Development, Optimization, and Structure-Activity Relationships of Covalent-Reversible JAK3 Inhibitors Based on a Tricyclic Imidazo[5,4- d]pyrrolo[2,3- b]pyridine Scaffold

Article abstract of DOI:10.1021/acs.jmedchem.8b00571

Janus kinases are major drivers of immune signaling and have been the focus of anti-inflammatory drug discovery for more than a decade. Because of the invariable colocalization of JAK1 and JAK3 at cytokine receptors, the question if selective JAK3 inhibition is sufficient to effectively block downstream signaling has been highly controversial. Recently, we discovered the covalent-reversible JAK3 inhibitor FM-381 (23) featuring high isoform and kinome selectivity. Crystallography revealed that this inhibitor induces an unprecedented binding pocket by interactions of a nitrile substituent with arginine residues in JAK3. Herein, we describe detailed structure-activity relationships necessary for induction of the arginine pocket and the impact of this structural change on potency, isoform selectivity, and efficacy in cellular models. Furthermore, we evaluated the stability of this novel inhibitor class in in vitro metabolic assays and were able to demonstrate an adequate stability of key compound 23 for in vivo use.

Full text of DOI:10.1021/acs.jmedchem.8b00571

Subscriber access provided by University of Winnipeg Library
Article
Development, Optimization and Structure–Activity
Relationships of Covalent-Reversible JAK3 Inhibitors Based
on a Tricyclic Imidazo[5,4-d]pyrrolo[2,3-b]pyridine Scaffold.
Michael Forster, Apirat Chaikuad, Teodor Dimitrov, Eva Döring, Julia Holstein, Benedict-Tilman
Berger, Matthias Gehringer, Kamran Ghoreschi, Susanne Müller, Stefan Knapp, and Stefan A. Laufer
J. Med. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.jmedchem.8b00571 • Publication Date (Web): 31 May 2018
Downloaded from http://pubs.acs.org on May 31, 2018
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 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 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.
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 64
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Development, Optimization and Structure–Activity
Relationships of CovalentꢀReversible JAK3
Inhibitors Based on a Tricyclic Imidazo[5,4ꢀ
d]pyrrolo[2,3ꢀb]pyridine Scaffold.
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
†
§,&
†
†
#
Michael Forster , Apirat Chaikuad , Teodor Dimitrov , Eva Döring , Julia Holstein , Benedictꢀ
§
,&
†
#
,&
Tilman Berger , Matthias Gehringer , Kamran Ghoreschi , Susanne Müller , Stefan
§
,&,%
†
Knapp
, Stefan A. Laufer *
†
Department of Pharmaceutical / Medicinal Chemistry, Eberhard Karls University Tübingen, Auf
der Morgenstelle 8, 72076 Tübingen, DE
&
Institute for Pharmaceutical Chemistry, Johann Wolfgang Goethe University, MaxꢀvonꢀLaueꢀ
Str. 9, Dꢀ60438 Frankfurt am Main, DE
§
Structure Genomics Consortium, Johann Wolfgang Goethe University, Buchmann Institute for
Molecular Life Sciences, MaxꢀvonꢀLaueꢀStr. 15, Dꢀ60438 Frankfurt am Main, DE
#
Department of Dermatology, University Medical Center, Eberhard Karls University Tübingen,
Liebermeisterstr. 25, 72076 Tübingen, DE
%
German Cancer Consortium, DKTK, Site Frankfurt/Mainz, DE
ACS Paragon Plus Environment
1

Journal of Medicinal Chemistry
Page 2 of 64
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
KEYWORDS: Janus kinases, kinome selectivity, covalentꢀreversible inhibitor, inflammation
ABSTRACT. Janus kinases are major drivers of immune signaling and have been the focus of
antiꢀinflammatory drug discovery for more than a decade. Because of the invariable coꢀ
localization of JAK1 and JAK3 at cytokine receptors, the question if selective JAK3 inhibition is
sufficient to effectively block downstream signaling has been highly controversial. Recently, we
discovered the covalentꢀreversible JAK3 inhibitor FMꢀ381 (23) featuring high isoform and
kinome selectivity. Crystallography revealed that this inhibitor induces an unprecedented binding
pocket by interactions of a nitrile substituent with arginine residues in JAK3. Herein we describe
detailed structure activity relationships necessary for induction of the arginine pocket and the
impact of this structural change on potency, isoform selectivity and efficacy in cellular models.
Furthermore, we evaluated the stability of this novel inhibitor class in in vitro metabolic assays
and were able to demonstrate an adequate stability of key compound 23 for in vivo use.
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. The four Janus kinases JAK1, JAK2, JAK3 and TYK2 form a small
subfamily of cytosolic tyrosine kinases with pivotal importance for the signal transduction of
1
interleukins (ILs), interferons (IFNs) and other cytokines. As nonꢀreceptor tyrosine kinases,
JAKs bind to the activated cytokine receptors transmitting extracellular signals by
phosphorylation of signal transducers and activator of transcription (STAT) proteins. The
phosphorylated STAT proteins dimerize, translocate to the nucleus and act as transcription
2
factors affecting many cellular functions. Although each JAK fulfills important functions in
hematological and immunological pathways, the JAK3 isoform features some unique properties
compared to the other family members. While JAK1, JAK2 and TYK2 are found in many
different cell types and tissues, the expression of JAK3 is mainly restricted to leukocytes having
ACS Paragon Plus Environment
2

