Hit-to-lead studies: The discovery of potent, orally active, thiophenecarboxamide IKK-2 inhibitors

Article abstract of DOI:10.1016/j.bmcl.2004.03.058

A hit-to-lead optimisation programme was carried out on the thiophenecarboxamide high throughput screening hits 1 and 2 resulting in the discovery of the potent and orally bioavailable IKK-2 inhibitor 22.

Full text of DOI:10.1016/j.bmcl.2004.03.058

Bioorganic & Medicinal Chemistry Letters 14 (2004) 2817–2822
Hit-to-lead studies: the discovery of potent, orally active,
thiophenecarboxamide IKK-2 inhibitors
Andrew Baxter,a,* Steve Brough,a Anne Cooper,a Eike Floettmann,a Steve Foster,b
Christine Harding,b Jason Kettle,b Tom McInally,a Craig Martin,a Michelle Mobbs,b
Maurice Needham,b Pete Newham,b Stuart Paine,a Steve St-Gallay,a Sylvia Salter,a
John Unitta and Yafeng Xuec
aAstraZeneca R&D Charnwood, Bakewell Road, Loughborough LE11 5RH, UK
bAstraZeneca R&D AlderleyPark, Cheshire, Macclesfield SK10 4TG, UK
cStructural ChemistryLaboratory, AstraZeneca R&D M o€lndal, S-431 83 M€olndal, Sweden
Received 12 December 2003; revised 5 March 2004; accepted 19 March 2004
Abstract—A hit-to-lead optimisation programme was carried out on the thiophenecarboxamide high throughput screening hits 1
and 2 resulting in the discovery of the potent and orally bioavailable IKK-2 inhibitor 22.
Ó 2004 Elsevier Ltd. All rights reserved.
The use of high throughput screening (HTS) is now
widespread in the pharmaceutical industry. There was
an expectation that once a screen was established for a
particular target then potent lead compounds or candi-
date drugs would be found. The reality is often far from
this. Bridging the gap between the end of a HTS and the
start of a full lead optimisation (LO) project has been
described as hit-to-lead (HtL).1 Hits from HTS are
profiled and compared to generic lead criteria. The lead
target profile used is shown in Figure 1. Lead series then
have a balance of properties––potency, SAR and an
encouraging metabolic and selectivity profile––such that
rapid lead optimisation should provide a candidate drug
(CD) in less than 2 years.
Nuclear factor jB (NF-jB) transcription factors are
composed of homo- and heterodimers of the Rel family
of DNA-binding proteins.2 A key role of these tran-
scription factors is to induce and coordinate the
expression of a broad spectrum of pro-inflammatory
genes including cytokines, chemokines, interferons,
MHC proteins, growth factors and cell adhesion mole-
cules.3 NF-jB is normally retained in the cytoplasm by
IjB, however, upon cellular activation, IjB is phos-
phorylated by an IjB kinase (IKK) and is subsequently
degraded. Free NF-jB then translocates to the nucleus
where it mediates proinflammatory gene expression.
There are three classical IjB’s: IjBa, IjBb and IjBe; all
require the phosphorylation of two key serine residues
Figure 1. Hit-to-lead generic lead target profile.
Keywords: Hit-to-lead; IKK inhibitor; Kinase.
* Corresponding author. Tel.: +44-(0)1509-644772; fax: +44-(0)1509-645513; e-mail: andrew.jg.baxter@astrazeneca.com
0960-894X/$ - see front matter Ó 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2004.03.058

2818
A. Baxter et al. / Bioorg. Med. Chem. Lett. 14 (2004) 2817–2822
before they can be degraded. Two major enzymes
appear to be responsible for IjB phosphorylation: IKK-1
and IKK-2.4 Dominant-negative (DN) versions of either
IKK-1 or IKK-2 (where ATP binding is disabled by the
mutation of a key kinase domain residue) were found to
suppress the activation of NF-jB by TNF-a, IL-1b, LPS
and CD3/CD28 crosslinking; importantly IKK-2 DN
was found to be a far more potent inhibitor than IKK-1
DN.5–7 Furthermore, the generation of IKK-1 and
IKK-2 deficient mice has established the requirement of
IKK-2 for activation of NF-jB by proinflammatory
stimuli and reinforced the dominant role of IKK-2
suggested by biochemical data. Indeed it was demon-
strated that IKK-1 was dispensable for NF-jB activa-
tion by these stimuli.8–10 Thus, inhibition of IKK-2
represents a potentially attractive target for modulation
of immune function and hence the development of drugs
for the treatment of auto-immune diseases.
possesses good TNF inhibitory activity. The focus of the
present study then became the remaining issues of sol-
ubility, mediocre metabolic stability and understanding
of SAR with particular reference to the likely binding
mode.
