Orally active and brain permeable proline amides as highly selective 5HT2c agonists for the treatment of obesity

Article abstract of DOI:10.1016/j.bmcl.2010.01.107

Brain-penetrable proline amides were developed as 5HT2c agonists with more than 1000-fold binding selectivity against 5HT2b receptor. After medicinal chemistry optimization and SAR studies, orally active proline amides with robust efficacy in a rodent foo

Full text of DOI:10.1016/j.bmcl.2010.01.107

Bioorganic & Medicinal Chemistry Letters 20 (2010) 2365–2369
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry Letters
journal homepage: www.elsevier.com/locate/bmcl
Orally active and brain permeable proline amides as highly selective 5HT2c
agonists for the treatment of obesity
b
b
b
b
Kevin K.-C. Liu a, , Bruce A. Lefker , Mark A. Dombroski , Phoebe Chiang , Peter Cornelius ,
Terrell A. Patterson b, Yuan Zeng b, Stephanie Santucci b, Elizabeth Tomlinson b, Colleen P. Gibbons b,
Ravi Marala b, Janice A. Brown b, Jimmy X. Kong b, Eunsun Lee b, Wendy Werner b, Zane Wenzel b,
Craig Giragossian b, Hou Chen b, Steven B. Coffey b
*
a Pfizer Global Research and Development, Chemistry Department, 10770 Science Center Drive, La Jolla, CA 92120, United States
b Pfizer Global Research and Development, Groton Laboratories, 558 Eastern Point Road, Groton, CT 06340, United States
a r t i c l e i n f o
a b s t r a c t
Article history:
Brain-penetrable proline amides were developed as 5HT2c agonists with more than 1000-fold binding
selectivity against 5HT2b receptor. After medicinal chemistry optimization and SAR studies, orally active
proline amides with robust efficacy in a rodent food intake inhibition model were uncovered.
Ó 2010 Published by Elsevier Ltd.
Received 10 December 2009
Revised 19 January 2010
Accepted 20 January 2010
Available online 25 January 2010
Keywords:
5HT2c agonist
Obesity
Binding selectivity
The GPCR 5HT2c sub-type of serotonin receptors has received
considerable interest as a therapeutic target for the treatment of
a wide variety of conditions including obesity, anxiety, depression,
obsessive compulsive disorder, schizophrenia, migraine, and erec-
tile dysfunction.1 The 5HT2c receptor has been implicated in the
regulation of body weight in both rodents and humans. 5HT2c
knockout (KO) mice are hyperphagic and gain excess weight (pre-
dominantly adipose tissue) compared to wild-type (WT) animals.2
These 5HT2c knockout mice are resistant to the anorectic effects of
mCPP (a non-selective 5HT2c receptor agonist, 1) and partially
resistant to the anorectic effects of dexfenfluramine (a 5HT-reup-
take inhibitor and releaser). In humans, these and other serotoner-
gic drugs inhibit food intake and promote weight loss without
losing lean body mass. For example, the non-selective 5HT2c ago-
nists, mCPP and PNU-22394 (legacy Pharmacia-Upjohn, 2a), have
both caused significant weight loss in short-term clinical trials
(14–24 days).3 In addition, dexfenfluramine (ReduxÒ, American
Home Products; AHP) produced sustained body weight reduction
of 10% after 1 yr in a subset of patients. Similarly, a 5HT2C selective
agonist (i.e., BVT933, Biovitrum) was reported to cause body
weight reduction (2.2 kg in 4 wks) in a Phase II study.4 Recently,
another 5HT2c agonist, APD-356 (Lorcaserin (2b), Arena), has been
shown to inhibit food intake, reduce body weight, and is in phase
III clinical trials.5
N
NH
N
Cl
Cl
N
N
H C
3
mCPP (1)
PNU22394 (2a)
APD-356 (Lorcaserin, 2b)
To find a safe 5HT2c agonist for the treatment of obesity, the key
hurdle is the selectivity against 5HT2a and 5HT2b receptors. Acti-
vation of 5HT2a receptors in humans has been associated with hal-
lucinations as the notorious effects of lysergic acid diethylamide
(LSD).6 Since the withdrawal of dexfenfluramine and fenfluramine
due to increased incidence of valvular heart disease (VHD), a signif-
icant body of evidence has accumulated that links the activation of
5HT2b receptor with this pathology.7 However, there are currently
no known animal models of drug-induced VHD or validated mod-
els of drug-induced pulmonary hypertension.8 But, given the po-
tential safety concerns around the pharmacology of 5HT2b
receptor, we were seeking 5HT2c agonists with more than 1000-
fold binding selectivity over 5HT2b receptor and without 5HT2a
agonistic activity. We decided that the best approach would be
to prioritize compound series by 5HT2c versus 5HT2b binding
selectivity. It was our belief that it would be easier to optimize
5HT2c potency than sub-type selectivity based on our experience
and literature data in this area.
