Rational approach to highly potent and selective apoptosis signal-regulating kinase 1 (ASK1) inhibitors

Article abstract of DOI:10.1016/j.ejmech.2017.12.041

Many diseases are believed to be driven by pathological levels of reactive oxygen species (ROS) and oxidative stress has long been recognized as a driver for inflammatory disorders. Apoptosis signal-regulating kinase 1 (ASK1) has been reported to be activ

Full text of DOI:10.1016/j.ejmech.2017.12.041

Accepted Manuscript
Rational approach to highly potent and selective apoptosis signal-regulating kinase 1
(ASK1) inhibitors
Frank Lovering, Paul Morgan, Christophe Allais, Ann Aulabaugh, Joanne Brodfuehrer,
Jeanne Chang, Jotham Coe, WeiDong Ding, Heather Dowty, Margaret Fleming,
Richard Frisbie, Julia Guzova, David Hepworth, Jayasankar Jasti, Steve Kortum, Ravi
Kurumbail, Shashi Mohan, Nikolaos Papaioannou, Joseph W. Strohbach, Fabien
Vincent, Katherine Lee, Christoph W. Zapf
PII:
S0223-5234(17)31068-1
10.1016/j.ejmech.2017.12.041
EJMECH 10019
DOI:
Reference:
To appear in: European Journal of Medicinal Chemistry
Received Date: 27 October 2017
Revised Date: 11 December 2017
Accepted Date: 13 December 2017
Please cite this article as: F. Lovering, P. Morgan, C. Allais, A. Aulabaugh, J. Brodfuehrer, J. Chang,
J. Coe, W. Ding, H. Dowty, M. Fleming, R. Frisbie, J. Guzova, D. Hepworth, J. Jasti, S. Kortum, R.
Kurumbail, S. Mohan, N. Papaioannou, J.W. Strohbach, F. Vincent, K. Lee, C.W. Zapf, Rational
approach to highly potent and selective apoptosis signal-regulating kinase 1 (ASK1) inhibitors, European
Journal of Medicinal Chemistry (2018), doi: 10.1016/j.ejmech.2017.12.041.
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT

ACCEPTED MANUSCRIPT
Rational Approach to Highly Potent and Selective Apoptosis Signal-
Regulating Kinase 1 (ASK1) inhibitors
Frank Lovering1, Paul Morgan2, Christophe Allais1, Ann Aulabaugh1, Joanne Brodfuehrer3, Jeanne Chang1,
Jotham Coe1, WeiDong Ding1, Heather Dowty4, Margaret Fleming2, Richard Frisbie1, Julia Guzova2, David
Hepworth1, Jayasankar Jasti1, Steve Kortum1, Ravi Kurumbail1, Shashi Mohan2, Nikolaos Papaioannou1,
Joseph W. Strohbach1, Fabien Vincent1, Katherine Lee1, Christoph W. Zapf1
1Medicine Design, 2Inflammation and Immunology, 3Biomedicine Design, 4Drug Safety Research and
Development
******************
Pfizer, Inc.
1 Portland Street, Cambridge, MA 02139, USA
445 Eastern Point Road, Groton, CT 06340, USA
E-mail: frank.lovering@pfizer.com
******************
1

ACCEPTED MANUSCRIPT
Abstract
Many diseases are believed to be driven by pathological levels of reactive oxygen species (ROS) and
oxidative stress has long been recognized as a driver for inflammatory disorders. Apoptosis signal-
regulating kinase 1 (ASK1) has been reported to be activated by intracellular ROS and its inhibition leads
to a down regulation of p38- and JNK-dependent signaling. Consequently, ASK1 inhibitors may have the
potential to treat clinically important inflammatory pathologies including renal, pulmonary and liver
diseases. Analysis of the ASK1 ATP-binding site suggested that Gln756, an amino acid that rarely occurs
at the GK+2 position, offered opportunities for achieving kinase selectivity for ASK1 which was applied
to the design of a parallel medicinal chemistry library that afforded inhibitors of ASK1 with nanomolar
potency and excellent kinome selectivity. A focused optimization strategy utilizing structure-based
design resulted in the identification of ASK1 inhibitors with low nanomolar potency in a cellular assay,
high selectivity when tested against kinase and broad pharmacology screening panels, and attractive
physicochemical properties. The compounds we describe are attractive tool compounds to inform the
therapeutic potential of ASK1 inhibition.
Introduction
The mitogen-activated protein kinase (MAPK) pathway is an important intracellular signaling system that
regulates diverse cellular functions, such as proliferation, differentiation, and apoptosis.1 Apoptosis
signal-regulating kinase 1 (ASK1), also known as mitogen-activated protein kinase kinase kinase 5
(MAP3K5), is an ubiquitously expressed member of MAPK kinase kinases regulating Jun N-terminal
kinase (JNK) and p38 signaling.2-4 ASK1 is preferentially activated in response to various stimuli including
reactive oxygen species (ROS), lipopolysaccharide (LPS), tumor necrosis factor-α (TNF-α), mitochondrial
dysfunction and endoplasmic reticulum (ER) stress.5-8 The redox regulation of ASK1 occurs via
thioredoxin (TRX) which is covalently bound to the N-terminal region of ASK1 through a disulfide bond.9
2

