A novel series of potent and selective IKK2 inhibitors

Article abstract of DOI:10.1016/j.bmcl.2003.10.047

A novel series of aminopyrimidine IKK2 inhibitors have been developed which show excellent in vitro inhibition of this enzyme and good selectivity over the IKK1 isoform. The relative potency and selectivity of these compounds has been rationalized using QSAR and structure-based modelling.

Full text of DOI:10.1016/j.bmcl.2003.10.047

Bioorganic & Medicinal Chemistry Letters 14 (2004) 409–412
A novel series of potent and selective IKK2 inhibitors
Alistair H. Bingham, Richard J. Davenport,* Lewis Gowers, Roland L. Knight,
Christopher Lowe, David A. Owen, David M. Parry and Will R. Pitt
Celltech R&D Ltd, Granta Park, Great Abington, Cambridge CB16GS, UK
Received 8 August 2003; revised 9 October 2003; accepted 28 October 2003
Abstract—A novel series of aminopyrimidine IKK2 inhibitors have been developed which show excellent in vitro inhibition of this
enzyme and good selectivity over the IKK1 isoform. The relative potency and selectivity of these compounds has been rationalized
using QSAR and structure-based modelling.
# 2003 Elsevier Ltd. All rights reserved.
The NF-kB pathway is important in regulating the
expression of cellular genes that are involved in the
control of the immune and inflammatory response.1 The
activation of NF-kB induces the expression of more than
150 genes2 such as cytokines (TNF, IL-1, IL-6), chemo-
kines (IL-8, MCP-1), cell adhesion molecules (ICAM-1,
VCAM-1) and proteases. NF-kB induces the production
of proteins,3 for example TNF, that are themselves able
to stimulate NF-kB, hence leading to an amplification on
any physiological effect of the NF-kB pathway.
a study into the structure–activity relationship around
ring A of our lead structure 1 (Fig. 1).
The lead compound 1 was found to have an IKK2 IC50
of 146 nM, with no selectivity for IKK2 over IKK1
(IKK1 IC50 154 nM). No IKK crystal structures were
available so a homology model was created.14 Crystal
structures of analogues of 1 in complex with two other
kinases were available in-house (unpublished data).
Since these analogues showed a consistent binding mode
in all these crystal structures, we made the assumption
that the binding of 1 in IKK2 would be equivalent, and
manually docked the compound as shown in Figure 2.
This predicted binding mode is also shown in Figure 2.
In this configuration, the aminopyrimidine-N forms a
hydrogen bond with the backbone NH of Cys99 (IKK2)
and the aminopyrimidine NH forms a hydrogen bond
with the backbone carbonyl of the same residue.
Activation of NF-kB is mediated by the increase in the
activation of two kinases, IKK1 and IKK2.4À7 IKK2
(À/À) knockout mice data8À11 have shown that IKK2 is
more critical than IKK1 in activating the NF-kB path-
way for the inflammatory response, whilst data from
IKK1 (À/À) knockout mice have indicated a role for
IKK1 in skin and skeletal development. Hence, small
molecule inhibition of IKK2, with selectivity over IKK1
acting via the NF–kB pathway, could lead to novel
treatments for autoimmune and inflammatory diseases,
such as rheumatoid arthritis.12 Celgene has reported
recently13 that SPC839 (Fig. 1), a selective IKK2 inhi-
bitor, is undergoing pre-clinical development for the
treatment of rheumatoid arthritis.
In our laboratories, we have identified a novel series of
aminopyrimidines that show inhibitory action against
IKK2. In this communication, we disclose the results of
Keywords: Potent; Selective; IKK2; Inhibitors.
* Corresponding author. Tel.: +44-1223-896491; fax: +44-1223-
896400; e-mail: richard.davenport@celltechgroup.com
Figure 1. The Celgene IKK2 inhibitor SPC839 and our initial lead
compound 1.
0960-894X/$ - see front matter # 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2003.10.047

