Structure-Based Design of Selective Noncovalent CDK12 Inhibitors

Article abstract of DOI:10.1002/cmdc.201700695

Cyclin-dependent kinase (CDK) 12 knockdown via siRNA decreases the transcription of DNA-damage-response genes and sensitizes BRCA wild-type cells to poly(ADP-ribose) polymerase (PARP) inhibition. To recapitulate this effect with a small molecule, we sought a potent, selective CDK12 inhibitor. Crystal structures and modeling informed hybridization between dinaciclib and SR-3029, resulting in lead compound 5 [(S)-2-(1-(6-(((6,7-difluoro-1H-benzo[d]imidazol-2-yl)methyl)amino)-9-ethyl-9H-purin-2-yl)piperidin-2-yl)ethan-1-ol]. Further structure-guided optimization delivered a series of selective CDK12 inhibitors, including compound 7 [(S)-2-(1-(6-(((6,7-difluoro-1H-benzo[d]imidazol-2-yl)methyl)amino)-9-isopropyl-9H-purin-2-yl)piperidin-2-yl)ethan-1-ol]. Profiling of this compound across CDK9, 7, 2, and 1 at high ATP concentration, single-point kinase panel screening against 352 targets at 0.1 μm, and proteomics via kinase affinity matrix technology demonstrated the selectivity. This series of compounds inhibits phosphorylation of Ser2 on the C-terminal repeat domain of RNA polymerase II, consistent with CDK12 inhibition. These selective compounds were also acutely toxic to OV90 as well as THP1 cells.

Full text of DOI:10.1002/cmdc.201700695

Accepted Article
Title: Structure Based Design of Selective Non-covalent CDK12
Inhibitors
Authors: Jeffrey Johannes, Christopher R. Denz, Nancy Su, Allan Wu,
Anna C. Impastato, Scott Mlynarski, Jeffrey G. Varnes, D.
Bryan Prince, Justin Cidado, Ning Gao, Malcolm Haddrick,
Natalie H. Jones, Shaobin Li, Xiuwei Li, Yang Liu, Toan
B. Nguyen, Nichole O'Connell, Emma Rivers, Daniel W.
Robbins, Ronald Tomlinson, Tieguang Yao, Andrew D.
Ferguson, Michelle L. Lamb, John I. Manchester, and Sylvie
Guichard
This manuscript has been accepted after peer review and appears as an
Accepted Article online prior to editing, proofing, and formal publication
of the final Version of Record (VoR). This work is currently citable by
using the Digital Object Identifier (DOI) given below. The VoR will be
published online in Early View as soon as possible and may be different
to this Accepted Article as a result of editing. Readers should obtain
the VoR from the journal website shown below when it is published
to ensure accuracy of information. The authors are responsible for the
content of this Accepted Article.
To be cited as: ChemMedChem 10.1002/cmdc.201700695
Link to VoR: http://dx.doi.org/10.1002/cmdc.201700695
A Journal of
www.chemmedchem.org

10.1002/cmdc.201700695
ChemMedChem
COMMUNICATION
Structure Based Design of Selective Non-covalent CDK12
Inhibitors
Jeffrey W. Johannes,* Christopher R. Denz, Nancy Su, Allan Wu, Anna C. Impastato, Scott Mlynarski,
Jeffrey G. Varnes, D. Bryan Prince, Justin Cidado, Ning Gao, Malcolm Haddrick, Natalie H. Jones,
Shaobin Li, Xiuwei Li, Yang Liu, Toan B. Nguyen, Nichole O'Connell, Emma Rivers, Daniel W.
Robbins, Ronald Tomlinson, Tieguang Yao, Xiahui Zhu, Andrew D. Ferguson, Michelle L. Lamb, John
I. Manchester, Sylvie Guichard
Abstract: CDK12 knockdown via siRNA reduces transcription of
DNA damage response genes and sensitizes BRCA wild type cells to
PARP inhibition. To recapitulate this effect with a small molecule, we
sought a potent, selective CDK12 inhibitor. Crystal structures and
modeling informed hybridization between dinaciclib and SR-3029,
resulting in lead compound 5. Further structure guided optimization
delivered a series of selective CDK12 inhibitors, including compound
7. Profiling of this compound across CDK9, 7, 2 and 1 at high ATP
concentration, single point kinase panel screening against 352 targets
at 0.1 µM, and proteomics via kinase affinity matrix technology
demonstrated the selectivity. This series of compounds inhibit
phosphorylation of Ser2 on the C-terminal repeat domain of RNA pol
II, consistent with CDK12 inhibition. These selective compounds were
also acutely toxic to OV90 as well as THP1 cells.
