Biophysical mapping of the adenosine A2A receptor

Article abstract of DOI:10.1021/jm2003798

A new approach to generating information on ligand receptor interactions within the binding pocket of G protein-coupled receptors has been developed, called Biophysical Mapping (BPM). Starting from a stabilized receptor (StaR), minimally engineered for thermostability, additional single mutations are then added at positions that could be involved in small molecule interactions. The StaR and a panel of binding site mutants are captured onto Biacore chips to enable characterization of the binding of small molecule ligands using surface plasmon resonance (SPR) measurement. A matrix of binding data for a set of ligands versus each active site mutation is then generated, providing specific affinity and kinetic information (K_D, k_on, and k _off) of receptor-ligand interactions. This data set, in combination with molecular modeling and docking, is used to map the small molecule binding site for each class of compounds. Taken together, the many constraints provided by these data identify key protein-ligand interactions and allow the shape of the site to be refined to produce a high quality three-dimensional picture of ligand binding, thereby facilitating structure based drug design. Results of biophysical mapping of the adenosine A_2A receptor are presented.

Full text of DOI:10.1021/jm2003798

ARTICLE
pubs.acs.org/jmc
Biophysical Mapping of the Adenosine A2A Receptor
Andrei Zhukov, Stephen P. Andrews, James C. Errey, Nathan Robertson, Benjamin Tehan,
Jonathan S. Mason, Fiona H. Marshall, Malcolm Weir, and Miles Congreve*
Heptares Therapeutics Limited, BioPark, Broadwater Road, Welwyn Garden City, Hertfordshire AL7 3AX, U.K.
S Supporting Information
b
ABSTRACT: A new approach to generating information on
ligand receptor interactions within the binding pocket of G
protein-coupled receptors has been developed, called Biophy-
sical Mapping (BPM). Starting from a stabilized receptor
(StaR), minimally engineered for thermostability, additional
single mutations are then added at positions that could be
involved in small molecule interactions. The StaR and a panel of
binding site mutants are captured onto Biacore chips to enable
characterization of the binding of small molecule ligands using
surface plasmon resonance (SPR) measurement. A matrix of binding data for a set of ligands versus each active site mutation is then
generated, providing specific affinity and kinetic information (KD, kon, and koff) of receptorꢀligand interactions. This data set, in
combination with molecular modeling and docking, is used to map the small molecule binding site for each class of compounds.
Taken together, the many constraints provided by these data identify key proteinꢀligand interactions and allow the shape of the site
to be refined to produce a high quality three-dimensional picture of ligand binding, thereby facilitating structure based drug design.
Results of biophysical mapping of the adenosine A2A receptor are presented.
’ INTRODUCTION
mutated to differing residues so that the effect on binding of the
radioligand being used in the assay can be determined. This
technique is extremely powerful in identifying which residues are
involved in ligand recognition and binding or catalysis.11 Typi-
cally, mutations that cause significant changes in binding affinity
are consistent with the participation of their side chains in direct
ligand interactions, although confounding factors such as “second
shell” effects, changes in solvent structure, and conformational
rearrangements or perturbations may complicate interpretation.
Critically, the requirement for binding measurements using labeled
compounds typically limits studies to one or a few exemplary
ligands, and literature data is consequently often incomplete or
conflicting, comprising a sparse matrix of data. What is required,
and is described herein, is a rapid-turnaround direct-binding
method applicable to many ligands and multiple mutations so
that a broad matrix of data can be collected, leading to a self-
consistent and convergent modeling solution for any given
GPCR system.
GPCRs have 7 R-helical membrane spanning domains which
interact with lipids within the membrane bilayer. In order to carry
out biophysical or structural studies, it is necessary to transfer the
protein from the membrane into solution by solubilization with
detergent. This process can often result in unfolding and denatura-
tion of the protein due to its inherent instability. Recently, a new
technique has been developed which involves the selection of muta-
tions which both increase the thermostability of the receptor as
G protein-coupled receptors (GPCRs) are one of the most
important classes of protein targets due to their critical role in cell
signaling in response to hormones, neurotransmitters, and other
metabolites. GPCRs are the site of action for a broad range of
small molecule and biological drugs across many therapeutic
areas. Until recently, the only X-ray structure solved from this
family was that of rhodopsin.1 Knowledge of how drugs inter-
acted with GPCRs has been limited to models based on
homology with rhodopsin, enhanced by site-directed mutagen-
esis experiments (SDM) with radiochemically labeled ligands.2,3
In the last 2ꢀ3 years, a number of different technological break-
throughs have resulted in the structures of five new GPCRs, all
of which are important drug targets; the β1 and β2 adrenergic
receptors, the adenosine A2A receptor, the CXCR4 chemokine
receptor, and the dopamine D3 receptor.4ꢀ10 Nevertheless obtain-
ing novel GPCR structures in complex with multiple ligands is
still very challenging, with some approaches being limited to
highly potent antagonist ligands and requiring specialized crystal-
lization techniques in a lipid environment. In addition, crystal-
lography does not provide direct information regarding the energetic
and kinetic contributions of particular interactions. Improved
methods are therefore required to enable structure-based design
approaches to be applied to GPCRs that are complementary
to X-ray crystallography and can be rapidly obtained, thereby
experimentally enhancing in silico design within the short designꢀ
synthesizeꢀtest cycles essential for drug discovery.
SDM is a technique commonly used in molecular pharmacol-
ogy whereby residues of a protein with a known sequence are
Received: March 31, 2011
Published: June 10, 2011
r
2011 American Chemical Society
4312
dx.doi.org/10.1021/jm2003798 J. Med. Chem. 2011, 54, 4312–4323
|

Journal of Medicinal Chemistry
ARTICLE
Figure 1. General scheme for biophysical mapping technology. Mutations within the predicted binding site of the stabilized receptor (StaR) of interest
are introduced and the proteins expressed as transient transfections. Each receptor, appropriately tagged, is captured directly from the cell lysate onto
Biacore (SPR) chips. Each mutant receptor is then challenged with a panel of antagonists with different potencies and from different chemistries, and a
matrix of binding and kinetic data is generated. Molecular modeling of the binding site, utilizing these data, refines both the protein structure and ligand
binding interactions for each chemical series.
well as driving the receptor toward a particular conformational
state. This technique of conformational thermostabilization has
been demonstrated for the β1-adrenergic,12 adenosine A2A,13
neurotensin,14 and muscarinic acetylcholine receptors.15 The
resultant stabilized receptors, known as StaRs,15 typically contain
4ꢀ8 stabilizing mutations. Although these mutations can be
identified throughout the receptor, they are selected such that
they do not interfere directly with binding to the chosen ligand
class. In this study, we used an inverse agonist StaR (fully inactive
ground state) of the adenosine A2A receptor which has a greatly
increased thermostability compared to that of the wild type
receptor and could therefore be purified to homogeneity while also
retaining correct folding once immobilized on a chip or bead in
also has been shown to suppress L-DOPA-induced dyskinesias.17
Recently, the clinical agent Preladenant (SCH420814) met its
primary end points in phase II clinical trials in moderate to severe
PD patients and has progressed into phase III.18 This finding and
other encouraging clinical trial results help to validate that antagon-
ismof this receptor is a useful approach to the treatment of PD. In
addition, there is a growing body of literature on the potential for
adenosine A2A receptor antagonism to protect from neurode-
generation.19 As part of our ongoing efforts to develop adenosine
A2A receptor antagonists for the treatment of PD, both crystal-
lographic and SDM approaches have been a focus of effort in our
laboratories. With StaRs of the A2A receptor in hand, this system
was chosen to establish a new approach to SDM that we call
biophysical mapping (BPM). Figure 1 illustrates the approach.
Starting from the A2A-StaR1, we envisaged capturing the
receptor onto a SPR chip and establishing that antagonists could
be demonstrated to bind with reproducible potency. Next, a
panel of StaRs with additional modifications in the binding site
would be generated and each captured on chips in a similar
manner. This panel of receptors would then be cross-screened
with an array of antagonists of varying potencies and from a broad
range of chemical series, generating a matrix of SPR binding and
kinetic data. In this way, it was hoped that a map of the binding
site interactions would be generated that would be useful for
structure-based design applications. Specifically, the binding data
detergent.15 The StaR (A2A-STAR1, also known as Rant2113,15
used here contains 4 mutations spread throughout the receptor
)
but distinct from the antagonist binding site (A54L2.52, T88A3.36
,
K122A4.43, and V239A6.41; superscripts refer to Ballesteros-
Weinstein numbering16). The pharmacology has been previously
characterized15 and indicates that the receptor has been stabilized
in an inverse agonist conformation.
The adenosine A2A receptor is coexpressed with the dopamine
D2 receptor on the indirect striatopallidal pathway and acts to
oppose the actions of dopamine agonists. Blockade of the A2A
receptor in rodent and primate models of Parkinson’s disease (PD)
with antagonists alleviates the motor symptoms of the condition and
4313
dx.doi.org/10.1021/jm2003798 |J. Med. Chem. 2011, 54, 4312–4323

