N-Benzyl-indolo carboxylic acids: Design and synthesis of potent and selective adipocyte fatty-acid binding protein (A-FABP) inhibitors

Article abstract of DOI:10.1016/j.bmcl.2009.01.084

Small molecule inhibitors of adipocyte fatty-acid binding protein (A-FABP) have gained renewed interest following the recent publication of pharmacologically beneficial effects of such inhibitors. Despite the potential utility of selective A-FABP inhibito

Full text of DOI:10.1016/j.bmcl.2009.01.084

Bioorganic & Medicinal Chemistry Letters 19 (2009) 1745–1748
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry Letters
journal homepage: www.elsevier.com/locate/bmcl
N-Benzyl-indolo carboxylic acids: Design and synthesis of potent and
selective adipocyte fatty-acid binding protein (A-FABP) inhibitors
a
a
b
b
c
Tjeerd Barf a, , Fredrik Lehmann , Kristin Hammer , Saba Haile , Eva Axen , Carmen Medina ,
*
Jonas Uppenberg c, Stefan Svensson c, Lena Rondahl d, Thomas Lundbäck b
a Department of Medicinal Chemistry, Biovitrum AB, Research, 112 76 Stockholm, Sweden
b Department of Assay Development and Screening, Biovitrum AB, Research, 112 76 Stockholm, Sweden
c Department of Structural Chemistry, Biovitrum AB, Research, 112 76 Stockholm, Sweden
d Department of Biology, Biovitrum AB, Research, 112 76 Stockholm, Sweden
a r t i c l e i n f o
a b s t r a c t
Article history:
Small molecule inhibitors of adipocyte fatty-acid binding protein (A-FABP) have gained renewed interest
following the recent publication of pharmacologically beneficial effects of such inhibitors. Despite the
potential utility of selective A-FABP inhibitors within the fields of metabolic disease, inflammation and
atherosclerosis, there are few examples of useful A-FABP inhibitors in the public domain. Herein, we
describe the optimization of N-benzyl-tetrahydrocarbazole derivatives through the use of co-crystal
structure guided medicinal chemistry efforts. This led to the identification of a potent and selective class
of A-FABP inhibitors as illustrated by N-benzyl-hexahydrocyclohepta[b]indole 30.
Received 23 December 2008
Revised 22 January 2009
Accepted 23 January 2009
Available online 30 January 2009
Ó 2009 Elsevier Ltd. All rights reserved.
Fatty-acid binding proteins (FABPs) are tissue-specific, ꢀ15 kDa
cytoplasmic lipid chaperones, capable of binding and transporting
endogenous fatty acids from the cell surface to the various sites of
metabolism or storage.1 Consequently, the role of FABPs is directly
linked to lipid-mediated biology such as signalling pathways, traf-
ficking and membrane synthesis.2
the possibility to develop isoform selective compounds (see be-
low), the lipophilic and charged nature of the endogenous ligands
and how this translates to the drugability of the binding pocket.
Also, the high intracellular content of A-FABP, which may limit
the possibility to neutralise a sufficient amount of the target pro-
tein may pose a liability. The recent data illustrating the pharma-
cological impact of a potent A-FABP inhibitor on lipid-mediated
biology suggests that at least the latter obstacles can be overcome.6
Mice with a disruption in the gene encoding for the heart or
muscle variant H-FABP (or FABP3) were reported to suffer from
stress-intolerability, in a few cases leading to death. Hence, in
developing inhibitors for any FABP isoform, the potential need
for selectivity against H-FABP should be considered.10 The overall
sequence identity of human H-FABP and A-FABP is 65%, which is
the highest degree of homology among the known human FABPs.11
Comparison of the binding sites of these two FABPs reveals only a
handful of amino acid differences.
