Discovery of potent and selective CDK8 inhibitors from an HSP90 pharmacophore

Article abstract of DOI:10.1016/j.bmcl.2016.01.062

Here we describe the discovery and optimization of 3-benzylindazoles as potent and selective inhibitors of CDK8, also modulating CDK19, discovered from a high-throughput screening (HTS) campaign sampling the Merck compound collection. The primary hits with strong HSP90 affinity were subsequently optimized to potent and selective CDK8 inhibitors which demonstrate inhibition of WNT pathway activity in cell-based assays. X-ray crystallographic data demonstrated that 3-benzylindazoles occupy the ATP binding site of CDK8 and adopt a Type I binding mode. Medicinal chemistry optimization successfully led to improved potency, physicochemical properties and oral pharmacokinetics. Modulation of phospho-STAT1, a pharmacodynamic biomarker of CDK8, was demonstrated in an APC-mutant SW620 human colorectal carcinoma xenograft model following oral administration.

Full text of DOI:10.1016/j.bmcl.2016.01.062

Accepted Manuscript
Discovery of potent and selective CDK8 inhibitors from an HSP90 pharmaco-
phore
Kai Schiemann, Aurélie Mallinger, Dirk Wienke, Christina Esdar, Oliver
Poeschke, Michael Busch, Felix Rohdich, Suzanne A. Eccles, Richard
Schneider, Florence I. Raynaud, Paul Czodrowski, Djordje Musil, Daniel
Schwarz, Klaus Urbahns, Julian Blagg
PII:
S0960-894X(16)30060-9
DOI:
http://dx.doi.org/10.1016/j.bmcl.2016.01.062
BMCL 23527
Reference:
Bioorganic & Medicinal Chemistry Letters
To appear in:
Received Date:
Revised Date:
Accepted Date:
22 December 2015
18 January 2016
21 January 2016
Please cite this article as: Schiemann, K., Mallinger, A., Wienke, D., Esdar, C., Poeschke, O., Busch, M., Rohdich,
F., Eccles, S.A., Schneider, R., Raynaud, F.I., Czodrowski, P., Musil, D., Schwarz, D., Urbahns, K., Blagg, J.,
Discovery of potent and selective CDK8 inhibitors from an HSP90 pharmacophore, Bioorganic & Medicinal
Chemistry Letters (2016), doi: http://dx.doi.org/10.1016/j.bmcl.2016.01.062
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Discovery of potent and selective CDK8 inhibitors from an HSP90 pharmacophore
Kai Schiemanna*, Aurélie Mallingerb, Dirk Wienkea, Christina Esdara, Oliver Poeschkea, Michael
Buscha, Felix Rohdicha, Suzanne A. Ecclesb, Richard Schneidera, Florence I. Raynaudb, Paul
Czodrowskia, Djordje Musila, Daniel Schwarza, Klaus Urbahnsa and Julian Blaggb
a
Merck KGaA, 64293 Darmstadt, Germany
b Cancer Research UK Cancer Therapeutics Unit at The Institute of Cancer Research, London SW7 3RP, U.K.
Keywords: WNT pathway inhibitors; Mediator complex; CDK8; CDK19; HSP90; indazoles
*
Corresponding
author
Tel.:
+496151728139,
fax:
+496151723129,
e-mail:
kai.schiemann@merckgroup.com
Graphical abstract
18
CDK8 IC50 = 10 nM
Cell, 7dF3 IC50 = 65 nM
F (mouse/rat/dog) = 50/88/85%
Abstract – Here we describe the discovery and optimization of 3-benzylindazoles as potent and
selective inhibitors of CDK8, also modulating CDK19, discovered from a high-throughput
screening (HTS) campaign sampling the Merck compound collection. The primary hits with strong
HSP90 affinity were subsequently optimized to potent and selective CDK8 inhibitors which
demonstrate inhibition of WNT pathway activity in cell-based assays. X-ray crystallographic data
demonstrated that 3-benzylindazoles occupy the ATP binding site of CDK8 and adopt a Type I
binding mode. Medicinal chemistry optimization successfully led to improved potency,
1

