Fighting obesity with a sugar-based library: Discovery of novel MCH-1R antagonists by a new computational-VAST approach for exploration of GPCR binding sites

Article abstract of DOI:10.1021/ci4000882

Obesity is an increasingly common disease. While antagonism of the melanin-concentrating hormone-1 receptor (MCH-1R) has been widely reported as a promising therapeutic avenue for obesity treatment, no MCH-1R antagonists have reached the market. Discovery and optimization of new chemical matter targeting MCH-1R is hindered by reduced HTS success rates and a lack of structural information about the MCH-1R binding site. X-ray crystallography and NMR, the major experimental sources of structural information, are very slow processes for membrane proteins and are not currently feasible for every GPCR or GPCR-ligand complex. This situation significantly limits the ability of these methods to impact the drug discovery process for GPCR targets in real-time , and hence, there is an urgent need for other practical and cost-efficient alternatives. We present here a conceptually pioneering approach that integrates GPCR modeling with design, synthesis, and screening of a diverse library of sugar-based compounds from the VAST technology (versatile assembly on stable templates) to provide structural insights on the MCH-1R binding site. This approach creates a cost-efficient new avenue for structure-based drug discovery (SBDD) against GPCR targets. In our work, a primary VAST hit was used to construct a high-quality MCH-1R model. Following model validation, a structure-based virtual screen yielded a 14% hit rate and 10 novel chemotypes of potent MCH-1R antagonists, including EOAI3367472 (IC ₅₀ = 131 nM) and EOAI3367474 (IC₅₀ = 213 nM).

Full text of DOI:10.1021/ci4000882

Subscriber access provided by University of South Dakota
Article
Fighting Obesity with a Sugar-Based Library: Discovery
of Novel MCH-1R Antagonists by a New Computational-
VAST Approach for Exploration of GPCR Binding Sites
Alexander Heifetz, Oliver Barker, Geraldine P Verquin, Norbert Wimmer,
Wim Meutermans, Sandeep Pal, Richard J. Law, and Mark Whittaker
J. Chem. Inf. Model., Just Accepted Manuscript • DOI: 10.1021/ci4000882 • Publication Date (Web): 16 Apr 2013
Downloaded from http://pubs.acs.org on April 26, 2013
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a free service to the research community to expedite the
dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts
appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been
fully peer reviewed, but should not be considered the official version of record. They are accessible to all
readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered
to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published
in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just
Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor
changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers
and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors
or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
Journal of Chemical Information and Modeling is published by the American Chemical
Society. 1155 Sixteenth Street N.W., Washington, DC 20036
Published by American Chemical Society. Copyright © American Chemical Society.
However, no copyright claim is made to original U.S. Government works, or works
produced by employees of any Commonwealth realm Crown government in the course
of their duties.

Page 1 of 22
Journal of Chemical Information and Modeling
1
2
3
4
5
6
7
8
9
Fighting Obesity with a Sugar-Based Library: Discovery of Novel
MCH-1R Antagonists by a New Computational-VAST Approach for
Exploration of GPCR Binding Sites
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Alexander Heifetz,1* Oliver Barker,1 Geraldine Verquin,2 Norbert Wimmer,2 Wim Meutermans,2
Sandeep Pal,1 Richard J. Law,1 Mark Whittaker1
1 Evotec (UK) Ltd., 114 Milton Park, Abingdon, Oxfordshire OX14 4SA, United Kingdom;
2 Alchemia Ltd, Eight Mile Plains, Queensland 4113, Australia
ABSTRACT: Obesity is an increasingly common disease. While antagonism of the melanin-concentrating hormone-1
receptor (MCH-1R) has been widely reported as a promising therapeutic avenue for obesity treatment, no MCH-1R antag-
onists have reached the market. Discovery and optimization of new chemical matter targeting MCH-1R is hindered by
reduced HTS success rates and a lack of structural information about the MCH-1R binding site. X-ray crystallography and
NMR, the major experimental sources of structural information, are very slow processes for membrane proteins, and are
not currently feasible for every GPCR or GPCR-ligand complex. This situation significantly limits the ability of these
methods to impact the drug discovery process for GPCR targets in “real-time” and hence there is an urgent need for other
practical and cost-efficient alternatives. We present here a conceptually pioneering approach that integrates GPCR mod-
eling with design, synthesis and screening of a diverse library of sugar-based compounds from the VAST technology (Ver-
satile Assembly on Stable Templates), to provide structural insights on the MCH-1R binding site. This approach creates a
cost-efficient new avenue for structure-based drug discovery (SBDD) against GPCR targets. In our work, a primary VAST
hit was used to construct a high-quality MCH-1R model. Following model validation, a structure-based virtual screen
yielded a 14% hit rate and 10 novel chemotypes of potent MCH-1R antagonists, including EOAI3367472 (IC50 = 131 nM) and
EOAI3367474 (IC50 = 213 nM).
the class A family of GPCRs, which are integral membrane
proteins. MCHꢀ1R is widely distributed in the central nervꢀ
ous system and is particularly concentrated in regions that
are involved in rewarding behavior, feeding behavior and
metabolic regulation.2 The distribution of the MCHꢀ2R
receptor is less clear.8 Behavior studies have demonstrated
a physiological role of the MCH system in regulating food
intake.4 Acute central administration of MCH or MCHꢀ1R
synthetic agonists leads to a rapid and significant increase
in food intake in rodents.9 This increase in food consumpꢀ
tion is correlated with the functional activity of the agonist
for the receptor, suggesting that the MCHꢀ1R directly meꢀ
diates the appetite stimulant effects of MCH.2 The activaꢀ
tion of MCHꢀ1R by MCH or by synthetic agonists could be
a key factor in the development of obesity, increasing enerꢀ
gy intake while also reducing energy expenditure.2 Conꢀ
versely, MCHꢀ1R antagonists could reduce MCHꢀ1 funcꢀ
tional activity and as such slow down the development of
obesity.2
1. INTRODUCTION
Obesity is an increasingly problematic medical condiꢀ
tion, particularly in the developed world, associated with an
increased risk of developing typeꢀ2 diabetes, cardiovascular
disease, hypertension, stroke, and cancer.1 Obesity is charꢀ
acterized by a chronic imbalance between energy expendiꢀ
ture and energy intake.2 While many of the mechanisms
underlying obesity are far from being fully understood, it
has become clear that the condition is influenced by six
,
4
neuropeptides3 that are ligands of Gꢀproteinꢀcoupled reꢀ
ceptors (GPCRs); i.e. the 19ꢀaminoꢀacid polypeptide melaꢀ
ninꢀconcentrating hormone (MCH) of MCHꢀ1 (MCHꢀ1R)
and ꢀ2 (MCHꢀ2R) receptors,5 the ghrelin hormone that
binds to the growth hormone secretagogue receptor
(GHSR1a), orexinꢀA and ꢀB peptides of Orexin ꢀ1 and ꢀ2
receptors respectively and neuropeptides B/W that partner
with the neuropeptide B/W receptors. It was discovered
that small organic molecules antagonize obesityꢀrelated
GPCRs and as such have the potential to be used as antiꢀ
obesity drugs.2
The first nonꢀpeptide MCHꢀ1R selective antagonist reꢀ
ported was Tꢀ226296,10 which effectively blocked oral
food intake in rats. A second nonꢀpeptide antagonist,
SNAP7941,2 provided the first evidence that chronic oral
administration of an antagonist can cause sustained reducꢀ
Wide experimental evidence has identified MCH recepꢀ
tors as promising therapeutic targets in the treatment of
,
7
obesity.6 The MCHꢀ1R and MCHꢀ2R receptors belong to
ACS Paragon Plus Environment

Journal of Chemical Information and Modeling
Page 2 of 22
tions in body weight. A number of MCHꢀ1R antagonists
onists. This practical approach, integrating the VAST techꢀ
including GWꢀ80343011 have been reported to reduce the
amount of food intake in rats and cause weight loss in
obese mouse models. The chronic administration of other
MCHꢀ1R antagonists in rodents led to significantly reduced
food intake, significant reduction of body weight and fat
nology with molecular modeling, opens an additional aveꢀ
nue for structural exploration of GPCRs and for structureꢀ
based drug discovery against GPCR targets. Our method is
able to satisfy the immediate need of drugꢀdiscovery proꢀ
cess in the structural information on the target GPCR, parꢀ
ticularly when there is an absence of Xꢀray crystal data.
The general workflow for the discovery of novel MCHꢀ1R
antagonists from initial screening of a pyranoseꢀbased liꢀ
brary and subsequent development of an optimized MCHꢀ
1R model is shown in Figure 1a.
1
2
3
4
5
6
7
8
,
13
gain during the whole treatment period.12 These promisꢀ
ing experimental results maintain the high interest in furꢀ
ther development of MCHꢀ1R antagonists as potential
drugs for obesity.7 More than 2,700 MCHꢀ1R antagonists
have been discovered but none of them have been successꢀ
ful in clinical trials so far.7
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
(a) VAST-GPCR modeling workflow
A lack of direct structural information on the binding site
of MCHꢀ1R has limited the discovery and optimization of
Sugar-Based
library (VAST)
Screening
Structure-
Based Drug
Discovery
GPCR
Modeling
Model
Validation
,
14
MCHꢀ1R antagonists to a trial and error approach.7 This
observation has resulted in a significant reduction in the
reporting of discovery and optimization of new chemotypes
of MCHꢀ1R antagonists in recent years.6 However, some
success has been reported for structureꢀbased virtual
screening (VS) campaigns employing theoretical 3D modꢀ
els of MCHꢀ1R, indicating that structural information can
benefit the drug discovery process for MCHꢀ1R antagoꢀ
VAST Hit
(b) Structural diversity in VAST library
F
OMe
O
O
OMe
O
O
HO
HO
MeO
HO
O
,
nists.14 15
O
O
F
NH
F
NH
NH
W
O
O
The observation that MCHꢀ1R binds ligands with a vaꢀ
riety of chemotypes and sizes hints at the complex topology
F
F
R
R
K
14
and large size of its binding site. Xꢀray crystallography
A
B
C
and NMR, the major experimental sources of structural
information, are very slow processes for membrane proꢀ
teins, and are not currently feasible for every GPCR or
GPCRꢀligand complex.16 This situation significantly limits
the ability of these methods to impact the drug discovery
process for GPCR targets in “realꢀtime” and hence there is
an urgent need for other practical and costꢀefficient alternaꢀ
tives. Prediction of GPCR structures in the absence of diꢀ
rect experimentallyꢀderived structural information on target
receptor is possible but speculative.
Figure 1: (a) General strategy followed in this work, proceed-
ing from the initial VAST library screening to the discovery
of novel chemically unrelated antagonists, through the de-
velopment of an optimized MCH-1R model. (b) Examples of
structural diversity in the VAST library scanning the general
Aromatic-Aromatic-Positive Charge motif. Substituents on
the various positions of the sugar scaffold are represented by
a letter corresponding to the amino-acid side-chain mim-
icked (F=Phenylalanine, W=Tryptophane, R=Arginine,
K=Lysine). Compounds of type A and B share the same motif
(FRF) but differ in terms of chemoform (relative orientation
of each substituent defined by the sugar scaffold type and the
positioning of each substituent on the scaffold). Compounds
of type A and C share the same chemoform but differ in
terms of motifs (FRF and FKW respectively).
Discovery of small drugꢀlike GPCR antagonists is also
especially challenging for peptidergic GPCRs like MCHꢀ
1R because it depends on correctly transposing known
binding elements of the parent peptide onto small drugꢀlike
molecules while maintaining bioactivity.17 The binding of
peptides to GPCRs is defined by two key factors; the comꢀ
bination of unique binding elements (motifs) to generate
key interactions with specific residues of the receptor and
the relative orientation of these binding elements in3D
space (bioactive conformation). The peptide bioactive conꢀ
formations and motifs can be mimicked by appending subꢀ
stituents onto a rigid organic scaffold in a similar relative
orientation as in the original peptide. A significant amount
of work has been reported on the applicability of using a
monosaccharide scaffold in peptidomimetic design apꢀ
The VAST technology is ideally suited to the exploration
of large and/or topologically complex GPCR binding
,
21
sites.19 It is based on the use of a sugar (pyranose) scafꢀ
fold on which it is possible to append various substituents
in a systematic manner to obtain a structurally diverse liꢀ
brary. Many GPCR peptides can be minimized to tripeptide
molecules and still maintain biological activity.26 We
showed that the pyranose scaffold is ideally suited to mimic
these tripeptides because of its unique versatility in terms
of positioning 3 substituents in 3D space and the significant
overlay of topographical space thus accessible with the
corresponding tripeptide conformational space. An examꢀ
ple of a diverse VAST library, described in detail in a preꢀ
vious paper, was designed to systematically scan the genꢀ
eral AromaticꢀAromaticꢀPositive Charge motif.17 The comꢀ
pounds in this library each contained three substituents
representing various triꢀpeptide sequences of this general
motif (more precisely FKW, FKF, WKW, FRW, FRF or
-
proaches17 25 however the design of bioactive pepꢀ
tidomimetics, particularly for GPCRs, remains a challenge.
We present here a novel integrated experimentalꢀ
computational approach that enabled molecularꢀlevel exꢀ
ploration of the MCHꢀ1R structure, including prediction of
its antagonist binding site, its topology and the key residues
involved in antagonist binding followed by structureꢀbased
discovery of 10 new chemotypes of potent MCHꢀ1R antagꢀ
ACS Paragon Plus Environment

