Inhibitors of the tyrosine kinase EphB4. Part 2: Structure-based discovery and optimisation of 3,5-bis substituted anilinopyrimidines

Article abstract of DOI:10.1016/j.bmcl.2008.09.087

Crystallographic studies of a range of 3-substituted anilinopyrimidine inhibitors of EphB4 have highlighted two alternative C-2 aniline conformations and this discovery has been exploited in the design of a highly potent series of 3,5-disubstituted anilin

Full text of DOI:10.1016/j.bmcl.2008.09.087

Bioorganic & Medicinal Chemistry Letters 18 (2008) 5717–5721
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry Letters
journal homepage: www.elsevier.com/locate/bmcl
Inhibitors of the tyrosine kinase EphB4. Part 2: Structure-based discovery
and optimisation of 3,5-bis substituted anilinopyrimidines
*
Catherine Bardelle, Tanya Coleman, Darren Cross, Sara Davenport, Jason G. Kettle , Eun Jung Ko,
Andrew G. Leach, Andrew Mortlock, Jon Read, Nicola J. Roberts, Peter Robins, Emma J. Williams
AstraZeneca, Mereside, Alderley Park, Macclesfield, Cheshire SK10 4TG, UK
a r t i c l e i n f o
a b s t r a c t
Article history:
Crystallographic studies of a range of 3-substituted anilinopyrimidine inhibitors of EphB4 have high-
lighted two alternative C-2 aniline conformations and this discovery has been exploited in the design
of a highly potent series of 3,5-disubstituted anilinopyrimidines. The observed range of cellular activities
has been rationalised on the basis of physicochemical and structural characteristics.
Ó 2008 Elsevier Ltd. All rights reserved.
Received 4 September 2008
Revised 23 September 2008
Accepted 24 September 2008
Available online 27 September 2008
Keywords:
EphB4
Tyrosine kinase
Structure-based design
Pyrimidine
The field of kinase inhibition has seen an explosion in activity
over the last ten years, and much of this has been directed towards
inhibition of the receptor tyrosine kinases known to play a pivotal
role in the aberrant signalling that is characteristic of the uncon-
trolled proliferation of tumours. The erythropoietin-producing
hepatoma amplified sequence (Eph) family is the largest known
group of such kinases,1 and together with their ephrin ligands
are increasingly implicated in tumourigenesis in a wide variety
of human cancers, either on tumour cells directly, or indirectly
via modulation of vascularisation.2–4 In the preceding paper5
which referred to the first EphB4 crystal structures,6 we outlined
the discovery of a novel series of potent and selective 2,4-bis anili-
nopyrimidines as inhibitors of EphB4 derived from structural over-
lays of two alternative inhibitor series bound in the active site.7
Initial chemical optimisation led to the observation of a strong
preference for a 3-substituted anilino C-2 hinge-binding group,
with strong electron-withdrawing groups such as sulfonamide 1
and sulfone 2 most preferred. Substitution at the ortho position
in this ring was generally poorly tolerated, and all substituents at
para were less active than the same groups at the meta position.
In order to try and understand these preferences further, the
crystal structure of sulfonamide ligand 1 in complex with EphB4
was determined.8 The inhibitor adopts the expected conformation,
with the benzdioxole moiety buried in the selectivity pocket, and a
donor–acceptor hydrogen bond to the hinge region at Met696,
mimicking the binding of ATP (Fig. 1a and b). Intriguingly, how-
ever, the C-2 aniline of 1 is able to adopt a dual conformation with-
in the protein structure, with the sulfonamide alternatively
hydrogen bonding towards the hinge and Glu697, or back towards
the glycine-rich loop and Ile621 (Fig. 1a). The electron density in
Figure 1a suggests the two binding modes are approximately
equally populated (each has been modelled with an occupancy of
0.5, where 1.0 is full occupancy). A view along the hinge (Fig. 1b)
demonstrates the proximity of the hydrogen bonds to the protein,
and the near overlap of the C-2 phenyl rings in the differing bind-
ing modes. The initial SAR studies highlighted a strong preference
for groups at this position that were capable either of accepting or
donating a hydrogen bond (and in the case of 1, both simulta-
neously) and the determined structure of 1 may go some way to
explain this observation. A second striking feature of the structure
of 1 is that there appears to be no movement in the protein to
accommodate these alternative orientations, opening up the possi-
bility that bis-meta, that is, 3,5-disubstituted anilines would be tol-
erated at C-2 of the pyrimidine, and could potentially lead to
enhancements in potency.
