N-Phenylphenylglycines as Novel Corticotropin Releasing Factor Receptor Antagonists

Article abstract of DOI:10.1021/jm049974i

Screening of a computationally designed synthetic library led to the discovery of the N-phenylphenylglycines (NPPGs) as a novel class of human corticotropin releasing factor (h-CRF₁) antagonists. Several NPPGs with greater potency than the original hit 1 were rapidly identified, and resolution of the racemate demonstrated that only the R-enantiomer displays activity. This structural class represents the first example of a non-peptide CRF₁ antagonist with a stereochemically distinct receptor binding affinity.

Full text of DOI:10.1021/jm049974i

2426
J . Med. Chem. 2004, 47, 2426-2429
N-P h en ylp h en ylglycin es a s Novel
Cor ticotr op in Relea sin g F a ctor Recep tor
An ta gon ists
Valentina Molteni,† J ulie Penzotti,† Dean M. Wilson,†
Andreas P. Termin,† Long Mao,† Christine M. Crane,†
Fred Hassman,† Tao Wang,† Harvey Wong,‡
Keith J . Miller,‡ Scott Grossman,‡ and
Peter D. J . Grootenhuis*,†
F igu r e 1. Initial NPPG hit.
Bristol-Myers Squibb Pharma Research Laboratories,
4570 Executive Drive, Suite 400,
San Diego, California 92121, and
Bristol-Myers Squibb, Experimental Station,
E353/ 11B, Wilmington, Delaware 19805
Received J anuary 7, 2004
Abstr a ct: Screening of a computationally designed synthetic
library led to the discovery of the N-phenylphenylglycines
(NPPGs) as a novel class of human corticotropin releasing
factor (h-CRF1) antagonists. Several NPPGs with greater
potency than the original hit 1 were rapidly identified, and
resolution of the racemate demonstrated that only the R-
enantiomer displays activity. This structural class represents
the first example of a non-peptide CRF1 antagonist with a
stereochemically distinct receptor binding affinity.
F igu r e 2. Compound 30A (the R enantiomer) aligned to two
high information content pharmacophores. Each pharmaco-
phore contains one hydrogen-bond acceptor (magenta), an
aromatic group (yellow), and two hydrophobic groups (brown).
class of CRF antagonists utilizing computational design
tools and parallel synthesis to rapidly advance from “hit
to lead”.
Corticotropin releasing factor (CRF; also known as
corticotropin releasing hormone)1 is the primary physi-
ological regulator of the hypothalamic-pituitary-
adrenal axis, presiding over a large number of neuronal,
endocrine, and immune processes. This 41 amino acid
peptide is a major modulator of the body’s response to
stress,2 and regulation of its function has become an
important target for drug design. Studies of CRF
agonists have shown that abnormal hormonal levels
may be important in neuropsychiatric disorders such
as anxiety and depression,3 substance abuse,4 eating
disorders,5,6 premature parturition,7 and gastrointesti-
nal maladies.8 Peptide agonists and antagonists have
been available for many years, and they are valuable
tools for the investigation of CRF-mediated physiology.
Studies with peptide ligands provide support for a role
for CRF in coordinating the body’s response to stress.
In the past 10 years, considerable progress in the
identification of non-peptide modulators of CRF has
provided opportunities for clinical studies on treatment
for some of the disorders described above. Several
classes of antagonists have already been investigated
to date,9 but there is still a need for the identification
of structurally diverse CRF antagonists.
The initial hit from the N-phenylphenylglycine series,
1 (Figure 1), was discovered using computational library
design methods. A library design space biased toward
pharmacophores contained in compounds exhibiting
CRF1 antagonist activity was generated from an initial
data set that included 2973 actives (Ki < 100 nM) and
1212 inactives (Ki > 10µΜ). The library design space
was derived using molecular signatures composed of
three-dimensional pharmacophore descriptors described
in the Supporting Information.