The compounds were synthesised by the routes shown in
Scheme 1. The parent aminothiophenecarboxylates were
prepared by cyclisation of methyl mercaptoacetate with
aroylacrylonitriles that were in turn prepared from the
corresponding acetophenones.15 Ester hydrolysis gave
the aminoacids, which were converted to the amino-
amides via thionyl chloride treatment followed by
reaction with the appropriate amine or ammonia
(compound 5). The aminoamides were reacted with
either trimethylsilylisocyanate or sodium isocyanate in
aqueous acetic acid to give the unsubstituted ureas 3, 4,
16–22. Reaction of the aminoamide with alkylisocya-
nates gave the substituted ureas 8 and 9. Alternatively
the aminoacid could be converted by phosgene to the
oxazinedione, which reacted with amines to give the
ureidoacids. Thionyl chloride treatment followed by
reaction with ammonia gave the substituted ureas 10
and 11. The isomeric thiophene 13, imidazole 14 and
pyrazine 15 were prepared from the corresponding
documented amino heterocycles.
As part of an IKK-2 inhibitor program, the AZ com-
pound collection was screened in a scintillation prox-
imity assay (SPA) kinase assay that employed
recombinant IKK-2 and a GST-IjBa fusion protein.
The activity and potency of hits were confirmed using
‘filter wash’ kinase assays for both IKK-2 and IKK-
1.11;12 Whole cell activity was determined using the
inhibition of LPS TNF-a production in peripheral blood
mononuclear cells (PBMCs).11;13 Cell toxicity was also
measured in PBMCs using a WST-1 assay.13
Using the carboxamide 4 as a starting point, an explo-
ration of SAR was undertaken focusing in turn on the
amide and urea, central thiophene ring and phenyl
substitution. Table 2 details the effects of simple sub-
stitution on the urea and amide nitrogens. First it should
be noted that the potency increase observed for the
amine to urea transformation is almost identical for the
3-thienyl (1 to 3) and phenyl (5 to 4). Mono and dime-
thylation of the amide caused complete loss of activity (4
vs 6 and 7). Mono-methylation of the terminal urea
nitrogen caused some loss of potency (8 vs 4) but larger
groups (ethyl 9 or benzyl 10) and dimethylation (11)
ablated all activity. Parallel synthesis of a range of
amides (i.e., R3 ¼ COR) also gave only inactive com-
pounds (data not shown). Sulfonamides (e.g., 12) had
weak activity. The SAR of the amide and urea indicates
that all of the H-bond donors and acceptors are either
required for active site kinase interactions or that sub-
The profiles of the thiophenecarboxamides 114 and 2 are
shown in Table 1, compared with the key lead criteria––
potency against isolated enzyme and in cells, in vitro
metabolic clearance by rat hepatocytes and human liver
microsomes, molecular weight and lipophilicity. The
amides have reasonable potency especially for their low
molecular weights, and metabolic stability is acceptable
at this stage. Comparison of the structures of the two
hits 1 and 2 prompted the synthesis of 3. This compound
(3) was found to have an excellent combination of
potency, low molecular weight and log D. The 3-thienyl
group was potentially metabolically labile and so was
replaced by a phenyl group (compound 4) with no loss
of potency and improved in vitro metabolic stability.
