Discovery and design of molecular glue enhancers of CDK12-DDB1 interactions for targeted degradation of cyclin K

Abstract

The CDK12 inhibitor SR-4835 promotes the proteasomal degradation of cyclin K, contingent on the presence of CDK12 and the CUL4-RBX1-DDB1 E3 ligase complex. The inhibitor displays molecular glue activity, which correlates with its enhanced ability to inhibit cell growth. This effect is achieved by facilitating the formation of a ternary complex that requires the small molecule SR-4835, CDK12, and the adaptor protein DDB1, leading to the subsequent ubiquitination and degradation of cyclin K. We have successfully solved the structure of the ternary complex, enabling the de novo design of molecular glues that transform four different CDK12 scaffold inhibitors, including the clinical pan-CDK inhibitor dinaciclib, into cyclin K degraders. These results not only deepen our understanding of CDK12's role in cell regulation but also underscore significant progress in designing molecular glues for targeted protein degradation in cancers associated with dysregulated cyclin K activity.

Fig. 1. CDK12 inhibitor SR-4835 depletes cyclin K through proteasomal degradation and is dependent on the expression of CDK12 and the CUL4–RBX1–DDB1 E3 ligase complex. (A) Chemical structures of CDK12/13 inhibitors SR-4835 and THZ531. (B) Immunoblot depicting cyclin K levels in MDA-MB-231 cells treated with SR-4835 or THZ531 (500 nM) for the indicated time (left upper and lower panels) and docking poses of SR-4835 and THZ-531 in the active site of CDK12 (right upper and lower panels). (C) Degradation of HiBiT-tagged cyclin K using SR-4835 or THZ-531 (A549 cells). (D) Immunoblot analysis to assess if CDK12 is essential to provoke cyclin K degradation in the presence of SR-4835. (E) Transient knockdown of CUL4A/CUL4B gene expression led to reduced degradation of cyclin K upon treatment with SR-4835. (F) Rescue of cyclin K degradation by treatment of MBA-MD-231 cells with MLN7243; an inhibitor of the ubiquitin-activating enzyme. (G) Rescue of cyclin K degradation by treatment of MBA-MD-231 cells with pevonidistat (pevo), an inhibitor of NEDD8. (H) Immunofluorescence assay for visualizing loss of nuclear cyclin K upon treatment with SR-4835 (500 nM), THZ-531 (500 nM) and SR-4835 (500 nM) +MG132 (1 μM).

Fig. 2. The molecular glue activity of SR-4835 is correlated with its cytotoxicity in HEK-293T cells. (A) Cyclin K levels remained unchanged in MDA-MB-231_DDB1-shRNA cells treated with SR-4835 (dose range 0.01 μM to 3 μM). (B) Cyclin K levels remained unchanged upon treatment with a single dose of SR-4835 (1 μM) in HEK-293T_DDB1-shRNA cells compared to WT-HEK-293T cells. (C) Immunoblots showing genetic ablation of VHL and CRBN do not rescue cyclin K degradation in contrast to loss of DDB1 which does inhibit cyclin K degradation. (D) and (E) CellTiter-Glo® viability assays comparing the EC₅₀ values of SR-4835 and THZ531 in HEK-293T cells and its isogenic HEK-293T paired cell line with downregulated DDB1 expression (HEK-293T_DDB1-shRNA). (F) Quantification of difference found in the EC_50s of THZ531 and SR-4835 in the HEK-293T and HEK-293T_DDB1-shRNA cells after 72 hours. Error bars indicate standard deviation (SD). Comparison between the indicated groups was performed by Student's un-paired two-tailed t-tests and statistical significance is expressed as ****, p < 0.0001 compared with HEK-293T-WT group.

