2-Phenoxy-nicotinamides are Potent Agonists at the Bile Acid Receptor GPBAR1 (TGR5)

Article abstract of DOI:10.1002/cmdc.201200474

Potency with potential: 2-Phenoxy- nicotinamides were identified as potent agonists at the GPBAR1 receptor, a target in the treatment of obesity, type2 diabetes and metabolic syndrome. Extensive structure-activity relationship studies supported by homolog

Full text of DOI:10.1002/cmdc.201200474

CHEMMEDCHEM
COMMUNICATIONS
DOI: 10.1002/cmdc.201200474
2-Phenoxy-nicotinamides are Potent Agonists at the Bile
Acid Receptor GPBAR1 (TGR5)
Rainer E. Martin,*[a] Caterina Bissantz,[a] Olivier Gavelle,[a] Christoph Kuratli,[a]
Henrietta Dehmlow,[a] Hans G. F. Richter,[a] Ulrike Obst Sander,[a] Shawn D. Erickson,[b]
Kyungjin Kim,[b] Sherrie Lynn Pietranico-Cole,[b] Rubꢀn Alvarez-Sꢁnchez,[c] and
Christoph Ullmer[d]
G Protein-coupled bile acid receptor 1 (GPBAR1; synonyms:
TGR5-Takeda, M-BAR and BG37-Banyu, AXOR109-Glaxo,
GPCR19) was first reported by two Japanese groups in 2002[1]
and 2003[2] as a novel Gas G protein-coupled receptor (GPCR)
responsive to bile acids (BAs) exhibiting immunosuppressive
properties. The receptor is proposed to be mainly activated by
lithocholic acid (LCA) and deoxycholic acid (DCA) in peripheral
tissues and was recently identified as a neurosteroid receptor
in the brain,[3] where GPBAR1 is expressed in neurons and as-
trocytes. In the periphery, GPBAR1 is neuronally expressed in
the enteric nervous system,[4] but it is also abundant in tissues,
such as pancreas, intestine, muscle and brown adipose tissue,
as well as in the intestinal tract, lung, spleen, and placenta.
GPBAR1 has been identified as an important component of the
bile acid signaling network. GPBAR1 activation has been linked
to enhanced energy expenditure and regulation of intestinal
glucagon-like peptide-1 (GLP-1) secretion resulting in improved
glycaemic control.[5] Thus, GPBAR1 has recently become an at-
tractive therapeutic target for the prevention and/or treatment
of obesity, type 2 diabetes and metabolic syndrome.
flammation might serve as an attractive therapeutic tool for
immune and inflammatory liver diseases,[8] protection against
gastrointestinal injury caused by acetylsalicylic acid and non-
steroidal anti-inflammatory drugs,[9] and attenuation of athero-
sclerosis.[6] The most advanced GPBAR1-selective compound
with a potency of 0.9 mm against the human receptor is the
semi-synthetic bile acid 6a-ethyl-23(S)-methylcholic acid (S-
EMCA, INT-777), which was originally derived from the farne-
soid X receptor (FXR)-selective 6a-ethyl-chenodeoxycholic acid
(6-ECDCA, INT-747, obeticholic acid; Figure 1).[10,11] In preclinical
Expression of the receptor has also been shown in immune
cells, such as monocytes, macrophages,[2,6] and Kupffer cells,[7]
where GPBAR1 activation results in the suppression of pro-
inflammatory cytokines. GPBAR1 as a negative mediator of in-
Figure 1. Structures of semi-synthetic bile acids S-EMCA (INT-777) and 6-
ECDCA (INT-747), with their associated biological values taken from Ref. [11].
animal models of obesity-induced type-2 diabetes, administra-
tion of INT-777 induced GLP-1 secretion and insulin sensitivity,
increased basal energy expenditure (thus preventing weight
gain), and normalized glycemic control. Furthermore, INT-777
decreased lipid levels, as well as liver steatosis and fibrosis, and
thus might show potential in the treatment of diabetes and
associated disorders.[5]
[a] Dr. R. E. Martin, Dr. C. Bissantz, O. Gavelle, C. Kuratli, Dr. H. Dehmlow,
Dr. H. G. F. Richter, Dr. U. Obst Sander
Small Molecule Research, Medicinal Chemistry
Pharma Research & Early Development (pRED)
F. Hoffmann-La Roche AG
Grenzacherstrasse 124, 4070 Basel (Switzerland)
E-mail: rainer_e.martin@roche.com
[b] Dr. S. D. Erickson, Dr. K. Kim, Dr. S. L. Pietranico-Cole
Small Molecule Research, Medicinal Chemistry
Pharma Research & Early Development (pRED)
F. Hoffmann-La Roche Inc.
Our search for novel, small molecule GPBAR1 agonists was
initiated with a high-throughput screening (HTS) of the propri-
etary Roche compound collection. Nicotinamide 1 with EC50
values of 0.396 mm and 2.02 mm against the human (h) and
murine (m) receptors, respectively, was identified as an inter-
esting starting point for the hit-to-lead process as this com-
pound offers ample space for structural variations and has
high chemical tractability (Figure 2).
