Identification and Optimization of Benzimidazole Sulfonamides as Orally Bioavailable Sphingosine 1-Phosphate Receptor 1 Antagonists with in Vivo Activity

Article abstract of DOI:10.1021/acs.jmedchem.5b01078

We report here a novel series of benzimidazole sulfonamides that act as antagonists of the S1P₁ receptor, identified by exploiting an understanding of the pharmacophore of a high throughput screening (HTS)-derived series of compounds described previously. Lead compound 2 potently inhibits S1P-induced receptor internalization in a cell-based assay (EC₅₀ = 0.05 μM), but has poor physical properties and metabolic stability. Evolution of this compound through structure-activity relationship development and property optimization led to in vivo probes such as 4. However, this compound was unexpectedly found to be a potent CYP3A inducer in human hepatocytes, and thus further chemistry efforts were directed at addressing this liability. By employing a pregnane X receptor (PXR) reporter gene assay to prioritize compounds for further testing in human hepatocytes, we identified lipophilicity as a key molecular property influencing the likelihood of P450 induction. Ultimately, we have identified compounds such as 46 and 47, which demonstrate the desired S1P₁ antagonist activity while having greatly reduced risk of CYP3A induction in humans. These compounds have excellent oral bioavailability in preclinical species and exhibit pharmacodynamic effects of S1P₁ antagonism in several in vivo models following oral dosing. Relatively modest antitumor activity was observed in multiple xenograft models, however, suggesting that selective S1P₁ antagonists would have limited utility as anticancer therapeutics as single agents.

Full text of DOI:10.1021/acs.jmedchem.5b01078

Subscriber access provided by - Access paid by the | UCSB Libraries
Article
Identification and Optimization of Benzimidazole
Sulfonamides as Orally Bioavailable Sphingosine 1-
Phosphate Receptor 1 Antagonists with in Vivo Activity
Edward J Hennessy, Vibha B. Oza, Ammar Adam, Kate Byth, Lillian Castriotta, Gurmit Grewal,
Geraldine Hamilton, Victor M Kamhi, Paula Lewis, Danyang Li, Paul D Lyne, Linda Oster,
Michael T Rooney, Jamal C Saeh, Li Sha, Qibin Su, Shengua Wen, Yafeng Xue, and Bin Yang
J. Med. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.jmedchem.5b01078 • Publication Date (Web): 20 Aug 2015
Downloaded from http://pubs.acs.org on August 24, 2015
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a free service to the research community to expedite the
dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts
appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been
fully peer reviewed, but should not be considered the official version of record. They are accessible to all
readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered
to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published
in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just
Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor
changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers
and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors
or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
Journal of Medicinal Chemistry is published by the American Chemical Society. 1155
Sixteenth Street N.W., Washington, DC 20036
Published by American Chemical Society. Copyright © American Chemical Society.
However, no copyright claim is made to original U.S. Government works, or works
produced by employees of any Commonwealth realm Crown government in the course
of their duties.

Page 1 of 78
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Identification and Optimization of Benzimidazole Sulfonamides
as Orally Bioavailable Sphingosine 1ꢀPhosphate Receptor 1
Antagonists with in Vivo Activity
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Edward J. Hennessy,* Vibha Oza, Ammar Adam, Kate Byth, Lillian Castriotta, Gurmit
Grewal, Geraldine A. Hamilton, Victor M. Kamhi, Paula Lewis, Danyang Li, Paul Lyne,
†
Linda Öster, Michael T. Rooney, Jamal C. Saeh, Li Sha, Qibin Su, Shengua Wen, Yafeng
†
Xue, and Bin Yang
Oncology iMed, Innovative Medicines and Early Development, AstraZeneca R&D Boston, 35
Gatehouse Drive, Waltham, MA 02451, USA
†
Discovery Sciences, Innovative Medicines and Early Development, AstraZeneca R&D
Mölndal, S-43183, Mölndal, Sweden
Abstract
We report here a novel series of benzimidazole sulfonamides that act as antagonists of
1
the S1P receptor, identified by exploiting an understanding of the pharmacophore of an
HTSꢀderived series of compounds described previously. Lead compound 2 potently inhibits
S1Pꢀinduced receptor internalization in a cellꢀbased assay (EC50 = 0.05 µM), but has poor
physical properties and metabolic stability. Evolution of this compound through structureꢀ
activity relationship development and property optimization led to in vivo probes such as 4.
However, this compound was unexpectedly found to be a potent CYP3A inducer in human
hepatocytes, and thus further chemistry efforts were directed at addressing this liability. By
employing a PXR reporter gene assay to prioritize compounds for further testing in human
hepatocytes, we identified lipophilicity as a key molecular property influencing the likelihood
of P450 induction. Ultimately, we have identified compounds such as 46 and 47, which
demonstrate the desired S1P
induction in humans. These compounds have excellent oral bioavailability in preclinical
species, and exhibit pharmacodynamic effects of S1P antagonism in several in vivo models
1
antagonist activity while having greatly reduced risk of CYP3A
1
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 2 of 78
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
following oral dosing. Relatively modest antitumor activity was observed in multiple
1
xenograft models, however, suggesting that selective S1P antagonists would have limited
utility as anticancer therapeutics as single agents.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Introduction
Sphingosine 1ꢀphosphate (S1P), a signaling phospholipid found throughout blood and
bodily tissues, plays a role in a diverse variety of biological processes. Through its
interactions with a class of five G proteinꢀcoupled receptors (GPCRs) known as the S1P
receptors (S1P ꢀS1P ), S1P plays a role in cell proliferation and migration, cellular
1
5
1
architecture, and immune cell trafficking. Accordingly, agents that are capable of
modulating the interactions of S1P with its receptors may have utility as therapeutics in a
variety of diseases such as diabetes, inflammatory disorders, and cancer.
2
FTY720 (fingolimod), a sphingolipid mimetic first described in 1995, is
phosphorylated in cells to yield a potent agonist for four of the five isoforms of S1P receptor
3
,4
(
S1P , S1P , S1P , and S1P ). Through its interactions with S1P , this compound inhibits
1 3 4 5 1
lymphocyte egress from secondary lymphoid organs, resulting in an immunosuppressive
5
effect in vivo. As a result, it was approved for use in treatment of multiple sclerosis in 2010.
Upon binding to and activating S1P
1
, fingolimod causes prolonged downregulation of the
6
receptor by inducing receptor internalization followed by ubiquitinylation and proteasomal
7
,8
degradation. Therefore, despite being an S1P agonist, it is this “functional antagonism”
1
that is responsible for the compound’s beneficial therapeutic activity.
1
In addition to the effects on lymphocyte trafficking attributed to S1P function, this
receptor has been shown to play a key role in angiogenesis and the maintenance of vascular
integrity. For example, S1P ꢀknockout mice demonstrate impaired vascular maturation
1
9
during embryonic development, leading to lethal hemorrhaging in utero. Additionally,
ACS Paragon Plus Environment