ACCEPTED MANUSCRIPT
Under conditions such as oxidative stress, ROS induced liberation of TRX from ASK1 leads to ASK1
homodimerization10 and the phosphorylation of the critical threonine residue (Thr838) in the activation
loop of ASK1.11
A number of studies using ASK1-deficient mice suggest that ASK1 has important functions in oxidative
stress-related diseases. ASK1−/− mice display a protective phenotype in multiple disease models including
asthma12, acute kidney injury13, neurodegenerative disorders8, cardiovascular14, 15, inflammatory16, and
metabolic disorders17. Thus, it is thought that small molecule compounds inhibiting ASK1 could be used
for the treatment of these pathologies. Recent studies suggest that ASK1 inhibitors may be protective in
rodent models of ischemia-reperfusion injury,18 dilated cardiomyopathy,19 diabetic nephropathy,20 and
hepatotoxic liver injury.21
Several ASK1 inhibitors have been disclosed in the public domain (Figure 1). Imidazopyridine 1 was
recently reported to inhibit ASK1 with an IC50 of 14 nM.22 Subsequently, additional ASK1 inhibitors have
been described in the patent literature with compounds 223 and 324 as representative examples from
their respective filings, while analog 4 was the sole compound reported in a patent application25 and
was recently reported to be in clinical trials as GS-4997/selonsertib for diabetic kidney disease and
nonalcoholic steatohepatitis (NASH).26
N
N
N
HN
O
N
HN
O
N
N
N
N
O
N
N
HN
N
1
2
N
N
O
Br
OH
HN
O
N
N
N
H
N
HO
N
F
3
4
Selonsertib
Figure 1: Representative ASK1 inhibitors 1-4 from the literature.
3

ACCEPTED MANUSCRIPT
Structural Analysis
We sought tool compounds with selective inhibition of ASK1 to help build our understanding of the
pharmacology of ASK1. An analysis of the ATP binding site of ASK1 was conducted to identify unique
structural features in order to enhance the kinome selectivity of new compounds. Targeting distinctive
kinase features such as binding pockets or side chains in the design of kinase inhibitors is a strategy to
increase kinome selectivity.
Figure 2: Protein crystal structure of literature compound 1 (PDB 3VW6)22 reveals an unusual Gln756 at
the GK+2 position of the hinge which engages in an intra-protein hydrogen bond with the backbone
carbonyl of Ile706.
An analysis of the ATP-binding site revealed that most of the amino acids in each position of the ASK1
active site frequently occur in other kinases as well. For example, 182 of the 442 kinases examined,
including ASK1, share a methionine as the “gatekeeper” residue (Figure 2). In contrast, Gln756 (at the
GK+2 position using the nomenclature of Ghose et al27) has a very low abundance in this particular
position of the ATP binding site of kinases. Only 3 kinases feature a Gln at the GK+2 position: ASK1
(MEKK5, MAP3K5), TAK1 (MAP3K7), and PIK3R4. Inspection of a crystal structure22 of imidazopyridine 1
bound in the active site of ASK1 revealed that the side-chain of Gln756 engages the backbone carbonyl
4

ACCEPTED MANUSCRIPT
of Ile706 in an intra-protein hydrogen bond. It was speculated that this interaction widens the channel
at the hinge, allowing the t-Bu phenyl moiety of compound 1 to move past the Gln756 side chain and
access space commonly inaccessible.
This distinguishing feature of the ASK1 pocket within the kinome was effectively demonstrated from the
overlay of crystal structures of ~4800 kinase inhibitors from the Pfizer database into the ATP active site
of ASK1 (Figure 3; For clarity Figure 3 was prepared by clustering the 4800 inhibitors to 1000 clusters
and the cluster centers were kept). It was noted that inhibitor 1 emerges from this ensemble of kinase
inhibitors to occupy this unusual selectivity channel adjacent to Gln756. This observation, in addition to
the fact that the GK+2 position is rarely populated by the polar amino acid glutamine led to the design
hypothesis that (1) accessing this unique channel, as observed for compound 1, with small molecules
while simultaneously (2) interacting with Gln756 at the GK+2 position would lead to potent and highly
selective ASK1 inhibitors.
5

ACCEPTED MANUSCRIPT
Figure 3: Protein crystal structure of literature compound 1 (PDB 3VW6)22 overlayed with approximately
1000 kinase inhibitor co-crystal structures from the Pfizer database. (A) illustrates the traditional view of
the kinase site with the p-loop labeled. (B) is rotated 180o to illustrate that imidazopyridine 1 occupies
rather unique space across the kinome (red circle).
Analysis of ASK1 crystal structures as well as an investigation of how compound 4 might bind were
undertaken to understand opportunities for engaging Gln756. To this end a docking model derived from
PDB 3VW622 was used to predict the binding mode of amide 4 in the ASK1 active site as shown in Figure
4. Instead of interacting with the hinge through the hydrogen bond donor and acceptor motif of the
amino-pyridine moiety, amide 4 was predicted to act as a single-point hinge binder with its amide
carbonyl engaging the NH of Val757. In light of a recent analysis on kinome selectivity, a single-point
hinge binding motif may be preferable to a two-point hinge binder to increase the probability of
achieving kinome selectivity.28
6

ACCEPTED MANUSCRIPT
In addition to the hinge contact, the docking model suggested a productive interaction of the triazole
with Lys709 and a non-classical hydrogen-bond interaction between the imidazole C5-H and the carbonyl
of Gln756 (3.3 Å). Having identified amide 4 as an inhibitor which features key criteria for its target
engagement such as being a single-point hinge binder and displaying an interaction with Gln756, efforts
were directed toward expanding the chemical space around this compound. Specifically we sought to
maintain the single point hinge binding characteristics while exploring which other moieties might
engage Gln756.
Figure 4: Docking pose of compound 4 in ASK1. Amide 4 is predicted by the docking model to interact
with the amide carbonyl of Val757 of the hinge through its amide carbonyl, the conserved Lys709
through the triazole N3, as well as a non-classical hydrogen bond between the carbonyl of Gln756 and
the imidazole C5-H.
Library design
A virtual amide library was enumerated by acylating the iPr-triazole-substituted 2-amino-pyridine
moiety with available carboxylic acids (Figure 6). The goal of this library was to identify ASK1 inhibitors
which would engage either rotamer of the carboxamide side-chain of Gln756, thus targeting either the
7