410
A. H. Bingham et al. / Bioorg. Med. Chem. Lett. 14 (2004) 409–412
by absolute IKK2 activity). This compound is 3 fold
more potent against IKK2 and 24 fold less potent
against IKK1 than 1. In our model structure of 1 bound
to IKK2, the 4 position of the A ring is solvent exposed.
This fact led to the hypothesis that the improvement in
potency against IKK2 seen with 8q compared to 1 was a
consequence of electronic effects on the conjugated ani-
linopyrimidine core and not to changes in direct inter-
actions of the 4-substituent with the protein. To test this
hypothesis, analogues with a range of electronic prop-
erties, shapes and hydrophobicity were selected. A Craig
plot15 was used to help select the compounds that best
matched our criteria with all four Craig plot quadrants
being sampled. These compounds were then synthesized
in the manner described earlier and tested in vitro to
ascertain whether these properties could be a factor
affecting enzyme inhibition.
Figure 2. Predicted binding mode of compound 1 in IKK2 homology
model.
The analogues were synthesized from commercially
available starting materials. Sulfonyl chloride 2 was
reacted with Boc-piperazine (Scheme 1), to yield the
sulfonamide 3 (82–100%). The acetyl group was then
elaborated to the enaminone 4 by refluxing in dime-
thylformamide-dimethylacetamide (87–100%). This was
then reacted under basic conditions with a variety of
aryl guanidines 6 [which were derived from reaction of
commercially available anilines 5 with cyanamide and
nitric acid (85–98%)] to yield the Boc-protected amino-
pyrimidines 7 (60–98%). Finally, the Boc group was
removed using 10% trifluoroacteic acid in dichloro-
methane to yield the compounds 8 (82–100%).
The results of this study are represented in Table 1. In
general, it was found that the compounds with the
greatest IKK2 activity had substituents found in the top
left-hand quadrant of the Craig Plot.15 These sub-
stituents had high Hammett s electronic substituent16
values and low Hansch p hydrophobicity17 values.
However, multiple linear regression18 using these two
terms failed to produce an equation that could be used
to predict IKK2 inhibition for related molecules (cross
validated correlation coefficient, q2=0.08, n=19). Two
related molecular descriptors were then calculated and
used instead of s and p, these were hydrogen bonding
strength of N1 (Fig. 2) and ClogP, respectively. Hydro-
gen bonding strengths were calculated using the semi-
empirical method of Gancia et al.19 and ClogP was cal-
culated using BioByte software.20 This method enabled
us to include three molecules for which no s and p were
available (8a, 8n and 8t) and resulted in a more pre-
dictive equation (q2=0.32, n=22). When an insoluble
outlier was removed (8b) from the dataset the pre-
dictivity improved significantly (Correlation coefficient,
r2=0.65, q2=0.47, n=21). This final equation was
IKK2 pIC50=À1.2*H-BondÀ0.1 *ClogP+9.8 (pIC50=
One of the first compounds to be made in this series was
the 4-cyano analogue (8q, Table 1, compounds ordered
À0.3*H-BondÀ0.1*ClogP+7.0
for
standardized
descriptors). It can be seen from these equations that
the hydrogen bonding strength of N1 has the larger
influence on IKK2 activity. The weaker the potential
hydrogen bond to this atom, the greater the activity. In
fact, the calculated hydrogen bond acceptor strength of
the two nitrogens in the pyrimidine ring were effectively
the same (r=0.99, n=22) whilst the hydrogen bond
donating strength of the anilino NH was inversely cor-
related with the acceptor strength of these two atoms
(r=À0.97 and À0.98). This means that greater activity
against IKK2 tends to occur for molecules with a greater
N2 hydrogen bonding potential (Fig. 3). Fortunately,
the changes made to the A ring described above did not
increase the activity against IKK1 by the same amount.
As a result, the most active compounds against IKK2
tend to show an increase in selectivity over IKK1.
Replacing s and p with calculated hydrogen bonding
strength and logP gave a more predictive equation but
this may only be because this allowed the inclusion of
two or three extra data points as the two pairs of
Scheme 1. Synthesis of arylaminopyrimidine analogues. Reagents: (a)
BocPiperazine, Et3N, DCM; (b) DMF–DMA; (c) NC–NH2, HNO3;
(d) DMF, NaOH; (e) 10% TFA/DCM.