340 Kingsland Street, Nutley (USA)
[c] Dr. R. Alvarez-Sꢀnchez
Nonclinical Drug Safety
Pharma Research & Early Development (pRED)
F. Hoffmann-La Roche AG
Grenzacherstrasse 124, 4070 Basel (Switzerland)
The lower potency against the mouse receptor was not sur-
prising as the amino acid sequence identity between the
human and mouse receptors is only 83%.[2] Interestingly, com-
pound 1 was a single hit, as no other compound with the
same 2-phenoxy-nicotinamide tetrahydroquinoline substruc-
ture were tested in the HTS campaign. Therefore, no prelimina-
[d] Dr. C. Ullmer
DTA Vascular and Metabolic Diseases
Pharma Research & Early Development (pRED)
F. Hoffmann-La Roche AG
Grenzacherstrasse 124, 4070 Basel (Switzerland)
Supporting information for this article is available on the WWW under
http://dx.doi.org/10.1002/cmdc.201200474.
ꢂ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2013, 8, 569 – 576 569

CHEMMEDCHEM
COMMUNICATIONS
www.chemmedchem.org
with respect to lowering lipophilicity and increasing solubility,
such as dimethylamino, hydroxy, acetylamide or free carboxylic
acid moieties, were not tolerated at various positions on the
phenyl ring. However, activities in the mid-nanomolar range
were obtained in many instances by substitution of the meta
position of the phenyl ring with fluorine (3: hEC50 =0.488 mm),
chlorine (4: hEC50 =0.192 mm) or bromine (5: hEC50 =0.232 mm)
and small moieties like methyl (6: hEC50 =0.644 mm), ethyl (7:
hEC50 =0.473 mm) or isopropyl (8: hEC50 =0.625 mm) alike. Sub-
stitution of the phenyl group at the meta position relative to
the oxygen linker was clearly preferred over other locations as
exemplified by a direct comparison between the 2-chloro (9:
hEC50 =1.0 mm), 3-chloro (4: hEC50 =0.192 mm) or 4-chloro (10:
hEC50 >10 mm) derivatives. Sterically more demanding substitu-
ents such as ethoxy (11: hEC50 =2.31 mm) or phenyl (12:
hEC50 =2.75 mm) gave much less potent compounds. Further
substitution of the phenoxy group with a second halogen
turned out to be very advantageous as demonstrated by 2,5-
dichloro compound 13 displaying a potency of 10 nm and full
agonist activity (102% measured relative to LCA). However,
compound 13 still exhibited potency against the mouse recep-
tor (mEC50 =2.56 mm) in the same range as HTS hit 1 (mEC50 =
2.02 mm). Similarly, the physicochemical and metabolic proper-
ties of 13 were also not improved over 1. Compound 13 has
Figure 2. HTS hit structure 1 and its associated biological and pharmaco-
kinetic profile (in house data).
ry structure–activity relationship (SAR) could be derived from
the HTS results. The physicochemical profile of this isolated
hit was rather poor, displaying low aqueous solubility
(<1 mgmLÀ1),[12] high metabolic clearance (CL) as determined
in both human and mouse microsomal stability assays, as well
as some degree of CYP3A4 and CYP2C9 competitive inhibition.
Medicinal chemistry optimization was initiated with the goal to
identify selective and potent agonists endowed with physico-
chemical properties suitable for in vivo pharmacological proof-
of-concept studies.
a logD value of >3,[12] aqueous solubility <1 mgmLÀ1 [13]
,
At the onset of our structure–activity exploration, we investi-
gated the scope of the substitution pattern of the phenoxy
side chain. Interestingly, removal of the 3-trifluoromethyl
group in 1 leading to analogue 2 resulted in complete loss of
activity (Table 1). Similarly, replacement with heterocyclic moi-
eties like an unsubstituted 4-pyridyl ring gave inactive com-
pounds (hEC50 >10 mm; structure not shown). Furthermore, ad-
dition of polar side chains that would have been advantageous
CYP3A4, CYP2D6 and CYP2C9 liabilities (IC50 =1.1/2.6/0.4 mm)
as well as high clearance in both human (CL=
1418 mLminÀ1 kgÀ1) and mouse (CL=2134 mLminÀ1 kgÀ1) mi-
crosomes, which was further confirmed in mouse hepatocytes.
The 2,5-substitution pattern turned out to be essential as this
decoration delivered the most potent compounds as can be
seen by a direct comparison of 13 with the 2,3-dichloro (14:
hEC50 =1.15 mm), 2,4-dichloro (15: hEC50 =1.96 mm) or 3,4-di-
chloro (16: hEC50 =1.00 mm) congeners, which provid-
ed to be significantly less potent agonists. Exchange
of the 2-chloro with methyl (17: hEC50 =0.153 mm) or
the 5-chloro with trifluoromethyl (18: hEC50 =
0.286 mm) also resulted in compounds with 15- and
29-times lower potency, respectively. Surprisingly,
substitution of the two chloro substituents in 13 by
fluorine to give the 2-fluoro-5-chloro-, the 2-chloro-5-
fluoro- and the 2,5-difluoro-substituted compounds
had a very strong influence on agonist potency as
Table 1. In vitro data for analogous of compound 1 varying the substitution pattern
of the phenoxy side chain.
Compd
R
hGPBAR1
EC50[a] [mm] Eff[b] [%]
Compd
R
hGPBAR1
EC50[a] [mm] Eff[b] [%]
shown by
a comparison between 19 (hEC50 =
4.45 mm), 20 (hEC50 >10 mm) and 21 (hEC50 >10 mm).
The considerable loss of activity for the last two com-
pounds was particularly surprising as the switch of
chlorine in 3 to fluorine providing 4 resulted in
a modest potency decrease by only a factor of two.
From the many other substitution patterns explored,
only 3-chloro-4-fluoro derivative 22 (hEC50 =53 nm)
showed comparable potency to lead compound 13.