Journal of Medicinal Chemistry
ARTICLE
would be used to guide the selection of best poses from a set of in
silico docked possibilities for each ligand; docking methods
frequently find the correct pose for a ligand, but it can often be
buried within a list of 5ꢀ10 poorly prioritized options. In the case
of the A2A receptor, it would also later be possible to compare
these results with crystallographic information. The following
report details this series of studies on the A2A receptor and the
interpretation of these data.
agonist signaling (noted above) were found to be thermostabiliz-
ing as part of the inverse agonist complex and had unaltered
[3H]ZM241385 binding.
The final set of mutations selected for the SPR studies were
L85A3.33, L167AECL2, M177A5.38, N253A6.55, Y271A7.36, I66A2.64
,
N181A5.42, and S277A7.42. It was quickly identified that N253A6.55
was nonbinding for all compounds attempted; no ligands gave a
signal with this mutated StaR, consistent with previous studies
that suggest this residue has a critical role in ligand binding.22,23
However, despite evidence that the receptor did give low but
sufficient expression for capture on the chip, it cannot absolutely
be ruled out that nonbinding is instead due to the observed lower
level of expression.
’ RESULTS
Experimental Design/Selection of Mutants. At the time this
work was initiated, there was no structural information available,
and our understanding of ligand binding to the adenosine A2A
receptor was based on in silico modeling using homology models
derived from the stabilized β1-adrenergic receptor X-ray
structure.7 The models were built with both Modeler20 and
MOE,21 with manual adjustments of alignments where neces-
sary. No attempt was made to completely build extra-cellular
loop 2 (ECL2); only the residues between the top of transmem-
brane helix 5 (TM5) to the Cys166 involved in a disulfide bridge
were modeled. Supporting both the model building and, criti-
cally, the docking of ligands was a large set of literature and in-
house SDM data. Adenosine A2A has been the subject of a
number of previous studies to characterize the binding of
radioligands.22,23 Of particular note, Kim et al. have summarized
SDM results for both adenosine A2A and A1 receptors and the
effects of mutations on the binding of various ligands.23 By
comparing these data with our model of the binding site, a list of
residues likely to be directly involved in ligand binding was
derived. The residues highlighted by Kim et al. as having a large
Interpretation of Data and Generation of Biophysical
Maps. Table 1 summarizes the data collected for each SDM
modified StaR versus a panel of 21 ligands. The log difference in
binding affinity for each binding site mutant is given compared
with the unaltered StaR. Figure 2 shows a representative set of
SPR sensorgrams obtained with one of the compounds,
ZM241385. Figure 3 presents the same data graphically to
visualize the patterns characteristic of a particular compound
series. ZM241385 and SCH420814 are furan containing amino-
heterocyclic analogues, and the latter compound is in clinical
trials for the treatment of PD.18 Istradefylline (KW6002),
xanthine amine congener (XAC),25 caffeine, and theophylline
are xanthine derivatives.26 KW600227 has also been studied in
PD clinical trials. Compounds 1aꢀe, 2aꢀb, and 3aꢀh are from
three distinct compound series for which medicinal chemistry
optimization efforts have been undertaken in our laboratories.
The chemical structures of analogues in series 1 and 3 are not
disclosed here and will be the topic of future publications.
However, these data are included to illustrate a number of
aspects of the BPM technology.
effect on antagonist binding were V843.32, E151ECL2, E169ECL2
F1825.43, H2506.52, N2536.55, F2576.59, I2747.39, and H2787.43
,
.
Those residues thought to be involved in agonist binding and/or
signaling were T883.36, S2777.42, and S2817.46. Residues reported
as also having similar effects on antagonist or agonist binding for
First, it is clear that trends in the binding of all ligands
independent of their chemotype or potency can be drawn from
the data set. It is an important finding that these effects are not
dependent on affinity and that quite weakly active molecules can
be profiled and interrogated by this method. The most obvious
feature is that L85A causes an average change in binding affinity
by ꢀ1.3 log units relative to that of the unaltered StaR (StaR
background). In comparison, L167A and M177A have on
average only a small effect (ꢀ0.2 log units). Y271A, I66A, and
N181A have, on average, an intermediate effect (ꢀ0.4, ꢀ0.7, and
ꢀ0.5 log units, respectively). Interestingly, S277A, on average,
causes an increase of 0.3 log units to ligand binding. Second, each
chemotype has a distinct fingerprint, supporting the notion that
different classes of ligands have, in fine detail, different binding
modes in the site (Figure 3). For example, compounds in series 1
have a larger reduction in binding due to L85A and are affected by
I66A, but other mutations have less of an effect on the binding of
these molecules. In contrast, the binding of compounds in series
2 is affected to a similar extent by the majority of the SDM
changes, suggesting a quite distinct binding mode compared to
that of series 1. In series 3, where the most examples are
presented, there is a fairly consistent trend across the data;
L85A, Y271A, I66A, and N181A all reduce binding, and S277A
causes an increase in binding, especially when compared to other
series. These data were very informative in our optimization of
series 3, which was proposed to bind very deeply in the bottom of
the binding cavity, unusually forming interactions inside the
adenosine ribose binding pocket adjacent to Ser277. Finally, in
each series there is a definite structureꢀactivity relationship
the adenosine A1 receptor were T913.36, H2516.52, N2546.55
T2777.42, and H2787.43. Of these residues, we considered V843.32
,
,
F1825.43, H2506.52, N2536.55, I2747.39, and H2787.43 to be inside
the A2A antagonist binding pocket. Asn2536.55 was assigned as
the key H-bonding motif likely to be critical to antagonist
binding. Thr883.36 and Ser2777.42 were proposed to be located
below the antagonist binding site and involved in agonist signaling.
[3H]ZM24138524 was the ligand used during the StaR ther-
mostabilization process. During the generation of a StaR, large
numbers of mutations to the receptor are introduced iteratively,
and as well as determining which residues increase the thermal
stability of the receptor, binding data was generated on mutations
that affect the binding of the radioligand. The alanine scanning of
the adenosine A2A receptor identified a selection of mutants that
resulted in a loss of [3H]ZM241385 binding. These mutants
provided a profile of the antagonist binding site. Levels of
expression for these mutants were tested by Western blotting
with only two mutants, M177A5.38 and N253A6.55, showing reduced
levels of expression. The mutations L85A3.33, L167AECL2, I66A2.64
,
and N181A5.42 showed significantly reduced [3H]ZM241385
binding to below 20% of wild type levels, while retaining wild type
levels of expression. Mutant Y271A7.36 bound [3H]ZM241385 at
approximately 40% of wild type levels, while the mutants
N253A6.55 and H278A7.43 showed almost a complete loss of
[3H]ZM241385 binding at saturating ligand concentrations. The
two mutants T88A3.36 and S277A7.42 proposed to be involved in
4314
dx.doi.org/10.1021/jm2003798 |J. Med. Chem. 2011, 54, 4312–4323

