Kinase array design, back to front: Biaryl amides

Article abstract of DOI:10.1016/j.bmcl.2008.08.051

New kinase inhibitors can be found by synthesis of targeted arrays of compounds designed using system-based knowledge as well as through screening focused or diverse compounds. Most array strategies aim to add functionality to a fragment that binds in the

Full text of DOI:10.1016/j.bmcl.2008.08.051

Bioorganic & Medicinal Chemistry Letters 18 (2008) 5285–5289
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry Letters
journal homepage: www.elsevier.com/locate/bmcl
Kinase array design, back to front: Biaryl amides
*
Ian Baldwin, Paul Bamborough , Claudine G. Haslam, Suchete S. Hunjan, Tim Longstaff,
Christopher J. Mooney, Shila Patel , Jo Quinn à, Don O. Somers
GlaxoSmithKline R&D, Medicines Research Centre, Gunnels Wood Road, Stevenage, Hertfordshire SG1 2NY, UK
a r t i c l e i n f o
a b s t r a c t
Article history:
New kinase inhibitors can be found by synthesis of targeted arrays of compounds designed using system-
based knowledge as well as through screening focused or diverse compounds. Most array strategies aim
to add functionality to a fragment that binds in the purine subpocket of the ATP-site. Here, an alternative
pharmacophore-guided array approach is described which set out to discover novel purine subpocket-
Received 25 July 2008
Revised 15 August 2008
Accepted 16 August 2008
Available online 22 August 2008
binding groups. Results are shown for p38
a and cFMS kinase, for which multiple distinct series with
nanomolar potency were discovered. Some of the compounds showed potency in cell-based assays and
good pharmacokinetic properties.
Keywords:
Kinase array design
p38 MAP kinase
p38a
Ó 2008 Elsevier Ltd. All rights reserved.
cFMS
Inhibitors
Structure-based drug design
Most kinase inhibitors bind in the purine subpocket of the ATP-
site using conserved H-bonding patterns to the ‘hinge’. Typically,
arrays designed to target kinases have used a single purine sub-
pocket core, modified to extend into different parts of the ATP-
site.1 While often successful, such arrays only probe one purine
subpocket group at a time. Here, we report a different approach
to identify novel purine subpocket binders of kinases.
Some inhibitors interact with the lipophilic interior of the
ATP-site (or ‘back pocket’) in kinases where this is accessible
to gain potency and/or selectivity. Some of these, such as 1, bind
to the ‘DFG-in’ conformation of their targets, while others such
as Gleevec, BIRB-796 and 2 bind in the DFG-out mode.2,3 The
back-pocket binding groups of these compounds make extensive
interactions and contribute greatly to affinity. Here, we describe
the use of these back-pocket binding groups as starting points
from which to build out towards the front of the ATP-site. A
similar ‘back-to-front’ approach was used to optimise the biaryl
urea series to give BIRB-796.2c The work reported here differs in
its intent, to identify new classes of kinase inhibitors for lead
discovery as opposed to lead optimisation, and its use of arrays
and pharmacophore-guided reagent selection together with
multiple back-pocket groups utilising both DFG-in and DFG-out
binding modes.
O
Back
pocket
N
H
Me
Back
pocket
Gleevec
Purine
subpocket
N
1
Me
H
O
NH
N
N
O
Purine
subpocket
N
N
H
Me
O
NH
Purine
subpocket
Back
pocket
DFG-out
pocket
O
NH
DFG-out
pocket
2
N
MeN
N
N
O
Crystal structures of 1 and 2 bound to p38a showed a critical hydro-
gen-bond from their cyclopropylmethyl amide carbonyls to the pur-
ine subpocket donor NH of Met109 (the ‘hinge’).3 The short distance
between this and the tolyl ring, together with the prevalence of aro-
matic rings in the purine pocket of kinase inhibitors, dictated the
choice of chemistry used to build the arrays. Suzuki coupling al-
lowed the insertion of a wide range of haloaryl rings targeting the
purine subpocket.
* Corresponding author. Tel.: +44(0) 1438 763246; fax: +44(0) 1438 763352.
E-mail address: paul.a.bamborough@gsk.com (P. Bamborough).