Compound 4 is in fact close to the lead criteria and also
Table 1. Profile of thiophenes 1 to 4
NH2
NH
NH2
NH
NH2
O
O
O
NH2
NH
CONH2
S
CONH2
CONH2
S
CONH2
S
S
S
S
1
2
3
4
Generic lead criteria
1
2
3
4
IKK-2 enzyme IC50 < 0.1 lM
PBMC TNF inhib IC50 < 0.3 lM
Rat hepatocyte clearance <14 lL/min/106 cells
Human microsome clearance <23 lL/min/mg
Molecular weight <450
1.6
6.3
18
2.0
>10
14
0.063
0.50
21
0.025
0.25
3
30
49
43
27
224
2.2
185
0.04
267
1.7
261
2.4
log D < 3:0

A. Baxter et al. / Bioorg. Med. Chem. Lett. 14 (2004) 2817–2822
2819
NH2
CN
i
ii
Ar
COOMe
S
Ar
O
Ar
O
iii
NH2
NH2
NH2
vi
iv
Ar
COOH
Ar
CONH2
S
S
5
Ar
CONR1R2
S
v
viii
vii
v
NHCONH2
NHCONR4R5
CONH2
NHCONH2
CONH2
NHCONR4R5
iv
Ar
CONR1R2
Ar
Ar
Ar
COOH
S
S
8-11
S
S
6,7
3,4,16-22
Scheme 1. (i) POCl3, DMF then NH2OHÆHCl; (ii) HACH2CO2Me, NaOH, MeOH; (iii) NaOH, aq MeOH; (iv) SOCl2 then NH3, CH3CN; (v)
TMSNCO, DMF, CH2Cl2 or NaOCN, aq AcOH; (vi) SOCl2 then R1R2NH2; (vii) COCl2 then R4R5NH; (viii) R4NCO, toluene.
Table 2. IKK-2 inhibitory potencies––amide and urea variation
Table 3. IKK-2 inhibitory potencies––central thiophene variation
IKK-2 IC50 (lM)
R3
N
H
NHCONH2
4
0.025
O
S
S
CONH2
S
N
R2
R1
NHCONH2
CONH2
R1
R2
R3
IKK-2 IC50
(lM)a
13
14
0.013
2.0
4
H
H
H
H
Me
H
H
H
H
H
CONH2
H
0.025
3.2
5
H
NHCONH2
6
Me
Me
H
CONH2
CONH2
CONHMe
CONHEt
NA
NA
0.05
NA
N
F
7
N
H
8
CONH2
9b
10
11
12
H
H
CONHCH2Ph NA
NHCONH2
CONH2
N
H
CONMe2
SO2(4-FPh)
NA
1.6
H
15
10
N
a NA ¼ <50% inhibition at 10 lM.
b 4-Fluorophenyl.
stitution causes disruption of the optimum binding ori-
entation. Some confirmation of this hypothesis was
obtained from modelling studies and from an X-ray
crystal structure of a carboxamide inhibitor bound in a
related kinase (see below).
The focus of this study now shifted to explore variation
of the phenyl substituent on amide 4 and its effect on a
wider range of parameters. Table 4 has data on IKK-2
versus IKK-1 selectivity, TNF inhibition in cells versus
cell toxicity and primary in vitro DMPK data. The
profile of the unsubstituted phenyl compound 4 shows
good IKK-2 potency with 40-fold selectivity versus
IKK-1 together with good cellular activity associated
with very weak cell toxicity. Variation of the phenyl
substituents has only a modest effect on potency, selec-
tivity or cell activity. Addition of electron withdrawing
substituents in the 2, 3 or 4 position led to compounds
that were all equiactive (4 vs 19, 20 and 22). A slight fall
in enzyme potency was seen with electron donating
substituents (4 vs 16–18 and 21). This variation was only
Only limited variation of the central thiophene core was
undertaken at this stage (Table 3). Interestingly the
isomeric thiophene (13) retained all activity whilst the
imidazole (14) and pyrazine (15) had much reduced
potency. As stated previously, the orientation of the
urea and amide would be expected to be important and
it may be that internal hydrogen bonds between the urea
NH and heterocyclic ring nitrogens prevent this active
arrangement in the imidazole 14 and pyrazine 15.