We next turned our attention to the oxygen linker
between the 2,5-dichlorophenyl and the nicotina-
mide core of 13 looking at a number of structural
variations. From the numerous linkers explored, only
compounds containing ÀNHÀ (hEC50 =1.10 mm),
1
2
3
4
5
6
7
8
3-CF3 0.396
86
–
12
13
14
15
16
17
18
19
20
21
22
3-Ph
2.75
0.010
1.15
1.96
1.00
0.153
0.286
4.45
91
102
71
83
77
100
91
97
–
H
>10
0.322
0.192
0.232
2,5-di-Cl
2,3-di-Cl
2,4-di-Cl
3,4-di-Cl
2-CH3-5-Cl
2-F-5-CF3
2-F-5-Cl
2-Cl-5-F
2,5-di-F
3-Cl-4-F
3-F
3-Cl
3-Br
103
96
91
78
85
85
98
–
3-CH3 0.644
3-Et 0.473
3-iPr 0.625
9
10
11
2-Cl
4-Cl
0.99
>10
>10
>10
0.053
–
78
3-EtO 2.31
75
[a] EC50 values are the average of determinations (n=1) performed in triplicate. [b] Ef-
ficacy (eff) measured as maximum response relative to the response by stimulation
with 10 mm lithocholic acid (LCA).
ꢂ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2013, 8, 569 – 576 570

CHEMMEDCHEM
COMMUNICATIONS
www.chemmedchem.org
ÀCH2À (hEC50 =1.73 mm) and ÀCOÀ (hEC50 =4.01 mm) main-
tained some degree of potency albeit at the micromolar level.
Other variations such as exchange of oxygen to ÀCHOHÀ or
direct fusion of the dichlorophenyl to the nicotinamide core
were not tolerated (structures not shown). Furthermore,
moving the entire tetrahydroquinoline-carbonyl side chain of
compound 13 from position 3 of the pyridine core to posi-
tion 4 (para to the pyridine nitrogen) resulted in a considerably
less potent compound. However, shifting the nitrogen of the
pyridine core meta (hEC50 =0.007 mm) or para (hEC50 =
0.037 mm) to the oxygen bridge as well as its deletion yielding
a phenyl core (hEC50 =0.029 mm) provided similarly potent
compounds. Interestingly, only the meta-shifted compound
showed a significant increase in agonist potency against the
murine receptor, with the mEC50 value going from 2.56 mm to
0.317 (structures not shown). However, all these modifications
came with no added benefit in terms of physicochemical prop-
erties or metabolic stability.
this part of the molecule with the aim to improve metabolic
stability by blocking potential sites of metabolic attack and to
lower the logD value of prospect target structures. The addi-
tion of fluorine at position 6 (23: hEC50 =0.028 mm), position 7
(24: hEC50 =0.058 mm) or difluorination at positions 6,7 (25:
hEC50 =0.057 mm) and 6,8 (26: hEC50 =0.044 mm) was well toler-
ated and yielded compounds with good potency, but did not
improve in vitro metabolic properties (Table 2). Interestingly,
exchange of the 6-fluoro substituent in 26 with a methyl
group resulted in a considerably less potent molecule (27:
hEC50 =0.837 mm). Indeed, such a drop in potency was ob-
served for many other only slightly larger substituents, such as
methoxy, trifluoromethyl or isopropyl (structures not shown).
Similarly, replacement of fluorine in 23 and 24 with chlorine
providing analogues 28 (hEC50 =0.436 mm) and 29 (hEC50 =
0.377 mm) led to a 16- and 7-fold decrease in potency, respec-
tively.
A classical N-scan, introducing nitrogen atoms at various po-
sitions of the tetrahydroquinoline aromatic moiety, provided
inactive compounds and thus was not a viable option to im-
prove physicochemical properties. Surprisingly, opening of the
saturated alkyl bridge of the tetrahydroquinoline at various po-
sitions gave access to compounds with relatively high activity,
such as 30 (hEC50 =0.092 mm), 31 (hEC50 =0.067 mm) and 32
(hEC50 =0.070 mm). SAR studies around this motif quickly
showed that ortho substituents on the aryl unit are beneficial
for activity as exemplified by methoxy derivative 33 (hEC50 =
0.038 mm), although the size of this substituent seems to be
important as larger groups such as cyclopropyl caused a drop
in potency (34: hEC50 =0.321 mm). Unfortunately, addition of
a nitrogen atom to the phenyl side chain in compound 33 pro-
viding methoxy-pyridine derivative 35 (hEC50 =0.115 mm) was
accompanied by a threefold decrease in potency.
After identifying the preferred linker between the 2,5-di-
chlorophenoxy side chain and the nicotinamide core of 13, we
next turned our attention to the tetrahydroquinoline moiety.
