Both selective and neutral processes drive GC content evolution in the human genome

Abstract

Background: Mammalian genomes consist of regions differing in GC content, referred to as isochores or GC-content domains. The scientific debate is still open as to whether such compositional heterogeneity is a selected or neutral trait.

Results: Here we analyze SNP allele frequencies, retrotransposon insertion polymorphisms (RIPs), as well as fixed substitutions accumulated in the human lineage since its divergence from chimpanzee to indicate that biased gene conversion (BGC) has been playing a role in within-genome GC content variation. Yet, a distinct contribution to GC content evolution is accounted for by a selective process. Accordingly, we searched for independent evidences that GC content distribution does not conform to neutral expectations. Indeed, after correcting for possible biases, we show that intron GC content and size display isochore-specific correlations.

Conclusion: We consider that the more parsimonious explanation for our results is that GC content is subjected to the action of both weak selection and BGC in the human genome with features such as nucleosome positioning or chromatin conformation possibly representing the final target of selective processes. This view might reconcile previous contrasting findings and add some theoretical background to recent evidences suggesting that GC content domains display different behaviors with respect to highly regulated biological processes such as developmentally-stage related gene expression and programmed replication timing during neural stem cell differentiation.

Figure 1

Comparison of allele frequency spectra. (A) Quantile-quantile plots of GC->AT and AT-> GC derived allele frequencies for highly (red) and low (blue) recombining intronic regions after fixing GC content. (B) The same as (A), but in this case we fixed recombination rates and compared high (red) vs low (blue) GC regions. (C) The same as (A), but in this case we fixed both GC content and recombination rates in order to compare regions from highly (red) vs low (blue) expressed genes.

Figure 2

Analysis of fixed versus polymorphic retrotransposon insertions. (A) Analysis of average GC content flanking polymorphic (P, white) and fixed (F, gray) retrotransposons. GC content was calculated in 5 kb flanking the repeat. The number of repeat instances is also indicated. GC content is significantly higher for regions flanking fixed compared to polymorphic Alus ; the same holds for SVAs. (B) Analysis of polymorphic (white) and fixed (gray) retrotransposon relative frequency in different isochores (L1 to H3, ordered from 1 to 5, as described in [20]). Fixed Alus are significantly enriched in heavy isochores compared to polymorphic instances.

Figure 3

Analysis of GC content distribution in human introns with different isochoric location. Isochore definition is as described in [20]. (A) Scatter plot and loess fitting of intron size and GC content in light (blue) and heavy (red) isochores. (B) Analysis of GC₂₀₀ (see text). GC₂₀₀ significantly increases or decreases with residual size (percentile classes are shown) for introns located in heavy (red; breaks in bp = 681, 934, 1309, 1960, 3665) or light (blue; breaks in bp = 810, 1181, 1714, 2638, 5476) isochores, respectively (Kruskall Wallis Test, p = 1.3 × 10^-34 and 7.9 × 10^-7, respectively). The number of introns in each size class amounted to 2490 and 1567 for heavy and light isochores, respectively. (C) Distributions of within-gene correlation coefficients. For each gene having more than 15 introns (n = 500 and 1021 for light and heavy isochores, respectively) we calculated correlation coefficients between masked GC content and residual size. Hatched and dotted lines represent envelopes (1^st and 99^th percentiles, respectively) of correlation coefficient distributions obtained by randomization. (D) Scatter plot and loess fits of GC content over intron size (log₁₀ values) for introns (upper panel) and pseudointrons (lower panel). Spearman correlation coefficients (rho) are also shown (all p values were < 0.01). Introns and pseudointrons were divided on the basis of their isochoric location: blue for light isochores (501 introns-pseudointrons pairs), red for heavy ones (926 pairs).