Page 3 of 78
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
siRNA silencing of the S1P
1
gene within subcutaneous implants of Matrigel results in the
1
0
suppression of vessel formation in vivo. Furthermore, smallꢀmolecule abrogation of S1P
1
11
signaling (with fingolimod) inhibits angiogenesis in multiple in vivo models.
In clinical trials of fingolimod, transient bradycardia was frequently observed upon
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
1
2
initial dosing. This effect has also been observed preclinically in rodents, and has been
13,14
attributed to the compound’s effects on the S1P
that S1P agonism, resulting in activation of G proteinꢀcoupled inwardly rectifying potassium
GIRK) channels in mouse and human cardiomyocytes, results in the heart rate decreases
3
receptor.
Alternative findings suggest
1
15
16
(
16
observed in human patients. Taken together, these observations suggest that a compound
that selectively and effectively antagonizes the S1P receptor, without any agonism of S1P
1
1
or interaction with S1P
3
, should retain the desired pharmacological profile of fingolimod
without any adverse cardiovascular effects.
Given the documented role of S1P function on vessel formation, and the therapeutic
1
potential of compounds that inhibit angiogenesis in the treatment of various cancers, we
aimed to develop an orally bioavailable, selective S1P antagonist to assess the feasibility of
1
such a compound as an antitumor agent. While numerous examples of S1P
1
agonists that act
17
as functional antagonists have been described, at the time we initiated our program only a
18ꢀ20
few reports of pure antagonists (devoid of agonist activity) were available.
As mimetics
of the phospholipid S1P, these compounds are useful tools but likely have limited utility as
orally available agents. More recently, however, disclosures of orally available, smallꢀ
21ꢀ23
molecule S1P
1
antagonists have emerged in the literature.
We have recently disclosed
our initial efforts towards this goal, having identified a series of heterocyclic sulfonamides
24
1
with submicromolar potency in a S1P receptor internalization assay. In this Article, we
will detail our further evolution of this series of compounds, culminating in the identification
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 4 of 78
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
of compounds (such as 46 and 47) with promising biopharmaceutical properties that elicit in
1
vivo pharmacodynamic effects associated with antagonism of the S1P receptor.
Results and Discussion
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
24
In our previous communication, we highlighted the preliminary development of
structureꢀactivity relationships in a series of triazoleꢀcontaining sulfonamides, with the goal
of improving potency alongside physical properties and metabolic stability. These efforts
established that substitution at the 4ꢀ and 5ꢀpositions of the triazole ring with small alkyl
1
2
groups gave the most effective S1P antagonists (Figure 1, R and R ). As part of this work,
1
we generated data which suggested that the hydrogen bond acceptor at the 1ꢀposition of the
1
triazole ring is not required for interaction with the S1P receptor, since the corresponding
imidazole in which the 1ꢀposition nitrogen atom has been deleted retains the antagonist
activity seen in the triazole. However, the acceptor at the 2ꢀposition appears to be critical for
binding, since the matchedꢀpair imidazole was inactive in an assay measuring inhibition of
2
5
S1Pꢀinduced S1P receptor internalization. These observations were further supported by
1
data for other heterocycles in which the 2ꢀposition hydrogen bond acceptor was maintained
while the 1ꢀposition nitrogen atom was replaced with other heteroatoms. Unfortunately, the
above modifications resulted in poorer physical properties than for the corresponding
triazoles.
To capitalize on these findings, we proposed additional structural modifications to the
triazole scaffold in order to assess the potential for improvements in potency and properties.
In particular, we hypothesized that a bicyclic system (such as benzimidazole) encompassing
the 5ꢀposition substituent as well as the 2ꢀposition hydrogen bond acceptor would maintain
the critical elements of the pharmacophore, while offering additional opportunities for further
substitution that could influence receptor binding as well as physical properties (Figure 1).
ACS Paragon Plus Environment

Page 5 of 78
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 1. Scaffold hopping from triazole to benzimidazole core
To test this hypothesis, we first prepared benzimidazole 2 containing the structural
elements found to be optimal from our studies on the triazole series (ethyl substituent off 5ꢀ
2
4
membered ring, (R)ꢀstereochemistry at chiral center). We were pleased to find that this
compound was a very effective antagonist of human S1P in our receptor internalization
assay, with improved measured potency relative to the most optimized triazoles previously
prepared (exemplified by 1, Table 1). No antagonism of the S1P receptor was observed with
1
3
this compound, consistent with our desired selectivity profile. Despite this promising result,
however, this increased potency came at the cost of a significant increase in lipophilicity.
26
Indeed, utilization of the lipophilic ligand efficiency (LLE) metric indicates that triazole 1 is
a far more “efficient” antagonist that benzimidazole 2. Nevertheless, we were encouraged by
the level of potency not yet observed in the project to that point, and viewed lipophilicity as a
design parameter that could be readily addressed in future rounds of chemistry. We therefore
embarked on efforts to establish the likelihood of delivery of a candidate drug from the
benzimidazole scaffold.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 6 of 78
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Table 1. Comparison of lead triazole 1 with initial benzimidazole 2
S1P Translocation
b
1
Example
1
Structure
a
cLogP
LLE
EC50 (µM)
0.25 ± 0.037
0.05 ± 0.008
2.1
4.5
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
2
4.0
3.3
a
EC50 values are reported as the mean of at least three separate determinations ± standard
b
deviation; LLE = pEC50 – cLogP
An early goal of our work on this scaffold was to study the nature of the interaction
1
between the compound and the receptor. The cellular S1P translocation assay that we used
for primary screening measures the effects of compound treatment on S1Pꢀinduced S1P
1
receptor internalization; compounds that behave as receptor agonists would be expected to
increase the amount of receptor internalization relative to a fixed concentration of S1P alone,
while antagonists should suppress receptor internalization. Based on our observations with 2,
1
we were confident that this compound was acting as an S1P antagonist in the same manner
as we had observed with the triazoles. In order to further characterize the mode of inhibition,
we conducted this assay by varying the concentrations of S1P present at a series of fixed
concentrations of 2. Increasing the concentration of 2 results in a shift of the S1P agonism
doseꢀresponse curve to the right, and also results in a decrease in the maximal amount of
receptor internalization (Supporting Information, Figure S1). This observation, that
antagonism by 2 cannot be overcome by increasing the concentration of S1P, suggests that 2
1
and S1P bind at different sites on the S1P receptor, and that 2 is therefore a noncompetitive
antagonist. We have observed a similar mode of inhibition with structurallyꢀrelated
compounds prepared previously (unpublished data).
Our initial investigations to follow up compound 2 focused on determining whether the
structureꢀactivity relationships previously established for the triazole series could be
ACS Paragon Plus Environment

