Probing the Hydrophobic Binding Pocket of G-Protein-Coupled Lysophosphatidylserine Receptor GPR34/LPS1 by Docking-Aided Structure-Activity Analysis

Article

pubs.acs.org/jmc

Probing the Hydrophobic Binding Pocket of G‑Protein-Coupled

Lysophosphatidylserine Receptor GPR34/LPS₁by Docking-Aided

Structure−Activity Analysis

Misa Sayama,^†,∇Asuka Inoue,^‡,⊥,∇Sho Nakamura,^†Sejin Jung,^†Masaya Ikubo,^†Yuko Otani,^†

Akiharu Uwamizu,^‡Takayuki Kishi,^‡Kumiko Makide,^‡,⊥Junken Aoki,*^,‡,#Takatsugu Hirokawa,*^,§,∥

and Tomohiko Ohwada*^,†

^†Laboratory of Organic and Medicinal Chemistry, Graduate School of Pharmaceutical Sciences, The University of Tokyo, 7-3-1

Hongo, Bunkyo-ku, Tokyo 113-0033, Japan

^‡Laboratory of Molecular and Cellular Biochemistry, Graduate School of Pharmaceutical Sciences, Tohoku University, 6-3, Aoba,

Aramaki, Aoba-ku, Sendai, Miyagi 980-8578, Japan

^§Molecular Proﬁling Research Center of Drug Discovery (molprof), National Institute of Advanced Industrial Science and

Technology (AIST), 2-3-26 Aomi, Koto-ku, Tokyo 135-0064, Japan

^∥Division of Biomedical Science, Faculty of Medicine, University of Tsukuba, 1-1-1 Tennodai, Tsukuba-shi, Ibaraki 305-8575, Japan

^⊥PRESTO, Japan Science and Technology Agency (JST), 4-1-8, Honcho, Kawaguchi-shi, Saitama 332-0012, Japan

^#AMED-CREST, Japan Agency for Medical Research and Development, 1-7-1 Otemachi, Chiyoda-ku, Tokyo 100-0004, Japan

S

* Supporting Information

ABSTRACT: The ligands of certain G-protein-coupled

receptors (GPCRs) have been identiﬁed as endogenous lipids,

such as lysophosphatidylserine (LysoPS). Here, we analyzed

the molecular basis of the structure−activity relationship of

ligands of GPR34, one of the LysoPS receptor subtypes,

focusing on recognition of the long-chain fatty acid moiety by

the hydrophobic pocket. By introducing benzene ring(s) into

the fatty acid moiety of 2-deoxy-LysoPS, we explored the

binding site’s preference for the hydrophobic shape. A

tribenzene-containing fatty acid surrogate with modiﬁcations

of the terminal aromatic moiety showed potent agonistic

activity toward GPR34. Computational docking of these

derivatives with a homology modeling/molecular dynamics-based virtual binding site of GPR34 indicated that a kink in the

benzene-based lipid surrogates matches the L-shaped hydrophobic pocket of GPR34. A tetrabenzene-based lipid analogue

bearing a bulky tert-butyl group at the 4-position of the terminal benzene ring exhibited potent GPR34 agonistic activity,

validating the present hydrophobic binding pocket model.

adenocarcinoma,^6,7mucosa-associated lymphoid tissue

INTRODUCTION

■

(MALT) lymphoma,⁸and colorectal cancer.⁹

G-protein-coupled receptors (GPCRs) are membrane receptors

However, although GPR34 is related to immune pathology,

containing seven transmembrane helix (TM) domains and are

the underlying mechanisms remain unknown, and chemical

considered major therapeutic targets. During the past decade, a

tools, such as potent agonists, antagonists, or inverse agonists of

number of GPCRs have been deorphaned, and their ligands

GPR34, are needed to investigate these issues. We have

have been identiﬁed as endogenous lipids. Among them,

designed and synthesized several LysoPS analogues that exhibit

1,2

agonist activity toward GPR34.¹⁰⁻¹³LysoPS is an amphipathic

GPR34/LPS₁is speciﬁcally activated by lysophosphatidylser-

ine (LysoPS), a lysophospholipid generated by enzymatic

hydrolysis of a component of membrane bilayers, phosphati-

dylserine (PS).³GPR34 is highly expressed in immune cells,

such as mast cells, macrophages, and microglia.¹GPR34-

deﬁcient mice show a defective immune response after

pathogen challenge⁴and exhibit signiﬁcant changes in micro-

glial morphology and phagocytosis.⁵GPR34 also plays a role in

the progression of some tumors, such as human gastric

molecule that has a phosphoserine as a polar headgroup linked

via a glycerol moiety to a fatty acid (e.g., oleic acid (18:1)) as

the hydrophobic tail. Our previous structure−activity relation-

ship (SAR) study of LysoPS analogues suggested that the

phosphoserine headgroup is necessary for GPR34-activating

Received: May 10, 2017

A

DOI: 10.1021/acs.jmedchem.7b00693

J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

Figure 1. Structural development of lysophosphatidylserine and its 2-deoxy derivatives.

potency.¹²On the other hand, the length and the shape (i.e.,

RESULTS AND DISCUSSION

■

geometry and position of the oleﬁn) of the fatty acid moiety

also seemed to contribute to receptor activation.¹²However,

from the viewpoint of ligand−receptor recognition, the

interaction of such long-chain fatty acids with the hydrophobic

pocket of GPCRs, including GPR34, has been little studied.

Therefore, in this work, we set out to investigate the

molecular basis of the dependency of GPR34 activity on the

structure of the fatty acid moiety of LysoPS analogues. We

focused on the GPR34 agonist activity of 2-deoxy-LysoPS

analogues, which lack the sn-2 hydroxyl group of the glycerol

moiety, and explored the preferred shape of the hydrophobic

fatty acid moiety by introducing benzene ring(s) to restrict the

molecular conformation. Our ﬁndings indicated that GPR34-

activating potency depends strongly on the conﬁguration and

substitution of aromatic rings in these fatty acid surrogates. On

the basis of the obtained SAR, we probed the hydrophobic

binding pocket of GPR34 by means of computational docking

of these synthesized compounds with a homology modeling/

molecular dynamics-based virtual binding site of GPR34. We

found that the matching/mismatching of ligand analogues with

a putative L-shaped binding pocket between TM4 and TM5

was consistent with the observed SAR. The cis-oleﬁn moiety of

the endogenous oleoyl group (18:1) also ﬁt well to the L-

shaped binding pocket. To test the validity of this binding

model, we used it to design a tetrabenzene-based lipid analogue

bearing a bulky tert-butyl group at the 4-position of the terminal

benzene ring. We synthesized this compound and conﬁrmed

that it showed potent GPR34 agonistic activity.

Tactics for Designing LysoPS Analogues. In a previous

study, we had found that analogues of the fatty acid moiety with

an ortho-substituted benzene ring (ring A) had greater activity

than meta- or para-substituted analogues (Figure 1).¹²

Accordingly, in this study, we aimed to elaborate ortho-

substituted analogues by expanding the hydrophobic fatty acid

moiety of 2-deoxy-LysoPS analogues (1a → 2a → 3a → 4a)

through introduction of additional benzene ring(s) (Figure 1).

SAR analysis of the synthesized compounds indicated that

GPR34-activating potency depends on the conﬁguration and

substitution of aromatic rings in these fatty acid surrogates. In

particular, the tribenzene-containing fatty acid surrogate (3a)

showed moderate agonistic activity toward GPR34 (Table 1).

We further explored the SAR, focusing on the terminal moiety

of the fatty acid surrogate, by means of computational docking

of these compounds with a homology modeling/molecular

dynamics-based virtual binding site of GPR34. To test the

validity of the binding models, we used the virtual binding site

to design a tetrabenzene compound bearing a bulky tert-butyl

group at the 4-position of the terminal benzene ring D (4b) as

a candidate agonist, which we synthesized and tested (Figure

1). The chemical structures of all the compounds synthesized in

this work are compiled in Chart S1.

Synthesis of LysoPS Analogues. We previously estab-

lished synthetic routes to 2-deoxy-LysoPS analogues by means

of benzyl protection−hydrogenative debenzylation¹⁰and Boc

protection−acid-catalyzed deprotection procedures.¹²Because

B

DOI: 10.1021/acs.jmedchem.7b00693

J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

a

Table 1. SAR of 2-Deoxy-LysoPS Analogues Containing Fatty Acid Surrogates

a

Activities are represented in terms of EC₅₀. LogEC₅₀(M) and E_max(% AP-TGFα release) values are calculated from a sigmoidal concentration−

response curve and shown as mean SEM of indicated numbers of independent experiments (n). EC₅₀values are calculated from a mean value of

LogEC₅₀. RIA (relative intrinsic activity, a dimensionless parameter; ref 49) is an estimation of agonist activity and represented as an E_max/EC₅₀value

relative to that of a positive control compound (compound 6 in the previous work, ref 13). By deﬁnition, the RIA value of a positive control

compound is equal to 1. “NA” means “not available” because of very low activity.

the target fatty acid surrogates in the present work contain a

benzyl position sensitive to hydrogenation, we chose the Boc-

protection−acid-catalyzed deprotection procedure. The syn-

theses of 2-deoxy-LysoPS analogues with fatty acid surrogates

(4a−4f) are summarized in Scheme 1.

(12a−12f). Global deprotection with TFA furnished the 2-

deoxy-LysoPS analogues as the TFA salts (4a−4f).

Details of the synthesis of the acyl moiety of other

compounds (13−34) are given in the Supporting Information.

Activity of Fatty Acid Surrogate-Containing LysoPS

Analogues toward GPR34. The activities of the synthesized

compounds toward GPR34 were determined by means of

TGFα shedding assay,¹⁶as described in the Experimental

Section. We principally used mouse receptors in the cell-based

assay because of scalability to in vivo studies in mice in the

future (Tables 1−3). For selected analogues (3a, 3f, 3g, 3h, 3v,

4a, 4c, and 4e), we also examined the response to the

corresponding human GPR34, and we conﬁrmed that the

response of LysoPS analogues to human GPR34 matched well

to that of mouse GPR34 (Table 4). We consider the SAR for

mouse GPR34 activity in the following discussion.

Biarylethers (5a and 5b) were constructed by means of

Chan−Evans−Lam coupling.^14,15Reduction of the formyl

group, followed by condensation with the 3-phenylpropionic

acid methyl ester (7) by means of Mitsunobu reaction aﬀorded

the methyl ester intermediates (8a and 8b) of aromatic fatty

acid surrogates. Suzuki−Miyaura coupling aﬀorded compounds

containing an additional benzene ring (9a−9f). Basic hydrolysis

of the methyl ester (9a−9f) furnished the desired carboxylic

acids (10a−10f). Various fatty acid surrogates were synthesized

and coupled to the polar moiety intermediate 11, which was

synthesized through phosphoramidite coupling of the protected

phosphoserine part and 1,3-propandiol as previously de-

scribed¹²to furnish protected 2-deoxy-LysoPS analogues

When we introduced another benzene ring (ring B) near ring

A of our previously synthesized compound 1a, the resulting

compound 2a (Figure 1) exhibited very weak potency toward

C

DOI: 10.1021/acs.jmedchem.7b00693

J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

Scheme 1. General Synthetic Routes to 2-Deoxy-LysoPS Analogues

GPR34 (EC₅₀≥ 3 μM; Table 1 and Figures S1a and S2b).

