Discovery of VX-509 (Decernotinib): A Potent and Selective Janus Kinase 3 Inhibitor for the Treatment of Autoimmune Diseases

Article abstract of DOI:10.1021/acs.jmedchem.5b00301

While several therapeutic options exist, the need for more effective, safe, and convenient treatment for a variety of autoimmune diseases persists. Targeting the Janus tyrosine kinases (JAKs), which play essential roles in cell signaling responses and can contribute to aberrant immune function associated with disease, has emerged as a novel and attractive approach for the development of new autoimmune disease therapies. We screened our compound library against JAK3, a key signaling kinase in immune cells, and identified multiple scaffolds showing good inhibitory activity for this kinase. A particular scaffold of interest, the 1H-pyrrolo[2,3-b]pyridine series (7-azaindoles), was selected for further optimization in part on the basis of binding affinity (K_i) as well as on the basis of cellular potency. Optimization of this chemical series led to the identification of VX-509 (decernotinib), a novel, potent, and selective JAK3 inhibitor, which demonstrates good efficacy in vivo in the rat host versus graft model (HvG). On the basis of these findings, it appears that VX-509 offers potential for the treatment of a variety of autoimmune diseases.

Full text of DOI:10.1021/acs.jmedchem.5b00301

Page 1 of 101
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Wang, Jian; Vertex Pharmaceuticals, Inc.,
Wang, Tiansheng; Vertex Pharmaceuticals, Inc.,
Wannamaker, M; Vertex Pharmaceuticals, Inc.,
Winquist, Raymond; Vertex Pharmaceuticals, Inc.,
Zuccola, Harmon; Vertex Pharmaceuticals, Inc.,
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 2 of 101
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
Discovery of VX-509 (Decernotinib): A Potent and Selective Janus kinase (JAK) 3 Inhibitor for
the Treatment of Autoimmune Diseases
*
τ
*
Luc Farmer , Mark W. Ledeboer , Thomas Hoock, Michael J. Arnost, Randy S. Bethiel, Youssef L.
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
τ
υ
Bennani , James J. Black, Christopher L. Brummel , Ananthsrinivas Chakilam, Warren A. Dorsch, Bin
ε

Ψ
Fan , John E. Cochran, Summer Halas, Edmund M. Harrington , James K. Hogan, David Howe , Hui
Huang, Dylan H. Jacobs, Leena M. Laitinen, Shengkai Liao, Sudipta Mahajan, Valerie Marone, Gabriel
γ
ν
κ
MartinezꢀBotella , Pamela McCarthy, David Messersmith, Mark Namchuk , Luke Oh , Albert C.
γ
φ
Pierce, Scott A. Raybuck, Arthur Rugg, Francesco G. Salituro , Kumkum Saxena, Dean Shannon , Dina
ξ
Shlyakter, Lora Swenson, ShiꢀKai Tian , Christopher Town, Jian Wang, Tiansheng Wang, M. Woods
Wannamaker, Raymond J. Winquist, and Harmon J. Zuccola
*
To whom correspondence should be addressed:
(
LF) Telephone: (450) 680ꢀ4656. Fax: 450 978ꢀ7972. Eꢀmail: luc_farmer@vrtx.com;
(ML) Telephone: (617) 341ꢀ6309. Fax: (617) 444ꢀ7825. Eꢀmail: mark_ledeboer@vrtx.com.
Vertex Pharmaceuticals Incorporated, 50 Northern Avenue, Boston, Massachusetts 02210, United
τ
States; Vertex Pharmaceuticals (Canada) Incorporated, 275 ArmandꢀFrappier, Laval, Québec H7V
4A7, Canada
γ
Present address: Sage Therapeutics, 215 First St, Cambridge, Massachusetts 02141, United States
1
ACS Paragon Plus Environment

Page 3 of 101
Journal of Medicinal Chemistry
1
2
3
υ
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
Present address: RxDiscovery LLC, 38 Sidney Street, Marlborough, MA 01752, United States
ε
Present address: Agios Pharmaceuticals, Inc., 38 Sidney Street, Cambridge, MA 02139, United States
Present address: Novartis, 250 Massachusetts Avenue, Cambridge, MA 02139, United States
Present address: DE Synthetics, 30 Dineen Drive, Fredericton, NB, E3B 5A3, Canada

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
φ
ν
Present address: Alkermes, 852 Winter Street, Waltham, MA 02451, United States
κ
Present address: Mallinkrodt Pharmaceuticals, 6011 University blvd 260, Ellicott City, MD, USA
2
1043, United States
Ψ
Present address: LuminaCare Solutions, Inc, 100 Trade Center, Suite Gꢀ700, Woburn, MA 01801,
United States
ξ
Present address: Department of Chemistry, University of Science and Technology of China, Hefei,
Anhui 230026, China
Keywords. VXꢀ509, Janus kinase, JAK3, JAK2, autoimmune diseases, immunosuppression, rheumatoid
Arthritis, Inflammation, transplant rejection
Abstract. While several therapeutic options exist, the need for more effective, safe and convenient
treatment for a variety of autoimmune diseases persists. Targeting the Janus tyrosine kinases (JAKs),
which play essential roles in cell signaling responses and can contribute to aberrant immune function
associated with disease, has emerged as a novel and attractive approach for the development of new
autoimmune disease therapies. We screened our compound library against JAK3, a key signaling kinase
in immune cells, and identified multiple scaffolds showing good inhibitory activity for this kinase. A
2
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 4 of 101
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
particular scaffold of interest, the 1Hꢀpyrrolo[2,3ꢀb]pyridine series (7ꢀAzaindoles), was selected for
further optimization based in part on binding affinity (K ) as well as on cellular potency. Optimization of
i
this chemical series led to the identification of VX-509 (decernotinib), a novel, potent and selective
JAK3 inhibitor, which demonstrates good efficacy in vivo in the rat host versus graft model (HvG).
Based on these findings, it appears that VX-509 offers potential for the treatment of a variety of
autoimmune diseases.
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 Janus kinase family of protein kinases consists of four members: JAK1, JAK2, JAK3 and TYK2.
JAK3 expression is largely restricted to lymphocytes, which are cells central to the uncontrolled immune
1, 2, 3
response associated with many diseases, including rheumatoid arthritis (RA).
In these cells, JAK3 is
constrained physically and functionally to only the common gamma (γ) chain, which is shared by
4
, 5, 6, 7
receptors for interleukin (ILꢀ2, ILꢀ4, ILꢀ7, ILꢀ9, ILꢀ15 and ILꢀ21).
Observations from genetic
studies of mice and humans with loss of JAK3 function demonstrate that JAK3 is essential component
8
, 9, 10, 11
of cell signaling that underlies immune response/function (Figure 1).
Conversely, while JAK3
expression and function is restricted to the immune system, the other JAKs are much more broadly
expressed and utilized in a variety of cells and organs. Therefore, a potent and selective JAK3 inhibitor
1
2, 13
would be an attractive candidate molecule for the treatment of a variety of autoimmune diseases.
Inhibition of other JAK isoforms could potentially increase the chances of unwanted safety and
1
4, 15,
tolerability issues without adding any additional activity for controlling immuneꢀmediated disease.
1
6, 17, 18, 19, 20
3
ACS Paragon Plus Environment

