Combined single-sample metabolomics and RNAseq reveals a hepatic pyrimidine metabolic response to acute viral infection

Abstract

ObjectiveMetabolomics and RNA sequencing (RNAseq) each provide powerful readouts of phenotype, and integration of these data can provide information greater than the sum of their parts. The ability to conduct such analysis on a single sample has many practical advantages, especially when dealing with rare or difficult-to-obtain samples. While methods exist to isolate multiple biomolecular subclasses from the same sample, in-depth analysis of the suitability of these approaches for multi- omics readouts is lacking. MethodsMice were injected with lymphocytic choriomeningitis virus (LCMV) or vehicle (Veh) control and liver tissue was harvested 2.5-days later. RNA was isolated from aliquots of pulverized liver tissue either following metabolite extraction using 80% methanol (MetRNA) or directly from frozen tissue (RNA). RNA sequencing data was evaluated by differential expression analysis via edgeR and dispersion using Ginis mean differences. Differential metabolite abundance was assessed using LIMMA. Pathway enrichment analysis was conducted on metabolomics and RNAseq data using MetaboAnalysts joint-integration tools. ResultsPrior metabolite extraction had no deleterious effects on quality or quantity of isolated RNA. RNA and MetRNA generated from the same sample clustered together by principal component analysis, indicating that inter-individual differences were the largest source of variance. Of the 2,169 genes that were differentially expressed between LCMV and Veh, the vast majority (n=1,848) were shared between extraction method, with the remainder evenly divided between RNA (n=165) and MetRNA (n=156). These differentially expressed genes unique to extraction method were attributed to randomness around the false discovery rate (FDR) = 0.05 cutoff and stochastic changes in variance estimation. Gini analysis further revealed that extraction method had no effect on the dispersion of detected transcripts across the entire dataset. To demonstrate the power of multi-omics integration on interrogated metabolic phenotypes, we next performed integrated pathway enrichment analysis on RNAseq data and metabolomics data. Our analysis revealed pyrimidine metabolism as the most impacted pathway by LCMV infection. Plotting up- and down-regulated genes and metabolites on the Kyoto Encyclopedia of Genes and Genomes (KEGG) pyrimidine pathway exposed a pattern enzymatic degradation of pyrimidine nucleotides to generate the nucleobase uracil. Further, uracil was among the most differentially abundant metabolite in serum of LCMV infected mice, suggesting a novel mechanism of hepatic uracil export in acute infection response. ConclusionsWe demonstrate that prior metabolite extraction does not have a deleterious effect on RNAseq quality, which enables investigators to confidently perform metabolomics and RNAseq on the same sample. Implementation of this approach revealed a novel involvement of the hepatic pyrimidine metabolism during acute viral infection.