Present address: Biochemistry Department, Cambridge University, Tennis Court
Road, Cambridge CB2 1GA, UK.
Present address: Siena Biotech S.p.A., Strada del Petriccio e Belriguardo, 53100
Siena, Italy.
à
0960-894X/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2008.08.051

5286
I. Baldwin et al. / Bioorg. Med. Chem. Lett. 18 (2008) 5285–5289
Back-pocket groups found in potent kinase inhibitors were
chosen as starting points. A–C (Scheme 1) are substructures of
potent biphenyl amide p38 inhibitors, such as 1 (Ki = 12 nM)
a
and 2 (Ki = 4.4 nM).3 A is found within compounds that bind in
the DFG-in mode, while those containing B and C bind in the
DFG-out mode. D is a fragment of Gleevec, a potent inhibitor of
cAbl and other kinases including LCK and cFMS.2a,4 Other frag-
ments not included in this report were added to probe the back-
pockets of other kinases.
Boronic acid esters A–D were prepared and coupled to a selec-
tion of aryl halides as shown in Scheme 1. Halides were chosen
from a database of available reagents with the help of a pharmaco-
phore model.5 This defined the position of an aromatic ring on the
tolyl ring together with a hydrogen-bond from the hinge Met109
NH. Coordinates were taken from the X-ray complex of 2 with
p38
The unreacted aryl bromides used to make compounds 3, 6, 8,
14 and 19 all had Ki > 2 M against p38 , as did intermediate F
(Scheme 1). However, the products of arrays A–C were expected
to include potent p38 inhibitors. One hundred and seventy-two
a
(Fig. 1).3b
Figure 1. The pharmacophore search used to obtain the purine-site reagent list.5
The p38 X-ray complex of 2 (magenta) defines the binding site. Pharmacophoric
features (spheres) include the H-bond acceptor from Met109 NH (red, radius 1.25
and 1.75 Å for the projected point), tolyl ring aromatic atoms (blue, radii 0.5 Å) and
excluded volumes (black, radii 1.7 Å). A representative isoquinoline hit that led to
14 is shown in green.
a
l
a
a
compounds from these arrays were successfully purified and were
submitted for test against a panel of protein kinases. Table 1 shows
the number of products from each array that were assayed. The 87
sub-100 nM compounds spanned 60 different purine subpocket-
Table 1
Distribition of p38a
compound activities by array6
Array
A
B
C
A–C
Purine pocket groups
Number tested
Number Ki <1 lM
Number Ki <100 nM
Number Ki <10 nM
61
47
17
4
57
50
30
6
54
49
40
7
172
146
87
94
84
60
12
O
B
Me
O
B
Me
17
O
O
A
a
E
O
OH
O
NH
binding groups. Table 2 shows selected compounds and their
p38 activities.
X-ray structures of several examples have been solved in com-
plex with p38 . Though alternative binding modes cannot be ex-
Me
a
Me
O
B
Me
I
I
O
a
B
cluded for all compounds, especially the weaker inhibitors, the
proposed back-pocket groups occupy the back-pocket of the site
in all examples solved, and make the same extensive interactions
as previously described for DFG-in and DFG-out binding biphenyl
O
O
Cl
O
NH
b
c
O
N
NH
NH2
F
N
N
3
amides.
O
O
Three types of p38a inhibitors will now be discussed. The first
Me
Me
NH
Me
NH
O
B
Me
type includes aryl compounds para-substituted with a five-mem-
bered ring, for example the pyrazoles (3c), thiazoles (4a–c), and
1,2,4-oxadiazoles (5a–c).
