The human variome: genomic and epigenomic diversity.

Work from the HUGO Pan-Asian Single Nucleotide Polymorphism (SNP) consortium has mapped human genetic diversity in different subpopulations in Asia (Abdulla et al, 2009). Now this highlights the question whether disease-SNP associations identified by genome-wide association studies (GWAS) in one subpopulation can be applied to other subpopulations or whether separate GWAS should be considered for different subpopulations because of their different SNP distributions (Haiman & Stram, 2010).

DNA methylation is an epigenetic mark in the mammalian genome. ‘Epigenetic diversity’ encompasses variation in DNA methylation and like ‘genetic diversity’, ‘epigenetic diversity’ may also co-vary with phenotype (Feinberg et al, 2010). Mechanistically, disease-associated SNPs may be found in DNA regulatory regions where the genetic variation affects cognate binding of transcription factor complexes, effecting on the expression of disease-relevant genes (Harismendy et al, 2011). Similarly, variation in the methylation of a DNA regulatory region (either increased or decreased methylation) can also alter the binding of transcription factor complexes and regulate both distal and proximal gene expression (Choy et al, 2010; Jones & Takai, 2001; Phillips & Corces, 2009; Weaver et al, 2004). The availability of genome-wide maps of DNA methylation (Choy et al, 2010; Down et al, 2008; Lister et al, 2009) led us to examine whether a relationship exists between DNA methylation and regions of heterogeneous genetic variation.

The Fixation Index (Fst) is a measure of fixation rate of SNP between two subpopulations. SNPs with high Fst are ones with more heterogeneity in a subpopulation compared to another. Using DNA methylation maps that were generated from individuals of European ancestry, we made the interesting discovery that high Fst-SNPs (Duan et al, 2008) (N = 8751) from European ancestry showed an aggregate methylation density pattern of modulation centred on SNPs compared to flanks of the SNPs (Fig 1). Methylation maps from Europeans showed density modulation only at sites of European high Fst-SNPs (i.e. high Fst-SNPs for European vs. African + European vs. Asian), but not sites of Asian or African high Fst-SNPs (i.e. high Fst-SNPs for Asian vs. European; or African vs. European, Fig 1). Similarly aggregate methylation density modulation was also absent in a negative control set of randomly selected SNPs (N = 8000). On the basis that increased DNA methylation may be a means of regulating transcription factor complex binding and therefore mark genomic regions that are important to control by methylation (Choy et al, 2010; Jones & Takai, 2001), our findings support the notion that sites of heterogeneous genetic variation in one subpopulation are functionally relevant to the corresponding subpopulation, but sites of heterogeneous genetic variation of a different subpopulation, are not.

Figure 1

Methylation density across sites of SNPs

Methylation maps from hearts (blue) and sperm (green) from Europeans were used and aggregate methylation density was plotted against sites −1000 to +1000 bp of European, Asian, African high-Fst-SNPs, and random SNPs. European methylation density showed a statistically significant pattern of modulation only in European high Fst-SNPs but not high Fst-SNPs of other subpopulations (one-way ANOVA for methylation density at red and yellow bars, p < 0.05 for hearts and p < 0.001 for sperm). Aggregate methylation density (BATMAN score) was generated and plotted using algorithms previously described (Choy et al, 2010; Down et al, 2008). MeDIP-seq (Methylated DNA-immunoprecipitation followed by high-throughput sequencing) data were generated from four human hearts and spermatozoa and analyzed using a Bayesian deconvolution strategy (BATMAN score) (Choy et al, 2010; Down et al, 2008). BATMAN scores from four healthy human hearts were averaged using Perl script written by HGG and MKC as described in Choy et al, (2010). Average plots of methylation densities were calculated using an algorithm found in a public resource (Kolasinska-Zwierz et al, 2009).

This is consistent with the need to consider subpopulation differences in genetic variation when studying disease-SNP association. By extension, we propose that the epigenomes of subpopulations should also be considered as the International Human Epigenome Consortium gears up to generate 1000 reference epigenomes (Beck, 2010). Subpopulation differences will be important when drawing conclusions from associations between disease and the human variome.