Fig. 3. SR-4835 induces ternary complex formation between Cdk12/CycK and DDB1. (A) Analytical size exclusion chromatography of purified, recombinant DDB1ΔB and Cdk12/CycK proteins (20 μM) in presence of either 2% DMSO or 60 μM small molecular compounds CR8, SR-4835, or THZ531. Peak fractions were analyzed by SDS-PAGE, giving rise to a SR-4835-induced ternary complex between Cdk12/CycK and DDB1ΔB, compared to THZ531-treated negative control. Samples highlighted with asterisk were subsequently analyzed by DLS. (B) Dynamic light scattering analysis of peak middle fractions of previously conducted analytical size exclusion chromatography runs (in A). Hydrodynamic radii of CR8, and SR-4835-treated samples are significantly increased compared to negative controls, indicating ternary complex formation. Box-plots display the median of 6 data points per sample, each equivalent to the average of 20 DLS measurements. (C) Thermal stability analysis using the nanoDSF method of compound-treated Cdk12/CycK (left panel) and a mixture of compound-treated DDB1ΔB and Cdk12/CycK (right panel). All compounds including THZ531 stabilized Cdk12/CycK kinase. Degrader compounds elevated their stabilizing effects in presence of DDB1ΔB, indicating ternary complex formation, while THZ531 only stabilized Cdk12/CycK. (D) pulldown assays using GST-tagged variants of Cdk12 and CycK, identifying drug-induced recruitment of DDB1 in CR8- and SR-4835-treated samples. (E) Cartoon representation of the ternary complex structure of DDB1·SR-4835·Cdk12/CycK at a resolution of 3.9 Å (PDB: 9FMR). (F) Close-up of the SR-4835-linked interface between Cdk12 (C-lobe, blue; N-lobe, turquoise) and the DDB1ΔB BPC domain (bright orange), identifying key stacking interactions between the benzimidazole moiety of SR4835 with DDB1 residue R928. Key residues forming the interface are highlighted.

Fig. 4. SR-4784 and SR-4784-PAP require DDB1 to degrade cyclin K and induce proximity between DDB1–CDK12–cyclin K. (A) Chemical structures of CDK12/13 inhibitors SR-4784 and SR-4784-PAP. (B) Immunoblot depicting cyclin K levels in MDA-MB-231 cells treated with SR-4784 alone (1–2 μM for 2 hours) and/or with MG132 (1 μM for 30 minutes). (C) Immunoblots showing cyclin K degradation in parent WT MDA-MB-231 and MDA-MB-231_DDB1-shRNA cell lines exposed to SR-4784-PAP alone (for 2 hours at 2 μM) or with MG132 (1 μM for 30 minutes). (D) Neutravidin agarose pull-down shows DDB1 enrichment by SR-4784PAP compared to SR-4784. Detection of biotinylated proteins by anti-streptavidin antibody and DDB1 is enriched in the SR-4784-PAP treated sample (for 2 hours at 2 μM). (E) Immunoblots of biotin-labelled proteins after click chemistry reaction showing cyclin K, CDK12, and DDB1 levels (for 2 hours at 2 μM).

Fig. 5. Cyclin K degraders by design. (A) De novo design of cyclin K molecular glue degraders from CDK12/pan-CDK inhibitors (dinaciclib). (B)–(E) Microscale thermophoresis studies assessing CDK12/cyclin K + DDB1 ternary complex formation induced by compound of interest. (F)–(I) Degradation of HiBiT tagged cyclin K using with MR-1187, MR-1110, MR-1014 and MR-1226 (A549 cell lines). (J)–(M) Label-free TMT-based quantitative proteomics upon treatment with MR-1187 (500 nM), MR-1110 (200 nM), MR-1014 (500 nM), MR-1226 (200 nM) at 2 h. Plot shows the significantly differentially abundant proteins ranked according to their p value (Y-axis) as −log₁₀ and their relative abundance log₂ ratio (X axis) between DMSO control and the compound of interest. All samples were generated in triplicates and 0.5 fold changes cut off were used for avg log₂ ratio's to analyze the data. The p values < 0.05 were considered as significant.