Microsomal metabolite identification studies revealed this
moiety as the primary site of oxidative attack in both human
and mouse microsomes. Three metabolic pathways were iden-
tified: 1) oxidation of the alkyl bridge of the tetrahydroquino-
line moiety ([M+O]), 2) oxidation of the alkyl bridge followed
by elimination ([MÀ2H]) and hydroxylation of the phenyl ring
para to the amide nitrogen ([M+O]). As a first attempt, we
thus aimed for a structural replacement of the tetrahydroqui-
noline moiety ameliorating this insufficiency. Numerous struc-
tural alternatives were tested, but all were found to be consid-
erably less potent or entirely inactive (see the Supporting In-
formation). For instance, the following modifications were
poorly tolerated: removal of the annulated phenyl group of
the tetrahydroquinoline moiety, saturation of the phenyl group
to a decahydroquinoline system, or moving the linking nitro-
gen atom to the beta position or out of the ring system. Simi-
larly, elaboration of the phenyl ring into heterocylic systems
such as a tetrahydroimidazo[1,2-a]pyrimidine or contraction of
the piperidine to a dihydroisoindole group provided inactive
compounds. Modification of the tetrahydroquinoline moiety
into an indole or benzimidazole equally provided compounds
devoid of any activity. Only 2-methyl-2,3-dihydro-1H-indole,
2,3-dihydro-1H-indole or 2,3,4,5-tetrahydro-1H-benzo[b]azepine
side chains showed some modest potency, with hEC50 values
of 1.71, 8.97 and 3.56 mm, re-
Complete removal of the ortho substituent on the phenyl
ring was tolerated in some cases as evidenced by 36 (hEC50 =
0.283 mm), although compounds devoid of any substituents at
the amide nitrogen completely lost activity (37: hEC50 >10 mm),
demonstrating that small alkyl side chains such as present in
31–36 are absolutely mandatory in order to retain (at least
some) potency. This observation might be explained by a com-
parison of the predicted preferred conformation of the nicotin-
yl-tetrahydroquinoline substructure in compound 13 derived
as bioactive (lowest energy) conformation in the homology
model and the conformations of tertiary and secondary amides
(Figure 3). In the preferred conformation of 13, the pyridyl core
spectively
(structures
not
shown). However, the SAR
around the dihydroindole core
was found to be rather tight as
even small substituents like fluo-
rine or chlorine on the periphery
of the core were not tolerated.
Based on the SAR data reveal-
ing the unique superiority of the
tetrahydroquinoline moiety, we
Figure 3. a) Docked conformation of nicotinamide 13. Preferred conformation of tertiary (b) and secondary (c)
turned again our attention to amides according to the CSD.
ꢂ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2013, 8, 569 – 576 571

CHEMMEDCHEM
COMMUNICATIONS
www.chemmedchem.org
Table 2. In vitro data for analogous of compound 13 optimizing the tetrahydro-quinoline moiety.
Compd
R
hGPBAR1
EC50[a] [mm]
Compd
R
hGPBAR1
Eff[b] [%]
108
EC50[a] [mm]
Eff[b] [%]
136
23
24
25
0.028
0.058
0.057
35
36
37
0.066
139
101
0.283
97
–
>10
26
27
0.044
0.837
73
73
38
39
0.135
0.029
106
82
28
29
30
0.436
0.203
0.092
100
93
40
41
42
>10
0.062
0.002
–
103
87
119
31
32
0.067
0.070
78
43
44
0.004
0.010
107
121
131
33
34
0.020
0.321
91
45
46
0.014
0.307
96
82
116
[a] EC50 values are the average of determinations (n=1) performed in triplicate. [b] Efficacy (eff) measured as maximum response relative to the response
by stimulation with 10 mm lithocholic acid (LCA).
is predicted to form an intramolecular p–p stacking interaction
ing the amide into a s-cis conformation. A search of the Cam-
with the phenyl group of the tetrahydroquinoline moiety, forc-
bridge Structural Database (CSD) database clearly shows that
ꢂ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2013, 8, 569 – 576 572

CHEMMEDCHEM
COMMUNICATIONS
www.chemmedchem.org
tertiary amides (N-phenyl-benzamides with a methyl group at
the amide nitrogen) generally prefer a s-cis amide conforma-
tion, as illustrated in Figure 3b. This preference for the s-cis
conformation is independent of the presence of an ortho-
methyl substituent at the phenyl ring. However, secondary
amide structures (N-phenyl-benzamide with a free hydrogen
atom at the amide nitrogen) prefer an s-trans amide conforma-
tion as exemplified in Figure 3c. This three-dimensional ar-
rangement is predicted to lead to a more elongated geometry
compared with the s-cis conformation. Thus, secondary amide
analogue 37 likely adopts a more linear shape than compound
13 and cannot be aligned onto the original tetrahydroquino-
line moiety. Thus, tertiary amide analogues such as 30–36 are
predicted to have a similar conformation than 13 and conse-
quently retain activity while the secondary amide analogue 37
favoring a distinct conformation compared with 13 results in
complete loss of binding.
cholic acid (TLCA), the endogenous ligands of GPBAR1. The
alignment was generated by docking TLCA and nicotinamide
13 into the hydrophobic binding cavity of a GPBAR1 three-
dimensional homology model (see Supporting Information).
The docking pose of the bile acid TLCA, which was used for
the docking study, was oriented such that the hydrophobic
steroid scaffold was fitted into the hydrophobic pocket formed
by transmembrane domains 3, 5 and 6. According to the pre-
dicted binding mode, the lengthy side chain of the bile acid
then extended through a narrow cleft formed by transmem-
brane domains 2, 3 and 7, positioning the negatively charged
sulfate group proximal to the polar receptor side chains at the
outer part of the binding cavity (Figure 4). The tetrahydroqui-
A switch from the tetrahydroquinoline moiety to a 3,4-di-
hydro-2H-benzo[1,4]oxazine (38: hEC50 =0.135 mm) and particu-
larly to 3,4-dihydro-2H-benzo[1,4]thiazine (39: hEC50 =
0.076 mm) provided compounds with good potency. A de-
crease in lipophilicity was observed for O-analogue 38, with
the associated ClogP value decreasing from 5.23 (13) to 4.53
(38). This change was also reflected in the measured logD
values, decreasing from >4 (13) to 3.67 (38). Compared with
13, the ClogP value for 3,4-dihydro-2H-benzo[1,4]thiazine (39)
was only slightly decreased to 5.16, but for the corresponding
thiazine-1,1-dioxide 40, the ClogP value made a considerable
jump down to 3.2. Unfortunately, compound 40 was entirely
devoid of activity at the target receptor.