Page 3 of 64
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
3
a key role in the development and maturation of Tꢀcells, Bꢀcells and natural killer (NK) cells.
Therefore, JAK3 knockꢀout or lossꢀofꢀfunction mutations in the kinaseꢀ or receptorꢀgenes result
in the phenotype of severe combined immunodeficiency (SCID), a general immunological
disorder characterized by a complete absence of Tꢀ and NKꢀcells. Besides these serious
consequences in the immune system, SCID patients are not affected by other defects highlighting
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
4
ꢀ5
the isolated role of JAK3 in immune signaling. This suggests that JAK3ꢀselective inhibition
may be a valuable therapeutic strategy to achieve immunosuppression with less adverse effects
6ꢀ7
compared to known immunosuppressive panJAK inhibitors. However, none of the disease
causing JAK3 mutations are localized to the catalytic domain and the invariant coꢀlocalization of
JAK3 with JAK1 suggests that inhibition of JAK3 catalytic activity may not be sufficient for full
8
inhibition of this signaling pathway. In recent years it has been therefore fiercely discussed
among medicinal chemists whether specific inhibition of JAK3 catalytic activity would be
sufficient to achieve immunosuppression or if concomitant inhibition of JAK1 is required to
9
ꢀ11
abrogate cytokine signaling.
Mouse models largely confirmed the SCID phenotype, but
catalytically inactive JAK3 variants that could shed light on the scaffolding role of JAK3 in the
activated receptorꢀkinase complexes have not been generated in mice so far. To validate the
usefulness of selective inhibition, much effort has been undertaken in the last decade by industry
12ꢀ13
as well as academia to develop isoform specific inhibitors
and a number of inhibitors with
14
different selectivities within the JAK family are under clinical investigation. Currently, the
most prevalent inhibitors for interrogating JAK dependent signaling are baricitinib and
tofacitinib (1, Figure 1), which are both approved as secondꢀline treatment for rheumatoid
1
5
arthritis. Contrary to initial reports claiming it to be JAK3 specific, 1 is a panJAK inhibitor
1
0
making it unsuitable to address the JAK1 vs. JAK3 signaling issue. Another prominent JAK3
ACS Paragon Plus Environment
3

Journal of Medicinal Chemistry
Page 4 of 64
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
inhibitor is NIBR3049 (2, Figure 1), which shows a more favorable isoform selectivity profile
with a 130 to 1000ꢀfold specificity for JAK3 vs. JAK1, JAK2 and TYK2, respectively. However,
this compound potently inhibits several offꢀtargets within the kinome (namely PKCα, PKCθ and
1
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
GSK3β) in the lower nanomolar range. Due to the high structural similarity of the four JAKs,
the design of isoform specific inhibitors is a demanding task. One of the few essential structural
differences is a cysteine residue (C909) located in the solvent exposed front part of the ATP
binding pocket, which is exclusively found in JAK3. The latter is replaced by a serine in the
17
three other isoforms of the kinase family. The presence of a cysteine in the proximity of the
ATP binding site provides the opportunity to achieve selectivity by covalent targeting with
1
8ꢀ19
electrophilic inhibitors.
This concept was pursued by many groups in the last years and the
highly selective JAK3 inhibitors which originated from these efforts have been recently
2
0ꢀ26
reviewed.
JAK3 inhibitor with an outstanding selectivity within the JAK family and the whole kinome.
This compound has been included in the Structural Genomics Consortium’s (SGC) Chemical
In 2016, we reported the discovery of FMꢀ381 (23, Figure 1), a covalentꢀreversible
2
7
28
29
Probe pool and has been rated as a “4ꢀStarꢀProbe” in the Chemical Probe Portal as a selective
tool for the elucidation of JAK3 dependent signaling. In this paper we provide an expanded
insight into the structure activity relationships (SARs) and further investigations of
physicochemical and metabolic properties of the underlying inhibitor class.
ACS Paragon Plus Environment
4

Page 5 of 64
Journal of Medicinal Chemistry
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 1. Structures of known reference JAK inhibitors discussed in this study.
ACS Paragon Plus Environment
5

Journal of Medicinal Chemistry
Page 6 of 64
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 2. (a) Compound 23 bound to JAK3 (PDB: 5LWM) (b) Compound 1 bound to JAK3
(PDB: 3LXK). Comparison of the protein surfaces surrounding C909 (illustrated as ball/stick)
shows that 23 induces a distinct binding pocket by interaction of the nitrile substituent and R911.
The cavity is generated by significant rearrangement of the sidechains of R911 and R953.
RESULTS AND DISCUSSION. JAK3 Structure Activity Relationships. During the
analysis of the Xꢀray structure of JAK3 in complex with compound 23 (PDB: 5LWM) we
ACS Paragon Plus Environment
6

Page 7 of 64
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
discovered a new binding pocket induced by the nitrile substituent of the inibitor’s αꢀ
cyanoacrylamide warhead (Figure 2a). This cavity is located in close proximity to the covalently
targeted C909 and is formed by significant rearrangement of two arginine sidechains (R911 and
R953). So far, this inducedꢀfit binding pocket was only observed in the crystal structures of
compound 23 and a close analogue (PDB: 5LWN) while it is not present in any other JAK3
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
2
7
structure deposited in the PDB (Figure 2b). To get a more detailed insight into the SARs
regarding this “arginine pocket” and to evaluate its impact on potency and selectivity, we
iteratively modified the electrophilic warhead of compound 23 by removal of the nitrile or the
amide group from the αꢀcyanoacrylamide motif or by saturation of the α,βꢀdouble bond. These
structural modifications were applied to different linkers (paraꢀ/metaꢀphenyl and 2,5ꢀfuryl) to
cover all relevant angles between the hingeꢀbinding motif and the sidechain directed towards
3
0
R911. The compounds were initially tested in our activityꢀbased inꢀhouse ELISA to evaluate
JAK3 inhibition on the isolated enzyme (Table 1). In this assay format both reference
compounds 1 and 2 demonstrated excellent JAK3 potency in the lower nanomolar range with
IC values of 3.5 nM for 1 and 11 nM for 2, respectively. The first investigated compound series
5
0
were acrylamides and their corresponding saturated analogues. For both, the metaꢀ and paraꢀ
phenyl linkers, the corresponding acrylamide derivatives 3, 4, 6 and 7 exhibited only moderate to
weak inhibitory activity, ranging from micromolar to tripleꢀdigit nanomolar IC50 values.
However, it is worth mentioning that primary amide 3 precipitated under the assay conditions
due to an insufficient aqueous solubility. Saturation of the double bond and therefore inactivation
of the Michael system (compounds 12, 13, 15 and 16) gave IC values in similar range with a
5
0
maximally 3ꢀfold decrease in activity, which was not substantial enough to assume a covalent
targeting of C909 by this inhibitor class. These subtle changes are more probably caused by
ACS Paragon Plus Environment
7