Page 5 of 101
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 1. JAKꢀSTAT signaling
Thus, over the last decade there have been extensive efforts to identify and design novel transformative
small molecules JAK inhibitors with varied profiles of subtype selectivity in order to address unmet
medical needs such as transplant rejection, rheumatoid arthritis (RA), cancers and other autoimmune
2
1, 22
diseases (Figure 2).
Flanagan et al. have reported on tofacitinib (Xeljanz) as the first orally
2
3
available JAK1/JAK3 inhibitor to be clinically approved for rheumatoid arthritis (RA). Ruxolitinib, a
JAK1/JAK2 subꢀtype selective inhibitor, was first to be approved for the treatment of myelofibrosis
24
under the trade name Jakafi.
Baricitinib, a JAK1/2 inhibitor, and filgotinib, a JAK1 selective
2
5, 26
inhibitor, are currently in clinical trials for RA.
Figure 2. Selected Clinically Approved or Experimental JAK inhibitors
4
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 6 of 101
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
In this paper we report on the lead optimization process that led to the identification of decernotinib
2
7, 28
(VX-509), an orally available, selective inhibitor of JAK3 for the treatment of autoimmune diseases.
To date, this is the only selective JAK3 inhibitor that has been evaluated in clinical studies of RA and
data from this clinical program has demonstrated that selective inhibition of JAK3 is sufficient for
significant clinical response.
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
Results & Discussion
In our effort to discover potent and selective smallꢀmolecule JAK3 inhibitors, we conducted a high
throughput screen of our inꢀhouse compound collections for their affinities to bind JAK3. Compounds
with confirmed JAK3 inhibitory activity were then screened for selectivity against a small panel of
kinases. Our JAK kinase assays were performed using a truncated form of the protein including only the
active kinase domain. This was done primarily because the fullꢀlength protein was difficult to produce
and does not remain stable in solution making its’ utility as an enzyme screening reagent limited. It is
clear from our assays and published work that the isolated kinase domains of the various JAK isotypes
29,
does not accurately measure the relative potency and selectivity of small molecule inhibitors in cells.
3
0
The context of other JAK protein domains and the microenvironment at the subꢀplasma membrane in
the cell significantly affect these relative data. To assess JAK isoform selectivity we extended our
enzymatic data set on isolated JAK kinase domains with a number of cellꢀbased assays to provide data
3
1
points from more biologically relevant systems. We first investigated the ability of our compounds to
inhibit STATꢀ5 phosphorylation by blocking the JAK3/1 ꢀ STAT5 cascade in ILꢀ2 stimulated HTꢀ2 cells.
Likewise, the ability to block the JAK2 mediated phosphorylation of STAT5 was assessed in GMCSFꢀ
stimulated TFꢀ1 cells. This provided a measure of selectivity for inhibition of JAK3/1 vs JAK2
32
mediated signaling in cells. Further, assaying IFNꢀα mediated STAT2 translocation to the nucleus in
5
ACS Paragon Plus Environment

Page 7 of 101
Journal of Medicinal Chemistry
1
2
3
HeLa cells, which is dependent on JAK1 and TYK2, provides useful counterꢀscreening data to establish
each compounds specificity against these additional JAK isoforms.
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
One of the more promising classes of compounds resulting from our JAK3 screen was the
diaminotriazole class represented by 1. While optimizing this class of compounds, a parallel screening
effort was maintained to identify additional structurally diverse inhibitors of JAK3 from other ongoing
kinase inhibitor programs including SYK, FLT3 and AURA. As part of this latter strategy, a variety of
known kinase inhibitor structures were compared with the crystal structure of 1 (Figure 3). From this
approach a different chemotype (2) was identified based on the 1Hꢀpyrrolo[2,3ꢀb]pyridine (7ꢀazaindole).
The superposition of the crystal coꢀcomplexes of JAK2ꢀ1 and ERK2ꢀ2 showed that the exo-primary
amino group of the triazole and the NH of the 1Hꢀpyrrolo[2,3ꢀb]pyridine core are closely aligned, as are
the nitrogen hydrogen bond acceptors of the 1Hꢀpyrrolo[2,3ꢀb]pyridine and triazole. This overlay
provided a basis for replacement of the anilinoꢀaminotriazole hinge binding element of 1 with a smaller
1
Hꢀpyrrolo[2,3ꢀb]pyridine hingeꢀbinding element. Although the vectors from the triazole and 1Hꢀ
pyrrolo[2,3ꢀb]pyridine do not overlay perfectly, they are similar enough to suggest that the
anilinopyrimidine of 1 might form a similar set of interactions on the 1Hꢀpyrrolo[2,3ꢀb]pyridine core.
JAK2 was used as a surrogate for JAK3 because of difficulty in handling this protein and due to the
close sequence homology of the ATP binding sites of JAK2 and 3. As such, the JAK2 coꢀcomplex
should provide a reasonable approximation for binding of these compounds with the JAK3 isoform.
6
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 8 of 101
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. Compounds 1 and 2 overlaid in the JAK2 structure. Superposition created by overlaying the
hinge regions of the ERK2 and JAK2 crystal structures
Our overlay also suggested that a 1,3ꢀdisubstitution pattern in the central core pyrimidine ring of 1
would position the phenyl group more optimally. The resulting 1Hꢀpyrrolo[2,3ꢀb]pyridine based hybrid
compound 3 was prepared and exhibited moderate potency. Compound 3 became the starting point for
further exploration of this class of JAK3 inhibitors. With a minimal survey to explore the amineꢀ linked
side chain we identified benzylamineꢀderived 1Hꢀpyrrolo[2,3ꢀb]pyridine 4 with good affinity for JAK3
and a K of 14 nM. From the outset we desired compounds with a clean selectivity profile against the
i
kinome to avoid unwanted side effects. In particular, we needed to avoid binding to additional kinases
with potential for undesirable pharmacologic activity. Despite its potent FLT3 inhibition, compound 4
was used as the prototype for the central ring variations with the goal to further improve potency as well
as selectivity against FLT3, SYK and AURA. Changing the 4,6ꢀpyrimidinyl moiety to 2,6ꢀ and 2,4ꢀ
pyrimidinyl led to a decrease in affinity for JAK3, as exemplified by compounds 5 and 6 respectively.
However, 2,4ꢀpyrimidinyl regioisomer 6 exhibited the best selectivity against FLT3 (Table 1). The
7
ACS Paragon Plus Environment

