Combination of the MTA-Cooperative PRMT5 Inhibitor BMS-986504 and KRAS Inhibitors Is an Effective Treatment Strategy for MTAP-Deleted KRAS-Mutant Pancreatic Cancer

Abstract

Protein arginine methyltransferase 5 (PRMT5) is a synthetic lethal target in MTAP-deleted (MTAP-del) cancers. The MTA-cooperative PRMT5 inhibitor BMS-986504 exhibited potent and selective antitumor activity in MTAP-del preclinical models and demonstrated activity in MTAP-del patients without the toxicity associated with previous PRMT5 inhibitors. In this study, we focused on pancreatic ductal adenocarcinoma (PDAC), ∼22% of which are MTAP-del, and demonstrated that BMS-986504 suppressed PRMT5 function and cell growth in MTAP-del cells and xenograft models. CRISPR/Cas9 loss-of-function screens implicated cotargeting KRAS as a combination strategy. Concurrent inhibition of PRMT5 and KRASG12C/D enhanced and prolonged suppression of PDAC growth. RNA sequencing analysis revealed that PRMT5 inhibition disrupted RNA splicing of genes essential for PDAC growth. Although PRMT5 and KRAS regulated distinct transcriptomes, they converged on pathways governing cancer cell growth and expression of PDAC-essential genes. These findings provide rationale for combined inhibition of PRMT5 and KRAS in MTAP-del/KRAS-mutant PDAC.

Significance: MTAP deletion and mutational activation of KRAS create therapeutic vulnerabilities for MTA-cooperative PRMT5 and mutant-selective KRAS inhibitors, respectively, providing the rationale for their combination therapy for MTAP-deleted, KRAS-mutant pancreatic cancer. See related article by Knoll et al., p. 3518.

Figure 1.

PRMT5 inhibitor BMS-986504 (PRMT5i) inhibits PRMT5 function and suppresses growth in MTAP-deleted (MTAP-del) PDAC cell models. A, Immunoblot for SDMA in MTAP WT and MTAP-del PDAC cell lines treated with PRMT5i (48 hours). Names are color-coded by MTAP and KRAS-mutant status. Black: MTAP WT, KRAS^G12D; blue: MTAP-del, KRAS^G12D; red: MTAP-del, KRAS^G12C; green: MTAP-del, KRAS^G12V; purple: MTAP-del, KRAS^G12R. B, Proliferation following dose-response of PRMT5i (5 days), represented as mean percent cell growth ± SD (n = 3–5 independent experiments). C, GI₅₀ values from B. D, 10- to 14-day clonogenic growth assay (images representative of n = 3 independent experiments). E, Cell cycle distribution of PRMT5i-treated cells (100 nM, 5 days) determined by flow cytometry, percent in each phase ± SD (n = 3–4). F, Apoptosis in PRMT5i-treated cells (5 days) determined by flow cytometry. Shown are percentages ± SD of early (annexin V-FITC-positive/PI-negative) and late (annexin V-FITC-positive/PI-positive) apoptotic cells (n = 3–4). Statistical significance by two-way ANOVA with Dunnett’s multiple comparisons test. *, p < 0.05; **, p < 0.01; ****, p < 0.0001.

Figure 2.

PRMT5i BMS-986504 demonstrates anti-tumor activity in KRAS-mutant, MTAP-del PDAC. A, PRMT5i or vehicle control were administered daily by oral gavage to mice bearing cell-line derived xenografts (CDX, n = 5 mice/group). B, PRMT5i or vehicle administered to mice bearing patient-derived xenografts (PDX, n = 3 mice/group). KRAS^G12D-mutant: PA1233, PA1338. KRAS^G12V-mutant: PA3060, PAXF1872. Graphs depict mean tumor volume ± SEM. Statistics by two-way ANOVA with Šidák’s multiple comparison. Last observation for vehicle was carried forward for SU8686, PAXF1872. *, p < 0.05; ****, p < 0.0001. C, Immunoblots of SDMA signal from end-of-study tumors from three independent mice/tumors from A, B.

Figure 3.

RPPA profiling identifies signaling changes following PRMT5i treatment. A, Volcano plots depict log2 fold-change (log2FC) versus adjusted p-values (adj. p-val.) for 191 antibodies in PANC-1, KP4, SU8686, and MIA PaCa-2 cells treated (DMSO or PRMT5i, 100 nM) for 24, 72, or 144 hours. Differential expression and significance were determined by linear modeling of all cell lines together and empirical Bayes moderation in limma. Shown are significantly (adj. p-val. < 0.05, dotted line) upregulated (red) or downregulated (blue) proteins across all cell lines. B, Heatmap shows log2FC of top 15 highest/lowest significantly dysregulated antibodies (adj. p-val. < 0.05) from A in individual lines (PRMT5i, 144 hours). C, Log2FC of select antibodies from B. Boxplots and significance values calculated across all lines. D, Immunoblots of phospho-/total ERK in cells treated as in B.

Figure 4.

CRISPR/Cas9 loss-of-function screens identify potential PRMT5i combination targets. A, Schematic of screens. Created in BioRender. Drizyte-Miller, K. (2025) https://BioRender.com/c93z956. B, Log2FC of genes with sgRNA depletion or enrichment in MIA PaCa-2 cells treated for two weeks in vitro (DMSO or PRMT5i, 100 nM). C, Log2FC of genes as in B from MIA PaCa-2 CDX treated for two weeks in vivo (vehicle or PRMT5i, 100 mg/kg). Highlighted: control genes (orange, blue) and potential combination target genes of interest (red).