Page 3 of 22
Journal of Chemical Information and Modeling
WRW with F=Phenylalanine, W=Tryptophan, K=Lysine, region and 4 aligned prolines. The second extracellular
1
2
3
4
5
6
7
8
R=Arginine sideꢀchain mimetic) in a particular relative
orientation referred to as a chemoform, defined by the
choice of sugarꢀbased scaffold and substitution pattern on
the selected scaffold (Figure 1b). This concept can be exꢀ
tended to a wide range of different motifs, the premise of
our approach being that each GPCR will have a unique
pattern of motif and chemoform preference. The 490 VAST
compounds obtained from this library design were screened
against MCHꢀ1R, resulting in the discovery of a potent
MCHꢀ1R antagonist, ACL21823 (radioligand binding RLB
MCHꢀ1R IC50 = 306 nM, aequorin functional assay AEQ
MCHꢀ1R EC50 = 679 nM, Figure 2).
loop was modeled using ECL2 from bovine rhodopsin
(PDB entry 1F88), attached to the structure and optimized
using OPLSꢀAA energy function. We selected rhodopsin as
a template to model the ECL2 loop of MCHꢀ1R due to two
reasons: Firstly, the ECL2 loop of MCHꢀ1R has 24.1%
sequence similarity with rhodopsin and just 13.8% similariꢀ
ty with the β2AR ECL2 loop. Secondly, the conformation
of the β2AR ECL2 loop is defined by two disulfide bonds
formed between cysteine residues in the ECL2 loop and the
7TM domain. Both MCHꢀ1R and rhodopsin only have one
cysteine in the ECL2 loop.
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
(a) Model of MCH-1R with ACL21823 – Binding site location
Extracellular side
Extracellular side view
Figure 2: Structure of ACL21823 and schematic representa-
tion in terms of vectors of substituent positioning from the
pyranose scaffold (vectors V1, V2, V3 for the substituents at
position 1, 2 and 3 respectively on the sugar scaffold).
The discovery of ACL21823 was a key factor in the conꢀ
struction of a high quality MCHꢀ1R model and the location
of its antagonist binding site. The high quality of the MCHꢀ
1R model was demonstrated by a virtual enrichment experꢀ
iment and the modelꢀdriven structureꢀbased expansion of
ACL21823, which allowed the generation of a list of key
MCHꢀ1R residues involved in antagonist binding. The
utility of our method to drug discovery was demonstrated
by a structureꢀbased virtual screen, achieving a hit rate of
14% and yielding 10 new chemotypes of MCHꢀ1R antagoꢀ
nists including EOAI3367472 (IC50 = 131 nM) and
EOAI3367474 (IC50 = 213 nM).
Intracellular side
(b) Model of MCH-1R with ACL21823 – Key residues
2. RESULTS AND DISCUSSION
2.1 VAST library synthesis and screening
We reported on the method of design, synthesis and
screening of a structurally diverse sugarꢀbased VAST liꢀ
brary of 490 compounds.17 The screening of this library
against MCHꢀ1R resulted in the discovery of the potent
MCHꢀ1R antagonist ACL21823 (RLB MCHꢀ1R IC50 = 306
nM; Figure 2 and Table 1).
2.2 Modeling of MCH-1R structure
The structure of human MCHꢀ1R was modeled based on
a 2.8Å resolution crystal structure of the β2ꢀadrenergic
receptor (β2AR, PDB entry 2RH1). Multiple sequence
alignment, shown in Supplementary Figure S1, indicated
that β2AR had a higher sequence similarity with MCHꢀ1R
compared with other publiclyꢀavailable experimental
GPCR structures, with 69.5% similarity within the 7TMD
ACS Paragon Plus Environment

Journal of Chemical Information and Modeling
Page 4 of 22
A deep, narrow subꢀpocket SV1 is located between TM3,
(c) Model of MCH-1R with ACL21823 – Binding site surface
topology
5 and 6 and contains the aromatic residues Phe1973.37
,
1
2
3
4
5
6
7
8
Phe2865.47 Trp3386.48 and Tyr3426.52. Additional important
residues located in the “bottle neck” of SV1 subꢀpocket are
Gln1963.36 and Gln3456.55. The upper subꢀpocket SV3 is loꢀ
cated between TM5, 6, 7 and contains residues Phe358ECL3
,
Asn175ECL1 and Asn3637.36. The third, long subꢀpocket SV2
is generated by TM2, 3 and 7 and contains residues
Met1682.60 and Asp1923.32. The fourth central subꢀpocket
SV0 is located between subꢀpockets SV1, SV2 and SV3, is
formed by TM5, 6 and 7 and contains the residue
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Gln3486.58
.
The predicted binding mode of ACL21823 is shown in
Figure 3b. ACL21823 occupies all 4 subꢀpockets ꢀ SV1, SV2,
SV3 and SV0 ꢀ of the MCHꢀ1R binding site. In subꢀpocket
SV2, the guanidine moiety of ACL21823 can form a saltꢀ
bridge with Asp1923.32 and a hydrogen bond with the backꢀ
bone carbonyl of Met1682.60. In subꢀpocket SV3 the aromatic
moiety of ACL21823 can form a π−π interaction with
Phe358ECL3 and a halogen bond between the chlorine subꢀ
stituent and the sidechain carbonyl of Asn175ECL1. Experiꢀ
mental evidence that halogen bonds can play a key role in
binding and recognition in complexes between proteins and
ligands, has been previously reported.28 In the central subꢀ
pocket SV0 the sugar core can form a hydrogen bond with
Gln3486.48. In subꢀpocket SV1 we find an aromatic cluster
including the residues Phe1973.37, Phe2865.47 Trp3386.48 and
Tyr3426.52. The aromatic moiety of ACL21823 can form an
edgeꢀtoꢀface π−π interaction with Phe1973.37 while the
chlorine substituent can form a halogenꢀπ interaction with
the Trp3386.48 ring system.29 A nonꢀclassical hydrogen
bond can be formed between the aromatic hydrogen of
Figure 3: (a), (b) and (c) show the final model of MCH-1R
with the predicted docking pose of ACL21823. The ribbon
color coding of TMs 1 to 7 is dark orange, pink, red, purple,
dark red, orange and light yellow respectively, with ICLs and
ECLs in dark green. The carbon atoms of ACL21823 are
shown in orange and those of MCH-1R in yellow. Nitrogen
atoms are shown in blue, oxygen in red, sulfur in yellow and
chlorine in light green. (b) Interactions with key residues are
indicated by dotted lines. (c) The surface of MCH-1R binding
site is colored in green and clipped to allow a better view of
the 4 sub-pockets.
2.3 MCH-1R binding site and complex with
ACL21823
ACL21823 and the carbonyl oxygen of Gln3456.55 30
. Finalꢀ
ly, the sidechain of Gln1963.36 can form a hydrogen bond
with the oxygen linker of ACL21823.
With the MCHꢀ1R model and ACL21823 in hand, we
next turned to the prediction, refinement and exploration of
the MCHꢀ1R binding site. We docked ACL21823 into
MCHꢀ1R using flexible docking procedure. Analysis of the
10 topꢀscoring docked poses of ACL21823 revealed that
they were all very similar. Heavy atom RMSD between
docked poses did not exceed 1.5Å, indicating that a single
conserved pose had been identified. The top scoring pose
of ACL21823 was selected for further binding site refineꢀ
ment with "lowꢀmode" molecular dynamics (LowMoꢀ
deMD) simulation.
2.4 Validation of MCH-1R structure by virtual
‘enrichment’ experiment
To validate the model, a virtual ‘enrichment’ experiment
was carried out. We docked a diverse set of 126 known
MCHꢀ1R ligands, together within a library of 3,921 similar
decoy compounds. When docked compounds were ranked
by AMBER interaction energy an enrichment factor 5ꢀfold
better than random was observed at the point where 50% of
the known compounds were identified (Figure 5). Furꢀ
thermore, 90% of the known MCHꢀ1R antagonists were
ranked among the top 30% of the library based on AMBER
interaction energy. This ‘enrichment’ test demonstrated that
the MCHꢀ1R model was of high quality, as proven by its
ability to discriminate between known binders and very
similar decoys. ROCS shapeꢀbased similarity searches,
using both the docked pose of ACL21823 and its miniꢀ
mized structure as queries, failed to distinguish between
MCHꢀ1R binders and decoys (Figure 4).
27
As was previously published the MD simulation step
allows refinement of the GPCR models to a degree that is
not possible with static homology modeling. The proposed
docking pose of ACL21823 in MCHꢀ1R and the location of
the binding site are shown in Figure 3a. We observed that
the MCHꢀ1R antagonist binding site has a complicated
surface topology that allows it to accommodate large moleꢀ
cules such as ACL21823. The predicted ACL21823 bindꢀ
ing site in MCHꢀ1R is located between TM2, 3, 5, 6 and 7
(Figure 3b); it can be divided into 4 main regions (subꢀ
pockets), SV1, SV2, SV3 and SV0 (Figure 3c) which accomꢀ
modate the corresponding vectors V1, V2 and V3 and the
sugar scaffold (V0; core), shown in Figure 2, of
ACL21823.
ACS Paragon Plus Environment

Page 5 of 22
Journal of Chemical Information and Modeling
and R3, in compounds 8d and 8e did not cause a significant
1
2
variation in potency.
In compound 8f, the singleꢀpoint replacement of the R3
chlorine with fluorine resulted in a 28ꢀfold reduction in
MCHꢀ1R potency compared to ACL21823. The model
suggested several possible explanations for this. Firstly, a
potential halogen bond was observed formed between the
R3 chlorine and Asn175ECL1. Fluorine has not been obꢀ
served to form halogen bonds28, so the fluorine substitution
in 8f would result in the loss of this interaction. Secondly,
the electronic effects of the R3 substituent may also affect
the predicted π−π interaction with Phe358ECL3. While fluoꢀ
rine is more electronegative than chlorine, suggesting that it
should have a greater electronꢀwithdrawing effect, resoꢀ
nance effects mean that this is not the case in practice, and
indeed fluorine has been reported to have a weak electronꢀ
donating effect.32 Aromatic rings with electronꢀ
withdrawing groups favor faceꢀtoꢀface π−π interactions,
while rings with electronꢀdonating groups tend towards
edgeꢀtoꢀface π−π configurations. Therefore, the R3 fluorine
substituent may also affect affinity by reducing the strength
3
4
5
6
7
8
9
10
11
12
13
14
15
Figure 4: Enrichment plot for the in-silico screening of 126
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
known MCH-1R ligands embedded in a library of 3,921 similar
compounds with > 0.7 Tanimoto similarity after docking into
MCH-1R. The fraction of library compounds is ranked along
the x axis and fraction of known MCH-1R antagonists is
ranked along the y axis. The blue curve (best enrichment)
resulted from ranking by the AMBER protein-ligand interac-
tion energy, while the red and green curves resulted from
ligand-based shape similarity (ROCS) searches that used the
docked pose of ACL21823 (red) or ACL21823 minimized struc-
ture (green) as queries.
of the proposed π−π interaction with Phe358ECL3
.
Compound 8g is a hybrid of compounds 8c and 8e, with
the chlorines in positions R1 and R3 replaced by fluorine
and CF3, respectively. This compound retains similar poꢀ
tency range to the original ACL21823.
2.5 Validation of MCH-1R structure with
ACL21823 analogues
In compound 8h, as with compound 8f, a significant loss
of binding is caused by the replacement of chlorine with
fluorine in position R3. The R1 and/or R2 fluorine substituꢀ
ents appear to ameliorate this somewhat, as there is only a
11ꢀfold loss of binding for compound 8h (compared to
ACL28123), as opposed to the 28ꢀfold loss observed for 8f.
Analogues of the primary hit ACL21823 were syntheꢀ
sized for SAR exploration and validation of MCHꢀ1R. We
focused on modifications of the substituents R1, R2 (located
in subꢀpocket SV1) and R3 (located in subꢀpocket SV3), as
shown in Table 1.
Compound 8i (ACL22482) provided the highest potency
measurement of all of the VAST compounds tested. Modiꢀ
fication of the R1, R2 and R3 substituents to F, F and CF3,
respectively, resulted in a measured IC50 of 65 nM.
The compounds listed in Table 1 were synthesized as
part of a SAR library of 80 compounds, focused on the
structure of the initial hit with variation of the substituents
on the various positions of the sugar scaffold. These
ACL21823 analogues were prepared on solid phase
(Scheme 1) using standard procedures for glycosylation,
resin loading, library production and cleavage offꢀresin
In compound 8j, the singleꢀpoint replacement of the R3
chlorine with hydrogen resulted in a substantial decrease in
MCHꢀ1R binding compared to ACL21823. As with comꢀ
pound 8f, the model suggests two potential contributors to
this. Firstly, a potential halogen bond with Asn1752.67 is
lost. Secondly, as discussed previously, aromatic rings with
electronꢀwithdrawing groups, such as chlorine in the origiꢀ
nal ACL21823 hit, favor faceꢀtoꢀface π−π interactions.
The R3 hydrogen substituent may affect binding affinity by
reducing the strength of the π−π interaction with
Phe358ECL3. Toleration of R3 CF3 substituents in comꢀ
pounds 8e, 8g and 8i would appear to support this latter
hypothesis as CF3 is, like Cl, an electron withdrawing
group.
,
31
described previously.17
Based on the docking pose of
ACL21823 in MCHꢀ1R the R1, R2 and R3 substituents were
predicted to be particularly significant for the binding of
ACL21823 to MCHꢀ1R. They were predicted to affect
binding either through direct interaction with key MCHꢀ1R
residues or via nonꢀspecific contacts with the surface of the
MCHꢀ1R binding site, or through electronic effects exerted
on the aromatic rings in the ligand that modulate the
strength of the π−π interactions formed by these groups. As
the IC50s were determined from a single experiment, and
due to the inherent uncertainty of homology modeling, the
model was only used to rationalize large shifts in affinity (>
10 fold) compared to the reference compound ACL21823
(8a).
As predicted by the MCHꢀ1R model, the nature of R1, R2
and in particular R3 substituents appears to affect MCHꢀ1R
binding of the VAST hits. The independent and additive
effects of some of these substituents suggests that they inꢀ
teract with MCHꢀ1R via separate subꢀpockets, supporting
the predicted MCHꢀ1R binding site topology and modeled
binding mode in which the substituents occupy “geographꢀ
ically” distant subꢀpockets (SV1 and SV3).
In compounds 8b and 8c, the singleꢀpoint substitution of
chlorine in the R1 position with either CF3 or fluorine, reꢀ
spectively, resulted in a similar potency range to the referꢀ
ence ACL21823 (8a). Further substitution in positions R2
ACS Paragon Plus Environment