* Corresponding authors.
E-mail addresses: kevin.k.liu@pfizer.com (K.K.-C. Liu), dombroskima@mac.com
(M.A. Dombroski), phoebe.y.chiang@gmail.com (P. Chiang).
0960-894X/$ - see front matter Ó 2010 Published by Elsevier Ltd.
doi:10.1016/j.bmcl.2010.01.107

2366
K. K.-C. Liu et al. / Bioorg. Med. Chem. Lett. 20 (2010) 2365–2369
To achieve this goal, we identified 3a9 and 3b as lead structures
We hypothesized that potent and selective 5HT2c agonists could
be identified by replacing the urea or carbamates of 3a–c with
amides, while improving brain penetration. To this end, we de-
signed series of libraries. Our initial library was designed to iden-
tify novel groups which could mimic the basic amine found in all
known 5HT2c agonists as in compounds 1–2. We selected a basis
set of four amines (R1), and coupled these with a diverse set of
88 Boc-protected amino acids, followed by de-protection of the
Boc group (Scheme 1). Although we could have chosen to direct
our library towards products which would mimic the distance
and orientation of the basic amines of 3a/3b relative to the linking
carbonyl group, we chose to explore a broad range of chemotypes.
Measureable 5HT2c binding was detected with a number of differ-
ent amides from this first library. Of the highest interest were the
2- and 3-pyrrolidine analogs 3d–f with good ligand efficiency
(LE).10
with significant binding selectivity versus the 5HT2b receptor.
However, both of these leads suffered from significant in vivo brain
impairment and efflux in a MDR expressing cell line. Interestingly,
we also found that truncation of the isoxazole tail of 3a led to 3c
with similar 5HT2c potency.
N
O
F
F
O
NH
N
O
O
N
Cl
O
F
3a
N
NH
3b
5HT2c Ki = 4nM
5HT2b Ki = 3500 nM
5HT2c Ki = 43 nM
5HT2b Ki >10,000nM
O
O
N
To our delight, compound 3d showed no evidence of efflux in
our MDR efflux assay. We were also attracted to 3d over 3e due
to its lower pKa, and expected lower risk of phosholipidosis which
has been shown to correlate with increased basicity and lipophilic-
ity.11 In addition, compounds with reduced basicity are less
N
3c
5HT2c Ki = 83 nM
5HT2b Ki - not tested
Scheme 1. Library design strategy to identify leads 3d–f. (NT: not tested).
O
H
R2
R2
O
N
N
H
N
+
N
HO
R1
R1
O
O
Cl
F
O
N
N
O
O
O
7
Cl
N
N
N
O
O
N
N
H
N
H
4
5
6
Compound number
5HT2c Ki (nM)
5HT2b Ki (nM)
5ΗΤ2cΕC50 (nM)
5HT2c%max
4
5
6
7
4.4
4120
88
8
23
20
ND
183
65
2400
38
46
1720
73
40
53
Scheme 2. Library follow-up on lead compound 3d, and some representatives 5HT2c agonists from the library.

