DDB1 engagement defines the selectivity of S656 analogs for cyclin K degradation over CDK inhibition

Abstract

In efforts to identify additional therapeutic targets for Acute Myeloid Leukemia (AML), we performed a high-throughput screen that includes 56 primary specimens tested with 10,000 structurally diverse small molecules. One specific hit, called S656 acts as a molecular glue degrader (MGD), that mediates the CRL4-dependent proteolysis of cyclin K. Structurally, S656 features a moiety that binds to the ATP binding site of cyclin-dependent kinases (CDKs), allowing the recruitment of the CDK12-cyclin K complex, along with a binding site for DDB1 bridging the CRL4 complex. Structure activity relationship studies reveal that minimal modifications to the dimethylaniline moiety of S656 improve its cyclin K MGD function over CDK inhibition by promoting DDB1 engagement. This includes full occupation of the DDB1 pocket, preferably with hydrophobic terminal groups, and cation-π interaction with Arg928. Additionally, we demonstrate that despite structural diversity, cyclin K degraders exhibit similar functional activity in AML which is distinct from direct CDK12 inhibition.

Keywords: Acute Myeloid Leukemia; CDK12; Cyclin K; DDB1; Molecular Glue Degrader.

Figure 1. High-throughput screening identifies S656 as a selective anti-proliferative molecule in AML.

(A) Selection of candidate hits from the discovery screen on primary AML samples. (B) Chemical structure of the S656 hit. (C) Inhibitory profile of S656 compound across 56 primary AML specimens and 2 cord blood (CB) samples. Percentage of inhibition at 1 μM, after 6 days incubation, normalized to DMSO control treatment. (D) Dot plot distribution of S656 associated percentage of inhibition (at 1 μM) across the different AML specimen subgroups. Normalized to DMSO control treatment, median is represented in red (n = 56 primary specimens). (E) AML subtype classification of the 157 primary specimens used in validation screen. (F) Bi-modal distribution of S656 IC₅₀ values obtained in the validation screen (dose-response, 6 days incubation, normalized to DMSO control). (G) Dot plot distribution of S656 IC₅₀ values across primary AML specimens. Blue and orange dots represent more sensitive and more resistant specimens respectively as defined in the bi-modal distribution in (F) (n = 157 primary specimens). See also Appendix Fig. S1.

Figure 2. S656 mediates cyclin K degradation through the cullin RING ubiquitin ligase complex.

(A) Volcano plot representing results of whole genome CRISPR/Cas9 screen performed in EKO OCI-AML1 cells during exposure to 1 μM of S656 compound (n = 1, around 10 sgRNAs per gene). (B) Schematic representation of the Cullin4-RING E3 ubiquitin ligase (CRL4) complex. (C) Dose response experiment to determine IC₅₀ values of S656 in OCI-AML5 cells, in regular media or supplemented by 25 nM of the neddylation inhibitor MLN4924 to prevent CRL-mediated proteolysis. IC₅₀ values were determined after 4 days of incubation (mean +/− SD, n = 3, biological replicates, t test, P value = 0.0478). (D) Quantitative proteome-wide mass spectrometry analysis performed in OCI-AML5 cells exposed to 8 μM of S656 for 5 h (CCNK = cyclin K, n = 3, biological replicates). (E) Cyclin K degradation assessment by measuring the mean fluorescence intensity (MFI) of cyclin K_eGFP over mCherry by flow cytometry. OCI-AML5 G7 clone was treated for 3 h with 10 μM of the indicated compounds. Results are normalized to fluorescence in DMSO-treated cells (mean +/− SD, n = 5, biological replicates, t test, P value < 0.0001). (F) Immunoblot analysis of cyclin K and CDK12 protein levels in OCI-AML5 cells, after exposure to increasing concentrations of HQ461 or S656 for 6 h. Tubulin is used as a loading control. (G) Immunoblot analysis of cyclin K protein levels after exposure to HQ461 or S656 (5 μM for 5 h). Where indicated, OCI-AML5 cells were pre-treated 1 h with 500 nM of MLN4924 or MG132. (H) Immunoblot analysis of DDB1 and cyclin K protein levels after exposure to S656 (5 μM for 5 h) in OCI-AML5 Cas9 cells expressing inducible ( + DOX) sgRNAs targeting DDB1 or control region AAVS1. (I) Dose response experiment to determine S656 IC₅₀ values in inducible ( + DOX) OCI-AML5 Cas9 cells expressing two different sgRNAs targeting DDB1 gene or AAVS1 control region (mean +/− SD, 4 days incubation, n = 3, biological replicates, t test, P value = 0.0006 and 0.0118 in sgDDB1 #1 and #2, respectively). See also Appendix Fig. S2. Source data are available online for this figure.

Figure 3. S656 suppresses expression of DDR genes and induces DNA lesions.