To select the subset of all pharmacophore hypotheses
potentially associated with CRF1 activity, the signatures
of all compounds were systematically analyzed in the
context of their associated activity data. The pharma-
cophores were ranked on the basis of their ability to
discriminate between actives and inactives across the
entire data set. The ranking criterion was “information
content”, a function of the active and inactive com-
pounds (see ref 11 for the equation). To assess the
significance of the information content value, the activ-
ity data were randomly permuted (10 trials) to deter-
mine the average information content level expected
from random chance correlations in the data set. All
77 075 pharmacophores with information content values
greater than those expected by random chance were
utilized as our CRF1 library design space. Figure 2
shows two examples of pharmacophores with high
information content. The arrangement of the hydrogen
bond acceptor and aromatic and hydrophobic groups in
these high-ranking pharmacophores is consistent with
the general CRF1 antagonist SAR proposed by Gilligan.2
The lower information content pharmacophores encom-
pass a more diverse collection of feature combinations
Corticotropin releasing factor receptor antagonists
such as anilinopyrimidines, triazines, and bicyclic com-
pounds have been reviewed by Gilligan.2 In a previous
paper we described that arylamidrazones are a novel
class of CRF antagonists.10 Here, we report the discov-
ery of the N-phenylphenylglycines (NPPGs) as a new
* To whom the correspondence should be addressed. Phone: 858-
404-6676. Fax: 858-404-6713. E-mail: Peter_Grootenhuis@sd.vrtx.com.
† Bristol-Myers Squibb Pharma Research Laboratories.
‡ Bristol-Myers Squibb.
10.1021/jm049974i CCC: $27.50 © 2004 American Chemical Society
Published on Web 04/10/2004

Letters
J ournal of Medicinal Chemistry, 2004, Vol. 47, No. 10 2427
Sch em e 1a
F igu r e 3. NPPG’s synthetic strategy.
phores) and off-bits (inaccessible pharmacophores) in
their molecular signatures systematically investigated
the CRF1 pharmacophore space. While one would not
expect to test all 77 075 hypotheses with a small
combinatorial library for one chemical scaffold, the
NPPG library design in combination with informative
library designs for additional chemical series provided
sufficient information to refine our CRF-biased phar-
macophore space for subsequent rounds of design and
synthesis. (In the combined library designs for multiple
classes, ∼99% of the 77 075 pharmacophores were
matched by at least one compound and ∼60% were
covered by 50 or more compounds.) The final NPPG
combinatorial matrix design for synthesis was composed
of all products resulting from the combination of the 12
computationally selected amines with 6 anilines and 6
boronic acids preselected by medicinal chemistry intu-
ition. The NPPG compounds included the 4-ethyl-
aminopyridine building block and the product 1.
The N-phenylphenylglycine amides were synthesized
in a two-step process involving a boronic acid Mannich
reaction13 followed by amidation (Scheme 1). Condensa-
tion of boronic acids (1) and amines (2) with glyoxylic
acid afforded the intermediate acids (3) that were
converted to the NPPGs (4) by amidation.
The first NPPG library utilized 6 anilines (2), 6
boronic acid (1), and 12 amines. Of 36 expected acids
(3), 30 examples were successful and were converted to
amides (4). Of 180 possible racemic products, 174 were
isolated by LC/MS in >85% purity.14 One mixture with
modest affinity in the in vitro h-CRF1 binding assay was
identified. Racemic mixture 1 afforded a Ki of 1035 nM.
Mixture 1 was also tested against CRF2 and found to
be inactive at 10 µM.
To optimize the NPPG series, each of the three
branches of the molecule were evaluated independently
(Figure 3). Initially, B-1 was varied while retaining B-2
and B-3 (Table 1). Substitution of the bromo at the
2-position with other substituents (13 examples) led to
the 2-vinyl- and 2-methyl-substituted variants (4, Ki )
752 nM; 5, Ki ) 922 nM), which were slightly more
potent than the original example 1. Other substituents
(CF3, OPh, SCH3, OCH2CH3, Ph) in the 2-position
resulted in inactive racemates. Monosubstitution in the
3- or 4-position afforded inactive or weakly active
racemic mixtures as illustrated by comparison of the
three monomethyl products 5, 9, and 12 and the three
monochloro products 2, 8, and 11. Interestingly, the 2-
and 3-methoxy mixtures 7 and 10, respectively, were
inactive as were the dimethoxy variants 13-15. The
unsubstituted phenyl derivative 16 was inactive. Race-
mates derived from heteroaromatic boronic acids were
also inactive.
a
Conditions: (a) HCOCOOH, toluene, 80 °C, 12 h; (b) HATU,
DIEA, DMSO, 12 h.
and arrangements, representing alternative descrip-
tions of activity from the known SAR.
