Pyrimidine-based inhibitors of CaMKIIδ

Article abstract of DOI:10.1016/j.bmcl.2008.02.056

Non-ATP competitive pyrimidine-based inhibitors of CaMKIIδ were identified. Computational studies were enlisted to predict the probable mode of binding. The results of the computational studies led to the design of ATP competitive inhibitors with optimized hinge interactions. Inhibitors of this class possessed improved enzyme and cellular activity compared to early leads.

Full text of DOI:10.1016/j.bmcl.2008.02.056

Available online at www.sciencedirect.com
Bioorganic & Medicinal Chemistry Letters 18 (2008) 2404–2408
Pyrimidine-based inhibitors of CaMKIId
Babu Mavunkel, Yong-jin Xu, Bindu Goyal, Don Lim, Qing Lu, Zheng Chen,
Dan-Xiong Wang, Jeffrey Higaki, Indrani Chakraborty, Albert Liclican, Steve Sideris,
Maureen Laney, Ulrike Delling, Rosanne Catalano, Linda S. Higgins, Hui Wang,
Jing Wang, Ying Feng, Sundeep Dugar and Daniel E. Levy*
Scios, Inc., 6500 Paseo Padre Parkway, Fremont, CA 94555, USA
Received 6 December 2007; revised 21 February 2008; accepted 22 February 2008
Available online 4 March 2008
Abstract—Non-ATP competitive pyrimidine-based inhibitors of CaMKIId were identified. Computational studies were enlisted to
predict the probable mode of binding. The results of the computational studies led to the design of ATP competitive inhibitors with
optimized hinge interactions. Inhibitors of this class possessed improved enzyme and cellular activity compared to early leads.
Ó 2008 Elsevier Ltd. All rights reserved.
The importance of calcium signaling to the regulation of
biological pathways, along with the associated roles of
the Ca2+/calmodulin-dependant kinases (CaMK)s, was
summarized in previous reports from our group.1–3 In
these reports, maleimide-based compounds capable of
inhibiting a recombinant preparation of CaMKIId were
described. Concurrent to these studies, a class of pyrim-
idine-based CaMKIId inhibitors was identified. The
preparation of the initial lead series, shown in Scheme
1, began with a Suzuki coupling4 with 2,4-dichloropy-
rimidine generating compounds 3. Buchwald coupling5
of a-S-methylbenzylamine with 2-chloro-4-nitropyridine
followed by the reduction of the nitro group gave com-
pound 6. Compounds 3 and 6 were joined using Buch-
wald conditions.
pound, for the first time, demonstrated sub-micromolar
potency and was determined to bind competitively with
ATP. This indicated that we were, at that time, interact-
ing with the ATP site, itself. This hypothesis was further
supported through computational analyses where the
carboxylic acid of compound 7f was shown capable of
binding, via salt bridge, to Lys22 thus blocking the
entrance to the ATP site.
Building upon the initial SAR, we studied the parame-
ters surrounding the activity of 7f. In particular, could
the carboxylic acid be modified to alternate functional-
ities, isosteres or homologated analogs? To support this
SAR, compounds 7g–i were prepared according to
Scheme 1. Furthermore, compounds 7j and 7l were pre-
pared from the methyl ester (7e) using standard condi-
tions. Finally, the tetrazole analog 7k was prepared
from a nitrile generated according to Scheme 1 and
reacting the nitrile with TMS azide.8
As shown in Table 1, the initial SAR studies resulted in
a series of inhibitors with IC50 values ranging from 1 to
3 lM (Table 1) against the cloned enzyme.6 These com-
pounds were determined to be non-competitive inhibi-
tors with respect to ATP. Furthermore, using a
computationally generated homology model for CaM-
KIId,7 these compounds were predicted to bind in a
groove near the opening to the ATP site and pointing
towards the solvent front. With this binding hypothesis
in mind, one additional compound, 7f, was prepared in
order to provide better interactions with the solvent
front residues. As the breakthrough structure, this com-
The data associated with this extended SAR are summa-
rized in Table 2. As illustrated, the position of the car-
boxylic acid was critical with essentially all activity
being lost when this group was moved to the 3-position
(7g). Further supporting the importance of the position-
ing of the carboxylic acid is the loss in potency resulting
in one-carbon homologation (7h). Maintaining an acidic
pH was demonstrated necessary when noting the re-
duced activity resulting from replacing the carboxylic
acid with a dimethylamino group (7i). Like the ester
(7e, Table 1), the methyl amide (7j) failed to demon-
strate an advantage. This was also noted with the tetra-
Keywords: Calmodulin; Kinase; Inhibitors; Pyrimidines.