When the second benzene moiety of 2a was replaced with a

triple bond (1b), the agonist activity was also very weak (EC₅₀

≥ 3 μM; for structures, see Table 1, Chart S1, and Figures S1a

and S2a). The second benzene ring inserted into the terminal

moiety of 1a did not greatly increase the GPR34-activating

potency (2b, EC₅₀= 1.7 μM; 2c, EC₅₀= 1.2 μM; Table 1 and

Figure S2b). When we introduced a third benzene moiety (ring

C) into the remaining ﬂexible alkyl chain moiety of 2a,

aﬀording analogues 3a−3e (for structures, see Table 1 and

Chart S1), compound 3a was as active as LysoPS (EC₅₀= 220

nM; Table 1 and Figures S1b and S2c). However, a slight

change in the conﬁguration of ring C abrogated the GPR34

activity (3b, EC₅₀≥ 3 μM; 3c, EC₅₀≥ 3 μM; 3d, EC₅₀= 1.3

μM; Table 1 and Figures S1b and S2c). Therefore, an

appropriate conﬁguration of ring C is important for GPR34

activation. The activity of saturated cyclohexyl-substituted

analogue 3e was higher than those of compounds 3b−3d

(3e: EC₅₀= 800 nM; Table 1 and Figures S1b and S2c),

suggesting that the presence of a hydrophobic moiety at the

terminal position is important for GPR34 activation and also

that π interaction is involved in 3a.

agonistic activity than previously found 1a toward GPR34, we

focused on derivatization of 3a to further characterize the

preferred environment around ring C. Various substituents

were introduced into ring C (3f−3v; for structures and

activities, see Table 2 and Figure 2a and b).

Introduction of a methyl group at the 2- (3f), 3- (3g), or 4-

position (3h) changed the GPR34-activating potency in

diﬀerent ways: the 4-Me group markedly increased the activity

(3h, EC₅₀= 39 nM), whereas the 2-Me or 3-Me groups

decreased the activity (3f, EC₅₀= 650 nM; 3g, EC₅₀= 550 nM)

compared with unsubstituted 3a (EC₅₀= 220 nM) (Table 2,

Figure 2a, and Figure S2d). Introduction of a chloro group at

the 2-, 3-, or 4-position of ring C showed a similar trend: the 3-

Cl or 4-Cl group showed a similar or slightly increased activity

compared with that of 3a (3j, EC₅₀= 170 nM; 3k, EC₅₀= 120

nM), whereas a 2-Cl group diminished the activity (3i, EC₅₀

=

1.2 μM). 3,4-Dichloro substitution caused synergistic activation

of GPR34 (3l, EC₅₀= 110 nM) (Table 2 and Figures S1c and

S2d). The results for these small probes suggest that ring C is

accommodated in the receptor pocket and that substitution at

the 4-position can be accommodated but that substitution at

the 2-position results in steric conﬂict. Other substituents at the

4-position, such as ﬂuoro (3m, EC₅₀= 120 nM), bromo (3n,

We also synthesized 2-deoxy-LysoPS analogues linked at the

meta- and para-positions of ring A of 1a. These meta-

substituted analogues (1c, 2d−2f, 3w, and 3x) and para-

substituted analogue (3y) lacked activity toward GPR34 (EC₅₀

≥ 3 μM; Table S1 and Figure S2a −c). These results are similar

to the SARs found previously for LysoPS derivatives.^12,13

Substituent Eﬀects at the Terminal Benzene Ring on

GPR34-Activating Activity. Because 3a showed more potent

EC₅₀= 82 nM), iodo (3o, EC₅₀= 100 nM), ethyl (3p, EC₅₀

=

180 nM), and methoxy (3q, EC₅₀= 260 nM) resulted in similar

or greater GPR34-activating potency to the unsubstituted

parent compound (3a, EC₅₀= 220 nM), but they were less

eﬀective than 4-methyl substitution (3h, EC₅₀= 39 nM) (Table

2 and Figures S1d and S2d). A hydrophilic cyano or amino

group at the 4-position diminished the activity toward GPR34

D

DOI: 10.1021/acs.jmedchem.7b00693

J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

a

Table 2. SAR of 3a Derivatives with Three Benzene Rings

a

Activities are represented in terms of EC₅₀. LogEC₅₀(M) and E_max(% AP-TGFα release) values are calculated from a sigmoidal concentration−

response curve and shown as mean SEM of indicated numbers of independent experiments (n). EC₅₀values are calculated from a mean value of

LogEC₅₀. RIA (relative intrinsic activity, a dimensionless parameter; ref 49) is an estimation of agonist activity and represented as an E_max/EC₅₀value

relative to that of a positive control compound (previous work). By deﬁnition, the RIA value of a positive control compound is equal to 1. “NA”

means “not available” because of very low activity.

(3r (4-CN), EC₅₀= 2.0 μM; 3s (4-NH₂), EC₅₀= 2.7 μM)

(Table 2 and Figure S2d). Therefore, the region of the binding

pocket adjacent to the 4-position of ring C may be

hydrophobic.

When a bulkier hydrophobic substituent such as a tert-butyl

group was introduced at the 2- or 3-position, GPR34-activating

potency changed in a similar manner to the case of a small

methyl or chloro group: the 2-tert-butyl group decreased the

activity (3t, EC₅₀≥ 3 μM) and the 3-tert-butyl-substituted

compound had almost the same activity as the unsubstituted 3a

(3u, EC₅₀= 210 nM (Table 2)). On the other hand, when a

tert-butyl group was introduced at the 4-position, GPR34-

activating potency was decreased signiﬁcantly as compared with

that of 3a (3v, EC₅₀= 900 nM) (Table 2, Figure 2b, and

E

DOI: 10.1021/acs.jmedchem.7b00693

J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

binding models by molecular dynamics. For prediction of the

ligand−protein binding mode, our SAR information (Table 2)

should be especially useful to deﬁne the hydrophobic binding

pocket of the acyl moiety. Although the SAR information in

this work is based on mouse GPR34, mouse and human GPR34

share high sequence identity (90%), and we conﬁrmed that the

rank-orders of activation potency of selected compounds

toward mouse and human GPR34s were similar (Table 4 and

Figure 3).

Figure 2. SAR of 2-Deoxy-LysoPS Derivatives for Mouse GPR34.

HEK293 cells transiently transfected with mouse-GPR34-encoding

expression vector or an empty vector were treated with compounds.

Receptor-speciﬁc AP-TGFα release responses were determined by

subtracting background responses in the empty vector-transfected

cells. Data are mean and SEM (standard error of the mean) for 3−6

independent experiments.

Figures S1e and S2d). In contrast, when an anisotropic phenyl

group was introduced, the analogue (4a (4-Ph), EC₅₀= 45 nM)

was as active as the 4-methyl derivative (3h (4-Me), EC₅₀= 39

nM) (Figure 1, Table 2, Figure 2b, and Figure S2e). These

observations are all consistent with the idea that ring C is

recognized by GPR34 and that the binding site contains a

hydrophobic pocket that can accommodate nonbulky sub-

stituents, especially at the 4-position of ring C.

Figure 3. SAR of 2-deoxy-LysoPS derivatives for human GPR34.

HEK293 cells transiently transfected with human-GPR34-encoding

expression vector or an empty vector were treated with compounds.

Receptor-speciﬁc AP-TGFα release responses were determined by

subtracting background responses in the empty vector-transfected

cells. Data are mean and SEM (standard error of the mean) for 3 or 4

independent experiments.

Homology Modeling of GPR34 and Computational

Docking with Active Compound 3a. The present SAR data

showed that GPR34 strictly recognizes the fatty acid moiety of

3a. To understand this ﬁnding in terms of receptor−ligand

interactions, we aimed to explore the structure of the active site

of GPR34 by means of computational docking of ligands to the

protein. Because the crystal structures of LysoPS receptors have

not been reported so far, we adopted a homology modeling/

molecular dynamics approach to obtain the human GPR34

structure. Thus, our approach to docking compounds to the

LysoPS binding sites of human GPR34 involved the following

three main steps: (1) homology modeling of GPR34, (2)

computational docking of compound 3a, (3) optimization of

Homology Modeling of GPR34. GPR34 belongs to the P2Y

family, which in turn belongs to the δ group of class A

GPCRs.¹⁷So far, crystal structures of four members of the δ

group GPCRs, PAR1,¹⁸P2Y1,¹⁹P2Y12,^20,21and GPR40,^22,23

have been solved. Among them, P2Y12 has the highest identity

with human GPR34 (26% identity in terms of the full-length

protein; 30% identity in terms of the modeling region after

truncation of the N- and C-terminals; Figure S3). Thus, we

chose P2Y12 (PDB ID: 4PXZ) in the agonist-bound state as a

template for GPR34 homology modeling.

F

DOI: 10.1021/acs.jmedchem.7b00693

J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

Computational Docking of Compound 3a. We obtained

19 binding models in total by docking ionized 3a to three

homology models of human GPR34 with the induced ﬁt

docking protocol^24,25followed by a 10 ns molecular dynamics

(MD) simulation, as described in the Experimental Section. In

all 19 models, the polar headgroup of the serine-phosphate

moiety of ligand 3a was located close to the exterior membrane

surface, and the hydrophobic fatty acid tail was located deep

inside the receptor pocket. This orientation is similar to that of

the original ligand in the P2Y12 crystal structure, 2-methylthio-

adenosine-5′-diphosphate (2MeSADP; for the chemical

structure, see Chart S2), which binds with its hydrophilic

phosphate group directed to the extracellular region and its

hydrophobic 2-methylthioadenine moiety embedded inside the

receptor.²¹The geometrical orientation of 3a was also similar

to that of the complex of another lysophospholipid receptor,

S1P₁, with its antagonist.²⁶

Optimization of Binding Models by Molecular Dynamics.

On the basis of the ligand RMSD values, the 19 binding poses

of 3a were clustered into ﬁve groups. Two of them represent

major conﬁgurations, consisting of 10 (cluster 1) and 6 poses

(cluster 2), respectively (Table S2). Representative binding

models (top 30% of cluster members) having the best MM-

GBSA ΔG_bindvalues were chosen from clusters 1 and 2

(designated as models 1-1, 1-2, and 1-3 (cluster 1) and models

2-1 and 2-2 (cluster 2)) (see Figure 4). These models were

subjected to a 100 ns MD simulation in an explicit membrane

environment to conﬁrm their stability.²⁷Figure 4 shows the

results for the ﬁve models after this calculation. The ligands did

not change their position drastically in the binding pocket

during the 100 ns MD simulation (Figure S4 and Table S2). In

models 1-1, 1-2, and 1-3 (Figure 4a−c, left), the hydrophobic

tail of 3a (blue balls) is surrounded by TM3, TM4, and TM5,

whereas in models 2-1 and 2-2 (Figures 4d and e, left), the

hydrophobic tail of 3a (blue balls) is surrounded by TM3,

TM5, and TM6. In particular, the orientation of ring C of 3a is

diﬀerent between models 1-1, 1-2, and 1-3 and models 2-1 and

2-2. On the other hand, the polar headgroup of 3a (orange

balls, except ring A) occupies a similar position in all models.

An interaction diagram of model 1-1 is shown in Figure S5 as a

representative example.

Selection and Evaluation of 3a Bound GPR34 Models

Based on the SAR Data. We examined the compatibility of

ﬁve models with the experimental SAR data for analogues of 3a.

