LAUSR

OBJECTIVE We aimed to examine associations between the oral, fecal, and mucosal microbiome communities and adenoma formation. SUMMARY BACKGROUND DATA Data are limited regarding the relationships between microbiota and preneoplastic colorectal lesions. METHODS Individuals undergoing screening colonoscopy were prospectively enrolled and divided into adenoma and nonadenoma formers. Oral, fecal, non-adenoma and adenoma-adjacent mucosa were collected along with clinical and dietary information. 16S rRNA gene libraries were generated using V4 primers. DADA2 processed sequence reads and custom R-scripts quantified microbial diversity. Linear regression identified differential taxonomy and diversity in microbial communities and machine learning identified adenoma former microbial signatures. RESULTS One hundred four subjects were included, 46% with adenomas. Mucosal and fecal samples were dominated by Firmicutes and Bacteroidetes whereas Firmicutes and Proteobacteria were most abundant in oral communities. Mucosal communities harbored significant microbial diversity that was not observed in fecal or oral communities. Random forest classifiers predicted adenoma formation using fecal, oral, and mucosal amplicon sequence variant (ASV) abundances. The mucosal classifier reliably diagnosed adenoma formation with an area under the curve (AUC) = 0.993 and an out-of-bag (OOB) error of 3.2%. Mucosal classifier accuracy was strongly influenced by five taxa associated with the family Lachnospiraceae, genera Bacteroides and Marvinbryantia, and Blautia obeum. In contrast, classifiers built using fecal and oral samples manifested high OOB error rates (47.3% and 51.1%, respectively) and poor diagnostic abilities (fecal and oral AUC = 0.53). CONCLUSION Normal mucosa microbial abundances of adenoma formers manifest unique patterns of microbial diversity that may be predictive of adenoma formation.