physicochemical properties and oral pharmacokinetics. Modulation of phospho-STAT1, a
pharmacodynamic biomarker of CDK8, was demonstrated in an APC-mutant SW620 human
colorectal carcinoma xenograft model following oral administration.
The Mediator complex-associated kinase CDK8 and its closely related paralog CDK19 are
transcription-regulating kinases.1,2,3 In contrast to some other members of the CDK family (for
example CDK1, CDK2 and CDK4/6), CDK8 does not play a direct role in cell cycle progression.
Embryonic development is prevented by CDK8 knockout in embryonic stem cells due to its
essential role in the pluripotent stem cell phenotype, but the growth of normal cells is not affected
by CDK8 depletion.4, 5,6 CDK8 and CDK19 are cyclin dependent enzymes that have a conserved
function in transcription and form the kinase module of the Mediator complex along with MED12
and MED13.2,7 The Mediator complex is a multiprotein assembly comprising at least 30 subunits
that functions as a regulator of gene transcription in multiple contexts including stem cell function,
the immune response, inflammation, cell adhesion, the epithelial to mesenchymal transition and
development.8,9,10 CDK8 has been shown to regulate several signaling pathways that are key
regulators of cancer and has been reported to control basal transcription by phosphorylation of RNA
polymerase II11, to promote expression of -Catenin target genes12 and to phosphorylate E2F1
thereby activating WNT signaling.13 Gene expression of CDK8 correlates with activation of β-
catenin, a core transcriptional regulator of canonical WNT signalling, in colon and gastric
cancers.14,15 CDK8 has been identified to be essential in cell proliferation in melanoma16 and to act
as an oncogene in colon cancer6, the CDK8 gene being amplified in about 60% of colorectal
cancers. CDK8 gene expression also correlates with a worse prognosis in colon, breast and ovarian
cancer.14 Additional cancer-relevant activities of CDK8 include growth factor-induced transcription
and modulation of TGF signaling17 and phosphorylation of the Notch intracellular domain.18,19
As CDK8 is a regulator of the WNT/-catenin signaling pathway, it may be associated with
disorders and diseases where activation of the WNT/-catenin pathway plays a role. Thus, the
2

treatment of several medical conditions by inhibiting the WNT/-catenin signaling has been
suggested and discussed20,21 among these are hyperproliferative, inflammatory or degenerative
disorders, in particular several malignant diseases. 22,23
The function and role of CDK19 is less well explored and its role is subject of ongoing studies.7
Previously-reported small molecule ligands for CDK8 and its paralog CDK19 have been described
in a recent comprehensive review.24 In the patent literature other series of small molecule CDK8
inhibitors have also been reported.25,26,27,28
Small molecule inhibitors of CDK8 were found in a HTS of our in-house compound collection. An
in vitro FRET-based ATP competitive CDK8/Cyclin C assay measuring the compound’s ability to
displace a commercially available ligand was established, providing a simple and robust
biochemical assay to determine compounds with potent affinity for the Mediator complex-
associated kinase CDK8.29 In our HTS campaign of a diverse 600,000 compound library, around
190 benzylindazoles - a chemical class previously claimed as HSP90 inhibitors30 - were tested, of
which 160 confirmed hits were identified, which represents an extraordinarily high hit rate within
the series. R-group analysis clearly showed that the 5-position of the benzylindazoles plays an
important role in CDK8 affinity as indicated by compound 1 with an amide moiety (Figure 1,
CDK8 IC50 = 15 nM) compared to a similar bromine-substituted intermediate 2 (CDK8 IC50 = 8600
nM). In addition to the CDK8 affinity, HSP90 inhibition also decreased 150-fold by removing the
amide functionality30; a first indicator that selectivity over HSP90 binding may not be possible by
varying this group. Additional closely-matched pairs (data not shown) indicated that the position of
the methyl group on the C3-benzylic moiety only moderately influences CDK8 and HSP90 potency
and does not counteract the above conclusions.
IC50 (CDK8) = 15 nM
IC50 (CDK8) = 8.6 µM
IC50 (HSP90) = 17 µM
IC50 (HSP90) = 110 nM
1
2
3

