PRMT1 promotes pancreatic cancer development and resistance to chemotherapy

Abstract

Pancreatic ductal adenocarcinoma (PDAC) remains one of the most lethal types of cancer, and novel treatment regimens are direly needed. Epigenetic regulation contributes to the development of various cancer types, but its role in the development of and potential as a therapeutic target for PDAC remains underexplored. Here, we show that PRMT1 is highly expressed in murine and human pancreatic cancer and is essential for cancer cell proliferation and tumorigenesis. Deletion of PRMT1 delays pancreatic cancer development in a KRAS-dependent mouse model, and multi-omics analyses reveal that PRMT1 depletion leads to global changes in chromatin accessibility and transcription, resulting in reduced glycolysis and a decrease in tumorigenic capacity. Pharmacological inhibition of PRMT1 in combination with gemcitabine has a synergistic effect on pancreatic tumor growth in vitro and in vivo. Collectively, our findings implicate PRMT1 as a key regulator of pancreatic cancer development and a promising target for combination therapy.

Keywords: PRMT1; chemotherapy resistance; chromatin accessibility; combination therapy; drug synergy; gemcitabine; glycolysis; histone methylation; pancreatic cancer; protein arginine methyl transferase 1; tumor metabolism.

Graphical abstract

Figure 1

PRMT1 is highly expressed in pancreatic cancer cells (A) Unsupervised clustering of all viable cells from the pancreas of WT (normal pancreas [NP], gray) and KP^fC (pink) mice represented as a t-distributed stochastic neighbor embedding plot (GSE125588). (B) Unsupervised clustering of all viable cells in (A), depicting the indicated cell populations. (C) Violin plot depicting the expression level of Prmt1 in the duct/cancer cell population of WT and KP^fC mice. (D) Representative H&E staining, Alcian blue staining, and CK19 and PRMT1 IHC staining in the pancreas of WT mice, as well as in ADM, PanIN, and PDAC lesions in the pancreas of KP^fC^ER mice. Scale bars, 100 μm. (E) Box and whisker plot of PRMT1 expression in human PAAD tumors (n = 179) and adjacent normal tissue (n = 171). ∗p < 0.05. (F) Kaplan-Meier analysis of overall survival in human PAAD patients with high PRMT1 (n = 44) versus low PRMT1 (n = 66) expression. (G) Bar graph showing cell viability in KP^fC^ER mouse pancreatic tumor-derived organoids after treatment with 50 μM FM or Veh for 48 h. n = 3 independent experiments in triplicate. ∗∗∗p < 0.001. (H) Representative H&E staining and IF staining for tdTomato and Ki67 in KP^fC^ER organoids (G). Scale bars, 50 μm. See also Figure S1.

Figure 2

Loss of PRMT1 impairs PDAC development (A) Genotype of KP^fC^ER;Prmt1 mice. (B) Experimental strategy for pancreas-specific genetic modification and analysis of KP^fC^ER and KP^fC^ER;Prmt1 mice. (C) Representative H&E staining, Alcian blue staining, and CK19 IHC staining in ADM, PanIN, and PDAC lesions in the pancreas of KP^fC^ER and KP^fC^ER;Prmt1 mice 2 months after the administration of tamoxifen. Scale bars, 100 μm. (D) Quantification of transformed lesions in the pancreas of KP^fC^ER and KP^fC^ER;Prmt1 mice 2 months after the administration of tamoxifen (n = 4 per group). Statistical analyses were performed by two-way ANOVA. ∗∗∗∗p < 0.0001 (E) Kaplan-Meier analysis of overall survival for KP^fC^ER (red) and KP^fC^ER;Prmt1 (black) mice (n = 15 per group). The hazard ratio and p value determined with the log rank test are shown. (F) Representative IF staining of Ki67, tdTomato, and CK19 in transformed lesions of the pancreas of KP^fC^ER and KP^fC^ER;Prmt1 mice 2 months after the administration of tamoxifen. (G) Quantification of Ki67⁺tdTomato⁺ cells in the pancreas of KP^fC^ER (n = 5) and KP^fC^ER;Prmt1 (n = 4) mice from images in (F). Data are for a total of 63 and 34 lesions, respectively. Scale bars, 100 μm. ∗∗∗p < 0.001. See also Figure S2 and Table S1.

