Discovery of novel and orally active FXR agonists for the potential treatment of dyslipidemia & diabetes

Article abstract of DOI:10.1016/j.bmcl.2010.11.039

Herein we describe the synthesis and structure activity relationship of a new class of FXR agonists identified from a high-throughput screening campaign. Further optimization of the original hits led to molecules that were highly active in an LDL-receptor KO model for dyslipidemia. The most promising candidate is discussed in more detail.

Full text of DOI:10.1016/j.bmcl.2010.11.039

Bioorganic & Medicinal Chemistry Letters 21 (2011) 191–194
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry Letters
journal homepage: www.elsevier.com/locate/bmcl
Discovery of novel and orally active FXR agonists for the potential treatment
of dyslipidemia & diabetes
Hans G. F. Richter a, Gregory M. Benson a, Denise Blum a, Evelyne Chaput a, Song Feng b, Christophe Gardes a,
Uwe Grether a, Peter Hartman a, Bernd Kuhn a, Rainer E. Martin a, Jean-Marc Plancher a, Markus G. Rudolph a,
Franz Schuler a, Sven Taylor a, Konrad H. Bleicher a,
⇑
a F. Hoffmann-La Roche AG, Grenzacherstrasse, CH-4070 Basel, Switzerland
b Roche R&D Center (China) Ltd, 720 Cai Lun Road, Pudong, Shanghai 201203, China
a r t i c l e i n f o
a b s t r a c t
Article history:
Herein we describe the synthesis and structure activity relationship of a new class of FXR agonists
identified from a high-throughput screening campaign. Further optimization of the original hits led to
molecules that were highly active in an LDL-receptor KO model for dyslipidemia. The most promising
candidate is discussed in more detail.
Received 14 October 2010
Revised 4 November 2010
Accepted 4 November 2010
Available online 12 November 2010
Ó 2010 Elsevier Ltd. All rights reserved.
Keywords:
Farnesoid X receptor
Nuclear hormone receptors
Ugi-reaction
Benzimidazoles
OH
The farnesoid X receptor (FXR) belongs to the superfamily of
nuclear hormone receptors which are gene regulating transcription
factors involved in several diverse physiological functions.1 FXR is
mainly expressed in the liver, intestine, adrenal gland and kidney,
where, upon activation by certain bile acids (BA), it regulates,
among others, the expression of various genes involved in choles-
terol and BA synthesis, metabolism and enterohepatic circulation.
Hence, BA present in the enterohepatic circulation maintains the
size and composition of the BA pool partially by feedback regula-
tion via FXR. Activation of FXR by non-BA agonists leads to a re-
duced BA pool size and a change in its composition which again
results in reduced cholesterol absorption from the small intestine
and reduced plasma cholesterol levels. FXR agonists also regulate
genes involved in triglyceride synthesis and metabolism, leading
to a reduction in plasma triglyceride.2 More recently it has been
shown that activation of FXR in the liver plays a key role in gluco-
neogenesis, glycogen synthesis and insulin sensitivity making this
protein a very promising target for treating metabolic and vascular
diseases. Further pharmacological effects mediated through FXR
activation have been reported such as improved liver regeneration
and protection against hepatocarcino-genesis.3 Hence, FXR
agonists are of great interest and, indeed, several small molecule
agonists have previously been described in the literature. FXR
agonists have been proposed for the treatment of dyslipidemia,
OH
O
O
O
N
Cl
Cl
O
N
N
O
Cl
GW-4064
Fexaramine
O
O
O
F
OH
N
F
N
H
O
HO
OH
R
FXR-450, R=CH3
Reference, R=H
6α-ECDCA
⇑
Corresponding author.
E-mail address: konrad.bleicher@roche.com (K.H. Bleicher).
Scheme 1. Representative FXR agonists with reported in vivo activity.