Compounds from array A are generally less potent that those
from arrays B and C. The 1,3,4-oxadiazoles (6a–c) are good exam-
I
I
I
O
N
d
e
f
NH2
O
O
O
NH
C
O
OH
ples of this type. A p38a X-ray structure was solved of the complex
with 6a.7 It binds in the DFG-in mode in a similar way to 1.3a The
two structures are shown superimposed in Figure 2. As intended,
the tolyl cyclopropylamide back-pocket group of 6a mimics that
of 1. The amide group makes the same pair of hydrogen-bonds to
Glu71 and Asp168 in the back-pocket. The oxadiazole ring occupies
the purine subpocket and hydrogen-bonds to the hinge. One point
of additional interest is that the hinge region of Gly110 flips, as
previously reported for other structures, such that the backbone
NH atoms of Met109 and Gly110 can both donate H-bonds to the
oxadiazole nitrogens.8
Cl
N
N
N
N
Cl
N
Me
NH
OH Me
B
Me
I
I
HO
g,h
i
O
O
N
NH
HO
N
O
NH2
D
N
It is interesting to compare 6a to 20, one of the first compounds
from the biphenyl amide series (Fig. 3).8 During the evolution of
the series through several iterations, the oxadiazole in the
back-pocket of 20 was first replaced by a cyclopropyl amide. Later,
as described here, the purine-pocket amide was replaced by an
oxadiazole. The location and substitution pattern of the central
biphenyl group is maintained in both complexes, so this is not
simply an example of a flipped binding mode.
N
N
N
Me
Me
Me
Scheme 1. Reagents and conditions: (a) cyclopropylamine, HATU, DIPEA, 3.5 h, rt;
(b) DMF, Et3N, 80 °C, 12 h; (c) bis(pinacolato)diborane, Et3N, PdCl2(dppf), dioxane,
80 °C, 12h; (d) HATU, DMF, DIPEA, 80 °C, 12 h; (e) pyrrolidine, 80 °C, 12 h; (f)
bis(pinacolato)diborane, KOAc, PdCl2(dppf), 80 °C, 12 h; (g) SOCl2, CHCl3, 80 °C,
16 h; (h) Et3N, DMF, 50 °C, 16 h; (i) MeMgBr, sBuLi, (EtO)3B, THF, À78 °C–rt, 16 h. E
was prepared as previously described.3a A–D were coupled to aryl halides under
Suzuki conditions using Pd(PPh3)4, Na2CO3 (aq), DME.

I. Baldwin et al. / Bioorg. Med. Chem. Lett. 18 (2008) 5285–5289
5287
Table 2 (continued)
Table 2
p38
activities (Ki, nM) of compounds from arrays A–C (Scheme 1)6
a
Purine-site group
Array
Purine-site group
Array
A
B
C
D
A
B
C
D
N
H
N
17
18
87
110
N
N
O
(
(
3
4
5
6
14
H
N
H
O
N
110
4
19
12
>850
H2N
N
(
130
20
6.7
1.1
1.7
>850
>850
1400
H2N
(
S
O
N
N
O
(
(
(
19
1.4
0.8
98
10
17
N
Me
HN
N
N
Products that were impure or that gave inconsistent assay results are omitted,
although some were later reprepared (data not shown).
2.8
O
Me
Me
O
H
N
7
49
4.7
6.8
H2N
(
N
O
(
H N
8
9
2.9
N
S
O
(
18
6
19
11
2000
N
O
N
H
O
(
10
11
N
Figure 2. X-ray structure of 1 complexed with p38a (orange) superimposed on that
N
N
of 6a (green).
O
(
The second compound type contains fused 6,6-bicyclic sys-
tems. Some of these resemble cyclised biphenyl amides (for
example 7–11). These utilise the cyclic amide carbonyl oxygen
to accept the hinge Met109 H-bond. The 6,6-bicyclic rings in
which the cyclic amides are positioned at the 4-position to the
biaryl linker have greater potency than those with the amide
at the 3-position (compare 7a–8a). This can be rationalised by
their geometry being better able to H-bond to the hinge. Other
6,6-fused systems include those in which a ring nitrogen atom
makes the hinge H-bond to Met109, for example, the isoquino-
lines 14a–b. Once again, the importance of forming the hinge
H-bond with appropriate geometry can be seen by comparing
the potent isoquinoline 14a to the less potent quinoline 15a.
While still fitting within the tolerances of the pharmacophore,
the quinoline nitrogen of 15a is not well placed to form the
hinge H-bond. Similarly, compounds in which the hinge-binding
nitrogen preferentially adopts the wrong tautomer are weaker
(compare 14a–13a, or 14b–13b).