Journal of Medicinal Chemistry
ARTICLE
Table 1. SPR Data for a Panel of mutated StaRs versus a Panel of Ligandsa
compound
pKi
pKD
L85A
L167A
M177A
Y271A
I66A
N181A
S277A
ZM241385
8.7
8.8
6.7
7.6
5.3
<5
9.3
9.5
5.9
7.7
4.6
5.5
7.3
5.9
5.9
6.1
5.8
6.0
7.9
6.7
8.9
8.5
8.2
8.2
8.5
9.3
9.2
ꢀ1.8
ꢀ2.3
ꢀ1.1
ꢀ0.3
ꢀ1.6
ꢀ1.8
ꢀ1.3
ꢀ2.9
ꢀ0.8
ꢀ0.8
ꢀ2.8
ꢀ0.8
ꢀ0.9
ꢀ1.1
ꢀ0.8
ꢀ0.7
ꢀ0.8
ꢀ0.8
ꢀ0.9
ꢀ1.1
ꢀ1.1
0
ꢀ0.1
ꢀ0.9
ꢀ0.3
ꢀ0.2
nd
ꢀ0.6
ꢀ0.9
0
ꢀ0.6
ꢀ1.0
ꢀ0.2
ꢀ0.9
ꢀ1.6
ꢀ0.1
ꢀ0.7
ꢀ0.5
ꢀ0.7
ꢀ0.3
ꢀ0.5
ꢀ0.7
0.1
ꢀ0.9
ꢀ1.2
ꢀ0.3
ꢀ0.4
0.6
ꢀ0.4
ꢀ0.9
0.7
0.3
0.7
0
SCH420814
ꢀ1.0
ꢀ0.2
ꢀ0.1
0.1
KW6002
XAC
ꢀ0.5
0
caffeine
theophylline
0
nd
0.1
ꢀ0.4
ꢀ0.3
0.1
1a
1b
1c
1d
1e
2a
2b
3a
3b
3c
3d
3e
3f
8.5
7.6
7.5
6.5
5.8
7.9
9.0
7.8
8.9
8.2
8.1
7.9
8.0
8.6
8.6
ꢀ0.1
ꢀ0.1
0.1
0.1
ꢀ0.7
0.1
0.3
0.3
0.2
0.1
0
ꢀ0.3
0
0
ꢀ0.3
ꢀ0.5
ꢀ0.4
ꢀ0.5
ꢀ0.2
ꢀ1.3
ꢀ0.5
ꢀ0.7
ꢀ0.6
0.1
ꢀ0.5
ꢀ0.1
ꢀ0.4
ꢀ0.4
ꢀ0.1
ꢀ0.1
0.1
nd
ꢀ0.6
ꢀ0.3
ꢀ0.5
ꢀ0.1
ꢀ0.8
ꢀ0.3
ꢀ0.3
ꢀ0.3
ꢀ0.2
ꢀ0.1
ꢀ1.4
ꢀ1.1
ꢀ0.2
nd
ꢀ0.6
ꢀ1.0
1.0
0.3
0.8
0.9
1.2
1.4
0.7
0.8
ꢀ0.3
ꢀ0.6
0
ꢀ1.1
ꢀ0.5
ꢀ0.4
ꢀ0.3
ꢀ0.8
ꢀ0.7
ꢀ1.5
ꢀ1.2
0
0.1
0.1
0.1
nd
0.3
0.5
ꢀ0.6
ꢀ1.0
ꢀ0.6
3g
ꢀ1.4
ꢀ0.7
ꢀ0.8
ꢀ0.8
3h
a The SPR data for each mutated StaR is given as a log difference compared with the unaltered StaR background (pKD). The pKi value versus the wild
type receptor is also given in the Table for each compound, as determined in a radioligand binding assay (using [3H]-ZM241385) for comparison. In
most cases, the pKi and pKD values are very similar, and differences are likely due to the low solubility of certain ligands which, we have observed, can
cause differences between the assays. nd = not determined (due to low signal). Errors in pKD difference values are typically (0.1.
(SAR) related to changes in chemical structure. This additional
information was used to refine the binding of individual analo-
gues and to suggest hypotheses for new compounds to be
synthesized. However, we believe the power of the BPM method
is in the SAR trends within the data for each series rather than any
outliers. For example, compounds 3g and 3h seem to be more
adversely affected by the SDM than closely related examples. It is
likely that this is simply due to their very high affinity, whereupon
optimal interactions can be readily disrupted, rather than any
significant change in binding mode.
In addition to the determination of binding constants given in
Table 1 and discussed above, data for on- and off-rates (kon and
koff) for each compound against each mutant is generated by the
SPR method. These data are presented in Supporting Informa-
tion, Figure S2. Differences in binding affinity observed in the
data are most often derived from changes in koff rather than kon.
This is consistent with general observations for other systems.
However, it can be noted that in some cases changes in kon were
important, and in further cases, changes in kon and koff occur but
cancel each other, meaning that overall affinity did not change.
The data provided by the kon and koff values give additional
insight into the energetics of binding of the various compound
series and assist in generating models and hypotheses of the role
of each binding site residue on molecular interactions.
and Asn253 is in the plane. N253A6.55 is assumed to be
nonbinding (NB in the figure) for all compounds based on
literature SDM results and on binding experiments carried out on
the first 4 compounds in Table 1, all of which gave no signal. The
numbers shown below each residue label are the log difference in
binding affinity relative to the background StaR, and the residue
changes that had the largest effect on each ligand are highlighted
with a black oval, the second largest effect by a dotted circle,
and the third largest effect by a gray box. In cases where there is
a similar effect (either positive or negative) for the first, second,
or third tier of binding change, then more than one residue is
highlighted.
Comparison of ZM241385 with SCH420814 indicates that
each ligand interacts with the receptor in a similar way, as judged
by the pattern of binding to the panel of mutant StaRs. In
particular, the first and second tier effects are the same with the
mutation of Leu85 and Asn181, respectively. Similarly, both
compounds are affected by changes to Ile66 and Tyr271 to a
similar extent; however, SCH420814 is equally affected by
changes to the other residues in the panel. As noted earlier, very
potent molecules seem to be more generally sensitive to any
change in the binding site, and this is how these third tier effects
could be interpreted for this compound. Alternatively, the larger
tricyclic template and bulky phenylpiperazine substituent might
be responsible for these changes. Overall, these data support that
the furan moieties in these ligands bind in a pocket below Asn253
lined by Met177, Asn181, and Leu85, and we proposed key H-bond
contacts with Asn253 to both the furan and the heterocyclic ring
of these molecules. The other residues in the site have smaller or
no effect on the binding of these furan containing analogues,
suggesting no critical interactions with the ligands.
Figure 4 illustrates the chemical structures of the first six
molecules in Table 1. In each case, the residues mutated in the
binding site are displayed around the ligand in their relative
orientation according to a homology model we had developed
before any crystal structure data was available. Ile66, Leu85,
Met177, Asn181, and Leu167 are in front of the plane of the
paper (bold), Ser277 and Tyr271 are behind the plane (italic),
4315
dx.doi.org/10.1021/jm2003798 |J. Med. Chem. 2011, 54, 4312–4323

Journal of Medicinal Chemistry
ARTICLE
Figure 2. Kinetics of ZM241385 binding to A2A StaR and seven binding site mutants as analyzed by SPR. The adenosine A2A StaR proteins were
immobilized on NTA chips on a Biacore T200 instrument and a series of ligand concentrations (5ꢀ80 nM) injected in the single cycle format at 10 °C.
The data were fitted to a 1:1 interaction model to generate KD values, and the data are presented in Table 1 and Figure 3.
The four xanthine analogues display a more complex pattern
of interactions within the binding site. Careful examination of the
data seems to support a subtly different binding mode for the
xanthine moiety of caffeine and of KW6002 as compared with
that of theophylline and XAC. Caffeine has a methyl group on its
imidazole ring that would prevent any H-bonding with Asn253 to
that ring, while theophylline lacks this substituent, allowing for
this possibility. A comparison of caffeine with theophylline shows
effects on affinity in the first and second binding tier to Leu85 and
Asn181, suggesting the ligands sit relatively deep in the bottom of
the receptor, adjacent to Asn253. However, caffeine is affected by
changes to Ser277 (þ0.7 log) and Ile66 (ꢀ1.6 log), while
theophylline shows no such effect. Theophylline therefore seems
to be forming H-bonds to Asn253 with a positioning in the site
close to Leu85 and Asn181, with no further key direct or indirect
interactions with the other residues studied. Caffeine, in contrast,
appears to sit further from Asn253, presumably forming a single
H-bond to one carbonyl group and is apparently positioned
closer to S277 (þ0.7 log) and Ile66 (ꢀ1.6 log). Furthermore,
comparison of caffeine with KW6002 (which is also N-methy-
lated on the imidazole ring) shows a similar pattern of interactions
with Leu85 and Ser277. This indicates a similar overall binding
mode of the core fragment, but differences in the effects of
Asn181 (ꢀ0.3 log) and Ile66 (ꢀ0.2 log), suggesting, in fine
detail, a shift in the vector of the xanthine moiety or, perhaps
more likely, an effect of the ethyl (versus methyl) substituents on
the heterocyclic core. Finally, a comparison of theophylline with
XAC suggests that XAC forms a number of new interactions with
the receptor, but the data again overall suggests the molecule sits
adjacent to Asn253 and accesses the residues lining the bottom
of the pocket. The larger propyl substituents (as compared with
methyl groups on theophylline) may account for the interaction
with Ile66 (ꢀ0.9 log), which is diminished in the smaller
derivative. The amide group in XAC appears to be positioned
close to Tyr271 (ꢀ0.5 log) compared with the styryl group of
KW6002 which appears to form no interactions with the residues
on that face of the receptor (Tyr271, Leu167, and Ile66), again
suggesting a subtly different vector in the presentation of these
two molecules to the key H-bonding Asn253 residue.
Modeling of the Binding Modes of ZM241385 and XAC.
Figure 5 illustrates the structures of ZM241385 and XAC bound
to an adenosine A2A homology model optimized using the BPM
4316
dx.doi.org/10.1021/jm2003798 |J. Med. Chem. 2011, 54, 4312–4323