0960-894X/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2010.11.039

192
H. G. F. Richter et al. / Bioorg. Med. Chem. Lett. 21 (2011) 191–194
cholestasis, non-alcoholic steatohepatitis (NASH) as well as type-2
diabetes. The five best characterized molecules so far are shown in
Scheme 1.4–7
To identify valuable starting points for a chemistry program we
initiated a high-throughput screening campaign of our corporate
compound library. Several interesting hit clusters were identified
with the benzimidazolyl acetamides 1 showing the most promising
overall profile (Scheme 2).
which usually led to complete conversion of intermediate 6 to
product 1. All final compounds were purified via preparative HPLC.
Using the synthesis protocol described above we were able to
quickly generate a preliminary SAR and investigate the physico-
chemical properties of the products. All compounds described in
Tables 1–4 were tested as racemic mixtures. As a control we used
GW4064 which showed an IC50 = 0.11 lM in our assay.
From the compound set generated it became evident that the li-
gand binding site tolerates a variety of modifications in position R
as long as the residues are lipophilic. A wide range of residues were
investigated with 4-chlorophenyl being the most potent both in
terms of binding affinity as well as functional activity. Polar groups
were not tolerated at this position as exemplified in Table 1 where
four analogs are presented with R0 and R00 both being cyclohexyl.
With 4-chlorophenyl as a very potent residue for R we further
investigated the SAR at position R0. Using the synthesis protocol
described above we generated a follow-up library varying the
aldehyde part (R0). Four representatives are shown in Table 2
(R00 = cyclohexyl).
As already observed for R the introduction of polarity at R’ was
quite challenging. Lipophilic groups such as alkyl-, cycloalkyl- and
aryl- were tolerated (data not shown) but even minor changes such
as 1g led to a loss of binding affinity by ꢀtwo log units (compared
to 1a). Interestingly, by shifting the oxygen in 1g from the 4 to the
2 position of the tetrahydropyran ring, low nanomolar binding
The compounds originate from an earlier in-house combinato-
rial chemistry effort where library design was based on availability
of the building blocks rather than on drug-likeness of the final
products. While R varied extensively, the diversity of R0 and R00
was very limited. No modifications were made at the benzene part
(Ar) of the benzimidazole core. This compound class caught our
attention because benzimidazoles are a well known substructural
element in nature and thus appear quite frequently in pharmaceuti-
cal products, for example, astemizole, omeprazole and candesartan.
Although compounds with the same core structure are well
described in the literature, this particular substitution pattern led
to rather novel chemical structures allowing us to investigate the
structure activity relationship of the compound class more broadly.
A set of approximately one hundred analogs was retested for hit con-
firmation in a scintillation proximity assay (SPA) using human FXR
(hFXR) ligand binding domain with some compounds even display-
ing nanomolar binding affinities. The general synthesis route, based
on the four component Ugi reaction, is depicted in Scheme 3.
Compounds of type 1 can be generated easily by treating
BOC-protected phenylenediamine with carboxylic acids, aldehydes
and isonitriles. The resulting intermediates 6 are further treated
with neat TFA after evaporation of the organic solvent.8 We slightly
modified the procedure described by Hulmes9 using equimolar
amounts of components 2–5 to avoid the use of polymer supported
reagents. In some cases the cyclization to the imidazole was not
completed even after 12 h. Here, the TFA was evaporated and the
residue taken up in acetic acid and maintained at 80 °C overnight
Table 1
SPA-binding affinities of 1a–d with both R0 and R00 being cyclohexyl. Binding affinity
was assessed in a radioligand displacement assay as described.10 Values are means of
at least two experiments
Compound
R
c log P
IC50 (lM)
1a
1b
1c
1d
4-Cl-Ph
7.78
6.50
5.52
5.65
0.07
1.25
12.44
30.44
4-CN-Ph
4-MeSO2Ph
4-Pyridyl
N
Table 2
SPA-binding affinities of 1 with R being 4-chlorophenyl and R00 being cyclohexyl.