Figure 1: Initial HTS hits and their biological activities against CDK8 and HSP90.
Alongside favorable physicochemical properties, high affinity binding and cellular potency,
biochemical selectivity was also a major objective of the desired compound characteristics. In
former studies, we previously reported that HSP90 affinity depends upon the hydrogen bond
network of the phenolic OH-group, an important component of the HSP90 pharmacophore.30
Compounds 3 and 4 illustrate the importance of the C6-OH group for HSP90 but not for CDK8
affinity (Figure 2). By removing the hydroxyl group the HSP90 activity dropped from IC50 = 430
nM (compound 3) to >20 µM (compound 4), whereas the CDK8 inhibition was not affected. We
did not expect the gem-difluoro substitution on the pyrrolidine ring of 4 to have a strong effect on
the CDK8 activity compared to the non-substituted pyrrolidine derivative 3 (as demonstrated later
in the program by comparing the activity of 37 and 38). Cognizant of this SAR, we monitored
HSP90 affinity throughout the course of our medicinal chemistry program and optimized
compounds were consistently devoid of HSP90 affinity (Supplementary Table S1).
IC50 (CDK8) = 12 nM
IC50 (CDK8) = 16 nM
IC50 (HSP90) = 430 nM
IC50 (HSP90) > 20 µM
3
4
Figure 2: Influence of the phenolic group on CDK8 and HSP90 binding
Benzylindazoles could be readily assembled by a 6-step synthesis starting from commercially
available 1-benzyl-5-bromo-1H-indazole-3-carboxylic acid as shown in Scheme 1. After activation
of the acid 5, compound 6 was reacted with the Grignard reagent 7 to form the ketone 8 in good
yields. Instead of the Grignard reagent, lithiated aromatic compounds were also used (formed from
the corresponding bromides by treatment with n-BuLi). Pd-catalyzed carbonylation of the
arylbromide to the ester in MeOH was followed by an oxidative cleavage of the benzyl protecting
group which concomitantly hydrolyzed the methyl ester to give carboxylic acid 10. Reduction of 10
4

in SiHEt3/TFA resulted in precursor 11, which gave the corresponding amides under standard
amidation reaction conditions, in this case to form compound 4. Compound 4 showed potent
biochemical activity (IC50 = 15 nM), but was excluded due to its suboptimal metabolic stability
(Clint (human/rat/mouse) = 244/792/>1000 µL/min/mg respectively).
7
i.
ii.
iii.
5
6
8
9
iv.
HCl
12
vi.
v.
11
10
4
i. Methoxymethylamine, EDCI, DMAP, DMF, 2 h, RT, 88%; ii. THF, 0°C, 92%; iii. Pd(dppf)Cl2*CH2Cl2, KOAc, CO,
MeOH, 12 h, 80°C, 73%; iv. t-BuOK, O2, DMSO, 0°C-RT, 12 h, 82%; v. SiHEt3, TFA, 15 h, RT, 100%; vi. HATU,
DIPEA, DMF, 1 h, RT, 28%.
Scheme1: General synthetic route to benzylindazole derivatives, exemplified by the preparation of
compound 4
The crystal structure of the very potent HTS hit 13 (IC50 CDK8 = 6 nM) in CDK8/cyclin C was
solved (2.4 Å resolution, PDB code 5HNB, Supplementary Table S2). Compound 13 binds into the
ATP binding site in a Type I, DMG-In mode. The indazole N-2 nitrogen interacts with NH of
Ala100 in the hinge region with an additional interaction observed between the indazole NH and the
carbonyl of Asp98 (Figure 3a). The indazole 6-membered ring stacks against gatekeeper residue
Phe97 whilst the aromatic ring of the C-3-benzylic substituent engages in a cation--stacking
interaction with guanidine side chain of Arg356 which is recruited into the ATP binding site from
the C-terminus of the protein. This unusual insertion of Arg356 into the hinge region of CDK8 is
consistent with our previous observations31,34 and, as this C-terminal extension is not present in the
5

other CDKs, suggested that CDK8 selectivity may be achieved (vide infra). The C6-carbonyl
moiety, important for potency, interacts with Lys52, whereas the phenolic-OH does not interact
with the protein, supporting the fact that removal of this group does not influence the binding
affinity (see also compound 3 and 4, Figure 2). The pyrrolidine ring points in direction of the
aqueous surroundings and interacts with Asn156 und Ala155.
13
IC50 (CDK8) = 6 nM
a.
CCT251545
IC50 (CDK8) = 12 nM
b.
Figure 3: a. Chemical structure of 13 and CDK8/Cyclin C-bound crystal structure of 13 (resolution
2.4 Å, PDB code 5HNB); b. Overlay of 13 in CDK8/Cyclin C and of CCT251545 in CDK8/Cyclin
C (resolution 2.6 Å, PDB code 5BNJ)
We have previously reported the discovery of CCT25154532, a potent, orally bioavailable small
molecule inhibitor of WNT signalling from a cell-based pathway screen, identified as a
CDK8inhibitor with equipotent affinity for CDK19.31 The SAR of the CCT251545 series31,34
,
6