Figure 3

Loss of PRMT1 inhibits glycolysis-related gene expression in PDAC cells (A) Immunoblot analysis of PRMT1, H4R3me2a, H4, and β-actin (loading control) in PANC-1 and MiaPaca-2 cells infected with lentiviral vectors for 2 independent PRMT1 short hairpin RNAs (shRNAs) or a control shRNA (shNTC). n = 3 independent experiments. (B) Time course of cell proliferation for cells as in (A). n = 3 independent experiments. (C) Colony formation assay for cells as in (A), with quantitative data (top) and representative images (bottom) being shown. n = 3 independent experiments. (D) Schematic representation of the multi-omics analysis. (E) Pie chart showing the distribution of 858 differentially expressed genes (p < 0.05) from the RNA-seq analysis of PRMT1-depleted versus control (shNTC) MiaPaca-2 cells. (F) Volcano plot showing key downregulated (blue dot), upregulated (black dot), and nonsignificant (gray dot) genes in PRMT1-depleted MiaPaca-2 cells. Vertical line indicates |log₂ FC| = 0.3785 and horizontal line indicates −log₁₀ (p value) = 1. Key downregulated genes are highlighted in blue boxes. (G) Pathway analysis for genes downregulated by the loss of PRMT1 as in (E). (H and I) GSEAPreranked enrichment plot showing the enrichment of hallmark glycolysis genes (H) and Kyoto Encyclopedia of Genes and Genomes glycolysis/gluconeogenesis genes (I) in control versus PRMT1-depleted cells. FDR, false discovery rate; NES, normalized enrichment score; NOM p, nominal p value. (J) Heatmap for the expression of selected genes related to glycolysis in control and PRMT1-depleted cells. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗∗p < 0.0001. See also Figure S3 and Table S2.

Figure 4

Multi-omics analysis reveals PRMT1-dependent changes in chromatin accessibility (A) Total number of peaks identified by ATAC-seq analysis of control (shNTC) and PRMT1-knockdown MiaPaca-2 cells. (B) Heatmap of normalized ATAC-seq intensity and signal distribution histogram around the peak center of ATAC-seq reads in control and PRMT1-depleted cells. (C) Profile plot of ATAC-seq read density across H3K27ac and H3K4me3 ChIP-seq peak centers. (D) Violin plot of mean read density of ATAC-seq for differentially accessible chromatin regions (p < 0.05). Horizontal lines indicate the median. Statistical analyses were performed by 2-tailed paired Student’s t test.∗∗∗∗p < 0.0001. (E) Pie chart showing the genomic distribution of PRMT1-dependent open chromatin regions (FC <−1.5, p < 0.05). (F) De novo motif analysis of differentially accessible chromatin regions. (G) Pathway analysis for 206 genes showing both downregulation in RNA-seq experiments and signal loss in ATAC-seq experiments using MSigDB_Hallmark_2020 in response to the depletion of PRMT1. (H and I) Representative Integrative Genomics Viewer tracks of ATAC-seq and H3K27ac ChIP-seq peaks in control (shNTC) and PRMT1-knockdown MiaPaca-2 cells aligned with H3K27ac and H3K4me3 ChIP-seq peaks (GSE29611) in HK2 (H) and SLC2A1 (I).

Figure 5

PRMT1 regulates glycolysis in PDAC (A and B) qRT-PCR analysis of Hk2 (A) and Slc2a1 (B) expression in the pancreas of WT (n = 4), KP^fC^ER (n = 7), and KP^fC^ER;Prmt1 (n = 7) mice. (C) Representative IHC staining of HK2 and GLUT1 in the pancreas of WT mice and in transformed lesions of the pancreas of KP^fC^ER and KP^fC^ER;Prmt1 mice. Scale bars, 100 μm. (D and E) Scatterplots showing positive correlation of HK2 (D) and SLC2A1 (E) with PRMT1 expression in human PAAD patients. (F) Representative IHC staining of PRMT1, GLUT1, and HK2 in 131 human PDAC tissue samples classified as Low, Med, or High according to the area and intensity of staining. Scale bars, 100 μm. (G) Quantification of PRMT1, GLUT1, and HK2 immunostainings in (F). PRMT1: High, 44.27%; Med, 45.04%; Low, 10.69%. GLUT1: High, 38.93%; Med, 40.46%; Low, 20.61%. HK2: High, 25.95%; Med, 35.11%; Low, 38.93%. (H and I) Scatterplots of IHC scores showing a positive correlation of HK2 (H) and GLUT1 (I) expression with PRMT1 expression in 131 human PDAC tissue samples in (F). (J) Representative profile of the glycolytic rate in control (shNTC) and PRMT1-depleted MiaPaca-2 cells (shPRMT1). Vertical lines indicate the addition of rotenone (Rot), antimycin A (AA), and 2-DG. ECAR, extracellular acidification rate. (K and L) Basal glycolysis (K) and compensatory glycolysis (L) rates in control (shNTC) and PRMT1-depleted MiaPaca-2 cells from (J). GlycoPER, glycolytic proton efflux rate. (M) Representative profile of the glycolytic rate in TCE-treated MiaPaca-2 cells. (N and O) Basal glycolysis (N) and compensatory glycolysis (O) rates in TCE-treated MiaPaca-2 cells from (M). (P and Q) Cell proliferation assay for MiaPaca-2 (P) and MPC-1 (Q) cells incubated in the presence of the indicated concentrations of 2-DG or vehicle for 48 h. n = 3 independent repeats. R, Pearson’s correlation coefficient. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. Statistical analyses were performed by 1-way ANOVA with Tukey’s multiple-comparison test (A and B), 2-way ANOVA with Tukey’s multiple-comparison test (J and M), or 1-way ANOVA with Dunnett’s multiple-comparison test (P and Q). See also Figure S4.