O
O
HN
Cl
N
O
S
N
N
X
H
O
1 X = NH2
2 X = CH3
* Corresponding author. Tel.: +44 1625517920; fax: +44 1625586707.
E-mail address: jason.kettle@astrazeneca.com (J.G. Kettle).
0960-894X/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2008.09.087

5718
C. Bardelle et al. / Bioorg. Med. Chem. Lett. 18 (2008) 5717–5721
Figure 1. Binding of 3-substituted anilinopyrimidines to EphB4. For clarity, final 2Fo ꢀ Fc electron density contoured at 1
r is drawn around the ligand only. (a) Binding of
sulfonamide 1 to the catalytic domain of EphB4 showing the dual conformation adopted by the C-2 anilino group. (b) An alternative view of bound 1 along the hinge-axis
showing the sulfonamide hydrogen-bonding both towards the hinge and Glu697 and towards the glycine-rich loop and Ile621. (c) Structure of sulfone 2 shows an occupancy
ratio of approximately 0.7–0.3 (Glu697/Ile621). (d) Structure of aminosulfonamide 3 shows occupancy ratio of approximately 0.3–0.7 (Glu697/Ile621). (e) Structure of
morpholine 4 shows full occupancy of the binding mode oriented towards Glu697. (f) Structure of primary carboxamide 5 shows almost full occupancy of the binding mode
oriented towards Ile621.
Prior to investigation of which 3,5-disubstituted anilines might
be examined, two questions remained unanswered. Firstly, which
of the possible different potent meta-substituents identified from
the initial SAR studies would be optimal in combination, and sec-
ondly which other of these substituents, if any, demonstrated pro-
pensity for a dual binding mode? At the outset it was reasonably
expected that combination of a meta-substituent with preference
for one binding mode would not be enhanced by a second meta-
substituent known to prefer the same binding orientation. In an at-
tempt to understand which substituents might be thus combined
we obtained crystal structures of compounds 2–5 (Table 1) bound
to EphB4,9 and it is clear that the nature of the 3-substituent is key
to determining the orientation of the C-2 aniline within the active
site. Similar to 1, the closely related sulfone 2 showed a dual bind-
ing mode, with a small preference for binding towards Glu697
(occupancy 0.7)10 over Ile621 (occupancy 0.3; Fig. 1c). Conversely
the aminosulfonamide compound 3 demonstrated the opposite
preference in favour of Ile621 (occupancy 0.8; Fig. 1d). The mor-
pholino-substituted compound 4 showed exclusive binding ori-
ented towards Glu697 (Fig. 1e), and conversely the electron
density for the primary carboxamide 5 showed near-exclusive
binding towards Ile621 (Fig. 1f; occupancy modelled as >0.8). On
the basis of these observations, it was tempting to speculate which
might be the optimal combinations, for example complementary

C. Bardelle et al. / Bioorg. Med. Chem. Lett. 18 (2008) 5717–5721
5719
Table 1
EphB4 inhibition data for 3-substituted and 3,5-disubstituted anilinopyrimidines
O
O
HN
3
Cl
N
5
N
H
N
Binding to Glu697b (%)
Binding to Ile621b (%)
a
Compound
3-
5-
EphB4 IC50
(lM)
Cell IC50
(l
M)
1
2
3
4
5
6
7
8
–SO2NH2
–SO2Me
–NHSO2Me
1-Morpholinyl
–CONH2
–SO2Me
–SO2Me
–SO2Me
–SO2Me
–NHSO2Me
–NHSO2Me
–NHSO2Me
–NHSO2Me
1-Morpholinyl
1-Morpholinyl
–CONH2
H
H
H
H
0.040 0.020
0.090 0.050
0.172 0.073
0.190 0.070
0.741
50
70
30
100
<20
50
30
70
0
0.794 0.226
0.580 0.120
0.220 0.014
0.114 0.095
0.012 0.002
0.023 0.001
0.002 0.002
0.001 0.001
0.002 0.002
0.015 0.004
0.370 0.014
0.002 0.004
0.013 0.001
0.077 0.007
NT
H
0.900 0.100
0.209
2.054 0.268
22.200
0.032 0.02
14.117 6.095
0.170
7.175
>30
0.017 0.011
0.887 0.271
>30
>80
–SO2Me
–NHSO2Me
–CONH2
1-Morpholinyl
–SO2NH2
1-Morpholinyl
–NHSO2Me
–CONH2
1-Morpholinyl
–CONH2
–CONH2
9
10
11
12
13
14
15
16
a
For determinations where n P 2, standard deviation is given.
