Co-occurring mutations identify prognostic subgroups of microsatellite stable colorectal cancer

Abstract

Background: Co-occurring mutations in pairs of genes can pinpoint clinically relevant subgroups of cancer. Most colorectal cancers (CRCs) are microsatellite stable (MSS) and have few frequent mutations. Large patient cohorts and broad genomic coverage are needed for comprehensive co-mutation profiling.

Methods: Co-mutations were identified in a population-based Swedish cohort analyzed by whole-genome sequencing (n=819 stage I-IV MSS CRCs). Prognostic value was further evaluated in a publicly available dataset of clinically sequenced metastatic CRCs (MSK-IMPACT; n=934 MSS). Multivariable Cox proportional hazards analyses with clinicopathological parameters were performed for locoregional (stage I-III) and metastatic (stage IV and recurrent) cancers separately.

Results: Prevalent co-mutations were detected in 23 unique gene pairs, 20 of which included APC, TP53, KRAS and/or PIK3CA. Several co-mutations involving APC were associated with good overall survival in locoregional CRC, including APC-TCF7L2 (multivariable HR: 0.49, 95% CI 0.27-0.89). This co-mutation was prognostic also in metastatic cancers (multivariable HR: 0.49 and 0.37, 95% CI: 0.24-0.98 and 0.17-0.82 in the Swedish and MSK cohorts, respectively). APC-SOX9 co-mutations were mutually exclusive with APC-TCF7L2, and the co-mutations combined had stronger prognostic associations than APC alone in both metastatic cohorts. BRAF p.V600E-RNF43 co-mutations were associated with poor overall and recurrence-free survival in locoregional CRC (multivariable HR: 4.13 and 3.2, 95% CI: 1.78-9.54 and 1.53-8.04, respectively).

Conclusions: We report a genome-wide evaluation of co-occurring mutations in MSS CRCs, and suggest that co-mutations can improve the prognostic stratification compared to single mutations alone.

Keywords: CRC; Co-mutations; Locoregional; MSS; Metastatic; Prognosis; Survival.

Fig. 1

Comparison of clinicopathological parameters and mutations among cohorts. A Clinicopathological characteristics of the Swedish locoregional (n = 719; 1 patient with missing survival information), Swedish metastatic (n = 228; 128 patients with metachronous metastases overlap with the locoregional cohort), and MSK (all patients, n = 934) cohorts. B Kaplan–Meier plots of overall survival in each cohort. C-D Bar plots of the frequency of single (C) and co-occurring (D) gene mutations in each cohort. Fisher exact analyses were performed between the cohorts and statistically significant differences are indicated by asterisks (***p < 0.001, **p < 0.01, and *p < 0.05)

Fig. 2

Survival analyses of single mutations and co-occurring mutations in each cohort. A-B Forest plots for multivariable Cox proportional hazards models of overall survival according to each of the most frequent (A) single mutations and (B) co-mutations in each cohort. The reference group is cancers without mutations in the respective genes. Only co-mutations with significant prognostic associations (p < 0.05) in at least one cohort are shown. C-E Kaplan–Meier plots of overall survival according to selected co-mutations in each cohort. The wild-type group represents patients without mutations in any of the genes considered, and was used as reference group in statistical comparisons. Hazard ratios are from Cox proportional hazard analyses, and p-values from Wald’s tests. One patient in the Swedish locoregional cohort had missing survival information