Discovery of tertiary aminoacids as dual PPARα/γ agonists-I

Article abstract of DOI:10.1016/j.bmcl.2007.01.060

A novel series of potent dual agonists of PPARα and PPARγ, the alkoxybenzylglycines, was identified and explored using a solution-phase library approach. The synthesis and structure-activity relationships of this series of dual PPARα/γ agonists are descri

Full text of DOI:10.1016/j.bmcl.2007.01.060

Bioorganic & Medicinal Chemistry Letters 17 (2007) 2312–2316
Discovery of tertiary aminoacids as dual PPARa/c agonists-I
Pratik V. Devasthale,a,* Sean Chen,a Yoon Jeon,a Fucheng Qu,a Denis E. Ryono,a
Wei Wang,a Hao Zhang,a Lin Cheng,b Dennis Farrelly,b Rajasree Golla,b Gary Grover,b
Zhengping Ma,b Lisa Moore,b Ramakrishna Seethala,b Wei Sun,b Arthur M. Doweyko,c
Gamini Chandrasena,d Paul Sleph,b Narayanan Hariharanb and Peter T. W. Chenga,*
aDiscovery Chemistry, Bristol-Myers Squibb Pharmaceutical Research Institute, PO Box 5400, Princeton, NJ 08543-5400, USA
bDiscovery Biology, Bristol-Myers Squibb Pharmaceutical Research Institute, PO Box 5400, Princeton, NJ 08543-5400, USA
cComputer-Assisted Drug Design, Bristol-Myers Squibb Pharmaceutical Research Institute,
PO Box 5400, Princeton, NJ 08543-5400, USA
dMetabolism and Pharmacokinetics, Bristol-Myers Squibb Pharmaceutical Research Institute,
PO Box 5400, Princeton, NJ 08543-5400, USA
Received 28 September 2006; revised 17 January 2007; accepted 18 January 2007
Available online 25 January 2007
Abstract—A novel series of potent dual agonists of PPARa and PPARc, the alkoxybenzylglycines, was identified and explored using
a solution-phase library approach. The synthesis and structure–activity relationships of this series of dual PPARa/c agonists are
described.
Ó 2007 Elsevier Ltd. All rights reserved.
Peroxisome proliferator-activated receptors (PPARs)1
are nuclear hormone receptors which act as transcrip-
tion factors in the regulation of genes involved in glu-
cose and lipid metabolism, and vessel wall function.
They are, therefore, relevant targets in such disease
areas as diabetes mellitus, obesity, inflammation, and
atherosclerosis. PPARa (expressed highly in the liver
and involved in fatty acid oxidation and lipoprotein
metabolism) is the target of the fibrate class of hypolip-
idemic drugs such as fenofibrate2 and gemfibrozil.3
PPARc (predominantly expressed in adipose tissue
and implicated in insulin sensitization, glucose and fatty
acid utilization as well as adipocyte differentiation) is the
target of the thiazolidinedione (TZD)4 class of antidia-
betic drugs such as rosiglitazone5 and pioglitazone.6 It
has therefore been hypothesized that a dual PPARa/c
agonist that would improve insulin sensitivity, lower
glucose and correct lipid abnormalities would be highly
beneficial for the treatment of type 2 diabetes and asso-
ciated dyslipidemia. The clinical utility of dual PPARa/c
agonists has been demonstrated in patients with type 2
diabetes.7
In this paper, we report on the initial SAR ofthe novel alk-
oxybenzylglycines 1 and 2, which are shown to be PPAR
ligands with potent agonist activities at both PPARa and
PPARc. The concept and design of the alkoxybenzylgly-
cines as PPAR ligands has been discussed in an earlier
publication.8 We decided to initiate our studies on explor-
ing the alkoxybenzylglycine chemotype through the syn-
thesis of tertiary aminoacid analogs with a two-carbon
linker, commonly used in the PPAR literature, between
the phenyloxazole and the central phenyl ring.
The initial lead compound, benzylamine 2a (Table 2),
displayed modest binding affinity to and promising
functional activity at both PPARa and PPARc. We
decided to explore the SAR of this initial lead com-
pound through the preparation of two solution-phase
libraries, as described below.