refined by information from protein co-crystal structure data, highly influenced the optimization of
the indazole series. The pyridine moiety of CCT251545 in the complex with CDK8/Cyclin C
(resolution 2.6 Å , PDB code 5BNJ) interacts with the hinge region (Ala100) and the C-3-phenyl
ring forms cation--stacking interactions with the guanidine side chain of Arg356 similar to the
benzylic moiety of 13 (Figure 3b). Comparison of the hydroxypyrrolidine of 13 to the spirolactam
of CCT251545 showed hydrogen binding to Ala155 and Asn156 for the former and interaction with
Asp173, for the latter, suggesting that further optimization could be achieved; both compounds
show carbonyl-Lys52 hydrogen bonding interactions. We were keen to optimize biochemical and
cellular potency, using our previously described 7dF3 WNT-specific reporter gene assay33,
metabolic stability and kinetic solubility (Table 1). SAR at the C-3 position of the indazole series
was evaluated with a methyl amide at C-5 position. Replacement of the benzyl group at C-3 by an
isobutyl group (14) led to low biochemical potency, consistent with the importance of a cation--
stacking interaction with Arg356. Introduction of the electron-rich 3,4-dimethoxybenzyl substituent
(compound 15) was detrimental to potency possibly due to steric encumbrance arising from the
introduction of an acyclic meta substituent. Pleasingly, the introduction of bicyclic systems
(compounds 17 and 18) improved biochemical potency34 and 18 demonstrated good translation to
cell-based potency (IC50 = 65 nM), accompanied by an excellent improvement in metabolic stability
in all species tested. 4-Pyrazolyl derivative 19 (having the same substituent as CCT251545) and its
solubility-enhancing derivative 20 also proved potent and stable; indeed, the thermodynamic
solubility Sth = 0.006 mg/mL of 19 was increased 5-fold to 0.026 mg/mL for 20. In addition, we
noted that the indazole N1-H is crucial for potency (via interaction with the carbonyl of Asp98)
since the N-1-methyl derivative of 18 (not shown) had weaker biochemical potency (IC50 = 16 µM).
Table 1: SAR on the C-3 position
7

CDK8
IC50 [µM]
7df3
IC50 [µM]
Clint (H/R/M)) Kinetic solubility
entry
14
R
[µL/min/mg]
[µM]
0.840
2.50
>10
>10
45/33/411
>200
<10/47/782
176
167
15
16
0.170
0.280
1.90
2.00
34/142/>1000
10/23/35
<10/<10/39
19/18/23
192
>200
169
17
18
19
0.010
0.015
0.065
0.055
0.016
0.076
<10/15/30
182
20
H: human, R: rat, M: mouse
Although the methylindazole moiety at C-3 in 18 appeared to be equally attractive, we selected the
4-(1-methyl-1H-pyrazol-4-yl)-benzyl moiety (present in compound 19) to further explore the SAR
at the C-5 position (Table 2). An interaction of the C-5-carbonyl moiety with Lys52 was evident in
the protein crystal structure of compound 13 (Figure 3a) and we set out to examine its importance
(Table 2). The amide was substituted by aromatic bioisosteres (compounds 21-23) to mimic
interaction with Lys52. Compounds 21, 22 and 23 demonstrate weaker CDK8 potency compared to
19 (IC50 = 78, 32, 270 and 15 nM respectively). Due to the decreased potency, solubility and
stability in mouse liver microsomes, the compounds were not progressed further. The carbonyl
moiety of compound 18 seems to also have a preferable orientation compared to compounds 24-26
in which the carbonyl function is pushed further away from the aromatic system by introducing
8

retroamides or sulfonamides. The biochemical potency of 24 (IC50 = 29 nM) is slightly worse
compared to 19 (IC50 = 15 nM), also reflected by the 2-fold decrease in cell-based potency. A
similar trend can be seen with the sulfonamide 25. The similar potency of the lactam 26 is
compromised by its unfavorable metabolic stability in mouse in vitro. Further pushing the carbonyl
moiety away from the indazole scaffold or removing it completely diminished the potency as shown
for compound 27 and 28 (IC50 CDK8 = 0.68 and > 30 µM, respectively). The sulfonamides, e.g. 29,
are less potent and less stable compared to 19.
Table 2: SAR of indazole C-5 variations
CDK8
IC50 [µM] IC50 [µM]
7df3
Clint (H/R/M) Kinetic solubility
entry
19
R
[µL/min/mg]
[µM]
0.015
0.078
0.032
0.055
0.200
1.90
19/18/23
169
18/17/99
6
21
22
372/73/441
96
0.270
0.029
0.073
0.036
1.50
0.110
0.710
0.082
15/17/62
33/17/58
30/11/32
41/31/320
53
23
24
25
26
115
153
147
0.680
>30
>10
>30
14/14/37
nd
183
153
107
27
28
29
0.180
0.313
100/155/123
H: human, R: rat, M: mouse; nd: not determined
9

