Pervasive natural selection in the Drosophila genome?

Abstract

Over the past four decades, the predominant view of molecular evolution saw little connection between natural selection and genome evolution, assuming that the functionally constrained fraction of the genome is relatively small and that adaptation is sufficiently infrequent to play little role in shaping patterns of variation within and even between species. Recent evidence from Drosophila, reviewed here, suggests that this view may be invalid. Analyses of genetic variation within and between species reveal that much of the Drosophila genome is under purifying selection, and thus of functional importance, and that a large fraction of coding and noncoding differences between species are adaptive. The findings further indicate that, in Drosophila, adaptations may be both common and strong enough that the fate of neutral mutations depends on their chance linkage to adaptive mutations as much as on the vagaries of genetic drift. The emerging evidence has implications for a wide variety of fields, from conservation genetics to bioinformatics, and presents challenges to modelers and experimentalists alike.

Figure 1. The effect of positive and negative selection on linked neutral sites.

This cartoon depicts a population of ten chromosomes, subject to recurrent selective sweeps (RSS) or background selection (BGS). Neutral mutations are shown as gray circles, the beneficial mutation in green, and deleterious mutations in red. RSS: An adaptive mutation destined for fixation arises on a particular haplotype, i.e., linked to a specific combination of neutral alleles at polymorphic sites. As it increases in frequency in the population, so does that genetic background. All pre-existing alleles not on the selected background are lost from the population, unless they recombine onto chromosomes carrying the beneficial allele before fixation. Thus, a “selective sweep” causes a reduction in the level of polymorphism as well as a distortion of allele frequencies in the vicinity of the beneficial substitution ,,. After fixation, diversity will be reintroduced by mutation, but a footprint of the substitution may remain for a long time (up to N _e generations; [78]). BGS: The balance between a steady flux of deleterious mutations and purifying selection generates a stable partition of chromosomes in a population, depending on how many deleterious mutations they carry. Chromosomes with deleterious mutations will be eliminated relatively quickly from the population by purifying selection, but this class is constantly replenished by new deleterious mutations. In the absence of recombination, a new neutral mutation can remain in the population for a long period of time and rise to high population frequencies only if it appears on a gamete that is free of deleterious mutations, and hence is not destined to be rapidly eliminated. The effect of this “background selection” against deleterious mutations is a reduction in the level of neutral polymorphism , as well a downward shift in their population frequencies, because of the relative excess of short-lived (and hence low frequency) neutral mutations .

Figure 2. Correlations in polymorphism data from D. melanogaster.

(A) Levels of synonymous site diversity versus recombination rates. The effects of the rate of amino acid divergence (K_a) and the rate of synonymous site divergence (K_s) have been controlled for by partial regression, with negative values set to zero. (B) K_s versus recombination rates. The effect of K_a has been controlled for by partial regression, with negative values set to zero. (C) A summary of the allele frequency spectrum at synonymous sites versus recombination rates; more negative values of the statistic reflect a higher proportion of rare alleles. The numerator is Tajima's D and the denominator is the minimum value D (in absolute value) can take given the sample size and number of segregating sites . (A–C) are based on the polymorphism data of Shapiro et al. , and recombination rates estimated by Comeron et al. . For the Shapiro et al. data, 349 loci with >50 synonymous sites were used and only African individuals are included. (D) Levels of synonymous site diversity as a function of K_a. In red are the 137 X-linked loci surveyed by Andolfatto . In black are autosomal loci surveyed by Shapiro et al. . For both data sets, the effect of K_s has been controlled for by partial regression, with negative values set to zero. For the Shapiro et al. data, 265 loci with recombination rates >0.5 cM/Mb and >50 synonymous sites were included. The red and black dotted lines represent average levels of synonymous π in the Andolfatto and Shapiro et al. datasets, respectively. Thick red and black lines indicate Lowess fits to the data. All p-values are one-tailed.

Figure 3. Cartoon of the effects of recurrent selective sweeps on patterns of genetic variation along the genome.

In this cartoon, several beneficial substitutions have occurred within this region, reducing levels of diversity relative to background levels. The sweep labeled 1 was driven by strong selection and occurred very recently, leading to a sharp decrease in diversity at linked sites. Sweep 2 was associated with a similarly strong selective coefficient, but occurred further in the past, such that levels of polymorphism surrounding the site have had some time to recover through mutation and random genetic drift. Sweep 3 occurred recently, but was associated with a weaker selective coefficient, thereby reducing polymorphism in a smaller region. We emphasize that, in practice, diversity patterns alone are likely to be an unreliable indicator of selective sweeps, as there are numerous other sources of heterogeneity.

Figure 4. The effects of genetic draft on the trajectory of a neutral allele.

(A) Simulated trajectory of a neutral allele affected by recurrent selective sweeps, from its origin on a single chromosome to fixation in the population. The population mutation and recombination parameters for this simulation are loosely based on estimates from D. melanogaster; the rate of adaptation, ν = 5×10⁻¹¹, and strength of selection, N_es = 10³, were taken from the high end of existing estimates. The allele spent the first ∼30,000 generations drifting around low frequencies (<5%). Then, at approximately the 30,000th generation, it increased sharply and rapidly in frequency (to ∼55%) because of linkage to a strongly advantageous mutation located approximately 80 kb away; it did not reach fixation, because of recombination during the ascent of the favored allele. Subsequent to this first, dramatic change in frequency, the mutant allele experienced three hitchhiking events that increased its frequency (selective sweeps 3 through 5) and one that decreased it (sweep 2). In (B) is a simulated trajectory of a neutral allele affected solely by genetic drift, for the same population parameters. Note the difference in the time scale of the two plots.