Because our challenge was to identify novel classes
of CRF1 antagonists, we used the full 77 075 pharma-
cophore CRF1 design space to select chemical scaffolds
and design libraries for synthesis. New scaffolds cover-
ing a broad range of chemical diversity were evaluated
for their ability to produce virtual libraries that could
match pharmacophores in the CRF1 design space. Each
of the 77 075 pharmacophores was treated equally, i.e.,
no extra importance was placed on scaffolds that were
able to cover more high than low information content
pharmacophores. Also examined was whether each
scaffold covered pharmacophores that were not matched
by other scaffolds. Pharmacophore coverage in the CRF1
design space, complementarity of the pharmacophores
covered, and synthetic feasibility constraints were the
main criteria used to select a set of chemistries for
library design and synthesis. The NPPG class was one
of these selections.
To design libraries that systematically explore the
portion of CRF1 pharmacophore space covered by the
new chemical classes, we applied the strategy of infor-
mative design12 by which compounds are used to “inter-
rogate” the target receptor and to determine which
chemical features are required for activity. The method
constructs a library such that the maximum number of
conclusions can be drawn from the new biological
screening results. Informative design was implemented
as a maximum Shannon entropy criterion to search a
binary descriptor space (e.g., pharmacophore signatures)
and systematically identify descriptors important for
activity.
For the first NPPG combinatorial library, 12 amines
were selected computationally from a list of 367 possible
amines using the whole molecule (product) based three-
dimensional pharmacophore descriptors and informa-
tive design.12 The 77 075 bit pharmacophore signatures
were generated for the final molecular products of each
of the 367 amines in combination with the two sets of
additional building blocks, anilines and boronic acids
in the chemical reaction for the formation of NPPGs
(Scheme 1).
The result of this design was a selection of compounds
whose combination of on-bits (accessible pharmaco-
Next B-2 was varied, keeping B-1 and B-3 constant.
A variety of amines were examined including aromatic,

2428 J ournal of Medicinal Chemistry, 2004, Vol. 47, No. 10
Letters
Ta ble 1. Structure-Activity Relationship Data15
c
compda
R1
R2
R4
Ki (nM)
1
2
3
4
5
6
7
8
2-Br
2-Cl
2-F
2-CHCH2
2-CH3
2-CH(CH3)2
2-OCH3
3-Cl
3-CH3
3-OCH3
4-Cl
4-CH3
2,5-di-OCH3
2,4-di-OCH3
2,6-di-OCH3
H
2-Br
2-Br
2-Br
2-Br
2-Br
2-Br
2-Br
2-Br
2-Cl, 4-CH3
2-Cl, 4-CH3
2-Cl, 4-CH3
2-Cl, 4-CH3
2-Cl, 4-CH3
2-Cl, 4-CH3
2-Cl, 4-CH3
2-Cl, 4-CH3
2-Cl, 4-CH3
2-Cl, 4-CH3
2-Cl, 4-CH3
2-Cl, 4-CH3
2-Cl, 4-CH3
2-Cl, 4-CH3
2-Cl, 4-CH3
2-Cl, 4-CH3
2-Cl
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
B
C
D
E
1035
3346
>10000
752
922
3295
>10000
>5000
>10000
>10000
>10000
>10000
>10000
>10000
>10000
>10000
7334
9
10
11
12
13
14
15
16
17
18b
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
F igu r e 4. R4 substitution for compounds in Table 1.
