Abstract
Advances in genomics in recent years have provided key insights into defining cancer subtypes "within-a-tissue"-that is, respecting traditional anatomically driven divisions of medicine. However, there remains a dearth of data regarding molecular profiles that are shared across tissues, an understanding of which could lead to the development of highly versatile, broadly applicable therapies. Using data acquired from The Cancer Genome Atlas (TCGA), we performed a transcriptomics-centered analysis on 1494 patient samples, comparing the two major histological subtypes of solid tumors (adenocarcinomas and squamous cell carcinomas) across organs, with a focus on tissues in which both subtypes arise: esophagus, lung, and uterine cervix. Via principal component and hierarchical clustering analysis, we discovered that histology-driven differences accounted for a greater degree of inherent molecular variation in the tumors than did tissue of origin. We then analyzed differential gene expression, DNA methylation, and non-coding RNA expression between adenocarcinomas and squamous cell carcinomas and found 1733 genes, 346 CpG sites, and 42 microRNAs in common between organ sites, indicating specific adenocarcinoma-associated and squamous cell carcinoma-associated molecular patterns that were conserved across tissues. We then identified specific pathways that may be critical to the development of adenocarcinomas and squamous cell carcinomas, including Liver X receptor activation, which was upregulated in adenocarcinomas but downregulated in squamous cell carcinomas, possibly indicating important differences in cancer cell metabolism between these two histological subtypes of cancer. In addition, we highlighted genes that may be common drivers of adenocarcinomas specifically, such as IGF2BP1, which suggests a possible link between embryonic development and tumor subtype. Altogether, we demonstrate the need to consider biological similarities that transcend anatomical boundaries to inform the development of novel therapeutic strategies. All data sets from our analysis are available as a resource for further investigation.