1. Comparison of the Binding Pocket Environment

around Ring C of 3a. We characterized the shape and

physicochemical features of the receptor-binding pocket at

which the acyl moiety of 3a interacts by means of SiteMap

analysis.²⁸In all of the representative models, there is a

hydrophobic pocket around ring C, as shown in yellow in

Figure 4 (right). Ring C can interact hydrophobically with

aromatic amino-acid side chain(s) in models 1-1 (Tyr139^3.37

and Phe219^5.39), 1-2 (Tyr139^3.37and Phe219^5.39), and 2-1

(Tyr135^3.33) (Ballesteros−Weinstein nomenclature is also

shown).²⁹In particular, in models 1-1 and 1-2, the L-shaped

hydrophobic pocket ﬁts well to the ligand kink created by ether

oxygen between rings B and C. Furthermore, this L-shape

seems important for strict recognition of ring C because

derivative 3b, which is closely related to 3a, lacked activity

toward GPR34; because rings B and C of 3b are directly

connected, this molecule would not ﬁt the L-shaped pocket of

the receptor, impairing the hydrophobic interactions between

ring C and Tyr and Phe residues. These considerations are

Figure 4. Five representative binding models from clusters 1 and 2 (a,

model 1-1; b, model 1-2; c, model 1-3; d, model 2-1; e, model 2-2).

Each left ﬁgure presents the side view of the binding model of ligand

3a in complex with the human GPR34 model (ribbon). The

extracellular half of the receptor is shown. Each right ﬁgure presents

a close-up view of the acyl moiety of 3a (stick) in the receptor binding

site, calculated using SiteMap, and the receptor side chains near the

ligand terminal benzene ring (ring C). Rings B and C are colored blue,

and the other part of 3a is colored orange. The hydrophobic pocket is

shown as a yellow surface, and the hydrophilic pocket is shown as a

green surface. White dots represent the grid where placement of ligand

atoms is allowed in the SiteMap analysis. Leu181 and Ile222 in human

GPR34 are mutated to Val in mouse GPR34. As starting points for

molecular modeling of GPR34, the crystal structure of P2Y12 as the

agonist bound state (PDB ID: 4PXZ) was used.

G

DOI: 10.1021/acs.jmedchem.7b00693

J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

Figure 5. SiteMap analysis of methyl substitution of ring C in the ﬁve representative models. Only the regions accommodating ring C of ligands 3f−

3h are shown. Ligands 3f−3h are shown as blue sticks. The hydrophobic pocket is shown as a yellow surface, and the hydrophilic pocket is shown as

a green surface. White dots represent the grid where placement of ligand atoms is favorable in SiteMap analysis.

consistent with the SAR-based assessment that ring C of 3a is

strictly recognized by the receptor (Table 1).

To examine whether each of the ﬁve models could account

for the diﬀerence in potency between 3v and 4a, we examined

the binding poses obtained using Glide docking software

(Figure 6).^30,31

2. Inﬂuence of Methyl Substitution Position on Ring C.

The observed substituent eﬀects on ring C of 3a suggested that

substitution of a methyl or a chloro group at the 4-position on

ring C is favored over that at the 2- or 3-position for GPR34

activation (3f (2-Me), EC₅₀= 180 nM; 3g (3-Me), EC₅₀= 230

nM; 3h (4-Me), EC₅₀= 27 nM for human GPR34; Table 4,

Figure 3a, and Figure S2f). Methyl-substituted compounds 3f−

3h were docked to the ﬁve representative receptor models

shown in Figure 4 using the Glide docking program^30,31on the

assumption that the location of the hydrophilic serine-

phosphate moiety in the receptor pocket is the same in each

case (see Experimental Section) (Figure 5). In models 1-1 and

1-2, the 4-methyl compound (3h) occupied the hydrophobic

pocket of the receptor more eﬀectively than the 2- (3f) or 3-

methyl (3g) compound in accordance with the relative GPR34-

activating potencies. However, in models 1-3, 2-1, and 2-2, the

methyl group in all compounds 3f−3h interacted similarly with

the hydrophobic pocket of the receptor, and thus, these models

cannot account for the observed preference for 4-substitution.

3. Inﬂuence of Bulkiness of the 4-Substituent on Ring C. A

tert-butyl group at the 4-position on ring C (3v) markedly

decreased the GPR34 activity as compared with a 4-methyl

group (3h), whereas a phenyl group at the 4-position of ring C

(4a) was as eﬀective as the 4-methyl group (3h) for GPR34

activation (3h (4-Me), EC₅₀= 27 nM; 3v (4-t-Bu), EC₅₀= 300

nM; 4a (4-Ph), EC₅₀= 78 nM for human GPR34; Table 4,

Figure 3b, and Figure S2f). These results suggested that linear

and planar substitution at this position is compatible with

GPR34 activity.

In models 1-1 and 1-2, 3v was successfully docked to the

receptor (Figure 6a), but the tert-butyl group of 3v was not

accommodated well in the hydrophobic pocket region (yellow)

predicted by SiteMap.²⁸This might explain why 3v showed

decreased activity. In contrast, the phenyl group of docked 4a

was accommodated more eﬀectively in the hydrophobic pocket

of the receptor in models 1-1 and 1-2 in accordance with the

fact that 4a showed high activity. Detailed scrutiny of models 1-

1 and 1-2 revealed a narrowing of the hydrophobic pocket in

which the phenyl group could still be accommodated but the

tert-butyl group would clash (Figure 6b). On the other hand, in

model 1-3, the tert-butyl group of 3v and the phenyl group of

4a were both located outside the hydrophobic pocket region,

and thus, this model cannot explain the observed diﬀerence of

GPR34 activity between 3v and 4a. In model 2-1, the tert-butyl

group of 3v is fully accommodated in the hydrophobic pocket,

and thus this model cannot account for the decreased activity of

3v. Model 2-1 gave a binding pose for 4a, but it was completely

diﬀerent from the original binding mode of 3a in complex with

receptor model 2-1 as shown in Figure 4d (3a is also shown as

orange sticks in Figure 6a). Instead, the binding pose of 4a in

model 2-1 was similar to those in models 1-1 and 1-2: ring C

was placed between TM4 and TM5 (Figure 6a, shown as

purple sticks). On the other hand, model 2-2 did not give

binding poses for 3v and 4a.

These results indicate that models 1-1 and 1-2 are more

compatible with the SAR data than are models 1-3, 2-1, and 2-

2. Thus, we consider that an L-shaped binding pocket occupied

H

DOI: 10.1021/acs.jmedchem.7b00693

J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

Figure 6. SiteMap analysis of the eﬀects of 4-tert-butyl and 4-phenyl substitution in ﬁve representative receptor models. Only the regions

accommodating ring C of ligands 3v and 4a are shown. Ligands 3v and 4a are shown in stick form (3v, blue; 4a, purple). (a) The hydrophobic

pocket, analyzed by SiteMap, is shown as a yellow surface, and the hydrophilic pocket is shown as a green surface. White dots represent the grid

where ligand placement is favorable in SiteMap analysis. For 4a in complex with model 2-1, the hydrophobic moiety was placed in a diﬀerent pocket

from the pocket for the originally bound 3a (orange stick model) as shown in the model 2-1 column. (b) Top view of the narrowing between TM4

and TM5 in model 1-1. 1-Palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) molecules (gray lines) of the membrane bilayer are extracted

from the MD trajectory (after 100 ns) and superimposed onto the docking model. The binding pocket surface calculated using SiteMap is shown in

gray. White dots represent the grid where ligand atom placement is favorable in SiteMap analysis.

by the acyl moiety of analogues of 3a is likely to exist between

TM4 and TM5.

examine this idea, we docked 4b (Figure 1), which has a 4-tert-

butyl group on ring D, into model 1-1. In the resulting binding

model, aromatic ring D extends outside the receptor, and thus

4b should be GPR34 active (Figure 7). The same rationale is

4. Conﬁrmation of Selected Active-Site Models. It is

important to note that protein structure is ﬂexible, and

therefore, we carried out ensemble docking of 4a (4-Ph) with

multiple structures of GPR34 to conﬁrm that the hydrophobic

pocket is located between TM4 and TM5 (as in models 1-1

and 1-2) but not around TM3, TM5, and TM6 (as in models 2-

1 and 2-2). Speciﬁcally, 4a (4-Ph) was docked to ten structures

of GPR34 obtained from 100 ns MD simulations of ﬁve models

(ensemble docking, Table S3). Even taking protein dynamics

(i.e., the ﬂexibility of side chains of the receptor) into

consideration, binding of the acyl moiety around TM3, TM5,

and TM6 seems highly unlikely. These results provide further

support for the location of the hydrophobic binding pocket

between TM4 and TM5 as in models 1-1 and 1-2.

applicable to 4c (4-Ph), 4d (4-O-n-C₆H₁₃), and 4e (3-t-Bu).

Therefore, we next synthesized and evaluated derivatives

(4b−4e) bearing various substituents on ring D of the

Design of Agonists Based on the Selected GPR34−

Compound Binding Models. Finally, to conﬁrm the validity

of models 1-1 and 1-2, we used them to design a LysoPS

derivative as a candidate agonist. In models 1-1 and 1-2, an L-

shaped aperture exists between TM4 and TM5 of the receptor.

As a result, in the docking pose of 4a, the 4-phenyl group (ring

D) is placed out of the receptor to reach the membrane region

(Figure 6b). Thus, we considered that the activity would be

retained even if a bulky substituent (such as a tert-butyl group)

was attached to the terminal ring D of 4a because the

substituent would not disturb the ligand binding mode. To

Figure 7. Docking model of tetrabenzene-analogue 4b bearing a 4-tert-

butyl group on ring D. (a) Docking model of 4b (purple) with GPR34

model 1-1 (ribbon). (b) Top view showing rings C and D. POPC

molecules (gray lines) of the membrane bilayer are extracted from the

MD trajectory (after 100 ns) and superimposed onto the docking

model. The binding pocket surface calculated using SiteMap is shown

in gray. White dots represent the grid where ligand atom placement is

favorable in SiteMap analysis.

I

DOI: 10.1021/acs.jmedchem.7b00693

J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

a

Table 3. SAR of 4a Derivatives with Four Benzene Rings

a

Activities are represented in terms of EC₅₀. LogEC₅₀(M) and E_max(% AP-TGFα release) values are calculated from a sigmoidal concentration−

response curve and shown as mean SEM of indicated numbers of independent experiments (n). EC₅₀values are calculated from a mean value of

LogEC₅₀. RIA (relative intrinsic activity, a dimensionless parameter; ref 49) is an estimation of agonist activity and represented as an E_max/EC₅₀value

relative to that of a positive control compound (previous work). By deﬁnition, the RIA value of a positive control compound is equal to 1. “NA”

means “not available” because of very low activity.

tetrabenzene fatty acid surrogate (for structures, see Table 3

and Chart S1).

derivatives of 35, introduction of additional groups on the

benzene ring located outside the receptor has been reported to

have little disrupting eﬀect on receptor binding.^32,33Our result

is in good agreement with this ﬁnding in the case of GPR40.

