From RNAseq to Milk Using RNA molecular phenotypes to study lactation in New Zealand dairy cattle

Abstract

<Background> Bovine milk is an important source of human nutrition. It is understood that milk production is under considerable genetic control, and our knowledge of the genes and variants underlying this regulation is incomplete. In most farming systems, germplasm is delivered by artificial insemination, enabling the rapid distribution of high-quality genetics across a population. The purpose of this study was to add to our ability to predict the best available cattle from their genomes, in order to increase the rate of genetic gain for economically-important traits such as milk production and growth. The work presented herein used high-depth RNAseq data, derived from RNA extracted from mammary biopsies of 411 lactating cows. Molecular phenotypes were developed, enabling the identification of candidate genes that influence lactation and body size, contributing to the improvement in accuracy of genetically informed phenotypic selection in dairy cattle. <Results> As a first step, I developed phenotypes for gene expression measured both from unspliced pre-mRNA and from mature mRNA, as well as splicing efficiency. These were used to identify quantitative trait loci (QTL) for intron-expression (ieQTL: n=554), exon-expression (eQTL: n=2,699), and splicing efficiency (seQTL: n=170). Further investigation linked these molecular QTL to multiple milk and body size QTL. Further studies found additional eQTL signals for lactose phenotypes at twelve loci across the genome, and identified MATN3 as a candidate causative gene for a stature QTL on chromosome 11. I also identified 2,413 RNA editing sites in 649 genes. Developing these into an additional molecular phenotype yielded editing QTL (edQTL) affecting 187 sites in 89 genes. At fifteen sites, edQTL were co-regulated with significant cis-eQTL, implying effects of editing on levels of mature mRNA transcripts for the genes containing these sites. In another study, I combined both eQTL and edQTL methodologies to investigate a complex pleiotropic region on chromosome 5, highlighting CSF2RB as the likely causative gene underlying at least some of the milk QTL present at this locus. In the final study, I conducted a study to test the phenotypic predictive ability of variants that tag eQTL compared to existing variant sets, demonstrating the possible improvement they can provide in total animal selection prediction accuracy relative to currently-used methods. <Conclusions> Several molecular phenotypes were developed, and numerous QTL discovered. A subset of these were co-regulated with eQTL or edQTL, highlighting potential causative genes operating via regulatory mechanisms. Tag variants for eQTL related to milk production pathways provided a potential route to improve prediction accuracy in animal breeding.