Binding percentage is estimated on the basis of the cage electron densities of the structures shown in Figure 2 and is approximate.
b
substituents such as sulfone (as in 2) with aminosulfonamide (as in
3) or morpholinyl (as in 4) with carboxamide (as in 5) might be ex-
pected to give enhancements in potency through reinforcement of
the individual substituents preferred binding orientations.11
In the event, we synthesized as many of the combinations of
these five substituents as could be accessed in a reasonable time-
frame, and these combinations are highlighted in Table 1. Synthe-
sis of the requisite 3,5-disubstituted anilines is detailed in Schemes
1 and 2 and the final inhibitors were assembled as previously de-
scribed.5,12 Synthesis of the anilines was initiated for the most part
from commercially available 3,5-disubstituted nitrobenzenes and
involved a core group of reactions whose order was varied to in-
stall each of the substituents as required. Thus in each case, mor-
pholine was introduced by reaction with
a fluorobenzene
activated by a meta-nitro group13 (see anilines 20, 23, 32 and 48
used to give target compounds 9, 14, 15 and 11, respectively),
and methanesulfone was installed by copper-catalysed coupling
with methanesulfinate anion on the requisite iodobenzene14 (see
anilines 20 and 27 used in inhibitors 9 and 7, respectively). Amino-
sulfonamides such as those present in inhibitors 7, and 10–13 were
introduced by simple methanesulfonylation of the pre-cursor ani-
lines (see anilines 27, 36, 39, 43 and 48) and primary carboxamides
(as in inhibitors 8, 13, 15 and 16) accessed via HATU coupling of
NO2
NH2
X
Y
X
Y
17 X=F, Y=I
18 X=F, Y=SO2Me
19 X=1-morpholinyl, Y=SO2Me
20 X=1-morpholinyl, Y=SO2Me
(a)
(b)
(c)
21 X, Y=F
22 X, Y=1-morpholinyl
23 X, Y=1-morpholinyl
(c)
(c)
(b)
(a)
24 X=NO2, Y=I
25 X=NO2, Y=SO2Me
26 X=NH2, Y=SO2Me
27 X=NHSO2Me, Y=SO2Me
(d)
(d)
O
28 X=F, Y=CO2Et
32 X=1-morpholinyl, Y=CONH2
(b)
(e)
(f)
29 X=1-morpholinyl, Y=CO2Et
30 X=1-morpholinyl, Y=CO2H
31 X=1-morpholinyl, Y=CONH2
(c)
N
O S O
33 X=NO2, Y=NH2
34 X=NO2, Y=SO2NH2
35 X=NH2, Y=SO2NH2
36 X=NHSO2Me, Y=SO2NH2
(g)
(d)
(c)
(c)
X
Y
X
Y
37 X,Y=NH2
38 X,Y=NHSO2Me
39 X,Y=NHSO2Me
29 X=NO2, Y=CO2Et
44 X=NH2, Y=CO2Et
45 X=NHSO2Me, Y=CO2Et
46 X=NHSO2Me, Y=CO2H
47 X=NHSO2Me, Y=NHCO2-t-Bu
48 X=NHSO2Me, Y=NH2
49 X=H, Y=CO2H
(a)
(b)
(c)
(d)
(f)
(g)
(a)
50 X=NO2, Y=CO2H
51 X=NO2, Y=CONH2
52 X=NH2, Y=CONH2
40 X=NH2, Y=CO2H
41 X=NH2, Y=CONH2
42 X=NHSO2Me, Y=CONH2
43 X=NHSO2Me, Y=CONH2
(f)
(c)
(d)
(e)
Scheme 1. Synthesis of anilines used in inhibitors 7, 9–10, 12–15. Reagents and
conditions: (a) MeSO2Na, CuI, DMF, 110 °C; (b) Morpholine, DMSO, 100 °C; (c) H2,
10% Pd/C, EtOH; (d) MeSO2Cl, pyridine, DCM, 25 °C; (e) aq NaOH, MeOH, THF, 25 °C;
(f) NH4Cl, HATU, DIPEA, DMF, 25 °C; (g) i—NaNO2, concd HCl, water, CuCl, SO2; then
ii—NH3, MeOH.