Journal of Medicinal Chemistry
ARTICLE
Figure 3. Effect of active site mutations on binding affinities of compounds representing five different chemical series. These data are a graphical
representation of the data presented in Table 1 and show how different compound series have different fingerprints which map their interactions within
the receptor binding site. On- and off-rates of ligands binding to the target receptors were also measured and indicated that SDM changes typically affect
the off-rate rather than the on-rate (see Supporting Information).
data detailed above. Homology models were constructed from
the β1 adrenoceptor crystal structure in the protein data bank
(PDB; 2VT4),7 with manual readjustment and refinement based
on known conserved GPCR motifs and SDM data (published
and in-house, as discussed earlier). In Figure 5, the first, second,
and third tier BPM effects described above and shown in Figure 4
are colored, along with Asn253, which is essential for the binding
of ligands to the receptor. The two ligands share a significant
portion of the A2A binding pocket in both position and interac-
tions with residues lining the pocket. Both have πꢀπ interac-
tions with their central aromatic cores and Phe168 from ECL2,
hydrophobic contacts with Leu249 and Met270 (not shown in
the figure), and hydrogen bonding contact with Asn253. The
interactions at the bottom of the pocket seen for ZM241385 are
highlighted in Figure 5 as the most significant, designated as first
and second tiers in the BPM experiments. Most important of
these is the almost two log drop in activity for the L85A mutation.
The Leu85 forms one side of the hydrophobic cavity into which
the furan ring, known to be essential for ligand binding, sits. The
furan also forms a key hydrogen bond with Asn253. Lastly in this
region, the Asn181 forms part of the pocket where the furan ring
binds and thus has an effect when mutated to Ala, as shown by the
BPM in Table 1 and Figure 4.
At the entrance to the binding site, the model has an open
arrangement that allows stacking of the phenolic substituent of
the ligand with Tyr271 in a cleft at the extracellular ends of
helices 1, 2, and 7. This allows the second tier effect with Tyr271
to be accounted for in the binding pose with a sensible ligand
conformation. This flexible part of the ligand was best accom-
modated by the solution presented here, and we rationalized that
the pose was reasonable on the basis of the BPM data and
previously published SDM and SAR data that suggested the
importance of the substitution pattern of groups on the phenyl
ring for potency and selectivity, suggestive of engagement of the
side chain in a subpocket within the binding site.22,23,28,29
XAC and ZM241385 are overall proposed to sit similarly in the
model, but there are some significant differences that affect the
way they are positioned within the model of the A2A binding
cavity. The most significant difference between the two inverse
agonists is the interaction proposed for XAC with Ile66. One of
4317
dx.doi.org/10.1021/jm2003798 |J. Med. Chem. 2011, 54, 4312–4323

Journal of Medicinal Chemistry
ARTICLE
Figure 4. Representative BPM interpretation. Chemical structures of ZM241385, SCH420814, KW6002, caffeine, XAC, and theophylline annotated
with the BPM data for each ligand in a cartoon representation based on a homology model of the A2A receptor. The residues shown were in each case
mutated to Ala, and beneath the label is the log difference in binding for that compound. Residue labels are positioned to represent their position in the
binding site and relative to the ligand. Residues labeled in bold are in front of the plane, in italics are behind the plane, and in normal font in the plane of
the ligand. Asn253 is included in each case, and NB = nonbinding (see text). ND = not determined. The numbers below each residue label are the log
difference in binding relative to the StaR background. The residue changes that had the largest effect on each ligand are highlighted with a black oval,
second largest effect by a dotted circle, and third largest effect by a gray box. In cases where there is a similar effect (either positive or negative) for the 1st,
2nd, or 3rd tier of binding change, then more than one residue is highlighted.
the propyl chains traverses the width of the binding site where it
makes a significant hydrophobic interaction within a lipophilic
cavity in the binding site defined by Ile66, Ala63, Val84, and
Ala81. ZM241385 does not access this lipophilic cavity, and
mutation of Ile66 has a small effect on its binding. This significant
hydrophobic interaction with Ile66 for XAC is therefore con-
sistent with the marked decrease on binding seen with the
mutation of Ile66 to Ala (reduced by almost 1 log unit). This
major reduction in XAC’s affinity with the Ile66 to Ala mutation
is designated as first tier in Figure 5. It was very interesting to us
that the BPM highlighted this residue because all high scoring
docking poses did not predict any interaction with this residue. In
order to recapitulate the interactions the ligand is predicted to
make from the biophysical map, it was necessary to rotate the
carboxamide of the Asn253 in the model to better present the
NH2 donor to the C6 carbonyl of the xanthine. This illustrates
that docking of xanthine ligands such as XAC was very challen-
ging without the additional information provided by the BPM
experiments. This interaction with Ile66 may actually be as a
result of the steric bulk present on the ligand from the second
propyl substituent that serves to cause a rotation of the scaffold
relative to ZM241385. This has the effect that XAC has only one
H-bonding interaction with Asn253 and forms subtly different
interactions with the residues that form the bottom of the
binding site (Ser277, Leu85, Asn181, and Met177) compared
with ZM241385. These interactions for XAC with the base of the
binding site are third tier effects possibly because of the flexibility
of this lipophilic propyl chain accessing this part of the binding
site, allowing the ligand to adapt to SDM changes introduced in
this region. In finalizing the proposed binding mode, we took
into account that the S277A mutant actually increases binding of
the ligand (shown as green in the figure), consistent with
placement of the hydrophobic side chain in this region. Finally,
when examining the BPM binding mode of XAC, at the top of the
binding site the Tyr271 chi1 angle is in a trans conformation in
the model where it is able to make a weak πꢀπ stacking
arrangement with the phenyl ring of XAC. This is consistent
with the moderate (2nd tier) effect seen on the binding of XAC
to the Y271A mutant with a drop of half a log in binding.
Interestingly, in a recent paper by Cheng et al. a binding mode
for xanthines in adenosine A2A is proposed that has some
similarity to our BPM guided pose.30 Cheng et al. also conclude
4318
dx.doi.org/10.1021/jm2003798 |J. Med. Chem. 2011, 54, 4312–4323