ACCEPTED MANUSCRIPT
NH2 or the C=O group with a range of functional groups. In addition, substituents were sought that
provided a vector to access the selectivity channel occupied by the tert-butyl phenyl moiety in amide 1.
Figure 5: Library design to identify ASK1 inhibitors targeting Gln756.
To this end, a virtual library of more than 14,000 amide analogs was enumerated and docked into the
ATP-binding site of ASK1 using a docking model derived from 1 (pdb code 3VW6) wherein each of the
two Gln756 rotamers were sampled. This effort led to the prioritization of 29 chemically diverse analogs
for synthesis which were expected to deliver highly potent and selective ASK1 inhibitors.
Synthesis of Potential Inhibitors
The compounds in Tables 1-3 were prepared by amide bond coupling of the appropriately substituted
benzoic acid fragments with 6-(4-isopropyl-4H-1,2,4-triazol-3-yl)pyridin-2-amine (5). Preparation of the
substituted benzoic acid fragments not commercially available is outlined in Schemes 1 and 2. 2-
Methoxy-5-(methylcarbamoyl)benzoic acid (9) was prepared by the amide coupling of 3-bromo-4-
methoxybenzoic acid (6) with methylamine to afford 7. Palladium catalyzed carbonylation of 7 under 50
psi CO in methanol / DMF (1:1) provided ester 8. The ester hydrolysis was accomplished using lithium
hydroxide in aqueous methanol to give compound 9 (Scheme 1). 2-Methoxy-4-methyl-5-
sulfamoylbenzoic acid (12) was prepared by treatment of 10 with thionyl chloride and chlorosulfonic
acid to afford 11. The sulfonyl chloride was dissolved in ammonium hydroxide and acidified to give 12
(Scheme 2). 16 was prepared by treating 11 with acetic acid at 90°C followed by reduction with tin
8

ACCEPTED MANUSCRIPT
metal in hydrochloric acid at 45°C to give the thiol 13. The thiol was alkylated with iodomethane in
acetone with potassium carbonate as base to afford the thioether 14. The thioether was oxidized with
potassium permanganate in a biphasic reaction to give 15 followed by ester hydrolysis to provide 16
(Scheme 2).
Scheme 1. Synthesis of Intermediate 9a
aReagents and conditions: (i) EDCI, HOBt, diisopropylethylamine, methylamine hydrochloride, DCM, rt,
18 h, 100%. (ii) trimethylamine, palladium acetate, 1,3-bis(diphenylphosphino)propane, DMF, MeOH, CO
50psi, 80°C, 16 h, 91%. (iii) lithium hydroxide, MeOH, water, rt, 1 h, 78%.
Scheme 2. Synthesis of Intermediates 12 and 16a
aReagents and conditions: (i) ClSO3H, DCM, thionyl chloride, -5 to -20°C, 5 h, 61%. (ii) NH4OH, 0°C, 5 min.,
43%. (iii) AcOH, 90°C, 10 min, Sn, HCl, 45°C-60°C, 2 h, 99%. (iv) MeI, K2CO3, acetone, 50°C, 16 h, 26%. (v)
9

ACCEPTED MANUSCRIPT
benzoic acid, benzyltriethylammonium chloride, DCM, KMnO4, water, 20°C, 3 h, 75%. (vi) LiOH
monohydrate, THF, water, EtOH, 20°C, 5 h, HCl, 100%.
Compounds 18-21 were synthesized by conversion of the appropriate acid to the acid chloride using
oxalyl chloride and subsequent addition of 5 in pyridine (Scheme 3). Compounds 25-27 were
synthesized in a library format by coupling the commercially available acids with 6-(4-isopropyl-4H-
1,2,4-triazol-3-yl)pyridin-2-amine using propylphosphonic anhydride (T3P) with triethylamine at 100°C
(Scheme 4).
Scheme 3. Synthesis of Compounds 18-21a
aReagents and conditions: (i) MeCN, oxalyl chloride, DMF, rt, 1 h. (ii) 6-(4-isopropyl-4H-1,2,4-triazol-3-yl)pyridin-2-
amine (5), pyridine, rt, 18 h, 6-99%.
Scheme 4. Synthesis of Compounds 25-27a
O
O
i
R
N
R
N
N
OH
N
H
N
22, R = EtSO2NH
23, R = CONH2
25, R = EtSO2NH
26, R = CONH2
24, R = 3-(1H-imidazol-2-yl)
27, R = 3-(1H-imidazol-2-yl)
aReagents and conditions: (i) T3P (50% in EtOAc), triethylamine, 5, 100°C, 16 h.
10