A. H. Bingham et al. / Bioorg. Med. Chem. Lett. 14 (2004) 409–412
Table 1. IKK2/IKK1 experimental inhibition data and calculated properties
411
Compd
R-group
IKK2 (nM) IKK1 (nM) IKK1/IKK2
s
p
H-bond ClogP20 Aqueous solubility (pH 6.5 in mg/mL)
8a
550
3090
6
nd
nd
2.75
2.26
nt
8b
8c
8d
8e
8f
8g
8h
1
8i
8j
8k
8l
8m
4-CF3
4-NHCOCH3
4-N-Butyl
4-CH3
4-OCH3
4-NHCONH2
4-Cl
3,4,5-Trimethoxy
3,5-Difluoro
4-SOCH3
H
513
331
309
269
214
155
148
145
117
91
3981
1622
2692
2089
1230
1479
2344
155
4074
2344
1175
5129
6918
8
5
9
8
6
10
16
1
35
26
13
62
105
0.54
0.88
1.90
2.45
2.53
2.42
2.38
2.45
2.17
2.17
1.97
2.08
2.25
2.02
1.58
4.59
2.69
4.75
4.17
3.60
2.37
4.40
2.81
3.98
2.13
3.67
3.16
3.46
Not detected
64.0
0.00 À0.97
À0.51
1.16
0.56
nt
0.1
27.8
nt
4.8
nt
5.0
nt
nt
nt
À0.17
À0.27 À0.02
À0.24 À1.30
0.23
0.71
À0.03 À0.06
0.68 0.24
0.49 À1.58
0.00 0.00
91
83
66
3-CN
4-NO2
0.56 À0.57
0.78 À0.28
Not detected
8n
60
851
14
nd
nd
2.17
3.67
0.1
8o
8p
8q
8r
8s
4-COCH3
4-CONH2
4-CN
4-NHSO2CH3
4-SO2CH3
59
48
44
43
31
1096
3981
3802
331
19
83
87
8
0.50 À0.55
0.36 À1.49
0.66 À0.57
0.03 À1.18
0.72 À1.63
2.11
2.12
1.89
2.15
1.75
3.15
2.21
3.16
2.48
2.09
0.5
Not detected
0.2
27.0
209.0
1202
39
8t
26
11
1698
245
65
21
nd
nd
2.11
1.87
3.70
1.86
nt
nt
8u
4-SO2NH2
0.57 À1.82
nd refers to a value that was not determined; nt refers to a compound that was not tested; s/p values were obtained using TSAR-3-D program
(Accelrys Inc.);18 H-bond values were calculated using the procedure described in ref 19 for N1 (see Fig. 2).
descriptors are highly correlated (r=À0.92 and 0.98,
respectively). An added benefit of using calculated
properties such as hydrogen bonding strength and clogP
is it enables groups that have no reported measured s
and p values to be investigated.
This observed relationship between the activity and the
electron withdrawing properties of substituents attached
to the A ring could be due to a number of factors. One
possibility is that changes in intermolecular hydrogen
bonding strengths result in changes in the affinity of the
ligand for the target. This could be due to small changes
in the binding orientation of the compounds. The fact
that IKK1 activity is not correlated with that of IKK2
for this set of compounds suggests differences between
the binding sites of the two enzymes. Our homology
models show that the amino acid residues proximal to
the ATP binding site are the same in both proteins with
Figure 3. Hydrogen bonding strength of aminopyridimidine core N1
(see Fig. 2) plotted against IKK inhibition IKK1 & and IKK2 ~.
one exception (Ser99 in IKK1 equivalent to Gln100 in
IKK2).