Page 7 of 78
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
translated to this new scaffold. We therefore first systematically varied the group appended
to the benzimidazole nitrogen, including larger alkyl groups (linear and branched) as well as
substituents containing more polar moieties. However none of these compounds had
improved potency relative to 2 (data not shown), confirming the preference for ethyl
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
2
4
substitution at this position (in accord with our observations from the triazole series).
Similarly, exploration of substituents at the position adjacent to the sulfonamide nitrogen
reiterated our previous findings that an αꢀmethyl group provided the most potency per unit
lipophilicity, and that (R)ꢀstereochemistry at this position is critical to antagonist activity
(data not shown).
In spite of the observation that 2 represented a partially optimized early lead in this
series, this compound had apparent shortcomings that needed to be rectified. For example, 2
is rapidly cleared in vivo when dosed intravenously to Wistar rats, and has very limited oral
exposure (F% = 5). These observations are consistent with this compound’s poor in vitro
stability in the presence of human liver microsomes (CL >100 µL/min/mg). An analysis of
int
the metabolites formed in microsomal incubations pointed to Nꢀdeethylation of the
benzimidazole as the predominant product formed, along with compounds derived from
subsequent oxidation reactions.
Since our earlier data suggested that an Nꢀethyl group on the benzimidazole ring
seemed to be a critical component of the pharmacophore, modification of this group was not
considered to be a viable approach to address metabolic dealkylation. Instead, we next
focused our attention on the variation of substituents elsewhere on the scaffold. In addition to
the goal of furthering our understanding of structureꢀactivity relationships within this series,
we hoped that microsomal stability data for successive compounds would point us towards
drivers for reducing the overall extent of oxidative metabolism.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 8 of 78
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
A survey of various substitutions on the carbocyclic ring of the benzimidazole was
conducted, from which representative compounds are exemplified in Table 2. These data
highlight the effect of lipophilic groups on receptor antagonist activity. In particular, the
inclusion of small nonpolar groups at the 5ꢀ or 6ꢀpositions of this ring results in significant
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
increases in potency in the S1P translocation assay relative to 2 (e.g., 3ꢀ5), although in
1
general these improvements in potency do not offset the increased lipophilicity. As is the
case for 2, these compounds have very limited stability in the presence of human
microsomes. Reducing lipophilicity by substitution of this ring with more polar groups does
appear to reduce the rate of metabolism in vitro, however more polar groups also tend to
result in a reduction in potency (6ꢀ12). A notable exception to this trend is the 6ꢀ
hydroxymethylꢀsubstituted analog 8, which has both a 10ꢀfold improvement in potency
relative to 2 as well as ~1 log unit reduction in cLogP. Unfortunately, the microsomal
stability for this compound is still low, possibly due to the oxidationꢀprone benzylic alcohol
moiety. Substitution of the 7ꢀposition of this ring (as in compound 13) results in a notable
drop in potency relative to 2, which may be due to steric interactions adversely affecting the
orientation of the critical Nꢀethyl group.
Table 2. Representative substitutions on the benzimidazole ring
S1P
Translocation
1
Hu Microsomes
b
Example
Substitution
cLogP
LLE
CLint
a
EC (µM)
(µL/min/mg)
5
0
3
4
5
6
7
8
6ꢀOCH
6ꢀCF
5ꢀF, 6ꢀCl
5ꢀCN
3
0.009 ± 0.009
0.005 ± 0.004
0.001 ± 0.0001
0.13 ± 0.037
0.18 ± 0.14
4.2
4.9
4.9
3.5
2.9
2.9
3.8
3.4
4.1
3.4
3.8
5.4
>100
>100
>100
71
88
84
3
5ꢀCH
6ꢀCH
OH
OH
2
0.005 ± 0.0005
2
ACS Paragon Plus Environment