Journal of Chemical Information and Modeling
Page 6 of 22
Table 1: SAR analogues of ACL21823 and corresponding MCH-1R activity (radioligand binding).
1
2
3
4
5
6
OMe
R3
O
HO
O
R1
R2
NH
O
O
Core V0
7
Vector V1
8
9
HN
Vector V2
Vector V3
NH
H2N
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
MCH-1R
IC50 (nM)*
Compound
R1
Cl
R2
R3
8a (ACL21823)
H
H
H
F
Cl
Cl
306
417
8b
CF3
F
8c
Cl
832
8d
F
Cl
196
8e
Cl
Cl
F
H
H
H
F
CF3
F
143
8f
8724
377
8g
CF3
F
8h
8i (ACL22482)
8j
F
3410
65
N/D**
F
F
CF3
H
Cl
H
*: Each IC50 is the average result of a single experiment (2 replicates for each data point) **: N/D =
IC50 Not Determined (no activity detected at 3 ꢀM)
This study also identified a list of residues predicted to
tained for these 10 hits. MCH (Figure 5a) and ATCꢀ0175
(Figure 5b) were used as controls. The 10 VS hits showed
low 2D Tanimoto similarity (< 0.5) to the known MCHꢀ1R
antagonists and hence represent novel chemotypes of
MCHꢀ1R antagonists. EOAI3367472 (IC50 = 131 nM; Fig-
ure 5d) and EOAI3367474 (IC50 = 213 nM; Figure 5c)
were the most potent VS hits identified.
be involved in ligand binding to MCHꢀ1R: Asn175ECL1
Asp1923.32, Gln1963.36, Phe1973.37, Thr2003.40, Phe358ECL3
,
,
Phe2865.47, Gln3486.58, Trp3386.48, Tyr3426.52, Gln3456.55
and Asn3637.36. Siteꢀdirected mutagenesis studies targeting
these residues (particularly Asn175ECL1 and Phe358ECL3
)
will provide further experimental validation of the proꢀ
posed binding mode, as will biological testing of additional
VAST compounds designed to strengthen and weaken the
predicted interactions.
2.7 Prediction of key structural features involved
in binding small molecule MCH-1R antagonists
Next we used the model to generate a prediction of the
key structural features of MCHꢀ1R involved in the binding
of small organic antagonists. Information of this sort can be
useful for guiding the discovery, hit to lead and lead optiꢀ
mization phases of novel MCHꢀ1R antagonists. In order to
obtain this information we generated a training set that conꢀ
sisted of docking poses of (1) VS hits EOAI3367472 and
EOAI3367474, (2) a representative set of 32 known MCHꢀ
1R antagonists (the most potent representative per cluster
from 126 clusters of known MCHꢀ1R antagonists with Kis
or IC50s cutoff < 160 nM (details are in Supplementary
Table S1) and (3) a potent literature hit ATCꢀ017533 (IC50
= 1.11 nM; Figure 6b). Ten docked poses were produced
for each molecule and ranked by AMBER interaction enerꢀ
gy. The top ranked pose was retained for further study;
other poses were also inspected visually and were found to
be similar to the topꢀranked pose, or to have fewer interacꢀ
tions with the protein.
2.6 Structure-based virtual screening
A virtual screening (VS) library of 45K compounds with
≤ 0.7 Tanimoto similarity to the 126 known MCHꢀ1R anꢀ
tagonists was extracted from a commercial vendor library
of 16M compounds. This was done to ensure that only
MCHꢀ1R binders with novel chemotypes would be identiꢀ
fied. The same GOLD docking procedure and calculation
of AMBER interaction energy as used in the “enrichment”
validation test was applied to docking, scoring and ranking
of these 45K compounds. The top 70 compounds from the
ranked list, which showed AMBER interaction energies in
the range of the top 10% of ranked known MCHꢀ1R antagꢀ
onists, were selected for biochemical screening. These 70
compounds were purchased from vendors and experimenꢀ
tally tested in an aequorinꢀbased inhibition assay for MCHꢀ
1R. This in-vitro assay confirmed that 10 of the 70 comꢀ
pounds tested were hits with % inhibition > 80% at 10ꢀM,
reflecting a 14% hit rate. Dose response curves were obꢀ
ACS Paragon Plus Environment

Page 7 of 22
Journal of Chemical Information and Modeling
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
Figure 5: MCH-1R dose response curves. (a) MCH and (b) ATC-0175 used as controls, (c) EOAI3367474 (IC50 = 213 nM) and (d)
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
EOAI3367472 (IC50 = 131 nM). Points represent mean±SEM values from single experiments with 2 replicates.
Docked poses for the VS hits EOAI3367472 and
EOAI3367474 are shown in Figure 6a and Figure 6b, reꢀ
spectively. Both compounds were predicted to have distinct
binding modes that access three of the four subꢀpockets
occupied by the VAST hits. The docked pose of the potent
hit ATCꢀ0175 is also illustrated here for the purpose of
comparison with the VS hits (Figure 6c).
left unfilled by the VAST compounds. It also partially acꢀ
cesses SV3. In subꢀpocket SV0, the aromatic moiety of
EOAI3367474 can form an edgeꢀtoꢀface π−π interaction
with the SV3 residue Phe358ECL3. The carbonyl oxygen of
the amide linker can accept a hydrogen bond from the SV3
residue Asn3637.36, and the NH of the amide linker can
form a hydrogen bond with Gln3456.55. In subꢀpocket SV2, a
hydrogen bond is potentially formed between the pyridine
nitrogen of EOAI3367474 and Gln1963.36. There is a poꢀ
tential for a saltꢀbridge between the basic piperazine nitroꢀ
gen and Asp1923.32; however, the compound contains mulꢀ
tiple ionizable centers with similar pKas and it is likely that
multiple microspecies will be present. Not all of them will
have this nitrogen protonated, reducing the significance of
this interaction. A further hydrogen bond can be generated
between the sidechain oxygen of Tyr3707.43 and the OH
group of EOAI3367474. Finally, the isopropyl portion of
EOAI3367474 appears to pack against the sidechain of
Phe1612.53, making either a generic hydrophobic contact or
potentially a CHꢀπ interaction.
The literature antagonist ATCꢀ0175 adopts a similar
docked pose to EOAI3367472, occupying subꢀpockets SV0,
SV1, and SV3 (Figure 6c). In the central subꢀpocket SV0, the
cyclohexane scaffold of ATCꢀ0175 is predicted to form
several nonꢀspecific hydrophobic contacts with MCHꢀ1R,
and a hydrogen bond can be formed between the NH linker
and Gln3486.58. In subꢀpocket SV1 the aromatic moiety of
ATCꢀ0175 can form an edgeꢀtoꢀface π−π interaction with
Trp3386.48 while one fluorine substituent can form a hydroꢀ
EOAI3367472 is predicted to occupy subꢀpockets SV0,
SV1, and SV3 (Figure 6a); SV2 does not appear to be accessed
by this molecule. In subꢀpocket SV3, the docked pose shows
the aromatic moiety of EOAI3367472 forming a faceꢀtoꢀ
face π−π interaction with Phe358ECL3. Hydrogen bonds are
predicted to occur between the basic nitrogen of the piperaꢀ
zine moiety and Asn175ECL1, and also between Asn3637.36
and the nonꢀbasic piperazine nitrogen. In the central subꢀ
pocket SV0, the carbonyl oxygen of the amide linker can
accept a hydrogen bond from Gln3486.58. In subꢀpocket SV1
the aromatic moiety of EOAI3367472 can form an edgeꢀtoꢀ
face π−π interaction with Trp3386.48 and Phe1973.37while
the fluorine substituent potentially forms a weak hydrogen
bond with the OH group of the Tyr3426.52 side chain. It is
not clear from our model what types of interactions might
exist between the thiophene moiety of EOAI3367472 and
Gln1963.36 and Gln3456.55 residues.
EOAI3367474 is predicted to have a different binding
mode to EOAI3367472 (Figure 6b). It does not appear to
access SV1, but seems to occupy subꢀpockets SV0 and SV2. In
the case of SV2, docking shows EOAI3367474 occupying an
extension to the subꢀpocket between TM2 and TM3 that is
ACS Paragon Plus Environment

Journal of Chemical Information and Modeling
Page 8 of 22
gen bond with Tyr3426.52
. Two hydrogen bonds can be
(c) MCH-1R bound to ATC-0175 (MCH-1R IC50 = 1.11 nM)
formed by the amide linker of ATCꢀ0175, with Gln1963.36
and Gln3456.55. In subꢀpocket SV3 the quinazoline moiety of
ATCꢀ0175 can form a faceꢀtoꢀface π−π interaction with
Phe358ECL3 and a hydrogen bond with Asn175ECL1. While
ATCꢀ0175 and EOAI3367472 adopt similar docked poses,
they exhibit a 100ꢀfold difference in potencies. The model
suggests a possible rationale for this difference, as ATCꢀ
0175 is predicted to form 5 hydrogen bonds with MCHꢀ1R,
while EOAI3357472 would form only 3.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
While the VAST hits are predicted to occupy all four subꢀ
pockets of the MCHꢀ1R antagonist site, docked poses for
ATCꢀ0175, EOAI3367472 and EOAI3367474 only acꢀ
cessed three of them. In addition, EOAI3367474 occupied
a different subset of the subꢀpockets to ATCꢀ0175 and
EOAI336742.
Figure 6: Docked poses of MCH-1R with EOAI3367472 (a),
EOAI3367474 (b) and ATC-0175 (c).The ribbon color coding
of TMs 1 to 7 is dark orange, pink, red, purple, dark red, or-
ange and light yellow respectively, with ICLs and ECLs in
dark green. The ligands carbon atoms are shown in orange
and those of MCH-1R in yellow. Nitrogen atoms are shown in
blue, oxygen in red, sulfur in yellow and fluorine in light
green. Interactions with key residues are indicated by dotted
lines and the surface of MCH-1R binding site is color in green
and clipped to allow a better view of the four sub-pockets. In
(b), the surface of the extra space in the SV2 pocket is colored
in blue. Trp338 is marked as (*) in (a) and (c) to avoid ob-
scuring the image.
(a) MCH-1R bound to VS hit EOAI3367472 (MCH-1R IC50
=
131 nM)
A further 32 published and potent MCHꢀ1R antagonists
were docked into the model in order to investigate how
frequently each subꢀpocket was predicted to be involved in
ligand binding. The results of this analysis are shown in
Figure 7. This analysis showed that 61% of the 33 antagoꢀ
nists (ATCꢀ0175 + 32 other literature compounds) are preꢀ
dicted to occupy 3 of the four subꢀpockets. In 36% of cases
the antagonists only occupied 2 subꢀpockets. Potent small
molecule antagonists that accessed all four subꢀpockets
were, according to the model, very rare, only accounting
for a single example. The SV0 subꢀpocket was occupied by
the docked poses of all 33 antagonists. The central position
of the SV0 subꢀpocket makes it an important junction allowꢀ
ing access to neighboring subꢀpockets. Due to the relatively
small number of data points, and as the reported potencies
for the literature compounds were taken from different asꢀ
says, it was not possible to provide a clear prediction as to
which of the SV1, SV2 or SV3 subꢀpockets are most significant
with regards to binding potency, but it does suggest that
potent, smallꢀmolecule MCHꢀ1R antagonists do not need to
occupy all 4 subꢀpockets occupied by the VAST comꢀ
pounds.
(b) MCH-1R bound to VS hit EOAI3367474 (MCH-1R IC50
=
213 nM)
We also calculated the frequency of interaction of each
of the binding site residues with docked small molecule
ligands. This exercise was performed to identify a list of
residues predicted to be involved in small ligand binding,
which could be used to prioritize future siteꢀdirected mutaꢀ
genesis studies and to suggest further development of
EOAI3367472 and EOAI3367474. The same training set of
ACS Paragon Plus Environment