Furthermore, in the crystal structure of GPR40 in complex with

(5aR,6S,6aS)-3-({2′,6′-dimethyl-4′-[3-(methylsulfonyl)-

propoxy][1,1′-biphenyl]-3-yl-methoxy}-5,5a,6,6a-

tetrahydrocyclopropa[4,5]cyclopenta[1,2-c]pyridine-6-carbox-

ylic acid (MK-8666 (36)) and (2S,3R)-3-cyclopropyl-3-[(2R)-

2-(1-{(1S)-1- [5-ﬂuoro-2-(triﬂuoromethoxy)phenyl]-ethyl}-

piperidin-4-yl)-3,4-dihydro-2H-1- benzopyran-7-yl]-2-methyl-

propanoic acid (AgoPAM AP8 (37); for the chemical

structures of 36 and 37, see Chart S2),²³36 also protruded

between TM3 and TM4 as in the case of 35 and allosteric

agonist, and 37 was bound in the lipid-facing binding pocket

between TM4 and TM5. During the binding process of a lipid-

like ligand, the pathway of the ligand entry from the membrane

bilayer was suggested in the sphingosine 1-phosphate receptor

subtype 1 (S1P₁)^26,34and the cannabinoid receptor subtype 2

(CB₂).³⁵These reports support the viewpoint that the

hydrophobic part of the ligand can interact with the membrane

bilayer or the receptor membrane-embedded surface upon

docking as shown in our present models (Figure 7b and also

see Figure 8).

As expected, 4b−4e exhibited activity toward GPR34 (4b,

EC₅₀= 7.9 nM; 4c, EC₅₀= 18 nM; 4d, EC₅₀= 49 nM; 4e, EC₅₀

= 14 nM for mouse GPR34, 4c, EC₅₀= 41 nM; 4e, EC₅₀= 15

nM for human GPR34; Tables 3 and 4, Figures 2c and 3c, and

Figure S2e and f). This is in sharp contrast to the ﬁnding that

the activity toward GPR34 was decreased when a tert-butyl

group was attached directly to ring C of 3a (3v, EC₅₀= 900

nM). In the case of 2-methyl substitution on ring D (4f (2-

Me)), the GPR34 activity was not as high as those of 4b−4e

(4f, EC₅₀= 78 nM; Table 3, Figure 2c, and Figure S2e), but 4f

retained considerable activity. The reduction of activity of 4f

can be interpreted in terms of increased steric hindrance

around the 4-position of ring C. Compound 4d, which bears an

n-hexyloxy group at the 4-position of ring D, showed high

potency toward GPR34 (4d, EC₅₀= 49 nM). Replacement of

the terminal methyl group with an amino group diminished the

activity (data not shown), supporting the idea that the

hydrophobic environment of the terminal substituent R of 4a

derivatives is exposed to the membrane bilayer. These results

further support the view that models 1-1 and 1-2 are plausible

binding models for GPR34 activation.

In the recently solved crystal structure of GPR40 complexed

with 2-[(3S)-6-[[3-[2,6-dimethyl-4-(3-methylsulfonylpropoxy)-

phenyl]phenyl]methoxy]-2,3-dihydro-1-benzofuran-3-yl]acetic

acid (TAK-875 (35); for the chemical structure, see Chart S2),

the ligand protruded between TM3 and TM4.²²In the case of

Finally, we docked the endogenous LysoPS (18:1) into

preferred model 1-1 using the Glide docking program.^30,31The

obtained binding mode suggested the cis-oleﬁn moiety of the

(18:1) oleoyl group ﬁtted well to the relevant L-shaped binding

J

DOI: 10.1021/acs.jmedchem.7b00693

J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

a

Table 4. SAR of 2-Deoxy-LysoPS Analogues for Human GPR34

a

Activities are represented in terms of EC₅₀. LogEC₅₀(M) and E_max(% AP-TGFα release) values are calculated from a sigmoidal concentration−

response curve and shown as mean SEM of indicated numbers of independent experiments (n). EC₅₀values are calculated from a mean value of

LogEC₅₀. RIA (relative intrinsic activity, a dimensionless parameter; ref 49) is an estimation of agonist activity and represented as an E_max/EC₅₀value

relative to that of a positive control compound (compound 6 in the previous work, ref 13). By deﬁnition, the RIA value of a positive control

compound is equal to 1. “NA” means “not available” because of very low activity.

fatty acid selectivity of GPR34, which favors unsaturated fatty

acids over saturated ones.

CONCLUSIONS

■

Our present focused SAR study revealed that GPR34-activating

potency strongly depends on the conﬁguration and substitution

of aromatic rings in a series of fatty acid surrogates. In this

study, we explored the preference of the receptor’s binding site

for the shape of the hydrophobic tail of 2-deoxy-LysoPS

analogues bearing aromatic ring(s) in the fatty acid moiety.

Tribenzene-containing compound 3a showed increased ago-

Figure 8. Binding model of (18:1) Oleoyl-LysoPS. The left ﬁgure

presents the side view of the binding model of LysoPS (18:1) in

complex with GPR34 model 1-1 (ribbon). The extracellular half of the

receptor is shown. The right ﬁgure presents a close-up view of the

oleic acid of LysoPS in the receptor binding site, calculated using

SiteMap, and the receptor side chains near the ligand terminus. The

cis-oleﬁn in oleic acid is colored purple to highlight the kink in the

ligand shape. The hydrophobic pocket is shown as a yellow surface,

and the hydrophilic pocket is shown as a green surface. White dots

represent the grid where placement of ligand atoms is allowed in the

SiteMap analysis. Leu181 and Ile222 in human GPR34 are mutated to

Val in mouse GPR34.

nistic activity toward GPR34. Our SAR study of 3a and its

analogues revealed strong dependence of GPR34-activating

ability upon substituents at the terminal aromatic substituent

(ring C). In particular, substitution at the 4-position of ring C

aﬀorded potent GPR34 agonists.

To understand the hydrophobic interactions between the

ligand and the receptor, we carried out computational docking

to binding-site models obtained by homology modeling/

molecular dynamics calculations based on the crystal structure

of agonist-bound human P2Y12. The preferred model of the

GPR34 receptor contains an L-shaped hydrophobic pocket,

which is complementary to the kink in the molecular shape of

the fatty acid surrogates 3a. Docking calculations with 3v, the

analogue having a tert-butyl group at the 4-position of terminal

ring C in the tribenzene system, indicated that the group would

be located in a narrow hydrophobic passage between TM4 and

TM5, where it would sterically interfere with the binding

pocket (Figure 8). This is consistent with the fact that LysoPS

(18:0), which bears a saturated fatty acid chain, favoring trans

conformations, shows signiﬁcantly decreased GPR34-activation

activity. Thus, the L-shaped pocket can explain the origin of the

K

DOI: 10.1021/acs.jmedchem.7b00693

J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

The combined organic layer was washed with brine, dried over

Na₂SO₄, and ﬁltered. The solvent was evaporated to give a colorless

oil. The residue was column-chromatographed on an open column

with silica-gel (n-hexane:AcOEt = 3:1) to aﬀord 6a as a colorless oil

(582.7 mg, 2.088 mmol, 93%).

pocket. Indeed, 3v (4-t-Bu) showed decreased activity

compared with that of 3a. On the other hand, docking

calculations predicted that introduction of a bulky tert-butyl

group at the 3- or 4-position of terminal ring D of 4a should

have little disturbing eﬀect on the interaction with the active

site. Accordingly, we synthesized 4b−4e and found that these

analogues indeed retained high agonistic activity toward

GPR34. Therefore, the strong dependency of GPR34 agonist

activity on the shape of the fatty acid moiety of LysoPS

analogues found in this work can be explained in terms of

interactions between the fatty acid surrogates of the LysoPS

analogue and the hydrophobic binding pocket of the receptor.

In other words, matching/mismatching of the kink in the

molecular shape of the benzene-based lipid surrogate with the

L-shaped hydrophobic binding pocket of GPR34 is a key

determinant of the activity. It is noteworthy that the position of

the kink in the bioactive lipid surrogates corresponds well to

the location of the cis-oleﬁn moiety of (18:1) oleoyl LysoPS, an

endogenous ligand and activator of GPR34 receptor. We

believe this feature of the hydrophobic pocket of GPR34 will be

helpful in the design of potent and speciﬁc modulators for use

in detailed studies of the pathophysiological roles of this

receptor.

¹H NMR (400 MHz, CDCl₃): 7.428 (2H, d, J = 9.0 Hz), 7.326 (1H,

dd, J = 7.9 Hz, 7.9 Hz), 7.112 (1H, dd, J = 7.6 Hz, 0.3 Hz), 7.008 (1H,

brs), 6.916 (1H, dd, J = 7.9 Hz, 1.1 Hz), 6.885 (2H, d, J = 9.0 Hz),

4.671 (2H, s), 1.753 (1H, brs). ¹³C NMR (100 MHz, CDCl₃): 157.07,

156.33, 143.09, 132.69, 129.99, 121.94, 120.58, 118.01, 117.17, 115.77,

64.81.

Methyl 3-(2-((3-(4-Bromophenoxy)benzyl)oxy)phenyl)-

propanoate (8a). To a solution of 7 (564.8 mg, 3.134 mmol), 6a

(563.9 mg, 2.020 mmol), and PPh₃(1.0263 g, 3.9128 mmol) in

anhydrous toluene (30 mL) was added dropwise DEAD (2 M in

toluene, 2.0 mL) at room temperature. The yellow mixture was heated

at 70−80 °C with stirring for 3 h under an Ar atmosphere; the color

changed to orange. The solvent was evaporated to give an orange oil

containing a white powder. The residue was column-chromatographed

on an open column with silica-gel (n-hexane:AcOEt = 10:1) to aﬀord

8a as a pink oil (640.0 mg, 1.450 mmol, 72%).

¹H NMR (400 MHz, CDCl₃): 7.433 (2H, d, J = 8.9 Hz), 7.353 (1H,

dd, J = 7.9 Hz, 7.9 Hz), 7.187−7.147 (3H, m), 7.060 (1H, brs), 6.943

(1H, dd, J = 8.4 Hz, 1.0 Hz), 6.919−6.845 (4H, m), 5.066 (2H, s),

3.640 (3H, s), 2.973 (2H, t, J = 7.9 Hz), 2.610 (2H, t, J = 7.8 Hz). ¹³C

NMR (100 MHz, CDCl₃): 173.56, 157.11, 156.27, 156.21, 139.47,

132.72, 130.13, 130.04, 129.13, 127.57, 122.03, 120.90, 120.67, 118.07,

117.25, 115.86, 111.53, 69.19, 51.49, 33.98, 26.13. HRMS (ESI-TOF)

EXPERIMENTAL SECTION

General Synthetic Procedures. Melting points were determined

with a Yanaco micro-melting point apparatus without correction. H

■

+

[M + Na]⁺calcd for C₂₃H₂₁BrNaO₄: 463.0515, 465.0495; found:

1

463.0498, 465.0483.

(400 MHz) and ¹³C (100 MHz) NMR spectra were recorded on a

Bruker Avance400. Chemical shifts were calibrated with tetramethylsi-

lane as an internal standard or with the solvent peak and are shown in

ppm (δ) values; coupling constants are shown in hertz (Hz). The

following abbreviations are used: s = singlet, d = doublet, t = triplet, q

= quartet, dd = double doublet, dt = double triplet, dq = double

quartet, h = hextet, m = multiplet, and brs = broad singlet. An electron

spray ionization time-of-ﬂight (ESI-TOF) mass spectrometer (Bruker

micrOTOF-05) was used to obtain high-resolution mass spectra

(HRMS). All commercially available compounds and solvents were

used as received. Combustion analyses were carried out in the

Microanalytical Laboratory of the Graduate School of Pharmaceutical

Sciences, the University of Tokyo. All tested compounds showed

≥95% purity on the basis of combustion analysis.

