 0.094
0.0628 ± 0.0228
1
1
3
8
Species differences
To analyze the observed species differences in more detail, the pK values of all potent
i
compounds investigated in this study at human and rat A or A ARs, respectively, were
1
3
2
correlated using a linear regression analysis (see Fig. 8). A high correlation (R = 0.948; Fig.
A) was observed for the human and rat A AR, with consistently higher affinities for the rat
8
1
species of about one order of magnitude. However, for the A AR, for which less data were
3
available, no correlation was observed, all compounds being less potent at the rat as compared
to the human receptor (Fig. 8B).

2
4
Fig. 8. Correlation of the pK values of rat versus human ARs. (A) correlation between human
i
and rat A ARs, (B) correlation between human and rat A ARs.
1
3
In order to understand the observed species differences, amino acid sequences of the mouse,
rat and human AR subtypes were aligned (see Table S1 of Supporting Information (SI)).
Sequence differences between AR subtypes are ranging from 39-61% (human) and a very
similar range is observed for rat and mouse (see Table S1, left column). A and A ARs are
2
A
2B
the most closely related subtypes (>60% sequence identity in human, rat and mouse). A /A
2B
2
A
compared to A ARs are the most dissimilar subtypes with only 38-40% sequence identity.
3
Although the amino acid sequences of A , A and A2B ARs are highly conserved between
1
2A
rat, mouse and human (e.g. human, rat and mouse A ARs are 95% identical, Table S1, right
1
column), the A ARs of rodents (rat, mouse) and humans share only 72% identity, which
3
determines the observed differences in ligand affinities.

2
5
Molecular modeling studies
In order to rationalize the potency and selectivity profile obtained for the 2-amino-4-
phenylthiazole derivatives, we performed molecular docking and molecular dynamics
simulation studies using the crystal structure of the human A AR (PDB ID: 5N2S) and a
1
homology model of the human A AR [29]. We selected the two potent antagonists 17 (A -
3
3
selective), and 18 (dual A /A antagonist) (see Table 1).
1
3

2
6
Fig. 9. Putative binding modes of the dual A /A antagonist 18 in the crystal structure of the
1
3
human A AR (A, B) and in the homology model of the human A AR (C, D) depicting
1
3
residues that may be important for the interaction. The human A AR and the human A AR
1
3
are displayed in cartoon representation. Carbon atoms of 18 are colored cyan and green, the
important residues are colored in gray and pink, respectively. Oxygen atoms are colored in
red, nitrogen atoms in blue, and sulfur atoms in yellow.
1
Compound 18, a 2-amino-4-phenylthiazole derivative containing a furyl group as R and a 4-
2
methylbenzoyl group at position R shows high potency towards both the human A AR and
1
the human A AR. The selected compound was docked into the orthosteric binding site of
3
both receptors using Autodock 4.2. The putative binding mode of 18 and important residues
in the binding site of the human A AR are shown in Fig. 9A and B. The compound was
1
anchored inside the binding site by two strong hydrogen bonding interactions between N3 of
the thiazole ring and NH of the carboxamido group of 18 with the amide moiety of Asn246.
The thiazole ring is predicted to additionally form a strong hydrophobic π-π interaction with
Phe171. Upon further analysis of the binding pose of 18, the compound was predicted to
occupy three subpockets (A-C, see Fig. S2 in Supporting Information). The phenyl group at
position 4 of the thiazole ring occupies subpocket A formed by the hydrophobic residues V87,
L88, A91, F171, M180, W247 and L250 and by two polar residues H251 and N254, which
stabilize compound 18 in the model through hydrophobic interactions. The respective binding
subpocket in the human A AR is narrower and cannot accommodate a phenyl ring [29]. This
2
A
is supported by the biological investigations which showed low potency of 18 at the A AR
2
A
(
Table 1). The amino acid residues in the human A AR are similar to that in the A AR, and
2B 2A
this is a likely explanation why 18 was found to be also inactive at the A AR.
2
B
2
The 4-methylbenzoyl group (R ) is proposed to be positioned in subpocket B of the human A1
AR, which is limited in size, interacting with another set of residues including V62, L65, A66,