We have previously reported the structure–activity relation-
ships (SAR) around carbazole- and indole-based inhibitors of A-
FABP, leading to the discovery of submicromolar inhibitors.9 In
the present study we have continued to explore the SAR around
a novel tetrahydrocarbazole series of A-FABP ligands. This has been
achieved by structure-based drug design. Variation of substitution
patterns as well as exploration of new benzoic acid containing
scaffolds led to the identification of several ligands with nanomolar
potency.
Out of the nine known isoforms, adipocyte FABP (A-FABP,
FABP4 or aP2) is the predominantly present isoform in adipose tis-
sue and also highly expressed in macrophages. This seems to trans-
late in a prominent function in certain specific aspects of the
metabolic syndrome and cardiovascular disease.3 Disruption of
A-FABP in mice prevents development of diet-induced insulin
resistance and whole body, as well as macrophage-specific, dele-
tion of A-FABP leads to protection of atherosclerosis in apolipopro-
tein E-deficient mice.4,5 In line with these results, pharmacological
intervention using the apparently selective A-FABP inhibitor BMS-
309403 produced similar phenotypes as demonstrated in diabetic
and atherosclerotic mouse models.6 In human settings, reduced
A-FABP expression in adipose tissue, as a consequence of polymor-
phism at the FABP4 locus, is has been linked with a lower risk for
hypertriglyceridemia, type 2 diabetes and coronary heart disease.7
Collectively, these studies indicate that pharmacological modu-
lation of the fatty-acid binding function of A-FABP could be of ther-
apeutic benefit in disorders such as type
2 diabetes and
atherosclerosis. Despite this potential, literature is scarce with re-
gards to reports of small molecule inhibitors for this family of pro-
teins.8,9 The underlying reasons may include concerns related to
Pyrazole derivative 1 with an IC50 value of 1.1 lM was identi-
fied by means of high throughput screening. Although representing
a sufficiently good starting point for optimization, pyrazole 1
suffered from b-oxidation as a consequence of its fatty acid-like
* Corresponding author. Fax: +31 0 412662546.
E-mail address: tjeerd.barf@spcorp.com (T. Barf).
0960-894X/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2009.01.084

1746
T. Barf et al. / Bioorg. Med. Chem. Lett. 19 (2009) 1745–1748
character. Co-crystallization of 1 with human A-FABP revealed a
binding mode which suggested that ring-closure of the aliphatic
chain on the pyrazole was feasible (Fig. 1) and hence the tetrahyd-
rocarbazole 2a was synthesised.
The inhibitory activity of 2a for A-FABP was confirmed (IC50
0.59 lM), and this fueled a medicinal chemistry program with
the aim to optimize potency and selectivity for A-FABP. A co-crys-
tal structure of 2a with human A-FABP was then solved (Fig. 2) to
further guide the medicinal chemistry efforts.
HO
O
HO
O
a
[
]n
]n
+
[
[
]n
]n
N
O
NHNH2
H
3
b
MeO
O
HO
O
c
[
N
N
H
X-ray crystallographic analysis shows that compound 2a binds
at the same site as endogenous fatty acids and other reported A-
FABP inhibitors, clearly demonstrating the competitive nature of
this inhibitor (Fig. 2).12 The inhibitory activity is driven mainly
by the interaction of the carboxylic acid with Y128 and R126,
and several hydrophobic interactions.
2
4
Scheme 1. Reagents and conditions: (a) AcOH, reflux (17–76%); (b) MeOH, H2SO4,
reflux (70–98%); (c) i—Benzyl bromide, KOH, DMSO, rt; ii—THF/MeOH, aq. LiOH
(2 M) rt (5–80%).
Our initial objective in exploring the SAR of 2a was to challenge
the relevance of the position of the carboxylic acid. This was
accomplished by the Fisher indole synthesis using 2-, 3-, and 4-
benzoic acid hydrazines and cyclohexanone in refluxing acetic acid
(Scheme 1). Attempts to N-alkylate directly were unsuccessful,
since the carboxylic acid needed to be protected first as the corre-
sponding ester. However, the alkylation proceeded smoothly using
benzyl bromide and potassium hydroxide as the base in DMSO. Fi-
nally, hydrolysis of the esters with aqueous lithium hydroxide fur-
nished the desired final compounds.
