Mapping the kinase domain of janus kinase 3

Article abstract of DOI:10.1016/S0960-894X(03)00657-7

The utilization and impact of parallel synthesis on lead exploration around initial hit oxindole (1) are described. The emergent SAR, analogue design and functional impact will also be detailed.

Full text of DOI:10.1016/S0960-894X(03)00657-7

Bioorganic & Medicinal Chemistry Letters 13 (2003) 3105–3110
Mapping the Kinase Domain of Janus Kinase 3
Christopher Adams, David J. Aldous,* Shelley Amendola, Paul Bamborough,
Colin Bright, Sarah Crowe, Paul Eastwood, Garry Fenton, Martyn Foster,
Trevor K. P. Harrison, Sue King, Justine Lai, Christopher Lawrence,
Jean-Philippe Letallec, Clive McCarthy, Neil Moorcroft, Kenneth Page, Sudha Rao,
Jane Redford, Shazia Sadiq, Keith Smith, John E. Souness, Sukanthini Thurairatnam,
Mark Vine and Barry Wyman
Aventis Pharmaceuticals, 1041 Route 202–206, PO Box 6800, Bridgewater, NJ 08807, USA
Received 24 April 2003; accepted 22 May 2003
Abstract—The utilization and impact of parallel synthesis on lead exploration around initial hit oxindole (1) are described. The
emergent SAR, analogue design and functional impact will also be detailed.
# 2003 Published by Elsevier Ltd.
There has been intensive effort to discover and exploit
selective kinase inhibitors.1,2 The ubiquitous nature of
kinases, coupled to the high homology in the ATP
binding domain, have presented medicinal chemists with
a myriad of difficulties in advancing inhibitors for
therapeutic evaluation.
In order to explore the therapeutic hypothesis, we
embarked upon a focused screening. A clone of the
kinase domain10 was generated and a HTRFassay for-
matted.11 The initial screen of the kinase biased set
yielded oxindole 1 as a mM hit (Fig. 1).
To further refine the inhibitor, a homology model was
generated. The model used as a template the complex
between cyclic adenosine monophosphate (cAMP)
dependent protein kinase (cAPK).12 cAPK was chosen
because at that time it was the only protein kinase
whose structure had been solved in its active, or closed,13
conformation. The overall sequence identity between
cAPK and human JAK3 (hJAK3) over the 256 aligned
positions in the catalytic domain is 26%. Some time
later, the insulin receptor tyrosine kinase (IRK), that
shares 32% sequence identity with JAK3, was crystal-
lized in an active conformation14 with an ATP analo-
We became interested in Janus Kinase 3 (JAK 3) pro-
tein tyrosine kinase, as a potential immune modulating
target due to the functional impact of down-regulated
JAK 3 signal in an immune compromised patient sub-
population.3ꢀ5 The sub-population of the severe com-
promised immune deficient (SCID) and X-linked severe
compromised immune deficient (X-SCID) patient pool
displayed abnormal function in the JAK 3 signaling
pathway either due to a mutation of the gc allele which
mutated the gc chain so JAK 3 did not bind or a muta-
tion directly impacting JAK 3.
gue, adenylyl imido-diphosphate (AMP-PNP).
A
Whilst the SCID and X-SCID patients were immune
compromised we postulated that in auto-immune dis-
eases where the immune system was up-regulated such
as arthritis or diabetes modulation of JAK 3 might be
an effective therapy.6ꢀ9
*Corresponding author. E-mail: david.aldous@aventis.com
Figure 1. Initial oxindole.
0960-894X/$ - see front matter # 2003 Published by Elsevier Ltd.
doi:10.1016/S0960-894X(03)00657-7