Methyl 3-(2-((3-([1,1′-Biphenyl]-4-yloxy)benzyl)oxy)phenyl)-

propanoate (9a). Compound 8a (300.4 mg, 0.6807 mmol) was

dissolved in EtOH (1.5 mL) and anhydrous toluene (3.0 mL).

Phenylboronic acid (133.5 mg, 1.095 mmol), H₂O (3 mL), Na₂CO₃

(317.8 mg, 2.998 mmol), PPh₃(13.3 mg, 0.0507 mmol), and

Pd(OAc)₂(6.2 mg, 0.028 mmol) were added to the solution. The

mixture was degassed by means of two freeze−pump−thaw cycles and

stirred at 80−90 °C under an Ar atmosphere for 17 h. H₂O (10 mL)

was added, and the aqueous layer was extracted three times with

AcOEt (15 mL). The combined organic layer was washed with brine,

dried over Na₂SO₄, and ﬁltered. The solvent was evaporated to leave a

colorless oil containing black droplets. The residue was column-

chromatographed on an open column with silica-gel (n-hexane:AcOEt

= 10:1) to aﬀord 9a as a colorless oil (256.9 mg, 0.5858 mmol, 86%).

¹H NMR (400 MHz, CDCl₃): 7.578−7.538 (4H, m), 7.447−7.404

(2H, m), 7.358 (1H, dd, J = 7.9 Hz, 7.9 Hz), 7.347−7.305 (1H, m),

7.183−7.143 (3H, m), 7.119 (1H, brs), 7.084 (2H, d, J = 8.7 Hz),

7.001 (1H, dd, J = 8.1 Hz, 0.9 Hz), 6.890 (1H, ddd, J = 7.1 Hz, 7.1 Hz,

1.0 Hz), 6.863 (1H, d, J = 7.7 Hz), 5.070 (2H, s), 3.607 (3H, s), 2.977

(2H, t, J = 7.7 Hz), 2.610 (2H, t, J = 7.7 Hz). ¹³C NMR (100 MHz,

CDCl₃): 173.67, 157.53, 156.46, 156.33, 140.48, 139.34, 136.50,

130.12, 129.94, 129.14, 128.76, 128.45, 127.56, 127.04, 126.88, 121.70,

120.84, 119.27, 118.05, 117.22, 111.55, 69.28, 51.44, 33.99, 26.15.

Details of the synthesis of compounds 1−3 and 13−34 are given in

the Supporting Information. The synthetic procedure for 4a (5a−12a)

is described below.

3-(4-Bromophenoxy)benzaldehyde (5a). To a 100 mL round-

bottomed ﬂask containing copper acetate (1.1055 g, 6.0864 mmol) in

anhydrous CH₂Cl₂(60 mL) were added pyridine (960 μL, 11.92

mmol), 4 Å molecular sieves, p-bromophenol (1.0568 g, 6.1084

mmol), and 3-formylphenylboronic acid (1.3488 g, 8.9956 mmol).

The resulting mixture was stirred for 47 h at room temperature.

Polyvinylpyridine (5.0550 g) was added, and stirring was continued for

10 min. The mixture was ﬁltered on Celite, and the solvent was

evaporated to give a green solid. The residue was column-

chromatographed on an open column with silica-gel (n-hexane:Et₂O

= 10:1) to aﬀord 5a as a colorless oil (644.7 mg, 2.326 mmol, 38%).

¹H NMR (400 MHz, CDCl₃): 9.964 (1H, s), 7.638 (1H, ddd, J =

7.6 Hz, 1.2 Hz, 1.2 Hz), 7.515 (1H, dd, J = 7.9 Hz, 7.9 Hz), 7.476 (2H,

d, J = 9.0 Hz), 7.452 (1H, dd, J = 2.4 Hz, 1.4 Hz), 7.277 (1H, ddd, J =

8.1 Hz, 2.6 Hz, 1.0 Hz), 6.919 (2H, d, J = 8.9 Hz). ¹³C NMR (100

MHz, CDCl₃): 191.38, 157.85, 155.43, 138.12, 133.00, 130.57, 125.20,

124.63, 121.06, 118.07, 116.73.

+

HRMS (ESI-TOF) [M + Na]⁺calcd for C₂₉H₂₆NaO₄: 461.1723;

found: 461.1733.

3-(2-((3-([1,1′-Biphenyl]-4-yloxy)benzyl)oxy)phenyl)propanoic

Acid (10a). Compound 9a (232.9 mg, 0.5311 mmol) was dissolved in

MeOH (1.0 mL) and THF (1.0 mL), and to the solution was added 2

M aqueous solution of NaOH (1.0 mL). The solution was stirred at

room temperature for 3.5 h, and then 2 M aqueous solution of HCl

was added to acidify the mixture to pH 2. The aqueous layer was

extracted three times with AcOEt (10 mL). The combined organic

layer was washed with brine, dried over Na₂SO₄, and ﬁltered. The

solvent was evaporated to leave a colorless oil. The residue was

column-chromatographed on an open column with silica-gel twice

(ﬁrst n-hexane:AcOEt = 2:1, second n-hexane:AcOEt = 6:1−2:1) to

aﬀord 10a as a white solid (189.5 mg, 0.4464 mmol, 84%).

¹H NMR (400 MHz, CDCl₃): 7.566−7.526 (4H, m), 7.429−7.391

(2H, m), 7.346 (1H, dd, J = 7.8 Hz, 7.8 Hz), 7.326−7.287 (1H, m),

7.186−7.147 (3H, m), 7.104−7.057 (3H, m), 6.994 (1H, dd, J = 8.1

(3-(4-Bromophenoxy)phenyl)methanol (6a). To a solution of 5a

(622.5 mg, 2.246 mmol) in MeOH (12 mL) was added portionwise

NaBH₄(169.1 mg, 4.470 mmol) at 0 °C. The reaction mixture was

stirred at 0 °C for 30 min and then diluted with CH₂Cl₂(10 mL),

saturated aqueous solution of NaHCO₃(5 mL), and water (15 mL).

The aqueous layer was extracted three times with CH₂Cl₂(20 mL).

L

DOI: 10.1021/acs.jmedchem.7b00693

J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

Hz, 0.9 Hz), 6.903−6.849 (2H, m), 5.056 (2H, s), 2.965 (2H, t, J = 7.7

Hz), 2.638 (2H, t, J = 7.7 Hz). ¹³C NMR (100 MHz, CDCl₃): 179.28,

157.56, 156.44, 156.33, 140.46, 139.25, 136.49, 130.12, 129.98, 128.80,

128.74, 128.44, 127.66, 127.02, 126.87, 121.69, 120.86, 119.32, 118.07,

117.19, 111.55, 69.29, 33.86, 25.84. HRMS (ESI-TOF) [M − H]⁻

2 × 10⁵cells/mL in 10% (v/v) fetal bovine serum (FBS)-

supplemented Dulbecco’s modiﬁed Eagle’s medium (DMEM) and

cultured in a CO₂incubator. After 24 h, the cells were transfected with

a mixture of plasmids encoding AP-TGFα (2.5 μg per dish), the

chimeric Gα_q/i1subunit (0.5 μg per dish) and GPR34 (2.5 μg per dish)

using Lipofectamine 2000 reagent (Life Technologies; 12.5 μL per

dish). As a control (mock transfection; for measuring receptor-

independent responses), cells were transfected with a mixture of

plasmids encoding AP-TGFα (2.5 μg per dish), the chimeric Gα_q/i1

subunit (0.5 μg per dish) and empty vector (2.5 μg per dish). After 24

h, cells were harvested with trypsin-EDTA solution, washed with PBS,

and suspended in Hank’s balanced salt solution (HBSS; 35 mL per

dish) containing 5 mM HEPES (pH 7.4). For GPR34-expressing cells

and the corresponding control cells, cell suspension was seeded in a

96-well plate (80 μL per well) and mixed with the LPA receptor

antagonist Ki16425 at a ﬁnal concentration of 10 μM (10 μL per well).

After 30 min incubation, cells were treated with compounds diluted in

0.01% (w/v) bovine serum albumin (BSA; essentially fatty acid-free

grade; Sigma-Aldrich) containing HBSS (10 μL per well; total volume

of 100 μL per well), and incubated for 1 h at 37 °C. The plate was

centrifuged for 2 min at 190g, and conditioned medium (CM) was

transferred to a blank 96-well plate (80 μL per well). AP solution (10

mM p-nitrophenyl phosphate (p-NPP), 40 mM Tris-HCl (pH 9.5), 40

mM NaCl, and 10 mM MgCl₂) was added to both the CM plate and

the cell plate (80 μL per well). Absorbance at 405 nm (OD₄₀₅) of the

two plates was measured before and after incubation for 1 h at room

temperature using a microplate reader. AP-TGFα release was

calculated according to the following equations, after subtracting

baseline responses in mock-transfected cells (also see Supporting

Information for details of the calculations)

−

calcd for C₂₈H₂₃O₄: 423.1602; found: 423.1631. Anal. Calcd for

C₂₈H₂₄O₄·0.1H₂O: C, 78.89; H, 5.72. Found: C, 78.80; H, 5.99. Mp

111.8−113.5 °C (recrystallized from n-hexane/CH₂Cl₂, colorless

needles).

tert-Butyl O-((3-((3-(2-((3-([1,1′-Biphenyl]-4-yloxy)benzyl)oxy)-

phenyl)propanoyl)-oxy)propoxy)(tert-butoxy)phosphoryl)-N-(tert-

butoxycarbonyl)-^L-serinate (12a). Compunds 10a (112.8 mg, 0.2657

mmol) and 11 (95.7 mg, 0.210 mmol) were dissolved in anhydrous

CH₂Cl₂(1.5 mL). DMAP (7.5 mg, 0.061 mmol) was added, and the

ﬂask was cooled to 0 °C in an ice−water bath. EDCI·HCl (79.0 mg,

0.412 mmol) was added at 0 °C, and the whole was stirred at room

temperature under an Ar atmosphere for 20 h. Additional EDCI·HCl

(39.8 mg) and MeOH (1.0 mL) were added to the reaction mixture,

and stirring was continued for an additional 24.3 h. H₂O (10 mL) was

added, and the aqueous layer was extracted three times with CH₂Cl₂

(15 mL). The combined organic layer was washed with brine, dried

over Na₂SO₄, and ﬁltered. The solvent was evaporated to give a

colorless oil. The residue was column-chromatographed on an open

column with silica-gel (n-hexane:AcOEt = 2:1 to 1:1) to yield 12a as a

colorless sticky oil (115.4 mg, 0.1339 mmol, 64%, a mixture of

diastereomers).

¹H NMR (400 MHz, CDCl₃): 7.581−7.542 (4H, m), 7.451−7.413

(2H, m), 7.383−7.309 (2H, m), 7.194−7.145 (3H, m), 7.112 (1H,

brs), 7.084 (2H, d, J = 8.7 Hz), 6.997 (1H, dd, J = 8.1 Hz, 0.8 Hz),

6.909−6.857 (2H, m), 5.499 (1H, d, J = 7.8 Hz), 5.084 (2H, s),

4.366−4.301 (2H, m), 4.231−4.176 (1H, m), 4.147−4.092 (2H, m),

4.027−3.978 (2H, m), 2.979 (2H, t, J = 7.6 Hz), 2.624 (2H, t, J = 7.7

Hz), 1.991−1.881 (2H, m), 1.473−1.432 (27H, m). ¹³C NMR (100

MHz, CDCl₃): 173.05, 168.32, 157.49, 156.43, 156.31, 155.21, 140.43,

139.34, 136.48, 130.09, 129.95, 129.01, 128.75, 128.43, 127.57, 127.03,

126.85, 121.64, 120.81, 119.25, 117.99, 117.20, 111.57, 83.68, 83.61,

82.64, 82.61, 79.90, 69.27, 67.38, 64.03, 63.98, 60.39, 54.43, 54.35,

34.00, 29.75, 29.74, 29.71, 29.51, 29.44, 28.26, 27.90, 26.08. ³¹P NMR

(161 MHz, CDCl₃): −5.56, −5.70. HRMS (ESI-TOF) [M + Na]⁺

calcd for C₄₇H₆₀NNaO₁₂P⁺: 884.3745; found: 884.3743.

