A second-generation assembly of the Drosophila simulans genome provides new insights into patterns of lineage-specific divergence

Abstract

We create a new assembly of the Drosophila simulans genome using 142 million paired short-read sequences and previously published data for strain w(501). Our assembly represents a higher-quality genomic sequence with greater coverage, fewer misassemblies, and, by several indexes, fewer sequence errors. Evolutionary analysis of this genome reference sequence reveals interesting patterns of lineage-specific divergence that are different from those previously reported. Specifically, we find that Drosophila melanogaster evolves faster than D. simulans at all annotated classes of sites, including putatively neutrally evolving sites found in minimal introns. While this may be partly explained by a higher mutation rate in D. melanogaster, we also find significant heterogeneity in rates of evolution across classes of sites, consistent with historical differences in the effective population size for the two species. Also contrary to previous findings, we find that the X chromosome is evolving significantly faster than autosomes for nonsynonymous and most noncoding DNA sites and significantly slower for synonymous sites. The absence of a X/A difference for putatively neutral sites and the robustness of the pattern to Gene Ontology and sex-biased expression suggest that partly recessive beneficial mutations may comprise a substantial fraction of noncoding DNA divergence observed between species. Our results have more general implications for the interpretation of evolutionary analyses of genomes of different quality.

Figure 1.

Dotplot for chromosomes X (A) and 3L (B) comparing our assembly to that of D. melanogaster (light gray) and D. simulans reference assembly (dark gray) (Begun et al. 2007). An inversion on the X (circled) is an example of a misassembly detected in the Begun et al. (2007) D. simulans assembly spanning X:13361146–13723239.

Figure 2.

Sequence quality metrics by gene. The number of D. simulans orthologs for which the amount of informative non-N codons, frameshifts, and premature stop codons generated from Dsim_ref differs from Dsim_w501 (in total, 11,053 genes are compared).

Figure 3.

Comparison of estimated lineage-specific divergence rates using the two D. simulans assemblies. The fraction of genes, by chromosome and site class, for which the estimated divergence rate per gene is different (greater or less than) depending on the D. simulans assembly used. Estimated rates for D. melanogaster are shown on top and D. simulans on bottom. (Right) The ratio of the number of genes for which the estimate from Dsim_ref  >  Dsim_w501 relative to Dsim_ref  <  Dsim_w501. Note that putatively neutral intron_FEI sites correspond to bases 8–30 of introns shorter than 100 bp (Halligan and Keightley 2006; Parsch et al. 2010).

Figure 4.

Comparison of intralineage and interlineage clustering of nonsynonymous substitutions. For all comparisons, (the correlation in divergence between nonsynonymous substitutions) decreases with increasing distance separating two nonsynonymous substitutions, specific to the lineage from which the substitution arose (polarized), following Figure 4A of Callahan et al. (2011). The amount of intralineage clustering within D. simulans (red line) and D. melanogaster (blue line) relative to interlineage clustering (black line) is shown separately for the two D. simulans assemblies: Dsim_ref (A); Dsim_w501 (B). The excess of intralineage relative to interlineage clustering in D. simulans is represented by the area between the red and blue curves for the first 20 codons [, shaded in gray]. Note that the same set of genes from both assemblies is analyzed, and D. simulans orthologs from either assembly containing a premature stop codon or non-start or stop codons are excluded. (B) Unlike the pattern from Dsim_ref, a similar amount of intralineage clustering is found in both the D. simulans and D. melanogaster lineages when using Dsim_w501 (greater overlap between blue and red lines; shaded region in gray is smaller). The intralineage clustering excess relative to the extent of intralineage clustering is shown in the inset (following Fig. 4B of Callahan et al. 2011). A considerably larger intralineage clustering excess is found in the D. simulans lineage when using Dsim_ref, which is also circled in the inset.

Figure 5.

Boxplot distribution of lineage-specific divergence by gene across different site classes in autosomal (2 + 3) genes in D. simulans and D. melanogaster. For each site class, the top darker bar represents the distribution across D. simulans, and the bottom lighter bar for D. melanogaster (each gene must contain a minimum of 10 non-N sites for intron_FEI sites and 100 otherwise). (*) The weighted average (based on the number of sites corresponding to the annotated class for each gene) across all genes. See also Table 2.

Figure 6.

X/autosome divergence ratio in D. simulans and D. melanogaster. The X/A divergence ratio is expected to be unity assuming that the X and autosomes have the same effective population size in the ancestor of these species. In support of this, synonymous nucleotide diversities on the X and autosomes are approximately equal in African populations of both species (Andolfatto 2001). The distribution for X/autosome divergence ratios reflects 10,000 bootstrap samples with replacement by gene across the various site classes, separately for genes on the X and autosomes. Refer to Supplemental Table 7 for bootstrap P-values.