Development of Highly Potent and Selective Pyrazolopyridine Inhibitor of CDK8/19

Abstract

CDK8 and its paralog CDK19 are cyclin-dependent kinases that are core components of the so-called Mediator complex that has essential roles as a positive and negative regulator of gene expression. Several efforts to develop inhibitors have yielded natural and synthetic ATP-competitive compounds including cortistatin A, Sel120, BCD-115, CCT251921 (1), and MSC2530818 (2). Here, we used a hybridization approach starting from CCT251921 and MSC2530818 to derive new inhibitors with the aim of developing highly potent and selective inhibitors of CDK8/19. Initial compounds suffered from rapid aldehyde oxidase-mediated metabolism. This liability was overcome by utilizing a pyrazolopyridine hinge binder with a chlorine at the C-3 position. These efforts resulted in JH-XVI-178 (compound 15), a highly potent and selective inhibitor of CDK8/19 that displays low clearance and moderate oral pharmacokinetic properties.

CDK8 and its paralog CDK19 are cyclin-dependent kinases that form part of the Mediator complex, a multiprotein assembly composed of at least 30 subunits that functions as a regulator of gene transcription. Mediator complex function has been implicated in multiple contexts including stem cell function, immune response, inflammation, cell adhesion, and epithelial to mesenchymal transition and development.¹⁻⁵ Deregulation of CDK8 and its binding partner cyclin C (CCNC) is a common feature of many cancers, including cancers of the colon,⁶⁻⁸ breast,⁹⁻¹² prostate,¹³ and pancreas¹⁴ and melanoma¹⁵ and leukemias.^16,17 Several ATP-competitive, small-molecule CDK8/19 inhibitors showed in vivo efficacy and two (senexin B and SEL120) have entered clinical trials, with no significant toxicities reported.¹⁸ The results of these trials have not been disclosed at the time of this publication. However, multiple toxicities were observed in rats and dogs treated with compounds 1 and 2.¹⁹⁻²² A recent report suggested these toxicities could be due to off-target effects.¹⁸ Based on this report, we set out to develop CDK8/19 inhibitors with narrower selectivity profile that may overcome these toxicities. Herein, we report our initial SAR efforts. Toxicity studies will be reported in due course.

Our strategy was to make a hybrid structure of 1 and 2 since they are known to be potent inhibitors of CDK8/19 with moderate selectivity profiles and promising pharmacokinetic profiles (Figure 1). We imagined that the hybrid of these two compounds should maintain potent CDK8/19 inhibition and might have a narrower selectivity profile. As a starting point, we decided to change the amino-pyridine hinge binder to a variety of hinge binders similar to 2, including azaindole (3), pyrazolopyridine (4), imidazopyridine (5), and thienopyridine (6).

Figure 1.

Hybridization design strategy leading to the development of this series of CDK8/19 inhibitors.

As shown in Table 1, we found that the azaindole (3) was extremely potent in the Lanthascreen enzyme assay from Invitrogen with an IC₅₀ of 2 nM. The imidazopyridine (5) was also potent with an IC₅₀ of 10 nM, while the pyrazolopyridine (4) and thienopyridine (6) lost a great deal of potency with IC₅₀ values of 178 and 206 nM, respectively. This was particularly interesting given the similarity in hinge binders between 4 and MSC2530818 (2). However, the reported binding mode of 2(22) is quite different from the reported binding mode of 1(21) with the hinge contacts being made at the nitrogens at the 1 and 2 positions with the methyl group at the 3-position directed into a hydrophobic pocket. We undertook a docking study of 3 using the cocrystal structure of 1 in complex with CDK8/cyclin C (Figure 2). The docking study predicts that the azaindole (3) maintains the same critical interactions as 1 including hinge contact with A100 and D98, a π–cation interaction with R356 and the left-hand side indazole, and hydrogen bond interactions with K51 and D173 with the lactam of the spirocycle. In addition, our docking model predicts that 4 could not adopt the same binding mode as 2 described above since the spirocycle is too large to fit in the hydrophobic pocket occupied by the methyl group of 2 and would clash with the backbone of the protein. In this case, the nitrogen at the 2-position of compound 4 is directed into a hydrophobic pocket causing an unfavorable interaction, resulting in the observed loss in potency.