2-Cl
2,4-di-Cl
>10000
204
2-Cl, 4-CH3
2-Cl, 4-CH3
2-Cl, 4-CH3
2-Cl, 4-CH3
2-Cl, 4-CH3
2-Cl, 4-CH3
2-Cl, 4-CH3
2-Cl, 4-CH3
2,4-di-Cl
2,4-di-Cl
2,4-di-Cl
2,4-di-Cl
2-Cl, 4-CH3
2,4-di-Cl
2-F, 4-CH3
2-CF3, 4-Cl
2-CF3, 4-CH3
2,4-di-Cl
2,4-di-Cl
2,4-di-Cl
>10000
>10000
>10000
>10000
>10000
>10000
1075
F
2-Br
2-Br
2-Br
G
H
I
1066
330
338
370
512
752
1002
439
2-CH2CH3
2,6-di-CH3
2- CH3
2- CH3,4-F
2-CHCH2
2-CF3
2-CH3
2-CH3
2-CH3
2-CH3
2-CH3
2-CH3
2-Br
A
A
A
A
A
A
A
A
A
I
F igu r e 5. Resolved NPPGs.
469
641
387
(31). The free phenol 37 resulted in an active mixture
but only when paired with an obligate C4 methoxy
substituent (37 vs 38 and 39).
L
M
N
>10000
>10000
514
2,4-di-Cl
a
b
R3 ) H unless otherwise specified. R3 ) Me. c See Figure 4.
Racemic mixture 1 was resolved by chiral supercriti-
cal fluid chromatography (SFC; Figure 5). The absolute
configuration of the two enantiomers 1A (R) and 1B (S)
was established by determining the configuration of the
precursor acid Ac1B, which corresponds to the second
compound 1B eluted by SFC. The racemic mixture of
the intermediate acid was resolved into the two pure
enantiomers Ac1A and Ac1B (see Figure 5). Each of
the two acids was coupled with 2-pyridin-4-ylethyl-
amine, giving 1A from Ac1A and 1B from Ac1B. X-ray
analysis of Ac1B confirmed the S configuration, allow-
ing assignment of 1B as S and 1A as R (see Supporting
Information). The two pure enantiomers 1A and 1B
were tested for binding affinity to the CRF1 receptor.
Compound 1A showed a 2-fold increase in binding
affinity compared to the racemic mixture (Ki ) 546 nM),
whereas 1B was inactive. This is one of the first
examples of a non-peptide antagonist of the CRF1
receptor in which chirality plays a role.
heteroaromatic, aliphatic, primary, and secondary (74
reactions attempted). Only aniline containing racemic
mixtures were found to have activity. Removal of the
methyl in the 4-position (17) resulted in a decrease in
activity, and N-methylation resulted in an inactive
mixture (18). Ultimately 2,4-dichloro substitution re-
sulted in the most active racemate 19 (Ki ) 204 nM).
Finally, B-3 was investigated using aromatic, het-
eroaromatic, aliphatic, primary, and secondary nitrogen
substituents (210 reactions attempted). The 4-pyridyl
feature was strongly favored over other nitrogen regio-
isomers because 2- and 3-pyridyl variants were inactive
(1 vs 20 vs 21). Molecules featuring a phenyl (22),
imidazole (23), or a 4-fluorophenyl group (24) resulted
in inactive mixtures. Shortening the aliphatic chain
linker (25) resulted in a loss of activity as well. Racemic
mixtures 26 and 27 derived from anilines rather than
aliphatic amines showed weak binding affinities with
Ki values of 1075 and 1066 nM, respectively.
Using the resulting optimized side chains, a final
round of synthesis focused on small modifications of the
preferred substituents (Table 1, Figure 4). The optimal
R2 was 2,4-dichloro substitution (compare 30, 34-36,
and 5). The greatest affinity was observed for racemic
mixture 19 when R1 ) 2-Br (Ki ) 204 nM). Substituents
that also resulted in an active mixture were R1 ) 2-CH2-
CH3 (28), 2,6-di-CH3 (29), 2-CH3 (30), 2-CH3, and 4-F
Racemic mixtures 5 and 30 were also resolved,
yielding 5A and 5B and 30A and 30B, respectively. The
absolute configuration of each was assigned by correla-
tion of their retention time (chiral SFC) to those for 1A
and 1B. Also, as in the previous example, 5A and 30A
were more potent when compared to the racemic mix-
ture (Ki ) 244 and 154 nM, respectively). Compound
5B was inactive, and 30B showed a Ki of 3180 nM. To
establish the selectivity of this class of compounds, 30A
was also serotonin receptors 5-HT2A, 5-HT1A, 5-HT1B
,

Letters
J ournal of Medicinal Chemistry, 2004, Vol. 47, No. 10 2429
(3) Aguilera, G. Corticotropin releasing hormone, receptor regulation
and the stress response. Trends Endocrinol. Metab. 1998, 9,
329-336.