Journal of Medicinal Chemistry
ARTICLE
Figure 5. Docked structures of ZM241385 (left) and XAC (right) in an adenosine A2A receptor homology model. Asn253 is colored red as SDM of this
residue prevents binding of the ligands (see text). The 1st, 2nd, and 3rd tier effects of mutations within the binding site are colored according to the key in
the figure (also see text) and relate to those residues indicated in Figure 4 from the BPM experiments.
that the xanthine ligands only form one H-bonding interaction
with the Asn253 and that there is a significant interaction with
Tyr271. However, as the xanthine ligands are docked into a
published adenosine A2A crystal structure (PDB: 3EML) in
which Tyr271 is in the gauche negative conformation the binding
mode suggested here is not possible.30 This is because the top of
the binding site is partially occluded in the position occupied by
the substituted phenyl group. However, the binding pose of
xanthines proposed by Cheng et al. in A2B is very similar to the
results presented here because Y271 (A2A) is actually N273
(A2B), and thus, there is no residue obstructing this portion of the
binding site.
The later publication of an adenosine A2A receptor structure in
complex with ZM241385 allowed us to make a comparison of
our biophysical map of this ligand with an X-ray crystal data set.8
The difference between the backbone rmsd of the homology
model and the crystal structure is 2.78 Å over all residues
included in the model. Most binding site residues within the
TM region showed a high degree of similarity, as was also found
by Ivanov et al. in another recent comparison of a model with this
X-ray data.31 Within the binding site, there are small differences
comparing the residues from the top of TM2, notably Ile66,
being somewhat further away from Asn253 in our homology
model compared to the crystal structure, perhaps helping to
explain the importance of this residue (Table 1). Differences, not
surprisingly, were also seen in residues emanating from the loops
which are notoriously difficult to model, notably in the exact
position of Phe168 on ECL2, which forms stacking interactions
with the ligand in the site.11,32 However, the most important
difference affecting the binding pose of the ligand ZM241385
between our model and the crystal structure (PDB: 3EML) was
the chi1 angle of Tyr271. In the 3EML crystal structure, chi1 of
Tyr271 is in a gauche negative rotameric state, compared to
the trans conformation in our model, which causes the flexible
phenolic portion of ZM241385 to sit in quite different positions
comparing the two binding poses. This gauche negative
rotameric state of Tyr271 chi1 serves to effectively block the
channel between helices 2 and 7 where we proposed the flexible
phenolic portion of ZM241385 might bind. While we also
obtained docking poses that did place this portion of
ZM241385 in a similar place to that seen for 3EML, these poses
did not satisfy the BPM data that are presented here and thus
were discarded. The BPM guided docking of ZM241385, how-
ever, did show a very similar placement of the central core with an
approximately 10° clockwise rotation of our BPM guided pose
relative to the 3EML model, resulting in a heavy atom rmsd of the
central core of 1.78 Å. The most significant region of displace-
ment was where the flexible phenolic portion joins onto the
central core. Importantly, the B factor (or temperature factor, a
crystallographic measure of disorder) of the flexible phenol
group in the A2A crystal structure is high (>100 Å2) compared
to the rest of the ligand (∼50), suggesting the position of this
substituent should be interpreted with caution and may adopt
more than one conformation.33 In our ongoing research, we were
subsequently able to solve a crystal structure of an adenosine A2A-
StaR with ZM241385 bound. Interestingly, this structure has
Tyr271 positioned in a trans arrangement, different from its
position in 3EML, that opens the binding cleft between helices 2
and 7 and is in a position similar to that in the homology model
described in this article.34 In this structure, the flexible phenolic
portion of ZM241385 does indeed bind within this cleft between
helices 2 and 7, consistent with the results from the BPM study
presented here and also with the SAR seen for this well explored
series of compounds.28,35
In summary, the studies presented here show how a selection
of binding site mutants can be made and SPR binding data
subsequently generated for each mutant receptor against a panel
of ligands with a broad range of potency and diverse chemistry.
Interpretation of the binding modes of each chemical series and
individual molecules within each series can be made using in
silico docking against a homology model, itself refined and improved
using the BPM data. Comparison of the binding modes suggested
4319
dx.doi.org/10.1021/jm2003798 |J. Med. Chem. 2011, 54, 4312–4323

Journal of Medicinal Chemistry
ARTICLE
from the BPM matrix with subsequently solved receptorꢀligand
X-ray structures demonstrates the value of the approach, and the
BPM data additionally serves to improve our understanding of
the ligand interactions seen in the crystal structures.
biophysical method that is able to identify small (molecular
weight generally less than 250ꢀ300 Da), low affinity (high
micromolar to low millimolar) compounds that are ligand
efficient.37 Chemical elaboration of these fragments then enables
the identification of compounds with good physicochemical
properties and target selectivity. The important first step is to
screen a relatively small library of fragment-size compounds and
to identify compounds with a good (energetically favorable) fit to
the binding site, and this can be done, for example, by SPR or
NMR based screening using StaR proteins.38 This is in contrast
to high throughput screening (HTS) whereby much larger hits
are often obtained, with compounds that can be making sub-
optimal interactions with the protein. Many interesting new
binding motifs can be rapidly identified by the fragment-based
approach. The utility of the BPM technology is thus potentially
not limited to lead-sized potent molecules but due to the ability
of the SPR screening approach to detect very weak binders can be
used to provide structure-based design in the critical fragment to
hit/lead stage. The BPM data enables characterization of where
fragments are binding, distinguishing between orthosteric or
other (allosteric) site binders, being used to screen panels of
analogues in a SAR by catalogue approach or by guiding the
synthesis of derivatives. This capability is very important to
enable structure-based design for fragment hits before an X-ray
structure can be obtained, as computational docking approaches
are poor at providing high confidence poses for small com-
pounds, generally giving several plausible possibilities. The data
for caffeine and theophylline described in this article are exam-
ples of fragments binding to a GPCR, and the BPM method
allowed the binding modes of these molecules to be predicted in
a mode largely consistent with medium resolution X-ray struc-
tures of these compounds and of the larger analogue, XAC, that
have subsequently been solved in our laboratories.34 One con-
sideration is that in some cases a small fragment might still bind
with similar affinity to a mutant receptor due to a significant
compensating change in its binding mode; however, this situa-
tion would be expected to be uncommon.
In conclusion, the matrix of direct binding information
provided by the BPM approach provides early access to quality
experimental data that allows better prediction of correct ligand
binding poses and quantifies the effects on binding of changing
residues around the binding site. Structure-based design for a
receptor target is thus greatly enabled by having a more accurate
binding pose, using experimental information. Additionally, an
understanding of the relative importance of different interactions
with binding site residues, given by the BPM approach, can be
used to guide design in a SAR program. Coupled with the advances
in homology modeling made possible by new GPCR structures
becoming available, enhanced by improved force fields (taking more
account of polarization, etc.), increased desktop computational
power, and new in silico methods for improving docking predic-
tions, the BPM method has great potential to support virtual and
fragment hit identification campaigns and empower SBDD for
lead optimization. Without the BPM data on our adenosine A2A
antagonist lead compounds, it would not have been possible to
use SBDD methods with confidence until much later in the
project, when X-ray structures with our ligands became available.
’ DISCUSSION
The BPM approach described here is an important link
between computational (in silico) docking and X-ray structures
of ligands bound to their receptor sites. In the absence of any
X-ray crystal structure data, the information provided by the
BPM of known ligands can be used to validate and improve
homology models (including orientation of key binding site
residues) and the way they are used to dock and score com-
pounds. This potentially improves the ability of a given model to
identify new ligands, such as via virtual screening of large
databases of existing compounds, and also to predict the binding
orientation of known ligands. Generating BPM data on a ligand
of interest provides higher confidence in the binding pose/
orientation predictions made by in silico docking and can lead
to predictions of binding modes for compounds that are very
close to the experimental binding mode observed when the X-ray
crystal structure is ultimately obtained. This is not limited to fine
tuning of known binding modes but can be used to correctly
predict novel binding modes. For example, the compounds in
series 3 (Table 1) were identified by virtual screening of our
homology model of the adenosine A2A receptor, carefully refined
and validated using SDM and BPM data. Furthermore, the BPM
data and binding mode predictions were very useful to the
project team, enabling the rapid optimization of the series and
allowing identification of advanced lead molecules that have
excellent drug-like properties and have led to the identification of
a preclinical candidate for the treatment of PD. An important
validation of the BPM method comes from later crystal structures
of members of this series. The structures are highly consistent
with the BPM data presented in this article and the binding mode
derived from these data that was being used during the optimiza-
tion of the compound series (unpublished results).
The choice of which residues are to be mutated during the
BPM screening is a critical part of the process that is done by
careful analysis of the 3D structure or homology model of the
receptor together with potential ligand docking poses and any
available SDM data. Mutations to cover all plausible binding
modes are then considered, modulating the size or polarity of the
residue in a way that is likely to give the most informative data.
There are naturally limitations to the approach, most importantly
that allosteric binding sites may not be predictable at the outset,
such that no mutations are made to probe binding at that site.
However, the lack of binding energy changes by mutations to the
orthosteric binding site residues would at least inform that
binding was occurring at an alternative site. An additional
consideration is that if a mutation ablates binding for all
compounds studied, then expression and protein immobilization
on to the SPR chip need to be carefully examined. In the
experiments presented here, all mutations in the panel were to
alanine. In our ongoing BPM studies for other receptors, we now
make multiple mutations for each key residue within the site such
that we can identify mutations that affect but do not completely
block ligand binding, aiding the interpretation of results.
’ EXPERIMENTAL SECTION
An exciting approach to finding hits for targets of interest
is fragment-based drug discovery (FBDD).36 In this approach,
smaller compounds (fragments) are typically screened by a
Materials. The A2AR receptor StaR used as the background is a
thermostabilized mutant previously known as RANT21,13 which
4320
dx.doi.org/10.1021/jm2003798 |J. Med. Chem. 2011, 54, 4312–4323