Table 1. SAR of the Hinge Binding Group.

compd	X	Y	Z	CDK8 enzyme IC50 (nM)	CDK19 enzyme IC50 (nM)
3	NH	CH	CH	2	7
4	NH	N	CH	178	214
5	NH	CH	N	10	5
6	S	CH	CH	206	a

^aNot determined.

Figure 2.

Docking study of 3 using the cocrystal structure of 1 in complex with CDK8 (PDB ID 5HBJ). Dashed lines suggest important interactions.

Based on the promising results with azaindole 3, we decided to pursue this series further by exploring the SAR at the 5 position while keeping the same spirocycle at the 4 position as shown in Table 2. We found that most fused 6–5 ring systems were able to maintain potent inhibition of CDK8 with the exception of the triazolopyridine (11) and the imidazopyridine (13). Our docking model predicted that a chlorine at the 3 position could form an edge-on halogen−π interaction with F97. We therefore installed a chlorine at the 3 position of the azaindole (compound 14, Table 2) and the 3-position of the pyrazolopyridine (compound 15, Table 2), hoping to improve the potency further. While compound 14 showed similar potency against CDK8 compared to 3, compound 15 showed a drastic increase in potency (178 nM for 4 compared to 1 nM for 15). To explain this finding, we docked compound 15 into our model shown in Figure 3 and confirmed that the 2 position of the hinge binder is positioned in a hydrophobic pocket, so having a polar nitrogen in this position results in an unfavorable interaction. However, introduction of a chlorine at the 3 position, could result in a halogen−π interaction with F97 as well as a decrease in polarity around that site. Taken together, these modifications result in an increase in potency on CDK8. We then chose to explore various spirocyclic substituents at the 4 position of the azaindole again hoping to further improve potency. We found that a variety of spirocycles (compounds 16–22, Table 2) were quite potent against CDK8; however, none of them showed a significant improvement in potency compared to 3, so we decided to use the original spirocycle moving forward.

Table 2. SAR of the 4 and 5 Position of Azaindole 3.

Figure 3.

Docking study of 15 using the cocrystal structure of 1 in complex with CDK8 (PDB ID 5HBJ). Dashed lines suggest important interactions.

The representative preparation of JH-XVI-178 (15) is presented in Scheme 1. Pyrazolopyridine 23 was chlorinated using NCS in CH₃CN at 70 °C followed by SEM protection to give 24. Compound 24 then underwent nucleophilic substitution with the spirocycle 25 in NMP at 180 °C under microwave conditions to give 26, which then underwent a Suzuki coupling with the corresponding N-methyl indazole boronic ester followed by removal of the SEM group with HCl to give the desired compound 15.

Scheme 1. Preparation of JH-XVI-178.

Reagents and conditions: (a) NCS, CH₃CN, 70 °C; (b) SEM-Cl, NaH, DMF; (c) NMP, TEA, 180 °C μW; (d) Pd(dppf)Cl₂, t-BuXPhos, 1-methyl-5-(4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)-1H-indazole, Na₂CO₃, 1,4-dioxane, H₂O, 100 °C; (e) HCl, MeOH, 70 °C.

We then tested the potency of these compounds in a cellular environment by measuring their inhibition of phosphorylation of S727-STAT1, a known substrate of CDK8, in comparison with 1. As seen in Figure 4, compound 3 showed much less inhibition of pS727-STAT1 compared to 1 and 2 with an IC₅₀ of 46 nM compared to 5 nM for 1 and 3 nM for 2. Compound 15 showed slightly stronger inhibition of pS727-STAT1 with an IC₅₀ of 2 nM. Western blot data for all compounds can be found in the SI, Figure S1. Satisfied with the level of potency of 15 in both biochemical and cellular environments, we then decided to evaluate the kinase selectivity by performing KINOMEscan binding analysis against a near comprehensive panel of 468 kinases at a concentration of 1 μM as shown in Figure 5. Compound 15 displayed exceptional kinome selectivity as measured by a KINOMEscan selectivity score of S(10) = 0.01 at a concentration of 1 μM. Compound 15 exhibited off-target inhibition of only two additional kinases, STK16 and FLT3 (D835V). Dose response analysis revealed an IC₅₀ of 107 nM for STK16. Biochemical assays for FLT3 (D835V) were not commercially available at the time of writing this manuscript. Full profiling results for 15 can be found in the SI.