ACCEPTED MANUSCRIPT
Compound 31 was synthesized by reacting 5-amino-2-methoxybenzoate (28) with ethanesulfonyl
chloride in the presence of pyridine in DCM to give sulfonamide 29. Hydrolysis of 29 by aqueous sodium
hydroxide afforded the acid 30 which was converted to the acid chloride by heating in thionyl chloride.
Residual thionyl chloride was removed under reduced pressure and the crude acid chloride reacted with
5 in acetonitrile at room temperature. The resulting precipitate was recrystallized from methanol to
provide compound 31 (Scheme 5). Compound 36 was synthesized by the palladium catalyzed
carbonylation of 3-bromo-4-methoxybenzonitrile (32) in methanol under 60 psi CO to give ester 33.
Hydrolysis with lithium hydroxide and amide coupling with 5 using Mukaiyama’s reagent29 afforded the
nitrile 35 which was treated with hydrogen peroxide and potassium carbonate to provide 36 (Scheme 6).
Compound 38 was synthesized by converting the commercially available acid 2-methoxy-5-
(methylsulfonyl)benzoic acid (37) to the acid chloride with thionyl chloride and reacting with 5 in
pyridine at room temperature (Scheme 7).
Scheme 5. Synthesis of Compound 31a
aReagents and conditions: (i) pyridine, ethanesulfonyl chloride, rt, 2 h, 96%. (ii) NaOH, THF, water, rt, 4 h,
78%. (iii) thionyl chloride, 50°C, 1 h. (iv) 5, rt, 16 h, MeOH, 33%.
11

ACCEPTED MANUSCRIPT
Scheme 6, Synthesis of Compound 36a
aReagents and conditions: (i) triethylamine, Pd(dppf)Cl2, MeOH, CO (60 psi), 100°C, 4 h, 22%. (ii) lithium
hydroxide monohydrate, THF, water, rt, 4 h, 81%. (iii) 5, diisopropylethylamine, Mukaiyama’s reagent,
THF, 70°C, 4 h, 97%. (iv) DMSO, K2CO3, H2O2, rt, 30 min., 10%.
Scheme 7, Synthesis of Compound 38a
aReagents and conditions: (i) thionyl chloride, 1 h, 55°C. (ii) 5, pyridine, rt, 2 h, NaOH, 60°C, 6N HCl, rt, 1
h, 35%.
Compound 45 was synthesized by bromination of methyl-2-hydroxy-4-methyl benzoate (39) followed by
phenol alkylation with methyl iodide to give 41. Palladium catalyzed cyanation with zinc cyanide was
used to obtain 42. Ester hydrolysis using sodium hydroxide followed by amide coupling and treatment of
the nitrile with hydrogen peroxide and potassium carbonate afforded the primary amide 45 (Scheme 8).
Scheme 8, Synthesis of compound 45a
12

ACCEPTED MANUSCRIPT
O
O
O
O
O
Br
Br
NC
i
ii
iii
iv
O
O
O
O
OH
O
OH
39
40
41
42
O
O
O
O
O
NC
NC
v
N
N
vi
N
N
OH
H2N
N
H
N
N
N
H
N
N
O
O
43
44
45
aReagents and conditions: (i) Br2, CHCl3, 0°C, 3 h, 95%. (ii) MeCN, MeI, K2CO3, 40°C, 16 h,17%. (iii)
Zn(CN)2, Xantphos, Pd2(dba)3, N,N,N’,N’-tetramethylenediamine, DMF, 160°C, 200 s, 92%. (iv) MeOH,
water, NaOH, rt, 30 h, HCl, 95%. (v) MeCN, DMF, oxalyl chloride, 1 h, rt, 5, pyridine, rt, 2 h, 85%. (vi)
H2O2, K2CO3, DMSO, rt, 30 min., 5%.
In-Vitro Assays
Inhibitor efficacy was evaluated using a homogeneous time-resolved fluorescence (HTRF) assay that
measures the phosphorylation of a peptide substrate by recombinant human full-length ASK1. In order
to best reflect the physiological intracellular environment, a 1 mM concentration of ATP in the assay was
employed. As some compounds were potent and approached the limits of the assay as defined by the
concentration of enzyme (3 nM), the Kiapp was calculated by fitting the enzyme inhibition versus
compound concentration data using the Morrison equation (equation 9.6 in Copeland30) or the
competitive inhibition equation (equation 8.1 in Copeland30) for less potent inhibitors (see Experimental
Section). This approach allows a more accurate determination of affinity for the most highly potent
inhibitors approaching the functional limitations of the assay.31 For SAR interpretation purposes, the
Kiapp values were converted to IC50 values using the Cheng-Prusoff equation (equation 3 in Cheng and
Prusoff32) employing an experimentally determined ATP Km of 48 uM.33
To evaluate intracellular inhibitor potency, a new cell assay was developed. Human full-length ASK1 was
over-expressed in HEK293 cells containing a stress-activated luciferase reporter downstream of the
13

ACCEPTED MANUSCRIPT
transcription factor Activator Protein-1 (AP-1). AP-1 acts downstream of the stress-activated protein
kinase/Jun N-terminal kinase (SAPK/JNK), hence the expectation was that activation of overexpressed
ASK1 would lead to a robust reporter cell line for ASK1 kinase activity via MMK4 or MKK7.7 It was
discovered that upon expression, full-length ASK1 was constitutively active in the HEK293/AP-1luc cells,
and did not require exogenous stimulation for activation, as evidenced by the autophosphorylation of
ASK1 at Thr838 (Figure 5, lanes 1 and 2). This basal autophosphorylation was minimally detected in cells
overexpressing kinase-inactive ASK1 K709R (Figure 5, lane 3). ASK1 autophosphorylation was blocked by
ASK1 selective inhibitors such as 38 dosed at 10 µM (Figure 5, lane 4). Despite the constitutive activation
of ASK1 in the overexpressing cells, AP-1-induced luciferase expression was not detected in cells. This
was supported by the failure to detect phosphorylated JNK (data not shown). Instead, in this cell system,
constitutively activated ASK1 resulted in elevated phosphorylation of p38, the other ASK1-dependent
substrate via MKK3 and MKK6.7
Thus, phosphorylation of p38 was used to measure the intracellular potency of ASK1 inhibitors. In this
assay, HEK293/AP-1luc cells expressing human full-length ASK1 were incubated with compound for 18
hours and then lysed and the level of phospho-p38 was quantified using the HTRF assay.33
14