Journal of Medicinal Chemistry
ARTICLE
includes the following thermostabilized mutations: A54L2.52, T88A3.36
,
(m, 2H), 6.74 - 6.78 (m, 2H), 6.67 - 6.73 (m, 1H), 4.31 - 4.43 (m, 2H),
3.89 - 3.99 (m, 2H), 3.53 - 3.62 (m, 2H), 3.25 (s, 3H), 2.87 - 2.98 (m, 4H),
2.76 - 2.83 (m, 2H), 2.52 - 2.60 (m, 4H). LCMS 5.13 min, 95.04% purity
(UV); m/z 504.45 (ESþ).
K122A4.43, and V239A6.41 and was used as the background for the
additional site-directed mutagenesis. [3H]-ZM241385 was supplied by
American Radiochemicals (USA). Ni-nitroltriacetic acid (Ni-NTA)
agarose was obtained from Qiagen (UK). All detergents were from
Anatrace, Affymetrix (USA). Transient transfection reagent Genejuice
was supplied by Merck Biosciences (USA).
ZM241385 (4-(2-(7-Amino-2-(furan-2-yl)-[1,2,4]triazolo[1,5-a]-
[1,3,5]triazin-5-ylamino)ethyl)phenol; Tocris Cookson), XAC (N-(2-
aminoethyl)-2-[4-(2,3,6,7-tetrahydro-2,6-dioxo-1,3-dipropyl-1H-purin-
8-yl)phenoxy]-acetamide; Sigma Aldrich), caffeine (Sigma Aldrich) and
theophylline (Sigma Aldrich) are commercially available.
Mutagenesis. Mutants were generated by PCR using the KOD
PCR system (Merck Biosciences, USA) and the expression plasmid as
template. PCRs were transformed into DH5R ultracompetent cells
prepared according to the Inoue method, and individual clones were
fully sequenced to check that only the desired mutation was present.
Different mutations were combined by PCR as described above and
sequenced to check plasmid integrity and desired mutations.
Protein Expression. HEK-293T mammalian cells were used to
express adenosine A2A and its mutants. Briefly, HEK293T cell cultures in
10 cm cell culture plates at 50% confluency (4.5 ꢁ 106 cells) were
transiently transfected with 6 μg of plasmid and left to grow in DMEM
(Lonza) supplemented with 10% fetal bovine serum (Sigma) at 37 °C
and 5% CO2. Cells were harvested 40 h later by dissociation into
phosphate buffer saline (Sigma) and pelleted at 600g for 5 min (aliquots
of 2 mL), and stored at ꢀ20 °C.
LCMS analysis of these commercial samples was performed at Heptares
under the following conditions: instrument, Waters Alliance 2795, Waters
2996 PDA detector, Micromass ZQ. Column: Phenomenex Gemini-NX
C-18, 3 μm, 2.0 ꢁ 30 mm. Gradient [time (min)/solvent B in A (%)]:
0.00/2, 0.10/2, 8.40/95, 9.40/95, 9.50/2, 10.00/2 (solvent A = 1.58 g
ammonium formate in 2.5 L of water þ 2.7 mL of 28% aqueous ammonia
solution; solvent B = 2.5 L of acetonitrile þ132 mL of solvent A þ
2.7 mL of 28% aqueous ammonia solution). Injection volume, 5 μL; UV
detection, 230 to 400 nM; column temperature, 45 °C; and flow rate =
1.5 mL/min. The LCMS purity was found to be >95% in all cases.
LCMS results for commercial samples (retention time, UV purity,
observed ion in ESþ mode): ZM241385 (2.42 min, >95%, m/z 336.3),
XAC (1.83 min, >95%, m/z 429.3), caffeine (1.07 min, >95%, m/z 194.9),
and theophylline (0.13 min, >95%, m/z 181.0).
Radioligand Binding. Competition binding studies were incu-
bated in buffer containing 50 mM Tris-HCl (pH 7.4) in a final volume of
0.25 mL with 10 ꢁ 0.5 log unit dilutions of test compounds and a final
[3H]-ZM241385 concentration of 0.5 nM. After incubation for 90 min
at room temperature, assays were terminated by rapid filtration through
96-well GF/B UniFilter plates presoaked with 0.1% PEI followed by
washing with 5 ꢁ 0.25 mL of ddH2O. Plates were dried, 50 μL of Ultima
Gold-F added per well, and bound ligand measured using a Packard
Microbeta counter. Data were normalized as % specific binding followed
by fitting according to a four parameter logistic fit to determine pIC50;
this was converted to pKi using the KD value generated from saturation
binding studies.
Determination of Protein Concentration. The protein con-
centration was determined using a detergent-compatible Bradford assay
(BioRad).
Chemical Compounds. SCH420814 and KW6002 were synthe-
sized by Jubilant Chemsys Ltd. (India), according to published proce-
dures. Their purities were determined by LCMS and were found to be
>95%.
SPR Analysis. SPR analysis was carried out at 10 °C on Biacore
T100 and Biacore T200 instruments using sensor chip NTA
(GE Healthcare). PBS was used as running buffer with 0.1% dodecylmalto-
side, 0.05 mM EDTA, and 5% DMSO, pH 7.4. To solubilize the
adenosine A2A receptor from the membranes of HEK-293T cells, the
pelleted membranes were mixed with an equal volume of ice-cold
40 mM Tris-HCl and 3% decylmaltoside, pH 7.4, and homogenized
in a Dounce homogenizer. The mixture was centrifuged (13200 rpm, 5
min) and the pellet discarded. The supernatant was supplied with 5 mM
imidazole and injected over a nickel-loaded NTA chip (1 μL/min, 10
min) to capture the His-tagged receptor. The capture levels of the wild
type and mutant receptors varied in the range of 4000ꢀ10000 resonance
units (RU). To assay the receptor ligands for the affinity and kinetics of
their interaction with the receptor, 2-fold dilution series (five concentra-
tions) were prepared and injected in either multi- or single-cycle format.
The actual concentrations as well as contact and dissociation times varied
between the ligands depending on the kinetics and affinity of interaction.
The data were generally fitted to the 1:1 interaction model. Some low-
affinity ligands, which needed to be assayed at higher concentrations,
exhibited a distinct secondary interaction component. This secondary
interaction was characterized by low affinity and was often super-
stoichiometric. In cases with significant contribution of the secondary
component, the data were fitted to the two-site interaction model, and
only the primary, high affinity, component was taken into consideration.
Binding constants were compared with affinities measured by radioligand
binding (the above method and Table 1), and compounds that per-
formed poorly due to solubility issues were discarded from the screen.
Computational Chemistry. Homology models were constructed
from the avian β1 adrenergic GPCR crystal structure bound to cyano-
pindolol (PDB: 2VT4).7,41 Owing to the relatively low percentage
identity between the two proteins (25% overall, less than 20% around
the ligand binding site), two initial homology models of the adenosine
A2A receptor were generated, using different methods. This provided a
means to assess consistency in the alignments, the variability within the built
structures, and which regions of the models had lower confidence associated
with them. One model was constructed using Modeler,20 while the other
was constructed using MOE,21 with manual readjustment of the
clustalW alignment where necessary.42 The alignment in each case
was checked to ensure consistency with known GPCR conserved
motifs43 and particularly the conserved disulfide bond, common to
family A GPCRs, which is located between the top of helix 3 and the
1H NMR spectra of these compounds were recorded at 400 MHz on a
Jeol instrument. Chemical shift values are expressed in parts per million,
i.e., δ values. The following abbreviations are used for the multiplicities
of the NMR signals: s = singlet, d = doublet, t = triplet, q = quartet,
dd = doublet of doublets, and m = multiplet. Coupling constants are
listed as J values, measured in Hz.
LCMS analysis of these compounds was performed by Jubilant Chemsys
Ltd. under the following conditions: column, Aquity BEH C-18, 1.7 μm,
2.1 ꢁ 100 mm; column temperature = 30 °C; and flow rate = 0.3 mL/min.
Gradient [time (min s)/solvent B in A (%)]: 0.00/90, 1.00/90, 2.00/
3
85, 4.30/45, 6.00/10, 8.00/10, 9.0/10, 10.00/10 (solvent A = acetoni-
trile; solvent B = 5 mM aqueous ammonium acetate).
KW600239. (8-[(E)-2-(3,4-Dimethoxyphenyl)vinyl]-1,3-diethyl-7-methyl-
3,7-dihydro-1H-purine-2,6-dione) 1H NMR (400 MHz, DMSO-d6) δ:
7.58 (d, J = 15.80 Hz, 1H), 7.39 (d, J = 1.83 Hz, 1H), 7.27 (dd, J = 1.72,
8.36 Hz, 1H), 7.18 (d, J = 15.80 Hz, 1H), 6.95 (d, J = 8.47 Hz, 1H), 4.03
(q, J = 7.10 Hz, 2H), 3.99 (s, 3H), 3.88 (q, J = 7.10 Hz, 2H), 3.81 (s, 3H),
3.76 (s, 3H), 1.22 (t, J = 7.10 Hz, 3H), 1.09 (t, J = 6.98 Hz, 3H). LCMS
5.95 min, 97.92% purity (UV); m/z 385.23 (ESþ).
SCH42081440. (2-(Furan-2-yl)-7-[2-[4-[4-(2-methoxyethoxy)phenyl]-
piperazin-1-yl]ethyl]-7H-pyrazolo[4,3-e][1,2,4]triazolo[1,5-c]pyrimidin-
5-amine) 1H NMR (400 MHz, DMSO-d6) δ: 8.09 - 8.16 (m, 1H), 7.99 -
8.07 (broad s, 2H), 7.86 - 7.94 (m, 1H), 7.16 - 7.23 (m, 1H), 6.78 - 6.83
4321
dx.doi.org/10.1021/jm2003798 |J. Med. Chem. 2011, 54, 4312–4323