Page 9 of 22
Journal of Chemical Information and Modeling
33 MCHꢀ1R potent binders detailed above was used. In
Phe358ECL3, Phe2865.47, Gln3486.58, Trp3386.48, Tyr3426.52
,
subꢀpocket SV3, 100% of the potent antagonists that occuꢀ
pied the SV3 subꢀpocket were predicted to form a π−π interꢀ
action with Phe358ECL3. Phe358ECL3 was also involved in
π−π interactions with 11% of the antagonists predicted to
occupy subꢀpockets other than the SV3 subꢀpocket.
Asn175ECL1 was predicted to be involved in a hydrogen or
halogen bond with 53% of the antagonists with docked
poses that occupied SV3. Asn3637.36 is located at the border
between subꢀpockets SV3 and SV0 and observed to form hyꢀ
drogen bonds with 61% of antagonists that dock in SV3, and
also with 22% of antagonists where the pose did not access
SV3. In the SV0 subꢀpocket the residue Gln3486.58 was preꢀ
dicted to form a hydrogen bond with antagonists in 35% of
cases. Involvement of Tyr275ECL3 in binding of MCHꢀ1R
antagonists was also predicted in 8% of cases. The SV1 subꢀ
Gln3456.55 and Asn3637.36. The fact that MCHꢀ1R binds
ligands of many chemotypes and sizes can be explained by
the observation that MCHꢀ1R antagonists have a large seꢀ
lection of key residues to bind to and the freedom to occuꢀ
py multiple combinations of subꢀpockets SV1, SV2 or SV3.
1
2
3
4
5
6
7
8
3. CONCLUSION
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Obesity is a chronic global epidemic disease and MCHꢀ
1R receptor antagonism remains a viable approach for obeꢀ
sity treatment.7 However the progress in discovery of
MCHꢀ1R antagonists depends on the availability of molecꢀ
ularꢀlevel insights on the MCHꢀ1R binding site. Here we
present a new practical approach which opens an alternaꢀ
tive avenue for the structural exploration of GPCRs and
structureꢀbased GPCR drug discovery. The practicality and
efficiency of this approach is enhanced by generation of
structural information on the binding site of a targeted
GPCR, to satisfy the immediate need of the drugꢀdiscovery
process for such information and to drive structureꢀbased
drug design. This conceptually pioneering approach was
applied to exploration of the MCHꢀ1R structure and resultꢀ
ed in the discovery of 10 new antagonist chemotypes, obviꢀ
ating the need for running an expensive and time consumꢀ
ing HTS. The usefulness of this approach to drug discovꢀ
ery, demonstrated in the complicated case of MCHꢀ1R,
should now allow SBDD even for some of the most chalꢀ
lenging GPCR targets. The results presented here are comꢀ
parable to those from reports which demonstrate that
GPCR modeling supported by experimental data in the
absence of a crystal structure can be a valid replacement for
structural and functional exploration of GPCRs and for
structureꢀbased drug discovery.34ꢀ38
pocket contains the aromatic residues Phe1973.37
,
Phe2865.47 Trp3386.48 and Tyr3426.52, which formed interacꢀ
tions with all docked poses of small MCHꢀ1R antagonists
that occupied the SV1 subꢀpocket. Unfortunately, the MCHꢀ
1R model could not separate the individual contributions of
each of these residues. The residues Gln1963.36 and
Gln3456.55, located on the border between subꢀpockets SV0
and SV1, were observed to be involved in hydrogen bonding
in 71% and 64% of cases respectively when docked poses
accessed the SV1 subꢀpocket, and in 32% and 15% of cases
respectively when SV1 was not occupied. In the SV2 subꢀ
pocket, the analysis suggested that potent MCHꢀ1R antagꢀ
onists predicted to occupy this pocket would mainly interꢀ
act with residues Asp1923.32 (72% frequency), Met1682.60
(35% frequency), Phe1612.53 (25% frequency) and
Tyr3707.43 (32% frequency).
Ligand SAR data, when combined with modeling, can
provide a useful source of structural information on GPCR
binding sites. The VAST compounds are designed to be
highly efficient in exploring GPCR binding sites and the
experimental information they provide allows the construcꢀ
tion of high quality GPCR models. The large size, comꢀ
plexity and rigidity of the VAST hits improves the predicꢀ
tion of binding site location, topology and key residues by
limiting the number of feasible locations and orientations in
which these hits can be docked into the receptor. In the
case of MCHꢀ1R, only one feasible binding site and bindꢀ
ing mode was found for the VAST hits. Optimization of the
GPCR model using VAST hits does not bias the GPCR
binding site towards a single ligand chemotype, a common
problem observed when performing the optimization with
known ligands.14 For this reason, we obtained high enrichꢀ
ment for all chemotypes of known MCHꢀ1R antagonists
over chemically similar decoys, and high performance in
virtual screening.
Figure 7: Plot based on analysis of docking poses for 33 high-
ly potent MCH-1R antagonists to determine sub-pocket oc-
cupancy.
We have demonstrated that the peptide’s bioactive conꢀ
formation and motifs can be mimicked by appending subꢀ
stituents onto a pyranoseꢀsugar scaffold in a way that posiꢀ
tions these elements in a similar relative orientation. These
compounds are ideal for exploring the structural and chemꢀ
This exercise provided a shortlist of potential key resiꢀ
dues, identified from the modeling of the MCHꢀ1R strucꢀ
ture with literature antagonists, that were predicted to be
involved in small molecule ligand binding to MCHꢀ1R:
Asn175ECL1, Asp1923.32, Gln1963.36, Phe1973.37, Thr2003.40
,
ACS Paragon Plus Environment

Journal of Chemical Information and Modeling
Page 10 of 22
ical features required for compounds to interact with GPCR ment (2) Template selection (3) Homology Modeling of
1
2
3
4
5
6
7
8
targets – particularly those GPCRs with large endogenous
ligands, e.g. peptides or hormones – and for the discovery
of ligands with novel chemotypes. The work described here
was based on the original VAST library of 490 compounds
that was biased toward specific motifs (represented by the
general motif AromaticꢀAromaticꢀPositive Charge), targetꢀ
ing selected receptors including MCHꢀ1R. A larger VAST
library that is systematically diverse in motif and chemoꢀ
form is now available (14,000 compounds), thus providing
an exciting opportunity to broaden the applicability of this
approach to any other GPCR.
MCH-1R (4) Docking procedure and generation of
ACL21823 complex (5) MCH-1R binding site optimization
with "lowꢀmode" molecular dynamics (LowModeMD)
simulation.
Multiple-Sequence Alignment is required for selection of
the optimum template from the available GPCR crystal
structures for further homology modeling of MCHꢀ1R. The
amino acid sequence for human MCHꢀ1R (Q99705) was
retrieved from the SwissProt database (Figure 8). The seꢀ
quence of MCHꢀ1R was aligned with four published GPCR
crystal structures [human dopamine D3 receptor (D3, PDB
entry 3PBL), β2ꢀadrenergic receptor (β2AR, PDB entry
2RH1), human A2A adenosine receptor (A2A, PDB entry
3EML) and bovine rhodopsin (PDB entry 1F88)], using a
multipleꢀsequence alignment tool implemented in MOE
version 2010.10 (Chemical Computing Group). In this apꢀ
proach, originally introduced by Needleman in 1970,41
alignments were computed by optimizing a function based
on residue similarity scores obtained from applying an
amino acid substitution matrix (blosum62)42 to pairs of
aligned residues and gap penalties. Penalties were imposed
for introducing and extending gaps in one sequence with
respect to another. The gap start penalty was set at 7 and
the penalty for gap extension was set at 1. The conserved
residues and conserved GPCR motifs were constrained to
ensure their proper alignment. The position of each amino
acid was identified by its sequence number and by the geꢀ
neric number proposed by Ballesteros and Weinstein.
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
In this work we aimed to use homology modeling, reꢀ
fined by VAST ligand directed modeling, to identify the
MCHꢀ1R structural features involved in the binding of
small organic antagonists. In spite of the inherent limitaꢀ
tions of modeling, such information can be usefully applied
to the future discovery and lead optimization of MCHꢀ1R
antagonists and of GPCR ligands in general. Based on our
analysis it seems that MCHꢀ1R antagonists can be classiꢀ
fied not only based on their chemotype but also on the subꢀ
pockets that they occupy and the unique interactions that
they form with the MCHꢀ1R binding site residues. While
the model suggests that there are four distinct subꢀpockets
in the MCHꢀ1R binding site, modeling of known antagoꢀ
nists indicates that it is not necessary for compounds to
occupy all of these four subꢀpockets. Two or three subꢀ
pockets are sufficient for potent antagonism, allowing the
design of smaller compounds that occupy a more drugꢀlike
property space in comparison with larger molecules. Deꢀ
sign and optimization of novel MCHꢀ1R antagonists that
already access two or three subꢀpockets can therefore be
focused on optimizing interactions within those subꢀ
pockets, rather than expansion into additional parts of the
site.
Template selection of the best template for modeling of
any GPCR is usually very challenging.43 Class A GPCRs
share the same arrangement of the seven helices44 and their
7TMD sequence similarities are relatively high.45ꢀ48 Howꢀ
ever, even small sequence differences can lead to signifiꢀ
cant differences in overall structure and particularly in the
topology of the ligand binding site.49 This renders each
GPCR unique to its exclusive biological function. The most
critical sequence difference in GPCRs is the difference in
the positions of proline residues.50ꢀ52 The prolines force
kinks in TM secondary structure and, as a result, even the
smallest difference in the positions of prolines in the seꢀ
quence alignment of the modeled GPCR and the template
can result in a significant decrease in the accuracy of the
model.51 We ranked the quality of crystal structures as poꢀ
tential templates for homology of MCHꢀ1R based on the
maximum number of correctly aligned prolines in the
7TMD. The crystal structure that had the highest number of
aligned prolines was chosen as a template for further hoꢀ
mology modeling.
EXPERIMENTAL SECTION
Computational Methods
MCH-1R Residue Numbering. The position of each amiꢀ
no acid residue of MCHꢀ1R was identified both by its seꢀ
quence number and by its generic number proposed by
Ballesteros and Weinstein.39 Briefly, in this numbering
system, amino acid residues in the 7 transꢀmembrane doꢀ
main (7TMD) are given two numbers; the first corresponds
to the transꢀmembrane helix (TM) number (1 to 7), while
the second indicates the residue position relative to a highly
conserved residue in class A GPCRs in that TM, which is
arbitrarily assigned to 50. The numbering of the loops is
done in a similar manner, for example extracellular loop 2
(ECL2) is labelled 45 to indicate its location between heliꢀ
ces 4 and 5, and the conserved cysteine (thought to be part
of a disulfideꢀbond) is given the index number 45.50. The
residues within the ECL2 loop are then numbered relative
to this position.
Homology Modeling of MCH-1R. Modeling of MCHꢀ1R
was performed using the homology modeling tool as imꢀ
plemented in the MOE software package (Chemical Comꢀ
puting Group, version 2010.10), based on the template seꢀ
lected in the previous stage. Due to insertions of the MCHꢀ
1R sequences with respect to the template, some residues,
particularly in the loop regions, did not have assigned
backbone geometries based on the template. These inserꢀ
tions were modeled from segments of highꢀresolution
chains from the protein data bank (PDB) which superposed
MCH-1R modeling procedure
We performed our hierarchical GPCR modeling protoꢀ
col40 in 5 sequential steps: (1) Multiple-Sequence Align-
ACS Paragon Plus Environment