Journal of Medicinal Chemistry
Page 8 of 64
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
entropic effects due to the increased flexibility of the saturated alkyl chain. The replacement of
the phenyl linker by a 2,5ꢀdisubstituted furan moiety generally resulted in better inhibitory
activities. While a similar 3ꢀfold decrease in potency was observed for the saturated
dimethylamide analogue 17 (IC = 181 nM) compared to the parent dimethylacrylamide 8 (IC
50
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
5
0
=
51 nM), the primary amides 5 and 14 surprisingly showed the reversed trend with IC values
50
of 129 nM for acrylamide 5 and 59 nM for the saturated derivate 14. This somewhat better
potency of 14 may be explained by the formation of an intramolecular hydrogen bond between
the primary amide and the furan oxygen atom. This would result in a reduced flexibility and
hydration and therefore having a positive entropic effect and less desolvation penalty.
In a second series the amide function was replaced by a cyano substituent leading to
acrylonitriles 9-11 and the corresponding saturated analogues 18-20. The paraꢀsubstituted phenyl
linker again provided only moderate inhibition levels with IC values around 400 nM for the
5
0
unsaturated and saturated derivatives 9 and 18. In contrast a 4ꢀfold difference in JAK3 inhibition
was determined for the corresponding metaꢀphenyl linked acrylonitrile 10 (IC = 135 nM) and
5
0
its saturated analog 19 (IC50 = 505 nM). Moreover, 10 exhibited a significantly lower IC50
compared to the corresponding amide derivatives 4 and 7 (IC values of 470 nM and 548 nM,
50
respectively). Again, the furyl linker yielded the most potent JAK3 inhibition within this series
with IC values of 27 nM for acrylonitrile 11 and 129 nM for alkylnitrile 20. To elucidate the
5
0
role of the nitrile substituent, Xꢀray structures were determined for compounds 10, 11 and 20 in
complex with JAK3 (Figure 3). All compounds showed a binding mode similar to 23 (Figure 1a)
with two hydrogen bonds formed with the backbone of the hinge residues E903 and L905. The
cyclohexyl moiety occupies the region of the ATP binding pocket. The linker is orientated
towards the solvent exposed front part of the catalytic cleft, positioning the nitrile substituents in
ACS Paragon Plus Environment
8

Page 9 of 64
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
close proximity to C909, R911 and R953. The alkylnitrile sidechain of 20 showed no directed
interactions to these residues (Figure 3c) and the conformation of R911 and R953 was
comparable to that seen in the Xꢀray structure of 1 (Figure 2b). In sharp contrast, both
acrylonitriles 10 and 11 induce the formation of the arginine pocket and interact with R911
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 3a,b) in the same way as observed in the crystal structure of 23 (Figure 2a). These
structural models together with the assay data demonstrated that both, the furan linker as well as
the acrylonitrile substituent and the resulting formation of the arginine pocket, are two distinct
factors positively influencing JAK3 inhibitory potency.
Table 1. ELISAꢀbased JAK3 inhibition data
JAK3 IC50
JAK3 IC50
Compd.
R =
n.a.
Compd.
R =
n.a.
[
nM]
[nM]
Tofacitinib
a
NIBR3049
b
3.5 ± 0.6
11 ± 1
(1)
(2)
Series 1: amideꢀsubstituted inhibitors
b,c
b
3
> 1000
12
13
14
499 ± 48
b
b
4
5
470 ± 36
490 ± 9
b
b
129 ± 8
59 ± 2
ACS Paragon Plus Environment
9

Journal of Medicinal Chemistry
Page 10 of 64
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
b
b
b
b
6
7
8
226 ± 31
548 ± 39
15
16
17
642 ± 63
678 ± 25
181 ± 14
b
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
b
51 ± 6
Series 2: nitrileꢀsubstituted inhibitors
b
b
b
b
9
390 ± 156
18
19
20
436 ± 15
505 ± 41
129 ± 13
b
1
1
0
1
135 ± 3
b
27 ± 1
Series 3: covalentꢀreversible inhibitors
b
b,d
b,d
2
1
571 ± 24
25
26
27
28
30 ± 6
11 ± 1
b
22
205 ± 8
b
b,d
2
2
3
4
9 ± 1
4 ± 1
b
b
23 ± 3
14 ± 2
3
0 a
b
IC values are calculated from the results of an ELISA. average ± SD (n = 18) average ±
SEM (n = 3) precipitation from aqueous assay buffer arylidene dinitriles show unspecific
reactivity with assay buffer components.
50
c
d
ACS Paragon Plus Environment
10

Page 11 of 64
Journal of Medicinal Chemistry
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 3. (a) Compound 10 bound to JAK3 (PDB: JAK3ꢀ10) (b) Compound 11 bound to JAK3
(PDB: JAK3ꢀ11) (c) Compound 20 bound to JAK3 (PDB: JAK3ꢀ20). Acrylonitriles 10 and 11
induce the arginine pocket as observed in the JAK3 complex with compound 23. Saturated
alkylnitrile 20 was not able to induce this cavity and R911 and R953 shared similar conformation
as observed in the crystal structure of 1.
In a third series we combined the nitrile and amide substituents resulting in cyanoacrylamides,
3
1
a compound class, which was previously described as covalentꢀreversible Michael acceptors.
Comparison of inhibitory activities of compounds 21, 22 and 23 with the corresponding
acrylonitriles 9, 10 and 11 reveals that only the furan linked compound 23 benefits from the
additional electronꢀwithdrawing effect of the amide group. While phenyl linked compounds 21
and 22 only show low to moderate JAK3 inhibition (IC50 values of 571 nM and 205 nM,
respectively), compound 23 exhibits an IC value of 9 nM, which is close to the lower resolution
50
limit of the applied ELISA. Further modifications of the dimethyl amide of 23, like truncation of
a methyl group (24) or replacement by an Nꢀmethylpiperazine (28), were also well tolerated and
resulted in low nanomolar IC50 values (23 nM and 14 nM for 24 and 28, respectively). The
observed significant difference in the inhibitory potency of 23 compared to 22 suggested that
covalent bond formation with C909 was limited to the furanꢀlinked 23 and not possible with the
ACS Paragon Plus Environment
11