Page 9 of 78
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
9
5ꢀC(O)NH
6ꢀCH N(CH
2
0.19 ± 0.037
23.5 ± 2.9
2.42 ± 0.81
1.84 ± 0.34
0.80 ± 0.37
2.9
3.8
3.8
2.8
4.7
3.8
0.8
1.8
2.9
1.4
<5
30
45
<5
>100
1
0
1
2
3
2
3
)
2
1
1
1
5ꢀmorpholino
6ꢀSO CH
7ꢀCl
2
3
a
EC50 values are reported as the mean of at least three separate determinations ± standard
deviation; LLE = pEC50 – cLogP
b
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Since we observed a trend towards improved microsomal stability with decreasing
lipophilicity of the benzimidazole substituent, we next surveyed the effects of incorporating a
nitrogen atom into the benzimidazole 6ꢀmembered ring (Table 3). Due to the more forcing
reaction conditions required to effect ring formation in these examples (as discussed in the
Chemistry section), we decided to conduct this initial exploration with racemic compounds
since our previous studies had shown that this tactic could be effectively applied to SAR
24
development. Gratifyingly, azaꢀsubstitution into this ring does not completely abrogate
1
S1P antagonism, with a clear preference for inclusion of nitrogen at the 5ꢀ and 4ꢀpositions of
the benzimidazole (16 and 17, respectively). Moreover, a notable increase in microsomal
stability was observed for these two compounds relative to 2, reiterating our observations
regarding the improved stability of compounds containing polar groups appended to this ring
(Table 2). Not surprisingly, however, these compounds were found to have varying levels of
cytochrome P450 inhibitory activity, likely owing to the ironꢀcoordinating ability of the
27
unhindered pyridine moiety. For example, compound 16 inhibits both CYP2D6 (IC50
=
0.23 µM) and CYP2C19 (IC50 = 0.91 µM). Nevertheless, we were encouraged by the data
for this initial set of azabenzimidazoles, and felt that these could be revisited at a later time as
a means to potentially decrease lipophilicity and improve properties.
Table 3. Structureꢀactivity relationships for racemic azabenzimidazoles
S1P1
Translocation
EC50 (µM)
Hu Microsomes
CLint (µL/min/mg)
Example
Structure
a
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 10 of 78
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
1
1
1
4
5
6
2.34 ± 1.06
1.17 ± 0.34
0.21 ± 0.15
94
93
35
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
7
0.31 ± 0.09
15
a
EC50 values are reported as the mean of at least three separate determinations ± standard
deviation
Earlier in the project, we had established that small, hydrophobic groups in the 3ꢀ and
ꢀpositions of the aryl sulfonamide ring were critical to maintaining measurable antagonist
4
activity, and the majority of our initial studies kept the 4ꢀchlorophenylsulfonamide moiety as
a constant to allow for direct comparisons between compounds with modifications elsewhere
on the scaffold. Given the increased potency with benzimidazole 2, however, we felt it was
timely to challenge these prior assertions in the hopes that a less lipophilic sulfonamide group
could be identified without resulting in a drastic loss of receptor antagonism. This work was
wellꢀsuited to a parallel synthesis campaign, given the facile and highꢀyielding formation of
the sulfonamide moiety (see Chemistry section). Therefore, employing the potent 6ꢀ
trifluoromethylꢀsubstituted compound 4 as a benchmark, we prepared a large number of
compounds from readily available sulfonyl chlorides, representative examples of which are
presented in Table 4.
Table 4. Effects of variation of the sulfonamide group
ACS Paragon Plus Environment

Page 11 of 78
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
S1P1
Translocation
EC50 (µM)
b
Hu Microsomes
CLint (µL/min/mg)
Example
18
R
cLogP
2.6
3.5
3.9
4.4
4.4
4.0
4.9
4.0
3.1
5.9
4.8
LLE
2.4
2.2
4.2
3.8
3.2
3.9
2.7
3.2
1.8
ꢀ0.2
1.8
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
a
9.0 ± 4.2
1.8 ± 0.52
<3
not tested
14
19
20
21
0.008 ± 0.002
0.007 ± 0.008
0.024 ± 0.004
0.012 ± 0.014
0.027 ± 0.005
0.057 ± 0.022
12.4 ± 5.1
26
22
23
H3C
>100
34
2
4
5
>100
>100
>100
>100
19
2
26
27
1.8 ± 0.45
28
0.24 ± 0.089
a
EC50 values are reported as the mean of at least three separate determinations ± standard
b
deviation; LLE = pEC50 – cLogP
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 12 of 78
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Analysis of the data for compounds in Table 4 reveals that while a range of
arylsulfonamides are tolerated, replacement of the aryl moiety with a cycloalkyl group (18,
1
9) results in a substantial loss of antagonist activity. With respect to substitution on the aryl
ring, the aforementioned preference for smaller, hydrophobic groups was reinforced here.
For example, the 4ꢀfluoro, 4ꢀmethyl, and 4ꢀcyano substituted compounds (21, 22, and 23,
respectively) have potency in the receptor internalization assay comparable to the 4ꢀchloro
analog (4). Incorporation of these moieties at the 3ꢀposition (24, 25) similarly results in
potent antagonists. Interestingly, even the unsubstituted phenyl sulfonamide 20 maintains
this level of potency. There do appear to be limitations with respect to the range of tolerated
substituents, however. For example, incorporation of a more polar amide group at the 4ꢀ
position (26) causes a >3000ꢀfold drop in measured potency. In addition, while the 4ꢀ
isobutylꢀsubstituted analog (27) is clearly outside the realm of acceptable lipophilicity, this
compound serves to illustrate that placement of larger groups on the aromatic ring results in a
loss of antagonist activity. Furthermore, comparison of the matched pair 21 and 28 highlights
that arenes without orthoꢀsubstituents are preferred.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
The data in Table 4 also suggests that the substitution pattern on the sulfonamide ring
can address the issues of metabolic instability observed within this series. For example,
changing the 4ꢀchlorophenyl ring of 4 to a 4ꢀfluoro or 4ꢀcyanoꢀsubstituted aryl (21 or 23)
results in reduced human microsomal CL , as does removal of the 4ꢀsubstituent altogether
int
(
20). These improvements cannot be simply attributed solely to lipophilicity reduction,
however, as comparably polar compounds with moieties prone to benzylic oxidation (22, 26)
have poorer stability.
While some of the compounds prepared in this effort had improved LLE relative the 4ꢀ
chlorophenylꢀsubstituted matched pair, we recognized that these compounds still were
relatively lipophilic in comparison to the triazoleꢀbased compounds. In order to introduce
ACS Paragon Plus Environment