3106
C. Adams et al. / Bioorg. Med. Chem. Lett. 13 (2003) 3105–3110
second hJAK3 model was constructed using this as a
template. The main differences between the two models
arise from the more open pocket in the IRK-based
model. Although IRK is more closely related to hJAK3
than is cAPK overall, when only the 29 residues closer
than 5.5 A from the AMP-PNP in the 1IR3 structure
are considered, 17 (59%) in IRK are identical to their
aligned positions in hJAK3, compared to 19 (66%) in
cAPK. The cAPK-based model was used to carry out
most of this work and that is referred to in the text.
of the SB203580–p38 kinase complex,19 have shown a
hydrophobic part of the inhibitors projecting into a
similar area at the rear of the ATP pocket. The hydro-
phobic residues Leu 875, Leu 900 and Met 902 lie
around this region. However, acidic and basic side
chains (Lys 855 and Glu 871) are also nearby. In many
complexes of kinases with ATP, the conserved equiva-
lent of Lys 855 binds to the ATP phosphates and forms
a salt bridge with Glu 871.20 A diverse set was defined to
probe this region (Fig. 3).
In JAK 3, ATP and the oxindole inhibitor were pre-
dicted to donate a hydrogen bond to the backbone car-
bonyl of residue Glu903 and to accept a hydrogen bond
from the backbone NH of Leu 905 (Fig. 2).
We identified proprietary and commercially available
aldehydes to deliver a two-dimensional array by con-
densation of oxindole cores (Scheme 1).
The product precipitated out resulting in >85% pure
product.21 The final array consisted of 700 inhibitors
which were then screened against the JAK 3 kinase
domain.
The oxindole 1 was docked into the modeled protein
active site, initially by fitting the acceptor–donor pair of
1 onto the acceptor–donor pair of ATP. As observed in
the Fibrobalst Growth Factor (FGF) Receptor com-
plex,15 the double bond was modeled in the Z-isomer
configuration. The Z-isomer is the dominant configur-
ation due to an internal hydrogen bond between the
oxindole oxygen and the pyrrole NH.16 When docked,
the oxindole core sits in the space between the two
kinase lobes, making contact with Val 836, Ala 853, Val
884, Met 902 and Leu 905.
We analyzed the impact of the substitution at the 5-
position of the oxindole core keeping the pyrrole con-
stant (Table 1).
Nitro afforded no improvement whilst the keto, acid
and amide improved potency approximately 5-fold.
However, reversing the amide reduced potency, imply-
ing that electron withdrawal was important which was
reinforced further by the potent keto thiophene. Inter-
estingly, the reduced hydroxy thiophene was potent,
albeit an order of magnitude less than the keto analo-
gue, implying a strong aryl interaction in this region.
The 3-pyridyl reinforced the need for electron with-
drawal with bulk at the 5-position.
Analysis of the docked inhibitor indicated that the pyr-
role ring projects outwards in the direction of solvent,
and contacts Leu 956 and Gly 908. Although vectors
leading away from the pyrrole ring at the 40- and 50-
position point towards solvent, the ring itself lies within
the ATP pocket. Variation at this region in the library
was intended to explore this space fully, to identify
possible interaction sites both inside the pocket and
outside on the lip, but also to explore the potential for
the attachment of groups to give better pharmacological
properties without necessarily interacting with the
kinase.
In contrast, before and after refinement, it could be seen
that the vector of the CH atom at the 5-position of the
oxindoles core pointed into an unfilled region of the
active site. Crystal structures of complexes of CDK2–
staurosporine17 and cAPK–staurosporine,18 as well as
Figure 3. 5-Position probes.
Scheme 1. Condensation to target inhibitors (IMS, industrial methyl-
ated spirit).
Figure 2. JAK 3 homology model.

C. Adams et al. / Bioorg. Med. Chem. Lett. 13 (2003) 3105–3110
Table 1. Selected 5-position substitution with pyrrole constant
3107
Table 2. Selected aldehyde variation with keto thiophene constant
Compd
10
Aldehyde R0
JAK 3 inhibition
(IC50, mM)
Compd
5-Position substitution R
JAK 3 inhibition
(IC50, mM)
2.3
4.9
1
2
H
NO2
0.900
0.710
11
3
4
5
0.168
0.213
0.21
12
13
7.1
CO2H
0.70
6
0.870
14
0.20
7
8
9
0.190
0.026
0.027
15
8
0.26
0.026
When the keto thiophene was constant, the impact of
varying the aldehyde was defined (Table 2).
Figure 4. 30-Pyridyl oxindole inhibitor.
Electron-donating aromatics increased potency with the
most potent being pyrrole.
One of the most active compounds in the library, IC50
27 nM, had a 30-pyridyl at the 5-position (Fig. 4).
The 30-pyridyl inhibitor 9 was docked into the active site
of JAK 3 (Fig. 5).
The pyridyl ring is positioned deep within the pocket,
surrounded by Leu 875, Leu 900, Met 902 and the
hydrophobic part of the Lys 855 side chain. Interest-
ingly, this part of the crystal structure of FGF receptor
complexed with SU5402 is occupied by a group of three
water molecules, probably hydrogen-bonding to Lys-
514 and Glu-531, which are the residues in the FGF
receptor corresponding to Lys 855 and Glu 871 of JAK
Figure 5. Docked 30-pyridyl inhibitor 9.