splicing. It has been shown that siRNA-mediated knockdown of
CDK12 reduced expression of a small number of genes, with a
disproportional effect on long genes and genes with a high
number of exons. Systematic analysis of these genes with
reduced expression found that many of them were associated
with DNA damage repair (DDR) networks, with BRCA1 as a key
gene showing reduced expression.3 In a related work, mutations
to CDK124 found in ovarian cancer samples that impaired kinase
activity had a detrimental effect on homologous repair, and tended
not to co-occur with BRCA1 mutations.5
Table 1. Published pan-CDK inhibitor dinaciclib16 and dual CK1δ/CK1ε inhibitor
SR-3029.17
Cyclin-dependent kinase (CDK) 12 forms a catalytically active
enzyme complex upon association with cyclin K (CycK).1 The
CycK/CDK12 complex phosphorylates the C-terminal domain
(CTD) heptapeptide repeat (Tyr-Ser-Pro-Thr-Ser-Pro-Ser) of
RNA pol II, specifically recognizing pSer7 and phosphorylating
Ser2 and Ser5 of the CTD.2 The CTD of RNA pol II recruits
proteins for RNA processing, including the spliceosome for RNA
dinaciclib ( )
SR-3029 ( )
1
2
1
2
[*]
Dr. J. W. Johannes, Dr. J. Cidado, C. Denz, Dr. S. Guichard, Dr. M.
L. Lamb, Dr. J. I. Manchester, Dr. S. Mlynarski, Dr. D. W. Robbins,
J. G. Varnes
Oncology, IMED Biotech Unit
AstraZeneca, Boston, MA
E-mail: jeffrey.johannes@astrazeneca.com
Dr. A. D. Ferguson, N. Gao, A. C. Impastato, N. H. Jones, Dr. T. B.
Nguyen, Dr. N. O'Connell, D. Bryan Prince, N. Su, R. Tomlinson, A.
Wu, X. Zhu,
Discovery Sciences, IMED Biotech Unit
AstraZeneca, Boston, MA
Dr. E. Rivers
Discovery Sciences, IMED Biotech Unit
AstraZeneca Pharmaceuticals LP
Unit 310 Darwin Building, Cambridge CB4 0WG (UK)
Dr. M. Haddrick
Discovery Sciences, IMED Biotech Unit
AstraZeneca Pharmaceuticals LP
Alderley Park, Macclesfield SK10 4TG
Dr. T. Yao, Y. Liu, S. Li, X. Li
Pharmaron Beijing Co., Ltd.
CDK12 Km* ATP IC50 (µM)
CDK12 high† ATP IC50 (µM)
CDK9 Km ATP IC50 (µM)
CDK7 Km ATP IC50 (µM)
CDK2 Km ATP IC50 (µM)
CDK1 Km ATP IC50 (µM)
0.020
0.050
<0.003
0.042
<0.003
0.012
0.086
>30
11.3
>16.5
>30
>30
[b]
[c]
[d]
[e]
* For Km assays [ATP] ranges from 10 to 50 µM (see SI for specific values
used). † High assays [ATP] = 5000 µM.
Interest in targeting DNA damage repair mechanisms in
tumors has grown following the approval of the PARP inhibitor
olaparib.6 PARP inhibitors are effective in tumors that harbour
mutations to BRCA1, and are hence deficient in homologous
repair (HR). Cells that are deficient in HR have difficulty in
correctly repairing double strand breaks. PARP inhibitors which
trap PARP1 on DNA cause excess production of double strand
breaks, which is toxic to HR deficient cells.7 The link between
6 Taihe Road BDA, Beijing 100176 (P. R. China)
Supporting information for this article is given via a link at the end of
the document.
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700695
ChemMedChem
COMMUNICATION
CDK12 and BRCA1 leads to the hypothesis that synergy may be
observed between CDK12 knockdown and PARP inhibition.
Indeed, Bajrami et al. found that knockdown of CDK12 in OV90
ovarian cancer cells rendered the cells susceptible to the PARP
inhibitor olaparib.8
CDK7 as transcriptional CDKs, and CDK2 and 1 as representative
and well known examples of the cell cycle CDKs.