(A) Immunoblot analysis of cyclin K and RNA polymerase II phospho-ser2 protein levels in OCI-AML5 cells, after exposure to HQ461 or S656 (5 μM for 5 h). Where indicated, cells were pre-treated 1 h with 500 nM of MLN4924 or MG132. Tubulin is used as a loading control. (B) Monitoring of BRCA2, ATR and BLM mRNA expression by qPCR in OCI-AML5 cells treated 4 h with 5 μM of S656 or 200 nM of the CDK12/13 inhibitor THZ531. Normalized to HPRT (mean +/− SD, n = 3, biological replicates, t test, P value < 0.0001 (DMSO vs S656) and P value = 0.0002 (DMSO vs THZ531) in BRCA mRNA, P value < 0.0001 (DMSO vs S656) and P value = 0.0292 (DMSO vs THZ531) in ATR mRNA, P value = 0.0003 (DMSO vs S656) and P value = 0.0003 (DMSO vs THZ531) in BLM mRNA). (C) Cell cycle profile (left) and quantification of the percentage of OCI-AML5 cells in cell cycle phases (right, mean +/− SD, n = 3, biological replicates), 24 h after exposure to S656. (D) Quantification of cell death by Annexin V/PI staining of OCI-AML5 cells treated for 24 h with increasing concentrations of S656 (mean +/− SD, n = 3, biological replicates). (E) U2OS cells were treated for 24 h with DMSO, 1 μM of THZ531 or 2 μM of S656, fixed with PFA and immunostained with γH2AX antibody. Quantification of the number of γH2AX foci per nucleus is presented (left, median is depicted in red, n = 3, biological replicates, t test, P value < 0.0001) with representative images (right, red: γH2AX, blue: DAPI, scale bar: 10 μm). (F) Immunoblot analysis of cyclin K, MYC and MCL-1 protein levels in OCI-AML5 cells after exposure to indicated cyclin K MGDs (5 μM for 4 h). Tubulin is used as a loading control. See also Appendix Fig. S3. Source data are available online for this figure.

Figure 4. Cyclin K degraders and CDK12/13 inhibitor show distinct fingerprints on primary AMLs.

(A) Inhibitory profile of S656, CR8, SR4835, HQ461, NCT02, THZ531 and SNS-032 compounds across 40 primary AML specimens in dose response assays. IC₅₀ values were determined after 6 days of incubation and normalized to DMSO control. (B) Dot plot summarizing the IC₅₀ values and medians (red) obtained in the panel of 40 primary AMLs. (C) Heatmap representation and hierarchical clustering of IC₅₀ values from (A). Correlation between S656 and the other molecules is indicated. Relative color scheme uses the minimum and maximum IC₅₀ values in each row. Source data are available online for this figure.

Figure 5. S656 mediates interaction between CDK12 and DDB1.

(A) 3D representation of S656 (turquoise, stick representation) bound to CDK12 (grey ribbons) and DDB1 (green surface representation). H-bonds are depicted as yellow dashed line and aromatic H-bonds are shown as turquoise dashed lines. Met816 provides two backbone interactions, a H-bond donor, and a H-bond acceptor, respectively. The CH on the thiazole core participates in an aromatic H-bond with Glu814 backbone carbonyl. S656’s pyridine ring also engages in a H-bond with Lys756. The phenyl ring of S656 establishes a cation-π interaction with Arg928 on DDB1, shown in green dashed lines and an aromatic H-bond to Asp817 on CDK12. (B) 2D interaction diagram of S656 bound to CDK12 and DDB1, in which purple arrows represent H-bonds and the red line stands for a cation-π interaction. (C) Values of IC₅₀ in proliferative assay (OCI-AML5, 4 days incubation) and percentage of cyclin K_eGFP degradation is displayed for S656 compound and its linear analog UOM-005845 (missing the methylene linker). Structural features of S656 are highlighted in different colors: methylene linked moiety (referred as “left-hand side”) of the molecule in green, aminothiazole core in blue and the “right-hand side” in grey. (D) In vitro inhibition of CDK12/Cyclin K activity by increasing concentrations of indicated molecules (mean +/− SD, n = 3, biological replicates). THZ531 is used as a positive control. (E) Immunoblot analysis of cyclin K protein levels after exposure to CR8, HQ461 or S656 (5 μM for 5 h). Where indicated, OCI-AML5 cells were pre-treated 1 h with 5 μM of THZ531. (F) Schematic representation (top) of the NanoLuc Binary Technology used to monitor the interaction between CDK12 fused to the long bait and DDB1 fused to the short bait. Luminescence intensity (bottom) generated by the reconstitution of the NanoLuc was monitored over 1 h right after the addition of S656 compound at increasing concentrations in HEK293 cells (mean +/− SD, n = 4, biological replicates). (G) Structure of the pull-down probes UOM-005790 and UOM-005839 derived from S656 and UOM-005628 respectively (carries an additional nitrogen highlighted in red). Probes were designed by functionalizing the dimethylaniline moiety with a PEG linker (grey) which was then immobilized in beads. (H) Overlay of compounds UOM-005628 (yellow, stick representation) and S656 (turquoise, stick representation), when bound to CDK12 and DDB1 (proteins not shown). The rotation of the thiadiazole in UOM-005628 when compared to the thiazole in S656, exhibits a different binding motif towards the kinase hinge region. (I) Graphical representation of the pull-down enrichments obtained for the indicated proteins (analyzed using the R package DEP). Pull-down assays were performed using the UOM-005790 probe in absence or presence of excess S656 (competition condition), the negative control probe UOM-005839 or beads with no probes in OCI-AML5 total protein extracts (mean +/− SD, n = 3, biological replicates). See also Appendix Fig. S4. Source data are available online for this figure.