Page 9 of 101
Journal of Medicinal Chemistry
1
2
3
addition of a 5ꢀchloro atom on the 1Hꢀpyrrolo[2,3ꢀb]pyridine ring led to an overall improvement in
potency for JAK3 as seen for 7 and 8. The replacement of pyrimidine central core for a pyridine ring in
compound 9 led to a similar activity profile as for analog 5. Replacing the central heteroaryl ring with a
phenyl ring gave a significant drop in affinity for JAK3 as exemplified with analog 10.
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
8
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 10 of 101
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
33
Table 1. Central Core Ring Variations
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
JAK3 Ki
ꢁM)
JAK2 Ki
(ꢁM)
SYK Ki
(ꢁM)
FLT3 Ki
(ꢁM)
AURA Ki
(ꢁM)
HTꢀ2ꢀIL2
IC50 (ꢁM)
Compound
X
A
Y
Z
(
4
5
6
7
8
CH
CH
CH
CCl
N
CH
N
CH
N
N
N
0.014
0.096
0.38
0.035
0.11
2.9
< 0.06
< 0.06
0.7
> 0.8
0.38
0.8
9.85
> 3.33
> 3.33
0.2
> 20
N.A.
5.4
N
CH
N
0.35
N
CH
N
0.003
0.043
0.006
0.033
< 0.06
< 0.06
0.062
0.02
CCl CH
CCl CH
N
> 3.33
> 20
9
N
CH
0.055
0.15
0.07
0.22
> 3.33
1.9
0.09
0.049
N.A.
> 20
N.A.
10
CCl CH CH CH
0.078
Most of the compounds exhibited FLT3 activity and had limited or poor cellular activity in the JAK3/1
HTꢀ2 assay. Compound 6 was the least potent inhibitor of JAK3 in this set, yet it exhibited only modest
potency for FLT3, and poor affinity for both SYK and AURA. As a result, compound 6 was the only
regioisomer considered for further exploration. To follow up on compound 6, we identified additional
potent analogs with a small set of compounds containing Nꢀlinked small branched or cyclic alkyl groups.
Several examples containing hydrogen bond donor and acceptor groups were also examined. These
results are presented in Table 2.
9
ACS Paragon Plus Environment

Page 11 of 101
Journal of Medicinal Chemistry
1
2
3
Table 2. Side Chain Variations
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
JAK3 Ki
ꢁM)
JAK2 Ki
(ꢁM)
SYK Ki
(ꢁM)
FLT3 Ki
(ꢁM)
AURA
Ki (ꢁM)
HTꢀ2ꢀIL2
IC50 (ꢁM)
Compound
R1
R2
(
11
F
F
0.005
0.011
0.005
0.012
0.39
0.43
0.33
0.05
0.030
N.A
> 20
N.A.
1
2
13
14
15
16
17
F
H
F
F
F
0.016
0.022
0.14
0.023
0.022
0.22
1.1
0.67
> 4
0.078
0.29
0.007
N.A.
N.A.
N.A.
N.A.
> 20
N.A.
0.08
N.A.
0.004
0.029
0.009
0.045
0.15
0.108
0.009
0.006
0.023
0.086
1
1
2
8
9
0
F
F
F
0.068
0.005
0.14
0.077
0.011
N.A.
3.2
0.2
3.6
0.19
0.032
0.045
0.004
0.019
0.27
N.A.
> 20
N.A
2
1
H
0.006
0.008
1.1
0.032
0.13
2.1
In general, derivatives with small, branched alkyl groups (11), and cycloalkyl groups (12) were well
tolerated as these compounds exhibited good affinity for JAK3, yet in most cases they lacked adequate
selectivity versus FLT3 (Table 2). Alkylation of the amine nitrogen on compound 13, as shown in 15
led to a 10ꢀfold decrease in JAK3 inhibition. From the Xꢀray structural data for related JAK3 inhibitors,
10
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 12 of 101
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
it appears unlikely that the aniline NH acts as a hydrogen bond donor to the protein. The introduction of
the Nꢀmethyl in combination with the adjacent F causes the alkyl substituent to twist out of plane
relative to the pyrimidine ring, leading to what we believe to be a modestly less favorable binding
conformation. Incorporation of polar atoms was well tolerated and led to improved potency in some
cases (16 and 19). However, addition of strongly basic amines as in compounds 17 and 20 did not
provide the same enhancements in affinity. Among these analogs, piperazine amide derivative 21 was
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 most potent example in our HTꢀ2 cellular assay, with an IC of 2.1 ꢀM and exhibited a modest 21ꢀ
50
fold selectivity vs AURA. Most examples showed reasonable selectivity against SYK and in the case of
2
1, AURA as well (>20 fold). Finally, in this exercise no appreciable selectivity was achieved versus
FLT3 for any of the compounds while also retaining potency in the cell.
Glu 930
Leu 932
Figure 4. Crystal structure coꢀcomplex of 22 in JAK2. Hinge hydrogen bonds to Leu 932 highlighted in
green.
NꢀAcetyl piperazine based compound 21 was the most active in our HT2 cell assay and it contained an
amide moiety in the side chain. We found that a related compound (22) from our AURA program,
11
ACS Paragon Plus Environment