(4) Koob, G. F. Stress, corticotropin-releasing factor, and drug
addiction. Ann. N. Y. Acad. Sci. 1999, 897, 27-45.
(5) Pelleymounter, M. A.; J oppa, M.; Carmouche, M.; Cullen, M. J .;
Brown, B.; Murphy, B.; Grigoriadis, D. E.; Ling, N.; Foster, A.
C. Role of corticotropin-releasing factor (CRF) receptors in the
anorexic syndrome induced by CRF. J . Pharmacol. Exp. Ther.
2000, 293, 799-806.
(6) Kaye, W. H. Neuropeptide abnormalities in anorexia nervosa.
Psychiatry Res. 1996, 62, 65-74.
(7) Keller, P. A.; Elfick, L.; Garner, J .; Morgan, J .; McCluskey, A.
Corticotropin releasing-hormone: therapeutic implications and
medicinal chemistry developments. Bioorg. Med. Chem. 2000,
8, 1213-1223.
(8) McCarthy, J . R.; Heinrichs, S. C.; Grigoriadis, D. E. Recent
advances with the CRF1 receptor: design of small molecule
inhibitors, receptor subtypes, and clinical indications. Curr.
Pharm. Des. 1999, 5, 289-315.
5-HT1D, 5-HT6, and 5-HT7 and resulted inactive at 10
µM in all cases.
To assess the pharmacokinetics of an early lead in
this series, we included 30A in a dog cassette dose
study.16 Briefly, 30A was coadministered intravenously
with seven other compounds to beagle dogs (n ) 2) at a
dose of 0.2 mg/kg for each compound. In addition, the
same set of compounds was administered orally to
beagle dogs (n ) 2) at a dose of 1 mg/kg. Plasma samples
were collected at 0, 0.1, 0.25, 0.5, 1, 2, 4, 6, 8, and 24 h
postdose for the intravenous experiment and at 0, 0.5,
1, 2, 4, 6, 8, and 24 h postdose for the oral experiment.
Drug concentrations were determined in the plasma
samples using LC/MS/MS. The following pharmacoki-
netic parameters were derived noncompartmentally
from the described study: total body clearance, 1.3 (L/
h)/kg, volume of distribution at steady state, 1.8 L/kg;
half-life, 1.7 h; oral bioavailability, 7%. Although the
parameters are not optimal, the total body clearance
estimate categorizes 30A as a moderate clearance
compound. Thus, 30A provides a compelling starting
point for further potency and bioavailability optimiza-
tion.
We have described the discovery of a novel series of
CRF1 antagonists by using a computational library
design strategy. Optimization of the core structure was
accomplished through parallel synthesis of small, tar-
geted libraries. In addition, resolution of the most active
mixture in the series (30, Ki ) 370 nM) showed that
the receptor seems to have a stereochemical bias, since
only the R enantiomer retained activity (30A, Ki ) 154
nM). This structural class represents the first example
in which the enantiomers of a non-peptide molecule
demonstrated differences in CRF1 binding affinity. In
addition, preliminary pharmacokinetic studies on 30A
were encouraging and provide incentive for additional
work in this series.
(9) Saunders, J .; Williams, J . New developments in the study of
corticotropin releasing factor. Annu. Rep. Med. Chem. 2001, 36,
21-30 and references therein.
(10) Wilson, D. M.; Termin, A. P.; Mao, L.; Ramirez-Weinhouse, M.
M.; Molteni, V.; Grootenhuis, P. D. J .; Miller, K.; Keim, S.; Wise,
G. Arylamidrazones as novel corticotropin releasing factor
receptor antagonists. J . Med. Chem. 2002, 45, 2123-2126.