412
A. H. Bingham et al. / Bioorg. Med. Chem. Lett. 14 (2004) 409–412
As these amino acid residues are found close to the
predicted location of Tyr98 in IKK2 (equivalent to Tyr99
in IKK1, Fig. 2), it is not unreasonable to postulate that
this Gln/Ser difference gives rise to the differences in
activity observed. The amino acids at this position could
interact differently with this tyrosine residue, resulting
in differences in its orientation, which in turn could lead
to a change in shape of the ATP binding pocket.
6. Regnier, C. H.; Song, H. Y.; Gao, X.; Goeddel, D. V.;
Cao, Z.; Rothe, M. Cell 1997, 90, 373.
7. Woronicz, J. D.; Gao, X.; Cao, Z.; Rothe, M.; Goeddel,
D. V. Science 1997, 278, 866.
8. Hu, Y.; Baud, V.; Delhause, M.; Zhang, P.; Deerick, T.;
Ellisman, M.; Johnson, R.; Karin, M. Science 1999, 284,
316.
9. Li, Q.; Lu, Q.; Hwang, J. Y.; Buscher, D.; Lee, K. F.;
Izpisua-Belmonte, J. C.; Verma, I. M. Genes Dev. 1999,
13, 1322.
In conclusion, a large number of novel analogues has
been synthesized with a variety of substituents on the
aromatic ring A. The choice of substituents has been
guided by a Craig plot to enhance the range of electro-
nic and hydrophobic properties explored by our analo-
gues. This has enabled a correlation between these
properties and IKK2 activity to be investigated. A large
increase in absolute IKK2 potency has been obtained,
without increasing the IKK1 activity, thereby increasing
the observed selectivity. Our starting compound 1 had
an original IKK2 IC50 of 146 nM, and this series has
been optimized to 11 nMin the case of analogue 8u.
Our starting lead also showed no selectivity for IKK2
over IKK1, whereas we have now achieved a selectivity
of 105-fold in the case of 8m.
10. Li, Q.; Van Antwerp, D.; Mercurio, F.; Lee, K. F.;
Verma, I. M. Science 1999, 284, 321.
11. Takeda, K.; Takeuchi, T.; Itami, S.; Adachi, O.; Kawai,
T.; Sanjo, H.; Yoshikawa, K.; Terada, N.; Akira, S.
Science 1999, 284, 313.
12. Griffiths, R. J. Rheumatoid Arthritis 1997, 1, 473.
13. Bhagwat, S. S.; Brydon L.; Bennett, Y. S.; Eoin, C.;
O’Leary, C.; Leisten, J.; Firestein, G. S.; Boyle, D. S.;
Dreano, M.; Anderson, D. W.; Grimshaw, C. E. The
Small Molecule IKK2 Inhibitor SPC 839 is efficacious in
an Animal Model of Arthritis. Paper was presented at
ACR Annual Scientific Meeting 2001 (San Francisco,
CA).
14. Homology models of IKK1 and IKK2 were built at York
University by George Dickinson and Pavel Herzyk using
MODELLER (Sali, A. and Blundell, T. L., J. Mol. Biol.
1993, 234, 779) and based on structures of PKA, CaMK,
ERK2, P38 and CDK2, all of which have approximately
30% sequence identity to IKK1 and IKK2.
References and notes
15. Craig, P. N. J. Med. Chem. 1971, 14, 680.
16. Hammett, L. P. J. Am. Chem. Soc. 1937, 59, 96.
17. Hansch, C.; Maloney, P. P.; Fujita, T.; Muir, R. M.
Nature 1962, 178.
18. Multiple linear regression with no F-test stepping, using
the TSAR-3-D program (Accelrys Inc., CA, USA).
19. Gancia, E.; Montana, J. G.; Manallack, D. T. J. Mol.
Graphics Mod. 2001, 19, 349.
1. Baldwin, A. S. Ann. Rev. Immunol. 1996, 14, 649.
2. Baeuerle, P. A.; Baltimore, D. Cell 1996, 87, 13.
3. Siebenlist, U.; Franzoso, G.; Brown, K. Annu. Rev. Cell
Biol. 1994, 10, 405.
4. DiDonato, J. A.; Hayakawa, M.; Rothwarf, D. M.;
Zandi, E.; Karin, M. Nature 1997, 388, 548.
5. Mercurio, F.; Zhu, H.; Murray, B. W.; Shevchenko, A.;
Bennett, B. L.; Li, J.; Young, D. B.; Barbosa, M.; Mann,
M. Science 1997, 278, 860.
20. BioByte Corp. (CA, USA), ClogP version 4.2 as imple-
mented in Tripos Inc. (MO, USA) software.