Scheme 2. Synthesis of anilines used in inhibitors
8 and 11. Reagents and
conditions: (a) H2, 10% Pd/C, EtOH; (b) MeSO2Cl, pyridine, DCM, 25 °C; (c) aq
NaOH, MeOH, THF, 25 °C; (d) DPPA, DIPEA, t-BuOH; (e) concd HCl, MeOH, 70 °C; (f)
concd HNO3, concd H2SO4, 85 °C; (g) NH4Cl, HATU, DIPEA, DMF, 25 °C.

5720
C. Bardelle et al. / Bioorg. Med. Chem. Lett. 18 (2008) 5717–5721
the appropriate benzoic acid derivatives (see anilines 32, 43 and
52). For aniline 48 (Scheme 2, used in inhibitor 11) the aniline
moiety was introduced via Curtius rearrangement of acid 46. Ani-
line 52 (present in inhibitor 8) was introduced via reduction of a
nitro group installed via nitration of commercially available
sulfone acid 49.
Introduction of a second sulfone group to 2 to give bis-sulfone 6
led to comparable enzyme activity15 but was clearly not detrimen-
tal. When the sulfone was combined with a second aminosulfona-
1.5
1
0.5
0
-0.5
-1
mide group as in 7, an increase in enzyme activity to 0.012 lM was
observed, a figure significantly more active than the contribution
from each substituent on its own (in 2 and 3), and appears to lend
support to the strategy for combining substituents with comple-
mentary binding modes. Similarly compound 8, comprising a
sulfone favouring binding towards Glu697, and a primary carbox-
amide favouring binding to Ile621 resulted in an inhibitor signifi-
cantly more potent than either substituent alone (compare to 2
and 5). Other examples of the synergistic nature of specific combi-
nations can be found in compounds 11 (combining a morpholinyl
-1.5
-2
1
2
3
4
5
6
7
Number H Donors
1.5
1
and aminosulfonamide to give a 0.002
bining morpholinyl with carboxamide to give a 0.013
l
M inhibitor) and 15 (com-
M inhibi-
l
0.5
0
tor) both of whose C-2 anilines might be expected to be potent
in combination, from the structural studies outlined above. Simi-
larly the one group for which there was no overall binding prefer-
ence, the sulfonamide (as in 1) also shows potent activity in the
single combination examined (here with aminosulfonamide as in
-0.5
-1
10) at 0.001 lM.
Surprisingly, it was observed that many other combinations
also led to potent inhibition, including combinations which singly
at least, were shown to prefer the same binding orientation. The
most striking example of this is bis-morpholine 14 that in 4 favours
a single conformation, but in combination leads to a very potent
-1.5
-2
3
3.5
4
4.5
5
inhibitor, at 0.002 lM nearly 300-fold more active than with a sin-
CLogP
gle substituent. It is noteworthy that this substantial potency in-
crease occurs in tandem with a 0.5 log unit decrease in ClogP
relative to 4, following introduction of a second morpholine. Mor-
pholine in combination with a sulfone (as in 9) was highly active
despite both substituents independently showing a favoured bind-
ing towards Glu697 and similarly bis-aminosulfonamide 12 and
bis-carboxamide 16 were also very potent. The inhibitor with car-
boxamide and aminosulfonamide groups, 13, showed enzyme
activity intermediate between the contributions from each substi-
tuent in isolation (3 and 5). The majority of 3,5-disubstituted ani-
lines examined are not only tolerated, but positively enhance
enzyme inhibition potency. Despite our initial assumption that
only substituents with complementary binding modes might be
expected to enhance potency, it is clear from the data that two
meta-substituents are favoured over one, and the observation that
the global protein structure does not change to accommodate the
different aniline orientations may in part explain this.
