* FWA and JSB contributed equally
Regulatory variation is an important source of genetic variation for many traits and can be identified as “expression quantitative trait loci” (eQTL). To date, all eQTL studies have been hampered by low statistical power due to limited sample sizes. As a consequence, the full extent and nature of regulatory variation remains unknown.
We have addressed this limitation in the yeast Saccharomyces cerevisiae. We used mRNA sequencing to profile gene expression in more than 1,000 recombinant individuals generated from a cross between two yeast isolates. The statistical power of this dataset is high enough to map thousands of previously “missing” eQTL that together account for ~80% of the heritability of gene expression. Thus, our data provide a nearly exhaustive view of how genetic variation influences the transcriptome.
We identified 34,318 eQTL for 6,210 transcripts. A typical transcript is influenced by a median of 6 and up to 20 eQTL, several fold more than previously seen. While 43% of all genes had a local eQTL that is located close to the gene, most eQTL are located elsewhere in the genome and influence gene expression in trans. The aggregate effect of the trans eQTL was larger than that of the local eQTL, illustrating the importance of trans acting variation.
Rather than appearing randomly across the genome, the newly discovered trans eQTL were highly structured, such that the vast majority fell into one of 111 hotspot regions that each affect the expression of many genes. Some of these hotspots have extraordinarily wide-reaching effects and influence thousands of transcripts across all cellular processes, while others specifically influence certain pathways. By combining information from all genes that map to a given hotspot, we can fine-map the causal hotspot location with high precision, in 26 cases to single-gene resolution of less than 1.5 kb. This permits – for the first time – a systematic and unbiased analysis of the types of genes that act as trans eQTL. Trans-acting variation generates structure in the yeast transcriptome such that groups of genes are affected by multiple eQTL in a combinatorial fashion. Finally, despite our high statistical power, many local eQTL did not act as trans eQTL for other genes. This might indicate that the expression changes these eQTL cause at their local genes do not further affect cellular physiology, at least not in ways that are reflected in the transcriptome. This raises important questions about the characteristics of genes where local regulatory variation does have cellular and phenotypic consequences.