Amphipathic 3-phenyl-7-propylbenzisoxazoles; human PPaR γ, δ and α agonists

Article abstract of DOI:10.1016/S0960-894X(02)01029-6

A series of amphipathic 3-phenylbenzisoxazoles were found to be potent agonists of human PPARα, γ and δ. The optimization of acid proximal structure for in vitro and in vivo potency is described. Results of po dosed efficacy studies in the db/db mouse mod

Full text of DOI:10.1016/S0960-894X(02)01029-6

Bioorganic & Medicinal Chemistry Letters 13 (2003) 931–935
Amphipathic 3-Phenyl-7-propylbenzisoxazoles; Human
PPaR ꢀ, ꢁ and ꢂ Agonists
Alan D. Adams,a,* Winston Yuen,a Zao Hu,a Conrad Santini,a A. Brian Jones,a
Karen L. MacNaul,b Joel P. Berger,b Thomas W. Doebberb and David E. Mollerb
aDepartments of Basic Chemistry, Merck Research Laboratories, Merck & Co. Inc., PO Box 2000, Rahway, NJ 07065, USA
bMolecular Endocrinology, Merck Research Laboratories, Merck & Co. Inc., PO Box 2000, Rahway, NJ 07065, USA
Received 25 July 2002; accepted 7 October 2002
Abstract—A series of amphipathic 3-phenylbenzisoxazoles were found to be potent agonists of human PPARa, g and d. The opti-
mization of acid proximal structure for in vitro and in vivo potency is described. Results of po dosed efficacy studies in the db/db
mouse model of type 2 diabetes showed efficacy equal or superior to Rosiglitazone in correcting hyperglycemia and hyper-
triglyceridemia. Good functional receptor selectivity for PPARa and g over PPARd can be obtained.
# 2003 Elsevier Science Ltd. All rights reserved.
The utility of the peroxisome proliferator activated
receptor g (PPARg, NR1C3) nuclear hormone receptor
agonists in the treatment of hyperglycemia associated
with type 2 diabetes1,2 (DM2) and of PPARa (NR1C1)
agonists as hypolipidemics has been amply demon-
strated.3 Three thiazolidinedione PPARg or PPARg
and a agonist insulin sensitizers have been marketed for
treatment of DM2 over the past 5–7 years. The search
for better PPAR agonist drugs is driven by limited effi-
cacy and side-effects including instances of lethal hepa-
totoxicity associated with the first marketed drug
Troglitazone.4 The development of an amphipathic car-
boxylate lead 1 from a Merck Frosst Ltd4 antagonist
program into a series of benzisoxazole PPARa/d/g ago-
nists has been reported recently.5 The phenylacetic acid
2 is a potent but nonselective PPARg, d and a agonist
and an efficacious insulin sensitizer in insulin resistant
db/db (leprdb-3J/leprdb-3J) mice. Further optimization of
2 and the optimization of the series for PPARa/g selec-
tivity are reported here (Fig. 1).
ments for acid structure with the objective of maximum
in vivo potency and determination of best attainable
selectivity among the three known PPAR receptors for
this series. Several varied acid structures give good
potency and widely varied selectivities in this series. Good,
but not complete, functional selectivity for PPARg and
a dual agonist activity over PPARd (NR1C2) can be
obtained while retaining reasonable potency.
Figure 1.
Typical SAR studies of PPAR agonists hold an acid or
heterocyclic head fragment relatively fixed and explore
the SAR for a lipophilic tail. This study examines the
opposite relation, holding a lipophilic 3-phenyl-7-propyl-
benzisoxazole tail fixed and studying minimum require-
Retaining the lipophilic anchor of the lead 2, the total
separation from the distal acid residue, as well as struc-
ture and topology of the connecting chain were varied
to study effects on affinity and selectivity among the
three PPAR receptors. The position of the acid
proximal phenyl residue of the lead was shifted both in
terms of its position in the linking tether and regio-iso-
merism. Synthesis of some benzisoxazole fragments and
*Corresponding author. Fax: +1-732-594-9556; e-mail: alan_adams@
merck.com
0960-894X/03/$ - see front matter # 2003 Elsevier Science Ltd. All rights reserved.
doi:10.1016/S0960-894X(02)01029-6

932
A. D. Adams et al. / Bioorg. Med. Chem. Lett. 13 (2003) 931–935
d and a receptor subtypes, with generally better intrinsic
potency demonstrated in the PPARg transfectant,
resulting in apparent functional selectivity superior to
the affinity selectivity (Fig. 3).