K. K.-C. Liu et al. / Bioorg. Med. Chem. Lett. 20 (2010) 2365–2369
2367
susceptible to hERG inhibition as well.12 With promising 5HT2b
binding selectivity, modest potency, good LE and lower risk of
brain impairment with 3d, we next designed a series of libraries
to optimize the tail (amine) region of the proline amide series,
while keeping the proline amide region constant (Scheme 2 and
Fig. 1a and b). Our library design goals for this round were to ex-
plore a broad range of topologies in different amines, increase po-
tency, maximize selectivity, and minimize efflux as measured in
our MDR assay.
None of the secondary amines that we synthesized showed sig-
nificant binding to the 5HT2c receptor when they were screened at
500 nM (Fig. 1a). However, the primary amines in this library pro-
duced several single-digit nanomolar 5HT2c binders with 5HT2c
agonistic activity. The majority of actives were benzylic primary
amines with a para-substituent (Scheme 2 and Fig. 1a). This tail-
modified amine library was quickly followed up by another more
focused library based on the actives from the previous library
(Fig. 1b vs a). This follow-up library expanded the active chemical
space, and improved 5HT2c potency as well. Compared with origi-
nal lead compound 3d, we identified compounds that are more
than 60-fold more potent at 5HT2c binding, while maintaining
>1000-fold selectivity vs. the 5HT2b receptor.
However, as shown in Scheme 2, compounds 4–7 are partial
agonists in our functional 5HT2c assay (5HT2c %max). We also
found that a number of these compounds are monoamine oxidase
(MAO) inhibitors. Inhibition MAO of a 5HT2c agonist will have
some potential concerns on either serotonin syndrome13 or
5HT2c receptor de-sensitization since MAO is the primary enzyme
responsible for serotonin metabolism in human. Subsequent SAR
studies, determined that modifications of the central phenyl ring
increased 5HT2c agonism, and concomitantly reduced these com-
pounds’ MAO inhibition as well. The SAR also revealed that the
lipophilic tail alkoxyl group led to good 5HT2c/2b binding selectiv-
ity, and the SAR around the head proline piece is very tight as illus-
trated in Scheme 3. The corresponding C-linked tail piece
compounds were also made but these compounds are less stable
in human liver microsomes.
Figure 1. (a) Amines from the initial proline amide library were clustered and
plotted versus 5HT2c IC50 (nM). (b) Additional follow-up around the active chemical
space provided improved potency and expansion of the active chemical space.
With this SAR knowledge, our focus was then turned to increas-
ing 5HT2c agonism while keeping 5HT2b binding away and
2b/2c selectivity
lipophilic pocket
no polar groups
F
issue: MAO inhibitor
O
R
O
N
H
O
Increase 5HT2c agonism
R
O
(tail)
H
N
N
very tight SAR
no substitutent allowed
H
N
(head)
modification core structure
F (OMe)
F
O
R
R
O
O
N
N
F
F
C-linked compounds
not stable in HLM
H N
H N
not MAO inhibitor
good 2c potency
Scheme 3. SAR summary of proline amides.

2368
K. K.-C. Liu et al. / Bioorg. Med. Chem. Lett. 20 (2010) 2365–2369
reducing 5HT2a agonistic activity. Analogs were then made to fill
the matrix of combination of promising tail alkoxy groups and
central phenyl rings. Four compounds are highlighted in Table 1.
from this exercise. Compound 8 was chosen for further characteri-
zation because of its fine balance between 5HT2c, 2a, and 2b
binding potency and agonistic activity. As shown in Table 1, com-
pound 8 has the best 5HT2c binding but with minimum 5HT2a and
5HT2b agonistic activity.14 Noteworthy, compound 8 has 5HT2b/2c
binding selectivity about 10,000-fold.