The most intriguing finding was revealed when switching to
tetrahydroquinoxalines. Although compound 41 (hEC50 =
0.062 mm) initially showed a sixfold decrease in potency com-
pared with 13, substitution of the free NH with small alkyl
groups, such as methyl (42: hEC50 =0.002 mm), cyclopropyl (43:
hEC50 =0.004 mm), cyclobutyl (44: hEC50 =0.010 mm) or oxetanyl
(45: hEC50 =0.014 mm), provided potency in the low nanomolar
range. In addition, all four compounds showed significantly
higher potency towards the mouse receptor (42: mEC50 =
0.218 mm; 43: 0.231 mm; 44: 0.217 mm; 45: 0.528 mm). Unfortu-
nately, compared with compound 13, the physicochemical and
metabolic properties of 42–45 did not improve significantly.
However, the most improved physicochemical profile was ob-
tained for oxetanyl derivative 45 with a logD value of 3.27 and
an aqueous solubility of 14 mgmLÀ1, but this compound still
carried some CYP3A4, CYP2D6 and CYP2C9 (IC50 =0.3/10.0/
8.2 mm) liabilities, as well as high clearance in human micro-
somes (CL=578 mLminÀ1 kgÀ1).
Figure 4. Taurolithocholic acid (cyan) and nicotinamide 13 (magenta)
docked into the homology model of the human hGPBAR1 receptor. Carbon
atoms of the protein are shown in green.
noline structure (13) that was used for the docking exercise
was placed into the hydrophobic cavity and thus overlaid with
the A and B rings of the bile acid scaffold, while the pyridine
core overlaid partially with the C ring, and the phenoxy sub-
stituent points into the direction of the taurine conjugate
chain.
According to this alignment and docking hypothesis, the
best position to incorporate polar side chains should be at the
2,5-dichlorophenyl group para to the oxygen linker. Substitu-
ents at this position are predicted to overlap with the conju-
gated side chain of the bile acid and reach through the narrow
cleft to the region containing polar amino acid side chains. In
order to extend to the same position of the TLCA sulfate
group, a linker length of four to five atoms should be optimal.
However, shorter chains might also be accepted (one- to two-
atom linker) since the unconjugated bile acid LCA is itself an
endogenous agonist. In this case, the free carboxylic acid
group is proposed to be located close to a serine at position 7.
As our strategy to decrease the high metabolic clearance of
the tetrahydroquinoline moiety by introducing nitrogen and
fluorine atoms did not lead to compounds with a significantly
improved profile, we next focused on the identification of posi-
tions that would allow attachment of polar and (under physio-
logical conditions) charged side chains such as carboxylic
acids. In order to guide our carboxy screen to identify the opti-
mal connection point, we used an alignment of our nicotina-
mide structures with bile acids such as DCA, LCA or taurolitho-
ꢂ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2013, 8, 569 – 576 573

CHEMMEDCHEM
COMMUNICATIONS
www.chemmedchem.org
On the other hand, it cannot be ruled out that longer linkers
might be tolerated since the polar region is located close to
the receptor surface and could thus reach into the extracellular
space.
considerably lower logD values compared with neutral com-
pounds from this series, but they all displayed medium stabili-
ty in human microsomes. The incorporation of five-membered
heterocycles bearing acidic functionalities like 56 or 57 was
also very well tolerated, with thiazole 56 showing nearly equal
potency on the human and mouse receptor. Furthermore,
compound 56 displayed a balanced logD value of 1.86 as well
as an improved CYP3A4, CYP2D6 and CYP2C9 inhibitory profile
(IC50 =11/ >50/11 mm). Interestingly, removal of the methyl
group on the thiazole core in 56 resulted in a considerable
loss of potency (hEC50 =1.43 mm), probably due to a change of
pre-organization of the carboxylic acid. Moving away from car-
boxylic acids to tetrazole led to a slight loss of potency on the
human and mouse receptor, but provided with 57, a com-
pound with medium stability in both human and mouse mi-
crosomes. These optimized structures were also tested in
a functional human GPBAR1 assay using NCI-H716 cells with
endogenous GPBAR1 expression and a hGPBAR1 binding
assay. Whereas in the functional assay, both compounds 56
and 57 showed nearly equal potency (EC50 =0.62 and 0.64 mm),
a slightly larger shift was observed in the case of the binding
assay (Ki =0.016 and 0.098 mm).
In order to probe the modeling hypothesis, we first attached
carboxylic acids along the periphery of the aromatic unit of
tetrahydroquinoline moiety in 13. Not surprisingly and fitting
to our alignment hypothesis, all these analogues failed and
provided inactive compounds. However, modification of the
free nitrogen of the tetrahydroquinoxaline moiety in 41 with
a carboxymethyl group yielded zwitterionic compound 46 that
retained some potency against the human receptor (hEC50 =
0.307 mm), an observation that can be rationalized by the
ligand adopting a 1808 rotated binding mode. More interest-
ingly, compound 46 has a considerably improved profile, with
a
ClogP value of 4.36 (logD=À0.18),
a solubility of
>540 mgmLÀ1 and medium clearance in human microsomes
(CL=23 mLminÀ1 kgÀ1). In addition, the compound displayed
high passive permeability (Pe) in a parallel artificial membrane
permeation assay (PAMPA) (Pe=2.56ꢃ10À6 cmsÀ1 [14]
and no
)
CYP3A4, CYP2D6 and CYP2C9 inhibition (IC50 =14/>50/>
50 mm), as well as hERG issue (IC20 >10 mm). Notably, elonga-
tion of the methylene chain to a hexanoic acid on a structure
containing an additional fluorine at the 5-position of the pyri-
dine core yielded an even more potent compound (hEC50 =
0.104 mm, structure not shown), but again with high microso-
mal clearance. Evidently, in this case, the carboxy group was
too remote to protect the tetrahydroquinoxaline core efficient-
ly from microsomal attack.