AP‐TGFα_CM(%) = (OD405_CM/(OD405_CM+ OD405_cell))

× 1.25 × 100

where factor 1.25 was used to obtain AP-TGFα in total CM (100 μL)

from measured AP-TGFα in transferred CM (80 μL).

AP‐TGFα release = AP‐TGFα_CMunder stimulated conditions (%)

− AP‐TGFα_CMunder vehicle‐treated conditions (%)

Typically, AP-TGFα_CMunder vehicle-treated conditions ranged from

10 to 25% depending on the transfected plasmids and assay

conditions.

O-((3-((3-(2-((3-([1,1′-Biphenyl]-4-yloxy)benzyl)oxy)phenyl)-

propanoyl)oxy)propoxy)(hydroxy)phosphoryl)-^L-serine (4a). Com-

pound 12a (98.4 mg, 0.114 mmol) was dissolved in TFA (1.0 mL) at 0

°C, and the solution was stirred at room temperature for 1.5 h, diluted

with CH₂Cl₂(6 mL), and evaporated. The residue was subjected to

ﬂash column chromatography (CHCl₃:MeOH:AcOH = 9:1:0 to

8:1:1). The resulting colorless oil was dissolved in CH₂Cl₂and TFA at

0 °C, and the solution was evaporated at 40 °C overnight to aﬀord 4a

as the TFA salt (white powder, 41.8 mg, 0.0644 mmol, 56%).

¹H NMR (400 MHz, CDCl₃/TFA-d): 7.578−7.550 (4H, m),

7.452−7.434 (2H, m), 7.380−7.315 (2H, m), 7.213−7.162 (2H, m),

7.116−7.056 (4H, m), 7.011 (1H, dd, J = 8.1, 1.0 Hz), 6.917−6.883

(2H, m), 5.095 (2H, s), 4.600−4.587 (2H, m), 4.474 (1H, brs), 4.169

(2H, t, J = 6.3 Hz), 4.018 (2H, brs), 2.968 (2H, t, J = 7.4 Hz), 2.716

Data are shown as receptor-speciﬁc responses calculated as

receptor‐specific AP‐TGFα release (%) = AP‐TGFα

release in receptor‐transfected cells (%) − AP‐TGFα

release in mock‐transfected cells (%)

LogEC₅₀(EC₅₀) and E_maxvalues were calculated for active compounds

with plateau or semiplateau responses by ﬁtting data to a four-

parameter sigmoid curve using GraphPad Prism 6 (GraphPad, USA).

Homology Modeling. GPR34 homology models were built using

MODELER 9.13³⁶based on the crystal structure of P2Y12 (PDB ID:

4PXZ) bound to the P2Y12 full agonist 2-methylthio-adenosine-5′-

diphosphate (2MeSADP).²¹The human GPR34 sequence was

obtained from UniProt (code: Q9UPC5).³⁷The N-terminal (from

Met1 to Asn42) and C-terminal (from Arg345 to Thr381) regions of

the GPR34 sequence were deleted because template structure was not

available for these regions. Sequence alignment was conducted with

(2H, t, J = 7.5 Hz), 1.943−1.914 (2H, m). ³¹P NMR (CDCl₃/TFA-d):

−

−1.36. HRMS (ESI-TOF) [M − H]⁻calcd for C₃₄H₃₅NO₁₀P

:

648.2004; found: 648.2014. Anal. Calcd for C₃₄H₃₆NO₁₀P·

1.5CF₃CO₂H: C, 54.15; H, 4.61; N, 1.71. Found: C, 54.21; H, 4.86;

N, 1.78. Mp 112−144 °C (colorless cubes).

TGFα Shedding Assay. TGFα shedding assay of the synthesized

compounds was performed using mouse GPR34 LysoPS receptors, as

described previously.¹⁶This assay detects activation of GPCRs by

measuring ectodomain shedding of a membrane-bound proform of

alkaline phosphatase-tagged TGFα (AP-TGFα) into conditioned

media. Because AP-TGFα shedding response occurs almost exclusively

downstream of Gα_12/13and Gα_qsignaling, we additionally transfected

a chimeric Gα_q/i1subunit to detect activation of G_i-coupled GPR34.

Brieﬂy, agonistic activities of all synthesized compounds toward

LysoPS receptors were measured as follows. HEK293A cells were

seeded in 100 mm dishes with 10 mL of cell suspension at a density of

ClustalW³⁸using Multiple Sequence Viewer (Schrodinger, LLC, NY)

̈

as a graphical interface. The alignment was manually modiﬁed to

match the cysteine residues that form two conserved disulﬁde bonds in

GPR34 and P2Y12 (Cys46-Cys299 and Cys127-Cys204 in GPR34)³⁹

and to ﬁll the gap in the transmembrane helices. The ﬁnal alignment is

shown in Figure S3 with the percentage identity (30%). The region

from Ser331 to Phe343 in GPR34 was constrained to α-helix form to

create helix 8. Ten models were constructed and three models that had

the best MODELER DOPE scores were used in the next docking

process.

M

DOI: 10.1021/acs.jmedchem.7b00693

J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

Docking of GPR34 Agonist 3a. The constructed GPR34

homology models were reﬁned for docking simulations with Protein

for 3f and 3h, respectively, the docking pose of 3g in which ring C was

placed around TM3, TM5, and TM6 was selected.

In the postdocking minimization stage, strain correction terms were

applied. SiteMap (version 3) analysis²⁸was conducted with each

representative model (models 1-1, 1-2, 1-3, 2-1, and 2-2) using default

parameters and compared with the binding pose of each compound.

The ﬁgures were created on the graphical user interface of Maestro or

Preparation Wizard in Maestro (Schrodinger, LLC, NY)⁴⁰using

̈

PROPKA with pH 7.0 in the H-bond assignment step.⁴¹Suitable 3D

coordinates and states of compound 3a were generated using the

OPLS2005 force ﬁeld and the Epik calculation with pH 7.0 on LigPrep

(version 3, Schrodinger, LLC, NY). The most suitable ionization states

̈

PyMOL Molecular Graphics System (version 1.7, Schrodinger, LLC,

̈

of the carboxyl, amino, and phosphodiester groups to give the total

molecular charge −1 were used.

NY).

Evaluation of 4a Binding Pose by Ensemble Docking. To

further evaluate the docking pose of 4a, we extracted 9 additional

structures from the trajectories of the 100 ns MD simulation, every 5

ns from 55 to 95 ns. Protein structures were prepared, and 4a was

docked to each structure in the same way as mentioned above. The

position of the acyl moiety of 4a in each docking model was visually

checked, and the results are summarized in Table S3.

Docking of LysoPS (18:1). LysoPS (18:1) was docked to the

reﬁned receptor model after 100 ns MD simulation using Glide SP

mode. The centroid of complexed 3a was deﬁned as the grid center.

Postdocking minimization was conducted for 25 poses, and among

them, 5 poses were written. The docking pose with the best docking

score was selected (Figure 8).

Compound 3a was docked to the three GPR34 homology models

using the standard induced ﬁt docking (IFD) protocol (Schrodinger,

̈

LLC, NY).^24,25Box center was deﬁned as the centroid of the P2Y12

agonist implemented in GPR34 homology models. In the IFD

protocol, 3a was ﬁrst docked to the rigid protein using Glide (version

6, Schrodinger, LLC, NY)^30,31with the softened potential of van der

̈

Waals radii scaling of 0.5 for the protein and 0.5 for the ligand. The

resulting top 20 poses of ligands were used in the next Prime side-

chain prediction step. Residues within 5 Å of any ligand pose were

subjected to conformational search and minimization with an implicit

membrane using Prime (version 3, Schrodinger, LLC, NY).^42,43The

̈

resulting 20 protein conformers were used for the redocking process in

Glide SP mode with the default potential of van der Waals radii scaling

of 1.0 for the protein and 0.8 for the ligand. Nineteen complex

structures were thus generated from the three homology models.

Molecular Dynamics. The 19 models of GPR34 complexed with

3a were subjected to 10 ns MD-based energy minimization using

Desmond 3.8.5.18.⁴⁴The OPLS2005 force ﬁeld was used. Initial

ligand−receptor complex models were aligned to the P2Y12 structure

obtained from PDBTM⁴⁵⁻⁴⁷and embedded in the POPC membrane

placed on the prealigned structure. The system was placed in TIP3P

water molecules solvated with 0.15 M NaCl. The system was ﬁrst

subjected to relaxation MD on CPU Desmond and then to production

MD on GPU Desmond for 10 ns. Production MD was performed in

the NPγT ensemble at 300 K with 1.01325 bar and 4000 bar·Å as

surface tension using Langevin dynamics. Long-range electrostatic

interactions were computed using the Smooth Particle Mesh Ewald

method.⁴⁸

ASSOCIATED CONTENT

* Supporting Information

■

S

The Supporting Information is available free of charge on the

ACS Publications website at DOI: 10.1021/acs.jmed-

chem.7b00693.

Additional graphics, structures of compounds, and full

biological data, including details of the syntheses of the

compounds (PDF)

Molecular formula strings (CSV)

Compound 3a in model 1-1 (PDB)

Compound 3a in model 2-1 (PDB)

AUTHOR INFORMATION

Corresponding Authors

The GPR34−3a complex models after 10 ns MD simulation were

extracted and reﬁned using Protein Preparation Wizard into Maestro

■

(Schrodinger, LLC, NY). All 19 complex structures were aligned using

̈

*Chemical study (T.O.): phone, +81-3-5841-4730; fax, +81-3-

5841-4735; e-mail, ohwada@mol.f.u-tokyo.ac.jp.

*Biological study (J.A.): phone, +81-22-795-6860; fax, +81-22-

795-6859; e-mail, jaoki@m.tohoku.ac.jp.

the Protein Structure Alignment tool in Maestro (Schrodinger, LLC,

̈

NY). Ligands were extracted from the complexes, and ligand poses

were clustered according to the ligand heavy atom RMSDs in place in

Maestro. The 19 poses were divided into ﬁve clusters of which two

were major (namely, cluster 1 containing 10 poses and cluster 2

containing 6 poses).

*Computational study (T.H.): phone, +813-3599-8638; fax,

+81-3-3599-2064; e-mail, t-hirokawa@aist.go.jp.

Prime MM-GBSA ΔG_bindwas calculated with the VSGB solvation

model using input ligand partial charges and an implicit membrane.

The MM-GBSA ΔG_bindscores are listed in Table S2 for the 19

complexes. The top scored three complexes from cluster 1 (models 1-

1, 1-2, and 1-3) and two complexes from cluster 2 (models 2-1 and 2-

2) were considered for further MD-based energy minimization.