Journal of Medicinal Chemistry
Page 12 of 64
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
phenyl bearing 22. Although there was no covalent bond observed in the Xꢀray structure of 23
(Figure 2a), covalent bond formation was observed for a close analogue of 23 by mass
2
7
spectroscopic and crystallographic data as reported earlier. Moreover, comparison of the Xꢀray
structures of 10, 11 (Figure 3a,b) and 23 (Figure 2a) revealed significant differences in the
dihedral angles along the α,βꢀdouble bond of the Michael system. While the usual trans angle 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
1
80° (measured as 179° and 178°) was retained in the structures of 10 and 11, the double bond of
2
3 was fairly distorted with a dihedral angle of 149°. This outꢀofꢀplane conformation of 23 may
be due to interaction with the electrophilic βꢀcarbon atom with the thiol group of C909 leading to
bond formation.
We further enhanced the electrophilicity of the Michael acceptor in the compounds 25-27 by
replacing the amide function by a second nitrile group. With IC values in the lower nanomolar
5
0
range, these benzylidene malononitriles were remarkably potent JAK3 inhibitors and contrary to
the corresponding cyanoacrylamides 21-23, the linker geometry had a significantly lower
influence on potency. It should be mentioned that the structural element of arylidene dinitriles is
relatively promiscuous and shows a high reactivity towards any nucleophile. Indeed, subsequent
chemical incubation experiments revealed an unspecific reactivity for the highly electrophilic
inhibitors 25-27 towards thiol nucleophiles. Dithiothreitol (DTT) is a common reductant and an
important ingredient of standard kinase assay buffer systems. As an aliphatic thiol, it possesses a
nucleophilicity comparable to cysteine and is therefore capable to react in a similar way via
Michael addition with electrophilic inhibitors. While no decomposition was observed for
inhibitors 25-27 in methanolic solution, the compounds degraded over time in presence of DTTꢀ
containing kinase assay buffer. In contrast, cyanoacrylamides showed reversible addition of the
thiol without any signs of further decomposition (data not shown). Because of the high and
ACS Paragon Plus Environment
12

Page 13 of 64
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
unspecific reactivity of the arylidene dinitrile moiety the inhibitory activity of 25-27 may not
only be attributed exclusively to the interaction with C909 and data obtained with this type of
compounds should be interpreted carefully. Nevertheless, compounds from this class showed
remarkable JAK3 potency and selectivity even in cells or in the presence of thiolꢀcontaining
buffer (vide supra).
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
For covalent inhibitors, the binding of the ligand can be divided into two steps: first a
reversible binding event and second the formation of the covalent bond between electrophile and
32
the cysteine sidechain. Since the initial reversible inhibitor binding is required before a
covalent bond to C909 can be formed, it is important that the inhibitor shows not only a high
reactivity towards sulfur nucleophiles, but also successfully competes with the high intracellular
ATP concentrations. In order to evaluate the contributions of reversible versus covalent binding
to the overall inhibitory potency of compound 23, we modified our lead compound with the goal
of modulating the reversible binding affinity while retaining the electrophilic cyanoacrylamide
warhead (Figure 4). The truncation of the cyclohexyl sidechain (compound 29, Figure 4), which
fills the ATP binding pocket towards the glycine rich loop, results in an IC value of 368 nM,
50
corresponding to a more than 40ꢀfold loss in potency compared to 23. As expected, an even more
drastic drop in potency resulted from the methylation of the pyrrole nitrogen atom (compound
3
0, Figure 4). This modification prevents the formation of key hydrogen bonds to the hinge
region and caused a complete loss of inhibitory activity towards JAK3 even at concentrations up
to the micromolar range. These data demonstrated that the high potency of 23 was not only
driven by the covalent interaction with C909, but also originates from strong reversible
interactions.
ACS Paragon Plus Environment
13

Journal of Medicinal Chemistry
Page 14 of 64
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 4. Structures and inhibition data of compounds 29 and 30. IC values are calculated from
5
0
3
0
the results of an ELISA as average ± SEM (n = 3)
Kinase Selectivity Profiles. As a next step in the development process, we determined JAK
family selectivity profiles of selected compounds of the furyl series in a radiometric assay format
carried out at Reaction Biology Corp. (Table 2). In this more sensitive assay, reference
compound 1 showed a subnanomolar JAK3 IC value of 0.29 nM and, as expected, poor JAK
5
0
1
0
isoform selectivity as reported earlier. While 1 inhibited JAK1, JAK2 and JAK3 with a similar
potency, TYK2 was inhibited to a lower extent. Maleimide 2 showed a better JAK3 selectivity,
1
1, 16
comparable with the previously reported data.
Meanwhile acrylamide 8 and its saturated
analogue 17 showed significantly higher IC values compared to reference compounds 1 and 2
5
0
and did not exhibit a noteworthy JAK3 selectivity as well. In contrast, a huge difference in
selectivity and potency was observed between acrylonitrile 11 and its saturated counterpart 20.
While 20 featured a similar selectivity profile and moderate potency as the amide derivatives 8
ACS Paragon Plus Environment
14

Page 15 of 64
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
and 17, compound 11 showed potent JAK3 inhibitory activity equal to 1 (IC = 0.21 nM) with a
50
5
7ꢀ, 250ꢀ and 210ꢀfold selectivity against JAK1, JAK2 and TYK2, respectively. This remarkable
leap in selectivity together with the crystallographic structural information disfavoring covalent
binding of 11 and 20 (Figure 3b,c) provides strong evidence that the induction of the arginine
pocket may be an additional key driver for JAK3 selectivity. Nevertheless, the excellent
selectivity profile was even further improved when a covalentꢀreversible Michael acceptor was
introduced. Compound 23 provided an extraordinarily potent JAK3 inhibition (IC = 0.13 nM)
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
5
0
combined with good isoform selectivity (400ꢀ, 2700ꢀ and 3600ꢀfold vs. JAK1, JAK2 and TYK2,
respectively). As mentioned before we hypothesize that this improved isoform selectivity of 23
compared to 11 can be attributed to a synergism between the formation of the arginine pocket
and a covalentꢀreversible interaction to C909 enabled by the additional electron withdrawing
effect of the amide group of 23.
It is known, that JAK3 has a higher affinity to ATP compared to the other JAK isoforms,
which causes a shift in selectivity when the assays are performed at constant and higher ATP
11, 20
concentrations compared to the typically determined data at K ATP.
To investigate the
m
ATP dependency of the isoform selectivity of 23, IC50 values were determined at an ATP
concentration of 200 µM and revealed a retention of the favorable selectivity profile at this
physiologically more relevant ATP concentration. This 20ꢀfold higher ATP concentration better
reflects the typical intracellular conditions with even higher ATP levels in the low millimolar
range. While under these conditions the JAK3 inhibitory potency of 23 remained single digit
nanomolar (IC = 2 nM), an at least 330ꢀfold selectivity window over JAK1 was observed (IC
50
5
0
=
668 nM) and the IC values for JAK2 and TYK2 even increased up to the micromolar range
50
2
7
(Table 2). As reported earlier, this compound was also screened against the activity based
ACS Paragon Plus Environment
15