The synthesis of nicotinamide GPBAR1 agonists is straight-
forward and summarized in Scheme 1. Commercially available
2-chloro nicotinic acid (58) was coupled with dichlorophenol
59 under Ullmann-type conditions to give intermediate 60,
which provided after amide-coupling with cyclopropyl-substi-
tuted quinoxaline 61, the pyridine building block 62. Methoxy-
carbonylation of 62 gave facile access to methyl ester 63. Ester
We then incorporated polar side chains para to the
oxygen linker of the 2,5-dichlorophenyl group as sug-
gested by the modeling hypothesis. Indeed, addition
of a propargyl alcohol (47: hEC50 =0.036 mm), hydrox-
yethylcarboxamide (48: hEC50 =0.068 mm), propropyl-
carboxylic acid (49: hEC50 =0.047 mm) or para-ethyle-
nebenzoic acid (50: hEC50 =0.069 mm) side chain de-
livered all very potent compounds with improved
physicochemical properties (e.g., solubility, logD), but
unfortunately did not resolve the issue of high micro-
somal clearance (Table 3). Interestingly, polar side
chains directly attached to the phenyl core like a car-
boxylic acid functionality (hEC50 =7.047 mm, structure
not shown) were not tolerated. However, molecules
bearing a carboxylic acid group linked via an alkylcar-
boxamide spacer, and thus being further remote
from the phenyl core, such as propionylamino acetic
acids 51 (hEC50 =0.063 mm) and 52 (hEC50 =
0.096 mm), butyrylamino propionic acid 53 (hEC50 =
0.041 mm) or its homologue 54 (hEC50 =0.091 mm)
provided very active agonists and thus are fully in
line with our alignment hypothesis. Amazingly, even
Scheme 1. Synthesis of nicotinamides 56 and 57. Abbreviations: O-(7-azabenzotriazol-1-
yl)-N,N,N’,N’-tetramethyluronium hexafluorophosphate (HATU), N,N-diisopropylethylamine
proline derivative 55 (hEC50 =0.071 mm) with its sig-
nificantly rigidified geometry compared to 51 gave (DIEA), (diphenylphosphino)ferrocene (dppf). Reagents and conditions: a) [Cu(CH3CN)4]PF6,
Cs2CO3, toluene, microwave, 1408C, 2 h, 73%; b) HATU, DIEA, CH2Cl2, RT, 12 h, 63%; c) CO
(70 bar), Pd(dppf)Cl2, MeOH/EtOAc (1:1), 1008C, 20 h, 52%; d) 1. LiOH·H2O, THF/MeOH/
H2O (2:2:1), RT, 18 h, 92%; For 56: 2. 2-amino-4-methyl-thiazole-5-carboxylic acid ethyl
an equally potent compound. Obviously, there is
some flexibility in the channel geometry allowing for
some side chain flexibility. Compounds 51–55 not
ester, HATU, Et3N, CH2Cl2, RT, 4 h; 3. LiOH·H2O, THF/MeOH/H2O (2:2:1), RT, 17 h, 37%
only showed excellent high aqueous solubility and (steps 2–3); For 57: 2. 5-amino-1H-tetrazole, HATU, Et3N, CH2Cl2, RT, 4 h, 44%.
ꢂ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2013, 8, 569 – 576 574

CHEMMEDCHEM
COMMUNICATIONS
www.chemmedchem.org
Table 3. In vitro data for analogous of compound 43 looking for the most improved polar side chain.
Compd
R
hGPBAR1/mGPBAR1
Aq solubility
Permeability
(PAMPA[14]
[10À6 cmsÀ1
ClogP
logD[12]
Microsomal CL
(Lysa[13]
)
)
[mLminÀ1 mgÀ1 protein]
EC50[a] [mm]
Eff[b] [%]
134/168
[mgmLÀ1
]
]
human
265
mouse
757
47
48
0.036/1.048
–
–
4.28
3.07
–
0.068/0.808
0.047/0.436
0.069/0.078
120/165
89/170
110
234
<1
3.23
–
2.82
569
65
2235
481
49
50
5.23
7.65
1.89
3.23
123/169
–
105
495
51
52
53
54
55
0.063/0.207
0.096/0.140
0.041/0.033
0.091/0.163
0.071/0.100
131/277
124/272
119/171
108/139
109/139
565
570
620
355
590
0.33
0.44
0.63
1.23
0.5
4.51
4.99
4.44
4.82
5.05
0.35
À0.14
0.54
46
52
35
22
46
290
318
242
280
215
1.44
0.49
56
57
0.026/0.032
0.076/0.350
114/200
89/215
80
1.18
0.95
4.73
2.81
1.86
0.33
48
56
226
56
535
[a] EC50 values are the average of determinations (n=1) performed in triplicate. [b] Efficacy (eff) measured as maximum response relative to the response
by stimulation with 10 mm lithocholic acid (LCA).
hydrolysis of 63 under basic conditions followed sequentially
by O-(7-azabenzotriazol-1-yl)-N,N,N’,N’-tetramethyluronium hex-
afluorophosphate (HATU)-mediated amine coupling with 2-
amino-4-methyl-thiazole-5-carboxylic acid ethyl ester and
cleavage of the ethyl ester on the pendent thiazole yielded
final target nicotinamide 56. In a similar fashion, tetrazol-nico-
tinamide 57 was produced.
