Pan-cancer analysis of exogenous (microbial) sequences in tumor transcriptome data from the ORIEN consortium and their association with cancer and tumor microenvironment.

Abstract

3113 Background: The tumor microbiome holds great potential for its ability to characterize various aspects of cancer biology and as a target for rational manipulation. For many cancer types, little is known about the role of microbes and in what contexts they affect clinical outcomes. Non-human (i.e. exogenous) sequences can be observed in low abundance within high throughput sequencing data of tumors. Here, we describe a collaboration among members of The Oncology Research Information Exchange Network (ORIEN) to leverage tumor biopsy RNAseq data collected under a shared protocol and generated at a single site to better understand the tumor microbiome, its association with prognostic features of the tumor microenvironment (TME) such as hypoxia, and how it may be used to improve clinical outcomes. Methods: Tumor RNAseq samples from 10 primary source locations including the tissues colon, lung, pancreas, and skin from ORIEN and similar cancers from The Cancer Genome Atlas (TCGA) were processed through the exoTIC (exogenous sequencing in tumors and immune cells) pipeline to identify and count exogenous sequences, filter contaminants, and normalize across datasets. Gene expression signatures of the TME, such as hypoxia, were calculated using ‘tmesig’. Microbe relative abundances were modeled with primary tumor location and hypoxia score using a gamma-distributed generalized linear regression via the stats package in R. Results: We analyzed RNAseq data of 2892 and 2720 tumors from ORIEN and TCGA, respectively. Patients’ ages were significantly greater in the ORIEN than the TCGA dataset (62 vs 58 yo, t-test p<0.001). The ORIEN data contained more sarcoma samples than TCGA (n = 691 vs 259) with roughly equivalent numbers in other cancer types. Fewer microbes were significantly associated with the hypoxia score than with cancer type (n = 32 vs 210). This trend was observed in both the ORIEN and TCGA datasets. The largest effect sizes were observed between microbes and small cell lung cancer. Conclusions: We found microbial sequences in all ORIEN and TCGA tumor RNAseq samples tested. Cancer type showed more significant associations with microbes than a hypoxia signature. These observations merit further investigation into the interaction between microbes and the TME. [Table: see text]