The key intermediate aminoacids 8 and 9 were prepared
in good yield, respectively, from 3- and 4-hydroxybenz-
aldehyde via: (1) reductive amination9 of 6 and 7 with
glycine tert-butyl ester/NaBH4 and (2) acidic deprotec-
tion of the tert-butyl ester (Scheme 1).
Keywords: PPAR; PPAR a/c dual agonists; Antidiabetic;
Oxybenzylglycines.
*
Corresponding authors. Tel.: +1 609 818 4974; fax: +1 609 818
3460; e-mail addresses: pratik.devasthale@bms.com; peter.t.cheng@
bms.com
0960-894X/$ - see front matter Ó 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2007.01.060

P. V. Devasthale et al. / Bioorg. Med. Chem. Lett. 17 (2007) 2312–2316
2313
O
5
1. glycine t-Bu ester;
NaBH4
CHO
CHO
N
OMs
O
N
K2CO3, CH3CN, 80 oC
2. 50% TFA:CH2Cl2
HO
O
3, 3-Hydroxybenzaldehyde
4, 4-Hydroxybenzaldehyde
6, 1,3-substitution
7, 1,4-substitution
RCHO, NaBH(OAc)3
CH2ClCH2Cl; SPE
N
H
CO2H
N
CO2H
O
O
N
R
N
O
O
8, 1,3-substitution
9, 1,4-substitution
1, 1,3-substitution
2, 1,4-substitution
Scheme 1. Synthesis of secondary amines.
We began our SAR studies by synthesizing a number of
tertiary amine analogs individually via reductive amina-
tion of the corresponding methyl or tert-butyl esters of 8
or 9.10a Subsequently, solution-phase libraries of tertiary
amines 1 and 2 in both the 1,3- and 1,4-oxybenzylglycine
series, respectively, were synthesized by reductive amina-
tion of aminoacids 8 and 9 with a diverse set of aromat-
ic, heteroaromatic and aliphatic aldehydes in a parallel
format. It is notable that satisfactory purification of
the crude reductive amination products was achieved
using solid-phase extraction (anion-exchange SAX
cartridges).10b
EC50 (GW-2331) = 71 nM]. Overall, analogs with an
arylalkyl group (e.g., benzyl, heteroarylmethyl) at R1
showed comparable potencies in terms of PPARa and
PPARc binding affinity as well as functional activity.
In the 1,3-alkoxybenzylglycine series 1, increasing the
linker length between the phenyl group and the glycine
amine (analogs 1a–c) results in only a modest change
in binding or functional activity even though 1c displays
a PPARa EC50 which is 19-fold more potent than its
PPARc EC50. Annulation of the N-benzyl analog (1a)
to provide the 1-naphthylmethyl compound 1g resulted
in increased binding and functional activity at both
receptors with nearly equivalent activity in their ability
to differentiate 3T3L1 preadipocytes. However, the
regioisomeric 2-naphthylmethyl analog (1h) showed
decreased transactivation activity at both receptors.