In order to screen and optimize for a selective non-covalent
CDK12 inhibitor, we developed a battery of CDK enzyme assays
using Caliper technology that could be run at low (Km) and high
(5 mM) ATP concentrations for CDK12, 9, 7, 2, and 1. We felt
that this panel covered both transcriptional and cell cycle CDKs
sufficiently for screening. Our goal was to discover a compound
with a CDK12 IC50 <500 nM at high ATP, with 10-fold high ATP
selectivity to the other CDKs in the panel. Utilizing 5 mM ATP
mimics the cellular concentration of ATP and provides an initial
picture of kinase selectivity in a cellular context.
These results suggest that a small molecule CDK12 inhibitor
used in combination with a PARP inhibitor such as olaparib could
be efficacious in BRCA1 wild type cells, and extend the utility of
PARP inhibition beyond a BRCA1 mutant background.9 To test
this hypothesis and validate CDK12 as an oncology target we set
out to discover a highly selective CDK12 small molecule inhibitor
to ensure that the observed effects were due to the target of
interest. Recently, Zhang et al. reported the discovery of THZ531,
a small molecule covalent inhibitor of CDK12 that contains an
acrylamide group that reacts with Cys1039 in CDK12.10 THZ531
reduced the expression of DDR associated genes in a manner
similar to siRNA CDK12 knockdown experiments.
We initially screened a number of CDK family and other
related Ser/Thr kinase inhibitors across this panel. One of our
initial hits was dinaciclib (1, Table 1) a known CDK inhibitor in
clinical trials in cancer.17 Dinaciclib was very potent in our CDK12
assays at both Km (20 nM) and high ATP (50 nM). However,
dinaciclib is a pan CDK inhibitor, showing potent inhibition of
CDK9, 7, 2, and 1 at Km ATP. In the same screen, we identified
SR-3029 (2, Table 1), a published CK1δ and CK1ε inhibitor,18 as
having some modest CDK12 inhibition at Km (86 nM).
Unfortunately, this compound didn’t show any inhibition of CDK12
at high ATP. Strikingly, SR-3029 did not inhibit CDK9 ,7, 2, nor 1
at Km ATP, demonstrating that selective CDK12 inhibition is
possible within the CDK family.
With these two inhibitors in hand, one very potent but non-
selective and the other very selective but lacking potency, we
hypothesized that these two compounds adopt a common binding
mode to CDK proteins. We docked SR-3029 to a published
structure of CDK12 (PDB 4CXA)19 and aligned and superposed
this structure with PDB 4KD1, the crystallographically observed
binding mode of dinaciclib bound to CDK2 (Figure 1).20 Indeed
the two compounds’ heterocyclic hinge-binding cores of both
compounds overlap, in a way which which perfectly aligns the
peripheral groups. This model suggested that hybridization of the
two compounds was possible.
At first, it was unclear which portions of dinaciclib conferred its
exquisite potency and which moieties on SR-3029 gave high
selectivity for CDK12. From reported SAR on the chemical series
that culminated in dinaciclib, we knew that the hydroxyethyl
piperidine group occupying the ribose pocket was very important
for binding, but the effects of the core heterocycle or other
substituents were unknown.
Figure 1. Overlay of dinaciclib in blue bound to CDK2 in orange (PDB 4KD1)
with SR-3029 in yellow docked to CDK12 structure PDB 4CXA (CDK12 protein
not shown for clarity).