Page 13 of 101
Journal of Medicinal Chemistry
1
2
3
bearing an amide containing side chain with our preferred core, also exhibited good binding affinity for
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
JAK3 (K = 2 nM). More importantly, in our IL2 driven HTꢀ2 cell assay (Table 3), compound 22
i
exhibited an IC of 0.074 ꢀM. To understand the binding of this compound, analysis of the xꢀray coꢀ
5
0
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
complex of 22 and JAK2, indicated a different orientation of the 1Hꢀpyrrolo[2,3ꢀb]pyridine at the hinge
than was anticipated in the original design concept from Figure 3 (see Figure 4 also). While both
orientations would accept a hydrogen bond from the backbone NH of Leu 932, the orientation of 22
anticipated from the ERK2 structure would have the 1Hꢀpyrrolo[2,3ꢀb]pyridine NH donating a hydrogen
bond to the backbone carbonyl of Glu 930. Instead, the 1Hꢀpyrrolo[2,3ꢀb]pyridine NH of 22 is donating
a hydrogen bond to the carbonyl oxygen of Leu 932 (Figure 4). While unexpected, this difference in
binding modes is not too surprising given the significant structural differences between compounds 2
and 22 and given that the compounds are complexed to different kinases. Interestingly, the amide of 22
does not appear to be involved in H bonding to the JAK2 protein.
Relative to earlier 1Hꢀpyrrolo[2,3ꢀb]pyridine derivatives, 22 exhibited excellent selectivity
against SYK (Ki > 4 ꢀM) and provided a good selectivity window against FLT3 as well as JAK2 as
reflected by the more relevant cellular endpoints (HT2 vs TF1). In order to minimize potential offꢀtarget
activity of our JAK3 inhibitors, further improvements in selectivity were still desired, particularly
AURA, while maintaining the promising selectivity profile exhibited by 22. Thus, a series of aminoꢀ
acid containing analogs, capped with a trifluoroethyl amide, were prepared and the results are
summarized in Table 3. Examples containing amides capped with ꢀNHMe, ꢀNEt or ꢀNMeEt were also
assessed, but these were all less selective for FLT3 and often accompanied by significant loss of cell
potency (data not shown). As such, we chose to focus on amides containing the trifluoroethyl group.
12
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 14 of 101
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
Many natural and unnatural amino acidꢀbased analogs were explored. As demonstrated by
compounds 23, 24 and 27, smaller and larger amino acids such as glycine, norvaline, as well as
methionine, were all tolerated and maintained good enzyme affinity for JAK3 with retention of
selectivity for SYK. However, SAR with respect to cellular potency for these compounds was quite
sensitive to variation of the side chain.
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
*
Table 3. 5ꢀSubstituted 1Hꢀpyrrolo[2,3ꢀb]pyridine with Aminoacid Side Chain Variations
13
ACS Paragon Plus Environment

Page 15 of 101
Journal of Medicinal Chemistry
1
2
3
TFꢀ1
JAK3
Ki
JAK2
Ki
(ꢁM)
HTꢀ2ꢀ
IL2 IC50
(ꢁM)
AURA
IC50
(ꢁM)
4
Cmpd
GMꢀCSF FLT3 Ki
RLM
%
X
Z
(aa)
5
6
7
IC50
ꢁM)
(ꢁM)
(
ꢁM)
(
2
2
2
3
CCl
CCl
CCl
CCl
CCl
CCl
CCl
CF
CH
CF
CH
CH
CH
CF
(S)ꢀAla
Gly
0.002
0.001
0.002
0.006
0.013
0.04
0.007
0.005
0.023
0.037
0.069
0.03
0.074
0.16
0.75
1.5
2.67
2.9
3.6
20
0.55
0.29
1.3
0.004
0.007
< 0.025
0.05
95
75
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
24
(S)ꢀNva
(S)ꢀSer
65
2
2
2
5
6
7
0.093
0.49
0.39
1.603
n/a
n/a
n/a
63
(S)ꢀSerOMe
(S)ꢀMet
(S)ꢀPro
S)ꢀaminobutanoic
acid
8.3
n/a
n/a
11
0.26
2.9
0.02
28
0.009
0.026
2.5
0.024
(
2
9
CCl
CF
0.045
0.12
11.4
n/a
0.31
0.15
n/a
3
3
3
0
1
2
CCl
CCl
CCl
CF
CH
CH
(R)ꢀAla
(R)ꢀAla
(R)ꢀNva
0.002
0.001
0.002
0.011
0.005
0.01
1.3
>20
14.3
2.1
>4
0.063
0.027
0.23
n/a
96
47
0.11
0.22
0.61
0.41
(R)ꢀPropargyl
glycine
3
3
CCl
CH
0.002
0.011
0.51
5.1
0.37
0.032
44
34
CCl
CCl
CCl
CCl
CF
CH
CH
CH
CF
CF
CF
CF
CF
CF
CH
CF
(R)ꢀVal
(R)ꢀLeu
(R)ꢀPhe
(R)ꢀHis
(S)ꢀAla
(S)ꢀAla
(S)ꢀAla
(R)ꢀAla
(R)ꢀAla
0.002
0.004
0.006
0.018
0.007
0.031
0.006
0.007
0.019
0.001
0.001
0.001
0.023
0.058
0.081
0.042
0.15
0.11
1.7
2.93
10.3
> 20
2.3
0.56
1.5
0.16
0.405
< 0.025
0.15
78
n/a
n/a
n/a
n/a
n/a
91
3
3
3
5
6
7
2.8
1.6
2.6
1
38
5.5
n/a
>3.7
>4.0
0.48
0.61
>4.0
0.18
0.325
0.049
0.21
3
4
4
9
0
1
CH
6.1
n/a
COMe
COMe
N
0.034
0.038
0.33
1.5
2.2
0.033
0.21
1.3
n/a
79
42
>20
0.058
0.081
n/a
>0.8
n/a
30
4
4
3
4
CCl
CCl
0.001
0.002
0.41
0.96
0.10
0.082
88
4
5
CCl
N
CH
CH
0.003
0.003
0.002
0.006
0.081
0.52
0.65
3
0.2
1.3
0.62
52
46
> 0.8
100
VX-509
CH
CCl
CH
CH
CH
CH
CH
N
0.002
0.002
0.003
0.003
0.013
0.028
0.004
0.024
0.099
0.9
2.59
7.9
1.04
0.29
0.54
1.39
> 0.8
0.070
> 0.8
0.49
70
n/a
63
97
4
7
48
0.072
0.33
1.65
3.7
4
9
*
All compound showed Ki > 4 uM for SYK
In the natural (S)ꢀconfiguration, the amino acid variations largely failed to significantly reduce AURA
affinity. Only methyl serine derived 26 showed modestly improved selectivity against AURA (20ꢀfold),
14
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 16 of 101
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
but because of its very weak potency in our HTꢀ2 cell assay, this compound did not meet the
requirements for further investigation. The rigid prolineꢀbased analog 28 exhibited favorable selectivity
against both SYK and FLT3, but no improvement in selectivity against AURA was observed. A
homologated version of 22 containing a betaꢀalanine group as in 29 exhibited poorer JAK3 affinity as
well as 150ꢀfold reduction of cellular potency. Inversion of the absolute stereochemistry of Sꢀ22,
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
(compound Rꢀ30), did not show any significant improvement in the overall profile relative to its
enantiomer while an 18ꢀfold loss of cellular potency was observed. Interestingly, the desꢀ
fluoropyrimidyl analog 31 with the same R configuration exhibited a similar selectivity profile as 30 and
improved cellular potency (HTꢀ2 IC 1.3 ꢀM vs. 0.11 ꢀM). Compounds bearing unnatural amino acids
5
0
with sp and sp2 character such as analogs 33, 36 and 37 did not show selectivity against AURA. These
compounds also exhibited weak cellular activities. Fortunately, incorporation of bigger unnatural
hydrophobic amino acid side chains, such as (R)ꢀnorvaline 32, (R)ꢀvaline 34 and (R)ꢀleucine 35, led to a
significant improvement in selectivity versus AURA. Several of these compounds, 34 and 32, also
retained good discrimination against SYK and FLT3 and displayed potent JAK3 mediated cellular
activity in our HT2 assay with IC of 0.11 and 0.22 ꢀM respectively. It is interesting to note that
50
differences in enzyme affinity do not always translate to potent cellular activity. This may be attributed
to subtle differences in the affinity of our compounds for the truncated JAK kinase enzyme domains
versus fullꢀlength enzymes present in the cells.
A brief survey of substituents at the Cꢀ5 position of the R/Sꢀalanine based 1Hꢀpyrrolo[2,3ꢀb]pyridine
indicated that smaller substituents such as H, F or OMe (compounds 38 to 41) were all tolerated for
JAK3 binding activity with good selectivity, but lacked the potency that the 5ꢀchloro analog 22
displayed in the HTꢀ2 cellular assay. Incorporation of larger Cꢀ5 substituents such as carbamates, or
15
ACS Paragon Plus Environment