To further improve cellular potency we concentrated our work on amide variation. In the protein X-
ray crystal structure of 13 in CDK8/Cyclin C (Figure 3a) the (S)-3-hydroxypyrrolidine amide
showed hydrogen-bonding interactions with Asn156 and to Ala155. We combined the favored 4-(1-
methyl-1H-pyrazol-4-yl)-benzyl substituent at C-3 with the (S)-3-hydroxypyrrolidine amide at C-5
of the indazole scaffold (compound 30); pleasingly, we observed improved CDK8 biochemical
potency (IC50 = 3 and 55 nM respectively) with good kinetic solubility and metabolic stability.
Enantiomer 31 showed similar potency suggesting tolerance of functional groups and their
orientation in this region of the ligand. Unfortunately both 30 and 31 proved to be P-gp substrates
with a decreased passive permeability shown in the Caco-2 assay (Table 3). Since the non-
substituted amide 32 exhibited weak biochemical potency (CDK8 IC50 = 110 nM), we focused our
attention on small aliphatic and cyclic alkyl amides bearing HBD or HBA functionality with the
potential to generate additional polar interactions with the protein. The hydroxyazetidine derivative
35 with a proposed interaction of the hydroxyl function with Ala155 and Asn156 exhibited potent
biochemical activity (IC50 = 4 nM) but relatively poor cellular potency (IC50 = 94 nM) consistent
with lower passive permeability and an efflux ratio of 140. To overcome this high P-gp affinity we
removed one HBD by synthesizing the methoxy derivative 36; gratifyingly, the cellular potency
was restored (IC50 = 9 nM) and the passive permeability increased (from Papp a>b + I = 5.6 to 28.8 x
10-6 cm/s) with reduced efflux ratio (11.5). Compounds with a decreased number of H-bond donors
and/or acceptors; for example, difluoro derivative 34, mitigated P-gp substrate liabilities but showed
decreased solubility and metabolic stability, most notably in rat (Clint = 190 µL/min/mg). A similar
observation was noted between the hydroxypyrrolidines 30 and 31 and non-hydroxy derivatives 37
and 38, where improved passive permeability and reduced efflux ratio was achieved at the expense
of in vitro metabolic stability. Introduction of morpholine 39 and piperazine 40 to increase
solubility also resulted in potent CDK8 inhibitors in the low nM range; however, the piperazine
basic center resulted in reduced passive cell permeability and high efflux ratio resulting in poor
10

cellular potency. Compound 31 was also tested for its kinetic binding affinity to CDK8 and CDK19
in a commercially available reporter displacement assay (Proteros Biostructures GmbH)35 and
demonstrated equipotent affinity versus CDK8/Cyclin C (IC50 = 6 nM) and CDK19/Cyclin C (IC50
= 4 nM) with residence times of around 60 min, in good agreement with the biochemical activity
reported here (IC50 CDK8 = 2 nM). Compounds 31 and 39 were also tested against a panel of 307
kinases; 31 inhibits none and 39 only one kinase by more than 50% inhibition at 1 µM (SAPK2A:
52% inhibition, Supplementary Table S3). Compounds 19, 30, 34 and 39 were inactive in the
biochemical HSP90 filter assay (Supplementary Table S1), confirming loss of HSP90 affinity due
to removal of the C6-hydroxy group.
Table 3: SAR of indazole C-5 amide variations
Kinetic
solubility
[µM]
Caco-2
Papp a>b [10-6
cm/s]/ER
CDK8
IC50 [µM]
7df3
IC50 [µM]
Clint (H/R/M)
[µL/min/mg]
entry
19
R
0.015
0.003
0.055
0.053
19/18/23
15/14/13
169
41.2/4.4
200
8.0/497
30
31
0.002
0.042
11/13/16
181
8.7/487
0.110
0.008
0.270
0.100
<10/<10/13
51/52/65
173
76
nd
nd
32
33
0.007
0.025
31/190/73
89
39.0/2.6
34
0.004
0.006
0.094
0.009
<10/15/15
40/103/59
196
170
5.6/140
35
36
28.8/11.5
11