Additional 100 ns production MD on GPU Desmond was performed

from the last snapshot of the previous 10 ns MD simulation (Figure

S4). The conditions were the same as those of the 10 ns simulation.

Ligand Docking and SiteMap Analysis. The structure of the

receptor model after the 100 ns MD simulation was extracted from the

MD trajectory and reﬁned using Protein Preparation Wizard in

ORCID

Yuko Otani: 0000-0002-8104-283X

Tomohiko Ohwada: 0000-0001-5390-0203

Author Contributions

^∇M.S. and A.I. contributed equally to this work.

Author Contributions

M.S., S.N., S.J., M.I., Y.O., and T.O. (The University of Tokyo)

performed the chemical studies including design and synthesis

of compounds. A.I., A.U., T.K., K.M., and J.A. (Tohoku

University) performed biological study using the TGFα

shedding assay. M.S. (The University of Tokyo) and T.H.

(AIST and Tsukuba University) performed the computational

studies of protein−ligand binding models.

Maestro (Schrodinger, LLC, NY). We selected key compounds (3f−

̈

3h, 3v, 4a, and 4b) based on the observed SAR for docking

experiments. The selected 3a derivatives were docked to the reﬁned

structures using Glide SP mode. The centroid of the complexed 3a was

deﬁned as the grid center. In ligand docking, the docking position was

restricted to the reference position including the serine-phosphate

moiety, as shown in Figure S6. Three types of reference core atoms

were tried, and the pose with the best docking score was selected

except in the case of 3g docked to model 2-1, where the docking pose

in which ring C was placed between TM4 and TM5 gave the best

docking score. However, for easy comparison with the docking poses

Notes

The authors declare no competing ﬁnancial interest.

ACKNOWLEDGMENTS

■

This research was supported by grants from the Ministry of

Education, Science, Sports, and Culture of Japan to J.A., T.O.,

and T.H. We thank Prof. Dr. Masaji Ishiguro, Niigata

N

DOI: 10.1021/acs.jmedchem.7b00693

J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

(7) Jin, Z. T.; Li, K.; Li, M.; Ren, Z. G.; Wang, F. S.; Zhu, J. Y.; Leng,

X. S.; Yu, W. D. G-protein coupled receptor 34 knockdown impairs

the proliferation and migration of HGC-27 gastric cancer cells in vitro.

Chin. Med. J. (Engl) 2015, 128, 545−549.

University of Pharmacy and Applied Life Sciences, Niigata,

Japan, for his initial guidance of the computational study. This

work was supported by the Platform for Drug Discovery,

Informatics, and Structural Life Science from the Ministry of

Education, Culture, Sports, Science and Technology, Japan.

This work was supported by JST, PRESTO Grant Number

JPMJPR1331 (to A.I.); the PRIME from Japan Agency for

Medical Research and development, AMED (to A.I); Japan

Society for the Promotion of Science (JSPS) Grant Number

16H05085 (J.A.); AMED-CREST, AMED (to J.A.).

(8) Ansell, S. M.; Akasaka, T.; McPhail, E.; Manske, M.; Braggio, E.;

Price-Troska, T.; Ziesmer, S.; Secreto, F.; Fonseca, R.; Gupta, M.; Law,

M.; Witzig, T. E.; Dyer, M. J.; Dogan, A.; Cerhan, J. R.; Novak, A. J.

t(X;14)(p11;q32) in MALT lymphoma involving GPR34 reveals a

role for GPR34 in tumor cell growth. Blood 2012, 120, 3949−3957.

(9) Iida, Y.; H Tsuno, N.; Kishikawa, J.; Kaneko, K.; Murono, K.;

Kawai, K.; Ikeda, T.; Ishihara, S.; Yamaguchi, H.; Sunami, E.;

Kitayama, J.; Yatomi, Y.; Watanabe, T. Lysophosphatidylserine

stimulates chemotactic migration of colorectal cancer cells through

GPR34 and PI3K/Akt pathway. Anticancer Res. 2014, 34, 5465−5472.

(10) Iwashita, M.; Makide, K.; Nonomura, T.; Misumi, Y.; Otani, Y.;

Ishida, M.; Taguchi, R.; Tsujimoto, M.; Aoki, J.; Arai, H.; Ohwada, T.

Synthesis and evaluation of lysophosphatidylserine analogues as

inducers of mast cell degranulation. Potent activities of lysophospha-

tidylthreonine and its 2-deoxy derivative. J. Med. Chem. 2009, 52,

5837−5863.

(11) Uwamizu, A.; Inoue, A.; Suzuki, K.; Okudaira, M.; Shuto, A.;

Shinjo, Y.; Ishiguro, J.; Makide, K.; Ikubo, M.; Nakamura, S.; Jung, S.;

Sayama, M.; Otani, Y.; Ohwada, T.; Aoki, J. Lysophosphatidylserine

analogues differentially activate three LysoPS receptors. J. Biochem.

2015, 157, 151−160.

(12) Ikubo, M.; Inoue, A.; Nakamura, S.; Jung, S.; Sayama, M.; Otani,

Y.; Uwamizu, A.; Suzuki, K.; Kishi, T.; Shuto, A.; Ishiguro, J.;

Okudaira, M.; Kano, K.; Makide, K.; Aoki, J.; Ohwada, T. Structure-

activity relationships of lysophosphatidylserine analogs as agonists of

G-protein-coupled receptors GPR34, P2Y10, and GPR174. J. Med.

Chem. 2015, 58, 4204−4219.

(13) Jung, S.; Inoue, A.; Nakamura, S.; Kishi, T.; Uwamizu, A.;

Sayama, M.; Ikubo, M.; Otani, Y.; Kano, K.; Makide, K.; Aoki, J.;

Ohwada, T. Conformational constraint of the glycerol moiety of

lysophosphatidylserine affords compounds with receptor subtype

selectivity. J. Med. Chem. 2016, 59, 3750−3776.

(14) Chan, D. M.; Monaco, K. L.; Wang, R.-P.; Winters, M. P. New

N-and O-arylations with phenylboronic acids and cupric acetate.

Tetrahedron Lett. 1998, 39, 2933−2936.

(15) Evans, D. A.; Katz, J. L.; West, T. R. Synthesis of diaryl ethers

through the copper-promoted arylation of phenols with arylboronic

acids. An expedient synthesis of thyroxine. Tetrahedron Lett. 1998, 39,

2937−2940.

ABBREVIATIONS USED

■

LysoPS, lysophosphatidylserine; PS, phosphatidylserine; LPS,

lysophosphatidylserine receptor; S1P, Sphingosine-1-phos-

phate; TGFα, transforming growth factor-alpha; AP-TGFα,

alkaline phosphatase-tagged TGFα; TM, transmembrane helix;

EL, extracellular loop; 2MeSADP, 2-methylthio-adenosine-5′-

diphosphate; DOPE, discrete optimized protein energy; MD,

molecular dynamics; MM-GBSA, molecular mechanics with

generalized Born and surface area solvation; SP, standard

precision; Mp, melting point; EDCI, N-(3-(dimethylamino)-

propyl)-N-ethylcarbodiimide hydrochloride; MS, molecular

sieves; HBSS, Hanks’ balanced salt solution; HEPES, 4-(2-

hydroxyethyl)piperazine-1-ethanesulfonic acid; CM, condi-

tioned media; DMEM, Dulbecco’s modiﬁed Eagle’s medium;

DAPI, 4′,6-diamidino-2-phenylindole; CI, conﬁdence interval;

SEM, standard errors of mean; IFD, induced ﬁt docking; GPU,

graphics processing unit; POPC, 1-palmitoyl-2-oleoylphospha-

tidylcholine; PDBTM, Protein Data Bank of Transmembrane

Proteins; NPγT, constant normal pressure and lateral surface

tension of membranes and constant temperature

REFERENCES

■

(1) Sugo, T.; Tachimoto, H.; Chikatsu, T.; Murakami, Y.; Kikukawa,

Y.; Sato, S.; Kikuchi, K.; Nagi, T.; Harada, M.; Ogi, K.; Ebisawa, M.;

Mori, M. Identification of a lysophosphatidylserine receptor on mast

cells. Biochem. Biophys. Res. Commun. 2006, 341, 1078−1087.

(2) Kitamura, H.; Makide, K.; Shuto, A.; Ikubo, M.; Inoue, A.; Suzuki,

K.; Sato, Y.; Nakamura, S.; Otani, Y.; Ohwada, T.; Aoki, J. GPR34 is a

receptor for lysophosphatidylserine with a fatty acid at the sn-2

position. J. Biochem. 2012, 151, 511−518.

(16) Inoue, A.; Ishiguro, J.; Kitamura, H.; Arima, N.; Okutani, M.;

Shuto, A.; Higashiyama, S.; Ohwada, T.; Arai, H.; Makide, K. TGFα

shedding assay: an accurate and versatile method for detecting GPCR

activation. Nat. Methods 2012, 9, 1021−1029.

(3) Makide, K.; Kitamura, H.; Sato, Y.; Okutani, M.; Aoki, J.

Emerging lysophospholipid mediators, lysophosphatidylserine, lyso-

phosphatidylthreonine, lysophosphatidylethanolamine and lysophos-

phatidylglycerol. Prostaglandins Other Lipid Mediators 2009, 89, 135−

139.

(17) Fredriksson, R.; Lagerstrom, M. C.; Lundin, L.-G.; Schioth, H.

̈

B. The G-protein-coupled receptors in the human genome form five

main families. Phylogenetic analysis, paralogon groups, and finger-

prints. Mol. Pharmacol. 2003, 63, 1256−1272.

(4) Liebscher, I.; Muller, U.; Teupser, D.; Engemaier, E.; Engel, K. M.

̈

(18) Zhang, C.; Srinivasan, Y.; Arlow, D. H.; Fung, J. J.; Palmer, D.;

Zheng, Y.; Green, H. F.; Pandey, A.; Dror, R. O.; Shaw, D. E.; Weis,

W. I.; Coughlin, S. R.; Kobilka, B. K. High-resolution crystal structure

of human protease-activated receptor 1. Nature 2012, 492, 387−392.

(19) Zhang, D.; Gao, Z.-G.; Zhang, K.; Kiselev, E.; Crane, S.; Wang,

J.; Paoletta, S.; Yi, C.; Ma, L.; Zhang, W.; et al. Two disparate ligand-

binding sites in the human P2Y1 receptor. Nature 2015, 520, 317−

321.

Y.; Ritscher, L.; Thor, D.; Sangkuhl, K.; Ricken, A.; Wurm, A.; Piehler,

D.; Schmutzler, S.; Fuhrmann, H.; Albert, F. W.; Reichenbach, A.;

Thiery, J.; Schoneberg, T.; Schulz, A. Altered immune response in

̈

mice deficient for the G protein-coupled receptor GPR34. J. Biol.

Chem. 2011, 286, 2101−2110.

(5) (a) Preissler, J.; Grosche, A.; Lede, V.; Le Duc, D.; Krugel, K.;

̈

Matyash, V.; Szulzewsky, F.; Kallendrusch, S.; Immig, K.; Kettenmann,

H.; Bechmann, I.; Schoneberg, T.; Schulz, A. Altered microglial

̈

(20) Zhang, K.; Zhang, J.; Gao, Z.-G.; Zhang, D.; Zhu, L.; Han, G.

