Indole RSK inhibitors. Part 2: Optimization of cell potency and kinase selectivity

Article abstract of DOI:10.1016/j.bmcl.2011.10.029

A series of inhibitors for the 90 kDa ribosomal S6 kinase (RSK) based on an 1-oxo-2,3,4,5-tetrahydro-1H-[1,4]diazepino[1,2-a]indole-8-carboxamide scaffold were optimized for cellular potency and kinase selectivity. This led to the identification of compound 24, BIX 02565, an attractive candidate for use in vitro and in vivo to explore the role of RSK as a target for the treatment heart failure.

Full text of DOI:10.1016/j.bmcl.2011.10.029

Bioorganic & Medicinal Chemistry Letters 22 (2012) 738–742
Contents lists available at SciVerse ScienceDirect
Bioorganic & Medicinal Chemistry Letters
journal homepage: www.elsevier.com/locate/bmcl
Indole RSK inhibitors. Part 2: Optimization of cell potency and kinase selectivity
Thomas M. Kirrane a,b, , Stephen J. Boyer a, , Jennifer Burke a, Xin Guo a, Roger J. Snow a,
⇑
Lida Soleymanzadeh a, Alan Swinamer a, Yunlong Zhang a, Jeffery B. Madwed b,à
,
Mohammed Kashem a, Stanley Kugler a, Margaret M. O’Neill b
a Department of Medicinal Chemistry, Boehringer-Ingelheim Pharmaceuticals, Inc., 900 Ridgebury Road, PO Box 368, Ridgefield, CT 06877, USA
b Department of CardioMetabolic Diseases Research, Boehringer-Ingelheim Pharmaceuticals, Inc., 900 Ridgebury Road, PO Box 368, Ridgefield, CT 06877, USA
a r t i c l e i n f o
a b s t r a c t
Article history:
Available online 18 October 2011
A series of inhibitors for the 90 kDa ribosomal S6 kinase (RSK) based on an 1-oxo-2,3,4,5-tetrahydro-1
H-[1,4]diazepino[1,2-a]indole-8-carboxamide scaffold were optimized for cellular potency and kinase
selectivity. This led to the identification of compound 24, BIX 02565, an attractive candidate for use
in vitro and in vivo to explore the role of RSK as a target for the treatment heart failure.
Ó 2011 Elsevier Ltd. All rights reserved.
Keywords:
RSK
MAPKAP-K1
Kinase inhibitor
Activation of the sodium–hydrogen exchanger (NHE) in
myocardial tissue is crucial for pH maintenance during the early
phase of cellular stress, such as post myocardial infarction.
However, over longer periods, NHE activation leads to Ca2+ over-
load, and eventually cardiac hypertrophy.1,2 It has been demon-
strated that NHE inhibitors can block this process and are
efficacious in preclinical models of ischemia-reperfusion injury
and heart failure.3 Several NHE inhibitors reached the clinic, but
most were plagued with adverse events that limited their useful-
ness.1 It has been hypothesized that the side effects associated
with NHE inhibitors could be due to inhibition of NHE’s role main-
taining basal intracellular pH levels. One way to address this would
be to target NHE activation instead of NHE directly. Phosphoryla-
tion of NHE by the 90 kDa ribosomal S6 kinase (RSK, also known
as mitogen-activated protein kinase-activated protein kinase-1
[MAPKAP-K1]) has been shown to activate NHE under stress condi-
tions and could be a potential point for therapeutic intervention.4
In a previous paper, we described the identification of a series of
indole lactam RSK inhibitors via a high-throughput screen and
their optimization into compounds with moderate cellular potency
and acceptable kinase selectivity profiles.5 In this Letter, we will
discuss the SAR for three classes of heteroaromatic amides with a
focus on further improvement of cellular potency, kinase
selectivity, and physicochemical properties.