Journal of Medicinal Chemistry
Page 16 of 64
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
ProQinase panel of 410 kinases and proved to be highly JAK3 selective showing no offꢀtargets at
a concentration of 100 nM and only eleven hits with more than 50% inhibition at a higher 500
nM cutꢀoff (Table S2). Moreover, there was no significant inhibition of the other ten human
19
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
kinases which are carrying a cysteine residue at a position equivalent to JAK3. Derivatization
of the amide residue of 23 maintains a good JAK3 selectivity, as shown by compound 28
carrying an Nꢀmethylpiperazinyl amide. While 28 shows a slightly weaker inhibition of JAK3
(
IC50 = 0.74 nM) compared to 23, it still exhibits a favorable selectivity against the other
isoforms with a 160ꢀfold window over JAK1 and an at least 1000ꢀfold selectivity towards JAK2
and TYK2. Despite showing a certain instability in thiolꢀcontaining buffer systems as described
above, the arylidene dinitrile 27 also exhibits an evident JAK3 selectivity with a subnanomolar
IC of 0.39 nM and a 10ꢀ to 470ꢀfold selectivity over the three other isoforms. The moderate
5
0
isoform selectivity of 27 may be explained by a fast inactivation of JAK3 via covalent
interaction before the highly reactive compound is decomposed in the DTTꢀcontaining kinase
buffer. Based on these results the compounds with the most favorable isoform selectivities
(compounds 11, 23, 27 and 28) were selected for further evaluation in cellular assays and
metabolic stability experiments.
Table 2. JAK isoform selectivity for chosen compounds.
a
IC [nM]
5
0
Compd.
JAK1
0.50
344
19
JAK2
2
JAK3
0.29
1
TYK2
1
2
9
b
578
113
53
2110
8
22
78
44
11
12
0.21
48
1
2
7
0
45
183
48
139
25
16
10
ACS Paragon Plus Environment
16

Page 17 of 64
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
2
3
52
346
0.13
459
11800
14
c
b,c
c
b,c
6
68
4
3460
188
2
d
27
0.39
0.74
2
8
117
765
791
a
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
IC values were calculated from the results of a radiometric assay. Data were obtained as 5ꢀ
50
dose singlicate IC50 with 10ꢀfold serial dilution starting at 1 µM. [ATP] = 10 µM starting
concentration 50 µM [ATP] = 200 µM arylidene dinitriles show unspecific reactivity with
assay buffer components.
b
c
d
Cellular Activity and Target Engagement. We used different assay types to assess JAK3
inhibitory potency and selectivity of the selected compounds in a cellular setting. First we used a
NanoBRET assay system to demonstrate target engagement and to confirm isoform selectivity
3
3
on full length Janus kinases in cells. This assay system utilizes ectopically expressed nanoꢀ
luciferaseꢀtagged fullꢀlength kinases, which interact with a fluorescentꢀlabelled tracerꢀmolecule
binding to the ATP binding site. While the tracer is bound to the kinase, the spatial proximity of
the luciferase and the fluorescence dye results in a measurable optical signal via bioluminescence
resonance energy transfer (BRET) upon substrate addition. The luminescence signal is
diminished when the tracer is replaced from the ATP binding site by increasing concentrations of
the tested inhibitor, hence allowing the determination of EC values on fullꢀlength kinases in
5
0
cellular environments. We perfomed these assays for JAK1, JAK2 and JAK3 to confirm isoform
selectivity and additionally for Bruton’s tyrosine kinase (BTK), which is a likely off target of
covalent JAK3 inhibitors since it possesses a cysteine at an equivalent position to C909 in
18ꢀ19
JAK3.
For compounds 23 and 28, which showed the best selectivity profiles within the JAK
family (Table 2), isoform selectivity was confirmed in the cellular environment. While JAK3
EC values of 237 nM and 257 nM, respectively, were observed, the activities of the other two
50
JAK isoforms, JAK1 and JAK2, were not influenced up to inhibitor concentrations of 10 µM
ACS Paragon Plus Environment
17

Journal of Medicinal Chemistry
Page 18 of 64
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
(
Figure 5 and Table 3). Inhibition of BTK was detected only at very high concentrations of 28
above 50 µM) but not for compound 23, which is in accordance with the kinome selectivity data
(
reported previously. For compounds 11 and 27 the JAK3 EC50 values shifted to the higher
nanomolar range in this cellular assay (745 nM and 670 nM, respectively, Figure 5 and Table 3),
suggesting a lower permeability for 11 and 27 when comparing these data to the IC values on
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
5
0
the isolated enzyme (Table 1). While 11 still shows no significant binding to JAK1, it only
exhibits a less pronounced 10ꢀfold selectivity over JAK2. On the other hand, the arylidene
dinitrile 27 demonstrated an only 10ꢀfold selectivity over JAK1, while it hits JAK2 with a similar
potency than JAK3 (EC = 0,63 µM, Table 3). Compared to the cyanoacrylamides 23 and 28,
50
these compounds show a slightly higher affinity towards BTK with EC values of 13.3 µM and
5
0
5
.8 µM for 11 and 27, respectively. The observed BTK selectivity of the tested compounds may
2
7
be achieved by the residue inducing the arginine pocket. As we reported earlier, none of the ten
other human kinases with a cysteine equivalent to C909 in JAK3 possesses an arginine at the
position of R911, which may be a mandatory requirement for the induction of this cavity. The
slightly less pronounced selectivity of acrylonitrile 11 compared to cyanoacrylamides 23 and 28
may be explained by the missing covalent interaction to C909 as shown by the Xꢀray
experiments described above. In summary, these data clearly demonstrated cellular target
engagement along with a high selectivity for JAK3 for the cyanoacrylamide based inhibitors.
ACS Paragon Plus Environment
18