W.; Moss, S. M.; Paoletta, S.; Kiselev, E.; Lu, W.; Fenalti, G.; Zhang,

W.; Muller, C. E.; Yang, H.; Jiang, H.; Cherezov, V.; Katritch, V.;

Jacobson, K. A.; Stevens, R. C.; Wu, B.; Zhao, Q. Structure of the

human P2Y₁₂receptor in complex with an antithrombotic drug. Nature

2014, 509, 115−118.

phagocytosis in GPR34-deficient mice. Glia 2015, 63, 206−215.

(b) Tokizane, K.; Konishi, H.; Makide, K.; Kawana, H.; Nakamuta, S.;

Kaibuchi, K.; Ohwada, T.; Aoki, J.; Kiyama, H. Phospholipid

localization implies microglial morphology and function via Cdc42

in vitro. Glia 2017, 65, 740−755.

(6) Yu, W.; Ma, S.; Wang, L.; Zuo, B.; Li, M.; Qiao, Z.; Pan, X.; Liu,

Y.; Wang, J. Upregulation of GPR34 expression affects the progression

and prognosis of human gastric adenocarcinoma by PI3K/PDK1/AKT

pathway. Histol. Histopathol. 2013, 28, 1629−1638.

(21) Zhang, J.; Zhang, K.; Gao, Z.-G.; Paoletta, S.; Zhang, D.; Han,

G. W.; Li, T.; Ma, L.; Zhang, W.; Muller, C. E.; Yang, H.; Jiang, H.;

̈

Cherezov, V.; Katritch, V.; Jacobson, K. A.; Stevens, R. C.; Wu, B.;

O

DOI: 10.1021/acs.jmedchem.7b00693

J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

Zhao, Q. Agonist-bound structure of the human P2Y₁₂receptor.

Nature 2014, 509, 119−122.

(37) UniProt Consortium. UniProt: a hub for protein information.

Nucleic Acids Res. 2015, 43 (D1), D204−D212.

(38) Thompson, J. D.; Higgins, D. G.; Gibson, T. J. CLUSTAL W:

improving the sensitivity of progressive multiple sequence alignment

through sequence weighting, position-specific gap penalties and weight

matrix choice. Nucleic Acids Res. 1994, 22, 4673−4680.

(22) Srivastava, A.; Yano, J.; Hirozane, Y.; Kefala, G.; Gruswitz, F.;

Snell, G.; Lane, W.; Ivetac, A.; Aertgeerts, K.; Nguyen, J.; Jennings, A.;

Okada, K. High-resolution structure of the human GPR40 receptor

bound to allosteric agonist TAK-875. Nature 2014, 513, 124−127.

(23) Lu, J.; Byrne, N.; Wang, J.; Bricogne, G.; Brown, F. K.;

Chobanian, H. R.; Colletti, S. L.; Di Salvo, J.; Thomas-Fowlkes, B.;

Guo, Y.; Hall, D. L.; Hadix, J.; Hastings, N. B.; Hermes, J. D.; Ho, T.;

Howard, A. D.; Josien, H.; Kornienko, M.; Lumb, K. J.; Miller, M. W.;

Patel, S. B.; Pio, B.; Plummer, C. W.; Sherborne, B. S.; Sheth, P.;

Souza, S.; Tummala, S.; Vonrhein, C.; Webb, M.; Allen, S. J.; Johnston,

J. M.; Weinglass, A. B.; Sharma, S.; Soisson, S. M. Structural basis for

the cooperative allosteric activation of the free fatty acid receptor

GPR40. Nat. Struct. Mol. Biol. 2017, Pub ASAP.

(39) Ritscher, L.; Engemaier, E.; Staeubert, C.; Liebscher, I.; Schmidt,

̈

P.; Hermsdorf, T.; Roempler, H.; Schulz, A.; Schoeneberg, T. The

̈

ligand specificity of the G-protein-coupled receptor GPR34. Biochem. J.

2012, 443, 841−850.

(40) Sastry, G. M.; Adzhigirey, M.; Day, T.; Annabhimoju, R.;

Sherman, W. Protein and ligand preparation: parameters, protocols,

and influence on virtual screening enrichments. J. Comput.-Aided Mol.

Des. 2013, 27, 221−234.

(41) Olsson, M. H.; Søndergaard, C. R.; Rostkowski, M.; Jensen, J. H.

PROPKA3: consistent treatment of internal and surface residues in

empirical pKa predictions. J. Chem. Theory Comput. 2011, 7, 525−537.

(42) Jacobson, M. P.; Pincus, D. L.; Rapp, C. S.; Day, T. J. F.; Honig,

B.; Shaw, D. E.; Friesner, R. A. A Hierarchical approach to all-atom

protein loop prediction. Proteins: Struct., Funct., Genet. 2004, 55, 351−

367.

(24) Sherman, W.; Day, T.; Jacobson, M. P.; Friesner, R. A.; Farid, R.

Novel procedure for modeling ligand/receptor induced fit effects. J.

Med. Chem. 2006, 49, 534−553.

(25) Sherman, W.; Beard, H. S.; Farid, R. Use of an induced fit

receptor structure in virtual screening. Chem. Biol. Drug Des. 2006, 67,

83−84.

(26) Hanson, M. A.; Roth, C. B.; Jo, E.; Griffith, M. T.; Scott, F. L.;

Reinhart, G.; Desale, H.; Clemons, B.; Cahalan, S. M.; Schuerer, S. C.;

Sanna, M. G.; Han, G. W.; Kuhn, P.; Rosen, H.; Stevens, R. C. Crystal

structure of a lipid G protein−coupled receptor. Science 2012, 335,

851−855.

(27) Sabbadin, D.; Ciancetta, A.; Moro, S. Bridging molecular

docking to membrane molecular dynamics to investigate GPCR−

ligand recognition: The human A_2Aadenosine receptor as a key study.

J. Chem. Inf. Model. 2014, 54, 169−183.

(28) Halgren, T. New method for fast and accurate binding-site

identification and analysis. Chem. Biol. Drug Des. 2007, 69, 146−148.

(29) Ballesteros, J. A.; Weinstein, H. Integrated methods for the

construction of three-dimensional models and computational probing

of structure-function relations in G protein-coupled receptors. Methods

Neurosci. 1995, 25, 366−428.

(30) Friesner, R. A.; Banks, J. L.; Murphy, R. B.; Halgren, T. A.;

Klicic, J. J.; Mainz, D. T.; Repasky, M. P.; Knoll, E. H.; Shaw, D. E.;

Shelley, M.; Perry, J. K.; Francis, P.; Shenkin, P. S. Glide: a new

approach for rapid, accurate docking and scoring. 1. Method and

assessment of docking accuracy. J. Med. Chem. 2004, 47, 1739−1749.

(31) Halgren, T. A.; Murphy, R. B.; Friesner, R. A.; Beard, H. S.;

Frye, L. L.; Pollard, W. T.; Banks, J. L. Glide: a new approach for rapid,

accurate docking and scoring. 2. Enrichment factors in database

screening. J. Med. Chem. 2004, 47, 1750−1759.

(32) Negoro, N.; Sasaki, S.; Mikami, S.; Ito, M.; Suzuki, M.;

Tsujihata, Y.; Ito, R.; Harada, A.; Takeuchi, K.; Suzuki, N.; Miyazaki, J.;

Santou, T.; Odani, T.; Kanzaki, N.; Funami, M.; Tanaka, T.; Kogame,

A.; Matsunaga, S.; Yasuma, T.; Momose, Y. Discovery of TAK-875: A

Potent, Selective, and Orally Bioavailable GPR40 Agonist. ACS Med.

Chem. Lett. 2010, 1, 290−294.

(43) Jacobson, M. P.; Friesner, R. A.; Xiang, Z.; Honig, B. On the

role of crystal packing forces in determining protein sidechain

conformations. J. Mol. Biol. 2002, 320, 597−608.

(44) Bowers, K. J.; Chow, E.; Xu, H.; Dror, R. O.; Eastwood, M. P.;

Gregersen, B. A.; Klepeis, J. L.; Kolossvary, I.; Moraes, M. A.;

Sacerdoti, F. D.; Salmon, J. K.; Yibing, S.; Shaw, D. E. Scalable

algorithms for molecular dynamics simulations on commodity clusters.

Proceedings of the ACM/IEEE Conference on Supercomputing (SC06);

Tampa, Florida, November 11−17, 2006.

́ ́

(45) Tusnady, G. E.; Dosztanyi, Z.; Simon, I. Transmembrane

proteins in the Protein Data Bank: identification and classification.

Bioinformatics 2004, 20, 2964−2972.

́ ́

(46) Tusnady, G. E.; Dosztanyi, Z.; Simon, I. PDB_TM: selection

and membrane localization of transmembrane proteins in the protein

data bank. Nucleic Acids Res. 2005, 33, D275−D278.

́

(47) Kozma, D.; Simon, I.; Tusnady, G. E. PDBTM: Protein Data

Bank of transmembrane proteins after 8 years. Nucleic Acids Res. 2013,

41, D524−D529.

(48) Darden, T.; York, D.; Pedersen, L. Particle mesh Ewald: An N·

log (N) method for Ewald sums in large systems. J. Chem. Phys. 1993,

98, 10089−10092.

(49) Ehlert, F. J.; Griffin, M. T.; Sawyer, G. W.; Bailon, R. A simple

method for estimation of agonist activity at receptor subtypes:

comparison of native and cloned M3 muscarinic receptors in guinea

pig ileum and transfected cells. J. Pharmacol. Exp. Ther. 1999, 289,

981−992.

(33) Negoro, N.; Sasaki, S.; Mikami, S.; Ito, M.; Tsujihata, Y.; Ito, R.;

Suzuki, M.; Takeuchi, K.; Suzuki, N.; Miyazaki, J.; Santou, T.; Odani,

T.; Kanzaki, N.; Funami, M.; Morohashi, A.; Nonaka, M.; Matsunaga,

S.; Yasuma, T.; Momose, Y. Optimization of (2,3-dihydro-1-

benzofuran-3-yl)acetic acids: discovery of a non-free fatty acid-like,

highly bioavailable G protein-coupled receptor 40/free fatty acid

receptor 1 agonist as a glucose-dependent insulinotropic agent. J. Med.

Chem. 2012, 55, 3960−3974.

(34) Stanley, N.; Pardo, L.; Fabritiis, G. D. The pathway of ligand

entry from the membrane bilayer to a lipid G protein-coupled

receptor. Sci. Rep. 2016, 6, 22639.

(35) Hurst, D. P.; Grossfield, A.; Lynch, D. L.; Feller, S.; Romo, T.

D.; Gawrisch, K.; Pitman, M. C.; Reggio, P. H. A Lipid pathway for

ligand binding is necessary for a cannabinoid G protein-coupled

receptor. J. Biol. Chem. 2010, 285, 17954−17964.

̌

(36) Sali, A.; Blundell, T. L. Comparative protein modelling by

satisfaction of spatial restraints. J. Mol. Biol. 1993, 234, 779−815.

P

DOI: 10.1021/acs.jmedchem.7b00693

J. Med. Chem. XXXX, XXX, XXX−XXX

Article Doi

Probing the Hydrophobic Binding Pocket of G-Protein-Coupled Lysophosphatidylserine Receptor GPR34/LPS1 by Docking-Aided Structure-Activity Analysis

DOI: 10.1021/acs.jmedchem.7b00693

Source and publish data:

Authors:

Article abstract of DOI:10.1021/acs.jmedchem.7b00693

Full text of DOI:10.1021/acs.jmedchem.7b00693

Products guided by the article

R&D Labs maybe for 65295-67-2

Relevant to this article

Hot Product