Since we were interested in a heart failure treatment that could
be administered chronically, we were particularly focused on
selecting compounds with the best possible safety profile. Thus, a
significant part of our effort centered on optimizing for kinase
selectivity. Previous internal experience with several kinase targets
had shown that utilizing a single representative off-target kinase to
assess selectivity often led to surprising selectivity issues with ad-
vanced compounds. For this program, we chose instead to monitor
selectivity for key compounds against a panel of 20 kinases (Table
1). The panel was determined by profiling several compounds from
the early optimization effort at a single concentration against 226
unique kinases offered by Invitrogen.6 Those with the highest
degree of cross-reactivity with our lead series were selected for
the routine panel. Interestingly, we found that affinity profile
was driven largely by chemotype and not by kinase homology, con-
sistent with other observations reported in the literature.7
It was most efficient to track kinase activity across a panel as
percent inhibition at a single concentration of 3 lM. Additionally,
through correlation with dose–response experiments, we found
that inhibition greater than 80% typically returned an IC50 under
1 lM, which would highlight a compound with a potential
concern. To aid the comparison of compound kinase selectivity
profiles, we instituted a selectivity count, defined as the number
of kinases with greater than 80% inhibition. Shown in Table 1 are
a non-selective HTS hit 1 along with an example of the oxazole
2, pyrazole 3, and benzimidazole 4 classes of amides. It is impor-
tant to note that this type of qualitative assessment has the great-
est utility when comparing compounds of similar RSK potency
Compounds in Tables 1–3 were made via amide coupling of in-
dole lactam carboxylic acid with the corresponding heteroaromatic
⇑
Corresponding author.
E-mail address: tom.kirrane@boehringer-ingelheim.com (T.M. Kirrane).
Present address: Global Regulatory Sciences, Bristol-Myers Squibb, 5 Research
Parkway, Wallingford, CT 06492, USA.
à
Present address: Department of Cardiovascular Diseases, Merck
126 E. Lincoln Avenue, PO Box 2000, Rahway, NJ 07065, USA.
& Co. Inc.,
0960-894X/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2011.10.029

T. M. Kirrane et al. / Bioorg. Med. Chem. Lett. 22 (2012) 738–742
739
Table 1
Table 2
Kinase selectivity for selected compounds
SAR of isoxazolyl and pyrazole amides
O
O
H
N
H
N
R1
NH
( )n
N
NH
N
R
R
O
O
R
Compound
R
RSK2
IC50
(nM)
HLR-
CREB
IC50
Selectivity
countb
Aqueous
solubility
(lg/mL)
a
11
(nM)
2
8
19
200
9
8
2
N
N
O
>10,000
O
N
9
18
26
250
460
9
5
21
O
10
>35
N
O
11
86
360
2
<0.1
N
O
12
3
13
11
7
130
790
120
0.1
>32
0.6
N
O
%
N
N
9
9
N
N
13
N
N
N
14
15
10
2
460
90
>40
>40
N
N
N
a
Values are means of at least two experiments, standard deviation is typically
50% reported value.
b
Selectivity count = number of kinases with >80% inhibition @ 3 lM, 20 total
kinases in panel shown in Table 1.
a 2-fluoro-nitrobenzene (5) to give aniline 6. Following reduction
of the nitro group, the desired 2-aminobenzimidazole 7 was ob-
tained by condensation with cyanogen bromide. Substituted lac-
tam coupling partners were prepared as previously reported,
with the exception that enantiopure (R)- and (S)-1-Me-lactams
were made using enantiopure aminoalcohols.5
Homology of ATP pocket as defined by Vieth, et al. versus RSK2
(N-terminal domain) in ().
a Green represents <80% inhibition, red is P 80% inhibition at 3
lM.
As previously mentioned, both the isoxazole and pyrazole LHS
amides give potent RSK inhibitors with moderate cellular potency
and have similar SAR trends.5 It is worth noting that substitution
on these aminoheterocycles adjacent to the amide results in a com-
plete loss of potency, likely as a result of lost planarity between the
heteroaromatic ring and the amide (compare 2 and 8). However,
docking studies showed potential for substitution at the 3-postion
amine left hand side (LHS).5 4-Aminopyrazoles were prepared from
commercially available 4-nitropyrazole following the literature
procedure.8 The synthesis of 2-aminobenzimidazoles (Scheme 1)
began with SNAr addition of ammonia or a substituted amine to
O
N+
O
N+
N
a
b, c
R1
O
O
R1
R1
NH2
N
R2
NH
R2
F
5
6
7
Scheme 1. Reagents and conditions: (a) amine, Hunig’s base, DMSO, 60 °C, 18 h; (b) ammonium formate, Zn, EtOH, 60 °C, 2 h; (c) cyanogen bromide, EtOH, 18 h, 10–50%
(three steps).