A second set of analogues retaining a similar range of
overall lengths was prepared based on alkyloxy-linked
dihydrocinnamic and cinnamic acids. A fairly sharp and
slightly greater magnitude (ÁÁG ꢀ1.5 kcal7 for 15 vs
23) affinity maximum for the PPARg receptor was
observed but now favoring the para isomer. The cinna-
mate analogues 27 and 28 showed relatively small
penalties in PPARg and a affinity for the restriction of
rotation in the acid proximal chain. The para-cinnamate
isomer 28 shows, in fact, the best PPARg/a dual agonist
selectivity of the analogues studied here based on affinity.
Both a,a- and b,b-dimethylation in the dihydrocinnamate
proved to be most effective in generating PPARg/a recep-
tor subtype selectivity. These result in 10- to 15-fold selec-
tivity for PPARg/a over PPARd with good functional
selectivity in the GAL4-PPAR transactivation assay.
Figure 2.
elaboration to the typical carboxylate products have
been previously reported.5,6
The synthesis of the 6-hydroxy-3-phenylbenzisoxazole
proceeds from the commercially available allyl ether as
illustrated in Figure 2 for example 9. Alkylation of this
benzisoxazole with the appropriate dibromide followed
by O-alkylation of a phenolic acid and subsequent ester
cleavage led to the analogues reported below. The
required phenolic acid fragments were prepared by
known methods. These various substrates include ben-
zoic, phenylacetic, phenylpropionic, phenoxyacetate,
cinnamic and phenylbutyric acid residues. Reasonably
potent PPAR ligands could be found in all series with
widely varying selectivity for the three PPAR subtypes.
Results are reported in Table 1.
Incorporation of one of the best known PPAR agonist
acid residues, a phenoxyacetate,11 also yielded potent
agonists on all three receptors. Some modest 7- to 9-fold
selectivity is evident for PPARg/a over PPARd affinity
in analogues 20 and 21. Functional selectivity found in
this series is similar to that for the a,a- and b,b-dimethyl
analogues above.
The tolerance for changes in the carbon skeleton prox-
imal to the acid residue is limited as demonstrated by
the 4-phenylbutyrate analogues 29 and 30. These com-
pounds, with a rigid aryl element nearer the center of
the molecule, showed substantially poorer affinity for
PPARg and PPARd and begin to show some
functional selectivity for PPARa.
The initial set of analogues 5–12 retained the lead
structure phenyl acetic acid terminus with the metaboli-
cally vulnerable sulfur replaced by oxygen. A fairly
sharp affinity maximum was found for compound 9
with the m-butyloxy link to the benzisoxazole fragment.
The apparently isosteric replacement of the sulfur in this
linking residue with a methylene oxy unit, yielding a
four carbon linking chain, is consistent with the many
known examples of substitution of sulfur for two car-
bon units.10 A substantial sensitivity to regioisomeric
substitution was observed. The difference in affinity for
the PPARg receptor implies an apparent ÁÁG of
approximately ꢀ1.3 kcal7 comparing 9 with meta sub-
stitution versus 6 with para. While 9 is modestly more
potent than 2 it is still nonselective. Some indication of
PPAR g/a selectivity over PPARd was seen in the para
analogue 6, but much more credible selectivity results
from alpha methylation as in 7 and 10.
