Vishnu Kumarasamy

Activating mutations in KRAS drive pancreatic ductal adenocarcinoma (PDAC) and non–small cell lung cancer (NSCLC). Although mutant-selective KRAS inhibitors and pan-RAS inhibitors provide clinical benefits, the development of resistance limits durable responses. In this study, transcriptomic and proteomic analyses revealed that, despite effective suppression of mutant KRAS signaling, resistant cells sustain cell-cycle progression. Distinct orthogonal mitogenic pathways were engaged in a context-dependent manner to bypass KRAS inhibition. Although these pathways could be broadly inhibited using the pan–RAS-ON inhibitor RMC-6236, cells remained capable of developing acquired resistance, in which cell proliferation was uncoupled from RAS signaling. Combinatorial drug screens and genome-wide CRISPR–Cas9 screens showed that perturbing cell-cycle nodes via targeting cyclin-dependent kinases CDK4/6 and CDK2 restored sensitivity to KRAS/RAS inhibitors. Cotargeting CDK4/6 induced G1 arrest and suppressed E2F-regulated proteins across all resistant models. In contrast, cotargeting CDK2 exerted a broader effect by impairing DNA replication, inducing G2 arrest, preventing mitotic entry, and yielding a more durable cytostatic response that delayed cellular outgrowth after drug withdrawal. Finally, concurrent inhibition of KRAS with either CDK4/6 or CDK2 yielded durable tumor control in vivo in xenografts derived from models with acquired resistance. In conclusion, these findings identify sustained cell-cycle activity as a defining feature of resistance to KRAS-directed therapies and establish cell-cycle cotargeting as an effective strategy to overcome KRAS/RAS inhibitor resistance. Significance: Acquired resistance to KRAS/RAS inhibitors limits clinical benefit. We identify active cell cycle machinery as a defining feature of resistance and demonstrate distinct combinatorial strategies to target cell cycle progression.

Figure S5. Cooperative and antagonistic combination strategies with BLU-222. A, BLISS synergistic analysis for antagonism between BLU-222 and CHK1 checkpoint kinase inhibitors in MDA-MB-231 cells (n = 3 per group; BLISS scores calculated with SynergyFinder v3). B, Live cell proliferative monitoring of the indicated cell lines treated with BLU-222, palbociclib, or in combination (n = 3–4 per group; one-way ANOVA; error bars represent SEM). C, Quantification of mVenus-DHB and mCherry-RB CDK activity sensor localization to the cytosol or nucleus in MDA-MB-231 and HCC-1806 cells following treatment with either 500 nM BLU-222, palbociclib, or combination. D, Quantification of mVenus-DHB and mCherry-RB CDK activity sensor localization to the cytosol or nucleus in MDA-MB-231 and HCC-1806 cells following treatment with siCCNE1, siCCND1 or combination. E, Live-cell imaging endpoint stills for HCC-1806 cells treated for 5 days with 500 nM BLU-222, palbociclib, or combination (scale bar = 100 µm). ns, not significant; *, P < 0.05; **, P < 0.01; ****, P < 0.0001.

Figure S7. Prevalence of Cyclin E1-P16INK4Ahigh tumors that express RB. A, Kaplan Meier curves based on CCNE1 status generated from The Cancer Genome Atlas (TCGA) serous ovarian carcinoma database. HR, hazard ratio; DFS, disease-free survival; OS, overall survival. Amp, amplification. B, Oncoprint of the serous ovarian carcinoma TCGA database for the indicated cell cycle genes. C, Normalized intensity for Cyclin D1, Cyclin E1, or P16INK4A expression within clusters 1, 3, and 4 (n = 82002 for cluster 1, n = 150037 for cluster 3, and n = 34088 for cluster 4; unpaired t-test between clusters 1 and 4). D, SCIMAP analysis of unique spatial single cell protein expression signatures of all 159 cores from the patient-derived gynecological tumor microarray (TMA). E, Computational algorithm-based identification of Cyclin E1high/P16high/RB+ cores from the gynecological TMA expected to be sensitive to CDK2 inhibition.

Supplemental File 1. BLU-222 combinatorial drug screen data from MDA-MB-231 (tab 1) and COV-504 cells (tab 2) reported in Figure 5. Data reported as percent growth compared to sole treatment with 0.1% DMSO.

Differential vulnerabilities to CDK2 versus CDK4/6 inhibition. A, Table of top identified predictive genes in breast and ovarian carcinomas for response to CDK2 inhibition from the publicly available DepMap dataset using CRISPR (top) and RNAi (bottom) screens. B, Cell Titer Glo values normalized to 0 nM treatment of ovarian and breast cancer cells treated with BLU-222 for 5 days. C, Table summarizing breast and ovarian cancer cell line response to BLU-222 treatment. D, COV-504 and PE-O1 ovarian cancer cell lines treated with serial concentrations of BLU-222 or palbociclib for 5 days and monitored with Cellcyte or Incucyte live cell proliferative software (n = 3–4 per group; error bars represent SEM).