In comparison with the enzyme data, the activity in a cellular
assay of EphB4 inhibition16 for these compounds highlights a much
broader spread of data, ranging from potent inhibition for 9 and 14
to activity higher than the top concentration tested for 13 and 16.
Analysis of the correlation of the difference in enzyme to cell activ-
ity and parameters such as number of hydrogen-bond donors and
lipophilicity (ClogP) is informative (Fig. 2). The cell activity is seen
to drop off rapidly as a function of increasing number of hydrogen-
bond donors or decreasing ClogP.17 These data may be indicative of
a problem of permeability (since adequate lipophilicity is required
for permeation and higher numbers of donors may impair perme-
ability across the cell membrane) or efflux (since increased hydro-
gen bonding capability may increase recognition by transporters)18
or both. Compounds 7 (large cell drop-off) and 14 (small cell drop-
off) were examined in the cell assay in the presence of an efflux
inhibitor.19 Consistent with active efflux, the cell activity of 7
Figure 2. Plots of Log (enzyme to cell) difference versus number of hydrogen-bond
donors and ClogP for compounds 1–16. Closed circles represent 3,5-disubstituted C-
2 anilines, stars represent 3-substituted anilines. Points on top line are out of range
values.
was increased to 0.175
of the already potent 14 remained unchanged at
0.010 0.001
M.20
lM in this system, whereas the cell activity
l
In summary, crystallographic studies of a range of 3-substituted
anilinopyrimidine inhibitors of EphB4 have highlighted two alter-
native aniline conformations are available in the active site of the
kinase. In an attempt to exploit these two interactions simulta-
neously, a set of 3,5-disubstituted anilinopyrimidines has been
prepared and these show potent enzyme inhibition. The observed
range of cellular activities has been rationalised on the basis of
physicochemical and structural characteristics, and has been
linked, for some inhibitors, to a potential for efflux. Further studies
on optimisation of both the C-2 and C-4 anilines are underway and
will be reported in due course.
Acknowledgments
We thank Jonathan Wingfield for technical assistance, Eileen
McCall, Isabelle Green and Anna Valentine for construct design, in-
sect cell expression and protein preparation, and Claire Brassington
for crystallisation.
Supplementary data
Supplementary data associated with this article can be found, in
the online version, at doi:10.1016/j.bmcl.2008.09.087.

C. Bardelle et al. / Bioorg. Med. Chem. Lett. 18 (2008) 5717–5721
5721
9. Complex crystal structures with compounds 2, 3, 4 and 5 were determined at
0
0
References and notes
1.65 ÅA (except compound 4 at 2.1 ÅA) and refined to R = 18.7%, 16.8%, 18.4% and
17.1%. PDB deposition codes are 2VWY, 2VWZ, 2VX0, 2VX1, respectively.
Statistics and protocols for both parts 1 and 2 of this publication series are
included in the Supplementary Material.
1. Kullander, K.; Klein, R. Nat. Rev. Mol. Cell Biol. 2002, 3, 475–486.
2. Xia, G.; Kumar, S. R.; Masood, R.; Zhu, S.; Reddy, R.; Krasnoperov, V.; Quinn, D.
I.; Henshall, S. M.; Sutherland, R. L.; Pinski, J. K.; Daneshmand, S.; Buscarini, M.;
Stein, J. P.; Zhong, C.; Broek, D.; Roy-Burman, P.; Gill, P. S. Cancer Res. 2005, 65,
4623–4632.