To verify whether binding affinity in this SAR series
is still driven by the chosen anchor residue, the mag-
nitude of the 7-propyl-3-phenylbenzisoxazole fragment
contribution to the binding energy for the optimized
analogue 9 was assessed. Simplification of the distal
heteroaryl to 2-propylphenyl by the complete
removal of the fused 5-phenylisoxazole ring resulted
in titratable affinity for only the PPARd (Ki 830 nM)
and PPARg (Kiꢁ5000 nM) subtypes. Further trun-
cation of the distal heteroaryl to unsubstituted phenyl
resulted in loss of titratable affinity for all PPAR
subtypes. The fused phenylisoxazole ring accounts
for some 4.6 kcal of apparent binding energy on PPARd
and >5.5 kcal on PPARg. Given a contribution of
this magnitude, it is reasonable to assume that
these compounds have in common a binding mode
largely determined by the distal heteroaryl binding
affinity.12
Agonist competence was assayed in the COS cell trans-
fected with a GAL 4-PPAR ligand binding domain chi-
mer expression vector and 5X-UAS-luciferase reporter
plasmid as described in Berger et al.9 Most of the ana-
logues in these series were competent full agonists in this
GAL4-PPAR transactivation assay. Where the titration
curves cannot be fit to generate an EC50, maximum
observed activation at the highest nontoxic dose is
reported. Apparent intrinsic potency varies within the g,
One obvious avenue open to affect selectivity in this
series is offered by the introduction of rigidifying unsa-
turation into the center of the four methylene spacer of the
highest affinity analogues. Both 20,30 olefin isomers and a
20,30 acetylenic analogue of 9 were prepared to investigate
this question. Retention of PPARg and PPARd binding is

A. D. Adams et al. / Bioorg. Med. Chem. Lett. 13 (2003) 931–935
Table 1. PPAR binding, Ki: agonist efficacy in the transfected COS Cell, EC50
933
Isomer
Tether
length
PPAR binding apparent Ki nMa
Transfected COS-1 cell EC50 nM or
Max% at 3 mM b
n
g
d
a
g
d
a
Rosiglitazone
136
11
>33,300
>33,300
21
4
0%
3
3%
3
2
Lead Phenylbenzisoxazole
para
3
3.7
14.0
A—— Acid proximal structure
3
4
–COOH
–COOH
para
meta
4
4
>5300
100.0
>14,000
508.0
920.0
5.2
ND
193
ND
34%
ND
8
5
6
7
–CH2COOH
–CH2COOH
–C(CH3)2COOH
para
para
para
3
4
4
32
30
17
5
93
203
3
7
23
194
60
10
147
86%
59%
3
8
2
8
9
10
–CH2COOH
–CH2COOH
–C(CH3)2COOH
meta
meta
meta
3
4
4
662
3.1
26
26
0.3
165
76
7.26
46
ND
ND
400
ND
50
0
2 5
52
11
12
–CH2COOH
–CH2COOH
ortho
ortho
3
4
498
348
62
47
489
86
ND
ND
ND
ND
ND
ND
13
14
15
16
17
–CH2CH2COOH
–CH2CH2COOH
–CH2CH2COOH
–C(CH3)2CH2COOH
–CH2C(CH3)2COOH
para
para
para
para
para
2
3
4
4
4
304
32
8.9
52
11
2
37.5
529
317
11
7.3
7.6
213
13
ND
99%
485
37
ND
74%
74%
18%
860
ND
42%
429
53
20
20
4
22
23
–CH2CH2COOH
–CH2CH2COOH
meta
meta
3
4
448
115
21
52
53
329
ND
ND
ND
ND
ND
ND
26
–CH2CH2COOH
ortho
3
1,330
3,520
>33,330
ND
ND
ND
27
28
–CH=CHCOOH
–CH=CHCOOH
para
para
3
4
152
13
18
265
50
18
254
29
148
640
56%
13
18
19
20
21
–O–CH2COOH
–O–CH(CH3)COOH
–O–CH(CH3)COOH
–O–CH(CH3)2COOH
para
para
para
para
3
3
4
4
34
22
20
2.4
3.7
4.6
136
23
8.1
9.3
47
75
27
16
2
16
13
117%
172
4
2
10
2
17
24
25
–O–CH2COOH
–O–CH(CH3)2COOH
meta
meta
4
4
52
41
47
110
133
255
87
100
127%
85%
248
92
29
30
–CH2CH2CH2COOH
-CH2CH2CH2COOH
para
para
2
3
616
145
156
38
305
121
72%
156
69%
183
33
3
ND, Not run.