Figure 5.

Combined inhibition of PRMT5 and KRAS effectively suppresses PDAC growth in vitro and in vivo. A, Cells were treated for 5 days with single agents or combined PRMT5i and G12Di (MRTX1133) / G12Ci (adagrasib). Data represented as mean percent cell growth ± SD (n = 3–4). Area under the curve (AUC) values for PRMT5i are plotted. Statistical significance determined by one-way ANOVA with Dunnett’s test. **, p < 0.01; ***, p < 0.001; ****, p < 0.0001. B, 10- to 14-day clonogenic assay from cells as in A (images representative of n = 3–4). C, Immunocompromised mice bearing KP4 xenografts were treated with single agents (PRMT5i daily, oral gavage; G12Di twice daily, intraperitoneally) or the combination (Comb.) for ~20–60 days. Graph depicts mean tumor volume ± SEM (n = 5/group). D, Graph depicts percent tumor growth inhibition or regression in individual tumors from C. Statistical significance determined by unpaired t-test. **, p < 0.01; ****, p < 0.0001.

Figure 6.

PRMT5i regulates alternative RNA splicing. A, Schematic illustrating experimental design: PANC-1 cells treated in duplicate with DMSO, PRMT5i (100 nM), G12Di (100 nM), or combination (Comb.; 100 nM each) for 24, 72, or 144 hours were analyzed by RNA-Seq. Some images created in BioRender. Drizyte-Miller, K. (2025) https://BioRender.com/26ksfid. B, Transcripts analyzed for alternative RNA splicing (AS) using SplAdder algorithm. Differential splicing refers to time-matched DMSO- and inhibitor-treated samples. Shown are global differential splice events at 144 hours; individual points represent events with FDR adj. p-val. <0.05. The largest magnitude log2FC values were thresholded for clarity of display. MXE, mutually exclusive exons; RI, retained introns; SE, skipped exons; A5SS, alternative 5’ splice site; A3SS, alternative 3’ splice site. C, Genes were ranked by median genetic dependency of 41 KRAS-mutant PDAC cell lines (Cancer Cell Line Encyclopedia, CCLE) in DepMap. Genes with PRMT5i-driven AS events were filtered by either log2FC>0.25 (SplAdder) or IncLevelDifference (change in proportion of reads representing splice event; rMATS) > 0.2 and minimum read coverage of 50 counts in PRMT5i alone or Comb. Adj. p-val. is Bonferroni scaled BH-adjusted p-value based on number of transcripts for genes identified in Ensembl, down-weighting differential splicing events with a higher likelihood of occurring by chance in larger, more complicated, and more variable transcripts. Genes with similar events in DMSO or G12Di alone were removed. Top 36 scoring genes (144 hours) are labeled. When more than one AS event was identified (e.g., AURKB), graph depicts only the most significant. D, AURKB genome locus plot showing read coverage (grey peaks) and read spanning (grey, cyan, and magenta lines) after PRMT5i versus DMSO (144 hours). Individual reads spanning exons 5–6 (cyan) and spanning the AS site of exons 5–7 (magenta) were based on read R1 or R2 when containing at least 10 bp overlap on each side of splice boundary. Below: Ensembl transcripts; colored IDs indicate support based on spanning reads. Red box: region with the most differential splicing events, beginning with alternative exon 5 splice site through exon 6. E, Immunoblot for AURKB in PDAC cell lines treated with DMSO, PRMT5i (100 nM), G12Di/G12Ci (100 nM), or both (Comb.; 100 nM each). F, relative protein levels in E were quantified by densitometry after normalizing to vinculin and then to DMSO. Shown are data ± SD (n = 3–6). Statistical significance determined by one-way ANOVA with Dunnett’s test. *, p < 0.05; **, p < 0.01; ***, p < 0.001; ****, p < 0.0001.

Figure 7.

Inhibition of PRMT5 and KRAS alters distinct PDAC transcriptomes and suppresses expression of PDAC essential genes. A, Graph depicts differentially expressed (DE) genes after PRMT5i, G12Di, or Comb. treatment. Positive values indicate upregulated (up) genes; negative values indicate downregulated (down) genes. B, Venn diagrams represent overlap of DE genes from A. C, GSEA for Hallmark, KEGG, and PDAC KRAS UP/DN gene sets within DE genes following Comb. treatment. Shown are gene sets with top 10 highest/lowest normalized enrichment scores (adj. p-val. < 0.05). D, Network analysis depicts distinct and shared gene clusters between PRMT5i, G12Di, and Comb. networks. Edges represent the product of |log2FC| and min-max scaled adj. p-val. (scaled log2FC), colored by treatment and direction of change (up/down). Edge cutoff was +/− one standard deviation of all edges. Nodes are genes or treatment conditions. Genes are colored and sized by median CRISPR dependency scores from 44 KRAS-mutant pancreatic cancer cells in DepMap. E, Isolated gene cluster downregulated only by Comb. treatment from D. Genes with median CRISPR dependency < −1.5 are labeled. F, Genes unique to Comb. down gene cluster (from E) were assessed by overrepresentation analysis of Hallmark, KEGG, and Reactome gene sets; top 15 most significant gene sets are shown. G, Heatmap depicts scaled log2FC values from the network shown in E. H, Immunoblot for p4EBP1 (S65), pS6 (S235/236), and total S6 levels in cells treated with DMSO, PRMT5i (100 nM), G12Di/G12Ci (100 nM), or combination (Comb.; 100 nM each).