*
Corresponding author. Tel.: +1 650 704 3051; e-mail: del345@gmail.com
0960-894X/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2008.02.056

B. Mavunkel et al. / Bioorg. Med. Chem. Lett. 18 (2008) 2404–2408
2405
Cl
Cl
(HO)2B
N
N
N
N
+
R
(a)
R
Cl
1
2
5
3
Cl
HN
N
H2N
+
N
(b,c)
O2N
H2N
4
6
N
N
R
3
+
6
N
N
H
(b)
N
H
7
Scheme 1. Reagents and conditions: (a) Pd[P(Ph)3]4, Na2CO3, CH3CN, 90 °C, 4 h, 50–60%; (b) Pd(OAc)2, Cs2CO3, BINAP, dioxane, 90 °C, 20 h,
65–80%; (c) H2, 10%Pd/C, EtOH, 40 psi, 24–48 h, 95%.
Table 1. SAR of initial pyrimidine-based leads
structure of this series. This was accomplished by sys-
tematically carving away the structural components of
N
N
7f and testing the activity of the resulting analogs. The
compounds of this new SAR were prepared according
to Scheme 2. As illustrated, dichloropyrimidine was cou-
pled with a boronic acid under Suzuki conditions. Sub-
sequent Buchwald coupling of anilines with the resulting
chloropyrimidine followed by the base hydrolysis of the
methyl esters gave target compounds 11. For compound
11a, the aniline utilized was prepared by coupling
p-methoxybenzylamine with compound 4 followed
by the reduction of the nitro group to an amine (Scheme
1). Subsequent cleavage of the p-methoxybenzyl
group was achieved by treating the ester analog with tri-
fluoroacetic acid. Compounds 11b, 11c, and 11d were
prepared using commercially available anilines and
aminopyridines.
X
N
H
N
N
H
Compound
X
IC50 (lM)
F
7a
1.28 (n = 1)
7b
3.14 (n = 1)
7c
7d
7e
7f
0.92 (n = 1)
As shown in Table 3, elimination of the methylbenzyl
group resulted in a 3-fold increase in activity against
CaMKIId. However, when the resulting amine was
removed, the activity trended lower. The true break-
through came when the pyridyl group was replaced
with a phenyl. This modification resulted in a 3-fold
increase in activity over our previous lead 7f. Fur-
thermore, by utilizing our homology model, 11c was
predicted to be capable of entering the ATP site
and making a dual hydrogen bond interaction be-
tween the hinge region and the aminopyrimidine unit.
Additionally, modeling also suggested that the car-
boxylic acid was now forming a new salt bridge with
Lys43 within the ATP site while the phenyl group
was filling a hydrophobic pocket. Additional enhance-
ment of activity was noted by introducing a chloro
group to the meta position of the phenyl ring in or-
der to maximize the occupation of the hydrophobic
pocket. This resulted in an additional benefit to
activity.