aThe apparent Ki was calculated from the Cheng–Prusoff equation7 using the IC50 determined in the previously described SPA binding assay at
15 ꢂC.8 The radioligand for the PPARa and g determination was tritiated 9.
bThe EC50 refers to the concentration yielding a 50% response relative to the standard in the previously described PPAR GAL4 chimer COS-1 cell
transactivation assay.9
highest in the same cis olefin conformation offering little
encouragement for further analogues. All three analo-
gues retain very good PPARd affinity.5 The remarkable
tolerance toward ligand structure observed for the
PPARd receptor here and above may be explained by
X-ray structure observations of ligands bound in multi-
ple conformations in the PPARd binding pocket.13
mM-Hr for a 2 mpk po dose of the sodium salt in rats.14
The isomeric para phenylacetic acid analogue 6, eval-
uated in a high throughput nonquantitative PK screen,
showed similar behavior.15 Both were chosen as good
candidates for efficacy studies. Several analogues
including the most selective, b,b-dimethyl analogue 16,
and the cinnamate analogue 28 could not be evaluated
in the db/db mouse model due to very poor oral bioa-
vailability. Efficacy studies for 6 and 9 were performed
in 7–8 week old highly insulin resistant leptin receptor
defective db/db mice as described by Berger et al.9 Effi-
cacy was assessed relative to Rosiglitazone which shows
variable (40–67%) correction of hyperglycemia (degree
Pharmacokinetics and Efficacy Studies
Compound 9 showed promising PK parameters with
76% bioavailability and dose normalized AUC of 4.2

934
A. D. Adams et al. / Bioorg. Med. Chem. Lett. 13 (2003) 931–935
assay.16 There is evidence to indicate that the added
human PPARa mediated potency would enhance effi-
cacy in humans.17 Complete receptor binding selectivity
over PPARd was not obtained in this series, and
remains the most challenging problem.
Acknowledgements
We would like to thank James Pivnichny, Kwan Leung
and Raul Alvaro for pharmacokinetic support on this
project. Additional technical suport for biological eval-
uation was provided by; Margaret Wu, John Ventre,
Roger Meurer, Chhabi Biswas and Neelam Sharma.
References and Notes
1. An excellent review; Willson, T. M.; Brown, P. J.; Stern-
bach, D. D.; Henke, B. R. J. Med. Chem. 2000, 43, 527. An
update in 2001 including a list of PPAR agonists in develop-
ment for diabetes; Sorbera, L. A.; Leeson, L.; Martin, L.;
Castaner, J. Drugs Future 2001, 26, 354.
2. Recent examples of PPARg/a mixed agonists; Nomura,
M.; Kinoshita, S.; Satoh, H.; Maeda, T.; Murakami, K.; Tsu-
noda, M.; Miyachi, H.; Awano, K. Bioorg. Med. Chem. Lett.
1999, 9, 533 (KRP-297). Lohray, B. B.; Lohray, V. B.; Bajji,
A. C.; Kalchar, S.; Poondra, R. R.; Padakanti, S.; Chakra-
barti, R.; Vikramadithyan, R. K.; Misra, P.; Juluri, S.;
Mamidi, N. V.; Rajagopalan, R. J. Med. Chem. 2001, 44, 2675
(NN-622). Brooks, D. A.; Etgen, G. J.; Rito, C. J.; Shuker,
A. J.; Dominianni, S. J.; Warshawsky, A. M.; Ardecky, R.;
Paterniti, J. R.; Tyhonas, J.; Karanewsky, D. S.; Kauffman,
R. F.; Broderick, C. L.; Oldham, B. A.; Montrose-Rafizadeh,
C.; Winneroski, L. L.; Faul, M. M.; McCarthy, J. R. J. Med.
Chem. 2001, 44, 2061.
