Biodiversity and biogeography of the multi-kingdom cancer microbiome

Microorganisms represent an important component of the tumor microenvironment, but conflicting reports have left the extent of microbial prevalence across cancer types unclear, necessitating more robust methods for characterizing tumor-associated microbiomes. We built and benchmarked a host-subtraction and classification pipeline to identify microbiota in whole-genome sequencing data and applied it to 16,369 high-depth tumor whole genomes from the UK 100,000 Genomes Project. After decontamination, microbial signatures were indistinguishable from the background in most cancer types. However, in orodigestive tumors, we detected multi-kingdom polymicrobial communities, including bacteria, fungi, viruses, archaea, and, in some cases, Trichomonas, a protozoan parasite. These communities varied by tumor site and subtype, with increased microbial colonization of microsatellite-instable and polymerase ε (POLE)/polymerase δ (POLD1)-mutated tumors, supported by a correlation between microbial load and tumor mutation burden observed across orodigestive cancers. This analysis helps to resolve pan-cancer microbial structure and links the tumor microbiome to host phenotype and tumor genomic context.