Page 11 of 22
Journal of Chemical Information and Modeling
well onto anchor residues on each side of the insertion area, affinities between ligand and protein and was therefore
after the method described by Fechteler et al.53 Following
the selection of appropriate loop templates, multiple model
candidates for each loop were constructed and scored using
a OPLSꢀAA energy function.54, 55 The coordinates of the
top ranked loop model were added to the global model.
After all of the loops had been added, the side chains were
modeled. Sidechain data is assembled from an extensive
rotamer library generated by systematic clustering of highꢀ
resolution PDB data. After all of the backbone segment and
sidechain conformations were chosen, the model was minꢀ
imized using the OPLSꢀAA54, 55 forceꢀfield.56
selected as a reliable method to reꢀscore and to rank dockꢀ
ing poses. The topꢀranked docking pose, according to the
AMBER interaction energy, was taken for further analysis.
Except when docking ACL21823 a rigid docking procedure
was used, with the backbone and sidechains of MCHꢀ1R
restrained during docking.
1
2
3
4
5
6
7
8
When docking ACL21823 a flexible GOLD docking
procedure was used, as ACL21823 was being docked into
an unrefined binding site. We assigned flexibility to all
aromatic residues using the GOLD rotamer library to enꢀ
sure that the docking procedure could adjust the site in orꢀ
der to accommodate large molecules such as ACL21823.
Since the exact location of the MCHꢀ1R binding site is not
known we initially defined a large region towards the exꢀ
tracellular end of the MCH1ꢀR 7TMD as a potential bindꢀ
ing site. Analysis of the 10 topꢀranked docking poses of
ACL21823 revealed that they all share high similarity;
pairwise heavyꢀatom RMSD was less than 1.5Å2 for all
pairs of poses. The topꢀranked pose of ACL21823, accordꢀ
ing to the AMBER interaction energy, was selected for
further binding site refinement.
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
MCH-1R binding site optimization. In the final MCHꢀ1R
modeling stage, we aimed to optimize the MCHꢀ1R bindꢀ
ing site of MCHꢀ1R by applying a “lowꢀmode” molecular
61
dynamics simulation (LowModeMD
)
to docked
ACL21823 pose. The LowModeMD protocol is a stochasꢀ
tic conformation generation protocol implemented in MOE.
For the LowModeMD refinement the MCHꢀ1R binding site
was defined by residues within a radius of 7.0Å of the
docked ACL21823. Flexibility was permitted for all atoms
within this radius when the rest of the atoms of MCHꢀ1R
and of ACL21823 were restrained. The radius of 7.0Å was
selected to insure that all the atoms of helical fraction that
include close to ACL21823 residues will be also included
in refinement. The OPLS–AA forceꢀfield was used for the
conformational search as the MOE AMBER implementaꢀ
tion did not contain appropriate atom types for the ligand.
The dielectric constant was set to 3. A LowModeMD conꢀ
stant temperature MD simulation was performed at 300K,
using the Berendsen thermostat62 and the velocity Verlet
algorithm. The default value of the energy minimisation
gradient (0.001 kcal.mol−1.Å−2) was used. The LowMoꢀ
deMD and stochastic searches were terminated after 200
failed attempts to generate a new conformation, with a
maximum of 10,000 iterations. To ensure that the TM heliꢀ
ces did not “unwind” during the optimization, simple harꢀ
monic distance constraints were applied to mimic the αꢀ
helical i, i + 4 carbonyl−amine hydrogen bonds.
Figure 8: Amino-acid sequence of human MCH-1R, the color
coding for TM 1 to 7 is dark orange, pink, red, purple, dark
red, orange and light yellow respectively. The conserved
residues in each TM that are assigned to 50 by the Ballester-
os-Weinstein numbering scheme are shown by black arrows.
Docking procedure and generation of ACL21823 com-
plex. In all docking experiments described in this manuꢀ
script we used the GOLD docking package (Cambridge
Crystallographic Data Centre, version 5.0), followed by a
consistent reꢀscoring and reꢀranking procedures. The dockꢀ
ing itself was performed once for each molecule, with the
10 top ranked docking poses scored by the GOLD default
scoring function57 retained. We then rescored and reꢀranked
Validation of the MCH-1R model
these 10 docked poses using AMBER interaction energy.
Virtual ‘enrichment’ experiment. For primary validation
of the MCHꢀ1R model a virtual enrichment experiment was
carried out. This step was designed to test the ability of the
MCHꢀ1R model to separate known MCHꢀ1R antagonists
from a library of chemically similar analogues, and to deꢀ
termine if the quality of the MCHꢀ1R model was high
enough to be used in structureꢀbased virtual screening. To
generate a representative library of MCHꢀ1R antagonists, a
total of 2,723 potent MCHꢀ1R antagonists were extracted
58
We used the MM_PBSA/GBSA approach
to calculate
the AMBER interaction energy (see details in Supplemen-
tary Materials), in order to take the entropy and salvation
energy change upon binding into account, as well as the
binding enthalpy between protein and ligand. As was reꢀ
cently published 59, 60 the AMBER interaction energy, while
subject to the same limitations as all force field based
methods, was able to accurately predict relative binding
ACS Paragon Plus Environment

Journal of Chemical Information and Modeling
Page 12 of 22
from the Liceptor database63 and clustered using a spectral
ly dried at 300ºC) under N2. After 1 h, a solution of
clustering method.64 One representative from each of the
126 clusters identified was retained, providing exhaustive
coverage of all available MCHꢀ1R antagonist chemotypes.
For the decoys, a set of 3,921 compounds was extracted
from commercial vendor libraries using Tripos UNITY
fingerprints,65 with a Tanimoto similarity cutoff to the
known 126 MCHꢀ1R binders > 0.7. The use of decoys simꢀ
ilar to the known compounds provided a challenging valiꢀ
dation exercise, success in which would indicate a highꢀ
quality MCHꢀ1R model. The library of 4,047 compounds
(126 known binders and 3,921 decoys) was docked into the
MCHꢀ1R model with the binding site refined based on
ACL21823. The docked poses were reꢀscored and ranked
by proteinꢀligand interaction energies calculated with the
AMBER force field as described in previous sections. For
comparison, compounds were also ranked with ROCS, a
shapeꢀbased similarity method (see details in Supplemen-
tary Materials), using both the docked pose of ACL21823
and its minimized structure as queries.
DMTST (0.5 M, 2.5 equiv.) in dry DCM was added to the
sugar and alcohol mixture and the resulting solution was
stirred under N2 at RT. Evolution of reaction was checked
by TLC and LC/MS. After complete reaction of starting
material (1ꢀ3 h), the reaction mixture was filtered on Celite,
and the Celite was washed twice with DCM. The organic
phase was washed three times with a saturated NaHCO3
solution. The aqueous phase was reꢀextracted once with
DCM. The combined organic phases were dried over
MgSO4 and, after filtration, evaporated under reduced presꢀ
sure to afford a crude residue, which contained desired α/β
anomers in both TBDPSꢀprotected and unprotected form.
To reprotect the 6ꢀposition with the TBDPS group, the
crude material was treated with imidazole (7 equiv. to startꢀ
ing building block 1) and TBDPSꢀCl (4.5 equiv. to starting
building block 1) in dry DMF (7 ml per gram of starting
material). The mixture was stirred under N2 at RT and evoꢀ
lution of reaction was checked by TLC and LC/MS. After
completion (typically 2 h), the reaction mixture was diluted
with DCM and the solution was washed with distilled waꢀ
ter. The water layer was reꢀextracted three times with DCM
and all organic layers were combined, dried over MgSO4
and, after filtration, evaporated under reduced pressure.
The αꢀ and βꢀanomers were separated and purified from
the recovered residue by column chromatography.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Validation of MCH-1R structure with ACL21823 ana-
logues. To further validate the model, we synthesized
ACL21823 analogues with various R1, R2 and R3 substituꢀ
ents (see Table 1), predicted to be highly significant for
MCHꢀ1R binding. Synthesis of these analogues is depicted
in Scheme 1 and described below.
(4’-Chlorobenzyl)-2-azido-3-O-benzoyl-6-O-(tert-
butyldiphenylsilyl)-2-deoxy-1-α-D-glucopyranoside
(2a). GlcSMe 1 (5.0 g, 8.7 mmol) and 4ꢀchlorobenzyl alcoꢀ
hol (6.2 g, 43.3 mmol) were used in the glycosylation reacꢀ
tion as described above. The crude material was purified by
column chromatography using PE/EtOAc (95/5) as eluent
Structure-based virtual screen
This step aimed to discover a novel chemotype for potent
MCHꢀ1R ligands. A virtual library of 45K compounds with
Tanimoto similarity ≤ 0.7 to the 126 known MCHꢀ1R anꢀ
tagonists was extracted from commercial vendor cataꢀ
logues. We used the same docking and scoring procedure
as the enrichment study to rank the library compounds and
select a shortlist of highly ranked virtual hits for testing.
These 70 compounds were purchased and tested in an aeꢀ
quorinꢀbased inhibition assay as described below.
1
to give the αꢀanomer 2a as a white solid (1.45 g, 25%): H
NMR (400 MHz, CDCl3)
JH4,OH = 3.5 Hz, OH), 3.40 (dd, 1H, JH1,H2 = 3.6 Hz,
H2,H3 = 10.6 Hz, Hꢀ2), 3.81ꢀ3.89 (m, 2H, Hꢀ4, Hꢀ5), 3.90ꢀ
δ 1.08 (s, 9H, CH3), 2.87 (d, 1H,
J
3.97 (m, 2H, Hꢀ6a, Hꢀ6b), 4.55 (d, 1H, JA,B = 12.2 Hz,
OCH2(A)), 4.72 (d, 1H, OCH2(B)), 5.04 (d, 1H,
Synthesis of VAST compounds
The 490 VAST library compounds as well as the
ACL21823 analogues were prepared following previously
J
H1,H2 = 3.5 Hz, Hꢀ1), 5.60 (dd, 1H, JH2,H3 = 10.4 Hz,
JH3,H4 = 8.4 Hz, Hꢀ3), 7.28ꢀ8.12 (m, 19H, HAr); MS (ES)
,
31
published solid phase synthetic methods.17 The synthetic
route for the preparation of the ACL21823 analogues is
depicted in Scheme 1.
m/z 694.14 [M+Na]+
(4’-Fluorobenzyl)-2-azido-3-O-benzoyl-6-O-(tert-
butyldiphenylsilyl)-2-deoxy-1-α-D-glucopyranoside
(2b). GlcSMe 1 (5.1 g, 8.8 mmol) and 4ꢀfluorobenzyl alcoꢀ
hol (4.8 mL, 44.5 mmol) were used in the glycosylation
reaction as described above. The crude material was puriꢀ
fied by column chromatography using PE/EtOAc (92/8) as
General procedure for glycosylation reaction
DMTST (dimethyl(methylthio)sulfonium triflate) prepa-
ration. Methyldisulfide (30 mL, 0.34 mol) was diluted in
300 mL dry DCM under N2. Methyl triflate (50 g,
0.30 mol) was added dropwise at RT to the stirred methylꢀ
disulfide solution, over 15 min. The resulting solution was
stirred at RT overnight under N2. The reaction mixture was
then poured into 1.4 L dry diethyl ether, the mixture was
stirred well and then left to stand for 30 min. The resulting
white precipitate was collected by filtration, washed with
dry diethyl ether and dried under high vacuum overnight
(74.4 g of white solid, 95%).
eluent to give the αꢀanomer 2b as a white solid (1.44 g,
1
25%): H NMR (400 MHz, CDCl3)
δ 1.08 (s, 9H, CH3),
2.89 (d, 1H, JH4,OH = 3.8 Hz, OH), 3.39 (dd, 1H,
JH1,H2 = 3.6 Hz, JH2,H3 = 10.6 Hz, Hꢀ2), 3.81ꢀ3.86 (m, 2H,
Hꢀ4, Hꢀ5), 3.92ꢀ3.94 (m, 2H, Hꢀ6a, Hꢀ6b), 4.55 (d, 1H,
JA,B = 11.9 Hz, OCH2(A)), 4.72 (d, 1H, OCH2(B)), 5.04 (d,
1H, JH1,H2 = 3.5 Hz, Hꢀ1), 5.60 (dd, 1H, JH2,H3 = 10.4 Hz,
JH3,H4 = 8.8 Hz, Hꢀ3), 7.01ꢀ7.06 (m, 2H, HAr), 7.31ꢀ8.11 (m,
17H, HAr); MS (ES) m/z 678.13 [M+Na]+.
Glycosylation reaction. Orthogonally protected building
block 1 (GlcꢀSMe) and alcohol R1OH (5 equiv.) were disꢀ
solved in dry DCM (6 mL per gram of 1), and the mixture
was stirred at RT over molecular sieves (AW300, previousꢀ
ACS Paragon Plus Environment

Page 13 of 22
Journal of Chemical Information and Modeling
Scheme 1: Synthesis of ACL21823 analogues on solid phase.
1
2
3
4
5
6
7
OTBDPS
O
OTBDPS
O
OTBDPS
O
2a - R1=Bn4Cl
2b - R1=Bn4F
2c - R1=Bn4CF3
2d - R1=Bn3,4diF
R1OH
HO
BzO
HO
BzO
HO
BzO
OR1
SMe
+
DMTST
MS AW300
DCM/Et2O
N3
N3
N3
OR1
2 (α)
2 (β)
1
8
9
O
CCl3
NH
OTBDPS
O
OTBDPS
O
OTBDPS
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
1. NaOMe, MeOH, THF
2. LiOtBu, R3X, DMF
O
O
HO
O
BzO
BzO
R3O
BF3.Et2O
DCM
N3
N3
N3
OR1
OR1
OR1
2 (α)
3
4
OMe
O
OMe
1. PSHF, AcOH, DMF
2. LiOtBu, MeI, DMF
1. LiOtBu, DTT, DMF
1. Piperidine, DMF
N
O
O
O
R3O
R3O
2. R2COOH, HBTU
DIPEA, DMF
N3
NH
OR1
OR1
O
2.
N
R2
NH .HNO3
5
6
H2N
(R2=(CH2)3NHFmoc)
DIPEA, DMF
OMe
OMe
O
1. TFA, TES, DCM
2. NH3, MeOH
O
O
HO
R3O
R3O
NH
NH
OR1
OR1
O
O
HN
HN
NH
NH
H2N
H2N
7
8
alcohol (3.65 mL, 31.9 mmol, 3.7 equiv.) were used in a
modified glycosylation procedure. Et2O/DCM (2:1) was
used as the solvent and DMTST (2 equiv.) was added as a
solid. After overnight reaction at 0°C, the reaction mixture
was diluted with EtOAc and worked up as described in the
general glycosylation procedure. TBDPS reprotection was
not necessary in this case. The crude material was purified
by column chromatography using PE/EtOAc (90/10 to
0/100) as eluent to give the αꢀanomer 2d as a white solid
(4’-Trifluoromethylbenzyl)-2-azido-3-O-benzoyl-6-O-
(tert-butyldiphenylsilyl)-2-deoxy-1-α-D-glucopyranoside
(2c).
GlcSMe
trifluoromethylbenzyl
1
(5.1 g,
alcohol
8.9 mmol)
(5.0 g,
and
4ꢀ
28.4 mmol,
3.2 equiv.) were used in a modified glycosylation proceꢀ
dure. Et2O (80 mL) was used as the solvent instead of
DCM and DMTST (2 equiv.) was added as a solid. After
overnight reaction at RT, the reaction mixture was diluted
with DCM and worked up as described in the general glyꢀ
cosylation procedure. TBDPS reprotection was not necesꢀ
sary in this case. The crude material was purified by colꢀ
umn chromatography using PE/EtOAc (95/5 to 90/10) as
(1.98 g, 34%): 1H NMR (400 MHz, CDCl3)
δ 1.08 (s, 9H,
CH3), 2.88 (d, 1H, JH4,OH = 3.6 Hz, OH), 3.42 (dd, 1H,
JH1,H2 = 3.2 Hz, JH2,H3 = 10.4 Hz, Hꢀ2), 3.80ꢀ3.88 (m, 2H,
Hꢀ4, Hꢀ5), 3.91ꢀ3.98 (m, 2H, Hꢀ6a, Hꢀ6b), 4.51 (d, 1H,
JA,B = 12.0 Hz, OCH2(A)), 4.69 (d, 1H, OCH2(B)), 5.03 (d,
1H, JH1,H2 = 3.5 Hz, Hꢀ1), 5.59 (dd, 1H, JH2,H3 = 10.4 Hz,
JH3,H4 = 8.0 Hz, Hꢀ3), 7.08ꢀ7.21 (m, 3H, HAr), 7.38ꢀ8.11 (m,
15H, HAr); MS (ES) m/z 696.14 [M+Na]+.
eluent to give the αꢀanomer 2c as a white solid (2.82 g,
1
45%): H NMR (400 MHz, CDCl3)
δ 1.08 (s, 9H, CH3),
2.87 (d, 1H, JH4,OH = 3.5 Hz, OH), 3.44 (dd, 1H,
JH1,H2 = 3.2 Hz, JH2,H3 = 10.4 Hz, Hꢀ2), 3.81ꢀ3.90 (m, 2H,
Hꢀ4, Hꢀ5), 3.91ꢀ3.95 (m, 2H, Hꢀ6a, Hꢀ6b), 4.64 (d, 1H,
JA,B = 12.6 Hz, OCH2(A)), 4.82 (d, 1H, OCH2(B)), 5.07 (d,
1H, JH1,H2 = 3.5 Hz, Hꢀ1), 5.61 (dd, 1H, JH2,H3 = 10.4 Hz,
JH3,H4 = 8.4 Hz, Hꢀ3), 7.36ꢀ8.12 (m, 19H, HAr); MS (ES)
m/z 728.10 [M+Na]+.
General solid-phase procedures
Resin loading. Commercial TCA Wang resin (Loadꢀ
ing = 1.08 mmol/g) and glycoside 2 were dried under high
vacuum over KOH overnight before use. The resin was
washed with dry THF (×1) and dry DCM (×2) under N2.
The sugar (2 equiv.) was dissolved in dry DCM (8 mL per
gram of resin) under N2 and the resulting solution was addꢀ
(3’,4’-Difluorobenzyl)-2-azido-3-O-benzoyl-6-O-(tert-
butyldiphenylsilyl)-2-deoxy-1-α-D-glucopyranoside
(2d). GlcSMe 1 (5.0 g, 8.7 mmol) and 3,4ꢀdifluorobenzyl
ACS Paragon Plus Environment