All compounds were tested for their inhibitory activity using a
fluorescence polarization (FP) assay as described earlier.13 Based
on the results presented in Table 1, it is clear that the 4-position
of the tetrahydrocarbazole moiety is the most favourable position
for the carboxylic acid. Moving the acid to the 3-position still keeps
an acceptable affinity for the fatty-acid binding site of A-FABP, but
a further shift to the 2-position renders a completely inactive com-
pound. Subsequently, the required nature of the aliphatic ring was
Table 1
IC50 (lM) values of compounds 2 and 5 for human A-FABP
O
2.
[
]
n
N
N
HO 3.
4.
HO
O
2a-e
5
Compound
COOH
n
IC50 (l
M)a
2a
2b
2c
2d
2e
5
4
3
2
4
4
—
1
1
1
0
2
—
0.59
0.75
>25
0.97
0.59
3.6
a
Values are means of triplicate fluorescence polarization experiments performed
on the same dilution.
investigated by ring expansion (cycloheptyl) or contraction (cyclo-
pentyl), respectively. In addition, the 2,3-dimethylindole derivative
5 was prepared starting from butan-2-one utilizing the same
chemistry as in Scheme 1.
N
N
N
CO2H
CO2H
As shown in Table 1, the cyclohexane and cycloheptane deriva-
tives were equipotent, and significantly more potent then their
cyclopentane and dimethyl counterparts. Thus, the 2,3,4,9-tetrahy-
dro-1H-carbazole-8-carboxylic acid and 5,6,7,8,9,10-hexahydrocy-
clohepta[b]indole-4-carboxylic acid series were considered
interesting enough for further modifications. Hence these scaffolds
were reacted (using the conditions described above) with a variety
of benzyl and alkyl halides chosen to include sterically demanding
as well as electron rich or poor substituents.
1
2a
Figure 1. Structure-guided information led to the discovery of tetrahydrocarbazole
derivative 2a.
Data presented in Table 2 demonstrated that the potencies of
the initial unsubstituted benzyl derivatives 2a and 2e (IC50
0.59 lM) could be improved. When introducing substituents on
the benzyl ring we found that more active compounds were ob-
tained when the substituent was in the ortho-position, hence the
ortho-trifluoromethyl substituted 17 (IC50 < 0.4
tent than its meta- or para-substituted counterparts (18 with
1.7 M and 19 with 2.4 M, respectively). This trend (ortho-substi-
lM) was more po-
l
l
tution > meta >> para) is general and applies to both electron with-
drawing (14–19) or electron donating (11–13) groups, especially
with increasing size. For the para-position it appears that sterically
demanding groups are forbidden, regardless of the electronic prop-
erties, hence the small para-fluoro (16) or methyl (20) substituents
gave better inhibition than the trifluoromethyl (19) or methoxy
(13) groups. With respect to combinations it is clear that the 2-flu-
oro-6-trifluoromethyl (21) combination is not optimal. In fact, the
IC50 of compound 21 was at least four times greater than either of
the corresponding monosubstitutions. It is also noteworthy that
Figure 2. Co-crystal structure of 2a in human A-FABP (PDB code: 3FR2).

T. Barf et al. / Bioorg. Med. Chem. Lett. 19 (2009) 1745–1748
1747
Table 2
IC50 M) values and selectivity data of compounds 2, 8 and 10–30 for human A-FABP
and H-FABP
This strategy proved successful in that plasma protein binding
indeed had been reduced from >99.9% for 2a to 90% for 8, however,
this modification also rendered an inactive compound. The oxim
analogue 10 showed still reasonable potency, showing the poten-
tial of this approach.