Page 19 of 64
Journal of Medicinal Chemistry
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 5. Representative BRET data measured on JAK3 kinase.
Table 3: Cellular EC values determined by nanoBRET.
50
a
EC [nM]
5
0
Compd.
JAK1
> 10 000
> 10 000
7660
JAK2
7710
JAK3
740
BTK
13 300
> 10 000
5790
1
2
1
3
> 10 000
630
237
27
678
2
8
> 10 000
> 10 000
257
> 10 000
a
Values were calculated form a nonꢀlinear least square fit using BRET ratios measured at 11
different inhibitor concentrations.
To study the effect of our compounds on the JAK/STAT pathway in a physiologically more
relevant model, we measured the dose dependent inhibition of STAT phosphorylation in primary
+
human Tꢀcells. Therefore, CD4 Tꢀcells were incubated with increasing compound
ACS Paragon Plus Environment
19

Journal of Medicinal Chemistry
Page 20 of 64
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
concentrations and stimulated with different cytokines that associate with certain JAKs. The
STAT phosphorylation levels were determined by immunoblotting after cell lysis (Figure 6).
2
7
We previously reported the results of 1 in this assay and confirmed the panJAK inhibitory
3
4
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
activity of this reference compound. While 1 diminished the JAK1 and JAK3 dependent STAT
phosphorylations after ILꢀ2 or ILꢀ4 stimulation in concentrations between 50 and 300 nM, it also
decreased JAK1/JAK2/TYK2 dependent signaling at concentrations above 300 nM. Compound
2
only showed an effect on JAK mediated signaling after stimulation with ILꢀ2 indicated by a
weak inhibition of STAT5 phosphorylation at a concentration of 300 nM, while no effect was
seen at concentrations up to 1 µM after stimulation with the three other cytokines. This rather
low cellular activities are consistent with the previously described data for this reference
compound, which initially seeded the controversial conclusion that JAK1 has a dominant role
9
, 16
over JAK3.
The putative reversible acrylonitrileꢀderived inhibitor 11 showed a moderate JAK3 selectivity
significantly inhibiting the JAK3ꢀdependent phosphorylations of STAT5 and STAT6 with
concentrations from 100 to 300 nM (Figure 6 A+B) while IFNꢀα induced STAT1
phosphorylation was only weakly inhibited at 1000 nM and ILꢀ6 dependent pSTAT3 levels were
not decreased at the same concentration (Figure 6 C+D).
The arylidene dinitrile 27 exhibited a somewhat weaker inhibition of the JAK1/3 dependent
STAT phosphorylation with a significant inhibition of ILꢀ2 signaling but no effect on the ILꢀ4
triggered pathway at the highest tested concentration of 300 nM (Figure 6 A+B). Since 27
showed significant binding to JAK2 in the BRET assays (Table 3) it is not surprising that it also
weakly inhibited JAK1/JAK2/TYK2ꢀdependent STAT3 phosphorylation at the highest
concentration of 1000 nM, while sparing JAK1/TYK2ꢀdependent STAT1 phosphorylation
ACS Paragon Plus Environment
20

Page 21 of 64
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
(
Figure 6 C+D). It should be noted, that despite the aforementioned unselective thiol reactivity,
compound 27 is still able to achieve a weak but evident selectivity for the inhibition of JAK3ꢀ
dependent signaling pathways in this functional cell assay.
2
7
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
A more pronounced JAK3 selectivity was demonstrated for cyanoacrylamides 23 and 28
showing significant inhibition of ILꢀ2 induced STAT5 phosphorylation at concentrations around
5
0 to 100 nM (Figure 6 A) and blockade or reduction of ILꢀ4 induced STAT6 phosphorylation at
concentrations around 300 nM (Figure 6 B). At the same time, the JAK3ꢀindependent ILꢀ6 and
IFNꢀα pathways are not influenced by 23 and 28 even at concentrations up to 1000 nM (Figure 6
C+D). These results are in good agreement to the initially shown cellular selectivity from the
BRET assays (Table 3) and the outcome from these stimulation experiments demonstrates in a
semiꢀquantitative manner that highly selective JAK3 inhibition is sufficient to effectively
abrogate the downstream signaling of γ cytokines.
c
ACS Paragon Plus Environment
21

Journal of Medicinal Chemistry
Page 22 of 64
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. Inhibition of cytokineꢀinduced STATꢀphosphorylation of selected compounds. Human
+
CD4 Tꢀcells were incubated with the indicated concentrations of 1, 2, 11, 23, 27 or 28 for one
ACS Paragon Plus Environment
22

Page 23 of 64
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
hour followed by 30 min stimulation with single cytokines. The cells were lysed and the pSTAT
levels were determined via immunoblotting. Actin was coꢀblotted to ensure equal loadings ILꢀ2
activates JAK1/3 dependent STAT5 phosphorylation (A), ILꢀ4 activates JAK1/3 dependent
STAT6 phosphorylation (B), ILꢀ6 activates JAK1/JAK2/TYK2 dependent STAT3
phosphorylation (C) and IFNꢀα activates JAK1/TYK2 dependent STAT1 phosphorylation (D).
The cellular assays/immunoblots for 1 and 23 were performed in a separate experiment as
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
2
6
published previously than the tests/blots for 2, 11, 27 and 28.
Table 4. in vitro stability and solubility of selected compounds.
a
b
c
MLM stability
%] ± SD
HWB stability
aqueous solubility
Compound
[
[%] ± SD
[mg/l] / [µM]
d
1
2
2
2
1
3
7
8
53.9 ± 2.1
58.0 ± 0.9
11.0 ± 0.7
6.9 ± 3.9
98.8 ± 3.6
49.4 ± 1.7
1.5 ± 0.1
n.d.
24 / 56
d
n.d.
47.0 ± 1.2
120 / 248
a
percentage of residual compound after 180 min incubation with mouse liver microsomes.
b
c
percentage of residual compound after 180 min incubation with human whole blood. aqueous
solubility in PBS puffer pH 7.4 n.d. = not determined.
d
Solubility and Metabolic Stability. Aiming for an in vivo use of our new JAK3 probes, we
next focused on physicochemical and metabolic properties of selected compounds. During the
work on this structure class, we recognized that the developed inhibitors showed limited
solubility. Indeed, the aqueous solubility of key compound 23 determined in phosphate buffered
saline (PBS) at a physiological pH of 7.4 proved to be fairly low (24 mg/l or 56 µM,
respectively). An improved solubility was achieved by replacing the dimethyl amide residue by a
more polar Nꢀmethylpiperazine amide (compound 28), which would be protonated under
physiological conditions. Since this residue is not responsible for critical interactions with the
ACS Paragon Plus Environment
23