Journal of Chemical Information and Modeling
Page 14 of 22
ed to the resin, which was shaken for 2 min at RT. were finally washed with DMF, MeOH and DCM (× 3
1
2
3
4
5
6
7
8
BF3.Et2O (0.3 equiv.) was then added to the resin, which
was shaken for 10 min. The addition of BF3.Et2O
(0.3 equiv.) was repeated once, followed by shaking for
10 min at RT. After draining, the resin was finally washed
with dry DCM (×5) to recover the glycoside. The filtrate
with unreacted glycoside was washed with saturated Naꢀ
HCO3 solution in water (×3). The organic phase was dried
over MgSO4 and, after filtration, evaporated under reduced
pressure to give crude recovered sugar 2, which was dried
under high vacuum. Resin 3 was further washed with dry
THF (×4) and dry DCM (×4) and then dried under high
vacuum.
each) and resin 4 was dried under high vacuum overnight.
Preparation of PSHF reagent. To a solution of proton
sponge
(1,8ꢀbisꢀ(dimethylamino)ꢀnaphthalene,
50 g,
0.233 mol) in dry THF (300 mL) was added HFꢀPyridine
(HF 70%, Pyridine 30%, 12.5 mL, 2 equiv.), in a plastic
vessel, dropwise over 10 min under N2. The mixture was
stirred for 15 min before filtration of the resulting white
precipitate. The precipitate was washed with dry THF
(4×50 mL) and dry Et2O (4×100 mL) and then dried under
high vacuum overnight (44 g, 81%).
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
TBDPS deprotection at O6. A solution of 0.5 M PSHF
and AcOH 5% in dry DMF was added to the Kans (2.5 mL
per Kan). The mixture was stirred and heated to 65°C
overnight. After cooling down, the solution was aspirated
and the Kans washed with DMF, THF and DCM (× 3
each). The resulting resin was dried under high vacuum
overnight.
Resin loading determination. The glycosideꢀresin 3 was
dried under high vacuum for a minimum of 16 h. A mixture
of anhydrous DCM/TES/TFA (7:2:1) was prepared under
N2. About 70 mg of the dried loaded resin was weighed
precisely, and the cleavage solution (1 mL) was added. The
resulting mixture was left at RT for 3 h. The solution was
then drained and the resin was washed with DCM
(4 × 2 mL). The filtrates were combined and diluted to a
known volume. The concentration of the corresponding
glycoside was then assessed using HPLC/UV chromatogꢀ
raphy comparison against a standard curve obtained from
five known glycoside concentrations (ranging from 0.06 to
1.2 mg/mL), and the resulting resin loading value was deꢀ
rived. Typical resin loading values were in the range 0.2ꢀ
0.4 mmol/g.
Methylation at O6. This reaction was performed followꢀ
ing the procedure described for Alkylation at O3 (see
above), using methyl iodide as the alkylating agent, to give
resin 5.
Azide reduction. A solution of 0.2 M LiOtBu and 0.2 M
DTT in dry DMF was added to the Kans (2.5 mL per Kan),
which were shaken overnight at RT. The solution was aspiꢀ
rated, the Kans washed with DMF, MeOH and DCM (× 3
each) and the resin dried under high vacuum overnight.
Acylation at N2. A solution of HBTU 0.5 M in dry DMF
was prepared. The acid was dissolved in dry DMF (0.5 M)
and treated with HBTU (0.5 M, same volume) and DIPEA
(final concentration 0.5 M). The resulting mixture was
swirled and allowed to stand for 10 min, then added to the
Kans (2.5 mL per Kan), which were shaken at RT overꢀ
night. The solution was then aspirated, the Kans washed
with DMF, MeOH and DCM (× 3 each), and resulting resin
6 dried under high vacuum overnight.
Library synthesis. The library compounds were syntheꢀ
sized following the “splitꢀandꢀpool” technique, using the
IRORI directed sorting process and the Synthesis Manager
software. Each individual resin was split in the required
number of Kan microreactors. Around 80 mg of resin was
introduced in each Kan, using the IRORI dry resin filler.
Each microreactor contained a unique tag identifier (ceramꢀ
ic 2Dꢀcoded cap). By pooling microreactors into common
building block groups, the directed sorting process made
the synthesis efficient while enabling the synthesis of one
discrete compound in each microreactor. Quality control
(QC) Kans were added to monitor evolution of each reacꢀ
tion along the library production process.
Fmoc deprotection. The Kans were treated with a soluꢀ
tion of 20% piperidine in dry DMF (2.5 mL per Kan), and
the resulting mixture was shaken for 2 h at RT. The soluꢀ
tion was aspirated, the Kans washed with DMF, MeOH and
DCM (× 3 each) and the resin dried under high vacuum
overnight.
Benzoate removal at O3. A solution of NaOMe 240 mM
in dry MeOH was prepared under N2, from Na metal. The
Kans were treated with a mixture dry THF / NaOMe soluꢀ
tion in MeOH (83:17, 2.5 mL per Kan, final NaOMe conꢀ
centration = 40 mM) and were shaken overnight at RT. The
solution was drained and the Kans washed with MeOH,
THF, MeOH and DCM (× 3 each). The resin was dried
under high vacuum overnight.
Alkylation at O3. A solution of LiOtBu in dry DMF
(0.25 M) was prepared under N2, as well as a solution of
alkylating agent in dry DMF (0.25 M). The Kans were
washed once with dry DMF, then treated with the LiOtBu
solution (2.5 ml per Kan), for 10 min at RT. The LiOtBu
solution was removed and the alkylating agent solution
(2.5 ml per Kan) added. The Kans were shaken for 20 min
at RT, then the solution was removed. This LiOtBu / alkylꢀ
ating agent cycle was performed 3 times, using fresh LiOtꢀ
Bu solution and fresh alkylating agent solution. The Kans
Guanidine formation. The Kans were treated with a soluꢀ
tion of DMPFN (3,5ꢀdimethylꢀ1ꢀpyrazolylformamidinium
nitrate, 0.25 M) and DIPEA (1 M) in dry DMF (2.5 mL per
Kan), and the resulting mixture was stirred and heated to
65°C overnight. After cooling down, the solution was aspiꢀ
rated, the Kans washed with DMF, MeOH and DCM (× 3
each) and resin 7 was dried under high vacuum overnight.
Cleavage off-resin. Each Kan of the library was placed in
a tube of a Bohdan block and treated with 2 mL of cleavage
solution (TFA/TES/dry DCM 10/20/70) for 3 hours at RT.
The cleavage solution was drained in the corresponding
collecting tube and the Kan washed with THF, DCM and
THF (1 mL each). The solvents were evaporated to obtain
crude cleavage products.
NH3 treatment. Each crude residue from the cleavage
step was treated with freshly prepared saturated ammonia
ACS Paragon Plus Environment

