Epithelial-to-mesenchymal transition signature assessment in colorectal cancer quantifies tumour stromal content rather than true transition.

Amy Mb Mccorry,Mark Lawler,Philip D Dunne,Daniel B Longley,Maurice B Loughrey

doi:10.1002/path.5155

Abstract

The process of epithelial‐to‐mesenchymal transition (EMT) in cancer is a well‐described process whereby epithelial tumour cells undergo molecular/phenotypic changes and transition to a mesenchymal biology. To aid in the transcriptional characterisation of this process, gene expression signatures have been developed that attribute a relative EMT score to samples in a given cohort. We demonstrate how such EMT signatures can identify epithelial cell line models with high levels of transition but also highlight that, unsurprisingly, fibroblast cell lines, which are inherently mesenchymal, have a higher EMT score relative to any epithelial cell line studied. In line with these data, we demonstrate how increased tumour stromal composition, and reduced epithelial cellularity, significantly correlates with increasing EMT signature score, which is evident using either in silico subtyping analysis (p < 0.00001) or in situ histopathological characterisation (p < 0.001). Considered together, these results reinforce the importance not only of interdisciplinary research to correctly define the nature of EMT biology but also the requirement for a cadre of multidisciplinary researchers who can analyse and interpret the underlying pathological, bioinformatic and molecular data that are essential for advancing our understanding of the malignant process. © 2018 The Authors. The Journal of Pathology published by John Wiley & Sons Ltd on behalf of Pathological Society of Great Britain and Ireland.

Highlights

Due to the decreasing cost of generating molecular information, large repositories [such as The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO)] provide access to semi-resolved molecular data
Pathologists perform diagnostic assessments of tissue following initial processing, the biological interrogation of these samples is frequently undertaken in a manner that fails to account for the fundamental tumour pathology by researchers with a predominantly computational biology or bioinformatics background, with input from molecular biologists
This study examines the use of transcriptional classification, with a focus on epithelial-to-mesenchymal transition (EMT) gene expression signatures in colorectal cancer (CRC) as Morris and Kopetz recently commented on how the ‘distinctions between mesenchymal and EMT signatures are commonly blurred in the gene expression literature’ [1]

Summary

Introduction

Due to the decreasing cost of generating molecular information, large repositories [such as The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO)] provide access to semi-resolved molecular data. This explosion in data availability has created a bottleneck in the cancer research pipeline as a large proportion of traditional ‘wet-lab’ molecular biologists are unable to independently perform the complex bioinformatic interrogation required to maximise the value of this freely available data. We demonstrate how a disconnect between pathological/experimental ‘wet-lab’ research and data processing/bioinformatics ‘dry-lab’ analysis can introduce potential confounding errors in the interpretation of gene expression signatures in tumour cohorts and the understanding of molecular subtypes in cancer

Materials and methods

Results

Discussion