Page 17 of 101
Journal of Medicinal Chemistry
1
2
3
amides provided compounds with modest enzyme activity and cellular potency at best (data not shown)
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
and compounds of this type were not pursued further. Changing the 1Hꢀpyrrolo[2,3ꢀb]pyridine
(azaindole) core to a deazapurine core (CH → N) led to a complete or partial loss of potency in the HTꢀ2
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
assay for compound 42 and diminished activity for compound 46. Taken together, these results suggest
that polar groups with available lone pairs at this Cꢀ5 position are less optimal for potency.
With the hope that we could retain the positive aspect of compounds 22 and 31 (HTꢀ2 cell potency) as
well as 30 (AURA selectivity) we prepared a series of compounds with germinal substitution, as
exemplified by 43 to 49. Both 43 and 44 retained the cellular activity as was seen for both the (S)ꢀ and
(R)ꢀalanine derivatives 22 and 31, with desꢀfluoro pyrimidine 43 being the most potent (HTꢀ2 IC50
.058 ꢀM). However, the selectivity with respect to AURA was modest overall.
=
0
Since the addition of a methyl group on the glycine side chain of 23 with the R stereochemistry, as in
R)ꢀ31, improved selectivity vs AURA (23 vs 31) we employed a similar strategy to follow up on 43 by
(
increasing steric bulk of the germinal R substituent. Thus, the (R)ꢀ2ꢀaminoꢀ2ꢀmethylbutanoic acid based
compound (R)ꢀ45 was prepared and tested for its affinity in our kinase panel and cellular assays. A
similar level of cellular potency (HTꢀ2 IC = 0.081 ꢀM) was observed for (R)ꢀ45, but more importantly,
50
a dramatic increase in selectivity against AURA (K = 0.620 ꢀM) was achieved. As expected from
i
previous results, its enantiomer, (S)ꢀ47, was 10ꢀfold less selective for AURA (K = 0.070 ꢀM).
i
Furthermore, optimization at the 5 position of the 1Hꢀpyrrolo[2,3ꢀb]pyridine core, as shown for
compounds, VX-509 and 48 led to the most kinaseꢀselective analogs of the series with excellent cellular
potencies (HTꢀ2 IC < 0.1 µM).
5
0
16
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 18 of 101
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
Gluꢀ930
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
Leuꢀ932
Figure 5. Crystal structure of 44 in JAK2. Hinge hydrogen bonds to Leu 932 and Glu 930 highlighted in
green.
Glu 930
Leu 932
Figure 6. Crystal structure of VX-509 in JAK2. Hinge hydrogen bonds to Leu 932 highlighted in green.
Interestingly, a crystal structure of 44 revealed that the orientation of the 1Hꢀpyrrolo[2,3ꢀb]pyridine
hingeꢀbinding element was flipped relative to 22 to give the same orientation as 2 (Figure 5). Given the
relatively modest structural differences between 44 and 22, this significant change in binding mode was
17
ACS Paragon Plus Environment

Page 19 of 101
Journal of Medicinal Chemistry
1
2
3
unexpected. However, as shown in the structure of VX-509, these significant binding mode changes
between structurally very similar compounds were fairly common with this compound class (Figure 6).
This phenomenon of flipping binding modes for closely related compounds obviously complicated the
interpretation of SAR. Given the interesting binding behavior of these compounds in JAK2, a separate
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
3
4
manuscript is in preparation to address the topic.
With promising candidates in hand we decided to investigate JAK isotype selectivity further. At the
3
5
outset of our efforts we were concerned about mitigating JAK2ꢀmediated pharmacological effects.
Significant hematopoietic consequences were expected (e.g. anemia) with potent JAK2 inhibition.
Therefore, we sought to optimize JAK isotypes selectivity, with a particular emphasis on JAK3 versus
JAK2. In our enzyme binding studies using isolated active kinase domains, selectivity for JAK3 versus
JAK2, was typically less than 10ꢀfold for most of the compounds made in this optimization process.
Given the high degree of structural similarity in the ATP binding pockets of JAK2 and JAK3 (identical
except for JAK3 Cys 909 vs JAK2 Ser 936 and JAK3 Ala 966 vs JAK2 Gly 993), our observations are
consistent with the expectation that the potential for JAK3 vs JAK2 selectivity would be limited.
Furthermore, it is known that other domains of the JAK kinases have some degree of regulatory function
for the kinase, and we had indications from cellular data sets that the selectivity window between JAK3
3
6
and JAK2 was greater in cellular assays. Thus, this limited JAK3/JAK2 enzymatic selectivity might be
explained by the fact that truncated isolated protein kinase domains were used for K determinations.
i
This does not fully account for the subtle structural differences in the fullꢀlength proteins, particularly in
the context of their natural microenvironment, when bound to the gamma chain of the ILꢀ2 family of
cytokine receptors in cells. Thus, we chose to emphasize JAK isotypes selectivity characterization in the
more physiologically relevant biologic systems with the hope of minimizing offꢀtarget JAK2 inhibition.
18
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 20 of 101
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
Comparison of inhibition of the JAK3/1ꢀ or JAK2ꢀmediated phosphorylation of STAT5 in HTꢀ2 or TFꢀ1
cells following brief stimulation with ILꢀ2 or GMCSF, respectively, provided a useful selectivity ratio
for the compounds. Further, assessment of selectivity against JAK1/TYK2 was obtained by counterꢀ
screening our compounds in HeLa cells stimulated with IFNꢀα and measuring inhibition of STAT2
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
translocation to the nucleus. Results from the HeLa assays show that VX-509 has an IC50 of >10 ꢀM,
demonstrating a high degree of selectivity against these additional JAK isoforms. However, a more
relevant predictor of in vivo selectivity was also needed. To confirm and expand upon these findings,
primary human and mouse cell assays were used. Specifically, Tꢀcell proliferative response in mixed
lymphocyte reactions (MLR), ILꢀ2ꢀstimulated Tꢀcell blast assays from isolated peripheral blood
mononuclear cells (PBMCs), and erythroid colony formation assays (CFUꢀE) from human bone marrow
after stimulation with erythropoietin (a JAK2 dependent signaling pathway) were used to assess JAK
isotype selectivity ratios in a more physiologically relevant manner (Table 4).
19
ACS Paragon Plus Environment

