Genomic variation: Copy number variation doesn't copy SNPs

Nature Reviews Genetics 8, 247 (April 2007) | doi:10.1038/nrg2088

Patrick Goymer

Recent studies have characterized the extent of copy number variation in the human genome, but how important is its contribution to complex phenotypes? A new study shows that copy number variants (CNVs) have a smaller contribution to gene-expression phenotypes than SNPs do, but the CNV and SNP contributions are largely independent, so both types of variation need to be measured to completely understand the genetic basis of phenotypic variation.

Manolis Dermitzakis, Matthew Hurles and colleagues carried out an association analysis of gene expression in cell lines from the HapMap individuals with both SNPs and CNVs, focusing on the genomic regions around the expressed genes. Of the 14,072 genes for which expression levels were measured, they found association with a local SNP in between 323 genes (in Europeans) and 411 genes (in Africans). For CNVs, the numbers were lower — local CNVs were associated with the expression levels of between 44 genes (in Chinese) and 96 genes (in Africans). On the basis of these data and corrections for the proportion of total copy number variation observed in this study, the authors estimate that 8.75–17.7% of heritable variation in gene expression is due to copy number variation.

So what types of effect are the CNVs having? Altering gene dosage through deletion or duplication is the obvious possibility, but over half the CNVs were outside the probed region of the gene with whose expression they were associated. This implies that they are affecting gene structure or regulation rather than dosage. There were even a small minority of cases in which increased transcription was associated with reduced copy number, implying a complex relationship.

In principle, it might not be necessary to measure both types of variation when carrying out an association study. Just as some SNPs are representative of others through shared ancestry and therefore linkage disequilibrium, the SNP variation might overlap with and therefore represent the CNV variation. However, the authors found that fewer than 20% of their CNV associations had a corresponding SNP association.

This study shows the importance of copy number variation in complex phenotypes, and the need to measure it without relying on SNPs to represent it. However, only a fraction of copy number variation was measured — future work will look at larger and more distant variants. It will also be important to test associations with phenotypes other than gene expression.