2
7
V83, A84, V87, F171, L250 and H278. Large substituents, e.g. quinazolin-4-one derivatives
included in a number of compounds (6-9, 15-17, 20-23), cannot be accommodated in
subpocket B of the human A AR (see Table 1 for structures and binding affinity values).
1
However the potency values of the compounds with large substituents is also depending on
1
2
the substitution at R which occupies subpocket C, because the smaller substitutions at R
might induce alternative binding poses and introduce steric clashes at subpocket C. Subpocket
1
C located below extracellular loop (ECL) 3 is occupied by the 2-furyl group substituted at R
of 18. Subpocket C is formed by a mixture of hydrophobic and polar residues which include
Phe171, Glu172, Met177, Leu250, Leu253, Thr257, Lys265 and Thr270. Specifically, the
amino acid residues in ECL3 are unique for each of the AR subtypes and in each species. The
residues of ECL3 contribute largely to subtype and species selectivity, and particularly large
differences are found between the ECL3 of the human and the rat A AR.
3
As shown in Fig. 9C and D, compound 18 follows a similar interaction pattern for the thiazole
ring in the human A AR as in the A AR. It forms hydrogen bond and hydrophobic
3
1
interactions with Asn250 and Phe168, respectively, and the substitutions at position 4
2
2
(
phenyl), R (p-methylbenzoyl group) and R (2-furyl group) of 18 were observed to occupy
the three subpockets A, B and C, respectively, in the model. These interactions are assumed to
stabilize 18 in the binding site of the human A AR with slight variations in the amino acid
3
residues in comparison to the human A AR. (See Fig. S1 for sequence alignment). Although
1
similar amino acids exist in the subpockets A and B of both receptor subtypes, subpocket C
located below the ECL3 of the human A AR differs largely in comparison to the human A1
3
AR. Furthermore, the polar residues Thr257 and Thr270 (human A AR) are replaced by
1
Ile253 and Leu264 (human A AR) which alters the polarity of the subpocket. This could
3
contribute to the approximately 6-fold decrease in binding affinity of 18 at the human A AR
3
(
K 42.2 nM) in comparison to the human A AR (K 7.57 nM). Thus in addition to key
i
1
i

2
8
hydrogen bond and hydrophobic interactions, the substitution which provides a tighter
binding in subpocket B is important for the potency, and substitution at the part binding in
subpocket C and interaction of amino acid residues from ECL3 are likely responsible for
selectivity.
As a next step, we performed molecular dynamics simulations of the two complexes for
explaining the stabilities of the predicted binding poses of 18 in the binding sites of the human
A AR and the A AR (see Supplementary Video S1 and S2). For both complexes, the
1
3
calculated root mean square deviation (RMSD) values of the Cα atoms of the complexes
rapidly reached an equilibrium state with approximately 1 Å deviation from the first frame of
1
00 ns simulations. The visual analysis of the trajectories obtained for the two complexes
shows that compound 18 was anchored inside the binding site with the interaction pattern
identified from the docking studies (Fig. S3 and Fig. S4). It maintains the key hydrogen bond
interaction with an asparagine residue (Asn246 and Asn250 in the human A AR and the A3
1
AR, respectively), and hydrophobic interactions with a phenylalanine residue (Phe171 and
Phe168). The root mean square fluctuation (RMSF) values of the two receptors, the human A1
AR and A AR, showed a profile with large fluctuations only appearing in the loop regions
3
indicating a similar mechanism of interaction of the ligand with both receptors. Furthermore,
an ionic lock is predicted to be formed in the human A AR between Glu172 from ECL2 and
1
Lys265 of ECL3 acting as a cap for subpocket C to which compound 18 binds, helping the
compound to bind tightly inside the binding pocket. Similarly, an ionic lock was previously
detected in the human A2A AR formed between Glu169 and His264, which contributed to
variations in dissociation kinetics [54]. On the contrary, the amino acids Glu172 and Lys265
of the human A AR are replaced by Gln167 and Val259 in the human A AR resulting in a
1
3
loss of ionic lock formation and formation of a binding cap in subpocket C. Due to the
absence of an ionic lock and loose, flexible binding of the p-methylbenzoyl group, the key

2
9
hydrogen and hydrophobic interactions with Asn250 and Phe168 may be weakened in the A3
AR.
Fig. 10. A. Putative binding modes of 17 in the homology model of the human A AR. B.
3
Residues predicted to be important for the binding interaction are shown. The human A AR
3
is displayed by cartoon representation. Carbon atoms of 17 are colored in magenta and the
important residues for binding to the A AR are colored in pink. For color coding of other
3
atoms see Fig. 9.
1
The second selected antagonist, 17, with a 4-methoxyphenyl residue as R and a 3-(4-
2
methoxyphenyl)quinazolin-4-one residue as R shows high potency and selectivity for the
human A AR (see Table 1). The docked pose of 17 and important residues in the binding
3
pocket of the human A AR model are shown in Fig. 10A and B. Similar to compound 18,
3
hydrogen bond interactions with Asn250 and a possible hydrophobic interaction with Phe168
are conceivable. Compound 17 is predicted to also occupy the three subpockets A, B and C in
2
the binding site. Due to larger substitution at position R in 17, subpocket B is almost