740
T. M. Kirrane et al. / Bioorg. Med. Chem. Lett. 22 (2012) 738–742
Table 3
series, we found it difficult to drive down potency while maintain-
ing a reasonable balance of selectivity and solubility.
SAR of 2-benzimidazoyl amides
O
Similar SAR was observed with pyrazole based amides. More-
over, the chemistry to make the pyrazoles allowed access to a
wider range of ring substituents that were not readily achievable
in the isoxazole series, such as the cyclohexylmethyl and pyridyl-
methyl fragments. The similar RSK potency for compounds 13
and 3 demonstrates the size of the lipophilic pocket and also high-
lights that pi-stacking does not play the same role in the lipophilic
pocket as it does with the amide substituent. Introduction of the
pyridylmethyl substituents in compounds 14 and 15 gave the ex-
pected improvement in aqueous solubility, and also provided the
4-pyridylmethyl compound 15 with a 5-fold boost in potency.
While compound 15 fulfilled most of our criteria, it was later dis-
covered to be a potent CYP3A4 inhibitor. Thus we turned to explore
other amide substituents.
We also looked at five-membered heterocycles fused to a
six-membered ring in the form of benzimidazole analogs such as
compound 4 (Table 3). One of the unique characteristics of the
benzimidazole amide series is its ability to tolerate substitution
adjacent to the amide (compare compound 8 with 4). This is likely
due to the formation of the 2-imino tautomer which simulta-
neously forces co-planarity as well as provides a hydrogen bond
donor interaction with Leu150 (Fig. 1), an effect that has been ob-
served for other amide substitutents in this series.5 Additionally,
the phenyl ring of the benzimidazole forms a key pi stacking inter-
action with Phe357.
H
N
NH
N
R
O
Compound
R
RSK2
IC50
(nM)
HLR-
CREB
IC50
Selectivity Aqueous
countb
Solubility
(lg/mL)
a
10
(nM)
N
4
9
1
200
40
13
13
21
3
N
N
16
N
N
N
17
18
20
61
180
4
9
12
N
N
1000
0.1
N
N
19
21
200
3
39
Since benzimadazoyl amides tolerated substitution adjacent to
the amide, in contrast to the other series, we set out to explore
the SAR at this position. Homologation of compound 4 to give 16
provided a significant boost in molecular and cellular potency,
which came at the expense of solubility. Docking suggests the
improvement results from the ethyl group fitting into a lipophilic
indentation on the floor of the pocket. The additional steric bulk
of isopropyl compound 18 shows a similar level of activity to 4
but has significantly improved kinase selectivity. Unfortunately, a
further increase in steric demand, as in compound 18, results in
micromolar cellular potency likely due to a clash with the wall of
the protein.
Molecular modeling suggested an extension of the ethyl side
chain could be tolerated and might allow for interaction with the
acidic residues on the floor of the binding pocket. We looked at a
number of basic amine containing sidechains of varying lengths
where the amine was either acyclic or cyclic, and discovered the
dimethylaminopropyl compound 19. Its RSK potency is maintained
at a similar level as methyl compound 4, while both the aqueous
solubility and kinase selectivity were measurably improved.
In addition to investigating substitution on the nitrogen, we
also wanted to explore the available space shown by modeling to
be adjacent to the phenyl ring of the benzimidazole. Chloro com-
pounds 20 and 21 showed very good potency and good to excellent
kinase selectivity, but both were withheld from further profiling
due to very poor aqueous solubility. We sought to improve solubil-
ity by combining the chloro substituent with the aminopropyl
sidechain to give compound 22. While its solubility is dramatically
improved and RSK potency maintained, the selectivity of com-
pound 22 was unexpectedly decreased relative to either 19 or 21.
We also attempted to modulate physicochemical properties by
introducing a nitrogen in the fused six membered ring of 16 to give
23. The compound maintained binding and cellular potency; how-
ever, it did not have the intended effect of improving solubility. Of
all the benzimidazole analogs we examined, the dimethylamino-
propyl substituent consistently offered the best balance of cellular
potency, kinase selectivity, and aqueous solubility.