In summary, we have presented extensive SAR data on
a novel nicotinamide series and identified potent GPBAR1 re-
ceptor agonists. While being highly potent on the human re-
ceptor, neutral compounds were flawed with high microsomal
clearance as determined in both human and mouse microso-
mal stability assays. Alignment of TLCA with our advanced lead
compounds enabled the identification of the optimal attach-
ment point for a polar vector. Zwitterionic tetrazole 57 com-
bines high potency against both the human and mouse recep-
tor, favorable physicochemical properties, and acceptable mi-
crosomal clearance. The good potency on the mouse receptor
eliminates the need for the generation of humanized knock-in
mice strains and thus 57 might serve as a useful tool com-
pound for further assessment of the biological role both in
vitro and in vivo of the GPBAR1 receptor.
ꢂ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2013, 8, 569 – 576 575

CHEMMEDCHEM
COMMUNICATIONS
www.chemmedchem.org
31.2, 40.2, 40.9, 48.7, 113.3, 115.4, 120.5, 121.7, 123.2, 124.7, 126.0,
126.7, 127.1, 130.0, 132.4, 134.3, 140.2, 140.9, 148.7, 150.0, 152.5,
156.2, 163.5, 164.2 ppm; HRMS (ESI+): m/z [M+H]+ calcd for
C25H21Cl2N8O3: 551.1104, found: 551.1097.
Experimental Section
2-{2,5-Dichloro-4-[3-(4-cyclopropyl-3,4-dihydro-2H-quinoxaline-
1-carbonyl)-pyridin-2-yloxy]-benzoylamino}-4-methyl-thiazole-5-
carboxylic acid (56): A solution of methyl ester 63 (56 mg,
0.11 mmol) in THF/MeOH/H2O (2:2:1, 2 mL) was treated with
LiOH·H2O (9.4 mg, 0.23 mmol, 2 equiv), and the reaction mixture
stirred at RT for 17 h. The solvent was evaporated in vacuo, a solu-
tion of 1m HCl (2 mL) added, and the crude reaction mixture was
extracted with EtOAc (3ꢃ5 mL). The combined organic phases
were collected, dried over MgSO4, filtered and concentrated in
vacuo to yield the corresponding free carboxylic acid (50 mg,
0.103 mmol, 92%). The crude intermediate was redissolved in
CH2Cl2 (2 mL), Et3N (29 mL, 0.21 mmol, 2 equiv) and HATU (50.9 mg,
0.134 mmol, 1.3 equiv) were added, and the reaction was stirred
for 15 min at RT. Then, 2-amino-4-methyl-thiazole-5-carboxylic acid
ethyl ester (23.0 mg, 0.12 mmol, 1.2 equiv) was added, and stirring
was continued for 4 h at RT. The reaction was diluted with CH2Cl2
(5 mL), and the organic phase was washed with saturated aq
NaHCO3 (2ꢃ5 mL). The organic phase was dried over MgSO4, fil-
tered and concentrated, and the residue was taken up in THF/
MeOH/H2O (2:2:1, 2 mL). LiOH·H2O (8.4 mg, 0.20 mmol, 2 equiv)
was added, and the reaction mixture stirred at RT for 17 h. The
crude reaction mixture was concentrated in vacuo and purified by
prep HPLC to give 56 as an off-white solid (37%): 1H NMR
(600 MHz, [D6]DMSO): d=0.69 (br s, 4H), 2.24 (br s, 1H), 2.56 (s,
3H), 3.12 (br s, 1H), 3.44 (br s, 2H), 3.70 (br s, 1H), 4.74 (br s, 1H),
5.74 (s, 1H), 6.35 (s, 1H), 6.47 (m, 2H), 6.94 (m, 1H), 7.01 (m, 1H),
7.36 (m, 1H), 7.94 (s, 1H), 8.16 (s, 1H), 8.22 (m, 1H), 13.04 ppm (br
s, 1H); 13C NMR (150 MHz, [D6]DMSO): d=6.9 (2C), 16.2, 30.2, 39.5,
47.6, 112.2, 114.3, 115.5, 119.9, 120.5, 122.8, 123.0, 123.5, 124.1,
125.6, 128.8, 130.2, 130.3, 139.4, 140.5, 147.8, 149.6, 154.7, 155.2,
158.2, 162.9 (2C), 163.1 ppm; HRMS (ESI+): m/z [M+H]+ calcd for
C29H25Cl2N5O5S: 624.0871, found: 624.0865.
Acknowledgements
We are grateful to Martin Ritter, Anneliese Stoller, Michel Alter-
matt, Philippe Huez, Aurꢁlien Merot and Anja Osterwald for ex-
cellent technical support, Isabelle Parrilla for aqueous solubility
(Lysa) and Bjçrn Wagner for logD measurements.
Keywords: bile acid receptors · GPBAR1 · metabolic disorders ·
nicotinamides · structure–activity relationships
[1] T. Maruyama, Y. Miyamoto, T. Nakamura, Y. Tamai, H. Okada, E. Sugiya-
ma, T. Nakamura, H. Itadani, K. Tanaka, Biochem. Biophys. Res. Commun.