6.7
O
S
H
N
(
12
13
290
18
49
>2500
>2500
O
H
O
N
420
2.4
(
N
14
15
11
4.5
830
(
N
490
>2500
(
The third compound type includes fused 5,6-bicyclic systems
(16–19). Among the most potent are the benzimidazole 17a–b
and the benzisoxazoles 18a–c and 19a–c. In this class of com-
pounds, additional bulk in the outer lipophilic pocket region is
beneficial for activity (compare 19b to 18b or 19c to 18c). A crystal
structure of the DFG-out binding benzisoxazole 19b in complex
(
O
16
1100
380
160
N
Me
with p38
a
was solved (Fig. 4).7 The benzisoxazole nitrogen fulfils
the intended role of hydrogen-bonding to the hinge of Met109.
The back-pocket binding part of the structure resembles that of

5288
I. Baldwin et al. / Bioorg. Med. Chem. Lett. 18 (2008) 5285–5289
Gly110
Met109
Met109
N
O
O
N
O
N
Me
HN
Me
Me
NH
N
H
O
N
H
Me
Me
Me
20
6a
Glu71
O
19b
2
O
N
NH
O
NH
N
O
Asp168
Phe169
Figure 3. p38
N
Asp168
Me
N
N
O
O
a
crystal structures of 6a (green) and 20 (orange).
Figure 4. X-ray complex of 19b (orange/green ribbon) superimposed on 2 (cyan).
the biphenyl amide 2 previously reported (Fig. 4), even though the
amide of 19b is of reversed direction to that of 2.3b
Selectivity profiles were assessed by screening against a panel
of protein kinases. Compounds from the same arrays with the same
back-pocket binding groups showed similar profiles (Table 3).
ray D (also inhibiting EGFR and ErbB4). It is unclear why the quin-
olines 14a and 14d are both less selective than other compounds
with the same back-pocket groups.
Selected compounds were evaluated for their ability to inhibit
cytokine production in cells. Table 3 shows IC50 values for inhibi-
Compounds binding to p38
C) also showed activity in a cRaf/MEK/ERK cascade assay.9 This
was unsurprising, given the p38 activity of the DFG-out binding
Raf inhibitor Sorafenib.10 In contrast, compounds from array A
were selective for p38 (Fig. 5).
Most other kinases tested were unaffected by compounds
from arrays A–C. One exception was the isoquinoline 14a, which
showed sub-lM inhibition of VEGFR2 (KDR) (Table 3). No signif-
a in the DFG-out mode (arrays B and
tion of TNF
a production in peripheral blood mononuclear cells
and in whole blood.6 Several compounds, for example 6a, 6c, 10a
and 14a, showed encouraging sub-micromolar activity, compara-
ble to or better than that seen with 1.3a Compound 6a was submit-
ted for pharmacokinetic analysis. Its rat PK profile was comparable
to that of 1, with low plasma clearance, moderate volume of
distribution and good bioavailability, but with a longer half-life
(Table 4).3a,11
a
a
icant VEGFR2 activity was seen with other 6,6-fused bicyclic
templates or with the other two substructural types reported
here.
In summary, a pharmacophore-guided array strategy has been
exploited to combine DFG-in and DFG-out back-pocket binding
groups with novel kinase purine subpocket binders. The example
Compounds from array D showed a completely different inhibi-
tion profile (Tables 2 and 3). As expected, they did not inhibit p38a.
Instead, as reported for Gleevec, members of this array inhibited
cFMS and LCK.2a,4 Compound 14d showed the greatest inhibition
of these two targets, with IC50 of 4 nM and 70 nM, respectively.