Values are means of at least two experiments
R: Broad range of lipophilic residues
R': alkyl-, cycloalkyl-, aryl
R
N
Compound
R0
c log P
IC50 (lM)
R'
O
1f
2-Methylbutyl
7.11
5.38
4.40
5.47
0.91
3.8
14.78
25.12
NH
1g
1h
1i
4-Tetrahydropyranyl
4-Acetylpiperidinyl
4-Pyridyl
R''
R'': isopropyl-, butyl-, cycloalkyl-
1
Scheme 2. Representative hit cluster identified from HTS.
Table 3
SPA-binding affinities of 1 with R being 4-chlorophenyl and R0 being cyclohexyl.
Values are means of at least two experiments
C
NHBoc
O
HO
O
N+
R''
+
+
R00
c log P
IC50 (lM)
+
Compound
R'
R
NH2
1k
1l
1m
1n
Ph
7.60
7.16
6.10
5.38
0.15
2.72
3.61
3.71
4,4-Difluorocyclohexyl
Cyclopropyl
4-Tetrahydropranyl
3
2
5
4
N
N
NHBoc
O
R
Table 4
N
R
SPA-binding affinities of 1 with modification at position 6 of the benzimidazole core
structure. R being 4-chlorophenyl and R0 & R00 being cyclohexyl. Values are means of at
least two experiments
R'
O
O
R'
NH
NH
R''
R''
Compound
Ar
c log P
IC50 (lM)
6
1
1p
1q
1r
1s
6-Methoxy–
6-Dimethylamino–
6-Carboxy–
8.02
8.34
7.79
6.74
0.45
5.17
5.72
6.19
Scheme 3. General synthesis route for benzoimidazolyl acetamides 1. Reagents and
conditions: (i) equimolar amounts of reagents in MeOH, rt, 1 h; (ii) neat TFA, rt,
16 h.
6-Hydroxymethyl–

H. G. F. Richter et al. / Bioorg. Med. Chem. Lett. 21 (2011) 191–194
193
affinity was regained. Obviously, burying the electronegative
oxygen within the molecule reduces the polar character of the
ligand. Unfortunately, this derivative did not show any improve-
ment in terms of physicochemical properties or microsomal stabil-
ity but rather increased the complexity of the molecule due to the
introduction of a second chiral center. In analogy to the above, R00
was investigated keeping R (4-chlorophenyl) and R’ (cyclohexyl)
constant. Again four representatives are depicted in Table 3.
Exploration of substitutions in this position showed that only a
phenyl group could be introduced as a substitute for the cyclohexyl
ring at position R00. All other, even minor changes led to a substan-
tial loss of binding affinity. We also investigated the aromatic part
of the benzimidazole core structure (Ar). Using proprietary N-Boc
protected phenylenediamines, originating from an earlier lead
optimization program. A library of over 100 analogs was generated
using state-of-the-art parallel synthesis and purification equip-
ment. From that series it became evident, that significant modifica-
tions were only tolerated at position 6 of the benzimidazole core
structure. Position 4, 5 and 7 were shown to be fairly conservative.
Four representatives with 6-modifications are shown in Table 4
with R being 4-chlorophenyl and R0 and R00 being cyclohexyl.
Here also, the functionalization of the aromatic moiety with po-
lar groups was very limited whereas the introduction of further
lipophilic residues (e.g., F or Cl) was well tolerated and even bene-
ficial in terms of binding and functional activity.
The crystal structure of compound 1a bound to the hFXR recep-
tor could be solved with a resolution of 2.3 Å (Fig. 1). The structure
nicely shows that the benzimidazole scaffold interacts through a
strong hydrogen bond (d = 2.6 Å) with Tyr373 and is well suited
to orient its substituents R, R0 and R00 into three binding pockets
in an almost perpendicular fashion. As observed in previous FXR
X-ray structures, the binding site is highly lipophilic, hence, pro-
viding a rational for the observed loss of affinity when introducing
polar groups. A second, weak polar interaction is observed for the
ligand amide NH which is engaged in a rather long hydrogen bond
with Ser336 (d = 3.6 Å). Fig. 1 shows the more active enantiomer of
the racemic mixture of 1a, which turned out to be S configured.