Figure S4. Loss of RB1 shifts phase of arrest and sensitivity to BLU-222. A, Heatmap of RNA sequencing data comparing differential gene expression between BLU-222-treated Kuramochi (125 nM BLU-222) and OVCAR-8 (250 nM BLU-222) cells normalized to respective DMSO-treated controls. B, Cell cycle profiling from univariate flow cytometric analysis from OVCAR-8 cells at baseline, after 1 to 5 days of treatment with 500 nM BLU-222, and after 1 to 5 days of drug washout. C, Western blot for abrogation of G2/M checkpoint signaling in OVCAR-8 cells following doxycycline (Dox) induction of the RB1 expression vector and treatment with BLU-222 compared with RB1 deficient controls. D, Live cell proliferative monitoring of OVCAR-8 cells following Dox induction of the RB1 expression vector (RB1 Dox +) with non-induced controls (RB1 Dox -) and treatment with serial concentrations of palbociclib (n = 3–4 per group; error bars represent SEM). E, Table summarizing effect of RB1 induction on IC50 to BLU-222 in OVCAR-8 cells. F, Live cell proliferative monitoring of OVCAR-8 cells following doxycycline induction of the RB1 expression vector (RB1 Dox +) with non-induced controls (RB1 Dox -) and transfection with siNT, siCCNE1, or siCCNA2 [n = 3–4 per group; two-way ANOVA, significance (shown as #) for combined RB1 and siRNA effect; error bars represent SEM]. G, Live-cell imaging endpoint stills for the OVCAR-8 cells in D (scale bar = 100 µm). H, Western blot of RB1-E2F pathway activity in Kuramochi cells with RB1 knockdown or siNT control and treatment with BLU-222. I, Cell cycle profiling from univariate flow cytometric analysis from Kuramochi cells with RB1 knockdown or siNT control and treatment with BLU-222. J, Western blot of RB1-E2F pathway activity in MDA-MB-157 cells with and without deletion (DEL) of RB1 and treatment with BLU-222. K, Cell cycle profiling from univariate flow cytometric analysis from MDA-MB-157 cells with and without deletion (DEL) of RB1 and treatment with BLU-222. L, Table summarizing effect of RB1 knockdown/deletion on IC50 to BLU-222 in Kuramochi and MDA-MB-157 cells, respectively. ####, P < 0.0001. PI, propidium iodide.

Figure S2. Contextual biochemical response to CDK inhibition. A, Western blot of RB1-E2F pathway activity in OVSAHO and MDA-MB-157 cells following BLU-222 or palbociclib treatment. B, Schematic depicting mVenus-DHB and mCherry-RB CDK activity reporter constructs. C, Representative images of mCherry-RB sensor activity in OVCAR-3 and T-47D cells treated with 500 nM BLU-222 or palbociclib (scale bar = 100 µm). D, Quantification of mCherry-RB CDK activity sensor localization to the cytosol or nucleus in the indicated ovarian and breast cancer cell lines following treatment with either 500 nM BLU-222 or palbociclib. E, Representative images of mCherry-RB sensor activity in OVCAR-3 cells transfected with siNT, siCCNE1 or siCCND1 (scale bar = 100 µm). F, Quantification of mCherry-RB CDK activity sensor localization to the cytosol or nucleus in OVCAR-3 and MDA-MB-157 cells following knockdown of CCNE1 or CCND1 with siRNA constructs.

Supplemental Figure S1-7 Legends

Figure S3. BLU-222 elicits potent G1 arrest in Cyclin E1 and P16INK4A-high models expressing RB. A, Cell cycle profiling from univariate flow cytometric analysis from the indicated cell lines following treatment with palbociclib. B, Bivariate flow cytometric analysis from the indicated cell lines following transfection with siNT or siCCNE1 constructs. C, RNA sequencing gene set enrichment plots for E2F targets and interferon alpha response pathways in Kuramochi and MDA-MB-157 cells following treatment with 125 nM BLU-222 or 100 nM, respectively, for 48 hours compared to DMSO-treated controls. PI, propidium iodide; FITC, anti-bromodeoxyuridine fluorescein.

Figure S6. P16INK4A is a key determinant of response to BLU-222. A, Heatmap analysis from the cancer dependency map (DepMap) database displaying relative expression of CCNE1 and several endogenous CDK inhibitor genes in a panel of established ovarian and breast cancer cell lines. B, Western blot of RB1-E2F pathway activity in MDA-MB-157 cells following overexpression (OE) of CDK4 with and without BLU-222 treatment. C, Live cell proliferative monitoring of CDK4 WT (left) and overexpressing (OE, right) of MDA-MB-157 cells treated with serial concentrations of BLU-222 [n = 3–4 per group; two-way ANOVA, significance (shown as #) comparing CDK4 OE versus CDK4 WT at 62.5 nM BLU-222 dose; error bars represent SEM]. D, Western blot of RB1-E2F pathway activity in MDA-MB-157 cells following overexpression (OE) of CDK6 with and without BLU-222 treatment. E, Live cell proliferative monitoring of CDK6 WT (left) and overexpressing (OE, right) MDA-MB-157 cells treated with serial concentrations of BLU-222 [n = 3–4 per group; two-way ANOVA, significance (shown as #) comparing CDK6 OE versus CDK6 WT at 250 nM BLU-222 dose; error bars represent SEM]. F, Table summarizing effect of CDK4 or CDK6 overexpression (OE) or CDKN2A knockdown on IC50 to BLU-222 in Kuramochi and MDA-MB-157 cells. G, Hematoxylin and Eosin (H&E) staining and of tumors excised from MDA-MB-157 xenografts in NOD scid gamma (NSG) mice treated with either vehicle or BLU-222; arrows indicate cells with enlarged morphology (scale bar = 100 µm). #, P < 0.05.