A number of tertiary amino acids in the 1,3-substituted
alkoxyphenyl series 1 are potent PPARa-selective ago-
nists (1a, 1c, 1e, 1g, 1h, and 1i). Compounds 1b and
1d (Table 1, R1 = 2-phenylethyl and 4-phenoxybenzyl,
respectively) both showed relatively equivalent activity
at PPARa and PPARc (<3-fold difference). Moreover,
compound 1d is 2-fold less potent than rosiglitazone at
PPARc [EC50 = 67 nM; EC50 (rosiglitazone) = 35 nM]
and 2-fold more potent than the reference PPARa/c
dual agonist GW-2331 at PPARa [1d EC50 = 30 nM;
The 1,4-alkoxyphenyl analog 2a (Table 2) was consider-
ably less potent (both in terms of binding affinity and
EC50 values in the PPARa and PPARc transactivation
assays) than its 1,3-alkoxyphenyl counterpart 1a. Sur-
prisingly, unlike corresponding literature analogs in
the a-alkoxy propanoic acid series,11 tertiary amines
Table 1. In vitro activities against PPARc and a in the 1,3-substituted series (11)14
O
N
N
CO2H
R1
O
1
Compound R1
PPARc
HEK (PPARc)
Kact 3T3L1
PPARa
HEK (PPARa)
IC50 (lM) EC50 (lM) cKact
(% at 1 lM)
(PPARc) (%) IC50 (lM) EC50 (lM) aKact
(% at 1 lM)
GW233115
Rosiglitazone
0.670
0.256
0.240
0.134
0.773
0.23
0.249 0.238 (90%)
0.035 0.0118 (108%) 101
0.295 0.1880 (82%) 104
0.108 0.0718 (102%) 147
105
0.348
Inactive
0.090
0.075
0.051
0.11
0.071 0.035 (79%)
Inactive
1a
1b
1c
1d
1e
1f
Benzyl10a
0.022 0.0063 (72%)
0.044 0.0346 (82%)
0.047 0.0080 (62%)
0.030 0.0184 (67%)
0.062 0.0641 (95%)
0.324 0.1679 (56%)
0.005a (85%)
2-Phenylethyl10a
3-Phenylpropyl10a
4-Phenoxybenzyl10a
3-Phenoxybenzyl10a
4-Benzyloxybenzyl10b
1-Naphthylmethyl10a
2-Naphthylmethyl10a
1H-indol-2-ylmethyl10a
0.901 0.2229 (43%)
0.067 0.0427 (73%)
0.359 0.0515 (90%)
0.412 0.0701 (59%)
118
116
115
88
0.11
0.083
0.387
0.034
0.149
0.246
0.080
0.183
0.067
0.145
0.540
1g
1h
1i
0.036 0.0105 (115%) 92
1.39 0.4374 (49%)
0.759 0.3733 (58%)
0.082a (32%)
94
0.313 0.1080 (65%)
0.197 0.0966 (62%)
0.153a (27%)
NDb
NDb
1j
1-[2,20]Bithiophenyl-5-ylmethyl10a 0.101
a No dose–response or the curve did not top out in 2 other assay runs.
b ND, not determined.

2314
P. V. Devasthale et al. / Bioorg. Med. Chem. Lett. 17 (2007) 2312–2316
Table 2. In vitro activities against PPARc and a in the 1,4-substituted series (2)14
O
N
CO2H
R1
N
O
2
Compound R1
PPARc
HEK (PPARc)
Kact 3T3L1
PPARa
HEK (PPARa)
IC50 (lM) EC50 (lM) cKact
(% at 1 lM)
(PPARc) (%) IC50 (lM) EC50 (lM) aKact
(% at 1 lM)
GI 262570 (Farglitazar)
Benzyl10a
0.217
1.77
2.30
0.0006 0.0007 (99%) 112
3.6c (28%)
125
0.136 0.0821 (66%) 87
2.87
1.4
0.321 0.0783 (59%)
4.02c (24%)
2a
2b
2c
2d
2e
2f
n-Heptyl10b
1.05
0.180 0.0936 (63%)
8%c
2-Ethyl-1-butyl10a
2-Benzyloxyethyl10b
Benzo[1,3]dioxol-5-yl-methyl10b
3-Phenoxybenzyl10b
4-Phenoxybenzyl10b
4-Benzyloxybenzyl10b
1-Naphthylmethyl10b
2-Naphthylmethyl10a
2-Pyridylmethyl10b
Inactive
4.56
3.29
1.8 0.3748 (32%)
1.18 0.3394 (41%)
NDa
NDa
1.8
154
0.726 0.0697 (56%)
0.128 0.0760 (77%)
0.004 0.0003 (107%)
0.035 0.0088 (59%)
0.189 0.1535 (106%)
0.021 0.0111 (90%)
0.618 0.1654 (39%)
0.692 0.3196 (12%)
0.307 0.1358 (72%) 140
0.391 0.0771 (78%) 117
0.052 0.0289 (70%) 152
0.221 0.0345 (82%) 40
0.840 0.7487 (63%) 73
1.21 0.05484 (29%) 108
0.230 0.0877 (20%) NDa
1.8
0.48
0.088
0.97
1.05
2g
2h
2i
0.227
0.284
0.767
1.21
0.14
0.68
2j
2k
2.9
Inactive
at 25 lM
1.8
2l
2m
5-(2-Chloro-phenyl)-furan-2-yl-methyl10a 0.280
1-[2,20]Bithiophenyl-5-ylmethyl10a
0.102
0.267 0.1580 (77%) 134
0.060 0.0369 (44%) NDa
0.871 0.4624 (55%)
0.078 0.0356 (38%)
Inactive
at 98 lMb
a ND, Not determined.