3108
C. Adams et al. / Bioorg. Med. Chem. Lett. 13 (2003) 3105–3110
3. The burial of hydrophobic groups into this pocket
and the consequent displacement of these waters prob-
ably accounts for the increased affinity of the 30-pyridyl
compared to the initial oxindole 1.
The introduction of nitrogen at the 4-position, 16,
resulted in a 10-fold loss in potency. The drop off in
potency for 16 can be rationalized by the introduction
of conformational mobility (Fig. 6).
Although the precipitation of the target analogue was a
great asset in the rapid construction of the library, it
highlighted solubility as a key property for analogue
progression and inhibitor design. The free base of the 30-
pyridyl inhibitor 9 had a solubility in water of <1 mg/L
whilst the mesylate salt had a solubility of 3600 mg/L.
The synthesis of the 30-pyridyl inhibitor 9 is outlined in
Scheme 2.
The NMR studies showed a single isomer to be present
for analogues 9, 17 and 18. In the case of analogue 16, a
mixture of conformations was observed. Others have
observed this conformational stability and mobility.16 We
rationalized the drop off in potency for 18 as a distribution
of the hydrogen bonding of the NH of the oxindole.
We designed a filtering cascade for progression of JAK
3 inhibitors that relied upon the inhibition of IL-2
induced proliferation on both mouse CTLL24 cells and
human T-cells. In addition, the analogues were required
to inhibit JAK 3 and STAT 5 phosphorylation.25
As the nascent SAR implied that electron withdrawing
imparted superior potency we decided to explore the
SAR further by the synthesis of aza oxindoles.22,23 The
aza analogues introduced an electron deficient aryl
along with an additional salt handle to enhance solubil-
ity (Table 3).
The basis for the filtering cascade stemmed from the
understanding of the function of Janus kinases being
cytoplasmic tyrosine kinases involved in signal trans-
duction. Binding of IL-2 to its receptor triggers oligo-
merization of the receptor chain, which activates Janus
kinases. Activated JAK 3 then autophosphorylates,
phosphorylates a member of the STAT family, STAT 5
Figure 6. Conformational studies on aza oxindoles 16.
Scheme 2. Synthesis 30-pyridyl oxindole: (a) KBr, Br2/H2O, 94%;
(b) Pd(PPh3)4, THF, 30%; (c) EtOH/piperidine, 80%; (d) MeSO3H,
THF, 87%.
Table 3. Selected 5-aryl substitution with pyrrole constant
Figure 7. Signaling through the IL-2 receptor.
1
Compd
X, X, 2X
JAK 3 inhibition
(IC50, mM)
1
9
16
X, X, 2X=CH
0.027
0.235
X=N
1X, 2X=CH
1X=N X,
2X=CH
2X=N X,
1X=CH
17
18
0.030
0.350
Figure 8. Anti-phopsphotyrosine antibody probed Western blot
showing concentration-dependent inhibition of IL-2 induced STAT 5
tyrosine phosphoryltion by inhibitor 17 in CTLL cells.