(l
X
Alternatively, potential H-bonding interactions A-FABP can be
anticipated via the amino acid triad made up by S53, S55 and
T60. To this end, carboxamides were introduced on the ortho-
and meta-position of the benzyl group of the hexahydrocyclohep-
ta[b]indole-4-carboxylic acid derivatives.14 Regioisomeric primary
amides 29 and 30 were approximately equipotent in the FP assay
[
]
[ ]
n
n
N
N
N
HO
O
HO
O
HO
O
R
2, 11-30
8: X = O
10: X = NOH
22
(ꢀ 0:4
lM). Since the protein concentration is also about 400 nM
a
a
Compound
n
R
A-FABP IC50
(l
M)
H-FABP IC50 (lM)
in this assay, our ability to distinguish between the most potent
compounds was limited. For this reason, a more sensitive scintilla-
tion proximity assay (SPA) assay had to be developed.11
2a
2e
8
1
2
–
–
1
1
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
–
–
–
–
0.59
0.59
>25
3.88
–
–
–
–
>10
>10
0.68
5.57
–
–
–
–
>10
–
–
–
–
The non-radiolabeled derivative 31 had previously been identi-
fied as a potent inhibitor of A-FABP and H-FABP.8 Therefore, also
the selectivity for both FABPs could potentially be addressed using
tritium-labeled 31 in a SPA-bead setting. The tritium-labels were
introduced on the 3- and 7-position of the carbazole moiety via
the reduction of the di-bromo derivative (Fig. 3).
Selected compounds were compared in this new SPA-bead as-
say and some of the inhibitors displayed indeed much better
potencies for A-FABP (Table 3). The sensitivity improved dramati-
cally, and could discriminate between submicromolar inhibitors,
such as the N-propyl analogues 22a and 22e on the one hand,
and low nanomolar inhibitors on the other hand. IC50 values down
to 49 nM for A-FABP were observed, as demonstrated for ortho-CF3
analogue 26. Unsubstituted 2e was two fold less potent (96 nM),
and introduction of a carboxamide on the ortho-position resulted
in a similar inhibitory activity. While 26 could not discriminate be-
tween A-FABP (49 nM) and H-FABP (34 nM), meta-carboxamide
derivative 30 actually was very selective for A-FABP (69 nM) versus
10
11
12
13
14
15
16
17
18
19
20
21
22a
22e
23
24
25
26
27
28
29
30
3.5
2-OMe
3-OMe
4-OMe
2-F
3-F
4-F
2-CF3
3-CF3
4-CF3
4-Me
2-F, 6-CF3
–
–
2-F
3-F
4-F
2-CF3
3-CF3
4-CF3
2-CONH2
3-CONH2
0.60
0.55
2.68
<0.4
<0.4
0.83
<0.4
1.66
2.39
1.09
1.81
0.73
0.43
0.71
0.72
0.78
0.65
1.04
2.03
<0.4
0.45
>10
>10
<0.4
>10
>10
–
>10
H-FABP (>10 lM) and also E-FABP (data not shown).
a
Again, X-ray crystallography provided an explanation for the
differences observed in selectivity between 26 and 30, because
the two compounds bind differently to A-FABP (Fig. 4). The CF3
group of 26 is accommodated in a lipophilic region, whereas the
carboxamide is optimally positioned for H-bonding interactions
with a serine residues S53 and S55 and backbone carbonyl of
K58, causing the benzyl group to flip by 180°. Interestingly, the
S53 of A-FABP resembles T53 in H-FABP and a repositioning of
Values are means of triplicate fluorescence polarization experiments performed
on the same dilution.
the N-propyl derivatives 22a and 22e were quite active indicating
that the carboxylic acid per se seems more important for binding
than the lipophilic benzyl part.