CDK12 is a member of the cyclin dependent kinase (CDK)
family of enzymes, which have been the subject of intense efforts
across academia and the pharmaceutical industry11 as oncology
agents.12 , 13 The family contains 20 members which can be
divided into the cell cycle kinases (including CDK1,2,3,4, and 6)
and the transcriptional CDKs (including CDK12,13, 9, 8, and 7).14
CDK12 is very closely related and highly homologous to CDK13,
and the ATP binding sites of these proteins are essentially
identical,15 however, knockdown of CDK13 affects different genes
compared to CDK12, having effects on growth signalling
pathways.16 From the outset, we concluded that selectivity
between CDK12 and CDK13 would be extremely difficult to obtain,
so we initially focused on obtaining selectivity over CDK9 and
The first hybrid that we made was compound 3 (Table 2),
which swapped the pyridine N-oxide solvent tail of dinaciclib for
the difluorobenzimidazole solvent tail of SR-3029. This hybrid
maintained similar potency compared to dinaciclib with improved
CDK7 and CDK1 selectivity. Keeping this bezimidazole solvent
tail fixed, we next explored the effects of other positions on
potency and selectivity. Our first example of this is illustrated by
compound 4 which compares the fluorophenyl group found in
SR30-29 with the ethyl group of dinaciclib. Although compound 4
is highly selective, it has lost CDK potency across the panel,
including potency at our primary target CDK12. Compound 5 is a
matched pair comparing the pyrazolopyrimidine core of dinaciclib
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700695
ChemMedChem
COMMUNICATION
with the purine core of SR-3029. The purine core further
improved selectivity against CDK9, 7 and 2 with only a marginal
loss of CDK12 activity. CDK enzyme potency and selectivity is
significantly affected by the hydrophobic group occupying the
selectivity pocket. Compound 5, with a CDK12 high ATP potency
of 103 nM, was quite selective against CDK7 and CDK1, had an
acceptable margin to CDK9, but only a 3-fold margin to CDK2 at
high ATP concentration.
would improve potency and possibly confer additional selectivity
for CDK12.
To this end, we synthesized numerous analogs exploring
substitutions at the N-9 position (varying R1 in Table 3). The
CDK12 high ATP potency and CDK panel selectivity at high ATP
was extremely sensitive to changes in the steric size and nature
of the hydrophobic group at R1.
For example, simple
monofluorinaton of the ethyl group in compound 5 to give
fluoroethyl analog 6 resulted in an almost 4-fold loss in CDK12
high ATP potency while having little impact on selectivity for CDK9
and CDK2 (still approximately 10-fold and 3-fold respectively).
R1 = 3-F-phenyl ( )
3
4
R1 = Et ( )
5
Table 2. Hybridization between dinaciclib and SR-3029 results in potent and
selective lead compound 5..
3
4
5
CDK12 Km ATP IC50 (µM)
CDK12 high ATP IC50 (µM)
CDK9 high ATP IC50 (µM)
CDK7 high ATP IC50 (µM)
CDK2 high ATP IC50 (µM)
CDK1 high ATP IC50 (µM)
Aq. Sol. pH 7.4 (µM)
0.022
0.047
0.240
21.2
0.040
3.41
9
0.065
3.62
>30.0
>30
>30
>30
3
0.014
0.103
1.09
>30
0.321
28.6
39
Figure 2. Crystal structure of compound 5 in yellow bound to CDK12 in blue
(PDB 6B3E).
Introduction of branching at R1 with an isopropyl group
improved the margin to CDK9 to approximately 20-fold at high
ATP. Importantly, this branching also improved the gap between
CDK12 and CDK2 to 5-fold. Overall, compound 7 is a 491 nM
inhibitor of CDK12 at high ATP with improved selectivity to CDK9
and CDK2 compared to ethyl analog 5. While the potency of
compound 7 is reduced approximately 5-fold compared to 5, the
added selectivity is desired to support the use of the compound
as an in vitro tool to validate small molecule inhibition of CDK12
as an oncology target. Difluoroethyl analog 8 showed a further
reduction in CDK12 high ATP potency, highlighting the sensitivity
to sterics at this position of the molecule. Compound 8 had a
similar margin to CDK9 and CDK2 compared to compound 7.
LogD pH 7.4
4.1
>3.3
4.1
To understand the subtleties of how the compound binds and
to confirm our hypothesized binding mode, we obtained a crystal
structure of compound 5 bound to CDK12 in complex with cyclin
K (Figure 2). Compound 5 binds to the backbone NH of hinge
residue Met816 via the purine N-7, and the backbone carbonyl of
Met816 via the NH at the purine 6-positon.