(11) Srinivasan, J .; Castellino, A.; Bradley, E. K.; Eksterowicz, J . D.;
Grootenhuis, P. D. J .; Putta, S.; Stanton, R. V. Evaluation of a
novel shape-based computational filter for lead evolution: ap-
plication to thrombin inhibitors. J . Med. Chem. 2002, 45 (12),
2494-2500.
(12) Miller, J . L.; Bradley, E. K.; Teig, S. L. Luddite: an information-
theoretic library design tool. J . Chem. Inf. Comput. Sci. 2003,
43 47-54.
(13) (a) Petasis, N. A.; Zavialov, I. A. A new and practical synthesis
of R-amino acids from alkenyl boronic acids. J . Am. Chem. Soc.
1997, 119, 445-446. (b) Petasis, N. A.; Goodman, A.; Zavialov,
I. A new synthesis of R-arylglycines from aryl boronic acids.
Tetrahedron 1997, 53, 16463-16470. (c) Petasis, N. A.; Zavialov,
I. Highly stereocontrolled one-step synthesis of anti-â-amino
alcohols from organoboronic acids, amines, and R-hydroxy al-
dehydes. J . Am. Chem. Soc. 1998, 120, 11798-11799.
(14) Zeng, L.; Kassel, D. B. Anal. Chem. 1998, 70, 4380.
(15) (a) Arvanitis, A. G.; Gilligan, P. J .; Chorvat, R. J .; Cheeseman,
R. S.; Christos, T. E.; Bakthavatchalam, R.; Beck, J . P.; Cocuzza,
A. J .; Hobbs, F. W.; Wilde, R. G.; Arnold, C.; Chidester, D.; Curry,
M.; He, L.; Hollis, A.; Klaczkiewicz, J .; Krenitsky, P. J .; Rescinito,
J . P.; Scholfield, E.; Culp, S.; De Souza, E. B.; Fitzgerald, L.;
Grigoriadis, D.; Tam, S. W.; Wong, Y. N.; Huang, S.; Shen, H.-
S. L. Non-peptide corticotropin releasing hormone antagonists:
syntheses and structure-activity relationships of 2-anilino-
pyrimidines and -triazines. J . Med. Chem. 1999, 42, 805-818.
(b) Mean Ki values were reported when more than one deter-
mination was performed. If the deviation between initial dupli-
cate Ki values exceeded 15%, additional duplicate studies were
performed until this criterion was met and determinations were
averaged. Internal control compounds were run on every plate.
(16) (a) Berman, J .; Halm, K.; Adkison, K.; Schaffer, J . Simultaneous
pharmacokinetic screening of a mixture of compounds in dog
using API LC/MS/MS analysis for increased throughput. J . Med.
Chem. 1997, 40, 827-829. (b) He, L.; Gilligan, P. J .; Zaczek, R.;
Fitzgerald, L. W.; McElroy, J .; Shen, H.-S. L.; Saye, J . A.; Kalin,
N. H.; Shelton, S.; Christ, D.; Trainor, G.; Hartig, P. 4-(1,3-
Dimethoxyprop-2-ylamino)-2,7-dimethyl-8-(2,4-dichlorophenyl)-
pyrazolo[1,5-a]-1,3,5-triazine: a potent, orally bioavailable CRF1
receptor antagonist. J . Med. Chem. 2000, 43, 449-456.
Ack n ow led gm en t. The authors thank R. E. Olson,
M. Ramirez-Weinhouse, M. Suto, P. Myers, R. Rourick,
A. Castellino, and P. Anderson for support and discus-
sions.
Su p p or tin g In for m a tion Ava ila ble: Experimental and
computational procedures, 1H NMR, LC/MS, HRMS, and
crystal structure data. This material is available free of charge
via the Internet at http://pubs.acs.org.
Refer en ces
(1) Vale, W.; Spiess, J .; Rivier, C.; Rivier, J . Characterization of a
41-residue ovine hypothalamic peptide that stimulates secretion
of corticotropin and beta-endorphin. Science 1981, 213, 1394-
1397.
(2) Gilligan, P. J .; Robertson, D. W.; Zaczeck, R. Corticotropin
releasing factor (CRF) receptor modulators: progress and op-
portunities for new therapeutic agents. J . Med. Chem. 2000, 43,
1641-1660 and references therein.
J M049974I