Page 21 of 101
Journal of Medicinal Chemistry
1
2
3
4
5
Table 4. Enzyme (K nM ± SD, n), CellꢀBased Potency (IC nM ± SD, n) & Selectivity (ratio)
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
i
50
JAK isoform
Involved
Assay
VX-509
44
34
43
45
tofacitinib
2 ± 0.7, 1 ± 0.07, 1 ± 0.09,
1 ± 0.3,
10
2 ± 0, 1
1 ± 0.5,
12
0.5 ± 0.07,
7
3 ± 0.4, 2
JAK3
JAK1
JAK2
TYK2
JAK3
JAK1
JAK2
TYK2
3 ± 2, 3
N.A.
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
10
5
N.A.
4 ± 009,
6
11 ± 0, 1 3 ± 0, 1
2 ± 0.5,
13 ± 0, 4
12
11 ± 2, 2 10 ± 0, 1
2 ± 1, 5
N.A.
1 ± 0.9, 5
11 ± 0, 1
N.A.
11 ± 0, 1
HTꢀ2/ILꢀ2/Pꢀ
STAT5
99 ± 50, 90 ± 30,
112 ±
70, 5
58 ± 40,
5
81 ± 70,
2
30 ± 20,
72
JAK3/1
4
10
TFꢀ1/
GMCSF/Pꢀ
STAT5
2600 ±
1664, 4
1060 ±
600, 9
2927 ±
1350, 7
409 ±
142, 6
651 ±
13, 2
190 ± 137,
35
JAK2
Mouse 2ꢀWay
MLR
170 ±
100, 4
160 ±
110, 3
170 ±
70, 2
JAK3/1
JAK3/1
280, 1
N.A.
N.A.
N.A.
N.A.
N.A.
N.A.
N.A.
N.A.
8
60, 1
o
1 Human ILꢀ2
Tꢀcell Blast
240 ±
180, 2
130 ± 28,
2
N.A.
N.A.
Human CFUꢀE
7700 ±
6100, 2
6100 ±
3400, 2
5400 ±
2000, 2
2500 ±
2200, 2
380 ±
160, 10
JAK2
3
U/ml EPO
Human CFUꢀE
0.3 U/ml EPO
5300 ±
740, 2
3600 ±
1200, 2
3000 ±
1000, 2
1700 ±
325, 2
320 ±
150, 10
JAK2
HeLa IFNꢀa
STAT2
11900 ± >20000,
2800 ±
2000, 11
JAK1/ TYK2
N. A.
22.1
14.2
26
N.A.
10.6
7.6
7
3650, 3
5.3
31.3
1
CFUꢀE / MLR
(3 U/ml EPO)
4
38.1
6.7
5.7
6
CFUꢀE / MLR
0.3 U/ml EPO)
22.5
12
(
TFꢀ1/HTꢀ2
ratio
2
6
The selectivity ratios of JAK3ꢀversus JAK2ꢀmediated endpoints (TFꢀ1 / HTꢀ2) were superior to that of
binding selectivity for all the above compounds tested, ranging from 7ꢀ to 26ꢀfold. The three most
selective compounds, 44 and VX-509 and 34, were further evaluated for their functional activity in
primary human cellꢀbased assays dependent on either JAK3 or JAK2, namely the MLR and CFUꢀE
assays. Analogs 44 and VX-509 exhibited robust selectivity windows between JAK3ꢀ and JAK2ꢀ
dependent cell assays, with 38ꢀfold and 45ꢀfold selectivity, respectively, in favor of JAK3 (CFUꢀ E /
MLR). As a clinical benchmark reference tofacitinib was similarly evaluated for selectivity in our
20
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 22 of 101
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
assays. Tofacitinib proved to be generally less selective in this assessment with a selectivity ratios of 6ꢀ
and 7ꢀfold (TFꢀ1 / HTꢀ2) and (CFUꢀE / MLR) respectively. In these assays, the superior selectivity for
compound VX-509 with respect to tofacitinib was consistent with the results obtained for the TFꢀ1 / HTꢀ
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
ratios. Furthermore, VX-509 was evaluated along with tofacitinib in a primary human ILꢀ2 Tꢀcell blast
proliferation assay, which provided an additional functional measure of the ability of compounds to
inhibit JAK3 activity. Both compounds showed similar potency with IC50s at 0.14 ꢀM and 0.15 ꢀM
respectively.
Table 5. Pharmacokinetic parameter determinations in SpragueꢀDawley rats (single IV bolus dose)
a
DN-AUCinf
CL
(mL/min/kg)
T1/2
(h)
Vdss
(L/kg)
Dose
mg/kg)
Compound
(
(hr*ꢀg/ml)
3
4
4
4
1.0
1.9
0.46
1.76
0.42
0.60
0.74
36.2
9.45
39.7
41.1
24.3
2.6
2.2
6.1
1.5
43
2.0
1.2
1.3
4
5
1.7
4.6
4.1
VX-509
0.98
5.57
4.98
a
The AUCINF values are normalized to a 1 mg/kg dose
During our lead optimization process we routinely determined the pharmacokinetic parameters for
potent and selective compounds following a single IV bolus dose to SpragueꢀDawley (SD) rats. The
results for the most promising analogs are summarized in Table 5. While compound 44 exhibited the
lowest clearance coupled with an acceptable T , VX-509 also demonstrated reasonable exposure, with
1
/2
a moderate clearance and extended T1/2 of 5.1 hr. In addition, VX-509 had a superior kinase selectivity
profile (particularly against JAK2 and AURA) relative to 44. Based on these promising results, VX-509
was selected for further evaluation.
21
ACS Paragon Plus Environment