The
difluorobenzimidazole occupies the solvent channel, with the
imidazole moiety engaging in hydrogen bonds to Tyr815 and
Asp819, while also interacting with the Trp1036 of the C-terminal
helix of CDK12. We posit that the benzimidazole drives selectivity,
since Tyr815 is a Phe in CDK9, 7, 2, and 1 and the other CDKs
completely lack a C-terminal helix containing the Trp. The
hydroxyethyl piperidine attached to the purine core at C-2
occupies the ribose pocket. The ethyl group attached to purine
core N-9 occupies the selectivity pocket, forming a hydrophobic
interaction with Phe813. Examining the surface of the binding
pocket (hydrophobic patches in green and hydrophilic patches in
purple) we noticed that there was a considerable amount of empty
space surrounding the ethyl group at N-9. We thought that
expanding the steric size of the group at N-9 to fill this space
We selected 7 for further profiling to understand the selectivity
of this compound across the kinome. The percent inhibition
results when tested at a single concentration of 0.1 µM across a
mixed panel of 352 ThermoFisher assays testing for kinase
binding or kinase enzyme inhibition at various ATP levels are
presented in Figure 3. Compound 7 is a selective inhibitor across
the kinome. The top his in the panel are presented in the inset of
Figure 3. Outside of the CDK family, inhibited kinases include
IRAK1 (32%), PI4KB (29%), and TAOK1 (27%) at 0.1 µM
compound concentration.21 We also tested the kinase selectivity
of compound 7 via affinity proteomics using a kinase affinity matrix.
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700695
ChemMedChem
COMMUNICATION
In this experiment, CDK12 was the most significantly competed
kinase off the solid matrix by compound 7, followed by CDK13
and CDK9 to a lesser extent.22 Compound 7 was also relatively
clean against a diverse panel of human secondary pharmacology
targets (see the supporting information table S5).
We next turned our attention to testing these compounds in
our mechanism of action assay, which detects inhibition of
phosphorylation of serine 2 (pSer2) on the C-terminal repeat
domain of RNA pol II (Table 4). Selective compounds 7 and 8
inhibited Ser2 phosphorylation in MCF7 cells. Other potent but
less selective CDK12 inhibitors such as dinaciclib (1) and
compounds 3, 5, and 6 also inhibited pSer2. Importantly, a
compound with weaker inhibition of CDK12 at high ATP, such as
4, inhibited pSer2 to a lesser extent. Compound 2 is potent for
neither CDK12 nor CDK9 at high ATP but still inhibits pSer2 with
an IC50 274 nM. Compound 2 was originally reported as a CK1δ
and CK1ε inhibitor (ref. 18) and is highly selective for these
kinases (see supporting information table S3 and S4). CK1
isoforms have been reported to associate with the CTD of RNA
pol II,23 and is possible that inhibition of CK1δ/ε kinase activity can
inhibit phosphorylation of pSer2. Overall there is a good
correlation between inhibition of CDK12 at high ATP and pSer2
inhibition.24
Figure 3. Percent inhibition results when tested at a single concentration of 0.1
µM across a mixed panel of 352 ThermoFisher assays testing for kinase binding
or kinase enzyme inhibition at various ATP levels of compound 7, with percent
inhibition data for the top 10 hits listed in the inset table.
Table 4. CDK12 and CDK8 high ATP enzyme inhibition data compared to the
cell assay results for inhibition of Ser2 phosphorylation in MCF7, 48 h growth
inhibition in OV90 cells, and THP1 cell cytotoxicity.
Cmpd
CDK12
high ATP
IC50 (µM)
CDK9
high ATP
IC50 (µM)
MCF7
pSer2 IC50
(µM)
OV90 GI50
48 h (µM)
THP-1
IC50 (µM)
3
1
0.047
0.050
0.103
0.389
0.491
0.845
3.62
0.240
<0.010
5.14
4.94
9.09
19.6
>30
0.012
<0.005
0.012
0.032
0.098
0.094
0.336
0.274
>3
0.104
0.021
0.099
0.044
0.204
0.067
0.373
0.321
0.067
0.106
0.251
1.54
5
6
7
R1 = CH CH F ( )
8
0.830
2.58
6
2
2
R1 = iPr ( )
7
4
R1 = CH CHF ( )
8
2
2
2
>30
>30
2.17
Table 3 Optimization of the purine N-9 substituent results in potent and selective
lead compound 7
9
>30
>250
0.093
6
7
8
10
0.213
>30
0.039
0.102
CDK12 Km ATP IC50 (µM)
CDK12 high ATP IC50 (µM)
CDK9 high ATP IC50 (µM)
CDK7 high ATP IC50 (µM)
CDK2 high ATP IC50 (µM)
CDK1 high ATP IC50 (µM)
Aq. Sol. pH 7.4 (µM)
0.017
0.389
4.94
>30
1.13
>30
53
0.031
0.491
9.09
>30
2.68
>30
18
0.023
0.845
19.6
>30
4.29
>30
50
‡ Ranked in order of CDK12 high ATP potency.