ACCEPTED MANUSCRIPT
Figure 6: Characterization of a cell line overexpressing human full-length ASK1.
Lane 1: HEK293/AP-1luc parental cell line; Lane 2: HEK293/AP-1luc cells overexpressing full-length ASK1;
Lane 3: HEK293/AP-1luc cells overexpressing catalytically inactive ASK1 K709R; Lane 4: HEK293/AP-1luc
cells overexpressing full-length ASK1 treated with 38 (10 µM).
Results and Discussion
Gratifyingly, several of the 29 prepared library analogs incorporating a variety of chemically diverse
substituents were found to display IC50 values in the recombinant ASK1 enzyme assay below 1 µM. Table
15

ACCEPTED MANUSCRIPT
1 highlights four key analogs from this initial library with the enzymatic potency, as well as lipophilicity
(SFlogD) and lipophilic efficiency (LipE) values shown.34, 35 Analogs 18 and 25-26 were attractive in that
they provided functional groups with synthetically straightforward design opportunities to explore
further potency enhancements and adjustment of physicochemical properties. Among the four analogs
shown in Table 1, sulfonamide 18 stood out with respect to its enzymatic potency and LipE. This analog
is about 30 times more potent than the reversed ethyl sulfonamide 25 which is reflected also in the LipE
value of 18 being 2 units higher than that of 25. Primary sulfonamide 18 was the only analog in the
library of 29 compounds which carries a methoxy group in the ortho-position to the amide. It was
hypothesized that this substituent led to a significant enhancement in potency as a result of rigidifying
the biaryl amide through an intra-molecular hydrogen-bond that lowered the rotational flexibility of the
molecule thus bringing the ground state conformation close to the bioactive conformation.
To ascertain whether the o-OMe substituent would consistently lead to a reproducible increase in
potency, the methoxy analogs corresponding to reverse sulfonamide 25 and primary amide 26 were
prepared and tested for their ability to inhibit ASK1 (Table 2). Indeed, the putative intramolecular
hydrogen bond established in analogs 31 and 36 resulted in a >200 fold and >500 fold enzyme potency
improvement compared to 25 and 26, respectively.
16

ACCEPTED MANUSCRIPT
Table 1: Representative analogs of the initial library.
O
N
N
N
R
N
H
N
Compound
R
ASK1 IC50 (nM)a
1.5 ± 0.87
SFlogDb
1.3
LipE
7.8
O
O
S
S
H2N
18
OMe
H
N
O
25
26
73 ± 27
1.7
1.2
5.7
5.4
O
O
580 ± 130
H2N
N
27
170 ± 38
2.4
4.7
N
H
aBiochemical assay with recombinant human full-length ASK1 in the presence of 1
mM ATP. All results are the mean of at least two independent determinations.
b Shake flask logD
It was recognized that the additional hydrogen bond donor substituents on analogs such as 18, 31 and
36 may have a detrimental effect on permeability. Methyl sulfone 38 was designed to minimize the H-
bond donor count while still displaying a functional group that could potentially engage Gln756 as
predicted in our docking model. Despite a 3-15 fold reduction in potency in the ASK1 enzyme assay
compared to the sulfonamides 18 and 31 as well as amide 36, the sulfone moiety of inhibitor 38
(IC50 = 5 nM), remained an attractive functional group to be considered for the design of ASK1 inhibitors
as the permeability as measured by RRCK cells (low efflux MDCK cells)36 was increased.
17

ACCEPTED MANUSCRIPT
Overall, the methoxy containing inhibitors are in a desirable lipophilicity range with SFlogD below
three34, 37 (Table 2) as measured in the shake flask logD (SFlogD) assay38 and have excellent LipE values.
These compounds are also characterized by low to moderate molecular weight and acceptable RRCK
permeability.36 As a result of their desirable physicochemical and ADME profile, these analogs were
advanced to the cell-based assay described above.
When evaluated in the cellular assay, analogs 18, 31, 36 and 38 are characterized by a significant shift in
potency compared to the biochemical enzymatic IC50 values (Table 2). As indicated above, the kinase
assay was carried out in the presence of 1 mM ATP to approximate intracellular ATP concentrations. As
a result, a modest enzyme-to-cell shift was expected, especially since the analogs appear to have
sufficient cell-permeability as predicted by the measured RRCK values. We postulate that this
unexpected shift is due the artificial nature of the engineered cells over expressing ASK1.
Table 2: Second generation ASK1 inhibitors.
18

