Jacek Majewski is an assistant professor in the Department of Human Genetics at McGill University in Montreal. GT's Jeanene Swanson spoke with him about his recent paper in PLoS Genetics exploring tissue-specific effects on expression QTLs and the future of linking GWAS and gene expression studies.
GT: Has there been progress linking GWAS data to gene expression?
JM: Many researchers strongly believe that eQTLs are largely responsible for subtle phenotypic variation and susceptibility to complex genetic disorders. The expectation is that many GWAS hits — SNPs associated with a phenotype — will be explained by underlying gene expression changes. However, to date only a handful of studies succeeded in connecting GWAS results with gene expression changes. [These] studies used primary tissues relevant to the trait of interest.
GT: Is it important to assay different cell and/or tissue types in GWAS or expression studies?
JM: I think that one of the crucial issues in linking GWAS and expression quantitative trait loci studies is to profile a large number of tissues. The human lymphoblast studies, carried out mostly on HapMap samples by a number of groups, showed us that eQTLs are frequent and shared across populations. The LCL [lymphoblastoid cell lines] studies suffered from two main weaknesses. The first one is the limited sensitivity with respect to gene expression. First, microarrays were a revolutionary tool, but while they could in theory be used to profile expression of every known and predicted gene, in practice their working range was biased toward highly expressed genes. We had very limited power to detect eQTLs among genes with low expression levels. Secondly, we had no means of detecting genes that were not expressed in LCLs.
GT: How does using primary cell lines change the nature of GWAS?
JM: It should be noted that not only is gene expression altered by genetic variants, but so are the isoform types produced by numerous loci. Our recent study used primary cells, human osteoblasts, to revisit our earlier eQTL isoform study. We found that only about 40 percent of all genes were expressed at a detectable level in both cell types. Within the shared genes, less than 50 percent showed a common eQTL effect, [meaning] expression of the same gene associated with a genotype of the same SNP. This effect was seen at the whole gene expression level and the isoform level. These results provide two major insights into the nature of eQTL studies: one, LCLs are a valid model for studying the effect of genetic variation on gene expression, and many of the eQTLs can be replicated in primary cultured cells; but two, a large fraction of eQTLs will be missed by studying only a single cell type, whether primary or immortalized. In order to consistently identify the genetic variants underlying eQTLs that are identified in GWAS, we will need a large number of tissues.
GT: What is the future of association studies in terms of mapping variation to gene expression changes?
JM: I expect that the scientific community will attempt to link eQTLs and GWAS by a two-pronged approach. Genomics groups will continue to expand the panel of tissues and cell lines, using a variety of emerging technologies, notably next-generation RNA sequencing, in order to a priori identify a large number of eQTLs and pinpoint the regulatory SNPs. These studies will result in databases containing SNPs associated with expression, isoforms, and splicing changes, which will in turn be used as candidates for GWAS studies. GWAS researchers will take their individual hits and subject them to a variety of expression tests in tissues that are particularly relevant to the disease of interest. I expect that we will be able to explain dozens of GWAS hits by underlying eQTLs and isoforms eQTLs in the next few years.