Page 23 of 101
Journal of Medicinal Chemistry
1
2
3
In vitro profiling indicated that VX-509 was moderately to extensively metabolized in liver microsomes
and S9 fractions in the mouse, rat, dog, monkey, and human (Table 6 and 7). This suggested that the
contributions to drug clearance from nonꢀCYP mediated metabolic pathways was minimal.
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
Table 6. Mean T1/2 and Predicted Hepatic Clearance for VX-509 at an Initial Concentration of 1 µM in
Liver Microsomes After Incubation at 37°C
Mean T1/2 (SD) and Mean Predicted Hepatic Clearances (SD)
Species
T1/2 (min)
Predicted Hepatic Clearance (mL/min/kg)
Mouse
Rat
19 ± 7.9
21 ± 1.6
34 ± 2.4
7 ± 0.4
69 ± 6.4
38 ± 0.9
24 ± 0.4
38 ± 0.3
17 ± 0.2
Dog
Monkey
Human
19 ± 1.2
Table 7. Mean T1/2 and Predicted Hepatic Clearance for VX-509 at an Initial Concentration of 1 µM in
Liver S9 After Incubation at 37°C
Mean T1/2 (SD) and Mean Predicted Hepatic Clearances (SD)
Species
T1/2 (min)
Predicted Hepatic Clearance (mL/min/kg)
Mouse
Rat
9 ± 1.01
21 ± 1.6
36 ± 3.4
6 ± 0.3
78 ± 1.2
37 ± 0.9
23 ± 0.6
39 ± 0.09
17 ± 0.9
Dog
Monkey
Human
20 ± 5.4
Other properties that favored selection of VX-509 included its high permeability in human colon
carcinoma (Cacoꢀ2) cell monolayer assays, indicating potential for good oral absorption, and
demonstration that it was a weak efflux pump protein substrate to Pꢀglycoprotein (Pꢀgp) and breast
cancer resistance protein (BCRP) with efflux ratios ranging between 2 and 3. Although, the efflux ratios
22
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 24 of 101
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
were small for Pꢀgp and BCRP, more than 50% reduction of these efflux ratios in the presence of
standard inhibitors was used to classify them as weak substrates. In addition, the hERG IC value for
50
VX-509 was determined to be 51.9 ꢁM and its solubility was found to be 6.38 ꢁg/mL in water.
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
Preclinical pharmacokinetics (Table 8A) of VX-509 was investigated in multiple species. Following a
nominal 1 mg/kg intravenous dose to rats and monkeys, plasma clearances were moderate at 24.3
mL/min/kg and 21.7 mL/min/kg, respectively; volume of distribution at steadyꢀstate (Vdss) ranged from
1.7ꢀ5.0 L/kg. In dogs, the clearance and Vdss values were significantly lower at 4.17 mL/min/kg and 0.4
L/kg, respectively. These lower values in dogs might be explained, at least in part, by differences in
aldehyde oxidase mediated metabolism in that species (results to be published elsewhere). The
percentage of VX-509 (1 ꢀM) bound to plasma proteins was 89, 93, 99, 80, and 81% in mouse, rat, dog,
monkey, and human plasma, respectively. Consistent with its high permeability and low to moderate
clearance, VX-509 exhibited high oral bioavailability in both rats and dogs (Table 8B).
Table 8. Mean Pharmacokinetic Parameters for VX-509
A) Single IV bolus dose
a
Dose
DN-AUCinf
CL
Vdss
T1/2
Species
(
mg/kg) (hr*ꢀg/ml) (ml/min/kg) (L/kg)
(hr)
Rat
0.98
0.88
0.74
4.06
24.3
4.17
4.98
0.39
5.57
1.58
Dog
Monkey
0.88
0.79
21.7
1.65
1.28
a
The AUCINF values are normalized to a 1 mg/kg dose
B) Single oral dose
a
Dose
DN-AUCinf
C
max
Tmax
(hr)
T1/2
F
Species
(
mg/kg)
(hr*ꢀg/ml)
(ꢀg/mg)
(hr)
%
Rat
10.7
9.65
6.75
44.6
1.73
7.46
1.50
1.00
5.87
5.82
91.8
100
Dog
a
The AUCINF values are normalized to a 10 mg/kg dose
23
ACS Paragon Plus Environment

Page 25 of 101
Journal of Medicinal Chemistry
1
2
3
4
5
Comparison of the estimated hepatic clearance values from liver microsomes in rats, dogs, and monkeys
Table 6: 38, 24, and 38 mL/min/kg, respectively) with clearance values obtained in vivo (Table 8: 24.3,
.2, and 21.7 mL/min/kg, respectively) showed that clearance was overestimated based on in vitro data
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
4
and suggested that VX-509 might exhibit acceptable clearance in humans. Due to its favorable potency,
selectivity and pharmacokinetic profile, VX-509 was selected for assessment of its potential for in vivo
efficacy in JAK3 mediated models of disease. Thus, VX-509 efficacy was evaluated in vivo in the rat
host versus graft (HvG) model (Figure 7). HvG response was elicited by the injection of allogeneic cells
isolated from Dark Agouti rat spleen into the footpad of the host Lewis rat. The HvG reaction is
primarily a Tꢀcellꢀmediated response, in which host T cells recognize major histocompatibility complex
(MHC) antigens presented by the antigen presenting cells on the graft cells. The graft cells migrate to
the host local popletial lymph node (PLN) and induce an immune response leading to hyperplasia of the
lymph node. The enlargement of host PLN peaks at Day 4, and the weight difference (delta) between
contralateral (nonꢀgrafted) and ipsilateral (grafted) PLNs was the efficacy assessment. Concomitantly,
ex vivo analysis of T cell activation of CD25 in whole blood from dosed animals was also evaluated as a
biomarker of JAK3 activity.
24
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 26 of 101
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 7. Dose dependent inhibition of popletial lymph node (PLN) hyperplasia by VX-509 in Rat HvG
Model.
VX-509 was evaluated at oral doses of 10 mg/kg BID, 25 mg/kg BID, 50 mg/kg BID, or 100 mg/kg QD.
Cyclosporine A was administered orally as a reference compound at a dose of 6 mg/kg QD. Popliteal
LNs were collected at termination. For ex vivo CD25 biomarker analysis, terminal blood samples at 1
and 14 hours (trough) post dose were used. VX-509 demonstrated statistically significant, doseꢀ
dependent inhibition on PLN hyperplasia as determined by delta measurement, which is the weight
difference between the grafted and nonꢀgrafted PLNs. As illustrated in Figure 7, VX-509 50 mg/kg BID
or 100mg/kg QD were significantly efficacious, inhibiting 66% (p = 0.019) and 94 % (p<0.01) of the
PLN hyperplasia, respectively relative to CyclosporinA (CsA). There was a partial but nonꢀsignificant
inhibition (50%, p = 0.101) at 25 mg/kg BID group relative to CsA. Tofacitnib, when evaluated in the
HvG model, also demonstrated statistically significant inhibition of PLN hyperplasia. The compound
25
ACS Paragon Plus Environment