in 87% yield which is subsequently reduced with sodium borohy-
dride to give 4,4-diemthylcyclohexanol (14) as the tail piece of
compound 8. Compound 14 is then introduced selectively to
2,4,5-trifluorophenylnitrile (15) under the basic condition with
potassium tert-butoxide at low temperature to give the region-
selective para-substituted compound 16 which is subjected to an-
other round of nucleophilic substitution with potassium tert-
butoxide, methanol in THF at low temperature to give methoxy
compound 17 in almost quantitative yield. The nitrile of 17 is re-
duced under a hydrogenation condition with Pd/C in acidic ethanol
solution to give benzyl amine, 18. Compound 18 is then coupled
A concise synthesis of compound 8 was developed as shown in
Scheme 4 and is described below. Dimethylcyclohexenone (12) is
hydrogenated with 10% Pd/C in ethyl acetetae to give ketone 13
with Boc-L-proline with carbonyldiimidazine (CDI) to bring the
Table 1
SAR of proline amides and data for compounds 8–11
O
O
H
H
N
H3C
H3C
H3C
F
F
N
N
H
N
H
OCH3
O
O
F
8
10
O
O
H
N
H
N
H3C
H3C
F
F
N
H
N
H
F
O
OCH3
O
9
11
Compound
8
9
10
11
Binding Ki, nM
5HT2C
5HT2B
5HT2A
0.9
3940
12.8
1.8
1670
13.7
25
NA
NA
1.7
NA
NA
Functional EC50, nM (% max)
5HT2C
5HT2B
5HT2A
2.3 (95–100)
>10,000
440 (76)
1 (88)
>10,000
130 (98)
44 (46)
>10,000
153 (99)
4.97 (85)
4027 (20)
500 (67)
H2, 10% Pd/C
EtOAc
H3C
H3C
H3C
H3C
H3C
H3C
NaBH4, CH3OH
88%
O
O
OH
87%
12
13
14
H3C
H3C
OH
H3C
F
H3C
H3C
F
F
CN
F
KOtBu, CH3OH
CN
F
F
CN
H2, 10% Pd/C
H3C
KOtBu, THF
-30 ºC
EtOH, HCl
75%
THF, -20 ºC
99%
O
O
OCH3
15
65%
16
17
O
CH3
CH3
CH3
O
Boc-L-Proline
CDI, 2-MeTHF
O
H3C
H3C
F
H3C
H3C
HCl, 2-MeTHF
98%
NH2
OCH3
F
N
N
H
O
99%
O
OCH3
18
19
O
H
N
H3C
H3C
F
N
H
O
OCH3
8
Scheme 4. Synthesis of compound 8.

K. K.-C. Liu et al. / Bioorg. Med. Chem. Lett. 20 (2010) 2365–2369
2369
Vehicle
100
25.0
22.5
20.0
17.5
15.0
12.5
10.0
7.5
10 mg/kg
20 mg/kg
30 mg/kg
60 mg/kg
80 mg/kg
10mg/kg
20mg/kg
30mg/kg
10
60mg/kg
80mg/kg
dose at t=0
1
5.0
2.5
0.0
dark Cycle
0.1
0
5
10
15
20
25
0
120 240 360 480 600 720 840
time (minutes)
hours
Scheme 5. Compound 8 in food intake model and its concentration-effect plot.
head piece of compound 8 to generate compound 19. Finally, com-
pound 19 is treated with HCl in 2-methyl THF to afford compound
8 in excellent yield.
References and notes
1. Hoyer, D.; Clarke, D. E.; Fozard, J. R. Pharmacol. Rev. 1994, 46, 157; Saxena, P. R.
Pharmacol. Rev. 1994, 66, 339.
In order to assess anorectic activity in the rat, a model of spon-
taneous nocturnal feeding was used.15 Rats are nocturnal feeders,
eating 80–90% of their daily intake during the dark phase. Com-
pound 8 showed robust efficacy in this food intake models. Fur-
thermore, dose–responsive inhibition of spontaneous nocturnal
food intake and concentration-effect correlation was also observed
as shown in Scheme 5.
