Metagenomic meta-analysis of the gut microbiome in the different primary locations of colorectal cancer

Abstract

BACKGROUND Recent studies have found a relationship between gut microbes and the primary location of colorectal cancer (CRC). However, most of these studies had limitations in sample size or sequencing methods. In this study, we collected metagenomic data from three studies and meta-analyzed the microbiological features according to the grouping of right-side colon cancer (RCC), left-side colon cancer (LCC), and rectal cancer (RC). METHODS We first identified confounding factors (except for tumor location) by two-way ANOVA and comparing species diversity. Subsequently, the microbial compositions were compared between different tumor locations. Microbial co-occurrence networks were established based on samples with different tumor locations. A prediction model for primary tumor location was constructed using a random forest algorithm based on microbial abundance features. Finally, tumor location and confounding factors were entered in the MAASLIN2 to identify differential species. Linear discriminant analysis (LDA) also identified the differential species. RESULTS Different study sources and BMI influenced gut microbiome and significantly altered α-diversity and β-diversity, bringing the confounding effect when analyzing gut microbial features in different tumor locations. However, α-diversity and β-diversity of gut microbiome had no significant difference in tumor locations. Species belonging to the Phylum of Actinobacteria, Firmicutes, and Proteobacteria played essential linkages in the three microbial networks, while Bacteroidetes were more critical in the microbial network of RCC. There are both the same hub species and different hub species among the three networks. The random forest classification model performed well in predicting RC (class error = 0.217) but poorly classified the RCC and LCC, with an overall classification error of 0.613. In comparing colon cancer (CC) with RC, MAASLIN2 and LDA identified six species significantly enriched in RC and thirteen in CC. In comparing RCC with LCC, MAASLIN2 identified nine species significantly enriched in RCC and six significantly enriched in LCC. Some of the differential species were reported to be associated with CRC location-related Molecular and immune features. CONCLUSION This study elucidated the relationship between gut microbiome and CRC location and confirmed that RCC, LCC, and RC had different enrich patterns of microbiota.