bioRxiv2025
BackgroundThe gut microbiota adapts to and shapes the hosts metabolic state through affecting circulating metabolites and consequent gene regulatory networks, resulting in systemic influences in diverse organs via connections such as the gut-liver axis. Numerous variables such as diet, age, and host genetics modulate the composition of the gut microbiome, but their interactions and specific associative and mechanistic links to host molecular phenotypes remain incompletely unannotated. Integrated multi-omics approaches in genetically diverse populations offer an opportunity to dissect these interactions and identify predictive microbial signatures for host phenotypes, such as body weight, and molecular associations with gene expression pathways in gut and liver. ResultsWe sequenced, aligned, and integrated the cecal metagenome, metatranscriptome, and host transcriptome from 232 mice across 175 distinct cohorts according to a low fat chow diet (CD) or a high-fat diet (HF), four adult ages (between roughly 180 to 730 days of age), and 43 distinct genotypes (inbred BXD strains). Genetics and diet exerted the strongest influence on microbiota abundance and activity, followed by age. HF feeding significantly reduced diversity across all ages and all genotypes, altering >300 species. Machine learning models based on microbial profiles reliably predicted body weight within dietary group (AUC = 0.84 for CD, 0.79 for HF) and chronological age (AUC = 0.84), with model performance rising of age prediction rising to 0.95 when integrating top microbial features with liver proteomics. Network analyses of expression data revealed links between genes, pathways, and specific microbes, including a negative association between cecal Ido1 expression and short-chain fatty acid (SCFA)-producing Lachnospiraceae, suggesting dietary fat may modulate host tryptophan metabolism through microbiota shifts. ConclusionsWhole metagenome and metatranscriptome sequencing approaches have massively expanded the landscape of microbiome analysis compared to earlier short-read 16S analyses. The resulting datasets quantify hundreds of uniquely identifiable microbes, which can be used to create sets of highly predictive microbial biomarkers for aging and obesity. When trained on controlled mouse populations, these results demonstrate that microbiome profiling can achieve high predictive capacity (AUC = 0.95 with multi-omics integration) for complex readouts such as age and body weight (AUC = 0.84), even considering genetic and dietary variation, establishing a framework for biomarker development. While at present many bacteria are still functionally unannotated at the species level, multi-omics approaches -- including gene expression from the host tissues -- provide insights into the functional associations of specific taxa in the microbiome.