The effect of patient demographics and polyp histologic characteristics on immune gene expression in microenvironment is mediated by infiltrating microbiome.

We apply new panomic distance mediation analysis with feature selection to understand the effect of polyp-infiltrating microbes on host microenvironment immune response. Multivariate omnibus distance mediation analysis (MODIMA) allows for testing for mediation in the context of multivariate exposure-mediator-response triples, such as polyp characteristics, microbiome composition, and expression, respectively. Here we extend MODIMA to allow for selection of individual components of the response and mediator responsible for mediation effects. Next we apply the method to new data from 30 colorectal polyps with simultaneous measurements of expression of 594 immune-related genes ascertained via the NanoString Immunology panel and microbiome composition via a 16S rRNA gene amplicon sequencing assay. The polyp specimens are derived from a cohort consisting of 11 female and 19 male subjects, of whom 17 are black and 25 are older than 55. The polyp biopsies come from proximal (17), distal (6), and rectal (7) locations of the colon. According to pathology evaluation, half of the polyps have evidence of high-grade dysplasia. The distribution across tubular, tubular villous, and villous histology is likewise approximately even (10, 9, and 11 specimens, respectively). Microbial DNA has been extracted from formalin-fixed, paraffin-embedded polyp tissues and analyzed using Illumina MiSeq 16S rRNA gene amplicon sequencing assay. After preprocessing, the species and genus level data have been normalized using central log ratio transformation with pseudo-counts. Species and genera with literature evidence for association with colorectal polyps and cancers have been selected for further analysis. Using these data, we evaluate the hypothesis that infiltrating microbiome mediates the immune gene expression in response to histodemographic characteristics. We expand the histodemographic data into a complete 6-way interaction design matrix. Euclidean distances have been used for all data types in all analyses. After feature selection, the joint expression of 140 immune genes is associated with polyp characteristics (dCor t-test, T = 2.9, bias corrected dCor = 0.14, p = 0.002). Likewise, 4 microbial features are jointly associated (dCor t-test, T = 2.1, bias corrected dCor = 0.10, p = 0.02). Within these data 9 genes (BCL2, CD19, CLEC6A, IRAK3, IRF3, NLRP3, PSMC2, TBX21, POLR2A) and 2 taxa (Bifidobacterium and Alistipes sp.) are jointly associated to each other (dCor t-test, T = 4.7, bias corrected dCor = 0.23, p Citation Format: Alexander V. Alekseyenko, Chentha Vasu, Brianna Bronsky, David N. Lewin, Shaoli Sun, Christine Bookout, John N. Baron, Kristin Wallace. The effect of patient demographics and polyp histologic characteristics on immune gene expression in microenvironment is mediated by infiltrating microbiome [abstract]. In: Proceedings of the AACR Special Conference on the Microbiome, Viruses, and Cancer; 2020 Feb 21-24; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2020;80(8 Suppl):Abstract nr B35.