Genome-Wide Co-Expression Analysis in Multiple Tissues

Ian C Grieve,Michal Pravenec,Vladimir Kren,Stuart A Cook,Enrico Petretto,Nicholas J Dickens,Norbert Hubner,Jonathan Mangion,Timothy J Aitman,Cathal Seoighe

doi:10.1371/journal.pone.0004033

Abstract

Expression quantitative trait loci (eQTLs) represent genetic control points of gene expression, and can be categorized as cis- and trans-acting, reflecting local and distant regulation of gene expression respectively. Although there is evidence of co-regulation within clusters of trans-eQTLs, the extent of co-expression patterns and their relationship with the genotypes at eQTLs are not fully understood. We have mapped thousands of cis- and trans-eQTLs in four tissues (fat, kidney, adrenal and left ventricle) in a large panel of rat recombinant inbred (RI) strains. Here we investigate the genome-wide correlation structure in expression levels of eQTL transcripts and underlying genotypes to elucidate the nature of co-regulation within cis- and trans-eQTL datasets. Across the four tissues, we consistently found statistically significant correlations of cis-regulated gene expression to be rare (<0.9% of all pairs tested). Most (>80%) of the observed significant correlations of cis-regulated gene expression are explained by correlation of the underlying genotypes. In comparison, co-expression of trans-regulated gene expression is more common, with significant correlation ranging from 2.9%–14.9% of all pairs of trans-eQTL transcripts. We observed a total of 81 trans-eQTL clusters (hot-spots), defined as consisting of ≥10 eQTLs linked to a common region, with very high levels of correlation between trans-regulated transcripts (77.2–90.2%). Moreover, functional analysis of large trans-eQTL clusters (≥30 eQTLs) revealed significant functional enrichment among genes comprising 80% of the large clusters. The results of this genome-wide co-expression study show the effects of the eQTL genotypes on the observed patterns of correlation, and suggest that functional relatedness between genes underlying trans-eQTLs is reflected in the degree of co-expression observed in trans-eQTL clusters. Our results demonstrate the power of an integrative, systematic approach to the analysis of a large gene expression dataset to uncover underlying structure, and inform future eQTL studies.

Highlights

The use of linkage analysis in combination with genome-wide expression profiling by microarray, known as ‘genetical genomics’ [1], enables the genetic control points of gene expression to be mapped to the genome
Of the significantly correlated pairs of cis-eQTLs that can not be explained by linkage disequilibrium (LD), 39.6 to 55.6% have correlated SDPs (Table S1), even though the eQTLs are often located on different chromosomes
Our genome-wide studies of expression levels of genes underlying cis- and trans-eQTLs provide strong support for the hypotheses that the categorisation of eQTLs has a genuine biological basis that can be detected in transcript expression levels, and that trans-eQTL clusters consist of functionally related and coordinately regulated transcripts

Summary

Introduction

The use of linkage analysis in combination with genome-wide expression profiling by microarray, known as ‘genetical genomics’ [1], enables the genetic control points of gene expression to be mapped to the genome. These have come to be referred to as expression quantitative trait loci (eQTLs) [2]. Similar procedures have been carried out in other model organisms [7,11,13,14,15] Using these approaches in humans, Deutsch et al [16] and Goring et al [17] identified candidate genes for Down’s syndrome phenotypes and plasma HDL cholesterol concentration respectively

Methods

Results

Conclusion