2.87 (n = 1)
Cl
1.69 0.47 (n = 3)
0.21 0.01 (n = 2)
CO2CH3
CO2H
zole (7k) introduced as a carboxylic acid isostere. In
fact, the only additional analog demonstrating sub-
micromolar activity was the hydroxamic acid (7l). All re-
sults can be explained by recognizing that non-acidic
groups or acidic groups in sub-optimal positions are
not able to optimally interact with Lys22 at the entrance
to the ATP site as predicted by our homology model.
With compound 7f remaining as our lead analog, our
next exercise was to determine the minimum active

2406
B. Mavunkel et al. / Bioorg. Med. Chem. Lett. 18 (2008) 2404–2408
Table 2. SAR surrounding the carboxylic acid of 7f
Table 3. SAR resulting from structural minimization of 7f
N
N
X
N
N
N
N
X
N
H
R
N
H
H
CO2H
Compound
X
IC50 (lM)
Compound
R
X
N
IC50 (lM)
7f
0.21 0.01 (n = 2)
CO2H
CO2H
7f
0.21 0.01 (n = 2)
N
H
7g
7h
7i
12.99 (n = 1)
11a
N
0.066 (n = 1)
H2N
H
11b
11c
11d
N
0.39 0.10 (n = 2)
0.064 0.031 (n = 2)
0.042 0.007 (n = 2)
CO2H
3.42 0.33 (n = 2)
2.22 (n = 1)
H
CH
CH
Cl
N(CH3)2
Table 4. SAR surrounding the carboxylic acid of 11c
N
7j
0.72 0.39 (n = 2)
2.26 0.21 (n = 2)
0.17 (n = 1)
CONHCH3
H
X
N
H
N
N
N
7k
7l
Compound
X
Enzyme
Cell-based
N
N
IC50 (lM) %I at 5 lM
CONHOH
11c
0.064 0.031 0
(n = 2)
CO2H
CO2H
Having identified 11c as the critical active structure, the
SAR applied to the carboxylic acid region of 7f was ap-
plied to our new lead. The compounds of this SAR were
prepared using chemistry illustrated or described in this
report and the data are summarized in Table 4. As illus-
trated and supported by the data in Table 2, the pres-
ence and positioning of an acidic group was essential
for activity. In fact, where a carboxylic acid and a
hydroxamic acid were essentially interchangeable in Ta-
ble 2 SAR, Table 4 shows this modification to result in a
5-fold loss in activity when applied to 11c.
11e
11f
11g
11h
11i
>20 (n = 1)
>20 (n = 1)
>20 (n = 1)
0
CO2CH3
CONHCH3
0.29 (n = 1) 90
CONHOH
H
N
N
Advancing from direct binding assays, CaMKIId-medi-
ated phosphorylation of vimentin was studied in whole
cell systems and modulation of this process was mea-
0.77 (n = 1)
0
N
N
Cl
Cl
(HO)2B
N
N
N
N
+
(a)
CO2Me
Cl
1
8
9
CO2Me
CO2H
R
N
X
X
N
9
+
N
H
R
(b,c)
H2N
10
11
Scheme 2. Reagents and conditions: (a) Pd[P(Ph)3]4, Na2CO3, CH3CN, 90 °C, 4 h, 50–60%; (b) Pd(OAc)2, Cs2CO3, BINAP, dioxane, 90 °C, 20 h,
75–85%; (c) NaOH, MeOH, H2O, 95%.

B. Mavunkel et al. / Bioorg. Med. Chem. Lett. 18 (2008) 2404–2408
Table 5. SAR surrounding amine-substituted aryl groups
2407
Glu100
O
N
N
H
H
Hydrophobic
Pocket
N
H
N
X
Cl
N
O
O
Br
Compound
X
Enzyme
IC50 (lM)
Cell-based
IC50 (lM)
11d
0.042 0.007 0% I at
(n = 2)
N
N
CO2H
H
5 lM
Hydrophobic
Pocket
H
H
NH2
N
15a
15b
15c
15d
15e
0.66 0.04
(n = 2)
O
O
Glu106
H
0.009
(n = 1)
0.32
(n = 1)
Figure 1. Amine-based maleimides interact with Glu106.