With selective compounds that can modulate pSer2 in cells,
we next sought to see if these compounds were suitable for
testing the hypothesis that a CDK12 enzyme inhibitor could
synergize with a PARP inhibitor (such as olaparib) in a 14-day
colony formation assay in OV90 cells. Prior to committing to a
long-term assay, we tested this series of CDK12 inhibitors in a 48
h growth inhibition assay to check for acute toxicity to OV90 cells
(Table 4). Interestingly, compounds 5, 6, 7, and 8 were toxic to
OV90 cells as single agents, with 48 h GI50’s ranging from 21 to
373 nM. The growth inhibition IC50’s correlated well with CDK12
inhibition and pSer2 inhibition, suggesting that CDK12 is driving
this growth inhibitory effect. As a negative control, we tested
ribociclib (9), a selective CDK4/6 inhibitor.25 Compound 9 lacks
CDK12 activity, is inactive in our pSer2 assay up to the highest
LogD pH 7.4
3.7
4.6
>4.0
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700695
ChemMedChem
COMMUNICATION
concentration tested, and is not toxic to THP1 cells (IC50 >250
µM). As a positive control, we included compound 10 (THZ531)10
which is a potent inhibitor of CDK12 (213 nM) and lacks CDK9
inhibition. In our hands, THZ531 inhibited pSer2 at 39 nM,
inhibited the growth of OV90 cells at 102 nM and had 93 nM IC50
against THP1 cells.
may keep cyclin-K sequestered via a CDK12:cyclinK interaction,
while knockdown of CDK12 may create additional free cyclin-K.
In conclusion, we started with dinaciclib as a highly potent
CDK12 inhibitor and SR-3029 as a weak but highly selective
CDK12 inhibitor vs CDK9, 7, 2, and 1. Using a known crystal
structure of dinaciclib bound to CDK2 and computational docking
of SR-3029 to a CDK12 crystal structure, we hypothesized that
hybridization of these two compounds might result in a potent and
Since the OV90 cell line does not harbor a BRCA mutation,
we did not expect single agent efficacy with a CDK12 inhibitor a
priori. Alternatively, it has been reported that modulation of pSer2
itself can have cytotoxic effects, as previously reviewed for CDK
inhibitors.26 In this vein, we hypothesized that these compounds
might be toxic to blood cells, so we tested them in a 24 h
cytotoxicity assay against THP1 cells, an AML cell line used
routinely in discovery phase safety departments to obtain an early
read on unwanted cytotoxicity. Indeed, this series of CDK12
inhibitors were toxic to THP1 cells. For example, highly potent
CDK12 compound 5, which was approximately 50-fold selective
against CDK9, had a THP1 IC50 of 106 nM. More importantly,
there was a relatively small difference between effects on pSer2
(12 nM) and OV90 growth inhibition (99 nM) and THP1
cytotoxicity for this compound. The compound with the biggest
difference between pSer2 and OV90 effects and toxicity was
compound 8, a very selective, but slightly less active CDK12
inhibitor. Even in this case the margin was only approximately
10-fold.
selective
CDK12
inhibitor.
We
found
that
the
difluorobenzimidazole moiety was key for CDK family selectivity,
and that the purine core lent further selectivity for CDK12.
Optimization of the purine N-9 group led to isopropyl compound 7
with 5-fold CDK2 and near 20-fold CDK9 selectivity. Lead
compound 7 was profiled across a panel of 352 kinases at 0.1 µM
followed by concentration response testing in 30 of the top hits,
showing that 7 is a highly selective CDK12 inhibitor, and this result
was confirmed by affinity proteomics using a kinase affinity matrix.
This series of compounds showed potent inhibition of
phosphorylation of Ser2 on the CTD repeat domain of RNA Pol II
as well as growth inhibition of OV90 cells and acute cytotoxicity to
THP1 cells.
Experimental Section
See supporting information for details of compound synthesis,
crystallographic methods, computational methods, and enzyme and cell
assay protocols.