Page 15 of 22
Journal of Chemical Information and Modeling
solution in MeOH (2 mL). The solution was gently shaken
(4’-Fluorobenzyl)-2-deoxy-2-(4’-guanidinobutyryl-
1
2
3
4
5
6
7
8
and left to stand for 2 hours at RT. The solution was evapoꢀ
rated and crude samples 8 were purified by preparative
LC/MS on a standard C18 column using gradient programs
of acetonitrile and water (containing 0.01% TFA). The puꢀ
rified samples were reanalysed by LC/MS with parallel
ELSD detection to determine purity (Column Zorbax C18
(4.6×50 mm); Solvent = H2O (0.01% TFA) / ACN (0.01%
TFA); Gradient = 5(0ꢀ1), 100 (7ꢀ12); Flow = 2 mL/min).
amido)-3-O-(4’-trifluoromethylbenzyl)-6-O-methyl-α-D-
glucopyranoside (8g). Resin loaded αꢀglycoside 2b
(80 mg resin, 0.44 mmol/g) was carried through the solid
phase synthetic process. The title product was isolated by
HPLC purification and MS detection as a white solid
(12.5 mg, 61%): C18 HPLC/ELSD Rt 5.14 min (97.34%),
MS (ES) m/z 587.06 [M+H]+.
(3’,4’-Difluorobenzyl)-2-deoxy-2-(4’-guanidino-
butyrylamido)-3-O-(4’-fluorobenzyl)-6-O-methyl-α-D-
glucopyranoside (8h). Resin loaded αꢀglycoside 2d
(80 mg resin, 0.43 mmol/g) was carried through the solid
phase synthetic process. The title product was isolated by
HPLC purification and MS detection as a white solid
(13.8 mg, 72%): C18 HPLC/ELSD Rt 4.93 min (99.77%),
MS (ES) m/z 555.00 [M+H]+.
9
(4’-Chlorobenzyl)-2-deoxy-2-(4’-guanidinobutyryl-
amido)-3-O-(4’-chlorobenzyl)-6-O-methyl-α-D-gluco-
pyranoside (8a). Resin loaded αꢀglycoside 2a (80 mg resꢀ
in, 0.44 mmol/g) was carried through the solid phase synꢀ
thetic process. The title product was isolated by HPLC puꢀ
rification and MS detection as a white solid (13.1 mg,
65%): C18 HPLC/ELSD Rt 5.14 min (99.82%); MS (ES)
m/z 568.99 [M+H]+.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
(3’,4’-Difluorobenzyl)-2-deoxy-2-(4’-guanidino-
butyrylamido)-3-O-(4’-trifluoromethyl-benzyl)-6-O-
methyl-α-D-glucopyranoside (8i). Resin loaded αꢀ
glycoside 2d (80 mg resin, 0.43 mmol/g) was carried
through the solid phase synthetic process. The title product
was isolated by HPLC purification and MS detection as a
white solid (11.8 mg, 57%): C18 HPLC/ELSD Rt 5.08 min
(99.2%), MS (ES) m/z 605.13 [M+H]+.
(4’-Trifluoromethylbenzyl)-2-deoxy-2-(4’-guanidino-
butyrylamido)-3-O-(4’-chlorobenzyl)-6-O-methyl-α-D-
glucopyranoside (8b). Resin loaded αꢀglycoside 2c
(80 mg resin, 0.40 mmol/g) was carried through the solid
phase synthetic process. The title product was isolated by
HPLC purification and MS detection as a white solid
(13.3 mg, 69%): C18 HPLC/ELSD Rt 5.33 min (99.99%),
MS (ES) m/z 603.02 [M+H]+.
(4’-Chlorobenzyl)-2-deoxy-2-(4’-guanidinobutyryl-
amido)-3-O-benzyl-6-O-methyl-α-D-glucopyranoside
(8j). Resin loaded αꢀglycoside 2a (80 mg resin,
0.44 mmol/g) was carried through the solid phase synthetic
process. The title product was isolated by HPLC purificaꢀ
tion and MS detection as a white solid (11.1 mg, 59%):
C18 HPLC/ELSD Rt 4.71 min (100%), MS (ES) m/z
535.20 [M+H]+.
(4’-Fluorobenzyl)-2-deoxy-2-(4’-guanidinobutyryl-
amido)-3-O-(4’-chlorobenzyl)-6-O-methyl-α-D-gluco-
pyranoside (8c). Resin loaded αꢀglycoside 2b (80 mg resꢀ
in, 0.44 mmol/g) was carried through the solid phase synꢀ
thetic process. The title product was isolated by HPLC puꢀ
rification and MS detection as a white solid (10.3 mg,
53%): C18 HPLC/ELSD Rt 5.05 min (99.37%), MS (ES)
m/z 553.02 [M+H]+.
Competition Binding Assay on MCH-1R
(3’,4’-Difluorobenzyl)-2-deoxy-2-(4’-guanidino-
butyrylamido)-3-O-(4’-chlorobenzyl)-6-O-methyl-α-D-
glucopyranoside (8d). Resin loaded αꢀglycoside 2d
(80 mg resin, 0.43 mmol/g) was carried through the solid
phase synthetic process. The title product was isolated by
HPLC purification and MS detection as a white solid
(12.4 mg, 63%): C18 HPLC/ELSD Rt 5.05 min (100%),
MS (ES) m/z 571.09 [M+H]+.
The competition binding assay was performed using
membranes prepared from CHO cells stably and selectively
expressing the cloned, human MCHꢀ1R. The assay was
carried out in an Optiplate (Perkin Elmer, 60052) wherein
test compound (50 ꢁL, used at increasing concentrations),
was combined with successive additions of 25 ꢁL of radiꢀ
oligand ([125I]ꢀMCH, Perkin Elmer, NEX 375,
2200 Ci/mMol) diluted in assay buffer (25 mM HEPES
pH 7.4, 1 mM CaCl2, 5 mM MgCl2, 0.2 % protease free
BSA) and 25 ꢁL of a premix of membrane/beads (PVTꢀ
WGA beads, Amersham, RPNQ0001, 0.25 mg/well). Plates
were incubated at room temperature for 2 hours before
counting for 1 minute/well in a TopCount (Packard). Pheꢀ
(D)TyrꢀMCH (Bachem, Hꢀ2218) was used as the reference
ligand. IC50 values were calculated by nonꢀlinear regression
analysis using a 4 parameter sigmoidal doseꢀresponse modꢀ
el (XLfit, IDBS Software).
(4’-Chlorobenzyl)-2-deoxy-2-(4’-guanidinobutyryl-
amido)-3-O-(4’-trifluoromethylbenzyl)-6-O-methyl-α-D-
glucopyranoside (8e). Resin loaded α−glycoside 2a
(80 mg resin, 0.44 mmol/g) was carried through the solid
phase synthetic process. The title product was isolated by
HPLC purification and MS detection as a white solid
(11.7 mg, 55%): C18 HPLC/ELSD Rt 5.25 min (98.44%),
MS (ES) m/z 603.02 [M+H]+.
(4’-Chlorobenzyl)-2-deoxy-2-(4’-guanidinobutyryl-
amido)-3-O-(4’-fluorobenzyl)-6-O-methyl-α-D-gluco-
pyranoside (8f). Resin loaded αꢀglycoside 2a (80 mg resꢀ
in, 0.44 mmol/g) was carried through the solid phase synꢀ
thetic process. The title product was isolated by HPLC puꢀ
rification and MS detection as a white solid (11.4 mg,
59%): C18 HPLC/ELSD Rt 5.02 min (98.03%), MS (ES)
m/z 553.02 [M+H]+.
Functional aequorin-based assay on MCH-1R
The aequorinꢀbased assay was performed in CHO cells
stably and selectively expressing the MCH1 receptor. Cells
in midꢀlog phase, grown in media without antibiotic 18
hours prior to the test, were detached by gentle flushing
with PBSꢀEDTA (5 mM EDTA), recovered by centrifugaꢀ
tion and resuspended in “BSA medium” (DMEM/HAM’s
F12 Gibco Cat#11039ꢀ021, with HEPES, without phenol
ACS Paragon Plus Environment

Journal of Chemical Information and Modeling
Page 16 of 22
protein-coupled receptor; MCH, melanin-concentrating
red + 0.1 % BSA). Cells were then counted, centrifuged
and resuspended in a 15 mL Falcon tube at a concentration
of 1.7 x 106 cells/mL. Coelenterazine h (Promega, Cat n°
X300X, stock solution: 500 μM in methanol) was added
1
2
3
4
5
6
7
8
hormone; VAST, Versatile Assembly on Stable Templates;
7TMD, 7 transmembrane domain; TM, transmembrane helix;
ECL, extracellular loop; PDB, Protein Data Bank; SAR, struc-
ture-activity relationship; VS, virtual screening; SBDD, struc-
ture-based drug discovery.
to the cells in suspension at a final concentration of 5 μM.
The Falcon tube, wrapped in aluminum foil, was then
placed on a vertical rotating wheel (6 rpm) and incubated at
room temperature (temperature should be maintained beꢀ
low 22°C) overnight in order to reconstitute active aeꢀ
quorin. Cells were counted, diluted in “BSAꢀmedium” and
incubated for 60 minutes prior to experimentation. Test and
reference compounds were similarly prepared and distribꢀ
uted in a 384ꢀwell plate (30 μl/well) prior to injection of
cells (5000 cells/30ꢁL). Cells were injected and agonist
activity was measured by recording the emission of light
(FDSS 6000 Hamamatsu) for a period of 90 seconds. Anꢀ
tagonist activity was tested on the same preparation followꢀ
ing the addition and 3 minute incubation with 30 ꢁL of the
reference agonist (MCH, Bachem, H1482), at a final conꢀ
centration equivalent to the EC80 determined for the same
experimental day (using the reader injection system). All
values were normalized relative to background (defined as
0 %) and the maximal response to MCH (defined as
100 %). EC50 values were calculated by nonꢀlinear regresꢀ
sion analysis using a 4 parameter sigmoidal doseꢀresponse
model (XLfit, IDBS Software).
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
ASSOCIATED CONTENT
Supporting Material on methods used to calculate AMBER MCHꢀ
1Rꢀligand interaction energy, Shape comparison and Statistical
comparison of subꢀpocket occupancy effects are included together
with multiple aminoꢀacid sequence alignment results of MCHꢀ1R
sequence aligned with four GPCR crystal structures [human doꢀ
pamine D3 receptor (D3, PDB entry 3PBL), β2ꢀadrenergic recepꢀ
tor (β2AR, PDB entry 2RH1), human A2A adenosine receptor
(A2A, PDB entry 3EML) and bovine rhodopsin (PDB entry
1U19)], and the table of 32 molecules used as training set for
statistical analysis. This material is available free of charge via the
Internet at http://pubs.acs.org.
AUTHOR INFORMATION
Corresponding Author
Alexander.Heifetz@Evotec.com
ACKNOWLEDGMENT
We would like to acknowledge Euroscreen, in particular Drs
Graeme Fraser and Sébastien Blanc, for the determination of
MCH-1R IC50 and EC50 values for the SAR analogues and Dr
Daniel Warner for his help in statistical analysis of MCH-1R
antagonists docking poses. We are also grateful to the Royal
Society for an Industry Award (AH) and to the Queensland
Government Smart State Fund for a NIRAP grant.
ABBREVIATIONS
MCH-1R, melanin-concentrating hormone receptor 1; MCH-
2R, melanin-concentrating hormone receptor 2; GPCR, G-
ACS Paragon Plus Environment

Page 17 of 22
Journal of Chemical Information and Modeling
Reference List
9. Della-Zuana, O.; Presse, F.; Ortola, C.;
1
2
3
4
5
6
7
8
9
Duhault, J.; Nahon, J. L.; Levens, N.
Acute and chronic administration of
melanin-concentrating
hormone
1. Forte, V.; Pandey, A.; Abdelmessih, R.;
Forte, G.; Whaley-Connell, A.;
Sowers, J. R.; McFarlane, S. I.
Obesity, Diabetes, the Cardiorenal
Syndrome, and Risk for Cancer.
Cardiorenal. Med. 2012, 2, 143-162.
enhances food intake and body
weight in Wistar and Sprague-
Dawley rats. Int. J. Obes. Relat Metab
Disord. 2002, 26, 1289-1295.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
10. Takekawa, S.; Asami, A.; Ishihara, Y.;
Terauchi, J.; Kato, K.; Shimomura,
Y.; Mori, M.; Murakoshi, H.; Kato,
K.; Suzuki, N.; Nishimura, O.;
Fujino, M. T-226296: a novel, orally
active and selective melanin-
concentrating hormone receptor
antagonist. Eur. J. Pharmacol. 2002,
438, 129-135.
2. Saito, Y.; Maruyama, K. Identification of
melanin-concentrating
hormone
receptor and its impact on drug
discovery. J. Exp. Zool. A Comp Exp.
Biol. 2006, 305, 761-768.
3. Xu, Y. L.; Jackson, V. R.; Civelli, O. Orphan
G protein-coupled receptors and
obesity. Eur. J. Pharmacol. 2004, 500,
243-253.
11. Cippitelli, A.; Karlsson, C.; Shaw, J. L.;
Thorsell, A.; Gehlert, D. R.; Heilig,
M. Suppression of alcohol self-
administration and reinstatement of
4. Schioth, H. B. G protein-coupled receptors
in regulation of body weight. CNS.
Neurol. Disord. Drug Targets. 2006,
5, 241-249.
alcohol
seeking
by
melanin-
concentrating hormone receptor 1
(MCH1-R) antagonism in Wistar
rats. Psychopharmacology (Berl)
2010, 211, 367-375.
5. Bednarek, M. A. Peptide ligands for the
melanin-concentrating
hormone
(MCH) receptor 1. Curr. Top. Med.
Chem 2007, 7, 1425-1432.
12. Borowsky, B.; Durkin, M. M.; Ogozalek, K.;
Marzabadi, M. R.; DeLeon, J.; Lagu,
B.; Heurich, R.; Lichtblau, H.;
Shaposhnik, Z.; Daniewska, I.;
Blackburn, T. P.; Branchek, T. A.;
Gerald, C.; Vaysse, P. J.; Forray, C.
6. Kowalski, T. J.; McBriar, M. D. Therapeutic
potential of melanin-concentrating
hormone-1 receptor antagonists for
the treatment of obesity. Expert.
Opin. Investig. Drugs 2004, 13, 1113-
1122.
Antidepressant,
anxiolytic
and
anorectic effects of
a
melanin-
concentrating hormone-1 receptor
antagonist. Nat. Med. 2002, 8, 825-
830.
7. Hogberg, T.; Frimurer, T. M.; Sasmal, P. K.
Melanin concentrating hormone
receptor 1 (MCHR1) antagonists-Still
13. Shearman, L. P.; Camacho, R. E.; Sloan, S.
D.; Zhou, D.; Bednarek, M. A.;
Hreniuk, D. L.; Feighner, S. D.; Tan,
C. P.; Howard, A. D.; Van der Ploeg,
L. H.; Macintyre, D. E.; Hickey, G. J.;
Strack, A. M. Chronic MCH-1
receptor modulation alters appetite,
body weight and adiposity in rats.
Eur. J. Pharmacol. 2003, 475, 37-47.
a
viable approach for obesity
treatment? Bioorg. Med. Chem Lett.
2012, 22, 6039-6047.
8. Miyamoto-Matsubara, M.; Chung, S.; Saito,
Y.
regulator of G protein signaling-2
with melanin-concentrating
Functional
interaction
of
hormone receptor 1. Ann. N. Y. Acad.
Sci. 2010, 1200, 112-119.
ACS Paragon Plus Environment