3. Ref 1 and; Guay, D. R Ann-Pharmacother. 1999, 33, 1083.
Haim, M.; Benderly, M.; Brunner, D.; Behar, S.; Graff, E.;
Reicher-Reiss, H.; Goldbourt, U. Circulation 1999, 100, 475.
4. Wagenaar, L. J.; Kuck, E. M.; Hoekstra, J. B. Neth-J-Med.
1999, 55, 4. Misbin, R. I. Ann-Intern-Med. 1999, 130, 330.
Herrine, S. K.; Choudhary, C. Ann-Intern-Med. 1999, 130,
163. Rosiglitazone appears to be safer, but see reports of liver
toxicity with Rosiglitazone; Forman, L. M.; Simmons, D. A.;
Diamond, R. H. Annals of Internal Medicine 2000, 132, 118.
Tolman, K. G. Intl. J. Med. Practice. Suppl. 2000, 113, 29.
5. Jones, A. B. Med. Res. Rev. 2001, 21, 540 Note that this
publication reports IC50 as opposed to Ki.
6. All new compounds gave consistent 400 MHz 1H NMR
spectra and satisfactory LCMS data.
7. Cheng, Y.-C.; Prusoff, W. H. Biochem. Pharmacol. 1973,
22, 3099 ÁG Values are calculated from Ki ratios using the
van’t Hoff reaction isotherm.
8. Human PPARg2, PPARa and PPARd receptors were
expressed as GST-fusion proteins in E. coli. The full length
human cDNAs for PPARd (provided by Dr. Azriel Schmidt,
MRL) and PPARa (provided by Dr. Tom Rushmore, MRL)
were subcloned into pGEX-KT expression vectors (Pharma-
cia), whereas pGEX-hPPARg2 was constructed as described
by Elbrecht (JBC 1999, 274, 7913). Expression and purifica-
tion of GST-PPAR proteins, and establishment of a Scintilla-
tion Proximity Assay (SPA)-based receptor binding assays was
similar that described by Elbrecht. For both PPARa and
PPARg, 5 nM of [3H2]L-797773 (specific activity of 34.3 Ci/
mmol) was used, and for PPARd 2.5 nM of [3H2]L-783483
(specific activity of 13.4 Ci/mmol) was used. Results are
Figure 3.
of mean glucose lowering as percentage of the difference
between vehicle treated db/db mice vs lean control mice)
at a dose of 10 mpk for 11 days. The partially optimized
lead compound 2 dosed at 30 mpk showed approxi-
mately 80% of the correction observed with Rosiglita-
zone dosed at 30 mpk. Both 6 and 9, dosed at 10 mpk/
day po, showed 110% of the observed Rosiglitazone
10 mpk correction for a significant improvement in in
vivo potency over the lead compound 2.
The fine adjustments of structure between the anchor
benzisoxazole and the carboxylate residue generated
high affinity PPARg agonists possessing binding affinity
surpassing the initial lead 2 on PPARg in series with 6
atom spacers with any of the original phenylacetic,
dihydrocinnamic or fibric acid residues. Functional and
binding selectivity was most effectively influenced by
introduction of either a-methyl substitution or b-sub-
stitution. Good oral bioavailability and efficacy equal or
superior to the benchmark, Rosiglitazone, can be
obtained in this series. The most potent analogues in
this series showed good antihyperglycemic efficacy at 10
mpk/day orally. The para substituted compounds
showing the highest PPARg/a selectivity in this series
showed a consistent trend toward poor bioavailability
which precluded efficacy testing. The observed in vivo
potency for this series in the mouse reflects only the
PPARg agonist potency as the series showed poor or no
agonist activity on the mouse PPARa receptor in a
PPAR homogenous time resolved fluorescence (HTRF)

A. D. Adams et al. / Bioorg. Med. Chem. Lett. 13 (2003) 931–935
935
expressed as inflection points calculated by a four-parameter
logistic equation. Kis are calculated by the equation of Cheng
and Prusofff, ref 7. Typical precision is represented by Rosi-
glitazone or compound 9. Four independent determinations
for rosiglitazone with PPARg, in duplicate, yield an average Ki
of 136 nM, Stdev 46, SEM 23. More than 400 determinations
of compound 9 with PPARg, in duplicate, yield an average Ki
of 3.1 nM, Stdev 1.1. For compound 9 on PPARa, Ki 7.2 nM,
Stdev 2.7.