Page 13 of 78
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
more favorable physical properties, therefore, we prepared additional matched pairs of 4
containing heteroaryl sulfonamides (Table 5). Unfortunately, the majority these compounds
were less effective antagonists than the corresponding arylsulfonamides. However, we were
pleased to observe that the 3ꢀpyridyl and 4ꢀpyridyl analogs (30 and 31) did maintain the level
of potency seen previously, suggesting that properties could be effectively attenuated with
selected heterocyclic sulfonamides without a substantial loss in antagonist activity. Similar
to our observations with the azabenzimidazoles (Table 3), however, these two compounds
were found to be cytochrome P450 inhibitors (CYP3A4/5 IC50 for 30 = 0.83 µM; CYP2C19
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
IC for 31 = 0.04 µM).
50
Table 5. Exploration of heteroaryl sulfonamides
S1P
Translocation
EC50 (µM)
1
b
Hu Microsomes
CLint (µL/min/mg)
Example
R
cLogP
3.6
LLE
1.4
4.0
4.1
3.0
2.5
a
29
30
31
32
9.2 ± 0.37
0.030 ± 0.005
0.026 ± 0.007
1.0 ± 0.27
<3
>100
45
3.5
3.5
2.9
28
33
0.42 ± 0.089
3.9
>100
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 14 of 78
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
3
3
3
4
5
6
0.59 ± 0.080
0.43 ± 0.074
0.80 ± 0.24
1.1 ± 0.073
4.0 ± 0.65
3.2
3.4
3.9
3.7
3.2
3.1
3.3
3.0
3.0
2.2
2.3
2.2
3.1
2.4
95
>100
>100
>100
43
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
37
38
39
0.65 ± 0.15
2.2 ± 0.79
58
40
15
a
EC50 values are reported as the mean of at least three separate determinations ± standard
deviation; LLE = pEC50 – cLogP
b
At the culmination of our investigation of various sulfonamide groups, the higher LLE
compounds (such as 20, 21, and 23) represented attractive candidates for exploratory in vivo
studies. These examples were among the most effective antagonists in the receptor
internalization assay prepared in the project at the time, and had improved microsomal
stability relative to most other compounds of comparable potency. In addition, we were
pleased to find that the instability of 4ꢀchloroꢀsubstituted analog 4 in the presence of human
microsomes (as well as human hepatocytes) did not translate to equally high turnover in
mouse or rat hepatocytes. Indeed, 4 has moderate in vivo clearance (33 mL/min/kg) and
acceptable oral bioavailability (43%) in rat. While metabolic instability in human would
clearly preclude further progression of this compound, these data suggested that 4 might be a
useful preclinical tool compound for rodent in vivo studies. Moreover, we profiled a number
of the more potent compounds from this series (including 4 and 21) in a cellular receptor
ACS Paragon Plus Environment

Page 15 of 78
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
internalization assay using mouse S1P , and observed a correlation with potency in the
1
1
human S1P assay (data not shown). Thus, we were confident that these compounds would
1
effectively act as S1P antagonists in mice if sufficient plasma concentrations could be
achieved.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
In preparation for in vivo experiments, we first wanted to accurately assess plasma
exposures at varying doses as well as tolerance of the animals to repeated dosing. Therefore,
compounds 4 and 21 were administered orally to mice at either 50 or 100 mg/kg, twice daily,
for a period of 5 days (Figure 2). We did not observe a significant increase in the plasma
exposure between the low and high dose for either compound, suggestive of poor solubility
and/or permeability limiting intestinal absorption for both compounds. More troubling,
however, we observed that plasma concentrations at 1 hour postꢀdose on day 5 were
markedly lower for both compounds than they were 1 hour postꢀdose on day 1.
Figure 2. Reduction in plasma exposures following repeat dosing. Female CB17 SCID mice
(
1
3 animals per cohort) were dosed orally with either compound 4 or 21, at either 50 mg/kg or
00 mg/kg twice daily, for 5 days. Plasma concentrations were measured at 1 hour after the
first dose on day 1 and day 5.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 16 of 78
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
In order to address the observation of decreased exposures upon repeat dosing, we first
ruled out pharmaceutical properties of the compounds and formulations used. No
degradation or change in concentration of the oral formulations was observed over time for
either compound, nor was there any observable change in solid form or particle size by
microscopic assessment. As a result, we considered whether the decreased exposure of these
compounds was occurring through increased expression of the cytochrome P450 enzymes
responsible for their metabolism. While the detailed nature of the oxidative metabolism in
mouse was not evaluated for these compounds, we recognized that a possible P450 induction
liability with this series was something that needed to be identified and addressed
immediately. Otherwise, the risk of not being able to achieve sufficient, durable plasma
concentrations would hamper testing in relevant preclinical disease models and toxicological
studies. In addition, while it is known that there are substantial interspecies differences in the
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
28,29
regulation of cytochrome P450 enzymes,
the possibility of a P450 induction risk in
humans (and the associated potential for adverse drugꢀdrug interactions) would likely
preclude advancement of these compounds towards clinical development.
To determine if this series carried a risk for P450 induction in humans, cultured primary
human hepatocytes were treated with varying concentrations of 4 for a period of 48 hours,
and were then treated with probe substrates known to be metabolized by specific isoforms of
cytochrome P450 (Figure 3). No significant increase in CYP1A2 or CYP2B6 activity was
observed when compared to hepatocytes treated with compounds known to be inducers of
these enzymes. However, 4 was found to result in a doseꢀdependent increase in CYP3A
activity, with a concentration of 2 µM resulting in an increase in metabolic activity
comparable to the reference compound rifampicin, a known potent inducer of CYP3A4 and
30
CYP3A5.
ACS Paragon Plus Environment