Various analogs of 1a bound to hFXR were successfully crystallized
and characterized showing that the hydrogen bonding of Tyr373 to
the N3 of the benzimidazole core is highly conserved making this a
key interaction for efficient ligand binding. This hypothesis was
also confirmed by generating the corresponding indole derivative
(N3 substituted by a C) which showed only very weak binding
affinity in the high micromolar range and completely lacking func-
tional activity (data not shown).
Using preparative chiral HPLC, we were able to separate most of
the racemic material into their pure enantiomers. The R enantio-
mers were usually at least 2–4 log units less active than their cor-
responding S analogs. For example, compound 7 (depicted in
Scheme 4) shows very high binding affinity and strong functional
potency in the cell based transactivation (TA) assay in contrast to
the corresponding R enantiomer which was inactive both in bind-
ing and function (data not shown). All compounds of this series
were partial agonists. Further investigations concerning the phys-
icochemical properties and the in vitro DMPK profile of this com-
pound were undertaken and are discussed in more detail below.
Not surprisingly the log D of 7 is very high resulting in negligible
aqueous solubility. Nevertheless, in simulated fasted and fed state
intestinal fluid (Fassif and Fessif, respectively) the solubility was
medium to high. The permeation data from the parallel artificial
membrane permeability assay (PAMPA) were low, but the activity
observed in the functional assay showed that this compound and
analogs thereof could easily penetrate cells. Concerning metabolic
stability of 7 the in vitro results were, unfortunately, not consistent.
Clearance by both, human and mouse microsomes were very high
whereas hepatic clearance was moderate to low. Cytochrome P450
interactions for isoforms 3A4, 2D6 and 2C9 were investigated as well
showing only weak affinities for all three enzymes. The compound
selectivity of 7 for FXR was evaluated using a reporter-gene tran-
scriptional assay versus a panel of other nuclear receptors including
PPAR-
a
, -b, -d, LXR-
a
, -b and RXR-a. No cross-reactivity was
observed at a concentration of 10
l
M (data not shown).
Based on the promising data obtained so far for this compound
class we investigated the in vivo effects of 7 on lipid lowering in
LDL-receptor knockout mice. The animals were fed a high fat diet
for 15 days prior to compound administration for an additional five
days. The reduction of total cholesterol (TC), low density lipopro-
tein (LDL) and triglycerides (TG) was measured. The ethyl ester
version of FXR-450 (see Scheme 1) was used as a reference. Both
compounds were orally dosed at 30 mg/kg. The efficacy data ob-
tained are shown in Table 5. PK/PD monitoring revealed a plasma
exposure for 7 of 0.47
lg/ml and 4.3-fold increased expression of
hepatic SHP (FXR regulated gene) 2 h after administration.
F
IC50 (SPA): 0.013 μM
EC50 (TA): 0.141 μM (43%)
logD: >4
Solubility: <1 μg/ml
N
N
Cl
Cl
H
Fassif/Fessif: 36/113 μg/ml
PAMPA (Pe): <0.2 x 10 -6cm/s
hMic(CL cat.): 17% (high)
mMic(CL cat): 4% (high)
hFH(CL cat): 100% (low)
mFH(CL cat): 66% (medium)
CYP3A4/2D6/2C9: 9.8/14.0/8.3 μM
O
NH
7
Scheme 4. In vitro profile of lead compound 7. TA: transactivation activity assay.