ACCEPTED MANUSCRIPT
O
R
N
N
H
OMe
N
N
N
Compound
R
ASK1 IC50 (nM)a
1.5 ± 0.87
cell IC50 (nM)b
736 ± 272
SFlogDc
1.3
LipE
7.8
MW
416
RRCKd
2.0
O
O
18
S
H2N
H
N
31
36
38
19
0.32 ± 0.12
1.0 ± 0.46
5.0 ±2.9
41 ± 4.2
94 ± 4.5
140 ± 38
272
2.2
1.5
1.3
1.8
7.6
7.8
7.3
7.1
445
380
415
394
5.6
3.5
7.9
9.8
S
O
O
O
H2N
O
O
O
S
2.4 ± 1.3
MeHN
a Biochemical assay with recombinant human full-length ASK1 in the presence of 1 mM ATP. All results are the mean of at least
two independent determinations.
bHuman full-length ASK1 over-expressed in HEK293/AP-1luc cells measuring inhibition of p38 phosphorylation. All results are the
mean of at least two independent determinations except for compound 19.
cShake flask log D
d(10-6 cm*s-1)
To identify compounds with improved enzymatic and cellular potency using structure-based design,
several X-ray co-crystal structures of these inhibitors bound in the active site of ASK1 were obtained. To
this end, efforts were focused on sulfonamides, amides and sulfones represented by examples 18, 36
and 38. The reverse sulfonamide chemotype 31 was deprioritized in order to avoid the potential of
forming a reactive p-quinone intermediate under conditions of oxidative metabolism.
Crystal structures were obtained of sulfonamide 18 and amide 19 as a derivative of amide 36 (Figure 7A
and B). The precise rotameric orientation of amide 36 was of significant interest to understand the
compound’s interaction with the protein. As it would be difficult to assign the rotomer of 36, the N-
methyl amide 19 was prepared and utilized for the X-ray crystal structure experiment along with
sulfonamide 18.
19

ACCEPTED MANUSCRIPT
Figure 7 shows the X-ray co-crystal structures for sulfonamide 18 and N-Me amide 19 bound to the ATP-
binding site of ASK1. The crystal structures confirmed that the analogs in this series are in fact single-
point hinge binders as predicted by our docking model. The amide carbonyl of the inhibitors engages the
backbone NH of Val757. As intended, the chemotypes interact with either rotamer of the carboxamide
of Gln756 depending on whether a hydrogen bond donor (Figure 7A) or acceptor (Figure 7B) is proximal
to the carboxamide side-chain residue. In addition to the Gln756 interaction, the inhibitors also engage
Gly759, either through its backbone NH in case of the sulfonamide 18 or through its backbone carbonyl
in case of the substituted amide 19.
Figure 5: X-ray co-crystal structures of analogs 18 (A) and 19 (B).
Having confirmed that these ASK1 inhibitors act as single point hinge binders demonstrating the
predictive capabilities of the docking model, X-ray protein crystal structure information was used to
guide efforts to design more potent inhibitors. As previously described, sulfonamides such as 18 as well
as amides such as 19 interact with Gly759. It was hypothesized that the strength of this hydrogen bond
could be enhanced by shielding it with a lipophilic substituent on the 4-position of the methoxy-
substituted benzamide ring.39 To test this hypothesis the corresponding 4-methyl analogs of 18, 36 and
20

ACCEPTED MANUSCRIPT
38 were prepared (Table 3). Improvements in enzyme and cell potencies were observed when
comparing sulfonamides 18 with 20 as well as amides 36 with 45. The potency increased as a result of
the addition of the methyl group in the case of the amides 45 and 36 was essentially lipophilic neutral as
evidenced by the LipE values of 8.0 vs 7.8. In contrast, sulfonamide 20 benefitted significantly more from
the additional methyl substituent beyond the increase in logD: its LipE value improved by a full unit
when compared to sulfonamide 18. This level of potency improvement clearly indicates that the 4-Me
group in the context of the sulfonamide substituent has a beneficial impact beyond increasing the
lipophilicity. However one must be careful in attributing the cause of the improved activity. Torsion
scans of the sulfonamides 18 and 20 and amides 36 and 45 were undertaken using quantum mechanics
(QM) as well as molecular mechanics (MM) to assess the effect of the methyl group on the energy of the
bound conformation (see Supplementary Materials, Figures SI-2 and SI-3 for corresponding dihedral
plots). The torsion scans suggest that the addition of the methyl group ortho to the sulfonamide to
afford 20 may have little effect (QM) to a 0.8 kcal stabilization (MM) of the conformational energy. In
the case of the amide, ortho methylation is predicted to destabilize the bound conformation by 0.8 kcal
(QM) to 0.21 kcal (MM). Thus in the case of the sulfonamides the beneficial effect of the methyl group
may not have been compromised by torsion strain, whereas in the case of the amide 45 the
conformational effect may have negated any benefit of shielding the H-bond. Sulfone 38 did not benefit
from the addition of the methyl group, as sulfone 21 is essentially equipotent to 38 in the enzyme
inhibition assay and only shows a twofold improvement in cell potency.
21

ACCEPTED MANUSCRIPT
Table 3: Effect of 4-methyl substitution on ASK1 potency.
While the additional methyl group in analogs 20 and 45 had a beneficial impact on the enzyme potency
of up to 15 fold and cell potency of up to 9 fold it did not have a significant impact on the ratio of
enzymatic to cellular IC50. The sulfonamides, which are characterized by the lowest permeability as
measured by their RRCK values, display the largest shift between the two assays. Conversely, the
sulfones, which have the highest permeability, also have the lowest enzyme to cell ratio. In an absolute
sense, amide 45 shows the best cell potency among this cohort of ASK1 inhibitors, presumably a
composite effect of its enzyme potency and its cell permeability.
At the outset of targeting the two Gln756 side-chain rotamers, it was unclear which rotational isomer
would lead to more potent ASK1 inhibitors. While such a preference might have been anticipated, the
22