Journal of Medicinal Chemistry
ARTICLE
extracellular loop 2. The model generated via MOE was selected for
further optimization as the Phe168 residue, implicated in antagonist
binding, was in closer proximity to the other residues highlighted from
literature SDM.
transmembrane helix; SAR, structureꢀactivity relationship;
PDB, protein data bank; FBDD, fragment-based drug discovery
’ REFERENCES
Initial validation and improvement of the homology models was
conducted with ZM241385, due to the wealth of SAR for this series and
the amount of SDM data available for the ligand at the adenosine A2A
receptor. The induced-fit docking protocol was used within Maestro
with an autogenerated box size around the residues highlighted by in-
house and external SDM as having a large effect on antagonist binding,
(1) Palczewski, K.; Kumasaka, T.; Hori, T.; Behnke, C. A.; Motoshima,
H.; Fox, B. A.; Le Trong, I.; Teller, D. C.; Okada, T.; Stenkamp, R. E.;
Yamamoto, M.; Miyano, M. Crystal structure of rhodopsin: A G
protein-coupled receptor. Science 2000, 289, 739–745.
(2) Patny, A.; Desai, P. V.; Avery, M. A. Homology modeling of
G-protein-coupled receptors and implications in drug design. Curr. Med.
Chem. 2006, 13, 1667–1691.
(3) Li, Y. Y.; Hou, T. J.; Goddard, W. A., III. Computational
modeling of structure-function of G protein-coupled receptors with
applications for drug design. Curr. Med. Chem. 2010, 17 (12), 1167–
1180.
(4) Rasmussen, S. G.; Choi, H.; Rosenbaum, D. M.; Thian, F.;
Kobilka, T.; Edwards, P. C.; Burghammer, M.; Ratnala, V. R. P.;
Sanishvili, R.; Fischetti, R. F.; Schertler, G. F. X.; Weis, W. I.; Kobilka,
B. K. Crystal structure of the human β2 adrenergic G-protein-coupled
receptor. Nature 2007, 450, 383–387.
(5) Cherezov, V.; Rosenbaum, D. M.; Hanson, M. A.; Rasmussen,
S. G. F.; Thian, F.; Kobilka, T.; Choi, H; Kuhn, P.; Weis, W. I.; Kobilka,
B. K.; Stevens, R. C.; Takeda, S.; Kadowaki, S.; Haga, T.; Takaesu, H.;
Mitaku, S.; Fredriksson, R.; Lagerstrom, M. C.; Lundin, L. G.; Schioth,
H. B.; Pierce, K. L.; Premont, R. T.; Lefkowitz, R. J.; Lefkowitz, R. J.;
Shenoy, S. K. High-resolution crystal structure of an engineered human
β2 adrenergic G protein-coupled receptor. Science 2007, 318,
1258–1265.
namely, I662.64, V843.32, L853.33, E151ECL2, L167ECL2, E169ECL2
N1815.42, F1825.43, H2506.52, N2536.55, F2576.59, Y2717.36, I2747.39
,
,
and H2787.43. The resultant docking solutions in combination with
the binding data from the thermostabilization experiments were then
used to select the in-house round of mutants for our biophysical mapping
experiments. These mutations were chosen so as to confirm or rule out
differing possible binding orientations within the A2A receptor.
The final docking experiments were guided by ligand SAR and our
iterative process of assessing SDM and then designing our own BPM
mutants to confirm or rule out possible binding modes. The protein
preparation and docking experiments were done within the Schrodinger
Maestro package.44 The grid generation necessary for docking was done
within Glide. The residues highlighted in SDM experiments (in-house
and external) were used to define the cavity of the grid; however, no
constraints were added in the grid generation to ensure that subsequent
dockings were not biased in any way. Glide XP docking was carried out
on all of the ligands in question with 10 poses per ligand being stored.
The poses were then assessed against the biophysical mapping data
(Table 1) and the best solution identified.
(6) Xu, F.; Wu, H.; Katritch, V.; Han, G. W.; Jacobson, K. A.; Gao, Z.;
Cherezov, V.; Stevens, R. C. Structure of an agonist-bound human A2A
adenosine receptor. Science 2011, 332, 322–327.
(7) Warne, T.; Serrano-Vega, M. J.; Baker, J. G.; Moukhametzianov,
R.; Edwards, P. C.; Henderson, R.; Leslie, A. G. W.; Tate, C. G.;
Schertler, G. F. X. Structure of a β1-adrenergic G-protein-coupled
receptor. Nature 2008, 454, 486–491.
(8) Jaakola, V. P.; Griffith, M. T.; Hanson, M. A.; Cherezov, V.;
Chien, E. Y. T.; Lane, J. R.; IJzerman, A. P.; Stevens, R. C. The 2.6
angstrom crystal structure of a human A2A adenosine receptor bound to
an antagonist. Science 2008, 322, 1211–1217.
(9) Wu, B.; Chien, E. Y. T.; Mol, C. D.; Fenalti, G.; Liu, W.; Katritch,
V.; Abagyan, R.; Brooun, A.; Wells, P.; Bi, F. C.; Hamel, D. J.; Kuhn, P.;
Handel, T. M.; Cherezov, V.; Stevens, R. C. Structures of the CXCR4
chemokine GPCR with small-molecule and cyclic peptide antagonists.
Science 2010, 330, 1066–1071.
’ ASSOCIATED CONTENT
S
Supporting Information. Representative SPR senso-
b
grams for three compound series, representative binding data
and rate constants for ZM241385 and three compound series,
information for compound series 1 and 3, and experimental
details for series 2. This material is available free of charge via the
Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION
Corresponding Author
*Phone þ44 (0)1707 358638. E-mail: miles.congreve@heptares.com.
Notes
The authors are employees of Heptares Therapeutics Ltd. who
are engaged in GPCR structure based drug discovery.
(10) Chien, E. Y.; Liu, W.; Zhao, Q.; Katritch, V.; Won Han, G.;
Hanson, M. A.; Shi, L.; Newman, A. H.; Javitch, J. A.; Cherezov, V.;
Stevens, R. C. Structure of the human dopamine D3 receptor in complex
with a D2/D3 selective antagonist. Science 2010, 330, 1091–1095.
(11) Garland, S.; Blaney, F. GPCR Homology Model Development
and Application. In Burger’s Medicinal Chemistry, Drug Discovery and
Development, 7th ed.; John Wiley and Sons, Inc.: New York, 2010; pp
279ꢀ300.
’ ACKNOWLEDGMENT
(12) Serrano-Vega, M. J.; Magnani, F.; Shibata, Y; Tate, C. G.
Conformational thermostabilization of the β1-adrenergic receptor in a
detergent-resistant form. Proc. Natl. Acad. Sci. U.S.A. 2008, 105,
877–882.
The authors thank Edward Hurrell for generating the radi-
oligand binding data, Chris Richardson and Bissan Al-Lazikani
for help with constructing the first generation of homology
models, and David Myszka for helping to establish conditions
for the use of A2A StaRs on the Biacore format as a precursor to
the work in this manuscript.
(13) Magnani, F.; Shibata, Y.; Serrano-Vega, M. J.; Tate, C. G. Co-
evolving stability and conformational homogeneity of the human
adenosine A2A receptor. Proc. Natl. Acad. Sci. U.S.A. 2008, 105,
10744–10749.
(14) Shibata, Y.; White, J. F.; Serrano-Vega, M. J.; Magnani, F.; Aloia,
A. L.; Grisshammer, R.; Tate, C. G. Thermostabilization of the neuro-
tensin receptor NTS1. J. Mol. Biol. 2009, 390, 262–277.
(15) Robertson, N.; Jazayeri, A.; Errey, J.; Baig, A.; Hurrell, E.;
Zhukov, A.; Langmead, C. J.; Weir, M.; Marshall, F. H. The properties of
thermostabilised G protein-coupled receptors (StaRs) and their use in
drug discovery. Neuropharmacology 2011, 60, 36–44.
’ ABBREVIATIONS USED
StaR, stabilized receptor; BPM, biophysical mapping; GPCR,
G protein-coupled receptor; SPR, surface plasmon resonance;
SDM, site-directed mutagenesis; XAC, xanthine amine conge-
ner; PD, Parkinson’s disease; ECL, extra-cellular loop; TM,
4322
dx.doi.org/10.1021/jm2003798 |J. Med. Chem. 2011, 54, 4312–4323

