Identification of potent and selective retinoic acid receptor gamma (RARγ) antagonists for the treatment of osteoarthritis pain using structure based drug design

Article abstract of DOI:10.1016/j.bmcl.2016.05.056

A series of triaryl pyrazoles were identified as potent pan antagonists for the retinoic acid receptors (RARs) α, β and γ. X-ray crystallography and structure-based drug design were used to improve selectivity for RARγ by targeting residue differences in

Full text of DOI:10.1016/j.bmcl.2016.05.056

Bioorganic & Medicinal Chemistry Letters xxx (2016) xxx–xxx
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry Letters
journal homepage: www.elsevier.com/locate/bmcl
Identification of potent and selective retinoic acid receptor gamma
(RARc) antagonists for the treatment of osteoarthritis pain using
structure based drug design
Norman E. Hughes a, Thomas J. Bleisch a, Scott A. Jones a, Timothy I. Richardson a, Robert A. Doti a,
Yong Wang a, Stephanie L. Stout a, Gregory L. Durst a, Mark G. Chambers b, Jennifer L. Oskins b,
Chaohua Lin b, Lisa A. Adams c, Todd J. Page c, Robert J. Barr a, Richard W. Zink a, Harold Osborne a,
Chahrzad Montrose-Rafizadeh a, Bryan H. Norman a,
⇑
a Discovery Chemistry Research and Technologies, Eli Lilly and Company, Lilly Corporate Center, Indianapolis, IN 46285, United States
b Musculoskeletal Research, Eli Lilly and Company, Lilly Corporate Center, Indianapolis, IN 46285, United States
c Toxicology/Drug Disposition, Eli Lilly and Company, Lilly Corporate Center, Indianapolis, IN 46285, United States
a r t i c l e i n f o
a b s t r a c t
Article history:
A series of triaryl pyrazoles were identified as potent pan antagonists for the retinoic acid receptors
Received 13 April 2016
Revised 18 May 2016
Accepted 19 May 2016
Available online xxxx
(RARs)
ity for RAR
resulted in the discovery of novel antagonists which maintained RAR
500-fold selective versus RAR and RARb. The potent and selective RAR
a
, b and
c. X-ray crystallography and structure-based drug design were used to improve selectiv-
c
by targeting residue differences in the ligand binding pockets of these receptors. This
c
potency but were greater than
a
c antagonist LY2955303 demon-
strated good pharmacokinetic properties and was efficacious in the MIA model of osteoarthritis-like joint
pain. This compound demonstrated an improved margin to RAR -mediated adverse effects.
Ó 2016 Elsevier Ltd. All rights reserved.
Keywords:
Retinoic acid receptor
a
RAR
c
Antagonist
Osteoarthritis
Pain
Retinoic acid receptors (RAR
a, b and
c
) are part of a superfamily
on the severe testicular effects observed in RARa
knock-out mice,9
of nuclear receptors (NRs) that behave as ligand-activated tran-
scription factors and operate as part of a complex signaling net-
work.1 It has been shown that the natural ligand for the RARs, All
Trans Retinoic Acid (ATRA), is deleterious to articular cartilage
health and is associated with the breakdown of cartilage in
osteoarthritis.2 Both natural and synthetic retinoids (RAR agonists)
are catabolic to cartilage, block early chondrogenesis and promote
chondrocyte hypertrophy via RAR-mediated signaling.3 It was also
demonstrated that retinoid levels are increased in the synovial
fluid of OA patients as a function of disease severity.4 Finally, ATRA
has been shown to cause nociceptive pain in rodents and this effect
can be blocked by a pan-RAR antagonist.5,6 It has been postulated
that RAR antagonists may prevent or reverse retinoid-mediated
cartilage destruction and mitigate OA pain.
this effect is largely believed to be RAR
a
-mediated. In contrast, we
agonist was sufficient to
recently showed that a selective RAR
c
cause an increase in markers of articular cartilage catabolism, such
as the proteolytic enzyme ADAMTS-5, in rats and a selective RAR
antagonist could reverse those effects.10 Thus, due to the apparent
importance of RAR in the joint, and the clear safety issues associ-
ated with RAR antagonism in other tissues, we sought to identify
RAR antagonists with very high selectivity versus RAR for
potential use in the treatment of OA and OA pain.