2820
A. Baxter et al. / Bioorg. Med. Chem. Lett. 14 (2004) 2817–2822
Table 4. IKK-2 inhibitory potencies––phenyl variation
NHCONH2
CONH2
R
S
R
IKK-2 IC50
(lM)
IKK-1 IC50
(lM)
TNF cell IC50 WSTtox IC
(lM)
Rat hepsa
Hu micsa
Sol (lg/mL)
50
(lM)
2
H
2.0
NA
5.0
1.0
10
NA
0.5
NA
NA
10
14
21
3
49
43
27
16
44
54
1
>50
34
6
3
3-Thienyl
Ph
0.063
0.025
0.20
0.30
0.05
0.1
4
0.25
0.79
16
17
18
19
20
21
22
2-MeOPh
3-MeOPh
4-MeOPh
2-ClPh
3-ClPh
3-HOPh
4-FPh
NA
17
13
10
21
24
23
7.9
5.0
6.3
2.5
3.2
6.3
0.25
NA
2
5
0.04
0.1
0.16
0.50
0.40
NA
NA
NA
10
45
0.063
2
8
1.2
a Units as in Figure 1.
small and disappeared in the whole cell assays––com-
pounds 3, 4, 10, 18 and 20–22 all being essentially
equiactive as inhibitors of TNF production. This level of
cell activity was normally not associated with any cell
toxicity.
cleared and the phenyl and fluorophenyl analogues (4
and 22) had lower clearance and were orally bioavail-
able, consistent with in vitro rat hepatocyte and Caco2
data.
A crystal structure of IKK-2 was not available, so a
model of the protein was constructed. A sequence
alignment was performed for all the kinase structures
available from the Brookhaven Database at the time of
the study, along with several AZ proprietary kinase
structures and human IKK-216 (accession code 080158).
Four template structures were selected: Brookhaven
A key component of the hit-to-lead process is to provide
lead compounds with acceptable DMPK. To this end
the more potent inhibitors in Table 4 were profiled in rat
hepatocytes and human liver microsome preparations.
Clearance of the parent thiophene 3 just exceeded the
lead criteria limits and the phenyl compound 4 was
borderline in human microsomes but stable in rat
hepatocytes. Other analogues though were found to give
improvements probably due to prevention of the pre-
dicted oxidative aryl metabolism. Substitution by simple
electron donating substituents caused increased meta-
bolic clearance consistent with increased susceptibility to
oxidative attack on the electron rich phenyl rings (4 vs 3
and 16–18). The expected clearance reductions were
observed, at least for the human microsomes, by the
addition of electron withdrawing substituents (4 vs 19,
20 and especially 22). Selected compounds in this study
were tested in vivo for pharmacokinetic properties
(Table 5). An initial in vitro assessment of the bio-
availability potential was undertaken using Caco2 per-
meability. Results obtained in vivo subsequently
confirmed the in vitro data. The thiophene 3 was rapidly
code 1AQ117 (a cell division protein kinase at 2.2 A
ꢀ
resolution); Brookhaven code 1HCK18 (a cell division
ꢀ
protein kinase at 1.9 A resolution); Brookhaven code
ꢀ
1AD519 (a protein tyrosine kinase at 2.6 A resolution)
and 2HCK19 (a protein tyrosine kinase at 3.0 A resolu-
ꢀ
tion). These were used to generate 10 models of human
IKK-2 using MODELLER v5.0,20 and the model with
the lowest objective function was selected. All hydrogen
atoms were then added to this model using CHARMm21
v25.2 with parameters v22. Automated docking pro-
grams did not place the compounds at the ATP binding
site without pre-specifying the binding mode with con-
straints. This is possibly due to the closed nature of the
pocket and the known conformational flexibility of
kinases that can open out and then close around a
ligand. It was decided to model the compounds in the
Table 5. DMPK parameters, Caco2 and plasma protein binding
NHCONH2
CONH2
R
S
R
Rat hepsa
Caco2
In vivo rat pharmacokineticsa
PPB (%)
Cl
Vss
T1/2
F
3
3-Thienyl
Ph
4-FPh
21
3
10.2
17.7
12.2
51
18
6
1.6
3.2
0.7
0.4
2.5
1.4
7
93.1
94.9
92.5
4
22
27
78
2
a Units as in Figure 1.

A. Baxter et al. / Bioorg. Med. Chem. Lett. 14 (2004) 2817–2822
2821
Figure 4. IKK-2 model complexed with compound 4 showing the
Connolly solvent accessible surface coloured by lipophilicity. Brown
represents more lipophilic regions, blue show more hydrophilic
regions.
Figure 2. Human IKK-2 model complexed with compound 4, showing
the key hydrogen bond network.
