Large-scale integrative analysis of fecal samples sequencing data supports novel miRNA-mediated host-microbiota interactions

INTRODUCTION AND AIMS: miRNAs are a class of small non-coding RNAs mainly involved in gene expression regulation2. The gut is an ecological niche for a variety of microorganisms (microbiota) involved in complex molecular interactions with the human cells. Moreover, there is evidence that miRNAs can be released in the gut lumen, modulate the expression of bacteria target genes and be detected in stool samples3. However, an extensive investigation of miRNAs and microbial species involved in such interactions in human is still missing. Our group recently showed that fecal miRNA levels distinguish patients with colorectal cancer (CRC) from healthy controls4 as well as subjects with specific dietary habits5. The concomitant profiling of fecal miRNAs and microbial species in a large cohort of individuals can then provide more insights on novel host-microbiota interactions. METHODS: Shotgun metagenomic sequencing and small RNA-Seq (sRNA-Seq) were performed on stool samples from 912 individuals from eight independent European cohorts. The cohorts included 409 healthy subjects, 125 patients with inflammatory or diverticular diseases, 43 with precancerous lesions, and 307 CRC. The study population also included subjects with different dietary habits, including, 40 omnivores, 40 vegetarian and 40 vegan individuals from5 were included in the control class. Microbial profiling was performed using MetaPhlAn 4.1 and identified 3,442 different species among which 632 passed a prevalence-based filter (nearZeroVar). Conversely, Docker4Seq was applied for fecal miRNA quantification resulting in 2,629 detected miRNAs. Among these miRNAs 232 were associated with a median level greater than 15 reads and selected for further analyses. Diversity analysis was performed considering alpha-diversity metrics (Inverse Simpson and Shannon-index) while miRNA and species richness were computed as number of features (miRNAs or species) detected in each sample. For the species, an oral score (expressed as number of oral species detected) was also computed. Rank regression models were applied to compute the miRNA-microbial associations, specifically, Spearman correlation analysis was computed for the definition of the significant associations (adj. p >0.05). Only association coherent among the different study cohorts were considered. RESULTS: Analysis of microbial species diversity showed differences among combination of microbial species from different diets, observing significant difference between them. Particularly, the number of detected species was higher in subjects with vegetarian diet with respect to vegan and omnivores. Microbial species diversity lowly increased from healthy to CRC patients’ samples, particularly, decreased in subjects with diverticular diseases and increased in subjects with inflammatory or hyperplastic conditions. Moreover, the number of oral species showed a significant increase from healthy to tumor samples. Despite the fecal miRNA heterogeneity significantly decreased from healthy to CRC, samples from vegan and vegetarian subjects showed higher miRNA diversity with respect to omnivores. Rank regression analysis highlighted 3,656 significant and coherent fecal miRNA-microbial species associations between cohorts (adj. p <0.05), including the known correlation between miR-1246 and Fusobacterium nucleatum6. Most associations were negative and observed prevalently in CRC patients. F. nucleatum, Peptostreptococcus stomatis, and Dysosmobacter welbionis were the species characterized by the highest number of associated miRNAs. Putative associations progressively change from healthy subjects to late-stage tumors, and some of them were confirmed also considering miRNA levels measured in tumor/adjacent tissue by sRNA-Seq. CONCLUSIONS: Our data suggest that a specific microbial composition may mirror the expression and release of miRNAs in the gut lumen, with their subsequent detection in stool. This was also reflected in the opposite trend between miRNA and oral species heterogeneity, especially in CRC. Correlation analysis between miRNA levels and species relative abundances showed a higher number of anti-correlations, particularly reported in CRC, and involved oral species, highlighting novel cross-kingdom molecular interactions.