N
Cl
N
20
21
6
130
90
7
3
<0.1
<0.1
N
N
12
N
Cl
Cl
N
N
22
23
20
4
160
80
10
12
29
2
N
N
N
N
a
Values are means of at least two experiments, standard deviation is typically
50% reported value.
b
Selectivity count = # of kinases with >80% inhibition @ 3 lM.
of the ring to reach a pocket that could be filled with larger, lipo-
philic substituents. The entrance to the lipophilic pocket appears
to be narrow, and potency drops off as the steric bulk increased
from methyl (2) and cyclopropyl (9) to isopropyl (10) and finally
phenyl (11). It is also noteworthy that kinase selectivity improves
along with steric bulk, as the selectivity count for 2 is 9 while that
for 11 is 2. Extending the substituent one carbon further, as in ben-
zyl compound 12, alleviates any steric clash and results in similar
activity to the corresponding methyl analog 2. The result suggests
that the benzyl group is tolerated, but filling the pocket does not
provide an improvement in inhibition. An improvement in aqueous
solubility is seen as the methyl group is substituted (compare 2, 9,
and 10); however, the addition of another aromatic ring in 11 and
12 results in very poorly soluble compounds.9 In the isoxazole
With the aminopropylbenzimidazole LHS in hand, we returned
to the lactam ring. It was known from earlier work with a pyridyl

T. M. Kirrane et al. / Bioorg. Med. Chem. Lett. 22 (2012) 738–742
741
O
H
N
N
NH
N
N
O
RSK2 1 nM
N
LRRK2
PRKD1
PRKD2
PRKD3
RET
16 nM
35 nM
139 nM
219 nM
161 nM
CLK1
CLK2
FGFR2
PDGFRa
FLT3
512 nM
112 nM
320 nM
956 nM
714 nM
Figure 2. Kinase selectivity profile of compound 24.
Figure 1. Docking of 3 into a crystal structure of RSK2 [30–350] (2.9 Å resolution,
Boehringer-Ingelheim unpublished results). The binding mode is based on an X-ray
structure for an analogous compound bound to MK2. Dashed yellow lines represent
H-bonds <2.5 Å.
At this point, we returned to the panel of 20 kinases used for
optimization to obtain a more accurate picture of kinase selectivity
for 24 (Fig. 2). Six kinases furnished an IC50 less than 1 lM: the five
that had shown greater than 80% inhibition in the selectivity count,
and PDGFRa which had shown 71% inhibition. Additionally, com-
pound 24 was more than 100-fold selective relative to RSK for
the kinase panel, with the exception of LRRK2 and PRKD1, for
which compound 24 showed 16- and 35-fold selectivity, respec-
tively. We then expanded the kinase panel to include over 200
additional unique kinases. We were pleased to find that 24 had
Table 4
Piperazinone and diazepinone SAR
H
N
Core
N
N
O
an IC50 of less than 1 lM for only four other family members
(PRKD2, PRKD3, RET, and FLT3). These results confirmed the power
of the selectivity screening methodology developed for this pro-
gram. By tracking a broader panel of potentially cross-reactive ki-
nases at a single concentration, we were able to identify a
compound with high selectivity vs. most of the readily testable
portion of the kinome.
N
Compound Core
RSK2
IC50
(nM)
HLR-
CREB
IC50
Selectivity
countb
Aqueous
solubility
(lg/mL)
a
10
(nM)
In summary, we followed up on the initial indole series leads
1–3 to further optimize potency, kinase selectivity and physical
properties. Based on its overall balanced profile, compound 24,
BIX 02565, is an attractive candidate for use in vitro and in vivo
to explore the role of RSK as a target for the treatment heart failure.
Further results will be reported in due course.10
O
24
1
20
5
5
26
NH
N
N
N
*
O
25
14
27
100
>45
NH
*
Acknowledgements
O
The authors would like to express their gratitude to Dr. John
Miglietta for determining the percent homology of the ATP pocket
for the kinase selectivity panel versus the N-terminal ATP pocket of
RSK2.
NH
26
290
3
>50
*
a
Values are means of at least two experiments, standard deviation is typically
50% reported value.
b
Selectivity count = # of kinases with >80% inhibition @ 3
lM.