Compound 14d was less selective than other compounds from ar-
of p38
a has been used to illustrate the approach. Inhibitors with
excellent enzyme potency and selectivity were produced without
additional iterations of optimisation. The strategy has been
validated by crystallographic confirmation of the intended binding
mode. Multiple diverse and novel series were identified as promis-
ing p38a backup series to the biphenyl amides. Some of these also
Table 3
Kinase selectivity profiles and inhibition of TNF
a production in PBMC cells and whole blood (IC50 l
M)6
Compound
Array
Purine-site
Raf/Mek/Erk
CDK2
cFMS
EGFR
ErbB4
JNK3
LCK
PLK1
SGK1
VEGFR2
PBMC
HWB
19b
6b
19c
6c
19a
6a
10a
14a
1
B
B
C
C
A
A
A
A
A
D
D
D
D
Benzisoxazole
1,3,4-Oxadiazole
Benzisoxazole
1,3,4-Oxadiazole
Benzisoxazole
1,3,4-Oxadiazole
1,2,3-Benzotriazinone
Isoquinoline
Biphenylamide
1,3,4-Oxadiazole
Isoquinoline
0.04
0.4
0.03
0.2
>50
20
>20
>20
>20
>20
>20
>20
>20
>20
>20
>20
>20
>20
>20
>20
0.7
7
>20
>20
>20
>20
0.6
>20
0.1
0.004
0.09
0.09
>25
>25
>25
>25
>25
>25
>25
>25
>25
>25
0.7
>25
>25
>25
>25
>25
>25
>25
>25
>25
>25
0.3
>16
>16
>16
>16
>16
>16
>16
>16
>16
>16
>16
>16
>16
8
8
6
>25
>25
>25
>25
>25
>25
>25
>25
>25
>25
>25
>25
>25
>33
>33
>33
>33
>33
>33
32
>33
>33
>33
>33
>33
>33
>20
6
>20
>20
>20
>20
>20
0.5
>20
>20
0.80
>20
>20
0.06
0.1
0.2
5
0.7
3
10
>16
>16
>16
3
>16
3
0.07
1
0.1
1
0.4
0.1
0.25
0.5
0.1
1
>50
6d
14d
18d
4d
Benzisoxazole
Thiazole
>25
>25
>25
>25
2

I. Baldwin et al. / Bioorg. Med. Chem. Lett. 18 (2008) 5285–5289
5289
Acknowledgments
The authors thank Tony Angell for generating cellular data and
Chun-wa Chung, Chris Edwards, Katherine Jones, Vipul Patel and
Ann Walker for critical reading of the manuscript.
References and notes
1. (a) Lai, J.; Langston, S.; Adams, R.; Beevers, R.; Boyce, R.; Burckhardt, S.; Cobb, J.;
Ferguson, Y.; Figueroa, E.; Grimster, N.; Henry, A.; Khan, N.; Jenkins, K.; Jones,
M.; Judkins, R.; Major, J.; Masood, A.; Nally, J.; Payne, H.; Payne, L.; Raphy, G.;
Raynham, T.; Reader, J.; Reader, V.; Reid, A.; Ruprah, P.; Shaw, M.; Sore, H.;
Stirling, M.; Talbot, A.; Taylor, J.; Thompson, S.; Wada, H.; Walker, D. Med. Res.
Rev. 2005, 25, 310; (b) Birault, V.; Harris, J.; Le, J.; Lipkin, M.; Nerella, R.;
Stevens, A. Curr. Med. Chem. 2006, 13, 1735; (c) Prien, O. ChemBioChem 2005, 6,
500.
2. (a) Schindler, T.; Bornmann, W.; Pellicena, P.; Miller, W.; Clarkson, B.; Kuriyan,
J. Science 2000, 289, 1938; (b) Pargellis, C.; Tong, L.; Churchill, L.; Cirillo, P.;
Gilmore, T.; Graham, A.; Grob, P.; Hickey, E.; Moss, N.; Pav, S.; Regan, J. Nat.
Struct. Biol. 2002, 9, 268; (c) Regan, J.; Breitfelder, S.; Cirillo, P.; Gilmore, T.;
Graham, A.; Hickey, E.; Klaus, B.; Madwed, J.; Moriak, M.; Moss, N.; Pargellis, C.;
Pav, S.; Proto, A.; Swinamer, A.; Tong, L.; Torcellini, C. J. Med. Chem. 2002, 45,
2994.
3. (a) Angell, R.; Aston, N.; Bamborough, P.; Buckton, J.; Cockerill, S.; deBoeck, S.;
Edwards, C.; Holmes, D.; Jones, K.; Laine, D.; Patel, S.; Smee, P.; Smith, K.;
Somers, D.; Walker, A. Bioorg. Med. Chem. Lett. 2008, 15, 4428; (b) Angell, R.;
Angell, T.; Bamborough, P.; Bamford, M.; Chung, C.; Cockerill, S.; Flack, S.; Jones,
K.; Laine, D.; Longstaff, T.; Ludbrook, S.; Pearson, R.; Smith, K.; Smee, P.; Somers,
D.; Walker, A. Bioorg. Med. Chem. Lett. 2008, 15, 4433.