From these data we surmised that selective CDK12 inhibition
causes effects on the phosphorylation state of Ser2 of RNA pol II,
and that this mechanism can have single agent toxic effects on
cells. This hints at a possible therapeutic use for CDK12
inhibitors; however, the effects on pSer2 may carry the risk of a
very tight therapeutic margin in vivo. The pioneering work by
Johnson et al.9 have demonstrated that it is possible to show
synergy between dinaciclib (a CDK2, 9, and 12 inhibitor) with the
PARP inhibitor veliparib in vivo in certain patient-derived tumor
xenograft models. There may also be inherent differences
between the biology of siRNA mediated knockdown and small
molecule inhibition of CDK12. Small molecule inhibited CDK12
Acknowledgements
The authors would like to thank Ieuan Roberts for helpful advice
and coordination of kinase panel testing at ThermoFisher.
Keywords: CDK • kinase • transcription• selective • oncology
1 Kohoutek, J.; Blazek, D. Cell Div. 2012, 7, 12.
Kochupurakkal, B.; Sun, H.; Unitt, C.; Moreau, L. A.; Sarosiek, K. A.;
Scaltriti, M.; Juric, D.; Baselga, J.; Richardson, A. L.; Rodig, S. J.;
D'Andrea, A. D.; Balmaña, J.; Johnson, N.; Geyer, M.; Serra, V.; Lim, E.;
Shapiro, G. I. Cell Rep. 2016, 17, 2367-2381.
2 Bösken, C. A.; Farnung, L.; Hintermair, C.; Merzel Schachter, M.; Vogel-
Bachmayr, K.; Blazek, D.; Anand, K.; Fisher, R. P.; Eick, D.; Geyer, M.
Nat. Commun. 2014, 5, 3505.
3 Blazek, D.; Kohoutek, J.; Bartholomeeusen, K.; Johansen, E.; Hulinkova, P.;
Luo, Z.; Cimermancic, P.; Ule, J.; Peterlin, B. M. Genes Dev 2011, 25,
2158-2172.
10 Zhang, T.; Kwiatkowski, N.; Olson, C. M.; Dixon-Clarke, S. E.; Abraham, B.
J.; Greifenberg, A. K.; Ficarro, S. B.; Elkins, J. M.; Liang, Y.; Hannett, N.
M.; Manz, T.; Hao, M.; Bartkowiak, B.; Greenleaf, A. L.; Marto, J. A.;
Geyer, M.; Bullock, A. N.; Young, R. A.; Gray, N. S. Nat. Chem. Biol.
2016, 12, 876-884.
4 Ekumi, K. M.; Paculova, H.; Lenasi, T.; Pospichalova, V.; Bösken, C. A.;
Rybarikova, J.; Bryja, V.; Geyer, M.; Blazek, D.; Barboric, M. Nucleic
Acids Res 2015, 43, 2575-2589.
11 Malínková, V.; Vylíčil, J.; Kryštof, V. Expert Opin. Ther. Pat. 2015, 25, 953-
970.
5 Joshi, P. M.; Sutor, S. L.; Huntoon, C. J.; Karnitz, L. M. J. Biol. Chem. 2014,
289, 9247-9253.
12 Asghar, U.; Witkiewicz, A. K.; Turner, N. C.; Knudsen, E. S. Nat. Rev. Drug
Discov. 2015, 14, 130-146.
6 O'Connor, M. J. Mol. Cell 2015, 60, 547-560.
7 Pommier, Y.; O'Connor, M. J.; de Bono, J. Sci. Transl. Med. 2016, 8, 362ps17.
8 Bajrami, I.; Frankum, J. R.; Konde, A.; Miller, R. E.; Rehman, F. L.; Brough,
R.; Campbell, J.; Sims, D.; Rafiq, R.; Hooper, S.; Chen, L.; Kozarewa, I.;
Assiotis, I.; Fenwick, K.; Natrajan, R.; Lord, C. J.; Ashworth, A. Cancer
Res. 2014, 74, 287-297.
13 Malumbres, M.; Pevarello, P.; Barbacid, M.; Bischoff, J. R. Trends
Pharmacol. Sci. 2008, 29, 16-21.
14 Malumbres, M. Genome Biol. 2014, 15, 122.
15 See supporting information table S1 for a comparison of sequence and
binding sites between CDK12 and CDK13, 9, 7, 2, and 1.