Cys99 in IKK-2 model is different from JNK-1 in the
same region due to nonhomologous sequences.
active site manually, using the available compound SAR
and knowledge of inhibitor binding to other kinases to
determine the initial binding mode. An energy minimi-
sation was performed using CHARMm for the protein
ligand complex with some restraints. The result for
compound 4 is shown in Figure 2.
Figure 4 shows the IKK-2 model complexed with
compound 4 highlighting the Connolly solvent accessi-
ble22 surface coloured by lipophilicity.23 Brown repre-
sents more lipophilic regions, blue show more
hydrophilic regions. This model helps to explain some of
the SAR seen in Tables 2 and 4. From Table 2, substi-
tution on the amide nitrogen (R1 and R2) disrupts the
hydrogen bond with the backbone carbonyl of residue
Tyr98, but also the space in this region is tight, so it is not
surprising that compounds 6 and 7 are inactive. Sub-
stitution off the urea group (R3) will put groups out into
a solvent accessible region, so the lipophilic substitu-
tions in compounds 8, 9 and 10 are less active. Removal
of the hydrogen bond and replacement with a methyl
group also produced an inactive compound (compound
11) that can be explained by the model in terms of steric
clash and loss of a key hydrogen bond. Table 4 shows
how a variety of lipophilic thiophene substitutions are
tolerated. This region is structurally rather featureless in
the plane of the aromatic ring, but there is little space
above and below the plane of the ring. The pocket is
generally quite lipophilic, becoming less so as the solvent
region is approached. This explains why a variety of
hydrophobic substitutions that maintain the flatness of
the molecule are well tolerated.
It was possible to obtain structural data from the related
kinase, JNK-1. Even though the sequence identity for
the kinase domains of human JNK-1 and human IKK-2
is only 16%, compound 4 has a JNK-1 IC50 of 1.6 lM
and the structure of this compound complexed with
ꢀ
human JNK-1 was obtained at 2.1 A resolution. Over-
laying the two kinase domains by the Ca atoms that are
similar between the protein sequences confirms the
modelled IKK-2 binding mode, with all the predicted
hydrogen bonds appearing in the X-ray structure (see
Fig. 3). There is some discrepancy since the loop around
Combination of the structural features present in the
HTS hits 1 and 2 led to the potent inhibitor 3. Alteration
of the metabolically labile thiophene to phenyl and then
to the more stable 4-fluorophenyl led to the identification
of the thiophenecarboxamide 22 as a potent, orally active
IKK-2 inhibitor that fulfilled almost all of the lead cri-
teria shown in Fig. 1. Poor solubility was the only
continuing deficiency that was progressed as a potential
problem for a future lead optimisation programme. The
poor solubility though did not affect the oral bioavail-
ability of this compound. The series was accepted as the
basis for a new IKK-2 lead optimisation project.
Figure 3. Overlay of the X-ray structure of compound 4 complexed
with human JNK-1 (blue) with the model of compound 4 IKK-2
complex (yellow).

2822
A. Baxter et al. / Bioorg. Med. Chem. Lett. 14 (2004) 2817–2822
addition of 25 lL Microscint-40 on a Packard topcount
microplate scintillation counter.
References and notes
13. Peripheral blood mononuclear cells (PBMCs) were pre-
pared from human whole blood using LymphoPrep
separation, counted using a haemocytometer and diluted
to 2.6 ꢀ 106/mL in cell medium {RPMI 1640 supplemented
with 2 mM glutamine, 1% (v/v) heat inactivated human A–
B serum and 100 IU/mL of penicillin and streptomycin}.
PBMCs (100 lL) were cultured with test compounds
(50 lL) in tissue culture treated 96-well plates 30 min prior
to the addition of bacterial endotoxin lipopolysaccharide
(LPS) (E. coli type 0111:B4: 50 lL of 4 lg/mL) and then
incubated for a further ꢁ16 h. The production of TNFa in
response to LPS by the PBMCs was stopped by removal of
the cell medium from the PBMCs by aspiration. TNFa in
the cell assay medium was quantitated using a conven-
tional enzyme-linked immunosorbent assay and nonspe-
cific compound effects on PBMC viability measured by
incubation of the remaining treated cells in a WST-1 based
cytotoxicity assay.