Journal of Chemical Information and Modeling
Page 18 of 22
14. Cavasotto, C. N.; Orry, A. J.; Murgolo, N. J.;
Mini. Rev. Med. Chem 2006, 6, 1299-
1309.
1
2
3
4
5
6
7
8
Czarniecki, M. F.; Kocsi, S. A.;
Hawes, B. E.; O'Neill, K. A.; Hine, H.;
Burton, M. S.; Voigt, J. H.; Abagyan,
R. A.; Bayne, M. L.; Monsma, F. J., Jr.
Discovery of novel chemotypes to a
G-protein-coupled receptor through
ligand-steered homology modeling
20. Meutermans, W.; Le, G. T.; Becker, B.
Carbohydrates as scaffolds in drug
discovery. ChemMedChem. 2006, 1,
1164-1194.
9
21. Halliday, J.; McKeveney, D.; Muldoon, C.;
Rajaratnam, P.; Meutermans, W.
and
structure-based
virtual
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
screening. J. Med. Chem. 2008, 51,
Targeting
transglycosylases.
Pharmacol. 2006, 71, 957-967.
the
forgotten
Biochem.
581-588.
15. Clark, D. E.; Higgs, C.; Wren, S. P.; Dyke, H.
J.; Wong, M.; Norman, D.; Lockey, P.
M.; Roach, A. G. A virtual screening
approach to finding novel and
potent antagonists at the melanin-
concentrating hormone 1 receptor. J.
Med. Chem 2004, 47, 3962-3971.
22. Tran, T. T.; McKie, J.; Meutermans, W. D.;
Bourne, G. T.; Andrews, P. R.;
Smythe, M. L. Topological side-
chain classification of beta-turns:
ideal motifs for peptidomimetic
development. J. Comput. Aided Mol.
Des 2005, 19, 551-566.
16. Tate, C. G. A crystal clear solution for
determining
receptor structures. Trends Biochem.
Sci. 2012, 37, 343-352.
G-protein-coupled
23. Le, G. T.; Abbenante, G.; Becker, B.;
Grathwohl,
Tometzki,
Meutermans, W. Molecular diversity
through sugar scaffolds. Drug
Discov. Today 2003, 8, 701-709.
M.;
G.;
Halliday,
Zuegg,
J.;
J.;
17. Abbenante, G.; Becker, B.; Blanc, S.; Clark,
C.; Condie, G.; Fraser, G.; Grathwohl,
M.; Halliday, J.; Henderson, S.; Lam,
A.; Liu, L.; Mann, M.; Muldoon, C.;
Pearson, A.; Premraj, R.; Ramsdale,
T.; Rossetti, T.; Schafer, K.; Le, T. G.;
Tometzki, G.; Vari, F.; Verquin, G.;
Waanders, J.; West, M.; Wimmer,
N.; Yau, A.; Zuegg, J.; Meutermans,
24. Gruner, S. A.; Locardi, E.; Lohof, E.; Kessler,
H. Carbohydrate-based mimetics in
drug design: sugar amino acids and
carbohydrate scaffolds. Chem Rev.
2002, 102, 491-514.
W. Biological diversity from
structurally
systematically
conformational space using
pyranose scaffold. J. Med. Chem
2010, 53, 5576-5586.
a
25. Gruner, S. A.; Keri, G.; Schwab, R.;
Venetianer, A.; Kessler, H. Sugar
amino acid containing somatostatin
analogues that induce apoptosis in
both drug-sensitive and multidrug-
resistant tumor cells. Org. Lett. 2001,
3, 3723-3725.
diverse
library:
scanning
a
18. Meutermans, W. D.; Bourne, G. T.; Golding,
S. W.; Horton, D. A.; Campitelli, M.
R.; Craik, D.; Scanlon, M.; Smythe,
M. L. Difficult macrocyclizations:
new strategies for synthesizing
highly strained cyclic tetrapeptides.
Org. Lett. 2003, 5, 2711-2714.
26. Beinborn, M. Class B GPCRs: a hidden
agonist within? Mol. Pharmacol.
2006, 70, 1-4.
27. Heifetz, A.; Morris, G. B.; Biggin, P. C.;
Barker, O.; Fryatt, T.; Bentley, J.;
Hallett, D.; Manikowski, D.; Pal, S.;
Reifegerste, R.; Slack, M.; Law, R.
Study of human Orexin-1 and -2 G-
19. Becker, B.; Condie, G. C.; Le, G. T.;
Meutermans, W. Carbohydrate-
based scaffolds in drug discovery.
ACS Paragon Plus Environment

Page 19 of 22
Journal of Chemical Information and Modeling
protein-coupled
novel and published antagonists by
receptors
with
Thomsen, W. Anxiolytic- and
antidepressant-like profile of
ATC0065 and ATC0175: nonpeptidic
and orally active melanin-
concentrating hormone receptor 1
antagonists. J. Pharmacol. Exp. Ther.
2005, 313, 831-839.
1
2
3
4
5
6
7
8
modeling,
simulations,
molecular
and
dynamics
site-directed
mutagenesis. Biochemistry 2012, 51,
3178-3197.
28. Auffinger, P.; Hays, F. A.; Westhof, E.; Ho, P.
S. Halogen bonds in biological
molecules. Proc. Natl. Acad. Sci. U. S.
A 2004, 101, 16789-16794.
9
34. Shoichet, B. K.; Kobilka, B. K. Structure-
based drug screening for G-protein-
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
coupled
receptors.
Trends
Pharmacol. Sci. 2012, 33, 268-272.
29. Yumi N.Imai; Yoshihisa Inoue; Isao
Nakanishi; Kazuo Kitaura Cl-pi
35. Congreve, M.; Langmead, C.; Marshall, F. H.
The use of GPCR structures in drug
design. Adv. Pharmacol. 2011, 62, 1-
36.
interactions
complexes. Protein Science 2008, 17,
1129-1137.
in
protein-ligand
30. Anand, S.; Anbarasu, A.; Sethumadhavan, R.
Non-classical hydrogen bonds in
interleukins: the role of C-H...O
interactions. Int. J. Biol. Macromol.
2008, 43, 468-473.
36. Congreve, M.; Marshall, F. The impact of
GPCR structures on pharmacology
and structure-based drug design. Br.
J. Pharmacol. 2010, 159, 986-996.
37. Congreve, M.; Langmead, C. J.; Mason, J. S.;
Marshall, F. H. Progress in structure
based drug design for G protein-
coupled receptors. J. Med. Chem
2011, 54, 4283-4311.
31. Thanh, G. L.; Abbenante, G.; Adamson, G.;
Becker, B.; Clark, C.; Condie, G.;
Falzun, T.; Grathwohl, M.; Gupta, P.;
Hanson, M.; Huynh, N.; Katavic, P.;
Kuipers, K.; Lam, A.; Liu, L.; Mann,
M.; Mason, J.; McKeveney, D.;
38. Carlsson, J.; Coleman, R. G.; Setola, V.;
Irwin, J. J.; Fan, H.; Schlessinger, A.;
Sali, A.; Roth, B. L.; Shoichet, B. K.
Ligand discovery from a dopamine
D3 receptor homology model and
crystal structure. Nat. Chem Biol.
2011, 7, 769-778.
Muldoon,
C.;
Pearson,
A.;
Rajaratnam, P.; Ryan, S.; Tometzki,
G.; Verquin, G.; Waanders, J.; West,
M.; Wilcox, N.; Wimmer, N.; Yau, A.;
Zuegg, J.; Meutermans, W.
versatile synthetic approach toward
diversity libraries using
A
39. Ballesteros, J. A.; Weinstein, H. Integrated
methods for construction three
monosaccharide scaffolds. J. Org.
Chem 2010, 75, 197-203.
dimensional
models
and
computational probing of structure-
function relations in G protein-
coupled receptors. Methods Neurosci
1995, 25, 366-428.
32. Carroll, T. X.; Thomas, T. D.; Bergersen, H.;
Borve, K. J.; Saethre, L. J. Fluorine as
a pi donor. Carbon 1s photoelectron
spectroscopy and proton affinities of
fluorobenzenes. J. Org. Chem 2006,
71, 1961-1968.
40. Heifetz, A.; Morris, G. B.; Biggin, P. C.;
Barker, O.; Fryatt, T.; Bentley, J.;
Hallett, D.; Manikowski, D.; Pal, S.;
Reifegerste, R.; Slack, M.; Law, R.
Study of human Orexin-1 and -2 G-
33. Chaki, S.; Funakoshi, T.; Hirota-Okuno, S.;
Nishiguchi, M.; Shimazaki, T.; Iijima,
M.; Grottick, A. J.; Kanuma, K.;
protein-coupled
receptors
with
Omodera,
K.;
Sekiguchi,
Y.;
novel and published antagonists by
Okuyama, S.; Tran, T. A.; Semple, G.;
ACS Paragon Plus Environment

Journal of Chemical Information and Modeling
Page 20 of 22
modeling,
simulations,
mutagenesis. Biochemistry 2012, 51,
3178-3197.
molecular
and
dynamics
site-directed
homologs. Protein Sci. 2006, 15, 509-
521.
1
2
3
4
5
6
7
8
49. Bettinelli, I.; Graziani, D.; Marconi, C.;
Pedretti, A.; Vistoli, G. The approach
of conformational chimeras to
model the role of proline-containing
helices on GPCR mobility: the fertile
case of Cys-LTR1. ChemMedChem.
2011, 6, 1217-1227.
41. Needleman, S. B.; Wunsch, C. D. A general
method applicable to the search for
similarities in the amino acid
sequence of two proteins. J. Mol.
Biol. 1970, 48, 443-453.
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
42. Styczynski, M. P.; Jensen, K. L.; Rigoutsos, I.;
Stephanopoulos, G. BLOSUM62
miscalculations improve search
performance. Nat. Biotechnol. 2008,
26, 274-275.
50. Langelaan, D. N.; Wieczorek, M.; Blouin, C.;
Rainey, J. K. Improved helix and kink
characterization
in
membrane
proteins allows evaluation of kink
sequence predictors. J. Chem Inf.
Model. 2010, 50, 2213-2220.
43. Fanelli,
F.;
Computational
approaches to structure-function
analysis of protein-coupled
De
Benedetti,
P.
G.
modeling
51. Hall, S. E.; Roberts, K.; Vaidehi, N. Position
of helical kinks in membrane protein
crystal structures and the accuracy
of computational prediction. J. Mol.
Graph. Model. 2009, 27, 944-950.
G
receptors. Chem Rev. 2005, 105,
3297-3351.
44. Baldwin, J. M. The probable arrangement of
the helices in G protein-coupled
receptors. The EMBO Journal 1993,
12, 1693-1703.
52. Yohannan, S.; Faham, S.; Yang, D.;
Whitelegge, J. P.; Bowie, J. U. The
evolution of transmembrane helix
kinks and the structural diversity of
G protein-coupled receptors. Proc.
Natl. Acad. Sci. U. S. A 2004, 101,
959-963.
45. Cobanoglu, M. C.; Saygin, Y.; Sezerman, U.
Classification of GPCRs using family
specific motifs. IEEE/ACM. Trans.
Comput. Biol. Bioinform. 2011, 8,
1495-1508.
53. Fechteler, T.; Dengler, U.; Schomburg, D.
Prediction
of
protein
three-
dimensional structures in insertion
and deletion regions: a procedure for
46. Gloriam, D. E.; Foord, S. M.; Blaney, F. E.;
Garland, S. L. Definition of the G
searching
data
bases
of
protein-coupled
receptor
representative protein fragments
using geometric scoring criteria. J.
Mol. Biol. 1995, 253, 114-131.
transmembrane bundle binding
pocket and calculation of receptor
similarities for drug design. J. Med.
Chem. 2009, 52, 4429-4442.
54. Jorgensen, W.; Maxwell, L. D. S.; Tirado, R.
J. Development and testing of the
47. Gao, Q. B.; Wang, Z. Z. Classification of G-
protein coupled receptors at four
levels. Protein Eng Des Sel 2006, 19,
511-516.
OPLS all-atom force field on
conformational
energetics
and
properties of organic liquids. J. Am.
Chem. Soc. 1996, 118, 11225-11236.
48. Wistrand, M.; Kall, L.; Sonnhammer, E. L. A
general model of G protein-coupled
55. Kaminski, G. A.; Friesner, J.; Tirado, R. J.;
Jorgensen, W. Evaluation and
reparameterization of the OPLS-AA
force field for proteins via
receptor
sequences
and
its
application
to detect remote
ACS Paragon Plus Environment

Page 21 of 22
Journal of Chemical Information and Modeling
comparison with accurate quantum
chemical calculations on peptides. J.
Phys. Chem. B. 2001, 105, 6474-6487.
63. Tiikkainen, P.; Franke, L. Analysis of
commercial and public bioactivity
databases. J. Chem Inf. Model. 2012,
52, 319-326.
1
2
3
4
56. Courcot, B.; Bridgeman, A. J. Modeling the
5
interactions
polyoxometalates
environment. J. Comput. Chem 2011,
32, 3143-3153.
between
and their
64. Brewer, M. L. Development of a spectral
clustering method for the analysis of
molecular data sets. J. Chem Inf.
Model. 2007, 47, 1727-1733.
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
57. Liebeschuetz, J. W.; Cole, J. C.; Korb, O.
65. Ma, C.; Wang, L.; Xie, X. Q. GPU accelerated
chemical similarity calculation for
compound library comparison. J.
Chem Inf. Model. 2011, 51, 1521-1527.
Pose
prediction
and
virtual
screening performance of GOLD
scoring functions in a standardized
test. J. Comput. Aided Mol. Des 2012,
26, 737-748.
58. Kollman, P. A.; Massova, I.; Reyes, C.; Kuhn,
B.; Huo, S.; Chong, L.; Lee, M.; Lee,
T.; Duan, Y.; Wang, W.; Donini, O.;
Cieplak, P.; Srinivasan, J.; Case, D.
A.; Cheatham, T. E., III Calculating
structures and free energies of
complex
molecular
molecules:
combining
and
mechanics
continuum models. Acc. Chem Res.
2000, 33, 889-897.
59. Homeyer, N.; Gohlke, H. FEW-A workflow
tool for free energy calculations of
ligand binding. J. Comput. Chem
2013.
60. Hoffer, L.; Horvath, D. S4MPLE--sampler for
multiple protein-ligand entities:
simultaneous docking of several
entities. J. Chem Inf. Model. 2013, 53,
88-102.
61. Labute, P. LowModeMD--implicit low-mode
velocity
conformational
filtering
applied
search
to
of
macrocycles and protein loops. J.
Chem Inf. Model. 2010, 50, 792-800.
62. Berensden, H. J. C.; Postma, J. P. M.; Van-
Gunsteren, W. F.; DiNola, A.; Haak,
J. R. MolecularꢀDynamics with
Coupling to an External Bath. J. Chem.
Phys. 1984, 81, 3684-3690.
ACS Paragon Plus Environment

Journal of Chemical Information and Modeling
Page 22 of 22
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Table of Contents artwork
22
ACS Paragon Plus Environment