2002, 298, 714–719.
[2] Y. Kawamata, R. Fujii, M. Hosoya, M. Harada, H. Yoshida, M. Miwa, S. Fu-
kusumi, Y. Habata, T. Itoh, Y. Shintani, S. Hinuma, Y. Fujisawa, M. Fujino,
J. Biol. Chem. 2003, 278, 9435–9440.
[3] V. Keitel, B. Gçrg, H. J. Bidmon, I. Zemtsova, L. Spomer, K. Zilles, D. Hꢄus-
singer, GLIA 2010, 58, 1794–1805.
[4] D. P. Poole, C. Godfrey, F. Cattaruzza, G. S. Cottrell, J. G. Kirkland, J. C.
Pelayo, N. W. Bunnett, C. U. Corvera, Neurogastroenterol. Motil. 2010, 22,
814–e828.
[5] C. Thomas, A. Gioiello, L. Noriega, A. Strehle, J. Oury, G. Rizzo, A. Mac-
chiarulo, H. Yamamoto, C. Mataki, M. Pruzanski, R. Pellicciari, J. Auwerx,
K. Schoonjans, Cell Metab. 2009, 10, 167–177.
[6] T. W. H. Pols, M. Nomura, T. Harach, G. Lo Sasso, M. H. Oosterveer, C.
Thomas, G. Rizzo, A. Gioiello, L. Adorini, R. Pellicciari, J. Auwerx, K.
Schoonjans, Cell Metab. 2011, 14, 747–757.
[7] V. Keitel, M. Donner, S. Winandy, R. Kubitz, D. Hꢄussinger, Biochem. Bio-
phys. Res. Commun. 2008, 372, 78–84.
[8] Y. D. Wang, W. D. Chen, D. Yu, B. M. Forman, W. Huang, Hepatology
2011, 54, 1421–1432.
[9] S. Cipriani, A. Mencarelli, A. Bruno, B. Renga, E. Distrutti, L. Santucci, F.
Baldelli, S. Fiorucci, Br. J. Pharmacol. 2012, accepted.
[10] R. Pellicciari, A. Gioiello, A. Macchiarulo, C. Thomas, E. Rosatelli, B. Natali-
ni, R. Sardella, M. Pruzanski, A. Roda, E. Pastorini, K. Schoonjans, J.
Auwerx, J. Med. Chem. 2009, 52, 7958–7961.
2,5-Dichloro-4-[3-(4-cyclopropyl-3,4-dihydro-2H-quinoxaline-1-
carbonyl)-pyridin-2-yloxy]-N-(1H-tetrazol-5-yl)-benzamide (57): A
solution of methyl ester 63 (45 mg, 0.090 mmol) in THF/MeOH/H2O
(2:2:1, 2 mL) was treated with LiOH·H2O (7.6 mg, 0.18 mmol,
2 equiv), and the reaction mixture was stirred at RT for 17 h. The
solvent was evaporated in vacuo, a solution of 1m HCl (2 mL)
added, and the crude reaction mixture was extracted with EtOAc
(3ꢃ5 mL). The combined organic phases were collected, dried over
MgSO4, filtered and concentrated in vacuo to yield the correspond-
ing free carboxylic acid (40 mg, 0.083 mmol, 92%). The crude inter-
mediate was redissolved in CH2Cl2 (2 mL), Et3N (23 mL, 0.17 mmol,
2 equiv) and HATU (40.5 mg, 0.107 mmol, 1.3 equiv) were added,
and the reaction was stirred for 15 min at RT. Then, 5-amino-1H-tet-
razole (8.5 mg, 0.10 mmol, 1.2 equiv) was added, and stirring was
continued for 4 h at RT. The crude reaction mixture was concen-
trated in vacuo and purified by prep HPLC to give 57 as a light
yellow solid (44%): 1H NMR (600 MHz, CDCl3): d=0.70 (br s, 4H),
2.27 (s, 1H), 3.19 (br s, 1H), 3.51 (br s, 1H), 3.59 (br s, 1H), 4.93 (br
s, 1H), 6.00 (s, 1H), 6.38–6.43 (m, 2H), 6.97–6.98 (m, 1H), 7.02–7.05
(m, 1H), 7.23–7.25 (m, 1H), 7.99 (s, 1H), 8.13–8.15 (m, 2H), 11.18
(br s, 1H), 13.50 ppm (br s, 1H); 13C NMR (150 MHz, CDCl3): d=7.9,
[11] A. Baghdasaryan, T. Claudel, J. Gumhold, D. Silbert, L. Adorini, A. Roda,
S. Vecchiotti, F. J. Gonzalez, K. Schoonjans, M. Strazzabosco, P. Fickert, M.
Trauner, Hepatology 2011, 54, 1303–1312.
[12] logD values were measured spectrophotometrically at pH 7.4 in a 1-oc-
tanol/50 mm TAPSO buffer system containing 5% (v/v) DMSO.
[13] Aqueous solubility was measured by lyophilization solubility assay
(Lysa). Solubility was measured from lyophilized DMSO stock solutions
spectrophotometrically at pH 6.5 in a 50 mm phosphate buffer.
[14] Parallel artificial membrane permeation assay (PAMPA): low: Pe<0.1,
medium: 0.1<Pe <1.0, high: Pe>1.0. See: M. Kansy, F. Senner, K. Gu-
bernator, J. Med. Chem. 1998, 41, 1007–1010.
Received: October 16, 2012
Published online on December 7, 2012
ꢂ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2013, 8, 569 – 576 576