3. Brantley-Sieders, D.; Schmidt, S.; Parker, M.; Chen, J. Curr. Pharm. Des. 2004, 10,
3431–3442.
4. Cheng, N.; Brantley, D. M.; Chen, J. Cytokine Growth Factor Rev. 2002, 13,
75–85.
10. Occupancies were roughly estimated by balancing the atomic temperature
factors in the two alternate ligand models and were not refined.
11. This simplistic argument ignores any impact on potency from changes in
lipophilicity and/or de-solvation penalties when adding an additional
substituent.
12. For detailed synthetic procedures see: Kettle, J. G.; Read, J.; Leach, A.; Barlaam,
B. C.; Ducray, R.; Lambert-Van Der Brempt, C. M. P.; PCT Int. Appl.
WO2007085833.
5. Bardelle, C.; Cross, D.; Davenport, S.; Kettle, J. G.; Ko, E. J.; Leach, A. G.; Mortlock,
A.; Read, J.; Roberts, N. J.; Robins, P.; Williams, E. J. Bioorg. Med. Chem. Lett. 2008,
18, 2776–2780.
13. Belfield, A. J.; Brown, G. R.; Foubister, A. J.; Ratcliffe, P. D. Tetrahedron 1999, 55,
13285–13300.
14. Baskin, J. M.; Wang, Z. Org. Lett. 2002, 4, 4423–4425.
6. Human EphB4 (598–892; Y774E) was crystallised by sitting drop vapour
diffusion using 2 ll 12 mg/ml protein with equivalent volumes of reservoir (25%
PEG, 5000 Mme, 0.1 M Tris, pH 7.5, 0.15 M MgCl2, 15% glycerol) and MPD in the
drop. Crystals were robust and of space group P21 with approximate cell
15. The assay detects inhibitors of recombinant EphB4-mediated phosphorylation of a
polypeptide substrate in presence of magnesium-ATP using Alphascreen (Packard
Bioscience) luminescence detection technology. For full details see Ref. 12.
16. CHO-K1 cells were engineered to stably express an EphB4-Myc-His construct.
The endpoint assay used a sandwich ELISA to detect EphB4 phosphorylation
status. Myc-tagged EphB4 from treated cell lysate were captured via an anti-c-
Myc antibody and the phosphorylation status of captured EphB4 was then
measured using a generic phosphotyrosine antibody. For further details see
Ref. 8.
0
dimensions a = 46.0, b = 53.5, c = 61.4 AÅ and b = 110.9°. The first complex
structure was solved by molecular replacement using a model of EphB2 (PDB
code 1JPA); subsequent structure solution used the newly determined EphB4
complex structures as models. Detailed protein preparation and other
experimental protocols are included in the Supplementary Material to this
paper.
7. Crystal structures0and structure factors for EphB4 complexes with compounds
1 (resolution 2.0 ÅA, R-facto0r 0.175; compound numbering from the 0preceding
paper),5 2 (resolution 1.7 ÅA, R-factor 0.238) and 7 (resolution 1.9 ÅA, R-factor
0.184) have been deposited in the PDB with accession codes 2VWU, 2VWV and
2VWW, respectively.
17. In this limited data set the correlation between number of hydrogen-bond
donors and ClogP is weak.
18. Seelig, A. Eur. J. Biochem. 1998, 251, 252–261.
19. N-[4-[3-(6,7-Dimethoxy-3,4-dihydro-1H-isoquinolin-2-yl)propoxy]phenyl]-9-
oxo-10H-acridine-4-carboxamide was used, an analogue of the known BCRP
and PGP inhibitor Elacridar, see: Dodic, N.; Dumaitre, B.; Daugan, A.; Pianetti, P.
J. Med. Chem. 1995, 38, 2418–2426.
20. This does not rule out the possibility of reduced permeability in conjunction
with efflux for the less cell potent inhibitors 10 and 16.
8. EphB4 crystals were soaked in 0.5 ll compound 1 (100 mg/ml) plus 4.5 ll
0
reservoir solution. The 1.65 ÅA structure was refined to R = 16.6% and deposited
with PDB accession code 2VWX. Data collection protocol and statistics are
available in the Supplementary Material.