Colorectal Cancer (CRC) is a heterogeneous disease, with variable molecular pathogenesis, natural history and response to treatments. Therefore, it is important to exploit all available molecular information to enable personalized management of the disease. To challenge such heterogeneity, research studies have been performed to exploit global transcriptional profiling of CRC samples, in three major ways:
- Stratification of CRC in molecular subtypes, to define subgroups with distinct molecular and clinical features
- Characterization of the stromal contribution to CRC pathogenesis and response/resistance to treatments
- Identification of rare events (“outliers”) associated with specific, therapeutically actionable molecular alterations
1. CRC Transcriptional Subtypes
To identify transcriptional subtypes of CRC, several independent research groups analyzed global gene expression profiles from large data series and organized tumor samples with similar expression profiles into subgroups. As a result, the various works identified variable numbers of CRC subtypes, from three to six. Subsequently, the research groups joined forces in the “Colorectal Cancer Subtyping Consortium”, which led to the definition of four CRC “consensus molecular subtypes”1. To search for microRNAs potentially driving these CRC subtypes, Enzo Medico’s group has developed an analytical pipeline, microRNA master regulator analysis (MMRA). Starting from a CRC dataset of 450 samples profiled by mRNA-seq and small RNA-Seq, MMRA combined statistical tests, microRNA-target prediction and joint mRNA/microRNA network analysis to successfully identify microRNAs potentially driving CRC subtypes. The candidates were then experimentally validated in CRC cell lines 2.
2. Stromal Expression Signatures in CRC
One of the CRC transcriptional subtypes, dubbed “stem/serrated/mesenchymal” (SSM), is endowed with low differentiation, epithelial-to-mesenchymal transition (EMT) and poor prognosis. These features have been interpreted as signs of a phenotypic switch whereby epithelial cancer cells acquire mesenchymal and stem cell-like features. However, Enzo Medico noted that genes upregulated in this subtype are also prominently expressed by stromal cells, suggesting that SSM transcripts could derive from stromal rather than epithelial cancer cells. To test this hypothesis, he analysed CRC expression data from patient-derived xenografts (PDXs), where mouse stroma supports human cancer cells. Species-specific expression analysis, carried out by both DNA microarray and RNA-Seq, showed that the mRNA levels of SSM genes were mostly due to stromal expression. Transcriptional signatures built to specifically report the abundance of cancer-associated fibroblasts (CAFs), leukocytes or endothelial cells all had significantly higher expression in human CRC samples of the SSM subtype. High expression of the CAF-signature was associated with poor prognosis of CRC after surgery, and high expression of all stromal signatures predicted resistance to radiotherapy in rectal cancer. At odds with tumor purity measures based on microscopic inspection or DNA analysis, transcriptional stromal scores do not simply reflect the fraction of stromal cells in the tumor; they also provide information on their composition and functional state. This may explain why evaluation of stroma abundance by pathological or DNA-based criteria has failed to highlight clinically relevant correlates. Collectively, these results reinterpret the EMT and stemness traits typically ascribed to epithelial cancer cells as prevalent stromal contributions and provide new stromal scores that may prove helpful in advancing CRC prognosis and response prediction 3.
3. Expression Outliers and Fusion Transcripts
Distinguishing molecular alterations activating a cancer driver from passenger alterations is mandatory to formulate effective targeted treatments. The common rule to define cancer drivers as the frequently altered ones, clashes with the fact that some alterations occur at low frequency but involve bona fide, therapeutically actionable oncogenic drivers. Among them, “outlier expression”, i.e. marked overexpression in a small fraction of cases has a high probability of pinpointing activation of a cancer driver, mostly by translocation or amplification, and an associated oncogenic dependence. By transcriptional outlier analysis of 151 CRC cell lines, Enzo Medico and colleagues identified tyrosine kinase genes whose outlier expression was associated with a fusion transcript, indicating a translocation through which the kinase gene acquired a strong promoter. Cell carrying an outlier kinase displayed exquisite sensitivity to the respective kinase inhibitor 4. This approach has the potential to pinpoint cancers with exquisite dependencies to individual kinases for which clinically approved drugs are already available.