Figure 6

Inhibition of PRMT1 synergizes with the therapeutic effect of Gem ex vivo (A and B) Correlation analysis of Gem sensitivity score (AUCpi) and either PRMT1 expression (A) or the average expression of 200 glycolytic genes (B) in human PDAC organoids. The mean and 95% confidence interval are indicated by the line and shading, respectively. R, Pearson’s correlation coefficient. (C) Kaplan-Meier analysis of overall survival for human Gem-treated PAAD patients showing high PRMT1 (n = 33) versus low PRMT1 expression (n = 32). (D) Schematics showing the establishment of PDOs and calculation of the synergy score. (E–G) Dose-response matrix of the percentage inhibition of cell viability of PDO6 (E), PDO10 (F), and PDO11 (G) in the presence of the indicated drug combinations. Values in each block represent the mean cell viability inhibition with SD. (H–J) Three-dimensional (3D) plot showing the Loewe synergy score of pairwise dose combinations in PDO6 (H), PDO10 (I), and PDO11 (J). z_axis, synergy score; x/y_axis, Gem+FM combination range. (K and L) Dose-response matrix of the percentage inhibition of cell viability of PDO6 (K) and PDO11 (L) in the presence of the indicated drug combinations. (M and N) 3D plot showing the Loewe synergy score of pairwise dose combinations in PDO6 (M) and PDO11 (N). z axis, synergy score; x/y axis, Gem/TCE combination range. See also Figures S5 and S6.

Figure 7

Combination of PRMT1 inhibition and Gem impairs pancreatic tumor growth in vivo (A) Schematic representation of subcutaneous implantation and drug administration in mice. (B) Relative growth curves for tumors of mice from the onset of drug treatment. Veh (n = 8), FM (n = 8), Gem (n = 6), or both (Gem+FM, n = 8). (C) The volume of tumors isolated from mice after 25 days of drug administration. (D) Weight of tumors in (C). (E) Representative IHC staining of HK2 and GLUT1 from tumors in (C). Scale bars, 100 μm. (F and G) qRT-PCR analysis of Hk2 (F) or Slc2a1 (G) mRNA abundance from tumors in (C). n = 5 tumors. (H) Schematic representation of subcutaneous implantation and drug administration in mice. (I) Relative growth curves for tumors of mice from the onset of drug treatment. Vehicle (Veh, n = 6), TC-E 5003 (TCE, n = 7), Gem (n = 6), or both (Gem+TCE, n = 6). (J) The volume of tumors isolated from mice at 19 days of drug administration. (K) Weight of tumors in (J). (L) Representative IHC staining of HK2 and GLUT1 from tumors in (J). Scale bars, 100 μm. (M and N) qRT-PCR analysis of Hk2 (M) or Slc2a1 (N) mRNA abundance from tumors in (J). n = 4–5 tumors. (O) Representative IHC staining of Ki67 and cCasp3 from tumors in (J). Scale bars, 100 μm. (P and Q) Quantification of Ki67 (P) and cCasp3 (Q) in tumors in (J). (Data are for a total of 37, 26, 32 and 29 lesions, respectively). Horizontal lines indicate the median or quartiles. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. Statistical analyses were performed by 2-way ANOVA with the Holm-Šídák test (B and I) or 1-way ANOVA with Tukey’s multiple-comparison test (C, D, F, G, J, K, M, and N).