Page 17 of 78
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 3. Compound 4 induces an increase in CYP3A activity in cultured human
hepatocytes. Human hepatocytes were seeded onto collagenꢀcoated plates and were cultured
in a collagenꢀMatrigel sandwich configuration. On day 3 of culture, the cells were treated
with either DMSO (control), compound 4, or a reference compound known to cause
induction of various CYP isoforms (3ꢀmethylcholanthrene for CYP1A2, phenobarbital for
CYP2B6, and rifampicin for CYP3A4/5). After 48 hours of treatment, the cultures were then
incubated with probe substrates (phenacetin for CYP1A2, bupropion for CYP2B6, and
testosterone for CYP3A4/5), and the amount of marker metabolite formed (acetaminophen
for CYP1A2, hydroxybupropion for CYP2B6, and 6ꢀβꢀhydroxytestosterone for CYP3A4/5)
was quantified. The amount of CYP activity for 4 is expressed as a percentage of the activity
observed for the individual reference compounds.
To determine whether these observations were specific to compound 4, other
compounds from this series were subsequently profiled in human hepatocytes, and similarly
revealed varying degrees of increased CYP3A activity (data not shown). In light of these
results, remediation of the human CYP3A induction liability within this series became the
priority for ongoing chemistry activities. While the human hepatocyte activity protocol
described in Figure 3 clearly indicated the magnitude of the issue, this assay wasn’t
considered to be suitable for routine screening purposes due to the labor intensive nature of
the assay, as well as issues surrounding the availability and variable quality of human liver
tissue. Therefore, we needed a higher throughput assay in order to identify design parameters
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 18 of 78
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
that would impact P450 induction, as well as to assess whether progress towards reducing
this liability was being made.
It has been established that compounds affecting expression of the genes encoding
P450 enzymes can do so through interactions with certain nuclear hormone receptors, such as
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
3
1ꢀ33
34
the pregnane X receptor (PXR)
and the constitutive androstane receptor (CAR). These
receptors recognize xenobiotics through direct or indirect binding interactions, and in turn
affect expression of the metabolic enzymes that result in the clearance of such compounds
from the body. Cellꢀbased reporter gene assays that measure the increase in P450 gene
32,35,36
expression associated with PXR activation by small molecules have been described,
our strategy therefore was to utilize such an assay as a filter to allow for prioritization of
and
compounds to be screened for P450 activity in human hepatocytes. While the generation of
EC50 values for PXR activation, derived from multiple concentrations of each compound,
would allow for rigorous rankꢀordering of the compounds with respect to CYP3A induction
risk, we instead opted to evaluate the extent of PXR activation at a single concentration (10
µM) of each compound in order to generate a larger data set. The resulting data was
anticipated to help guide further compound design as well as to prioritize compounds for
further evaluation in human hepatocyte studies.
Upon screening a set of compounds from this series in a PXR gene reporter assay, we
observed a loose correlation between the extent of PXR activation (at 10 µM) and the
experimentallyꢀdetermined LogD (Figure 4A), with more lipophilic compounds resulting in
greater PXR activation. This is in accord with previous literature observations, and is
consistent with more polar compounds having less favorable interactions with the relatively
3
7
hydrophobic binding site of PXR. While there were compounds with lower LogD that did
result in significant PXR activation, it was clear from this data that the more lipophilic
compounds tested had an increased likelihood of this unwanted activity (and to a greater
ACS Paragon Plus Environment

Page 19 of 78
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
extent). Indeed, roughly 70% of the compounds with LogD > 2 had >30% PXR activation at
a concentration of 10 µM (relative to rifampicin), a threshold considered to be indicative of a
37
high risk of CYP3A induction in humans (Figure 4B).
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 4. (A) Correlation between experimentallyꢀdetermined LogD and extent of PXR
activation. Compounds with greater than 30% PXR activation at 10 µM were considered to
be at risk for CYP3A induction. (B) Compounds with LogD > 2 are more likely to have
>30% PXR activation at 10 µM.
Given the LogD range that these data suggested would afford the best chance of
inductionꢀfree compounds, we stringently employed calculated lipophilicity as a primary
filter for all newly proposed compounds prior to initiating synthesis. In order to drive down
1
lipophilicity without sacrificing the potent S1P antagonism within this series, we assessed
which previous structural permutations resulted in increased LLE, and utilized this
information to design compounds that combined such modifications. Therefore, based on the
promising data presented in Table 3, the focus of our efforts centered on revisiting
azabenzimidazoles. In particular, we felt that compounds in which the 5ꢀposition was a
nitrogen (as in 16) would largely retain potency while effectively lowering lipophilicity. In
addition, the exploration around the aryl sulfonamide ring (Table 4) suggested the 4ꢀcyano
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 20 of 78
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
group (as in 23) contributed the least lipophilicity amongst the higher LLE compounds.
Combining these groups gave 41 (Table 6), a compound that validated our design strategy to
address P450 induction, combining potent antagonist activity with decreased activation of
PXR. Given the low LogD for this compound, we hypothesized that addition of a moderately
hydrophobic substituent at the 6ꢀposition would further enhance potency while staying within
the desired property space. Moreover, we anticipated that the inclusion of a substituent at
this position, adjacent to the pyridyl nitrogen, might attenuate the P450 inhibitory activity
observed previously with azabenzimidazoleꢀbased compounds. Thus, we prepared 42ꢀ44,
and were pleased to find that these compounds had measured potency comparable to previous
leads (4 and 21), but with reduced LogD and PXR activation.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
In addition to the compounds resulting from these combinations, we were able to
further exploit our previous finding that 3ꢀ or 4ꢀpyridyl groups were tolerated as replacements
for the substituted phenylsulfonamide (as shown in Table 5). In particular, the 3ꢀpyridyl
moiety still allowed for the incorporation of a potencyꢀboosting 4ꢀsubstituent on this ring, and
thus combination of all these features (3ꢀpyridyl, 4ꢀcyano arylsulfonamide; 6ꢀsubstituted 5ꢀ
azabenzimidazole) led to compounds such as 45ꢀ47. These examples represent a substantial
improvement in compound quality compared to our earlier lead 4, demonstrating comparable
potent S1P
1
antagonist activity while having a reduction in LogD of over two orders of
magnitude.
Table 6. Compounds with reduced lipophilicity and PXR activation
S1P1
b
PXR
Activation
Example
Structure
Translocation
EC50 (µM)
LogD
c
a
41
0.034 ± 0.014
0.9
14
ACS Paragon Plus Environment