0.004
0.005
0.005
0.027
70/233/167
40/75/111
164
104
27.8/9.2
36.5/6.0
37
38
0.004
0.007
0.006
0.011
0.140
0.013
26/48/59
<10/14/53
279/416/720
170
186
106
31.5/7.6
2.3/441
nd
39
40
41
H: human, R: rat, M: mouse; ER: efflux ratio; nd: not determined
In addition, a few analogs of 18 with a 1-methyl-1H-indazole-5-ylmethyl-substitution at C-3 of the
indazole were synthesized, which demonstrated the same trends in SAR, kinetic solubility,
metabolic stability and permeability (selected examples are shown in Table 4). Again, the passive
permeability and efflux ratio are highly influenced by the number of HBDs and HBAs. Whereas the
hydroxypyrrolidine derivative 42 has a low permeability and high efflux ratio, its
difluoropyrrolidine derivative 43 is a permeable without P-gp substrate liability. The same is true
for the hydroxyazetidine and its methoxyazetidine derivatives (44 vs 45). In contrast to the
suboptimal Caco-2 data, polar derivatives were found to be metabolically stable and vice versa.
Table 4: SAR of selected indazole C-5 amide variants, bearing a 1-methyl-1H-indazole-5-ylmethyl-
substitution at C-3, compared to 18
Kinetic
solubility
[µM]
Caco-2
Papp a>b [10-6
cm/s]/ER
CDK8
IC50 [µM]
7df3
Clint (H/R/M)
entry
18
R
IC50 [µM] [µL/min/mg]
0.010
0.004
0.065
0.043
<10/<10/29
>200
33.1/6.8
<10/<10/<10
>200
4.3/110
42
12

0.002
0.004
0.056
0.160
45/27/354
157
33.4/3.6
43
44
<10/<10/21
>200
12.1/39.7
0.007
0.018
44/126/114
>200
33.6/6.6
45
H: human, R: rat, M: mouse; ER: efflux ratio
Compound 18 proved the best compromise between favorable cellular potency, solubility,
metabolic stability and efflux ratio (Table 4) and was progressed to in vivo studies. Due to the
promising low dose in vivo PK in mouse (Table 5), we examined the low dose PK profile in higher
species where we observed low clearance and moderate to good oral bioavailability in all species
tested resulting in a favorable PK prediction to human. We estimated a human biologically
efficacious dose (BED) from in vivo exposures in mouse models that elicit robust
pharmacodynamic biomarker modulation (vide infra). At the human BED, the oral bioavailability of
18 was predicted to be low (approx. 10%) by PBPK simulation (GastroPlusTM) due to low
thermodynamic solubility (Sth = 0.006 mg/mL). We therefore consider 18 to be a BSC
(Biopharmaceutics Classification System) class 2 compound (solubility limited absorption)
although we anticipate that an increase in human oral bioavailability may be possible by appropriate
salt formation and/or pharmaceutical formulation to improve aqueous solubility.
Table 5: PK data of 18 in several species and allometric scaling to human.
Species
Clint [L/h/kg]
t1/2 [h]
Vss [L/kg]]
F [%]
Fub [%]
mouse
rat
dog
0.52 (10% LBF)
0.94 (28% LBF)
0.31 (17% LBF)
0.28 (11% LBF)
0.05 (4% LBF)
1.01
0.84
4.20
3.30
4.20
0.69
0.85
1.80
0.95
0.30
50
88
85
26
10*
10
2.4
3.3
19
monkey
Human (est.)
* Predicted by PBPK simulation (GastroPlusTM
1.6
)
In the commercially available reporter displacement assay,35 compound 18 is equipotent against
CDK8/Cyclin C (IC50 = 53 nM) and CDK19/Cyclin C (IC50 = 26 nM) each with a short residence
13

time (< 1.4 min), which is comparable to the biochemical activity reported here (IC50 CDK8 = 10
nM). 18 displayed high flux (Papp a>b + I = 33.1 × 10−6 cm/s), moderately increased efflux (ER =
6.8) and minimal hERG inhibition (IC50 >> 10 μM); aqueous thermodynamic solubilities were also
sufficient for further progression [0.006 mg/mL (PBS, pH 7.4), 0.032 mg/mL (FaSSIF, pH 6.4) and
0.059 mg/mL (FeSSIF, pH 5.2)]. When tested across a panel of 55 receptors, ion channels, and
enzymes at 1 μM (CEREP), compound 18 demonstrated activity on 5HT2A (IC50 = 240 nM).
Testing versus 307 kinases at 1 μM revealed weak inhibition of PASK (IC50 = 260 nM), as the only
potency observed below 1 µM (Supplementary Table S3); this level of activity is insufficiently
potent to account for the observed effects on cell-based WNT pathway activity. To confirm activity
on WNT signaling, compound 18 was tested in a human colorectal cancer cell line LS174T (β-
catenin-mutant) with constitutively activated WNT signaling and PA-1 human teratocarcinoma
cells that are WNT ligand-dependent; 18 inhibited reporter-based readouts of basal WNT pathway
activity in LS174T cells in the absence of WNT ligand or other stimulants (IC50 = 340 nM) and
demonstrated inhibition of WNT3a ligand-dependent reporter readout in PA-1 cells (IC50 = 710
nM). It was inactive in the biochemical HSP90 assay (0% inhibition @ 4 µM, Supplementary Table
S1).
Compound 18 was then assessed in vivo in an APC-mutant SW620 human colorectal carcinoma
xenograft model to assess pharmacokinetics and inhibition of STAT1SER727 phosphorylation which
we have previously demonstrated to be an in vitro cell-based and an in vivo pharmacodynamic
biomarker of CDK8 inhibition.31 Mice were treated orally with a single 20 mg/kg dose of 18 and
tumor STAT1SER727 phosphorylation was assessed by Luminex immunoassay.31 We observed a 75%
reduction in tumor phospho-STAT1SER727 at 2 and 6 h post-dose (Figure 4a) consistent with free
plasma exposure > 3 fold higher than calculated cell-based free IC50 in the same cell line (Figure 4b,
IC50 phospho-STAT1SER727 = 49 nM; SW620 cell-based assay contains 10% FCS, fub = 13% of 18 in
100% FCS) thus demonstrating robust inhibition of CDK8 at sufficient free plasma exposure.
14