Figure 4.

Western blot analysis and IC₅₀ determination following 24 h treatment of Jurkat cells with 1, 2, 3, or 15 monitoring phosphorylation of s727-STAT1. IC₅₀ curves can be found in the Supporting Information section S2.

Figure 5.

KINOMEscan results for 15 at a concentration of 1 μM with a cutoff of 90% inhibition.

With these promising results, we began investigating the metabolic stability of this series by assessing mouse hepatocyte stability.

As shown in Table 3, compounds 3, 5, and 14 suffered from poor metabolic stability. Recent reports in the literature have shown azaindoles to be substrates for aldehyde oxidase (AO).²³ More specifically, the 2-position on the bicyclic ring was found to be the labile site. Based on this finding, we tested 4 in mouse hepatocytes since the 2-position is a nitrogen and therefore should not be metabolized by AO. However, 4 was also metabolically unstable. We hypothesized that since we blocked the 2-position, the 3-position could likely become the new labile site. In order to check this possibility, we tested 15 with a nitrogen at the 2-position and a chlorine at the 3-position. Gratifyingly, this substitution greatly improved metabolic stability, and we chose 15 as our lead, and it was progressed to in vivo pharmacokinetic studies in C57Bl/6 male mice.

Table 3. In Vitro Metabolic Stability of Representative Hinge Binding Groups.

compd	T1/2 (min)	Clint (μL·min–1·mg–1)	CDK8 IC50 (nM)
3	9	78	2
4	11	59	178
5	5	158	10
14	11	64	2
15	46	15	1

Compound 15 exhibited a reasonable pharmacokinetic profile with a moderate C_max of 1.04 μM following oral dosing, and a low Cl of 17 mL·min^–1·kg^–1 (Table 4). Despite having a low oral bioavailability, the plasma concentrations of 15 at 4 and 6 h were 200 nM and 20 nM, respectively, which is well in excess of the concentration needed to cover the cellular IC₅₀ value.

Table 4. Summary of Compound 15.

In Vitro
CDK8 IC50 (nM)	1
CDK19 IC50 (nM)	2
Clinta (μL·min–1·mg–1)	15

In Vivo Mouse PKb
CMaxc (μM)	1.04
AUClast (min × ng/mL)c	51863
T1/2c (h)	1.3
Cl (mL·min–1·kg–1)	17
Vss (L/kg)	0.8
F (%)	8

^aMouse hepatocytes.

^bFormulation: 0.2 mg/mL (IV) and 1 mg/mL (PO) solution in 5/95 DMSO/30% Captisol; 2 mg/kg IV dose and 10 mg/kg PO dose.

^cFollowing PO dose.

In summary, we have discovered a highly potent and selective series of CDK8/19 inhibitors based on a hybridization approach starting from CCT251921 and MSC2530818. Compound 15 significantly inhibits phosphorylation of STAT1-S727 with an IC₅₀ of 2 nM. In addition, 15 exhibits excellent kinome selectivity, and shows moderate pharmacokinetic properties following oral dosing in mice. Further optimization of this chemotype especially in regard to in vivo bioavailability as well as toxicology studies will be reported in due course.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsmedchemlett.1c00300.

• Western blots, IC₅₀ curves, full Ambit profiling data for 15, experimental details, and in vivo PK plots of time vs plasma concentration (PDF)

Author Contributions

Conceptualization: J.M.H. and N.S.G. Compound design and synthesis: J.M.H. and P.S.V. Cellular experiments: E.W. and J.J. Molecular modeling: J.H. All authors have given approval to the final version of the manuscript.

The authors declare the following competing financial interest(s): Nathanael Gray is a founder, science advisory board member (SAB), and equity holder in Gatekeeper, Syros, Petra, C4, B2S, Aduro, Inception, Allorion, Jengu, Larkspur (board member), and Soltego (board member). The Gray lab receives or has received research funding from Novartis, Takeda, Astellas, Taiho, Janssen, Kinogen, Voronoi, Arbella, Deerfield, and Sanofi.