Figure 6. Hit compound S656 displays both cyclin K degrader function and CDK inhibitory activity.

(A) Dose response experiment to determine IC₅₀ values of the indicated compounds in OCI-AML5 cells expressing shRNAs targeting CDK1, 2, 4, 7, 9, cyclin K or a control region (4 days incubation, mean +/− SD, n = 3, biological replicates, t test, P value = 0.0011 (sh ctrl vs sh Cyclin K) in CR8, P value = 0.0029 (sh ctrl vs sh Cyclin K) in SR4835, P value = 0.0011 (sh ctrl vs sh Cyclin K) in HQ461, P value = 0.0025 (sh ctrl vs sh CDK9) in SNS-032, P value = 0.0038 (sh ctrl vs sh CDK1) in S656, P value = 0.0253 (sh ctrl vs sh CDK9) in S656 and P value = 0.0004 (sh ctrl vs sh Cyclin K) in S656). (B) Dot plot representation of the IC₅₀ values of all S656 analogs (black) and control molecules (colors) in OCI-AML5 cells along with their efficiency to degrade cyclin K_eGFP (percentage of degradation after 3 h incubation with 10 μM of compounds, compared to DMSO). Highest tested dose is 10000 nM. (C) Dot plot representation of the ratio of IC₅₀ values in OCI-AML5 expressing shRNA control or targeting CUL4A, and the percentage of cyclin K_eGFP degradation (3 h incubation with 10 μM of compounds, compared to DMSO) for all active (IC₅₀ < 10000 nM) S656 analogs (black) and control molecules (colors). (D) Working model in which molecules with high affinity for DDB1 behave as cyclin K degraders while modifications of the left-hand side of the molecule, lowering or abolishing the interaction with DDB1, induce cytotoxicity through CDK inhibition. See also Appendix Fig. S5. Source data are available online for this figure.

Figure 7. Minimal modifications in the DDB1 interacting surface optimizes cyclin K degrader function.

(A) Heatmap clustering S656 analogs and control molecules according to IC₅₀ values obtained in OCI-AML5 stably expressing shRNAs targeting CDK1, 2, 4, 7, 9 as well as cyclin K. IC₅₀ values in OCI-AML5 (4 days incubation) and the percentage of cyclin K_eGFP degradation (3 h, 10 μM, compared to DMSO) are displayed for each analog. For each class of molecules, 3 representative structures are depicted. Relative color scheme uses the minimum and maximum IC₅₀ values in each row. Highest dose tested is 10,000 nM. (B–E) 3D representation of the indicated molecule (stick representation) bound to CDK12 (grey ribbons) and DDB1 (green surface representation). The van der Waals surface of the molecules are shown in Mesh to showcase the volume of the molecule. In (B), UOM-005197 forms a cation-π interaction between the phenyl ring and Arg928 of DDB1, but the DDB1 pocket formed by the residues Asn907, Ile909 and Arg947 cannot be filled. Hinge region interactions on CDK12 are conserved. In (C), UOM-005636 shows interaction with the hinge region of CDK12: Met816 provides two backbone interactions, a hydrogen bond donor, and a hydrogen acceptor, respectively, while Glu814 participates in an aromatic H-bond. UOM-005636’s Pyridine ring also engages in a H-bond with Lys756. Regarding DDB1, UOM-005636’s phenyl ring establishes a cation–π interaction with Arg928 (green dashed line), and its pyridine ring filled DDB1’s pocket. (D) Overlay of the 3D representation of UOM-005197 (dark purple) and UOM-005636 (pink) bound to CDK12 and DDB1. UOM-005636 reaches into the pocket formed by DDB1 while UOM-005197 does not enter the DDB1 pocket. (E) Overlay of the 3D representation of S656 (turquoise) and UOM-005636 (pink) bound to CDK12 and DDB1 showing that UOM-005636 reaches further into the DDB1 pocket compared to S656. Hydrogen bonds: yellow dashed line, aromatic H-bonds: turquoise dashed line. See also Appendix Figs. S6 and S7.