c
c
a
c
a
Our lead generation began with the identification of 1 (Fig. 1),
which contained the tetramethyl-tetrahydronaphthalene sub-
structure (in red), common to many synthetic retinoids. This com-
pound was a potent RAR
selectivity versus RAR and b. Additionally, 1 demonstrated full
RAR antagonist function in
c
binder, but with only modest (ꢀ10Â)
a
A RAR pan antagonist (BMS-189453) was previously shown to
improve clinical scores in rodent models of joint pain and inflam-
mation,7 albeit with unacceptable adverse effects on testes.8 Based
c
a
cellular co-transfection assay
(KB = 2.2 nM, max inhibition >100%).11,12 Removal of the tetram-
ethyl-tetrahydronaphthalene substructure resulted in 2, which
maintained similar potency, but diminished selectivity.
Crystallography and structure-based drug design were used to
improve selectivity of this scaffold. With sufficient potency for
⇑
Corresponding author.
E-mail address: norman@lilly.com (B.H. Norman).
http://dx.doi.org/10.1016/j.bmcl.2016.05.056
0960-894X/Ó 2016 Elsevier Ltd. All rights reserved.
Please cite this article in press as: Hughes, N. E.; et al. Bioorg. Med. Chem. Lett. (2016), http://dx.doi.org/10.1016/j.bmcl.2016.05.056

2
N. E. Hughes et al. / Bioorg. Med. Chem. Lett. xxx (2016) xxx–xxx
Table 1
Residue differences in RAR subtype ligand binding domains
Helix 3
Helix 5
Helix 11
RAR
RARb
RAR
a
S232
A225
A234
I270
I263
M272
V395
V388
A397
c
binding, but maintain RAR
the C5 position of this aryl ring, which, as shown in Figure 2, was
less than 4 Å away from the sidechain atoms of Ile270 in RAR
c affinity, we added functional groups to
a.
The impact of C5-substitution is shown in Table 2. Increasing
the size of this substituent led to progressively poorer affinity
towards RAR
In addition to maintaining good RAR
and RARb, 5 also demonstrated
a
and RARb, but maintained high affinity to RAR
c.
c
binding affinity and
>100Â selectivity versus RAR
a
potent antagonist function in the cellular co-transfection assay
(KB = 35 nM, Max. inhibition >100%).
While we had optimized the selectivity via substitutions near
helix 5, the resulting antagonists, such as 5, demonstrated poor sol-
ubility (<100 ng/mL) in simulated intestinal fluid. To mitigate the
potential for solubility limiting absorption, we explored amides
and found that they were suitable sulfone replacements, as shown
in Table 3. In fact, compared to sulfone 5, substituted amides main-
Figure 1. RAR pan antagonist hits.
RARc
, we focused on the solution of a structure bound to RARa.
Our goal was to disrupt RAR
tially maintain RAR binding. Thus, we solved the structure of 2
bound to the ligand binding domain of RAR (Fig. 2). Based on pre-
a binding in a way that could poten-
c
tained RARc affinity and showed even higher selectivities versus
a
RAR and RARb, in contrast to pan antagonist BMS-189453, which,
a
viously reported RAR structures,13 we were not surprised by the
binding pose of 2 with the benzoic acid functionality anchored to
the subtype conserved arginine and the methylsulfonyl-phenyl
group bound near helix 12.14 Similar to other NHR antagonists, this
aromatic group perturbs helix 12 in a way that prevents coactiva-
tor binding, which is necessary for antagonist function.
as expected, was not selective in our hands. Furthermore, com-
pounds in this series were similarly potent and selective in the
RAR
c
co-transfection (CTF) functional assay in HEK293 cells
KB >4440 nM; RARb KB = 1510 nM; RAR
(LY2955303: RAR
a
c
KB = 7.11 nM). Additionally, being a zwitterion, LY2955303 showed
dramatic solubility improvement (>1.0 mg/mL in simulated
intestinal fluid).