Journal of Medicinal Chemistry
Page 24 of 64
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
kinase, this transformation maintained the good potency and selectivity as described before
(Table 1 and 2) while resulting in a 5ꢀfold increase in aqueous solubility to 120 mg/l or 248 µM,
respectively (Table 4).
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 in vitro stability of the compounds 11, 23, 27 and 28 was tested in mouse liver
microsomes (MLM) as well as in human whole blood (HWB). While compounds 27 and 28
undergo fast degradation in the microsomal stability assay with only a small amount (ca. 10 %)
remaining intact at the threeꢀhour end point, compounds 11 and 23 showed a favorable
microsomal stability with more than 50 % of the parent compound remaining after three hours of
incubation (Table 3). Mass spectrometric analysis of the metabolites suggested the main
metabolic transformations to be hydroxylation, hydrolysis of the nitrile to an amide and
dealkylation of the cyclohexyl sidechain or the amide residues of 23 and 28. The fast
decomposition of 28 is probably due to the cleavage of the metabolically vulnerable Nꢀ
methylpiperazine moiety, which most probably undergoes demethylation as well as ring opening.
Since this moiety is not required for critical interactions to the kinase, the metabolites originating
from transformation of the piperazine ring are likely to retain JAK3 inhibitory activity. However,
this remains speculative since the identification of possible metabolites of 28 and the
determination of their biological activity were not within the scope of this study.
To assess the general stability of the electrophilic inhibitors in a medium with a high content of
endogenous nucleophiles, we incubated the compounds with human whole blood (HWB) and
determined the residual amount of parent compound after three hours via mass spectrometry. As
in the microsomal stability assay, the arylidendinitrile 27 demonstrated a very low stability under
these conditions with only 1.5 % of the parent substance left after 3 h of incubation. This poor
overall stability of 27 can be most probably attributed to the very high electrophilicity of the
ACS Paragon Plus Environment
24

Page 25 of 64
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
dinitrile Michael system and the high intrinsic reactivity of this functional group. In contrast, the
simple acrylonitrile 11 showed a good stability in the HWB assay and was only marginally
degraded at the threeꢀhour endꢀpoint of the stability assay (Table 4). Cyanoacrylamides 23 and
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
2
8 both demonstrated a lower but still favorable whole blood stability with a halfꢀlife of around
three hours (Table 4).
Since key compound 23 exhibited favorable stability in both assays combined with an
excellent JAK3 potency and selectivity, we consider it the most promising candidate for in vivo
studies within this inhibitor class.
a
Scheme 1. Synthesis of JAK inhibitors 3-28
ACS Paragon Plus Environment
25

Journal of Medicinal Chemistry
Page 26 of 64
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
(
i) NaH, TsCl, THF, 0 °C to rt, 93%; (ii) Me N·NO , TFAA, DCM, 0ꢀ5 °C, 67%; (iii)
4
3
cyclohexylamine, Et N, iPrOH, reflux, 94%; (iv) Pd/C, H , MeOH / EtOAc, 50 °C, quant; (v)
corresponding aldehyde, KHSO , DMF / H O, rt, 68ꢀ72%; (vi) DMP, DCM, rt, 80ꢀ92%; (vii)
Cu(MeCN) ·OTf, TEMPO, bpy, NMI, O , MeCN, 50 °C, 87%; (viii) KOH, MeOH, rt, 54ꢀ75%
(ix) corresponding HWEꢀreagent, LiCl, DBU, MeCN / CHCl , 54ꢀ88%; (x) Pd/C, H , MeOH /
EtOAc, rt, 60ꢀ95%; (xi) corresponding cyanoacetamide or malononitrile, piperidine, MeOH,
0 °C, 20ꢀ93% (xii) cyanuric chloride, DMF, rt, 50ꢀ81%
3
2
5
2
4
2
3
2
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
6
27
CHEMISTRY. We previously reported a synthesis of 23 which was slightly modified to
allow a more convenient access to gram quantities of key intermediates and final compounds.
The synthetic route to compounds 3-28 is shown in Scheme 1 and starts from commercially
available 4ꢀchloroꢀ1Hꢀpyrrolo[2,3ꢀb]pyridine. After tosyl protection and nitration in position 5
under mild anhydrous conditions, the cyclohexyl sidechain was introduced via a nucleophilic
aromatic substitution. Catalytic hydrogenation afforded key intermediate 34, which is unsuitable
for long term storage under ambient conditions and should be used in the next step immediately.
Closure of the imidazole ring was achieved by condensation with the corresponding aldehydes
3
5
under oxidative conditions in wet DMF. Benzylic alcohols 35 and 36 were initially transformed
36
to the corresponding aryl carbaldehydes 38 and 39 via DessꢀMartinꢀOxidation. However, since
this reaction can be problematic on a multigram scale it was replaced by a catalytic Stahlꢀ
3
7
Oxidation for the conversion of furan derivative 37 to aldehyde 40. Intermediates 41-43 were
obtained by tosyl cleavage with methanolic KOH at ambient temperature. Acrylamides 3-8 were
accessible via a LiClꢀpromoted HornerꢀWadsworthꢀEmmons (HWE) reaction according to a
3
8
protocol modified from Blanchette et al. The saturated analogues were prepared from the
corresponding acrylamides by catalytic hydrogenation on Pd/C. Nitriles 9-11 and 18-20 were
synthesized from the corresponding primary amides by dehydration with cyanuric chloride in
3
9
DMF. Cyanoacrylamideꢀ and malononitrileꢀderived inhibitors 21-28 were obtained from
aldehydes 41-43 under Knoevenagel conditions with the corresponding cyanoacetamides or
ACS Paragon Plus Environment
26