a.
b.
Figure 4: a. PD modulation following treatment of an APC-mutant SW620 human colorectal
carcinoma xenograft model with a single dose of 18 (20mg/kg po); b. Free plasma concentration of
18 in the same experiment.
In conclusion, we have described the synthesis and properties of potent and selective CDK8
inhibitors with high affinity to CDK19 and high cellular activity. Starting from a high-throughput
screen, we were able to rapidly identify inhibitors with biochemical and cellular potencies in the
low nanomolar range that possess acceptable physical and biological properties, including good oral
pharmacokinetics in several species. The documented SAR is consistent with the X-ray structure of
CDK8/Cyclin C complex with 13, which binds to the ATP binding pocket in a Type I binding
mode. Oral administration of 18 enabled significant systemic exposure in mice leading to a
significant reduction of the target engagement biomarker phosphor-STATSER727. The canonical
WNT pathway is of significant interest in oncology and the described inhibitors of CDK8 offer the
potential for further progression subject to acceptable safety and toleration profiling.
Acknowledgement
We thank Dr. Christian Sirrenberg for providing the HSP90 inhibitory data and Dr. Mark Stubbs
and Dr. Rosemary Burke for determination of IC50 data in the LS174T assay. We also thank the
team of Proteros Biostructures GmbH, Bunsenstrasse 7a, D-82152 Martinsried, Germany, for the
Reporter Displacement Assay and in particular Dr. Elisabeth V. Schneider and Dr. Alfred Lammens
for the X-ray co-crystal structures with CDK8/Cyclin C. AM, SAE, FRa and JB are supported by
Cancer Research UK [grant number C309/A11566] and acknowledge NHS funding to the NIHR
15

Biomedical Research Centre at The Institute of Cancer Research and The Royal Marsden Hospital,
London.
Supplementary Data
Total synthesis of 18, including its analytical characterization, small molecule X-ray
crystallographic data analysis (Supplementary Table S2, including Fo-Fc omit map at 2.5 σ cutoff,
Figure S1) for compound 13, kinase selectivity data for compounds 18, 31 and 39 (Supplementary
Table S3), as well as HSP90 activity of compounds 13, 18, 19, 22, 26, 30, 34 and 39
(Supplementary Table S1). This material can be found in the online version, at
References and Notes
1
Taatjes, D. J. Trends Biochem. Sci. 2010, 35, 315.
2
Galbraith, M. D.; Donner, A. J.; Espinosa, J. M. Transcription 2010, 1, 4.
3
Firestein, R.; Hahn, W. C. Cancer Res. 2009, 69, 7899.
4
Porter, D. C.; Farmaki, E.; Altilia, S.; Schools, G. P.; West, D. K.; Chen, M.; Chang, B.-D.;
Puzyrev, A. T.; Lim, C.-U.; Rokow-Kittell, R.; Friedhoff, L. T.; Papavassiliou, A. G.;
Kalurupalle, S.; Hurteau, G.; Shi, J.; Baran, P. S.; Gyorffy, B.s; Wentland, M. P.; Broude, E. V.;
Kiaris, H.; Roninson, I. B. Proc. Natl. Acad. Sci. U.S.A. 2012, 109, 13799.
5
Westerling, T.; Kuuluvainen, E.; Makela, T. P. Mol. Cell. Biol. 2007, 27, 6177.
6
Adler, A. S.; McCleland, M. L.; Truong, T.; Lau, S.; Modrusan, Z.; Soukup, T. M.; Roose-
Girma, M.; Blackwood, E. M.; Firestein, R. Cancer Res. 2012, 72, 2129.
7
Tsutsui, T.; Fukasawa, R.; Tanaka, A.; Hirose, Y.; Okhuma, Y. Genes Cells 2011, 16, 1208.
8
Allen, B. L.; Taatjes, D. J. Nat. Rev. Mol. Cell Biol. 2015, 16, 155.
9
Kim, S.; Xu, X.; Hecht, A.; Boyer, T. G. J. Biol. Chem. 2006, 281, 14066.
10
Schiano, C.; Casamassimi, A.; Vietri, M. T.; Rienzo, M.; Napoli, C. Biochim. Biophys. Acta
2014, 1845, 66.
16