Page 27 of 101
Journal of Medicinal Chemistry
1
2
3
inhibited the PLN hyperplasia by 87% at 20 mg/kg TID, 88% at 20 mg/kg BID or 100% at 60 mg/kg
BID as compared to CsA (100%). (supplementary Figure 13). There was significant reduction of the
CD25 expression by tofacitinib at 1 hour post dose in all groups tested. (supplementary Figure 14)
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 8. Dose dependent decrease in ex vivo stimulated biomarker CD25 by VX-509 in Rat HvG
model.
Terminal whole blood samples at 1 and 14 hours postꢀdose were analyzed using ex vivo 4 betaꢀphorbol
12ꢀmyristate 13ꢀacetate (PMA) and antiꢀCD28 antibodyꢀstimulated CD25 expression. In the absence of
stimulation with PMA and antiꢀCD28 antibody, CD25 expression in CD3 (+) T cells was below 5%,
whereas CD25 expression reached approximately 85% within 24 hours after stimulation. There was
significant reduction of CD25 expression by VX-509 at 1 hour postꢀdose in all of the dose groups tested:
5
9 % (P<0.01) inhibition at 10 mg/kg BID, 51 % (P<0.01) at 25 mg/kg BID, 71% (P<0.01) at 50 mg/kg
26
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 28 of 101
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
BID, and 92% (P<0.01) was observed at 100 mg/kg QD. There was also significant reduction (66 %,
p<0.01) in CD25 expression in the 100 mg/kg QD group at trough (23 hours post dose)(Figure 8),
supporting the potential for a QD dosing regimen in clinical studies. In this study, VX-509 exhibited
significant doseꢀdependent immunosuppressive activity. Ex vivo whole blood analysis of CD25
expression demonstrates that VX-509 effectively attenuated Tꢀcell activation at 1 hour postꢀdose.
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
Given the potency, its unique selectivity profile favoring JAK3 as determined using cellular assays, the
pharmacokinetic profile and its efficacy in preꢀclinical pharmacologic models of aberrant immune
function, VX-509 appears to have potential clinical utility for treatment of a variety of immuneꢀbased
diseases and results from clinical studies will be reported in due course.
Synthesis
Synthesis of final compounds relied on ready access to common 1Hꢀpyrrolo[2,3ꢀb]pyridine and
deazapurine boronates 50. These key intermediates were all similarly prepared according to the general
route outlined in Scheme 1.
Scheme 1. Conditions: a) NaH, TsCl, THF; b) Br , CH Cl ; c) bis(pinacolato)diboron, Pd(PPh ) ,
2
2
2
3 4
KOAc, 1,4ꢀdioxane.
27
ACS Paragon Plus Environment

Page 29 of 101
Journal of Medicinal Chemistry
1
2
3
NꢀTosyl protection starting from known intermediates A, followed by bromination at Cꢀ3 enabled
efficient preparation of the 3ꢀbromoꢀ1Hꢀpyrrolo[2,3ꢀb]pyridine or deazapurines C. Treatment with
bis(pinacolato)diboron and palladium tetrakis(triphenylphosphine) provided the desired Suzuki coupling
partners, intermediates 50a-e.
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
Scheme 2a. Conditions: a) PhCH NH , DIPEA, THF, reflux, 1ꢀ2h or IPA, 90 °C, (54ꢀ79%); b)
2
2
PhCH NH , neat, (79%); c) PhCH NH , 150 °C (ꢀW) (40ꢀ55%) d) i: Pd(PPh ) , DME, 2 M Na CO ,
2
2
2
2
3 4
2
3
reflux ii: NaOtBu, 130ꢀ160 °C, 10ꢀ20 min (ꢀW) (25ꢀ89%)
Scheme 2b. Conditions: a) Pd(PPh ) , DME, 2M Na CO , reflux (51%; X = CCl); b) Oxone, CH OHꢀ
3
4
2
3
3
H O (1:1) (32%; X = CCl); c) i: PhCH NH , EtOH, reflux, 16 h (94%; X = CCl), ii: 3N NaOH, CH OH
2
2
2
3
(75%; X = CCl) or NaOtBu, 130ꢀ160 °C, 10ꢀ20 min (ꢀW)
28
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 30 of 101
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
The syntheses of analogs containing the benzylamine group and with central core ring variations in
Table 1 were achieved via the general route described in Scheme 2. The sequence involved Suzuki
coupling of protected boronates 50a,b with suitably substituted aryl halides followed by deprotection as
shown in scheme 2a. Thus, SnAr displacement of suitable dihalopyrimidine regioisomers with
benzylamine, provided haloꢀcoupling partners 51a,b and 52 in moderate to good yields. Suzuki
coupling under standard conditions was followed by removal of the tosyl group under basic conditions
to provide compounds 4, 6 and 7, as well as pyridine derived analogs 9. The corresponding phenyl
analog 10 was accessed using commercial Nꢀbenzylꢀ3ꢀbromoaniline following the same sequence.
Compounds 11−21 were prepared most conveniently by the same sequence starting from either 2,4ꢀ
dichloropyrimidine or 2,4ꢀdichloroꢀ5ꢀfluoropyrimidine and the corresponding amines. Thus, chloride
displacement followed by Suzuki coupling and tosyl deprotection delivered compounds 11-21.
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
Scheme 2b describes the chemistry to gain access to compounds 5 and 8. Interemediates 53 and 54 were
prepared in moderate yield via Suzuki coupling of boronic esters 50a and 50b with 4ꢀchloroꢀ2ꢀ
methylthiopyrimidine followed by oxidation of 53 to provide the corresponding sulfones 54. This
sequence allowed for facile SnAr displacement with benzylamine to obtain, after deꢀtosylation, the
desired analogs 5 and 8.
The syntheses of amino acidꢀbased compounds 22−49 depicted in Table 3 were accomplished in a 5ꢀstep
protocol from commercially available NꢀBoc amino acids M (Scheme 3). The requisite amide cap was
installed via the coupling of M with trifluoroethyl amine hydrochloride under standard conditions using
either HATU or EDC/HOBt to furnish, after Boc deprotection with CH Cl ꢀTFA, the corresponding
2
2
amides N in good overall yield. Nꢀarylation with the either 2,4ꢀdichloropyrimidine, 2,4ꢀdichloroꢀ5ꢀ
fluoropyrimidine or 2,4ꢀdichlorotriazine provided the desired coupling partners Q for assembly of the
29
ACS Paragon Plus Environment