b Poor solubility in test medium.
c No dose–response or the curve did not top out in 2 other assay runs.
with nonaromatic R1 groups (linear, branched alkyl,
and cycloalkyl, for example 2b, R1 = n-hexyl and 2c,
R1 = 2-ethyl-1-butyl) showed poor PPARa and PPARc
functional activity in this new oxybenzylglycine chemo-
type. Alkyl substitutions in the 1,3-alkoxybenzylglycine
core were, therefore, not explored. Overall, among these
two libraries, two of the most promising compounds
in vitro were the N-4-phenoxybenzyl analogs 1d (PPAR-
cEC50 = 67 nM; PPARaEC50 = 30 nM) and 2g (PPAR-
cEC50 = 52 nM; PPARaEC50 = 35 nM). Additionally,
both compounds showed a superior ability to differenti-
ate 3T3L1 pre-adipocytes into fat-loaded mature adipo-
cytes (a PPARc-specific activity) to the PPARc agonist
standard rosiglitazone. Due to their excellent in vitro
potency, 1d and 2g were evaluated in db/db mice
(10 mg/kg dosed po/qd for 2 weeks) for their antidiabet-
ic in vivo activity (with muraglitazar15b being used as an
internal reference standard). Both these compounds
caused an insignificant change in glucose, triglyceride,
free fatty acid, and insulin levels in this animal model.
Postulating that the inactivity of 2g (and possibly 1d)
in vivo was probably due to its poor permeability
(Caco-2 (apical to basal) < 15 nm/s at pH 6.5 for 2g)
and poor metabolic stability (rate of metabo-
lism = 0.16 nmol/min/mg; 46% remaining after 10 min
incubation in mouse liver microsomes), we decided to
evaluate 1a which showed excellent permeability and
good metabolic stability in mouse liver microsomes
(Caco-2 = 251 nm/s at pH 5.5 and 225 nm/s at pH 7.4;
rate of metabolism = 0.048 nmol/min/mg; % remaining:
COOH
O
HN
N
O
O
Farglitazar
Ph
Figure 1. (a) CPK depiction of the published X-ray structure of farglitazar in PPARc12 (as part of a heterodimer with RXRa). The carboxylate
group is positioned behind helices 3 and 12, while the pendant phenyloxazole group occupies a region behind and to the right of helix 3, forming a
U-shaped conformer. (b) The modeled structure of analog 2g is shown overlaid with farglitazar (in cyan). The carboxylate and phenyloxazole groups
of 2g and farglitazar occupy similar pockets.

P. V. Devasthale et al. / Bioorg. Med. Chem. Lett. 17 (2007) 2312–2316
2315
Table 3. In vivo data in db/db mice
by the lead compound 2a. A solution-phase library
approach was used to rapidly explore the SAR in this
series, which resulted in the discovery of: (1) analogs
such as 1d and 2g which are potent agonists of PPARa
and PPARc, (2) 1a as an analog which showed good
oral antidiabetic and antidyslipidemic activity in vivo,
and (3) several analogs with PPARa/c profiles that can
be used as tool molecules for PPAR research. Further
work in the development of the SAR of this lead series
based on the novel alkoxybenzylglycine core will be
described in a subsequent communication.
Compound
TG
Glucose NEFA Insulin
1a
% change ꢀ29
ꢀ31
ꢀ22
ꢀ40
P value
0.017 0.011
0.256
0.061
Mura
% change ꢀ50
ꢀ50
ꢀ69
ꢀ67
P value 0.000 0.000
0.000
0.005
84). Administration of 1a, which is somewhat less active
in vitro at PPARc (EC50 = 295 nM), did result in good
in vivo efficacy (Table 3).