N
N
N
NH
sured as a function of inhibitor concentration.9 As
shown in Table 4, there was initially no significant cellu-
lar activity noted for our inhibitors. Recognizing the va-
lue of tetrazoles and hydroxamic acids as carboxylic
acid isosteres, these groups were studied for their poten-
tial to improve cell penetrance. As shown, only the
hydroxamic acid exhibited interesting activity. There-
fore, with a narrow SAR for aryl functionality and the
poor cellular activity of acid-based inhibitors, we sought
new alternatives.
0.24 0.05
(n = 2)
O
0.93 0.04
(n = 2)
H
N
NH2 0.012
(n = 1)
0.11
(n = 1)
Turning to work presented in previous publications,1–3
we recognized that our homology model indicated that
the amine tether of our maleimide inhibitors interacted
with Glu106 (Fig. 1). Furthermore, our model placed
Glu106 in close proximity to Lys43. Using this informa-
tion, coupled with computational docking studies in the
homology model ATP site as well as solvation energy
calculations,10 compounds were designed with acidic
groups at the aryl 4-position replaced with basic groups
at the aryl 3-position. Preparation of these analogs,
illustrated in Scheme 3 involved Suzuki couplings of
boronic acids with dichloropyrimidine. Subsequent
Buchwald coupling with 3-chloroaniline gave the target
compounds 15. For analogs 15a, 15b, and 15e, the
sequence was executed using boronic acids with Boc-
protected amines. Formation and cleavage of the
Boc-protecting group was achieved under standard con-
ditions. Furthermore, the boronic acid required for 15e
was prepared by first treating 3-bromoaniline with Boc-
2-bromoethylamine followed by conversion of the bro-
mide to the required pinacolatoboronic ester.
Assay results of the amine-based compounds are shown
in Table 5 and compared to the parent carboxylic acid
analog, 11d. As illustrated, replacing the 4-carboxylic
acid with a 3-amino group resulted in a significant drop
in potency. This confirmed the modeling prediction that
a simple aniline could not reach Glu106. However, when
the amine was replaced with a piperazine, a significant
Cl
Cl
(HO)2B
R
N
N
N
N
+
R
(a)
Cl
1
12
13
Cl
N
R
N
N
13
+
Cl
H
(b)
H2N
15
14
Scheme 3. Reagents and conditions: (a) Pd[P(Ph)3]4, Na2CO3, CH3CN, 90 °C, 4 h, 50–60%; (b) Pd(OAc)2, Cs2CO3, BINAP, dioxane, 90 °C, 20 h,
75–85%.

2408
B. Mavunkel et al. / Bioorg. Med. Chem. Lett. 18 (2008) 2404–2408
5. Wolfe, J. P.; Buchwald, S. L. J. Org. Chem. 2000, 65, 1144.
6. Assays were performed with inhibitor or suitable control
solvent added 10 ll per well in a 96-well microtiter plate
(Corning, NY). CaMKIId was diluted in enzyme buffer
(50 mM PIPES, pH 7, 0.2 mg/ml BSA, 1 mM DTT) and
added 10 ll per well. Reactions were initiated with 30 ll
reaction buffer (62.5 mM PIPES, pH 7, 0.25 mg/ml BSA,
33.3 mM MgCl2, 83 lM ATP, 0.4 mM CaCl2, 8.3 lg/ml
calmodulin, 25 lM [His 5] Autocamtide-2, 120 nM [g-
33P]ATP) and incubated at RT for 3 min. Reactions were
terminated by transferring 25 ll to a UNIFILTER 96-well
P81microplate (Whatman, UK), pre-wet with 15 ll 1%
phosphoric acid. After 10 min, the plate was washed three
times with 1% phosphoric acid and one time with 95%
improvement in activity was noted. In fact, compound
15b demonstrated that we could overcome cell pene-
trance issues with sub-micromolar activity in our cell-
based assay. Furthermore, we validated the importance
of the extended basic group using morpholine or piper-
idine as piperazine variants. In both cases, enzyme activ-
ity was compromised. Finally, an unexpected benefit
resulted from relaxing the positioning of the extended
amine by replacing the piperazine with ethylene dia-
mine. While this compound showed no improvement
in enzyme activity, a 3-fold increase in cellular activity
was noted.
