Linkage disequilibrium of evolutionarily conserved regions in the human genome

Abstract

Background: The strong linkage disequilibrium (LD) recently found in genic or exonic regions of the human genome demonstrated that LD can be increased by evolutionary mechanisms that select for functionally important loci. This suggests that LD might be stronger in regions conserved among species than in non-conserved regions, since regions exposed to natural selection tend to be conserved. To assess this hypothesis, we used genome-wide polymorphism data from the HapMap project and investigated LD within DNA sequences conserved between the human and mouse genomes.

Results: Unexpectedly, we observed that LD was significantly weaker in conserved regions than in non-conserved regions. To investigate why, we examined sequence features that may distort the relationship between LD and conserved regions. We found that interspersed repeats, and not other sequence features, were associated with the weak LD tendency in conserved regions. To appropriately understand the relationship between LD and conserved regions, we removed the effect of repetitive elements and found that the high degree of sequence conservation was strongly associated with strong LD in coding regions but not with that in non-coding regions.

Conclusion: Our work demonstrates that the degree of sequence conservation does not simply increase LD as predicted by the hypothesis. Rather, it implies that purifying selection changes the polymorphic patterns of coding sequences but has little influence on the patterns of functional units such as regulatory elements present in non-coding regions, since the former are generally restricted by the constraint of maintaining a functional protein product across multiple exons while the latter may exist more as individually isolated units.

Figure 1

A moving average of the fraction of complete or nearly complete LD (r² > 0.8) versus distance between SNPs. All panels are those for CEU. See Additional file 1 for CHB, JPT, and YRI, which show the same tendency. (A) Plots of LD within DNA sequences conserved between the human and mouse genomes (in red with Xs), non-conserved regions (regions other than conserved ones; shown in red with circles), genic regions (in blue with Xs), and non-genic regions (in blue with circles). (B) Plots of LD within intersections of non-genic regions with conserved (in red with Xs) and non-conserved (in red with circles) regions, and of genic regions with conserved (in blue with Xs) and non-conserved (in blue with circles) regions. (C) Plots of LD within intersected regions of centromeric regions (the 10% definition, we only show plots in the 10% definition because of the same tendency in the 5% definition) with conserved (in red with Xs) and non-conserved (in red with circles) regions, of telomeric regions with conserved (in blue with Xs) and non-conserved (in blue with circles) regions, and of the residual regions (neither centromeric nor telomeric) with conserved (in green with Xs) and non-conserved (in green with circles) regions. (D) LD fractions for SNP pairs within highly conserved and less highly conserved regions (black and green), highly and less highly conserved non-coding regions (blue and light blue), and regions enriched (>20% in the bases) with highly and less highly conserved coding regions (red and pink). We selected only regions where the proportion of repeats was <20%, and since after this adjustment we found outliers of LD related to extreme GC-content, we further selected regions where the GC-content was 45–65%.