Since all synthesised compounds had been highly lipophilic so
far, attempts were made to develop more hydrophilic compounds
in order to improve on physicochemical properties, such as solubil-
ity and serum albumin binding. We envisaged that a conserved
water molecule, proximal to the 4-position of the carbazole, was
a suitable H-bonding partner or could even be replaced. To achieve
this, a keto- and oxim- functionality was introduced on this posi-
tion by selective oxidation of the methyl ester of 2a by DDQ
(Scheme 2). The acquired product was either hydrolysed, giving ke-
tone derivative 8 or reacted with hydroxylamine, followed by
hydrolysis to give oxim analogue 10.
O
N
OH
31
Figure 3. Structure and 3H-labeling position of 31.
Table 3
OH
IC50 (lM) values for selected compounds for human A-FABP as determined in a SPA-
bead assay
O
N
a
a
a
b
Compound
n
R
A-FABP IC50
(lM)
H-FABP IC50 (lM)
N
N
N
2e
22a
22e
26
29
30
2
1
2
2
2
2
–
–
–
0.096
0.67
0.58
0.049
0.098
0.069
–
–
–
MeO
O
MeO
O
MeO
O
6
2-CF3
2-CONH2
3-CONH2
0.034
–
> 10
7
9
c
c
10
8
a
Values are means of triplicate SPA-bead experiments performed on the same
Scheme 2. Reagents and conditions: (a) DDQ, THF/H2O, 0 °C (53%); (b) HONH2.HCl,
NaOAc, EtOH/H2O, reflux (71%); (c) KOH, EtOH/H2O, reflux (61–87%).
dilution.

1748
T. Barf et al. / Bioorg. Med. Chem. Lett. 19 (2009) 1745–1748
modes for this novel class of hexahydrocyclohepta[b]indole-based
A-FABP inhibitors. Inhibitors such as 30, will further help to delin-
eate the exact role of A-FABP in fatty-acid homeostasis in adipose
tissue and macrophages.
References and notes
1. (a) Banaszak, L.; Winter, N.; Xu, Z.; Bernlohr, D. A.; Cowan, S.; Jones, T. A. Adv.
Protein Chem. 1994, 45, 89; (b) Bernlohr, D. A.; Doering, T. L.; Kelly, T. J., Jr.;
Lane, M. D. Biochem. Biophys. Res. Commun. 1985, 132, 50; (c) Zimmerman, A.
W.; Veerkamp, J. H. Cell. Mol. Life Sci. 2002, 59, 1096.
2. Furuhashi, M.; Hotamisligil, G. S. Nat. Rev. 2008, 7, 489.
3. (a) Kazuhisa, M. Diabetes Res. Clin. Pract. 2007, 77, 17; (b) Hotamisligil, G. S.
Nature 2006, 444, 860; (c) Makowski, L.; Hotamisligil, G. S. Curr. Opin. Lipidol.
2005, 16, 543.
4. (a) Hotamisligil, G. S.; Johnson, R. S.; Distel, R. J.; Ellis, R.; Papaioannou, V. E.;
Spiegelman, B. M. Science 1996, 274, 1377; (b) Uysal, K. T.; Scheja, L.;
Wiesbrock, S. M.; Bonner-Weir, S.; Hotamisligil, G. S. Endocrinology 2000, 141,
3388.
5. (a) Makowski, L.; Boord, J. B.; Maeda, K.; Babaev, V. R.; Uysal, K. T.; Morgan, M.
A.; Parker, R. A.; Suttles, J.; Fazio, S.; Hotamisligil, G. S.; Linton, F. M. Nat. Med.
2001, 7, 699; (b) Boord, J. B.; Maeda, K.; Makowski, L.; Babaev, V. R.; Fazio, S.;
Linton, F. M.; Hotamisligil, G. S. Arterscler. Thromb. Vasc. Biol. 2002, 22, 1686.
6. Furuhashi, M.; Tuncman, G.; Görgün, C. Z.; Makowski, L.; Atsumi, G.;
Vaillancourt, E.; Kono, K.; Babaev, V. R.; Fazio, S.; Linton, M. F.; Sulsky, R.;
Robl, J. A.; Parker, R. A.; Hotamisligil, G. S. Nature 2007, 447, 959.