1. Michne, W. F. Pharm. News 1996, 3, 19.
2. Baldwin, A. S. Annu. Rev. Immunol. 1996, 14, 649.
3. Pahl, H. L. Oncogene 1999, 18, 6853.
4. Perkins, N. D. Trends Biochem. Sci. 2000, 9, 434.
5. Zandi, E.; Rothwarf, D. M.; Delhasse, M.; Hayakawa,
M.; Karin, M. Cell 1997, 91, 243.
6. Mercurio, F.; Zhu, H.; Murray, B. W.; Shevchenko, A.;
Bennett, B. L.; Li, J.; Young, B. D.; Barbosa, M.; Mann,
M.; Manning, A.; Rao, A. Science 1997, 278, 860.
7. Woronicz, J. D.; Gao, X.; Cao, Z.; Rothe, M.; Goeddel,
D. V. Science 1997, 278, 866.
8. Tanaka, M.; Fuentes, M. E.; Yamaguchi, K.; Durnin,
M. H.; Dalrymple, S. A.; Hardy, K. L.; Goeddel, D. V.
Immunity 1999, 10, 421.
9. Li, Q.; Van Antwerp, D.; Mercurio, F.; Lee, K. F.; Verma,
I. M. Science 1999, 284, 321.
10. Li, Z. W.; Chu, W.; Hu, Y.; Delhase, M.; Deerinck, T.;
Ellisman, M.; Johnson, R.; Karin, M. J. Exp. Med. 1999,
189, 1839.
11. For full details of the biological assays used see: Baxter,
A.; Brough, S.; Faull, A.; Johnstone, C.; McInally, T. Int.
Patent Appl. WO 01/58890.
12. Recombinant IKK enzymes and GST-IjBa were prepared
as described.11 IKK-2 kinase activity was determined using
a ‘filter kinase assay’. A GST-IjBa (1–53) fusion protein
was used as a substrate for IKK-1 and IKK-2. Briefly, 96-
well plates containing 50 lL reactions of 150 ng IKK-2
and 540 ng GST-IjBa in 20 mM HEPES pH 7.2, 10 mM
MgCl2, 0.35 mM MnCl2, 850 ng BSA, 2 lM ATP, 0.1 lCi
33Pc ATP and test compound at a suitable concentration
were incubated for 60 min at room temperature. The same
conditions were used for IKK-1, except 1 lg of GST-IjBa
(1–53) was used. The kinase reaction was then stopped by
addition of ice-cold 50 lL 20% (v/v) trichloroacetic acid
(TCA) containing 200 mM ATP. Proteins were allowed to
precipitate for 10 min before filtration through a Packard
Unifilter onto Packard GF/C filter 96-well plates. Wells
were washed six times with 2% (v/v) TCA and dried at 30–
40 °C for 60 min. Radioactivity was counted after the
14. Thiophenecarboxamide 1 has also been identified by
others as an IKK-2 inhibitor see Kishore, N.; Sommers,
C.; Mathialagan, S.; Guzova, J.; Yao, M.; Hauser, S.;
Huynh, K.; Bonar, S.; Mielke, S.; Albee, L.; Weier, R.;
Graneto, M.; Hanau, C.; Perry, T.; Tripp, C. S. J. Biol.
Chem. 2003, 278, 32861.
15. Davis, R. V.; Fraser, J. EP209977.
16. Hu, M. C. T.; Wang, Y.-P. Gene 1998, 222, 31.
17. Lawrie, A. M.; Noble, M. E.; Tunnah, P.; Brown, N. R.;
Johnson, L. N.; Endicott, J. A. Nat. Struct. Biol. 1997, 4,
796.
18. Schulze-Gahmen, U.; De Bondt, H. L.; Kim, S. H. J. Med.
Chem. 1996, 39, 4540.
19. Sicheri, F.; Moarefi, I.; Kuriyan, J. Nature 1997, 385, 602.
ꢁ
20. Sali, A.; Blundell, T. L. J. Mol. Biol. 1993, 234, 779.
21. Brooks, B. R.; Bruccoleri, R. E.; Olafson, B. D.; States,
D. J.; Swaminathan, S.; Karplus, M. Comp. Chem. 1983,
4, 187.
22. Connolly, M. L. Science 1983, 221, 709.
23. Ghose, A. K.; Crippen, G. M. J. Comp. Chem. 1986, 7,
565.