Population genomics of rapid adaptation by soft selective sweeps

Abstract

Organisms can often adapt surprisingly quickly to evolutionary challenges, such as the application of pesticides or antibiotics, suggesting an abundant supply of adaptive genetic variation. In these situations, adaptation should commonly produce 'soft' selective sweeps, where multiple adaptive alleles sweep through the population at the same time, either because the alleles were already present as standing genetic variation or arose independently by recurrent de novo mutations. Most well-known examples of rapid molecular adaptation indeed show signatures of such soft selective sweeps. Here, we review the current understanding of the mechanisms that produce soft sweeps and the approaches used for their identification in population genomic data. We argue that soft sweeps might be the dominant mode of adaptation in many species.

Figure Box 1

Figure 1

Definition of hard and soft sweeps. (A) In a hard sweep, all adaptive alleles in the sample arise from a single mutation (depicted by x) and coalesce after the onset of positive selection (dotted line). Note that even if the mutation had arisen prior to the onset of positive selection and was present as standing genetic variation, this would still be considered a hard sweep as long as only a single lineage is ultimately present in the sample. (B) In a soft sweep from recurrent de novo mutations, the adaptive alleles in the sample arose from at least two independent mutation events after the onset of positive selection and the lineages coalescence prior to the onset of positive selection. (C) In a soft sweep from the standing genetic variation, adaptive alleles were already present at the onset of positive selection. The different lineages in a population sample can originate from independent mutation events (i) or from a single mutation that reached some frequency prior to the onset of positive selection, such that several copies present at that time then swept through the population (ii). In this latter case, the population genetic signatures of the sweep will depend on the time τ between coalescence and onset of positive selection. If τ is short, the sweep will appear similar to a hard sweep, whereas when τ is large, it will be similar to a soft sweep from several de novo mutations.

Figure 2

Soft sweep examples in population genomic data. (a) Haplotypes of the HIV reverse transcriptase observed in two samples taken from the same patient prior to treatment (day 0; samples S1–S7) and after resistance had evolved (day 28; samples S8–S14) from [37]. Treatment resistance involves a single amino acid change from lysine to asparagine in the codon spanning positions 307–309 (grey columns). The original AAA codon was replaced by a mixture of AAC and AAT codons that both encode for asparagine. (b) Soft sweep in humans in the lactase gene from [38]. The top panel shows homozygosity tracts in African individuals that carry the persistent C-14010 allele (red) versus those that carry the non-persistent G-14010 allele (blue). The bottom panel shows tracts for Eurasian individuals that cary the persistent T-13910 allele (green) versus those that carry the non-persistent C-13910 allele (orange). (c) Soft sweep during the evolution of pesticide resistance in D. melanogaster from [41]. The table shows the observed haplotypes in a region of the Ace gene from flies sampled in North America and Australia. D. melanogaster evolved in Africa and then spread worldwide via Europe (lower panel). The A to G mutation at position 14870 of Ace increases resistance to several commonly used pesticides. NA1 and NA2 are commonly observed sensitive haplotypes in North America and samples S1–S9 show the haplotypes of nine resistant flies collected in North America. AUS is a commonly observed sensitive haplotype in Australia and sequences S10–S16 show the haplotypes of seven resistant flies collected in Australia. In both locations, resistance seems to have evolved on the locally prevailing sensitive haplotypes. (d) Haplotype homozygosity statistics. The top row depicts a hard sweep with a single common haplotype and several low-frequency variants; the bottom row depicts a soft sweep with two common haplotypes. The total grey area in the left panel specifies haplotype homozygosity $H_1={\sum }_i{x}_i^2$. The middle panel shows extended haplotype homozygosity H₁₂, obtained after combining the frequencies of the two most common haplotypes. The right panel shows haplotype homozygosity calculated after removing the most frequent haplotype. H₁ is larger (and H₂ smaller) for the hard sweep than for the soft sweep. H₁₂ is similar in both scenarios. (e) H₁₂ scan for chromosomes 2R and 3R of D. melanogaster from [64]. The three most prominent peaks coincide with three well-known cases of adaptation at the loci Cyp6g1, Ace, and CHKov1.

Figure 3

Likelihood of hard and soft sweeps and relevant timescales. (A) The red curve shows the frequency trajectory of an adaptive mutation. The blue curve shows the trajectory of another de novo adaptive mutation that successfully established before the first one became fixed in the population. This scenario is likely when establishment time T_e is shorter than fixation time T_f. (B) Adaptation in a subdivided population with two demes and migration. An adaptive mutation arises and sweeps through the first deme (red trajectory). The allele can subsequently migrate and also sweep in the second deme (dashed red trajectory), resulting in a global hard sweep. Alternatively, an independent de novo adaptive mutation can arise first and sweep in the second deme (dashed blue trajectory), resulting in a global soft sweep. (C) Adaptation in a spatially continuous population with limited dispersal. An adaptive mutation arises at one geographic location (red area) and then spreads through the population in a radial wave with speed ν (red circles). While this mutation is still spreading, another de novo adaptive mutation arises at a different location that has not yet been covered by the first mutation (blue area). The characteristic length χ specifies the average distance traveled by an adaptive mutation until another successful mutation is expected to have arise within its already covered area.

Figure 4

Soft sweeps and demography. (A) Probability of soft sweeps under recurrent population bottlenecks. Every ΔT generations the population size drops from N₁ to N₂ ≪ N₁. During the boom phase, Θ₁ > 1, but Θ₂ ≪ 1 during the bottleneck. Soft sweeps that emerge during a boom phase remain soft throughout the next bottleneck only if at least two mutations reached a frequency x = 1/N₂ such that they are likely to survive this bottleneck. (B) Difference in variance and coalescence N_e in the presence of a population bottleneck. The timescale of neutral coalescence (T_c) is primarily determined by the time since the bottleneck. The value of coalescence N_e inferred from the levels of neutral variation can thus be much smaller than the value of the present-day variance N_e, estimated over the much shorter timescale (T_a) relevant for recent adaptation.