Fig 2: SMARCA4-deficient SCCOHT cells are vulnerable to inhibition of CDK4/6 kinase activities. a Schematic outline of the shRNA screens for kinases whose inhibition is selectively lethal to SMARCA4-deficient SCCOHT cells (BIN-67) but not to SMARCA4-proficient control cells (IOSE80, OVCAR4). Cells were infected with the lentiviral shRNA library (T0) and cultured for selection for 14 days (T1). The relative abundance of shRNAs in the cell populations was determined by next-generation sequencing. b Analysis of the shRNA screens using the MAGeCK statistical software package31. CDK6 (magenta) and CDK4 (blue) are the first two ranked genes that were negatively selected in BIN-67 cells. All genes were ranked based on their RRA (robust rank aggregation, top) or raw p values (bottom) generated from the MAGeCK analysis. c, d Validation of CDK6 and CDK4 in SCCOHT cells (BIN-67, SCCOHT-1, COV434) and SMARCA4-proficient controls (IOSE80, OVCAR4). c Colony-formation assay of the indicated cell lines expressing pLKO control or shRNAs targeting CDK6 or CDK4 after 10–15 days of culturing. For each cell line, all dishes were fixed at the same time, stained, and photographed. d Western blot analysis of CDK6 and CDK4 and phosphorylated RB at serine 795 (pRB-S795) in the cells described in c. HSP90 was used as a loading control. e–j SCCOHT cells are more vulnerable to inhibition of CDK4/6 kinase activities, compared to SMARCA4-proficient control cells. e BIN-67 cells stably expressing pLX304-GFP, pLX304-CDK6, or pLX304-CDK6D163N were infected with viruses containing pLKO control or a shRNA targeting the 3’UTR of CDK6, selected for integration, and cultured for 14 days. All dishes were fixed at the same time. f Western blot analysis for CDK6, pRB-S795, and HSP90 in the cells described above. g, i BIN-67 (g) and OVCAR-4 (i) cells expressing pLX317-GFP, pLX317-CDK4, or pLX317-CDK4D158N were infected with viruses containing pLKO control or a shRNA vector targeting the 3’UTR of CDK4, selected for integration, and cultured for 14 days. For each cell line, all dishes were fixed at the same time. h, j Western blot analysis for CDK4, pRB-S795, and HSP90 in the cells described above

Fig 3: Compounds targeting the CDK6–INK4 complex inhibit CDK4/6i-resistant tumors. A, Immunoblotting of MCF7 parental cells and cells with high CDK6 expression [CDK6-overexpressing (OE) cells and CDK6-high cells with FAT1 loss] treated for 24 hours with increasing concentrations of bifunctional degrader compound, BSJ-03-123, demonstrating dose-dependent targeting of CDK6 but not CDK4. B, Assessment of a panel of degrader compounds that target CDK4 and/or CDK6. Immunoblotting after 24-hour drug treatment (500 nmol/L) in FAT1-loss cells shows varying selectivity for CDK4 versus CDK6. Representative blots from three independent experiments are shown. Among them, BSJ-05-017 and BSJ-03-096 show the most significant degradation of both CDK4 and CDK6. C, Immunoblot depicting dose–response effects of BSJ-05-017 in both CDK4/6i-sensitive (left) and CDK4/6i-resistant (right) cells in comparison with palbociclib (500 nmol/L) after 24-hour treatment. D, Percentage of growth plot showing that BSJ-05-017 inhibits sensitive MCF7 parental and resistant CDK6-high cells with equal potency, whereas palbociclib shows only partial inhibition of resistant cells. IC50 values were recorded at day 7. Data are shown as mean ± SD; n = 6. E, Assay for drug-induced senescence (Senescence Green) demonstrating number of senescence marker–positive cells induced by 8 days of treatment with DMSO, BSJ-05-017 (500 nmol/L), abemaciclib (100 nmol/L), and palbociclib (500 nmol/L). BSJ-05-017 induced a significantly higher number of cells into senescence compared with abemaciclib or palbociclib in CDK6-high cells. F, Immunoblotting showing the degradation of CDK4/6 and decreased phospho-RB1 and E2F1 levels in CDK6-high (FAT1 loss) tumor-bearing mice administered 25 mg/kg BSJ-05-017 intraperitoneally. Tumors were collected 6 hours after 3 consecutive days of vehicle or BSJ-05-017 treatment (n = 2). G, Growth curve plots of cell-derived xenografts of MCF7 parental, CDK6-overexpressing, and PTEN-loss cells. Mice were treated with vehicle, ribociclib (25 mg/kg, orally), BSJ-05-017 (50 mg/kg, i.p.), or BSJ-03-096 (50 mg/kg, orally) daily for 25 to 35 days. Tumor volumes were recorded every 3 to 4 days. Data are shown as mean ± SD; n = 4. See also Supplementary Fig. S4.