7. Tuncman, G.; Erbay, E.; De Vivo, I.; Campos, H.; Rimm, E. B.; Hotamisligil, G. S.
PNAS 2006, 103, 6970.
Figure 4. Overlays of co-crystal structures of 26 (in yellow) and 30 (in green) in
human A-FABP (PBD codes: 3FR4 and 3FR5).
8. (a) Ringom, R.; Axen, E.; Uppenberg, J.; Lundbäck, T.; Rondahl, L.; Barf, T. Bioorg.
Med. Chem. Lett. 2004, 14, 4449; (b) Sulsky, R.; Magnin, D. R.; Huang, Y.;
Simpkins, L.; Taunk, P.; Patel, M.; Zhu, Y.; Stouch, T. R.; Bassolino-Klimas, D.;
Parker, R.; Harrity, T.; Stoffel, R.; Taylor, D. S.; Lavoie, T. B.; Kish, K.; Jacobson, B.
L.; Sheriff, S.; Adam, L. P.; Ewing, W. R.; Robl, J. A. Bioorg. Med. Chem. Lett. 2007,
17, 3511; (c) McDonnell, P. A.; Constantine, K. L.; Goldfarb, V.; Johnson, S. R.;
Sulsky, R.; Magnin, D. R.; Robl, J. A.; Caulfield, T. J.; Parker, R. A.; Taylor, D. S.;
Adam, L. P.; Metzler, W. J.; Mueller, L.; Farmer, B. T. J. Med. Chem. 2006, 49, 501.
9. Lehmann, F.; Haile, S.; Axen, E.; Medina, C.; Uppenberg, J.; Svensson, S.;
Lundbäck, T.; Rondahl, L.; Barf, T. Bioorg. Med. Chem. Lett. 2004, 14, 4445.
10. Binas, B.; Danneberg, H.; McWhir, J.; Mullins, L.; Clark, A. J. FASEB J. 1999, 13,
805.
11. Veerkamp, J. H.; Peters, R. A.; Maatman, R. Biochim. Biophys. Acta 1991, 1081, 1.
12. Lundbäck, T.; Schultz, J.; Feldwisch, J.; Uppenberg, J. Rec. Res. Devel. Chem. 2004,
2, 165.
13. Van Dongen, M. J. P.; Uppenberg, J.; Svensson, S.; Lundbäck, T.; Åkerud, T.;
Wikström, M.; Schultz, J. J. Am. Chem. Soc. 2002, 124, 11874. The fluorescence
polarization assay described herein was used with minor modifications..
14. Carboxamide derivatives 29 and 30 were prepared via the respective 2- and 3-
cyanobenzyl intermediates. Hydrolysis of the methyl ester and nitrile to the
final compounds was effected in one step with KOH in refluxing EtOH/water
(5/1 v/v).
the hydroxyl group of S53 was observed while comparing co-crys-
tal structure complexes with 30, versus 26. T53 in H-FABP is pre-
sumably unable to adjust or simply causes a steric clash, which
most likely explains the superior selectivity of 30. The relative
higher polarity of 30 resulted in a slightly reduced plasma protein
binding (99.7%) as compared to the more lipophilic congeners.
In conclusion, 2a proved to be a viable starting point of a lead
optimization program, and using a structure-based drug design ap-
proach, several improvements were effected. Expanding the right-
hand ring of the tetrahydrocarbazole to a seven-membered ring
gave improved affinity. Introduction of a CF3 group in the ortho-po-
sition (26) or a carboxamide in the meta-position (30) of the N-
benzyl substituent, gave a significant improvement leading to
nanomolar inhibitors and with different selectivity profiles against
H-FABP. This difference could in part be explained by solving the
co-crystal structures showing two distinctly different binding