Journal of Medicinal Chemistry
ARTICLE
(16) Ballesteros, J. A.; Weinstein, H.; Stuart, C. S. In Methods in
Neurosciences; Academic Press: New York, 1995; Vol. 25, pp 366ꢀ428.
(17) Shah, U.; Hodgson, R. Recent progress in the discovery of
adenosine A2A receptor antagonists for the treatment of Parkinson’s
disease. Curr. Opin. Drug Discovery Dev. 2010, 13, 466–480.
(18) Salamone, J. D. Preladenant, a novel adenosine A2A receptor
antagonist for the potential treatment of parkinsonism and other
disorders. IDrugs 2010, 13, 723–731.
(36) Congreve, M.; Chessari, G.; Tisi, D.; Woodhead, A. J. Recent
developments in fragment-based drug discovery. J. Med. Chem. 2008,
51, 3661–3680.
(37) Hopkins, A. L.; Groom, C. R.; Alex, A. Ligand efficiency: a
useful metric for lead selection. Drug Discovery Today 2004, 9, 430–431.
(38) Congreve, M; Rich, R. L.; Myszka, D. G.; Figaroa, F.; Siegal, G.;
Marshall, F. H. Fragment screening of stabilised G protein-coupled receptors
using biophysical mapping. Methods Enzymol. 2011, 493, 115–134.
(39) Hockemeyer, J.; Burbiel, J. C.; M€uller, C. E. Multigram-scale
syntheses, stability, and photoreactions of A2A adenosine receptor
antagonists with 8-styrylxanthine structure: Potential drugs for Parkinson’s
disease. J. Org. Chem. 2004, 69, 3308–3318.
(40) Neustadt, B. R; Hao, J.; Lindo, N.; Greenlee, W. J.; Stamford,
A. W.; Tulshian, D.; Ongini, E.; Hunter, J.; Monopoli, A.; Bertorelli, R.;
Foster, C.; Arik, L.; Lachowicz, J.; Nga, K.; Feng, K. Potent, selective, and
orally active adenosine A2A receptor antagonists: Arylpiperazine deriva-
tives of pyrazolo[4,3-e]-1,2,4-triazolo[1,5-c]pyrimidines. Bioorg. Med.
Chem. Lett. 2007, 17, 1376–1380.
(41) Berman, H. M; Westbrook, J.; Feng, Z.; Gilliland, G.; Bhat,
T. N.; Weissig, H.; Shindyalov, I. N.; Bourne, P. E. The Protein Data
Bank. Nucleic Acids Res. 2000, 28, 235–242.
(42) Thompson, J. D.; Higgins, D. G.; Gibson, T. J. CLUSTAL W:
improving the sensitivity of progressive multiple sequence alignment
through sequence weighting, position-specific gap penalties and weight
matrix choice. Nucleic Acids Res. 1994, 22, 4673–4680.
(19) Chen, J. F.; Sonsalla, P. K.; Pedata, F.; Melani, A.; Domenici, M. R.;
Popoli, P.; Geiger, J.; Lopes, L. V.; de Mendonca, A. Adenosine A2A
receptors and brain injury: broad spectrum of neuroprotection, multifaceted
actions and “fine tuning” modulation. Prog. Neurobiol. 2007, 83, 310–331.
ꢀ
(20) Sali, A.; Blundell, T. L. Comparative protein modeling by
satisfaction of spatial restraints. J. Mol. Biol. 1993, 234, 779–815.
(21) Molecular Operating Environment (MOE 2008.10); Chemical
Computing Group, Inc.: Montreal, Quebec, Canada, 2008. www.chem-
comp.com.
(22) Dal Ben, D.; Lambertucci, C.; Marucci, G.; Volpini, R.; Cristalli,
G. Adenosine receptor modeling: What does the A2A crystal structure
tell us? Curr. Top. Med. Chem. 2010, 93, 993–1018.
(23) Kim, S. K.; Gao, Z.; Van Rompaey, P.; Gross, A. S.; Chen, A.;
Van Calenbergh, S.; Jacobson, K. A. Modeling the adenosine receptors:
comparison of the binding domains of A2A agonists and antagonists.
J. Med. Chem. 2003, 46, 4847–4859.
(24) Uustare, A.; Vonk, A.; Terasmaa, A.; Fuxe, K.; Rinken, A.
Kinetic and functional properties of [3H]ZM241385, a high affinity
antagonist for adenosine A2A receptors. Life Sci. 2005, 76, 1513–1526.
(25) Jacobson, K. A.; Ukena, D.; Kirk, K. L.; Daly, J. W. [3H]xanthine
amine congener of 1,3-dipropyl-8-phenylxanthine: an antagonist radi-
oligand for adenosine receptors. Proc. Natl. Acad. Sci. U.S.A. 1986,
83, 4089–4093.
(43) Mirzadegan, T.; Benko, G. Sequence analyses of G-protein
coupled receptor: Similarities to rhodopsin. Biochemistry 2003, 42 (10),
2759–2767.
(44) Maestro, version 9.1; Schr€odinger, LLC: New York.
(26) Daly, J. W. Alkylxanthines as research tools. J. Auton. Nerv. Syst.
2000, 81, 44–52.
(27) Jenner, P. Istradefylline, a novel adenosine A2A receptor
antagonist, for the treatment of Parkinson’s disease. Expert Opin. Invest.
Drugs 2005, 14, 729–738.
(28) Baraldi, P. G.; Cacciari, B.; Romagnoli, R.; Spalluto, G.;
Monopoli, A.; Ongini, E.; Varani, K.; Borea, P. A. 7-Substituted
5-Amino-2-(2-furyl)pyrazolo[4,3-e]-1,2,4-triazolo[1,5-c]pyrimidines as
A2A adenosine receptor antagonists: A study on the importance of
modifications at the side chain on the activity and solubility. J. Med.
Chem. 2002, 45, 115–126.
(29) Jaakola, V. P.; Lane, J. R.; Lin, J. Y.; Katritch, V.; Ijzerman, A. P.;
Stevens, R. C. Ligand binding and subtype selectivity of the human A2A
adenosine receptor: identification and characterization of essential
amino acid residues. J. Biol. Chem. 2010, 285 (17), 13032–13044.
(30) Cheng, F.; Xu, Z.; Liu, G.; YunTang, Y. Insights into binding
modes of adenosine A2B antagonists with ligand-based and receptor-
based methods. Eur. J. Med. Chem. 2010, 45, 3459–3471.
(31) Ivanov, A. A.; Barak, D.; Jacobson, K. A. J. Med. Chem. 2009,
52, 3284–3292.
(32) Barton, N.; Blaney, F. E.; Garland, S.; Tehan, B.; Wall, I. Seven
transmembrane G protein-coupled receptors: insights for drug design
from structure and modeling. Comp. Med. Chem. II 2007, 4, 669–701.
(33) Katritch, V.; Rueda, M.; Lam, P. C.; Yeager, M.; Abagyan, R.
GPCR 3D homology models for ligand screening: lessons learned from
blind predictions of adenosine A2a receptor complex. Proteins 2010, 78
(1), 197–211.
(34) Dorꢁe, A. S.; Robertson, N.; Errey, J. C.; Ng, I.; Hollenstein, K.;
Tehan, B.; Hurrell, E.; Bennet, K.; Congreve, M.; Magnani, F.; Tate,
C. G.; Weir, M.; Marshall, F. H. Structure of the adenosine A2A receptor
in complex with ZM241385 and the xanthines XAC and caffeine.
Structure 2011, in press.
(35) de Zwart, M.; Vollinga, R. C.; Beukers, M. W.; Sleegers, D. F.;
Von Frijtag Drabbe Kunzel, J. K.; De Groote, M.; Ijzerman, A. P. Potent
antagonists for the human adenosine A2B receptor. Derivatives of the
triazolotriazine adenosine receptor antagonist ZM241385 with high
affinity. Drug Dev. Res. 1999, 48, 95–103.
4323
dx.doi.org/10.1021/jm2003798 |J. Med. Chem. 2011, 54, 4312–4323