9. Reported EC50s are determined from titrations in triplicate.
Repeat titrations for Rosiglitazone with PPARg represent
typical precision. For 5 determinations, in triplicate, mean
EC50 is 21 nM stdev 5.6 SEM 2.5. Berger, J.; Leibowitz, M.;
Doebber, T. W.; Elbrecht, A.; Zhang, B.; Zhou, G.; Biswas,
C.; Cullinan, C. A.; Hayes, N. S.; Li, Y.; Tannen, M.; Ventre,
J.; Wu, M.; Berger, G. D.; Mosley, R.; Marquis, R.; Santini,
C.; Sahoo, S. P.; Tolman, R.; Smith, R. G.; Moller, D. E. J.
Biol. Chem. 1999, 274, 6718.
10. Sulfur is well known as a two carbon equivalent in ring
structures, saturated and aromatic. This is simply an acyclic
example. Burger, A. in Progress in Drug Research/Fortschritte
der Arzneimittelforschung; Jucker, E., Editor; Birkhauser:
Basel, 1991; Vol 37, p 287. See page 311.
11. The eponymous herbicide peroxisome proliferators are
2,4-D or (2,4-dichlorophenoxy)acetic acid and its congeners,
2,4,5-T and MCPA (4-Chloro-2-methylphenoxy)acetic acid.
Vainio, H.; Nickels, J.; Linnainmaa, K. Scand J. Work,
Environ. Health 1982, 8, 70.
required for agonist activity. For example, the nitriles corre-
sponding to 6 and 15 show no titratable affinity for any of the
receptors. The carboxy proximal fragment is not, however,
sufficient for measurable affinity. Phenyl acetic acid, 3-phenyl-
propionic acid and clofibric acid also show no measurable
affinity for any of the receptors. Exactly this paradox of attri-
bution and additivity of binding energy is discussed in Jencks,
W. P. Proc. Natl. Acad. Sci. USA 1981, 78, 4046.
13. Xu, H. E.; Lambert, M. H.; Montana, V. G.; Parks, D. J.;
Blanchard, S. G.; Brown, P. J.; Sternbach, D. G.; Lehmann,
J. M.; Wisely, G. B.; Willson, T. M.; Kliewer, S. A.; Milburn,
M. V. Mol. Cell 1999, 3, 397.
14. Pharmacokinetics were evaluated in male Sprague Dawley
rats with 0.5 mpk iv and 2.0 mpk po doses.
15. Pivichny et al. Unpublished.
16. Compounds were evaluated in an HTRF based assay
using the murine PPARa construct described in Zhou, G.;
Cummings, R.; Li, Y.; Mitra, S.; Wilkinson, H. A.; Elbrecht,
A.; Hermes, J. D.; Schaeffer, J. M.; Smith, R. G.; Moller, D. E.
Mol. Endo. 1998, 12, 1594.
17. See refs 118–120 in Willson et al. (ref 1) Many studies
show improvement of glucose homeostasis using Bezafibrate in
humans; Smud, R.; Sermukslis, B. Curr. Med. Res. Opinion 1987,
10, 612. Mikhailidis, D.; Mathur, S.; Barradas, M. A.; Dan-
dona, P. J Cardiovasc Pharm 1990, 16, S26. Durrington, P. N.;
Winocour, P. H.; Bhatnagar, D. J. Cardiovasc. Pharmacol.
1990, 16, S30. Attia, N.; Durlach, V.; Roche, D.; Paul, J. L.;
Soni, T.; Zahouani, A.; Landron, F.; Labrousse, F.; Leute-
negger, M.; Girard-Globa, A. Eur. J. Clin. Invest. 1997, 27, 55.
12. The carboxy residue present in all agonists reported here is