2. Nonogaki, K.; Strack, A. M.; Dallman, M. F.; Tecott, L. H. Nat. Med. 1998, 4, 1152.
3. Sargent, P. A.; Sharpley, A. L.; Willians, C.; Goodall, E. M.; Cowen, P. J.
Psychopharmacology 1997, 133, 309.
4. Halford, J. C.; Harrold, J. A.; Lawton, C. L.; Blundell, J. E. Curr. Drug Targets 2005,
6, 201.
5. (a) Smith, S. R.; Prosser, W. A.; Donahue, D.; Anderson, C.; Shanahan, W. 66th
Annu. Meet. Sci. Sess. Am. Diabetes. Assoc. (ADA): June 9–13, Washington, DC,
2006; Abst 344-OR.; (b) Smith, S. R.; Prosser, W. A.; Donahue, D. J.; Morgan, M.
E.; Anderson, C. M.; Shanahan, W. R. Obesity 2009, 17, 594.
Compound 8 is highly bound to plasma proteins (0.036–0.067%,
free fraction), has minimal blood partitioning (0.7–1.3) has good
brain penetration (B/P total = 10, B/P free = 0.4–0.7), and low bio-
availability in dogs and rats (5.5–13%, in dog and rat, respectively).
6. Nichlos, D. E. Pharmacol. Ther. 2004, 101, 131.
7. Loke, Y. K.; Derry, S.; Pritchard-Copley, A. BMC Clin. Pharmacol. 2002, 2, 6.
8. Setola, V.; Roth, B. L. Methods Principles Med. Chem. 2008, 38, 143.
9. Adams, D. R.; Bentley, J. M.; Davidson, J. E. P; Dawson, C. E.; George, A. R.;
Mansell, H. L.; Mattei, P.; Mizrahi, J.; Nettekoven, M. H.; Pratt, R. M.; Roffey, J. R.
A.; Roever, S.; Specklin, J.-L.; Stalder, H.; Wilkinson, K. WO 2002048124, 2002.
10. Hopkins, A. L.; Groom, C. R.; Alex, A. Drug Discovery Today 2004, 9, 430.
11. Pelletier, D. J.; Gehlhaar, D.; Tilloy-Ellul, A.; Johnson, T. O.; Greene, N. J. Chem.
Inf. Model. 2007, 47, 1196.
12. Price, D. A.; Armour, D.; de Groot, M.; Leishman, D.; Napier, C.; Perros, M.;
Stammen, B. L.; Wood, A. Bioorg. Org. Chem. Lett. 2006, 16, 4633.
13. Sternbach, H. Am. J. Psychiatry 1991, 148, 705.
14. Compound 8 did not exhibit any 5HT2b agonist activity in an ex vivo rat
stomach fundus assay, and in a monkey heart valve interstitial cell (mVIC)
mitogen activated protein kinase (MAPK) functional assay. Therefore, the
potential risks related to activation of the 5HT2b receptor are deemed to be
significantly reduced with compound 8. Compound 8 did not show any 5HT2a
agonist activity as well in the above mentioned MAPK functional assay, and in
the rat head twitch in vivo studies.
Compound 8 has good HLM and hepatocytes stability with T1/2
=
100 min and 440 min, respectively.
In summary, our chemistry strategy to focus on 5HT2c agonist
hits with good 5HT2c/2b binding selectivity instead of 5HT2c po-
tency allowed us to find selective 5HT2c agonists with more than
1000-fold binding selectivity against 5HT2b receptor. In the end,
the goal was achieved with proline amides based on this strategy.
In addition, MAO inhibition was overcame and 5HT2c agonistic
activity was improved by modifying the phenyl core ring. 5HT2c
agonists with single-digit nanomolar in both 5HT2c binding and
functional assays were identified with no 5HT2b agonistic activity.
The orally active compound 8 also demonstrated robust efficacy in
our rodent food intake model. Additional results including more
pharmacology data with this compound will be reported in due
course.
15. For the details of our biological assays and the spontaneous food intake model,
please see: Chen, H.; Coffey, S. B.; Lefker, B. A.; Liu, K. K.-C. WO 2006103511,
2006.