Homolog-Selective Degradation as a Strategy to Probe the Function of CDK6 in AML

Abstract

The design of selective small molecules is often stymied by similar ligand binding pockets. Here, we report BSJ-03-123, a phthalimide-based degrader that exploits protein-interface determinants to achieve proteome-wide selectivity for the degradation of cyclin-dependent kinase 6 (CDK6). Pharmacologic CDK6 degradation targets a selective dependency of acute myeloid leukemia cells, and transcriptomics and phosphoproteomics profiling of acute degradation of CDK6 enabled dynamic mapping of its immediate role in coordinating signaling and transcription.

Keywords: AML; CDK6; PROTAC; acute myeloid leukemia; molecular pharmacology; phosphoproteomics; selectivity; systems biology; targeted protein degradation; transcriptomics.

Figure 1.. Development of a homolog-selective degrader of CDK6.

(A) Structure of YKL-06–102. (B) Quantification of 5945 proteins following treatment with YKL-06–102 (500 nM, 5 h). FDR-adjusted p-values. Kinases inhibited by palbociclib in vitro are color-coded by magnitude of inhibition. (C) Structure of BSJ-03–123 (BSJ). (D) Immunoblot for CDK4, CDK6 and histone 3 after 4h BSJ treatment. (E) As in (D), but time-resolved at 200 nM BSJ. (F) Immunoblot for CDK6 and histone 3. 200 nM BSJ, 4h, in CRBN-deficient and wt MV4–11. (G) Quantification of 5995 proteins following BSJ treatment (100 nM, 2 h). FDR-adjusted p-values. Kinase labeling as in B. (H) Structure of BSJ-bump. (I) As (D), but for BSJ-bump. See also Figure S1 and Table S1–S2.

Figure 2.. BSJ-03–123 selectivity is explained by differential ternary complex formation.

(A) Immunoblot for CDK4 and CDK6 after thermal shift assay in MV4–11 CRBN^−/− cells after cellular treatment with DMSO, palbociclib, or BSJ (both at 20 µM). (B) Schematic representation of the luciferase complementation assay. (C) Measurement of CDK4:CRBN and CDK6:CRBN binding via NanoBiT® assay. Averages of 5 replicates are plotted, shaded areas represent 95% CIs. (D) Measurement of CDK6:CRBN complex formation after 1 h pre-treatment with 20 µM palbociclib or lenalidomide. Statistics as in (C).

Figure 3.. BSJ-03–123 exploits homolog-selective dependencies.

(A) Bottom: Waterfall plot of 391 cell lines ranked by CDK6 dependency as determined in genome-wide CRISPR/Cas9 screens. CERES essentiality score is normalized for copy number variation and scaled by setting the median of pan-essential genes to −1. Top: mRNA levels of CDK4, CDK6 and D-type cyclins. (B) Colony formation assays (12 days, refreshing the treatment every 2 d). (C) Growth curves of AML cell lines treated with 200 nM BSJ, BSJ-bump or DMSO. Cells were counted and treatment refreshed every 2 days. (D) Cell cycle after treatment with 200 nM BSJ or BSJ-bump for 24 h. (E) Percentage of apoptotic cells after 24 h treatment with 200 nM palbociclib, BSJ, BSJ-bump or DMSO (Caspase-Glo® 3/7 assay). BET protein degrader dBET6 served as positive control. (F) Growth curves of CDK4-deficient MV4–11 treated with 200 nM BSJ, palbociclib or DMSO. See also Figure S2.

Figure 4.. An integrated view of the effects of acute CDK6 degradation on cellular signaling and transcription.

(A) Immunoblot for p-RB S780, RB, CDK4, CDK6 and histone 3 after treatment with 200 nM BSJ, palbociclib or BSJ-bump. (B) Global phosphoproteomics. Heat map depicting fold changes in peptide phosphorylation after BSJ or palbociclib treatment (200 nM, 2 h) compared to DMSO for 305 differentially phosphorylated peptides. Hits (log2 FC > 0.5 or < −0.5, adj. p-value < 0.05) and peptides phosphorylated at canonical SP/TP CDK phosphorylation motifs are annotated. (C) Heat map of DMSO-normalized fold changes in gene expression after BSJ, YKL, or palbociclib treatment (200 nM, 6h) for 993 significantly deregulated genes (log2 FC > 1.5 or < −1.5, adj. p-value < 0.05). (D) Functional network of BSJ treatment. Nodes represent GO-terms enriched among genes that are differentially expressed upon treatment, scaled by magnitude and color coded by significance of enrichment. Edges represent parent-child relationships of GO-terms. Molecular network of BSJ treatment. Hits identified via global phosphoproteomics were mapped on a protein-protein interaction network and expanded to include first order neighbors limited to ENCODE transcriptional regulators. Node shape distinguishes transcriptional regulators (TR, diamonds) from phosphoproteomics hits (round). Node color represents the number of quantified phosphopeptides. Diamonds are scaled proportional to percentage of dysregulated TR target genes upon treatment. Proteins phosphorylated at CDK consensus motif (SP/TP) are annotated by edge color. Edges represent physical interaction between proteins. See also Figure S3–4.