ACCEPTED MANUSCRIPT
cohort of ASK1 inhibitors in Table 3 did not provide any guidance regarding which of the Gln756 amide
side chain rotamers is preferred. The results suggest that engaging either the carbonyl or the NH2 group
of the Gln756 side chain can lead to ASK1 inhibitors with sub nM biochemical potency as exemplified by
sulfonamide 20 and amide 45.
It was highlighted that the ASK1 inhibitors described herein were designed to productively engage
Gln756. The successful outcome of this effort was confirmed by X-ray crystallography. As a consequence
of the rather low occurrence of Gln in this position of the ATP-binding site across the kinome, it was
speculated that a productive interaction with this amino acid would lead to selective ASK1 inhibitors.
Figure 8 depicts the selectivity profile of the analogs described in Table 3 in a standard selectivity panel
of 40 kinases.40 Broadly, this set of potent ASK1 inhibitors displays a very high degree of selectivity when
assayed at 1 µM in the presence of 1 mM ATP to mimic physiologically relevant conditions. The less
potent analogs 18, 36 and 38 display a very clean overall kinome profile in this panel with a low to
moderate signal for CHEK2 which increases for the more potent analogs 20-21 and 25.
Figure 6: Kinome selectivity profiles for selected analogs using a standard Pfizer kinome selectivity
panel. Compounds were tested at 1 µM concentration in the presence of 1 mM ATP. The cells are
23

ACCEPTED MANUSCRIPT
colored on a sliding scale wherein zero percent inhibition is green, 50% inhibition is yellow and 100%
inhibition is red.
In order to study the pharmacokinetic (PK) profile of this chemotype, compounds 18 and 38 were
selected for a rat PK study. Methylsulfone 38 in particular is characterized by the lowest lipophilicity
and highest permeability among the ASK1 inhibitors in Table 3 while having low in vitro clearance as
measured in rat hepatocytes41 (Table 4). Compound 18 has comparable lipophilicity and in vitro
clearance but poorer permeability. However, following a 1 mg/kg IV dose, both compounds 18 and 38
were found to have moderate in vivo clearance in rat (Table 4, see Supplementary Materials for protocol
details). When dosed orally as a solution to minimize kinetic dissolution effects, both compounds 18 and
38 were shown to have low bioavailability (F = 1% and 8% respectively). It may be that though the
compounds were dosed as solutions, the low thermodynamic solubility42 of each compound hindered
absorption, but the precise source of the poor bioavailability could not be determined.
Table 4: Key in vitro and in vivo ADME parameters for 18 and 38.
Compound
18
38
SFlogD (pH 7.4)
1.2
1.3
thermodynamic solubility
1.2
5
pH 7.4 (µM)
RRCK (10-6 cm•s-1)
2.04
< 6
7.9
<6
rat hepatocyte Cl
(mL•min-1•10-6 cells)
rat IV Cl (mL•min-1•kg-1)
rat F (mg/kg)
35
41
< 1% (5)
8% (1 )
Conclusions
24

ACCEPTED MANUSCRIPT
A careful analysis of the ATP-binding site of ASK1 was carried out which identified the relatively unique
Gln756 position and thus provided an intriguing design hypothesis for the identification of selective
ASK1 inhibitors. An investigation of how compound 4 binds to ASK1 led to the hypothesis that it is a
single point hinge binder via the carbonyl oxygen. This binding hypothesis was later validated by protein
crystallography. This hypothesis informed the design of a virtual library of ~14K amides which was
enumerated in silico and triaged based on the analogs’ ability to engage either one of the two rotamers
of the Gln756 amide side chain. Several of the 29 amides targeted for synthesis were found to have IC50
values below 1 µM when tested in the full-length ASK1 biochemical kinase assay in the presence of 1
mM ATP as exemplified by amide 18 with an IC50 of 2 nM and LipE of 7.8. Subsequent SAR work led to
the discovery of a series of sulfonamides, amides and sulfones with sub-nanomolar potencies when
assayed against the ASK1 enzyme. In addition to displaying attractive enzymatic activity, several analogs
were found to be highly potent in a cell-based assay measuring inhibition of phosphorylation of p38, a
key kinase downstream of ASK1 involved in the regulation of inflammation and oncology pathways.
Though potent in the cell-based assay, there was a large right shift between enzyme and cellular
potency potentially due to the limited cell permeability as predicted by the low RRCK and over
expression of ASK1 in HEK293 cells. In addition to excellent potency, these inhibitors deliver a high
degree of kinome selectivity when assessed in a panel of 40 kinases. While lead compounds from this
program displayed encouraging in vitro ADME properties oral bioavailability in rat was poor. Further
efforts will have to be undertaken to address the series’ low bioavailability (as exemplified by
compounds 18 and 38). However, the ASK1 inhibitors reported here constitute attractive tool
compounds with the potential to support the elucidation of the pharmacological consequences of ASK1
inhibition.
25