C. Adams et al. / Bioorg. Med. Chem. Lett. 13 (2003) 3105–3110
3109
on specific tyrosyl residues. Upon phosphorylation of
specific, STAT proteins become activated and form
homo or hetero dimeric complexes that translocate to
the nucleus, bind to specific DNA sequences and alter
the expression of cytokine regulated genes (Fig. 7).26,27
analogues 16 and 18 did not inhibit STAT 5 phosphoryl-
ation and had reduced efficacy in the IL-2 proliferation
assays (Table 4). Neither analogue was progressed further.
The aza oxindole 17 was synthesized as outlined in
Scheme 3.
Inhibitors 9 and 17 inhibitited STAT 5 phosphorylation
(Fig. 8), JAK 3 kinase activity and JAK 3 phosphoryl-
ation (Table 4) with similar potency. Inhibitor 17
exhibited an improved efficacy profile in the IL-2 pro-
liferation assays, which we speculated might be due to
differences in physical form. The inhibitors 9 and 17 were
not profiled against mouse JAK 3i, therefore the com-
pression in the mouse CTLL assay can only be speculated
as being the result of orthologue differences. However,
Inhibitors 9 and 17 were profiled ip in cassette dosing
PK studies which clearly demonstrated the enhance-
ment in bioavailability due to the formation of the salt,
Table 4 (Table 5).
Inhibitor 9 attenuated the increase in ear weight in the
oxazolone-induced ear oedema model,28 indicating an
action upon the reactivation of the immune response
and/or upon the inflammatory response, which is pro-
duced. The efficacy of 9 was comparable to that of the
steroid dexamethasone (10 mg/ear) administered in ace-
tone (Fig. 9).
Table 4. Whole cell profiling
Compd
>JAK 3
inhibition
(IC50, mM)
Inhibition of
phosphorylation
at 3 mM
Inhibition of IL-2
induced proliferation
(IC50, mM)
In conclusion, we identified a weak inhibitor of JAK 3
which formed the basis for parallel synthesis to construct
STAT 5
JAK 3
CTLL
(mouse)
T-Cell
(human)
9
0.027
0.235
0.030
0.350
Yes
No
Yes
No
Yes
Yes
Yes
—
0.76
0.92
0.45
2.3
0.25
0.70
0.03
2.0
Table 5. CombiPK data following ip administration at 1 mg/kg
16
17
18
Compd
Cp max
(ng/mL)
Tmax (h)
AUC0ꢀ1
Terminal
T1/2 (h)
(h ng mLꢀ1
)
9
600–1400
(<50)a
700–1400
0.5
0.5
7686
(<500)a
4579
3.0
3.3
17
aFree base.
Scheme 3. Synthesis 30-pyridyl aza oxindoles 17: (a) Pd(PPh3)4, THF/
H2O, NaHCO3, 52%; (b) DMF, Á, 52%; (c) H2/Pd/C, IMS, 81%; (d)
NBS, tBuOH, H2O, 88%; (e) H2/Pd/C, IMS, Et3N, 36%; (f) EtOH/
piperidine, Á, 80%; (g) MeSO3H, THF, 87%.
Figure 9. Oxazolone induced ear oedema in mouse.

