Microbiome analysis of 940 lung cancers in never-smokers reveals lack of clinically relevant associations.

In spite of the growing interest in the microbiome in human cancer, there are currently only small-scale lung cancer microbiome studies conducted directly on tissue. As part of the Sherlock-Lung study, we studied the microbiomes of 940 lung cancers (4090 samples) in never smokers (LCINS) directly from lung tissue using three data types: 16S rRNA gene sequencing (16S), whole-genome sequencing (WGS) with paired blood, and RNA-seq. We observe very low biomass and few microbiome associations in LCINS using 16S and WGS tissue. Using RNA-seq, we observe more total microbial reads, and decreased relative abundance of several commensal bacteria at the genus and species levels in tumors relative to paired normal lung tissue. Among all datasets, we see no consistent associations between the lung tissue microbiome, or circulating bacterial DNA, and any available demographic and clinical features, including age, sex, genetic ancestry, second-hand tobacco smoking exposure, LCINS histology, stage, and overall survival. We also observe no microbiome associations with any human genomic alterations within the same samples. Every null result should be interpreted with caution given the possibility of future methodological breakthroughs. However, all together, using multiple data types in nearly 1000 patients, we find no substantive role for the lung cancer microbiome in treatment-naïve LCINS.