Molecular Characterization and Clinical Relevance of MGMT-Silenced Pancreatic Cancer

Abstract

Background: The identification of actionable molecular targets of pancreatic cancer (PAC) is key to improving patient outcomes. We hypothesized O6-methylguanine-DNA methyltransferase (MGMT) silencing may occur in a subset of PAC tumors, with unexplored clinical and molecular correlates.

Experimental design: We leveraged sequencing data from The Cancer Genome Atlas (TCGA), the Clinical Proteomic Tumor Analysis Consortium 3 (CPTAC-3), and the (Australian Pancreatic Cancer Genome Initiative) PACA-AU cohorts to characterize MGMT-silenced PAC. Genomic, transcriptomic, methylation, and clinical data were investigated, and findings were validated in silico using methylation, transcriptomic and drug sensitivity data from Cancer Cell Line Encyclopedia (CCLE) project, and in a real-world cohort of PAC patients profiled for MGMT status at Istituto Nazionale Tumori of Milan (INT).

Results: On the basis of Human Methylation 450k data, MGMT silencing was identified in ~6% of PAC cases and was enriched in tumors with non-ductal histology, with a trend toward longer overall survival. MGMT-silenced tumors were associated with a lower frequency of KRAS mutations and showed features of immune exclusion. In the INT cohort, MGMT-silencing was confirmed in ~7% of cases and prevalent in KRAS wild type tumors, with a favorable prognostic impact. In silico analysis suggested a higher sensitivity to alkylating and DNA damaging agents in MGMT-silenced PAC cell lines.

Conclusions: MGMT silencing occurs in a small subgroup of PACs and is enriched in KRAS wild type cases, with a favorable prognostic impact. Our findings provide the rationale to explore combinations of alkylating with DNA damaging agents in MGMT-silenced PAC.

Keywords: KRAS wild type; MGMT; biomarker; molecular profiling; pancreatic cancer; temozolomide.

FIGURE 1

Transcriptomic features of MGMT‐silenced PAC. (a) Volcano plot of differentially expressed genes in MGMT‐methylated versus not‐methylated PAC. The vertical dotted lines identify genes with absolute log2FoldChange > 0.58 (i.e., absolute fold change > 1). The horizontal dotted line identifies genes with adjusted p < 0.01. Each point represents a gene, and genes are colored based on up (red) and downregulation (yellow) in MGMT‐methylated versus not‐methylated groups. (b) Bar plot reporting results of GSEA performed on up‐ and downregulated genes in MGMT‐methylated versus not‐methylated PAC. The color of the barplot reports the significance of the p value of the hypergeometric tests (after Benjamini–Hochberg multiple test correction). Only pathways with adjusted p < 0.1 are reported. Most pathways show a negative enrichment score, representing a negative enrichment in MGMT‐methylated tumors. (c) Boxplots comparing ESTIMATE deconvolution scores between MGMT‐methylated versus not‐methylated PAC. Wilcoxon mean rank‐sum p values are shown. (d) Overlay of PAC cases according to MGMT‐methylation status (inner ring) with published PAC transcriptomic subtypes (outer rings), according to Moffitt, Collisson, and Bailey classifications. ADEX, Aberrantly Differentiated Endocrine Exocrine; GSEA, Gene Set Enrichment Analysis; MGMT, O⁶ ‐methylguanine‐DNA methyltransferase; PAC, pancreatic cancer.

FIGURE 2

Genomic features of PAC cases according to MGMT methylation status. (a) Oncoprint summarizing somatic SNVs/indels according to MGMT methylation status. The top 20 most frequently mutated genes are presented. Upper bars represent TMB levels. MGMT‐methylated tumors are shown on the far right. (b) Bar plot reporting results of enrichment analysis of SNVs/indels according to MGMT methylation status. On the x axis, enrichment score is reported as the z‐score scaled‐, natural logarithm‐transformed odds ratio of the comparison between MGMT‐methylated versus not‐methylated cases. Bars are colored according to the enrichment score and adjusted (after Benjamini–Hochberg multiple test correction) p value as orange (significantly enriched in not‐methylated, enrichment score ≤ 0 and adjusted p ≤ 0.05), yellow (enriched in not‐methylated, enrichment score < 0 and adjusted p value > 0.05), light red (enriched in methylated, enrichment score > 0 and adjusted p > 0.05), and dark red (significantly enriched in methylated, enrichment score > 0 and adjusted p ≤ 0.05). (c, d) Barplots reporting results of oncogenic signaling pathways analysis according to MGMT methylation status. Fraction of each affected pathway (c) and fraction of samples affected (d). The NRF2 pathway is not plotted as it was not affected in any cases. The * identifies the only significant different, as evaluated by Fisher's exact test. Del, Deletion—Ins, Insertion; MGMT, O⁶ ‐methylguanine‐DNA methyltransferase; NRF2, Nuclear Respiratory Factor 2; TMB, tumor mutational burden.

FIGURE 3

Prognostic impact of MGMT‐silencing in the public cohorts and in the INT validation cohort. (a) Overall survival represented through Kaplan–Meier curves according to MGMT methylation status in the pooled training cohort. Survival data are missing for one case among non‐methylated patients. (b) Forest plot of multivariable Cox regression analysis for overall survival adjusting for patients' age, sex, tumor histology, and tumor stage at diagnosis. (c) Overall survival represented through Kaplan–Meier curves according to MGMT methylation status in the INT cohort. (d) Forest plot of multivariable Cox regression analysis for overall survival adjusting for patients' age, sex, ECOG PS, and tumor stage at diagnosis. The hazard ratios are shown with 95% confidence intervals, and Wald p values are reported on the far right. Patient with MGMT‐methylated tumors show better overall survival after adjustment for other covariates. BR, Borderline Resectable; ECOG PS, Eastern Cooperative Oncology Group Performance Status; LAD, Locally Advanced Disease; MGMT, O⁶ ‐methylguanine‐DNA methyltransferase.