Fig 4: SCCOHT cells are highly sensitive to CDK4/6 inhibitors. a–c SMARCA4-deficient SCCOHT cells are highly sensitive to palbociclib treatment, similar to ER+ breast cancer cells. a Colony-formation assays in SCCOHT (BIN-67, SCCOHT-1, and COV434), SMARCA4-proficient ovarian (IOSE80, OVCAR4, and OVCAR8), and ER+ breast cancer (MCF7 and CAMA-1) cell lines. Cells were cultured in the absence or presence of palbociclib at the indicated concentrations for 10–21 days. For each cell line, all dishes were fixed at the same time. b Cell viability assay of the same cell line panel. Cells were treated with increasing concentrations of palbociclib for 5–10 days, and cell viability using CellTiter-Blue was determined by measuring the fluorescence (560/590 nm) in a microplate reader. Error bars: mean ± standard deviation (s.d.) of biological replicates (n = 4). c Palbociclib treatment suppresses RB phosphorylation in SCCOHT cells similar to ER+ breast cancer cells but not in IOSE80 and OVCAR4 cells. Levels of pRB-S795 in cells treated with 0, 100, and 300 nM of palbociclib for 24 h were documented by western blot analysis. d, e Transcriptome profiling in SCCOHT cells show that top ranked pathways affected upon palbociclib treatment are related to cell cycle regulation. RNA-Seq was performed in BIN-67 and SCCOHT-1 cells treated with 100 nM of palbociclib for 24 h. Common genes that significantly changed (p < 0.05) in both cell lines was analyzed using Gene Set Enrichment Analysis (GSEA). Top ten cellular processes by Gene Ontology (GO) term (d) and GSEA plots for the top three cellular processes (e) are shown. NES normalized enrichment score, FDR false discovery rate. f, g Transcriptome profiling in SCCOHT cells show that top ranked pathways affected upon CDK6 knockdown are related to cell cycle regulation. RNA-Seq was performed in BIN-67 and SCCOHT-1 cells expressing pLKO control or two independent shRNAs targeting CDK6. Common genes that significantly changed (p < 0.05) in both shRNAs and both cell lines were analyzed using GSEA. Top ten cellular processes by GO term (f) and GSEA plots for the top three cellular processes (g) are shown. NES normalized enrichment score, FDR false discovery rate

Fig 5: INK4–CDK6 complex promotes resistance to CDK4/6i in cells. A, Schematic for analysis of CDK4 and CDK6 interactions and activity via coimmunoprecipitation (co-IP) followed by ADP-Glo kinase assays and mass spectrometry. B, ADP-Glo kinase assay showing immunoprecipitated CDK4 and CDK6 (IP-CDK4 and IP-CDK6) kinase activity from MCF7 parental and CDK6-high cells [cells with FAT1 CRISPR knockout (CR) that have high CDK6 expression, previously shown to have resistance to CDK4/6i; ref. 8], with or without 100 nmol/L abemaciclib treatment. Data are shown as mean + SD of three biologically independent samples. P values were determined by unpaired two-sided Student t test. RLU, relative luminescence units. C, Venn diagram showing the number of unique proteins identified by mass spectrometry coimmunoprecipitated from IP-CDK4 and IP-CDK6 in FAT1-loss cells. Percentages were calculated by number of proteins identified in each subgroup divided by total proteins identified by IP of either CDK4 or CDK6. Data are shown as means of three replicates. D, Pathway analysis by Gene Ontology of proteins interacting with CDK6 but not CDK4 in the FAT1-loss cells. The proteins were grouped by their putative biological functions. E, Unique peptide counts of cyclin-dependent kinases and their endogenous inhibitor proteins identified in the co-IP/mass spectrometry associated with CDK4 or CDK6 in the FAT1-loss cells. N = 2. F, Co-IP and immunoblotting reveal association of p15INK4B and p18INK4C with CDK6, but not CDK4, in CDK6-high cells. G, Cell line screening results showing that models with high CDKN2A or low RB1 mRNA expression are correlated with poor response to palbociclib. H, Interface residues in CDK6 in close proximity with INK4 isoforms based on previous INK4-bound CDK6 structures in the Protein Data Bank (ref. 65; no available structure for p15INK4B). CDK6-HA was immunoprecipitated using HA beads in parental MCF7 cells and MCF7 cells expressing HA-WT-CDK6-, HA-V16D-, and R31C-mutant CDK6 (disrupted INK4–CDK6 interaction) or HA-K43M/D163N-mutant CDK6 (kinase dead), and interaction with INK4 proteins was determined by immunoblotting. I, Disruption of INK4s and CDK6 binding or impairment of CDK6 kinase activity restores the sensitivity of CDK6-overexpressing cells to CDK4/6i. Cells were treated with DMSO or 100 nmol/L abemaciclib for 24 hours prior to collection. J, Percentage of cell viability of cells overexpressing WT CDK6 or R31C- or D163N-mutant CDK6 treated with increasing concentrations of abemaciclib compared with parental cells. IC50 values were recorded on day 5 following treatment. Data are shown as mean ± SD; n = 6. K, Knockdown of p15INK4B and p18INK4C in FAT1-loss cells promotes suppression of RB phosphorylation in response to abemaciclib to a similar extent as in parental cells. Cells were collected 24 hours after 100 nmol/L abemaciclib treatment. Representative blots are shown, which were repeated independently three times. L, The growth rate of p15INK4B and p18INK4C knockout in FAT1-loss cells was inhibited by 100 nmol/L abemaciclib. The cell viability was recorded at day 14 and day 21. ****, P < 0.0001. Data are shown as mean ± SD; n = 6. See also Supplementary Fig. S1.