Page 21 of 78
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
4
4
4
4
4
4
2
3
4
5
6
7
0.009 ± 0.004
0.003 ± 0.002
0.006 ± 0.005
0.028 ± 0.014
0.007 ± 0.005
0.004 ± 0.004
1.4
1.6
1.9
1.0
1.3
1.6
18
32
27
11
15
20
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
a
EC50 values are reported as the mean of at least three separate determinations ± standard
b
c
deviation; Experimentally determined at pH 7.4; Expressed as a percentage of the activation
response seen with rifampicin when both compounds are used at a concentration of 10 µM.
Compounds 41ꢀ47 clearly demonstrate that the structureꢀactivity relationships for
interactions with S1P and PXR can diverge, and that potent S1P antagonists that do not
1
1
substantially activate PXR are accessible from this series. To rationalize this observation and
to potentially further inform our efforts at removing the CYP3A induction liability, we
obtained a crystal structure of the potent inducer 4 bound to human PXR (Figure 5). The
binding site occupied by this compound is indeed relatively nonpolar, and less lipophilic
compounds would be expected to have fewer favorable hydrophobic interactions in general.
In retrospect, however, our successful design approach to address PXR activation by
lowering overall lipophilicity could have serendipitously impacted specific binding
interactions with the receptor. For example, addition of a nitrogen atom at the 5ꢀposition of
the benzimidazole (as in 41ꢀ47) may disrupt contacts with the side chain of Met246.
Alternatively, our decision to switch to 4ꢀcyanoꢀsubstituted sulfonamides in order to reduce
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 22 of 78
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
LogD may have resulted in unfavorable steric interactions with Tyr306 in PXR, manifesting
in reduced activation in the gene reporter assay. Interestingly, compound 4 binds to PXR in a
“
bentꢀback” conformation, with the 4ꢀchlorophenylsulfonamide group engaged in πꢀstacking
interactions with both the benzimidazole ring as well as the indole side chain of Trp299. It is
likely that the introduction of nitrogen atoms into the arylsulfonamide and benzimidazole
moieties would impact the electron density in these rings, possibly reducing the ability of
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
38
these groups to participate in such πꢀstacking and resulting in decreased binding to the
receptor. Without detailed structural information regarding the nature of the interaction of
1
compounds from this series to S1P , however, the structural basis for these improvements in
PXR activation profile remains unclear.
Figure 5. Crystal structure of compound 4 bound to human PXR obtained at 2.3 Å resolution
(PDB ID 5A86). Coloration is as follows: compound 4 (yellow), Trp299 (green), Tyr306
(cyan), Met246 (brown).
Having identified a number of compounds with reduced PXR activation, we proceeded
to test whether these compounds in fact addressed the CYP3A induction liability in human
hepatocytes. At the time of our investigations, industry guidance on in vitro induction testing
ACS Paragon Plus Environment

Page 23 of 78
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
had suggested that compounds that cause increased CYP3A activity (at concentrations
approximating an in vivo Cmax) greater than 40% than that of rifampicin (at 10 µM) were
39
considered to be at risk for induction in vivo. A number of compounds meeting our PXR
activation criteria indeed resulted in <40% of the CYP3A activity caused by rifampicin in
human hepatocytes (Figure 6, blue circles). For compounds that did demonstrate increased
CYP3A activity (Figure 6, red circles), however, the singleꢀpoint PXR activation assay was
largely unpredictive, as a number of compounds perceived to be low risk for induction did
carry this liability. This may be attributed, at least in part, due to the variability inherent in
using single point PXR activation data or hepatocyte studies using liver tissue from different
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
40
donors. Alternatively, CYP3A induction resulting from activation of other xenobiotic
receptors (such as CAR) has not been considered in our approach. Nevertheless, we were
pleased to see that there were no “false negatives,” or compounds that were predicted to be
inducers in the PXR assay that did not show increased CYP3A activity. Thus, we felt
comfortable that by utilization of the PXR reporter gene assay at a single compound
concentration (instead of generating EC values) to prioritize compounds for further testing,
5
0
we had not unknowingly filtered inductionꢀfree compounds from further consideration.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 24 of 78
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 6. Correlation of PXR activation assay with CYP3A activity in human hepatocytes.
Compounds with <40% CYP3A activity observed with rifampicin (both compounds at 10
µM) are colored blue, while those with >40% activity are colored red.
At the culmination of our chemistry efforts, compounds 46 and 47 represented the top
candidates for continued evaluation due to their potency, physical properties, and
pharmacokinetic profiles in rat and dog (Table 7). These compounds did not show any
measurable inhibition of the hERG channel or of any of the major P450 isoforms, and the
human CYP3A induction risk was largely mitigated with compound 47, which did not show
any increase in CYP3A activity upon incubation with human hepatocytes. Compound 46 did
show a borderline increase (43%) in CYP3A activity compared to rifampicin in hepatocytes
from one donor, although on retest with a different batch of hepatocytes only a slight increase
(5%) in CYP3A activity was observed. In repeat dosing experiments in mouse (at 20 mg/kg)
and rat (at 25 mg/kg), neither compound showed appreciable decreases in plasma exposure (1
hour postꢀdose) between day 1 and day 5; higher doses (50 mg/kg) of these compounds in
ACS Paragon Plus Environment

Page 25 of 78
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
mice did result in observable exposure decreases over time, with the effect more pronounced
for 46 than for 47 (data not shown). Thus, despite known interspecies differences in
xenobioticꢀinduced P450 upregulation, by addressing the human P450 induction liability in
this series we had identified compounds and doses suitable for preclinical in vivo
pharmacological and toxicological evaluation.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Table 7. Physical properties and pharmacokinetic parameters for compounds 46 and 47
Compound 46
Compound 47
Solubility (µM, pH 7.4)
Plasma Protein Binding
765
>1000
5
3 (mouse)
32 (rat)
41 (mouse)
18 (rat)
(% free)
36 (dog)
50 (dog)
24 (human)
19 (human)
0
.3 (mouse)
2.1 (rat)
2.5 (mouse)
3.9 (rat)
Hepatocyte CLint
6
(
µL/min/10 cells)
2.4 (dog)
1.7 (dog)
1.2 (human)
0.2 (human)
CYP Inhibition (1A2, 2C9,
all >20 µM
>200 µM
all >20 µM
>200 µM
2
C19, 2D6, 3A4/5) IC50
hERG Inhibition (IC50
)
Hu Hepatocyte CYP3A
Induction (% activity
compared to rifampicin)
5, 43 (results from two
different donors)
0
Rat pharmacokinetics
IV dosing, 3 mg/kg
IV dosing, 4 mg/kg
CL (mL/min/kg)
8
10
V
dss (L/kg)
1/2 (h)
1.2
2.4
1.3
2.9
t
PO dosing (10 mg/kg)
PO dosing (10 mg/kg)
Cmax (µM)
Bioavailability (F%)
13.3
87
21.7
93
Dog pharmacokinetics
IV dosing, 2.5 mg/kg
IV dosing, 2 mg/kg
CL (mL/min/kg)
2
2
V
dss (L/kg)
1/2 (h)
0.6
4.9
0.7
4.6
t
PO dosing (10 mg/kg)
PO dosing (5 mg/kg)
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 26 of 78
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Cmax (µM)
Bioavailability (F%)
31.1
81
10.7
71
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 7. Inhibition of FGF/S1Pꢀstimulated angiogenesis by compounds 46 and 47. Matrigel
0.5 mL) containing 2 µg/mL βꢀFGF and 1 µM S1P was injected subcutaneously into the
ventral region of female 129s1/SvImJ mice. Compounds 46 and 47 were dosed orally at 10
mg/kg, once daily, on days 5ꢀ8 postꢀimplant (5 animals per cohort). Plasma
pharmacokinetics for both compounds was assessed 1 hour postꢀdose on day 4 (Supporting
Information, Figure S2). The plugs were then harvested, and the amount of hemoglobin per
gram of gel was quantified. Treated mice were compared with mice given vehicle used in the
dosed groups (0.5% HPMC, 0.1% Tween 80) as a control, as well as mice injected with
Matrigel not containing proangiogenic factors. The double asterisk (**) indicates p < 0.01.
(
To assess whether treatment with these compounds resulted in pharmacodynamic
effects expected from S1P
1
antagonism in vivo, we utilized a previouslyꢀdescribed Matrigel
plugꢀbased model that measured the ability of compounds to inhibit angiogenesis induced by
41
FGF and enhanced by S1P (Figure 7). In the absence of these growth factors, very little
angiogenesis (quantified by the amount of hemoglobin present in the Matrigel plug) occurs in
ACS Paragon Plus Environment