4. (a) Jacobs, M. D.; Caron, P. R.; Hare, B. J. Proteins 2008, 70, 1451; (b) b) Dewar, A.
L.; Cambareri, A. C.; Zannettino, C. W.; Miller, B. L.; Doherty, K. V.; Hughes, T. P.;
Lyons, A. B. Blood 2005, 105, 3127.
5.
A sub-set of a small-molecule X-ray crystal structure database (Allen, F.;
Acta Cryst., 2002, B58, 380) was converted into Catalyst format (Accelrys
Inc.) and searched using a 3D pharmacophore query built using the X-ray
complex of
retain only the subgraphs containing the query features. These were
converted into 2D SMARTS representation (Daylight Chemical Systems
Inc.) with generic atoms and bonds, with the exception of the hinge
H-bond acceptor atom and an attachment Br or atom replacing the
2 in p38a. Molecules hitting the query were processed to
a
I
tolyl ring of the query. The SMARTS queries were then used to search
the GSK database for available reagents, which were examined and
filtered by eye.
6. p38
a
Ki determinations, detecting displacement of
a
fluorescent ATP-
release from
competitive inhibitor, and IC50 values for inhibition of TNF
a
LPS-stimulated peripheral blood mononuclear cells (PBMCs) were carried
out as described in Angell, R. et al. Bioorg. Med. Chem. Lett. 2008, 18, 324.
Inhibition of TNF
described3a
7. Unphosphorylated p38
a
release from whole blood cells was carried out as
Figure. 5. Top: Plot of p38
selectivity of DFG-in compounds (array A, red diamonds) compared to DFG-out
(array B, black circles, array C, green squares). Bottom: p38
showing the greater cFMS activity of array D (blue stars).
a against cRaf/MEK/ERK activity showing the greater
.
a
was expressed, purified and crystallized as previously
a
against cFMS activity,
described (Angell, R. et al. Bioorg. Med. Chem. Lett. 2008, 18, 318). Apo-crystals
were soaked with 6a at 0.25 mM for 3 days or 19b at 0.1 mM for 1 day. Both
were cryoprotected for data collection at 100 K using an Oxford Cryostream. X-
ray diffraction data were collected and processed and the structures solved as
previously described.3a The final R-factor achieved for each complex was 17.2%
for compound 6a and 16.9% for compound 19b. The coordinates have been
deposited in the PDB as entries 3E92 and 3E93. Figures were produced using
Pymol (DeLano, W.L., DeLano Scientific, Palo Alto CA, USA. http://
www.pymol.org).
Table 4
Pharmacokinetic parameters of 6a measured in rat11
IV plasma clearance (mL/min/kg)
IV steady state volume of distribution (L/kg)
IV plasma terminal t1/2 (h)
PO AUC(0–12 h) (ng h/mL)
3
8. Angell, R.; Bamborough, P.; Cleasby, A.; Cockerill, S.; Jones, K.; Mooney, C.; Neu,
M.; Somers, D.; Walker, A. Bioorg. Med. Chem. Lett. 2008, 18, 318. and references
therein.
9. McDonald, O. B.; Chen, W.; Ellis, B.; Hoffman, C.; Overton, L.; Rink, M.; Smith,
A.; Marshalland, C.; Wood, E. Anal. Biochem. 1999, 268, 318.
1.2
4.6
4180
91%
PO bioavailability
10. Wilhelm, S.; Carter, C.; Lynch, M.; Lowinger, T.; Dumas, J.; Smith, R.; Schwartz,
B.; Simantov, R.; Kelley, S. Nat. Rev. Drug Discov. 2006, 5, 835.
11. Pharmacokinetic parameters derived from a composite blood sampling profile
in male Lewis rats were determined following intravenous (iv) and oral (po)
administration at 1 mg/kg. Compound was administered as a solution in 10%
DMSO/70% PEG200/20% water. Blood was collected over a 12 h time period.
Plasma preparation, sample analysis and data generation were carried out as
described in Ref. 3a.
showed favourable cellular activity and oral pharmacokinetic
properties. Further development of these series will be described
in future publications.