The ligand binding domain of RARs are highly conserved with
differences occurring in only 3 residues, as noted in Table 1.15
Our attention was drawn to the residues in helix 5. Specifically,
Having identified a potent and selective RARc antagonist, with
suitable properties, we sought to test our hypotheses around (a)
improved pharmacokinetic profile in rats, (b) efficacy in a rat
model of OA-like joint pain and (c) improved safety profile versus
we were interested in the single residue difference between RAR
and RARb versus RAR in this region. We hypothesized that the
branched sidechain of Ile270 (RAR ) and Ile263 (RARb) would
a
c
a
RARa mediated testicular effects. The pharmacokinetic profile for
occupy more three dimensional space and be less tolerant to
ligand-associated steric volume, than the linear sidechain of
Met272 found in RARc. Thus, to potentially disrupt RARa and RARb
10 (Fig. 3) showed proportional increases in Cmax and AUC at 10,
30 and 100 mg/kg after oral administration, with no accumulation
upon chronic dosing up to 14 days.16 Thus, we were ready to
evaluate this compound in efficacy and safety studies.
To assess the pharmacodynamic effects of selective antagonism
of RARc, we evaluated LY2955303 in the mono-iodoacetate (MIA)
model of OA-like joint pain in rats.17 This model assesses differen-
tial weight bearing as a measure of joint pain after a single intra-
articular injection of MIA into a hind limb joint. At nine days post
Table 2
Binding affinity and subtype specificity of RARc antagonists
R
Binding Ki + SEMa (nM)
RAR RARb
Selectivity
versus a
a
RARc
c
2
3
4
5
H
1.8 0.3
6.2 3.3
87 7.8
461 93
4.3 0.9
10.0 2.6
131 9.9
995 196
1.1 0.5
1.1 0.4
1.6 0.2
4.4 0.8
1.6Â
5.6Â
54Â
Me
i-Pr
t-Bu
Figure 2. 1.85 Å structure of 2 (grey) bound to the RAR
(green, PDB: 5K13), aligned and overlaid with the RAR
(orange) from the agonist structure (PDB: 1EXA) with helix 12 truncated.
a
c
ligand binding pocket
ligand binding pocket
106Â
a
All reported data derived from at least 3 replicates.
Please cite this article in press as: Hughes, N. E.; et al. Bioorg. Med. Chem. Lett. (2016), http://dx.doi.org/10.1016/j.bmcl.2016.05.056

N. E. Hughes et al. / Bioorg. Med. Chem. Lett. xxx (2016) xxx–xxx
3
Table 3
In vitro profile of compounds 5–9 and LY2955303
O
O
O
O
S
R
N
R
X
N
N
N
N
N
N
CO2H
CO2H
CO2H
Compound
R
Compound
R
NH2
NMe2
Compound
9
LY2955303
X
5
6
7
8
O
Me
NMe
NMe2
Compd
Binding Ki + SEMa (nM)
RARb
Selectivity
CTF functionala
KB + SEM (nM)
RARa
RARc
c
versus
a
c
versus b
RARc
Max. inhib. (%)
5
6
7
8
461 93
>1700
>1700
>1700
>1700
995 196
>2980
>2980
>2980
>2980
4.4 0.8
4.1 1.4
1.4 0.3
2.2 0.5
2.3 0.1
1.1 0.3
17.3 2.8
106Â
228Â
13.5 6.4
11.1 5.9
14.7 8.8
21.3 14.6
7.2 5.5
>100
>100
79
>100
>100
>100
>100
>414Â
>1259Â
>766Â
>752Â
>1560Â
>725Â
>2207Â
>1342Â
>1319Â
>2734Â
Not selective
9
LY2955303
BMS-189453
>1700
7.7 1.1
>2980
16.7 4.4
7.1 4.9
18.5 9.2
Not selective
a
All reported data derived from at least 3 replicates.
Previously, Beehler reported that RAR pan-antagonist BMS-
189453 was efficacious in the mouse collagen induced arthritis
(CIA) model and the streptococcyl cell wall induced arthritis
(SCWA) model, both at a dose of 15 mg/kg.7 It was also reported
by Schulze that this same compound demonstrated testicular
adverse effects in rats at doses as low as 2 mg/kg.8 Based on the
previous genetic studies that showed robust testicular degenera-
tion in RARa 9
learn if a selective RAR
,
but not RAR
antagonist could avoid the testicular
c
knockout mice,18 we were eager to
c
effects noted with a pan antagonist at efficacious exposures. Thus,
we tested LY2955303 in a 14 day toxicology study in rats and
found adverse testicular effects only at doses significantly exceed-
ing the MIA ED50 (Table 4). Based on drug exposure (AUC) in the rat
at the efficacious dose of 0.72 mg/kg, the margin of exposure at the
no effect dose for testicular effects (10 mg/kg) was 59 fold. Testic-
ular degeneration was observed at the 30 mg/kg dose, which had a
margin of exposure equal to 239 fold over the exposure at the MIA
ED50
.
