Diurnal transcriptome landscape of a multi-tissue response to time-restricted feeding in mammals
Time-restricted feeding (TRF) is an emerging behavioral nutrition intervention that involves a daily cycle of feeding and fasting. In both animals and humans, TRF has pleiotropic health benefits that arise from multiple organ systems, yet the molecular basis of TRF-mediated benefits is not well understood. Here we subjected mice to isocaloric ad libitum feeding (ALF) or TRF of a Western diet and examined gene expression changes in samples taken from 22 peripheral organs and brain regions collected every 2 h over a 24-h period. We discovered that TRF profoundly impacts gene expression. Nearly 80% of all genes show differential expression or rhythmicity under TRF in at least one tissue. Functional annotation of these changes revealed tissue- and pathway-specific impacts of TRF. These findings and resources provide a critical foundation for future mechanistic studies, and will help to guide human time-restricted eating (TRE) interventions to treat various disease conditions with or without pharmacotherapies. Dataset-1: TMM normalized counts of diurnal mRNA expression in 22 tissues from mice subjected to ad-libitum feeding (ALF) or 9 hr time-restricted feeding (TRF) of Western diet. Dataset-2: DE analysis for all 22 tissues using edgeR. Dataset-3: Rhythmic analysis for all 22 tissues under ALF and TRF conditions using Metacycle. Dataset-4: Raw and normalized metabolite levels in liver samples from ALF and TRF mice obtained from Metabolon, Inc., along with DE and rhythmic analyses.
Steps to reproduce
10 weeks old male C57BL/6J mice were subjected to 45% western diet (Research Diets, Inc #D12451) either in ad-libitum (ALF) or under 9h TRF from ZT13 to ZT22 for 7 weeks. 2-3 mice per treatment group were sacrificed every 2 h over a 24 h period (ZT-0, -2, -4, -6, -8, -10, -12, -14, -16, -18, -20, and -22). Twenty-two brain regions and peripheral tissues were collected and flash frozen within 1 h of dissection. All animal experiments were carried out in accordance with the guidelines and approved by the IACUC of the Salk Institute. Total RNA was extracted from samples depending on the method standardized for that tissue. Libraries were prepared using TruSeq Stranded mRNA kit (Illumina) as per manufacturer’s instructions. They were quantified using Quant-iT™ dsDNA HS Assay Kit, pooled and sequenced at the NGS Core Facility of the Salk Institute or at Novogene Co. Sequencing library read quality was assessed using FastQC, version 0.11.5. Libraries were mapped individually to the mm10 genome using STAR v2.5.3a. Gene expression levels were quantified across all exons using HOMER v4.10 and the mm10 UCSC genome annotation. Technical (re-sequencings) and biological replicates were mapped, aligned and quantified separately. Technical replicates were collapsed using DESeq2 v1.24.0 collapseReplicate function. Out of 1056 total samples (12 time-points X 2 independent animal replicates (biological replicates) X 2 treatment conditions X 22 tissues), 21 samples were found to be significantly divergent (outliers) and removed from downstream analysis. On average, ~36M mapped reads were obtained per sample. Differential gene expression analysis was carried out via edgeR v3.26.7. Differential expression results were corrected for multiple hypotheses testing using the Benjamini-Hochberg method. A FDR significance threshold of adjusted p-value ≤0.05 and an expression cutoff of logCPM >0 was set. Statistical analysis of rhythmicity was performed on the TMM normalized counts for each tissue using the Metacycle R package. Transcripts having the (combined JTK and LS) meta2d_BH.Q-value <0.05 were considered as statistically significant. The same liver tissue samples used for RNA-seq analysis were sent to Metabolon Inc. for metabolite analysis. Normalized metabolite counts from Metabolon were used for all analyses using the online tool Metaboanalyst 5.0. Metabolite Set Enrichment Analysis (MSEA) function was used to quantify fold change and FDR for metabolites, and the Metacycle package was used to identify rhythmic metabolites. Metabolites with MSEA FDR<0.05 and Metacycle meta2d_BH.Q <0.05 were considered as significant, and used as the input list for pathway analysis. To identify the combined interaction effect between TRF intervention and time of day, a Time-series + one experimental factor analysis was performed using Metaboanalyst, and metabolites with adjusted p-value<0.05 were considered as significant.