3110
C. Adams et al. / Bioorg. Med. Chem. Lett. 13 (2003) 3105–3110
a 700 component library. The library was designed
based on the docking into the homology model of the
hit 1. Low-nM inhibitors emerged from the library that
had a good correlation of JAK 3 inhibition linked to
signal related functional impact. Finally, the inhibitor 9
demonstrated an effect comparable to dexamethasone in
the ear odema model.
13. Zheng, J.; Knighton, D. R.; Xuong, N.-H.; Taylor, S. S.;
Sowadski, J. M.; Ten Eyck, L. F. Protein Sci. 1993, 2, 1559.
14. Hubbard, S. R. EMBO J. 1997, 16, 5572.
15. Mohammadi, M.; McMahon, G.; Sun, L.; Tang, C.;
Hirth, P.; Yeh, B. K.; Hubbard, S. R.; Schlessinger, J. Science
1997, 276, 955.
16. Sun, L.; Tran, N.; Tang, F.; App, H.; Hirth, P.; McMa-
hon, G.; Tang, C. J. Med. Chem. 1998, 41, 2588.
17. Lawrie, M. A.; Noble, E. M.; Tunnah, P.; Brown, N. R.;
Johnson, L. N.; Endicott, J. A. Nat. Struct. Biol. 1997, 4, 796.
18. Prade, L.; Engh, R. A.; Girod, A.; Kinzel, V.; Huber, R.;
Bossemeyer, D. Structure 1997, 4, 1627.
References and Notes
1. Sridhar, R.; Hanson-Painton, O.; Cooper, Denise, R.
Pharmaceut. Res. 2000, 17, 1345.
2. Fabbro, D.; Parkinson, D.; Matter, A. Curr. Opin. Phar-
macol. 2002, 2, 374.
3. Russell, S. M.; Johnston, J. A.; Noguchi, M.; Kawamura,
M.; Bacon, C. M.; Friedmann, M.; Berg, M.; McVicar, D. W.;
Witthuhn, B. A.; Silvennoinen, O. Science 1994, 266, 1042.
4. Noguchi, M.; Rosenblatt, H. M.; Filipovich, A. H.; Adel-
stein, S.; Modi, W. S.; McBride, O. W.; Leonard, W. J. Cell
1993, 73, 147.
19. Tong, L.; Pav, S.; White, D. M.; Rogers, S.; Crane, K. M.;
Cywin, C. L.; Brown, M. L.; Pargellis, C. A. Nat. Struct. Biol.
1997, 4, 311.
20. Johnson, L. N.; Noble, M. E.; Owen, D. J. Cell 1996, 85, 149.
21. 85% by LC/MS, was set as the lower limit for acceptance
of each library component.
22. Robinson, R. P.; Donahue, K. M.; Son, P. S.; Wagy, S. D.
J. Heterocycl. Chem. 1996, 33, 287.
23. Daisley, R. W.; Hanbali, J. R. Synth. Commun. 1981, 11,
743.
5. Leonard, W. J. Annu. Rev. Med. 1996, 47, 229.
6. Leonard, W. J.; O’Shea, J. J. Annu. Rev. Immunol. 1998, 16,
293.
7. Sugamura, K.; Asao, H.; Kondo, M.; Tanaka, N.; Ishii, N.;
Ohbo, K.; Nakamura, M.; Takeshita, T. Annu. Rev. Immunol.
1996, 14, 179.
8. Sudbeck, E.; Uckun, F. IDrugs 1999, 2, 1026.
9. Cetkovic-Cvrlje, M.; Dragt, A. L.; Vassilev, A.; Liu, X. P.;
Uckun, F. M. Clin. Immunol. 2003, 106, 213.
10. The protein was expressed with an N-terminal FLAG tag
using the in-house expression vector pYOUNG and the yeast
strain Kluveromycas lactis. The recombinant protein was
expressed intracellularly and was purified using M2 affinity
resin.
11. Homogenous time resolved fluorescence (HTRF) assay: A
selective synthetic tyrosine kinase containing substrate (biot-
bA-bA-bA-L-P-L-D-K-D-Y-Y-V-V-R-E-P-G-Q[NH]2 recog-
nized by JAK3 (catalytic domain) is labeled with biotin. The
substrate (1 mM) is phosphorylated by JAK3 (50 ng/assay
point) in the presence of ATP (8 uM), DTT and Mg2+, 10 min
at room temperature. APC-labeled streptavidin and a selected
cryptate-labeled anti-phosphotyrosine are then added into the
well in order to generate the specific signal that is proportional
to the phoshorylated tyrosine concentration. After 1 h of
incubation at room temperature, plates are read and the
%dmax is determined using Packard discovery counter. The
final DMSO concentration in the assay was 8%. The standard
for the assay was oxindole 1.
24. IL-2 induced murine CTLL cell proliferation: Seed CTLL’s
(2ꢁ104 cells/well) in 96-well plate, in the absence of IL-2, for 6
h. Add 10 U/mL recombinant human IL-2 at time zero. Add
JAK 3 inhibitors in DMSO (final concentration 0.5%) at time
zero. Add 1 mCi/well [3H]-thymidine at 18 h of culture. Har-
vest cells after 24 h, on to filter plates, and count scintillations
[3H]-thymidine incorporation into DNA reflects extent of
proliferation.
25. IL-2 induced JAK 3 and STAT5 tyrosine phosphorylation:
Pretreat CTLL cells with JAK 3 inhibitors for 4 h. Stimulate
cells with 10 U/mL IL-2 for 20 min, lyse cells and immuno-
precipitate with anti-mouse JAK 3 and STAT5 antibodies.
Subject immunoprecipitated samples to SDS-PAGE electro-
phoresis. Probe electrophoresized blots with anti-phosphotyro-
sine antibody.
26. Miyazaki, T.; Kawahara, A.; Fujii, H.; Nakagawa, Y.;
Minami, Y.; Liu, Z. J.; Oishi, I.; Silvennoinen, O.; Witthuhn,
B. A.; Ihle, J. N. Science 1994, 266, 1045.
27. Johnston, J. A.; Kamura, M.; Kirken, R. A.; Chen, Y. Q.;
Blake, T. B.; Shibuya, K.; Ortalso, J. R.; McVicar, D. W.;
O’Shea, J. J. Nature 1994, 370, 151.
28. Protocol for oxazolone induced ear odema: On day 1,
female BALB/c mice were sensitized by topical application of
oxazolone (100 mL of 2% solution in acetone) to the shaved
abdomen. On day 7, acetone (negative control group) or oxaz-
olone (2% solution in acetone, remainder of animals) was
applied topically to the right ear (10 mL on each side of the ear).
Test compounds or appropriate vehicles (acetone) were admi-
nistered topically to the right ear (10 mL on each side of the ear)
0.5 h after challenge. Mice were killed 24 h after challenge. A
6-mm disc of challenged (right) ear was removed and weighed.
12. Zheng, J.; Trafny, E. A.; Knighton, D. R.; Xuong, N.-H.;
Taylor, S. S.; Ten Eyck, L. F.; Sowadsky, J. M. Acta Crystal-
logr. 1993, D49, 362.