TA activity was measured by a luciferase transcriptional reporter gene assay as
described.11 Efficacies are relative to GW4064 (EC50 = 0.91
value is a mean. CL cat: clearance category.
lM set at 100%), the
Table 5
In vivo lipid lowering effects of 7 versus reference (30 mg/kg, p.o.) as determined in
high-fat diet fed LDLꢁ/ꢁ mice. (six animals per group, ANOVA followed by Dunnett’s
post-hoc tests: *p <0.05, **p <0.01)
Compound
TC
LDL
TG
Figure 1. X-ray crystal structure of 1a bound to hFXR (PDB id: 3OKI). Hydrogen
bonds between the ligand and the side chains of Tyr373 and Ser336, respectively,
are shown as red, dashed lines.
Reference
7
ꢁ37 5%**
ꢁ45 3%**
ꢁ35 5%*
ꢁ48 3%**
ꢁ54 7%
ꢁ52 9%*

194
H. G. F. Richter et al. / Bioorg. Med. Chem. Lett. 21 (2011) 191–194
4. Maloney, P. R.; Parks, D. J.; Haffner, C. D.; Fivush, H. M.; Chandra, G.; Plunket, K.
Further analysis of the available structural information revealed
D.; Creech, K. L.; Moore, L. B.; Wilson, J. G.; Lewis, M. C.; Jones, S. A.; Willson, T.
M. J. Med. Chem. 2000, 43, 2971.
5. Nicolau, K. C.; Evans, R. M.; Roecker, A. J.; Hughes, R.; Downes, M.; Pfefferkorn, J.
A. Org. Biomol. Chem. 2003, 1, 908.
a more polar, yet unexplored region in the direction of the R00
vector. Specifically targeting this area of the FXR binding site, we
subsequently refined our lead molecules to generate further ana-
logs with improved physicochemical properties, in vitro efficacy
and in vivo pharmacological characteristics. More detailed infor-
mation will be reported in due course.
6. Flatt, B.; Martin, R.; Wang, T. L.; Mahaney, P.; Murphy, B.; Gu, X.-H.; Foster, P.;
Li, J.; Pircher, P.; Petrowski, M.; Schulman, I.; Westin, S.; Wrobel, J.; Yan, G.;
Bischoff, E.; Daige, C.; Mohan, R. J. Med. Chem. 2009, 52, 904.
7. Pellicciari, R.; Fiorucci, S.; Camaioni, E.; Clerici, C.; Constantino, G.; Maloney, P.
R.; Morelli, A.; Parks, D. J.; Willson, T. M. J. Med. Chem. 2002, 45, 3569.
8. General procedure: mono-Boc protected phenylenediamines were dissolved in
MeOH and equimolar amounts of carboxylate, aldehyde and isonitrile added.
The reaction mixture was stirred at rt overnight. After evaporation of the
solvent, neat TFA was added and the reaction mixture again stirred at room
temperature. Spontaneous cyclization to the Benzimidazole was usually
observed after 4–12 h.
References and notes
1. Ottow, E.; Weinmann, H. Nuclear Receptors as Drug Targets; Wiley-VCH:
Weinheim, 2008.
2. (a) Crawley, M. L. Expert Opin. Ther. Pat. 2010, 1047; (b) Hageman, J. H.;
Herrema, H.; Groen, A. K.; Kuipers, F. Arterioscler. Thromb. Vasc. Biol. 2010, 30,
1519; (c) Cariou, B.; van Harmelen, K.; Duran-Sandoval, D.; van Dijk, T. H.;
Grefhorst, A.; Abdelkarim, M.; Caron, S.; Torpier, G.; Fruchart, J. C.; Gonzalez, F.
J. J. Biol. Chem. 2006, 281, 11039.
9. Tempest, P.; Ma, V.; Kelly, M. G.; Jones, W.; Hulme, C. Tetrahedron Lett. 2001, 42,
4963.
10. Nicholas, J. S.; Parks, D. J.; Consler, T. G.; Blanchard, S. G. Anal. Biochem. 1998,
257, 112.
3. Wang, Y.-D.; Chen, W.-D.; Moore, D. D.; Huang, W. Cell Res. 2008, 18, 1087.
11. Kanaya, E.; Shiraki, T.; Jingami, H. Biochem. J. 2004, 15, 913.