ACCEPTED MANUSCRIPT
Experimental section
Synthesis information
General Methods
Unless specified, all solvent were purchased as anhydrous sealed bottles and used as received. Standard techniques
for handling air-sensitive compounds were employed for indicated operations. Removal of solvents was
accomplished on a rotary evaporator at reduced pressure. The purity of compounds of the final compounds 18-21,
25-27, 31, 36, 38 and 45 was determined by either HPLC or UPLC analysis and determined to be >95% pure.
Low-Resolution Mass Spectrometry analyses were conducted on Waters Acquity UPLC (Acquity Binary Solvent
Manager, 2777C-Autosampler, Acquity PDA, Acquity ELS and Acquity Column Manager) and Waters Acquity SQ
systems from Waters Corporation, Milford, MA. Signal acquisition conditions included: Waters Acquity HSS T3,
2.1mmx50mm, C18, 1.7µm; Column Temperature 60 °C as the column; 0.1% formic acid in water (v/v) as the
mobile phase A; 0.1% formic acid in acetonitrile (v/v) as the mobile phase B; 1.25mL/min as the flow and ESCI
(ESI+/-, APCI+/-), 100-2000m/z scan, 0.4sec scan time, Centroid as the MS method.
High-Resolution Mass Spectrometry analyses were conducted on an Agilent 6220 TOF mass spectrometer (Agilent
Technologies, Wilmington, DE) in positive or negative electrospray mode. The system was calibrated to greater than
1ppm accuracy across the mass range prior to analyses according to manufacturer’s specifications. The samples
were separated using UHPLC on an Agilent 1200 (Agilent Technologies, Wilmington, DE) system prior to mass
spectrometric analysis. The resulting spectra were automatically lockmass corrected and the target mass ions and
any confirming adducts (Na+, NH4+) were extracted and combined as a chromatogram. The mass accuracy was
calculated for all observed isotopes against the theoretical mass ions derived from the chemical formula using
MassHunter software (Agilent Technologies, Wilmington, DE).
All NMR spectra were collected on either a Bruker 400 Avance III with a 5 mm BBFO probe (400 MHz for 1H; 101
MHz for 13C) or a Bruker 500 Avance III HD with a 5 mm BBO Nitrogen cryoprobe (500 MHz for 1H; 126 MHz for
13C). The proton signal for non-deuterated solvent (δ 7.27 for CHCl3, δ 2.50 for DMSO, δ 3.31 for MeOH) was used
1
as an internal reference for H NMR spectra. For 13C NMR spectra, chemical shifts are reported relative to the δ
26

ACCEPTED MANUSCRIPT
77.00 resonance of CDCl3 or δ 39.51 resonance of DMSO-d6. Deuterated solvents (CDCl3, CD3OD and DMSO-d6)
were purchased from Cambridge Isotope Laboratories Inc. and used as received.
27

ACCEPTED MANUSCRIPT
Procedures
2-methoxy-N-{6-[4-(propan-2-yl)-4H-1,2,4-triazol-3-yl]pyridin-2-yl}-5-sulfamoylbenzamide (18). To a mixture of 2-
methoxy-5-sulfamoylbenzoic acid (17, 500.0 mg, 2.16 mmol) in acetonitrile (10.0 mL) was added DMF (90 mg, 1
mmol, 0.1 mL) followed by oxalyl chloride (297 mg, 2.27 mmol, 0.204 mL). The mixture was stirred at ambient
temperature under nitrogen for 1 h. In a separate flask a mixture of 6-[4-(propan-2-yl)-4H-1,2,4-triazol-3-yl]pyridin-
2-amine25 (440 mg, 2.16 mmol) and pyridine (3.0 mL, c=0.2 M) was stirred at ambient temperature. After 1 h the
slurry of the amine in pyridine was added to the acid chloride reaction resulting in a clear yellow solution. The
mixture was stirred under nitrogen at ambient temperature for 4.5 h, at which point it became a slurry. The solids
were filtered, washed with acetonitrile and dried under vacuum to give a crude amorphous material. In an open
flask, the solids (~400 mg) were refluxed in ethanol (450 mL) and water (20 mL) until the volume reached ~ 100
mL. Further ethanol (300 mL) was added and the operation was repeated until a volume of ~ 50 mL. The mixture
was then allowed to slowly cool down to ambient temperature overnight, and the resulting crystalline solids were
filtered and dried under vacuum to provide the title compound (384 mg, 43%). 1H NMR (400 MHz, DMSO-d6): δ
10.69 (s, 1H), 8.89 (s, 1H), 8.29 (d, J=8.2 Hz, 1H), 8.26 (d, J=2.3 Hz, 1H), 8.05 (t, J=8.0 Hz, 1H), 8.01 - 7.96 (m,
1H), 7.90 (d, J=7.4 Hz, 1H), 7.43 (d, J=8.6 Hz, 1H), 7.38 (s, 2H), 5.56 (td, J=6.6, 13.3 Hz, 1H), 4.02 (s, 3H), 1.50 (d,
J=6.6 Hz, 6H). 13C NMR (500 Mhz, DMSO-d6): δ 163.77, 159.56, 151.20, 146.72, 143.81, 140.40, 137.11, 131.00,
128.70, 123.60, 119.75, 114.61, 113.39, 57.33, 48.64, 23.67 MS m/z 417 [M+H]+
3-(ethylsulfonamido)-N-(6-(4-isopropyl-4H-1,2,4-triazol-3-yl)pyridin-2-yl)benzamide (25). To a solution of 3-
(ethylsulfonamido)benzoic acid (22, 60 µmol) and 6-[4-(propan-2-yl)-4H-1,2,4-triazol-3-yl]pyridin-2-amine (5, 50
µmol) in EtOAC (1 mL) was added T3P (50% in EtOAc, 1 mL) and triethylamine (207 µmol). The vial was placed
under a positive atmosphere of nitrogen, capped and shaken at 100 °C for 16 h. The reaction was concentrated under
reduced pressure and purified by HPLC ( Phenomenex Gemini C18 250×21.2mm*8µm; Acetonitrile-NH4OH (pH
10); gradient from 0% to 40% in 8 minutes, flow rate 35 mL/min). MS m/z 415 [M+H]+
N-{6-[4-(propan-2-yl)-4H-1,2,4-triazol-3-yl]pyridin-2-yl}benzene-1,3-dicarboxamide (26) Compound 26 was
prepared following the procedure for compound 25 using 3-(aminocarbonyl) benzoic acid (23) and was purified by
28