The Colorectal Cancer Subtyping Consortium was formed to assess the presence or absence of core subtype patterns among existing gene expression–based colorectal cancer subtyping algorithms. The researchers also wanted to characterise the key biological features of the core subtypes, integrate and confront all other available data sources (mutation, copy number, methylation, microRNA and proteomics) and assess whether the subtype assignment correlated with patient outcome. Furthermore, their aim was to establish an important paradigm for collaborative, community-based cancer subtyping that will facilitate the translation of molecular subtypes into the clinic, not only for colorectal cancer, but for other malignancies as well.
The colorectal cancer subtypes are:
- CMS1 (microsatellite instability immune, 14%), hypermutated, microsatellite unstable and strong immune activation;
- CMS2 (canonical, 37%), epithelial, marked WNT and MYC signaling activation;
- CMS3 (metabolic, 13%), epithelial and evident metabolic dysregulation; and
- CMS4 (mesenchymal, 23%), prominent transforming growth factor–β activation, stromal invasion and angiogenesis.
Samples with mixed features (13%) possibly represent a transition phenotype or intratumoral heterogeneity.
Clinical and prognostic associations of the consensus molecular subtypes
The consortium also found important associations between the CMS groups and clinical variables. CMS1 tumours were frequently diagnosed in females with right-sided lesions and presented with higher histopathological grade.
Conversely, CMS2 tumours were mainly left-sided and, whereas CMS4 tumours tended to be diagnosed at more advanced stages (III and IV).
To determine whether the CMS groups differed in outcome, the researchers performed a Cox proportional hazards analysis on the combined data sets and separately in the subset of patients enrolled in a clinical trial with uniform follow-up (PETACC-3 clinical trial).
Irrespective of patient cohort, CMS4 tumours resulted in worse overall survival and worse relapse-free survival in both univariate and multivariate analyses, after adjustment for clinico-pathological features, MSI status and presence of BRAF or KRAS mutations.
The researchers also found superior survival rates after relapse in CMS2 patients, with a larger proportion of long-term survivors in this subset.
Notably, the CMS1 population had a very poor survival rate after relapse, in agreement with recent studies showing worse prognosis of patients with MSI and BRAF-mutated colorectal cancers that recur.
These differences in prognosis with unsupervised gene expression signatures confirm the clinical relevance of the intrinsic biological processes implicated in each CMS.
From a biological perspective, the consortium researchers were able to refine the number and interpretation of the 'non-MSI' subtypes, which represent nearly 85% of the primary colorectal cancer samples. They also described strong molecular associations, particularly for samples lacking a mesenchymal phenotype. From a clinical perspective, in colorectal cancer, as for many other cancer types, it remains unclear what features will provide the most relevant subclassification tool.
In colorectal cancer, few biomarkers (including RAS and BRAF mutations and MSI and CIMP status) have been translated to patient care. It is important to emphasize that although the CMS groups are enriched for some genomic and epigenomic markers, their associations described do not allow categorisation of gene expression subtypes, thus reinforcing the notion that transcriptional signatures allow refinement of disease subclassification beyond what can be achieved by currently validated biomarkers.
For example, although tumours with wild-type RAS are considered to be a homogeneous entity for the purpose of making therapeutic decisions in the setting of advanced cancer, they were found across distinct CMS groups with profound biological differences that are expected to translate into heterogeneous drug responses.
Thanks to the collaborative bioinformatics work on the largest collection of colorectal cancer cohorts with molecular annotation to date, and building upon previous efforts by the independent researchers, the analyses by members of the consortium resulted in a consensus molecular classification system that allows the categorisation of most tumours into one of four robust subtypes.
Marked differences in the intrinsic biological underpinnings of each subtype support the new taxonomy of this disease. The Consortium researchers believe that this new taxonomy will facilitate future research in the colorectal cancer field and should be adopted by the community for colorectal cancer stratification.