Figure 3. Pharmacokinetic profile of LY2955303.
The methods used to prepare triaryl pyrazoles have been previ-
ously described.11 As a representative example, Scheme 1 outlines
the procedure used to prepare compound LY2955303. Methyl 4-
acetylbenzoate (12) was treated with sodium hydroxide in metha-
nol, followed by the addition of 3,5-di-tert-butylbenzaldehyde
(11), resulting in formation of the intermediate chalcone acid,
which was esterified using methanesulfonic acid in methanol to
give chalcone ester 13. 13 was treated with hydrazine 14 and
acetic acid in n-butanol, condensing and cyclizing to the interme-
diate pyrazoline, which was oxidized to pyrazole 15 using man-
ganese dioxide in dichloromethane. 15 was treated with N-
methylpiperazine, HOBT and EDCI in dichloromethane to give the
corresponding amide-ester, which was saponified to complete
the synthesis of LY2955303.
MIA injection, the differential weight bearing between the con-
tralateral and ipsilateral limbs was about 22 g (Fig. 4). On the same
day, a single oral dose of LY2955303 demonstrated a dose respon-
sive effect whereby the rat reduced differential weight bearing
(ED50 = 0.72 mg/kg). The maximal analgesic response observed
was similar to the maximal response observed for other
analgesics.16
Table 4
Rat toxicological data for LY2955303
Dose (mg/kg)
AUC (ng h/ml)
MOE1
Testes effect
0.72 (ED50
10
30
)
1542
1
91003
36,8003
59Â
No adverse effect
Degeneration
239Â
1
2
3
Margin of exposure.
Determined from a rat exposure study at the ED50 of 0.72 mg/kg.
Determined from toxicokinetic study.
Figure 4. Dose responsive MIA analgesic data for LY2955303.
Please cite this article in press as: Hughes, N. E.; et al. Bioorg. Med. Chem. Lett. (2016), http://dx.doi.org/10.1016/j.bmcl.2016.05.056

4
N. E. Hughes et al. / Bioorg. Med. Chem. Lett. xxx (2016) xxx–xxx
O
O
O
a, b
HO
O
O
+
+
NH2
O
55%
N
H
O
O
N
11
12
13
14
O
O
HO
N
N
N
N
c, d
e, f
N
O
OH
O
78%
85%
O
15
LY2955303
Scheme 1. Synthesis of compound LY2955303. Reagents and conditions: (a) NaOH, CH3OH; (b) CH3SO3H, CH3OH; (c) HOAc, n-BuOH; (d) MnO2, CH2Cl2;
(e) N-methylpiperazine, HOBT, EDCI, CH2Cl2, (f) KOH, THF/water.
5. Romero-Sandoval, E. A.; Alique, M.; Moreno-Manzano1, V.; Molina, C.; Lucio, F.
J.; Herrero, J. F. Inflamm. Res. 2004, 53, 297.
6. Alique, M.; Lucio, F. J.; Herrero, J. F. Br. J. Pharmacol. 2006, 149, 65.
In conclusion, we have identified triaryl pyrazoles, a novel RAR
antagonist scaffold. We also utilized a ligand–protein x-ray crystal
structure to guide a structure-based drug design strategy to
improve selectivity. This approach culminated in the identification
of LY2955303, which displayed exquisite potency and robust selec-
7. Beehler, B. C.; Hei, Y.; Chen, S.; Lupisella, J. A.; Ostrowski, J.; Starrett, J. E.;
Tortolani, D.; Tramposch, K. M.; Reczek, P. R. J. Rheum. 2003, 30, 355.