In the 14-day db/db mouse model, 1a (at 10 mg/kg/day)
significantly decreased levels of fasted glucose (31%), tri-
glycerides (29%), insulin (40%), and non-esterified fatty
acids (22%).
References and notes
1. For general reviews on PPARs: (a) Torra, I. P.; Chinetti,
G.; Duval, C.; Fruchart, J.-C.; Staels, B. Curr. Opin.
Lipidol 2001, 12, 245; (b) Evans, A. J.; Krentz, A. J. Drugs
R&D 1999, 2, 75; (c) Sternbach, D. D. Annu. Rep. Med.
Chem. 2003, 38, 71; (d) Cheng, P. T. W.; Mukherjee, R.
Mini-Rev. Med. Chem. 2005, 5, 741.
2. (a) Packard, C. J. Eur. Heart J. 1998, A62; (b) Balfour, J.
A.; McTavish, D.; Heel, R. C. Drugs 1990, 40, 260; (c)
Despres, J.-P. Am. J. Cardiol. 2001, 88, 30N.
3. Spencer, C. M.; Barradell, L. B. Drugs 1996, 51, 982.
4. Mudaliar, S.; Henry, R. R. Annu. Rev. Med. 2001, 52, 239.
5. (a) Barman Balfour, J. A.; Plosker, G. L. Drugs 1999, 57,
921; (b) Goldstein, B. J. Int. J. Clin. Pract. 2000, 54, 333;
(c) Cheng-Lai, A.; Levine, A. Heart Dis. 2000, 2, 3.
6. Gillies, P. S.; Dunn, C. J. Drugs 2000, 60, 333.
7. Buse, J. B.; Rubin, C. J.; Frederich, R.; Viraswami-
Appanna, K.; Lin, K. C.; Montoro, R.; Shockey, G.;
Davidson, J. A. Clin. Ther. 2005, 27, 1181.
A possible binding model of the alkoxybenzylglycine
analog 2g in PPARc, which is based on its similarity
to the published structure of the tyrosine-derived
PPARc-selective agonist farglitazar,12 is illustrated in
Figure 1. The modeling was conducted by minimizing
structures of the ligand and PPARc site residues using
an Amber force field within the Flo molecular modeling
program.13 By allowing all residues in van der Waals
contact with the ligand to co-minimize with the ligand,
a molecular binding model was developed which is con-
sistent with observed SAR. According to this model, the
residues in electrostatic contact with the carboxylate
group of 2g represent a similar set to that found in the
farglitazar structure, namely, S289, H323, Y327, H449,
and Y473. In addition, as shown in Panel B, it is likely
that a bridging water molecule exists which provides
for a H-bond between the phenyloxazole moiety of the
ligand and the backbone NH of S342. The lower hydro-
phobic region is occupied by the pendant diphenyl ether.
Transactivation activity is thought to be modulated by
the complex and critical conformational changes that
helix 12 (H12) adopts to favor its interaction with co-ac-
tivators and disfavor interactions with co-repressors.
Functional activity differences, that is transactivation,
may occur due to the manner in which the ligand con-
tacts the bed of Phe residues lining the bottom of the
hydrophobic pocket (Phe 282/H3), Phe 360/H7, and
Phe 363/H7). These residues appear to make important
contacts with other hydrophobic residues on H12, espe-
cially Phe 282 which is in close contact with Met 463/
H12. Thus, the trajectories of ligand components into
this region may be critical for mediating functional
activity. For example, compounds 1g and 1h, represent-
ing regional isomers of a naphthylmethyl moiety in con-
tact with Phe 282, exhibit similar binding affinities with
differing degrees of transactivation. In the case of 1a and
2a, the extent of H12 contact with the pendant benzylic
groups is slightly different between the 1,3 and 1,4 phen-
yl linkers, resulting in a significant difference in transac-
tivation, while for the 4-phenoxybenzyl pair, 1d and 2g,
the difference in transactivation disappears, presumably
due to sufficient contact with H12 with the larger hydro-
phobic groups present in both compounds.
