The Molecular Landscape of the Aging Mouse Liver

Systems biology approaches often use networks of gene expression and metabolite data to identify regulatory factors and pathways connected with phenotypic variance. Separating upstream causal mechanisms, downstream biomarkers, and incidental correlations remains a significant challenge, yet it is essential for designing mechanistic experiments. To address this, we first designed a population following 2157 individual mice from 89 isogenic strains of BXD mice across their lifespans to identify molecular interactions between genotype, environment, age (GxExA) and metabolic fitness. Each strain was separated into two cohorts, fed low fat (6% cal/fat) or high fat (60% cal/fat) diets. One-third of individuals (662) were sacrificed at ~6, 12, 18, or 24 months-of-age, with the remainder monitored until natural death. Transcriptome, proteome, and metabolome profiles were generated from liver samples. These multi-omic measurements were deconvolved into metabolic networks, where we observed varying network connectivity as a function of GxExA. The multiple independent study variables permitted causal inference analysis for the network variants using stability selection. This calculates the strength and directionality of the interactions between molecular measurements and metabolic networks as a function of age, diet, and genotype, and assigns each gene a score for its relative position to the target pathway. At 1% FDR, 94% of novel connections were stable across age and diet, such as the connection between Rdh11 with cholesterol biosynthesis and Mut with mitochondrial translation. 6% of discovered candidate genes were unstable, indicating a clear causal relationship between the segregating independent variable, the gene, and the pathway. For instance, age drives variation in proteasomal genes (e.g. Psmb3, Psmb4), which in turn drive changes in the mitochondrial ribosome. Conversely, COX7A2L malformation drives variation in OXPHOS genes, but both are downstream of changes in mitochondrial translation. Finally, we examined all data for connections with the longevity and known longevity-related pathways, identifying several dozen novel candidate genes. Specific C. elegans orthologs for the top two candidates, Ctsd and St7, were knocked down with RNAi and found to reduce longevity both in wildtype worms and in mutant long-lived strains.