8. Schulze, G. E.; Clay, R. J.; Mezza, L. E.; Bregman, C. L.; Buroker, R. A.; Franz, J. D.
Toxicol. Sci. 2001, 59, 297.
9. Lufkin, T.; Lohnes, D.; Mark, M.; Dierich, A.; Gorry, P.; Gaub, M. P.; LeMeur, M.;
Chambon, P. Proc. Natl. Acad. Sci. U.S.A. 1993, 90, 7225.
10. Norman, B. H.; Geiser, A.; Chambers, M.; Moore, T.; Chandrasekhar, S.; Page, T.;
Bleisch, T.; Mitchell, P. Osteoarthritis Cartilage 2013, 21, S287.
11. Bleisch, T. J.; Coates, D. A.; Hughes, N. E.; Jones, S. A.; Norman, B. H. PCT Int.
Appl., 2013, WO 2,013,066,640.
tivity for RARc versus the other subtypes RARa and RARb in vitro.
This compound showed good pharmacokinetic properties and effi-
cacy in rodents at exposures significantly lower than the exposures
that produced testicular toxicity, consistent with our hypothesis
that a selective RARc antagonist could maintain efficacy with an
12. Binding and functional in vitro assays have been previously described (see Ref.
11). Functional activity was determined using human embryonic kidney HEK
293 cells transiently transfected with hybrid GAL4 DNA binding domain (DBD)
fused to RARgamma, alpha, or beta ligand binding domain (LBD) downstream
of an SV40 promoter, and a reporter plasmid containing 5Â GAL4 response
element upstream of a firefly luciferase gene. Antagonist activity was deter-
mined in the presence of all trans retinoic acid (ATRA) at the corresponding
EC80 concentrations (15 nM for RAR alpha and gamma receptors and 10 nM for
RAR beta). Percent inhibition was calculated based on the response obtained in
the presence (baseline) and absence (100% inhibition) of ATRA alone. IC50 was
calculated using a 4 parameter logistics. An apparent KB was calculated using
Cheng–Prusoff equation and the EC50 of ATRA determined within each assay.
13. Klaholz, B. P.; Mitschler, A.; Moras, D. J. Mol. Biol. 2000, 302, 155.
improved safety profile relative to RAR pan antagonists.
Acknowledgements
This research used resources of the Advanced Photon Source, a
U.S. Department of Energy (DOE) Office of Science User Facility
operated for the DOE Office of Science by Argonne National
Laboratory under Contract No. DE-AC02-06CH11357.
References and notes
14. The X-ray data for 2 bound to RARa (Fig. 2) have been deposited into the
Protein Data Bank with accession code 5K13.
15. Renaud, J.; Rochel, N.; Ruff, M.; Vivat, V.; Chambon, P.; Gronemeyer, H.; Moras,
D. Nature 1995, 378, 681.
1. Sporn, M. B.; Roberts, A. B.; Goodman, D. S. The Retinoids, Biology, Chemistry and
Medicine, 2nd ed.; Raven Press: NY, 1993.
16. A rat bioavailability study (2 mg/kg iv and 10 mg/kg po) demonstrated 26% oral
bioavailability.
2. Kubo, M.; Takase, T.; Matsusue, Y.; Rauvala, H.; Imai, S. Osteoarthritis Cartilage
2002, 10, 968.
17. Bove, S. E.; Calcaterra, S. L.; Brooker, R. M.; Huber, C. M.; Guzman, R. E.; Juneau,
P. L.; Schrier, D. J.; Kilgore, K. S. Osteoarthritis Cartilage 2003, 11, 821.
18. Lohnes, D.; Kastner, P.; Dierich, A.; Mark, M.; LeMeur, M.; Chambon, P. Cell
1993, 73, 643.
3. Johnson, K. A.; van Etten, D.; Nanda, N.; Graham, R. M.; Terkeltaub, R. A. J. Biol.
Chem. 2003, 278, 18824.
4. Davies, M. R.; Ribeiro, L. R.; Downey-Jones, M.; Needham, M. R. C.; Oakley, C.;
Wardale, J. Arthritis Rheum. 2009, 60, 1722.
Please cite this article in press as: Hughes, N. E.; et al. Bioorg. Med. Chem. Lett. (2016), http://dx.doi.org/10.1016/j.bmcl.2016.05.056