References and notes
1. Karmazyn, M.; Sawyer, M.; Fliegel, L. Curr. Drug Targets Cardiovasc. Haematol.
Disord. 2005, 5, 323.
2. Nakamura, T. Y.; Iwata, Y.; Arai, Y.; Komamura, K.; Wakabayashi Circ. Res. 2008,
103, 891.
3. (a) Karmazyn, M.; Kilic, A.; Javadov, S. J. Mol. Cell Cardiol. 2008, 44, 647; (b)
Karmazyn, M.; Gan, X. T.; Humphreys, R. A.; Yoshida, H.; Kusumoto, K. Circ. Res.
1999, 85, 777.
4. (a) Moor, A. N.; Gan, X. T.; Karmazyn, M.; Fliegel, L. J. Biol. Chem. 2001, 276,
16113; (b) Maekawa, N.; Abe, J.-I.; Shishido, T.; Itoh, S.; Ding, B.; Sharma, V. K.;
Sheu, S.-S.; Vlaxall, B. C.; Berk, B. C. Circulation 2006, 113, 2516; (c) Avkiran, M.;
Cook, A. R.; Cuello, F. Curr. Opin. Pharmacol. 2008, 8, 133.
5. Boyer, S. J.; Burke, J.; Guo, X.; Kirrane, T. M.; Snow, R. J.; Zhang, Y.; Sarko, C.;
Soleymanzadeh, L.; Swinamer, A.; Westbrook, J.; DiCapua, F.; Padyana, A.;
Cogan, D.; Gao, A.; Xiong, Z.; Madwed, J. B.; Kashem, M.; Kugler, S.; O’Neill M.
M. Bioorg. Med. Chem. Lett. 2011. doi:10.1016/j.bmcl.2011.10.030.
6. Compounds were tested in the Kinase SelectScreenÓ Kinase Profiling service,
www.invitrogen.com. Kinase inhibition data were determined using full length
human proteins.
7. Vieth, M.; Higgs, R. E.; Robertson, D. H.; Shapiro, M.; Gragg, E. A.; Hemmerle, H.
Biochim. Biophys. Acta 2004, 1697, 243.
8. Bouillot, AMJ; Daugan, AC-M; Dean, AW; Fillmore, MC. WO2008074824.
LHS amide that substitution of the lactam ring could enhance
potency as well as selectivity. Therefore, we combined the amino-
propylbenzimidazole with a number of mono-, di-, and spiro-
substituted lactams.5 In general, the SAR trends for lactam substi-
tution did not differ from what was previously observed for the
pyridyl amide series: mono-methylation tended to boost potency
while gem-dimethylation, as in 26, was detrimental (Table 4).
We were pleased to see a 14-fold enantiomeric preference for
the predicted eutamer (R)-24 over the distamer (S)-25. This result
further supports the docking hypothesis, which suggested one
enantiomer of a methyl group adjacent to the indole nitrogen
would occupy a small indentation in the roof of the binding pocket.
Of the substituted lactam analogs containing the aminopropyl-
benzimidazole LHS, compound 24 demonstrated the best combina-
tion of potency, selectivity, and solubility and was chosen for
further profiling.

742
T. M. Kirrane et al. / Bioorg. Med. Chem. Lett. 22 (2012) 738–742
9. Compound solubility was measured as follows: In duplicate deep well plates, 6
of a 10 mM DMSO stock solution was spiked into pH 7.4 buffer, targeting 50–
200 M final concentrations. The final DMSO concentration was 0.5%. The
samples were incubated for 16–18 h, filtered, and analyzed using
l
L
10. Shi, X.; Yan, C.; Brookes, P. S.; Berk, B. C. Manuscript in preparation.
11. Inhibition of the phosphorylation of the transcription factor CREB (cAMP
Response Element Binding) by RSK2 in cells was determined according to the
method described in: Boyer, S.; Gao, A.; Guo, X.; Kirrane, T.; Sarko, C.; Snow, R.;
Soleymanzadeh, F.; Zhang, Y. WO2011071716.
l
a
spectrophotometer. The UV spectra of samples and reference were scanned
from 230 to 500 nm. Solubility is measured by comparing the ratio of the areas
under the curves for reference and sample.