ethanol on
a
BiomekFX (Beckman–Coulter, CA)
In summary, novel pyrimidine-based inhibitors of CaM-
KIId were prepared. Homology model predictions were
used to advance the SAR from micromolar inhibitors to
low nanomolar inhibitors. Activity in cell-based assays
was improved by introducing basic groups at the phenyl
3-position and generating a predicted salt bridge with
Glu106. Our most potent inhibitor possessed a flexible
tether to the amine and exhibited an IC50 of 12 nM
against the isolated enzyme.
equipped with a vacuum manifold. Plates were dried for
approximately 60 min, scintillant was added to the wells,
and the plates were read on a TopCount NXT Microplate
Scintillation and Luminescence Counter (Perkin-Elmer,
MA).
7. An homology model of autoinhibited CaMKIId was built
based on the crystal structure 1A06 of autoinhibited rat
CaMKI. Because rat CaMKI shows high sequence
homology with CaMKIId, this model was used to study
inhibitors that were not ATP competitive. Due to the
lack of availability of a crystal structure of activated
CaMKIId, homology models were built based on crystal
structures 1CDK, 1PHK, and 1KOB. From these homol-
ogy models, we selected the one best explained the SAR.
The accuracy of this model was validated using point
mutation studies.
References and notes
1. Levy, D. E.; Wang, D-X.; Lu, Q.; Chen, Z.; Perumattam,
J.; Xu, Y-J.; Liclican, A.; Higaki, J.; Dong, H.; Laney, M.;
Mavunkel, B.; Dugar, S. Bioorg. Med. Chem. Lett. 2008,
18, 2390.
2. Levy, D. E.; Wang, D-X.; Lu, Q.; Chen, Z.; Perumattam,
J.; Xu, Y-J.; Higaki, J.; Dong, H.; Liclican, A.; Laney, M.;
Mavunkel, B.; Dugar, S. Bioorg. Med. Chem. Lett. 2008,
18, 2395.
3. Lu, Q.; Chen, Z.; Perumattam, J.; Wang, D-X.; Liang,
W.; Xu, Y-J.; Do, S.; Bonaga, L.; Higaki, J.; Dong, H.;
Liclican, A.; Sideris, S.; Laney, M.; Dugar, S.; Mavun-
kel, B.; Levy, D. E. Bioorg. Med. Chem. Lett. 2008, 18,
2399.
4. Zapf, A. In Transition Metals for Organic Synthesis:
Building Blocks and Fine Chemicals, 2nd ed.; Beller, M.,
Bolm, C., Eds.; Wiley-VCH: Weinheim, 2004; p 1344.
8. Amantini, D.; Beleggia, R.; Fringuelli, F.; Pizzo, F.;
Vaccaro, L. J. Org. Chem. 2004, 69, 2896.
9. HL-2 cells were pre-incubated with increasing concentra-
tions of an inhibitor for 1 h at 37 °C. CaMKII was
activated by cell treatment with 1 rmuM ionomycin for
15 min at 37 °C. Cells were then lysed in ice-cold M-PER
cell lysis buffer (Pierce) and frozen at À80 °C. Inhibitor
IC50 values were determined by an ELISA assay of
cell lysate measuring intracellular vimentin phosphory-
lated at Ser82 (Delling, Catalano, Wang, Higgins,
unpublished).
10. Solvation energy calculations were generated using ZAP
(Openeye Scientific Software).