Evaluation Only. Created with Aspose.PDF. Copyright 2002-2021 Aspose Pty Ltd.
11
Rickert, P.; Seghezzi, W.; Shanahan, F.; Cho, H.; Lees, E. Oncogene 1996, 12, 2631.
12
Firestein, R.; Bass, A. J.; Kim, S. Y.; Dunn, I. F.; Silver, S. J.; Guney, I.; Freed, E.; Ligon, A.
H.; Vena, N.; Ogino, S.; Chheda, M. G.; Tamayo, P.; Finn, S.; Shrestha, Y.; Boehm, J. S.; Jain,
S.; Bojarski, E.; Mermel, C.; Barretina, J.; Chan, J. A.; Baselga, J.; Tabernero, J.; Root, D. E.;
Fuchs, C. S.; Loda, M.; Shivdasani, R. A.; Meyerson, M.; Hahn, W. C. Nature 2008, 455, 547.
13
Morris, E. J .; Ji, J.-Y.; Yang, F.; Di Stefano, L.; Herr, A.; Moon, N.-S.; Kwon, E.-J.; Haigis, K.
M.; Naeaer, A. M.; Dyson, N. J. Nature 2008, 455, 552.
14
Firestein, R.; Shima, K.; Nosho, K.; Irahara, N.; Baba, Y.; Bojarski, E.; Giovannucci, E. L.;
Hahn, W. C.; Fuchs, C. S.; Ogino, S. Int. J. Cancer 2010, 126, 2863.
15
Kim, M. Y.; Han, S. I.; Lim, S. C. Int. J. Oncol. 2011, 38, 1375.
16
Kapoor, A.; Goldberg, M. S.; Cumberland, L. K.; Ratnakumar, K.; Sequra, M. F.; Emanuel, P.
O.; Menendez, S.; Vardabasso, C.; LeRoy, G.; Vidal, C. I.; Polsky, D.; Osman, I.; Garcia, B. A.;
Hernando, E.; Bernstein, E. Nature 2010, 468, 1105.
17
Alarcon, C.; Zaromytidou, A. I.; Xi, Q.; Gao, S.; Yu, J.; Fujisawa, S.; Barlas, A.; Miller, A. N.;
Manova-Todorova, K.; Macias, M. J.; Sapkota, G.; Pan, D.; Massague, J. Cell 2009, 139, 757.
18
Fryer C. J.; White J. B.; Jones K. A. Mol. Cell 2004, 16, 509.
19
Li, N.; Fassl, A.; Chick, J.; Inuzuka, H.; Li, X.; Mansour, M. R.; Liu, L.; Wang, H.; King, B.;
Shaik, S.; Gutierrez, A.; Ordureau, A.; Otto, T.; Kreslavsky, T.; Baitsch, L.; Bury, L.; Meyer, C.
A.; Ke, N.; Mulry, K. A.; Kluk, M. J.; Roy, M.; Kim, S.; Zhang, X.; Geng, Y.; Zagozdzon, A.;
Jenkinson, S.; Gale, R. E.; Linch, D. C.; Zhao, J. J.; Mullighan, C. G.; Harper, J. W.; Aster, J. C.;
Aifantis, I.; von Boehmer, H.; Gygi, S. P.; Wei, W.; Look, A. T.; Sicinski, P. Nat. Chem. Biol.
2014, 16, 1080.
20
Moon, R. T.; Kohn, A. D.; De Ferrari, G. V.; Kaykas, A. Nat. Rev. Genet. 2004, 5, 689.
21
Chien, A. J.; Conrad, W. H.; Moon, R. T. J. Invest. Dermatol. 2009, 129, 1614.
22
Giles, R. H.; Van Es, J. H.; Clevers, H. Biochim. Biophys. Acta 2003, 1653, 1.
17