8. (a) Devasthale, P. V. et al. J. Med. Chem. 2005, 48, 2248;
(b) Harrity, T. et al. Diabetes 2006, 55, 240.
9. Abdel-Magid, A. F.; Carson, K. G.; Harris, B. D.; Marya-
noff, C. A.; Shah, R. D. J. Org. Chem. 1996, 61, 3849.
10. Characterized by a) 1H NMR, LC–MS and HPLC; b)
LC–MS with HPLC purity >80%. Typical procedure for
the solution-phase library: To a solution of aminoacid 8 or
9 (0.074 mmol) in CH2Cl2 (2 mL) were added the
substituted benzaldehyde (0.37 mmol), NaBH(OAc)3
(0.37 mmol), and HOAc (0.1 mL). The reaction mixture
was stirred for 15 h at rt. The product was purified via
solid-phase extraction using a Varian SAX cartridge (3 g
of sorbent in a 6 mL column, 0.3 mequiv/g) by the
procedure outlined below: (a) The column was condi-
tioned with MeOH (10 mL) and CH2Cl2 (20 mL), (b) the
reaction mixture was loaded onto the SAX column, (c) the
column was rinsed with CH2Cl2 (10 mL), (d) the column
was rinsed with 1% TFA in MeOH (3 mL), and (e) the
product was eluted with 1% TFA in MeOH (20 mL). The
product solution was concentrated using a Speed Vac for
16 h to afford the desired crude product.
11. For example, Lohray, B. B.; Lohray, B. B.; Bajji, A. C.;
Kalchar, S.; Poondra, R. R.; Padakanti, S.; Chakrabarti,
R.; Vikramadithyan, R. K.; Misra, P.; Juluri, S.; Mamidi,
N. V. S. R.; Rajagopalan, R. J. Med. Chem. 2001, 44,
2675.
12. Gampe, R. T., Jr.; Montana, V. G.; Lambert, M. H.;
Miller, A. B.; Bledsoe, R. K.; Milburn, M. V.; Kliewer, S.
A.; Willson, T. M.; Xu, H. E. Mol. Cell 2000, 5, 545,
1FM9.pdb (Protein Data Bank of the Research Collabo-
ratory for Structural Bioinformatics, RCSB).
13. McMartin, C.; Bohacek, R. S. . J. Comput. Aid Mol. Des.
1997, 11, 333, Flo molecular modeling software, Thistle-
soft, Box 227, Colebrook, CT 06021 USA.
In conclusion, we have discovered a novel series of alk-
oxybenzylglycine dual PPARa/c agonists as exemplified

2316
P. V. Devasthale et al. / Bioorg. Med. Chem. Lett. 17 (2007) 2312–2316
14. a) IC50 and EC50 values are represented in micrometer and
are averages of 2–6 determinations. IC50 values for
PPARa and PPARc were determined by calculating
the amount of test compound required to cause
half-maximal inhibition of the binding of a fluoresce-
in-labeled PPARa or c agonist to the PPARa or c
ligand-binding domain, respectively. The fluorescence
polarization assay is described elsewhere (Seethala, R.
et al. Anal. Biochem., manuscript accepted). The EC50
value is defined as the concentration of the test ligand
required to elicit 50% of the maximal fold induction
of luciferase activity in HEK-293 cells transiently
co-transfected with GAL4-PPARa or GAL4-PPARc
and stably co-transfected with the luciferase reporter
gene. Kact (intrinsic activity) is defined as the activity of
a ligand at 1 lM relative to the activity of the primary
standards at 1 lM (GW2331[15] for PPARa and
rosiglitazone for PPARc, respectively) expressed as a
percentage. None of the compounds tested showed any
activity against PPARd.
15. Brown, P. J.; Smith-Oliver, T. A.; Charifson, P. S.;
Tomkinson, N. C. O.; Fivush, A. M.; Sternbach, D. D.;
Wade, L. E.; Orband-Miller, L.; Parks, D. J.; Blanchard,
S. G. Chem. Biol. 1997, 4, 909.