Consensus molecular subtyping of metastatic colorectal cancer expands biomarker-directed therapeutic benefit for patients with CMS1 and CMS2 tumors

Abstract

Background: We developed a whole transcriptome sequencing (WTS)-based Consensus Molecular Subtypes (CMS) classifier using FFPE tissue and investigated its prognostic and predictive utility in a large clinico-genomic database of CRC patients (n = 24,939).

Methods: The classifier was trained against the original CMS datasets using an SVM model and validated in an independent blinded TCGA dataset (88.0% accuracy). Kaplan-Meier estimates of overall survival (OS) and time-on-treatment (TOT) were calculated for each CMS (p < 0.05 considered significant).

Results: CMS2 tumors were enriched on left-side of colon and conferred the longest median OS. In RAS-wildtype mCRC, left-sided tumors and CMS2 classification were associated with longer TOT with anti-EGFR antibodies (cetuximab and panitumumab). When restricting to only CMS2, there was no significant difference in TOT between right- versus left-sided tumors. CMS1 tumors were associated with a longer median TOT with pembrolizumab relative to other CMS groups, even when analyzing only microsatellite stable (MSS) tumors.

Discussion: A WTS-based CMS classifier allowed investigation of a large multi-institutional clinico-genomic mCRC cohort, suggesting anti-EGFR therapy benefit for right-sided RAS-WT CMS2 tumors and immune checkpoint inhibitor benefit for MSS CMS1. Routine CMS classification of CRC provides important treatment associations that should be further investigated.

Fig. 1. Development of Caris Consensus Molecular Subtypes (CMS) classifier.

Flowchart illustrating the number of samples and models tested producing the indicated accuracy levels. The table shows sensitivity and specificity of the final selected model utilizing 600 genes per subtype among Caris CRC cohort.

Fig. 2. Molecular and clinical characteristics of study cohort by CMS classifier.

a Stacked bar charts show percentage of each CMS in all CRC (n = 24,393), left-sided (n = 7490), right-sided (n = 7323), liver (n = 4901), and lung metastases (n = 1278). b Prevalence of top alterations analyzed by next-generation sequencing in each CMS. q-values determined by Chi-square test. c Prevalence of top alterations in each CMS group for liver metastases, lung metastases, and all CRC. p values were calculated by Chi-square or Fisher exact tests. *p < 0.05; **p < 0.005; ***p < 0.0005; ****p < 0.0001. TMB-H tumor mutational burden=high, MSI-H microsatellite instability-high, dMMR mismatch repair deficient.

Fig. 3. Overall survival (OS) of CRC cohort.

a–c Kaplan–Meier curves depicting OS for each CMS among entire cohort (n = 24,939; a), among patients with primary or local tumors used for CMS classification (n = 14,153; b), and among patients with distant metastases used for CMS classification (n = 10,776, c). OS was calculated from time of tissue collection to last contact. d Forest plot shows log2 hazard ratios relative to CMS2 (circles) with 95% confidence intervals (whiskers), and p-values for the indicated cohorts.

Fig. 4. CMS is a predictive biomarker for anti-EGFR antibody treatment.

a Kaplan–Meier curve showing TOT with cetuximab or panitumumab for each CMS among RAS-wildtype (WT) mCRC. b Forest plot shows log2 hazard ratios relative to CMS2 (circles), 95% confidence intervals (whiskers), and p-values. Kaplan–Meier curves showing time-on-treatment (TOT) with cetuximab or panitumumab for left- and right-sided tumors among all RAS-WT mCRC (c) or only CMS2 RAS-WT mCRC (d). Given the sample size, the power to detect 0.7 HR (2-sided alpha of 0.05) is 85.6%. e Median TOT (black midlines) for left- and right-sided tumors among each CMS with 95% confidence intervals (boxes). *p < 0.05.

Fig. 5. CMS is a predictive biomarker for pembrolizumab treatment.

Kaplan–Meier curves showing TOT with pembrolizumab for each CMS among all mCRC (a) and microsatellite stable (MSS) CRC (b). c, d Forest plots show hazard ratios relative to CMS1 (circles), 95% confidence intervals (whiskers), and p-values. e, f Median IFNγ scores (black midlines) for each CMS among all mCRC (d) and MSS mCRC (e) with 95% confidence intervals (boxes). ****q < 0.0001. Kaplan–Meier curves show time-on-treatment (TOT) with pembrolizumab comparing MSS CMS1 versus MSS CMS2–4 in TMB-high (g) and TMB-low (h) cohorts.