Rapid genomic convergent evolution in experimental populations of Trinidadian guppies (Poecilia reticulata)

Abstract

Although rapid phenotypic evolution has been documented often, the genomic basis of rapid adaptation to natural environments is largely unknown in multicellular organisms. Population genomic studies of experimental populations of Trinidadian guppies (Poecilia reticulata) provide a unique opportunity to study this phenomenon. Guppy populations that were transplanted from high-predation (HP) to low-predation (LP) environments have been shown to evolve toward the phenotypes of naturally colonized LP populations in as few as eight generations. These changes persist in common garden experiments, indicating that they have a genetic basis. Here, we report results of whole genome variation in four experimental populations colonizing LP sites along with the corresponding HP source population. We examined genome-wide patterns of genetic variation to estimate past demography and used a combination of genome scans, forward simulations, and a novel analysis of allele frequency change vectors to uncover the signature of selection. We detected clear signals of population growth and bottlenecks at the genome-wide level that matched the known history of population numbers. We found a region on chromosome 15 under strong selection in three of the four populations and with our multivariate approach revealing subtle parallel changes in allele frequency in all four populations across this region. Investigating patterns of genome-wide selection in this uniquely replicated experiment offers remarkable insight into the mechanisms underlying rapid adaptation, providing a basis for comparison with other species and populations experiencing rapidly changing environments.

Keywords: Convergent evolution; Poecilia reticulata; experimental evolution; guppies; population genomics; rapid evolution.

Figure 1

Summary of experimental populations. Map (A) highlights the Guanapo river with the sampled populations, and an inset shows the experimental rivers. “Open” indicates sites with manipulated canopies, “closed” indicates intact canopies. (B) Summary of mark‐recapture census data (from Reznick et al. 2019) per sex, annotated with introduction sizes (in black), and years where phenotypic evolution was documented. Estimates and confidence intervals are from a POPAN survival model implemented in program MARK. (C) Census data for all four populations plotted on the same axis.

Figure 2

Population structure, runs of homozygosity, and neutral population statistics across the introduction sites and the HP source. PCA (A) with populations colored according to river; 75‐kb window‐based estimates of (B) expected heterozygosity (H _e), observed heterozygosity (H _o), nucleotide diversity (π), and Tajima's D (D) for each of the introduced populations and their HP source. (C) Number and sum of runs of homozygosity (ROH) in different size classes, per population.

Figure 3

Haplotype genome scans (A) XP‐EHH and (B) iHH12 for each introduced population across the genome. Chromosome 15 (C) XP‐EHH and (D) iHH12. Yellow points indicate unique outlier windows among the populations, pink points show overlapping outlier windows in two of the four populations, and orange points indicate windows overlapping in three of the four comparisons; there was no overlap among all four comparisons. Outliers are defined as windows with XP‐EHH values >2.5 and iHH12 values >5.

Figure 4

Eigenvector analysis. Transformed P‐values for eigenvector 1 across the genome. (B) Chromosome 15. (C) Chromosome 8. Red points indicate those at 99.9% quantile of the null distribution.

Figure 5

Forward simulated standardized allele frequency change, as estimated by selection coefficients based on neutral evolution and census size compared to observed selection coefficients. (A) Across the genome. (B) Across chromosome 15. Bars across the genome represent the expected number of outlier SNPs above the simulated 99.9% cutoff given the relative size of each chromosome compared with the observed number. Points along chromosome 15 show the per‐SNP summed selection coefficients, representing the summed estimates of standardized allele frequency change between the source (GHP) and each introduced population. Values are maximized when standardized allele frequency change is large in all populations and allele frequency change is in the same direction (same sign for estimated selection coefficient). Colors indicate SNPs with summed selection coefficients above the 99% (yellow), 99.9% (orange), and 99.99% quantiles defined by the neutral simulations based on census size estimates.