9 Johnson, S. F.; Cruz, C.; Greifenberg, A. K.; Dust, S.; Stover, D. G.; Chi, D.;
Primack, B.; Cao, S.; Bernhardy, A. J.; Coulson, R.; Lazaro, J. B.;
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700695
ChemMedChem
COMMUNICATION
19 Dixon-Clarke, S. E.; Elkins, J. M.; Cheng, S. W.; Morin, G. B.; Bullock, A. N.
Sci. Rep. 2015, 5, 17122.
16 Greifenberg, A. K.; Hönig, D.; Pilarova, K.; Düster, R.; Bartholomeeusen, K.;
Bösken, C. A.; Anand, K.; Blazek, D.; Geyer, M. Cell Rep. 2016, 14, 320-
331.
20 Martin, M. P.; Olesen, S. H.; Georg, G. I.; Schönbrunn, E. ACS Chem. Biol.
2013, 8, 2360-2365.
21 See supporting information for a full table of percent inhibition values.
22 See supporting information for the full dataset.
17 Paruch, K.; Dwyer, M. P.; Alvarez, C.; Brown, C.; Chan, T. Y.; Doll, R. J.;
Keertikar, K.; Knutson, C.; McKittrick, B.; Rivera, J.; Rossman, R.; Tucker,
G.; Fischmann, T.; Hruza, A.; Madison, V.; Nomeir, A. A.; Wang, Y.;
Kirschmeier, P.; Lees, E.; Parry, D.; Sgambellone, N.; Seghezzi, W.;
Schultz, L.; Shanahan, F.; Wiswell, D.; Xu, X.; Zhou, Q.; James, R. A.;
Paradkar, V. M.; Park, H.; Rokosz, L. R.; Stauffer, T. M.; Guzi, T. J. ACS
Med. Chem. Lett. 2010, 1, 204-208.
23 Gross, S. D.; Anderson, R. A. Cell. Signal. 1998, 10, 699-711.
24 See supporting information for correlation plots between CDK12 high ATP
IC50, and MCF7 pSer2, OV90 GI50, and THP1 IC50.
25 Chen, P.; Lee, N. V.; Hu, W.; Xu, M.; Ferre, R. A.; Lam, H.; Bergqvist, S.;
Solowiej, J.; Diehl, W.; He, Y. A.; Yu, X.; Nagata, A.; VanArsdale, T.;
Murray, B. W. Mol. Cancer Ther. 2016, 15, 2273-2281.
18 Bibian, M.; Rahaim, R. J.; Choi, J. Y.; Noguchi, Y.; Schürer, S.; Chen, W.;
Nakanishi, S.; Licht, K.; Rosenberg, L. H.; Li, L.; Feng, Y.; Cameron, M.
D.; Duckett, D. R.; Cleveland, J. L.; Roush, W. R. Bioorg. Med. Chem.
Lett. 2013, 23, 4374-4380.
26 Wang, S.; Fischer, P. M. Trends Pharmacol. Sci. 2008, 29, 302-13.
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700695
ChemMedChem
COMMUNICATION
Entry for the Table of Contents (Please choose one layout)
Layout 1:
COMMUNICATION
Text for Table of Contents
Author(s), Corresponding Author(s)*
Page No. – Page No.
Title
((Insert TOC Graphic here))
Layout 2:
COMMUNICATION
Jeffrey W. Johannes,* Christopher R.
Denz, Nancy Su, Allan Wu, Anna C.
Impastato, Scott Mlynarski, Jeffrey G.
Varnes, D. Bryan Prince, Justin Cidado,
Ning Gao, Malcolm Haddrick, Natalie H.
Jones, Shaobin Li, Xiuwei Li, Yang Liu,
Toan B. Nguyen, Nichole O'Connell,
Emma Rivers, Daniel W. Robbins,
Ronald Tomlinson, Tieguang Yao,
Xiahui Zhu, Andrew D. Ferguson,
Michelle L. Lamb, John I. Manchester,
Sylvie Guichard
Mix and Match: Structure guided hybridization of dinaciclib (1), a potent pan CDK
inhibitor, with SR-3029 (2) a compound with weak, but selective CDK12 activity,
resulted in lead compound 7, a highly potent and selective CDK12 inhibitor. The
selectivity of compound 7 amongst the CDK family and across the kinome, is
supported by high ATP enzyme assays, single point kinase panel data, and affinity
proteomics using kinase affinity matrix proteomics.
Page No. – Page No.
Structure Based Design of Highly
Selective Non-covalent CDK12
Inhibitors
This article is protected by copyright. All rights reserved.