Identification of enterotype and its predictive value for patients with colorectal cancer

Abstract

Gut microbiota dysbiosis involved in the pathogenesis of colorectal cancer (CRC). The characteristics of enterotypes in CRC development have not been determined. To characterize the gut microbiota of healthy, adenoma, and CRC subjects based on enterotype. The 16 S rRNA sequencing data from 315 newly sequenced individuals and three previously published datasets were collected, providing total data for 367 healthy, 320 adenomas, and 415 CRC subjects. Enterotypes were analyzed for all samples, and differences in microbiota composition across subjects with different disease states in each enterotype were determined. The predictive values of a random forest classifier based on enterotype in distinguishing healthy, adenoma, and CRC subjects were evaluated and validated. Subjects were classified into one of three enterotypes, namely, Bacteroide- (BA_E), Blautia- (BL_E), and Streptococcus- (S_E) dominated clusters. The taxonomic profiles of these three enterotypes differed among the healthy, adenoma, and CRC cohorts. BA_E group was enriched with Bacteroides and Blautia; BL_E group was enriched by Blautia and Coprococcus; S_E was enriched by Streptococcus and Ruminococcus. Relative abundances of these genera varying among the three human cohorts. In training and validation sets, the S_E cluster showed better performance in distinguishing among CRC patients, adenoma patients, and healthy controls, as well as between CRC and non-CRC individuals, than the other two clusters. This study provides the first evidence to indicate that changes in the microbial composition of enterotypes are associated with disease status, thereby highlighting the diagnostic potential of enterotypes in the treatment of adenoma and CRC. Three enterotypes (BA_E, BL_E, and S_E) were identified in healthy, adenoma, and CRC subjects. BA_E, BL_E, and S_E clusters were dominated by Bacteroide, Blautia, and Streptococcus, respectively. Differences in gut microbial composition were observed within the control, adenoma, CRC populations for each enterotype. S_E showed better performance in distinguishing three human cohorts than BA_E and BL_E. Disease prediction performance of enterotypes is no better than that of a classification model based on all samples.