Page 27 of 78
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
this model, whereas the combination of these results in the formation of wellꢀdeveloped
4
1
vessels. In control mice treated with vehicle only, this manifests as a significant increase in
hemoglobin content in the plug (Figure 7 and Supporting Information, Figure S3). When
mice are treated once daily with 10 mg/kg of either 46 or 47 for 4 days, a statistically
significant reduction in hemoglobin content within the plugs was observed compared to the
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
control group, consistent with these compounds acting as S1P
1
antagonists and suppressing
siRNA in this
S1Pꢀpromoted angiogenesis. Similar observations have been made with S1P
1
10
model.
To further establish that our compounds were exerting pharmacological effects in vivo,
we determined whether they induced changes in capillary integrity using an Evans blue dye
lung leakage model. Evans blue dye (EBD) binds tightly to the albumin present in plasma,
and when vessel function is compromised an increase in EBD in tissue would be observed as
a result of leakage of plasma proteins from the vasculature. Several previous studies have
demonstrated an increase in EBD content in rodent lungs upon treatment with S1P1
antagonists, suggesting these compounds cause an increase in pulmonary vascular
20,23,42
permeability.
In our hands, when female NCr nude mice were treated orally with either
compound 46 (20 mg/kg) or 47 (50 mg/kg), plasma concentrations of both compounds at 1
hour postꢀdose are more than sufficient to expect target engagement based on in vitro potency
and relatively low plasma protein binding (Supporting Information, Figure S4). In this strain
of mice, these doses result in a substantial increase in EBD content in the lungs relative to
vehicleꢀtreated controls (Figure 8 and Supporting Information, Figure S5), consistent with
1
these compounds acting as S1P antagonists in vivo.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 28 of 78
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 8. Induction of capillary leakage in mouse lungs by compounds 46 and 47. Female
NCr nude mice (4 animals per cohort) were dosed orally with either compound 46 (20
mg/kg), compound 47 (50 mg/kg), or vehicle alone (0.5% HPMC, 0.1% Tween 80). After 1
hour, the mice were injected with a solution of Evans blue dye (20 mg/kg). After an
additional 30 minutes, the mice were anesthetized and perfused with 0.9% normal saline.
The lungs were harvested, and the amount of Evans blue dye present was determined by
spectrophotometry. The amount of dye present was normalized to the amount found in
vehicleꢀtreated animals. The double asterisk (**) indicates p < 0.005.
Having demonstrated that oral doses of compounds 46 and 47 achieve plasma
1
concentrations that result in functional effects related to S1P antagonism, we next assessed
whether these compounds were efficacious in tumor xenograft studies. A number of
xenograft models were explored, with a focus on those in which previous antiangiogenic
4
3,44
agents have demonstrated effects on tumor vasculature.
Despite the observations of in
vivo activity described above, only modest antitumor activity was observed with our
compounds across the models evaluated. For example, compound 47 demonstrated 32%
tumor growth inhibition in a Calu6 xenograft model when dosed orally at 50 mg/kg twice
daily, while compound 46 (at 20 mg/kg twice daily) did not result in statistically significant
reductions in tumor volume (Figure 9). In contrast to the observation of marked increases in
ACS Paragon Plus Environment

Page 29 of 78
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
capillary leakiness in murine lungs, we did not observe statistically significant increases in
EBD accumulation in tumor tissue obtained from either Calu6 or Colo205 xenograftꢀbearing
mice treated with these compounds (data not shown), suggesting that these agents do not
substantially affect vascular integrity within the tumor. Even though decreases in plasma
exposure (1 hour postꢀdose) over time were observed with 47, both compounds achieved
plasma exposures throughout the duration of these studies comparable to those observed in
the previouslyꢀdescribed in vivo experiments (Supporting Information, Figure S6).
Therefore, given the disparity between the activity we observed in the pharmacodynamic
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
models and the various tumor xenograft studies, we conclude that a selective S1P
1
antagonist
would have only minimal effects on tumor growth inhibition as a single agent.
1
Figure 9. Evaluation of antitumor activity of lead S1P antagonists in a Calu6 xenograft
model in female NcRꢀnude mice. (A) Compound 46, dosed orally at 20 mg/kg BID (n=14)
on days 13ꢀ32, does not result in statistically significant tumor growth inhibition relative to
vehicle (HPMC)ꢀtreated controls (n=20). (B) Compound 47, dosed orally at 50 mg/kg BID
(
(
n=14) on days 11ꢀ33, results in 32% tumor growth inhibition (p=0.008) relative to controls
n=20).
ACS Paragon Plus Environment