Page 27 of 64
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
malononitrile. The cyanoacetamides for the synthesis of 24 and 28 were obtained via aminolysis
of ethyl cyanoacetate (see Supporting Information). Compound 29 was synthesized following the
same procedures described in Scheme 1, but using aqueous methylamine solution instead of
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
cyclohexylamine for S Ar (see Supporting Information for detailed procedures). Methylated
N
compound 30 was prepared by reacting 23 with methyl iodide after deprotonation with NaH in
DMF (see Supporting Information).
CONCLUSION:
Herein we report our efforts to elucidate the role of a ligandꢀinduced binding pocket in JAK3
formed by interactions of arginine residues with a nitrile substituent, which was initially
observed for cyanoacrylamideꢀbased inhibitors. Our recently described covalentꢀreversible
27
inhibitor 23 was iteratively modified to establish detailed SARs regarding this previously
unknown “arginine pocket” and we were able to show a correlation between inhibitor selectivity
and the formation of this cavity. With compound 11 we demonstrated via Xꢀray crystallography
and in vitro profiling that a significant JAK3 selectivity can be achieved by induction of the
arginine pocket even without covalent bond formation to the JAK3ꢀspecific C909. The
application of nitrileꢀbearing covalentꢀreversible warheads resulted in a positive synergism of
both selectivity vectors and yielded a set of JAK3 inhibitors with an extraordinarily high potency
and isoform selectivity. The favorable selectivity profile of the compounds 11, 23, 27 and 28 was
further validated in several cellular assays demonstrating target engagement in engineered cells
as well as selective inhibition of JAK3 dependent downstream signaling in functional human Tꢀ
cells. Our data also indicate that a highly selective inhibition of JAK3 is sufficient to abrogate
JAK1/3 dependent STATꢀphosphorylation without the need of a simultaneous JAK1 inhibition.
This is in good agreement with recent data from other groups, who investigated this issue and
ACS Paragon Plus Environment
27

Journal of Medicinal Chemistry
Page 28 of 64
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
20, 26, 40
even provided some in vivo data.
Compound 23 exhibited an acceptable stability profile in
two in vitro metabolism assays, while bioavailability remains to be shown. These studies are
ongoing and will be published in due course. In summary, our SAR study revealed that the well
characterized cyanoacrylamide 23 is currently the most optimized candidate from this series.
This compound is a promising tool, complementing the existing irreversible acrylamideꢀbased
inhibitors by a covalentꢀreversible mode of action and may be suitable to investigate the effects
of highly selective JAK3 inhibition in in vivo.
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
EXPERIMENTAL SECTION:
Protein expression and structure determination: Recombinant JAK3 was purified as
2
7
previously described. The protein was mixed with 1 mM inhibitors and 0.26% Nꢀphenylurea.
Crystallization was performed using sittingꢀdrop vapor diffusion method at 4 °C and the
reservoir solution containing 24ꢀ30% PEG 3350, 0.1 M MES, pH 5.5ꢀ6.0 and 0.1ꢀ0.2 M MgCl2.
Diffraction data were collected at Diamond Light Source, beamline i03 using Xꢀray at
4
1ꢀ42
wavelength of 0.97623 Å, and were processed and scaled with XDS and Scala, respectively.
43
Structure solutions were obtained by molecular replacement method using Phaser and the
2
7
published JAK3 coordinates. Manual Model rebuilding alternated with refinement were
4
4ꢀ45
performed in COOT and REFMAC.
Geometric correctness was verified using
46
MOLPROBITY. The data collection and refinement statistics are summarized in Supplemental
Table S1. Visualization of Xꢀray crystallographic data and preparation of images was performed
4
7
with PyMol.
JAK1/2/3 and BTK NanoBRET Assays: Assays were performed essentially as described
2
7
previously. In brief: Fullꢀlength JAK1, JAK2, JAK3 and BTK ORF (Promega) cloned in frame
with a Cꢀterminal NanoLucꢀfusion were transfected into HEK cells and proteins were allowed to
ACS Paragon Plus Environment
28

Page 29 of 64
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
express for 20h. Serially diluted inhibitor and NanoBRET Kinase Tracerꢀ04 (Promega) at 1 µM
for JAK1, Tracerꢀ05 (Promega) at 2 µM for JAK2/JAK3 and at 1 µM for BTK, respectively,
were pipetted into white 384ꢀwell plates (Greiner 781 207). The corresponding
JAK1/JAK2/JAK3 or BTKꢀtransfected cells were added and reseeded at a density of 2 x 105
cells per well after trypsinization and resuspending in OptiꢀMEM without phenol red (Life
Technologies). The system was allowed to equilibrate for 2 hours at 37°C/5% CO2 prior to
BRET measurements. To measure BRET, NanoBRET NanoGlo Substrate + Extracellular
NanoLuc Inhibitor (Promega) was added as per the manufacturer’s protocol, and filtered
luminescence was measured on a CLARIOstar plate reader (BMG Labtech) equipped with 450
nm BP filter (donor) and 610 nm LP filter (acceptor). Competitive displacement data were then
graphed using GraphPad Prism software using a 4ꢀparameter curve fit with the following
equation:
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
Y=Bottom + (TopꢀBottom) / (1+10^((LogIC50ꢀX)*HillSlope))
+
CD4 T-cell stimulations: The stimulation experiments were carried out following the same
27
+
procedure reported previously. CD4 T cells were purified from human PBMCs using magnetic
cell separation technology (Miltenyi Biotec) followed by activation with plateꢀbound antiꢀCD3
(clone UCHT1) and antiꢀCD28 (clone CD28.2) antibodies (each 5 µg/ml, Biolegend) in XVIVO
1
5 medium (Lonza) for 3 days and expanded with rhILꢀ2 for another 6 days. After expansion T
cells were washed and rested in fresh medium overnight. Equal numbers of T cells were
incubated with the indicated concentrations of compounds 1, 2, 11, 23, 27 and 28 in DMSO or
DMSO alone (control) for 1 hour and then stimulated with either rhILꢀ2 (50 ng/ml; Proleukin,
Novartis), rhILꢀ6 (50 ng/ml; Peprotech), rhILꢀ4 (50 ng/ml, Peprotech) or IFNꢀα (1000 U/ml,
Roche) for 30 min. After stimulation, cells were lysed in Triton Xꢀ100 lysis buffer containing
ACS Paragon Plus Environment
29