Search

Evolution can repeat itself, resulting in parallel adaptations in independent lineages occupying similar environments. Moreover, parallel evolution sometimes, but not always, uses the same genes. Two main hypotheses have been put forth to explain the probability and extent of parallel evolution. First, parallel evolution is more likely when shared ecologies result in similar patterns of natural selection in different taxa. Second, parallelism is more likely when genomes are similar because of shared standing variation and similar mutational effects in closely related genomes. Here we combine ecological, genomic, experimental and phenotypic data with Bayesian modelling and randomization tests to quantify the degree of parallelism and its relationship with ecology and genetics. Our results show that the extent to which genomic regions associated with climate are parallel among species of Timema stick insects is shaped collectively by shared ecology and genomic background. Specifically, the extent of genomic parallelism decays with divergence in climatic conditions (that is, habitat or ecological similarity) and genomic similarity. Moreover, we find that climate-associated loci are likely subject to selection in a field experiment, overlap with genetic regions associated with cuticular hydrocarbon traits and are not strongly shaped by introgression between species. Our findings shed light on when evolution is most expected to repeat itself.

Additional methods and results

Tables S3 and S6; Figures S1-S3

A fastSTRUCTURE analysis based on all SNPs and for all populations of R. pomonella included in this study, considering K=2–6 subpopulations.

Correlation coefficients between SNP allele frequency differences in the hawthorn host race and R. mendax eclosion time GWAS for all polymorphic loci genotyped in the two taxa and those mapping to chromosomes 1–5. Probability (P) level determined by Monte-Carlo non-parametric computer simulations (n=10,000 replicates). n SNPs=number of SNPs in a given comparison.

Studies assessing the predictability of evolution typically focus on short-term adaptation within populations or the repeatability of change among lineages. A missing consideration in speciation research is to determine whether natural selection predictably transforms standing genetic variation within populations into differences between species. Here, we test whether and how host-related selection on diapause timing associates with genome-wide differentiation during ecological speciation by comparing ancestral hawthorn and newly formed apple-infesting host races of Rhagoletis pomonella to their sibling species R. mendax that attacks blueberries. The associations of 57 857 single nucleotide polymorphisms in a diapause genome-wide-association studies (GWAS) on the hawthorn race strongly predicted the direction and magnitude of genomic divergence among the three fly populations at a field site in Fennville, MI, USA. The apple race and R. mendax show parallel changes in the frequencies of putative inversions on three chromosomes associated with the earlier fruiting times of apples and blueberries compared to hawthorns. A diapause GWAS on R. mendax revealed compensatory changes throughout the genome accounting for the earlier eclosion of blueberry, but not apple flies. Thus, a degree of predictability, although not complete, exists in the genomics of diapause across the ecological speciation continuum in Rhagoletis. The generality of this result is placed in the context of other similar systems.This article is part of the theme issue ‘Towards the completion of speciation: the evolution of reproductive isolation beyond the first barriers'.

Correlation coefficients (r) between SNP allele frequency differences in the R. mendax eclosion time GWAS versus the differences between R. mendax and the hawthorn host race (Blue-Haw) and R. mendax and the apple race (Blue-Ap) at the Fennville site. Results are given for all variable SNPs in the R. mendax eclosion study, as well as for loci mapping to chromosomes 1-5 and for high (H), medium (M), and low LD classes of loci on each chromosome (Chrome), as determined in Egan et al. (2015) and Ragland et al. (2017). Probability (P) levels determined by Monte-Carlo, non-parametric computer simulations (n=10,000 replicates), with significant relationships bolded. n SNPs=number SNPs in comparison.

Genomic correlation between eclosion variation and geographic differentiation in the R. pomonella hawthorn race in the Midwest USA, based on data from Ragland et al. (2017) and Doellman et al. (2019), and reanalyzed using the 44,749 SNPs called in this study. Scatterplots depict the relationship between the difference in allele frequency between late versus early eclosing samples of hawthorn flies (x-axis; eclosion time GWAS) and the difference in allele frequency between Urbana, Illinois (south) and Fennville, Michigan (north) study sites (y-axis; geographic cline). (A) graph for all SNPs genotyped in the study (r=0.446, PR. pomonella that has been hypothesized to take the form of clinal variation due to past geographic isolation, divergence, and secondary contact between hawthorn-infesting populations in Mexico and the USA (Feder et al. 2003).

Linkage disequilibrium (LD) network plots for SNPs mapping to chromosomes 1–5, demonstrating correspondence between high LD loci between R. pomonella and R. mendax. Loci are shown as nodes (circles) connected to one another in clusters by edges, when pairwise LD (r2) between loci exceeds 0.8. Tight clusters indicate this high pairwise LD between many SNPs, while unconnected SNPs have a pairwise LD below 0.8 with all other SNPs on the same chromosome. Within population LD is represented for the hawthorn host race of R. pomonella (panels A-E) and the blueberry-infesting population of R. mendax (panels F-J). Black circles represent high LD SNPs residing in the putative inversions segregating within R. pomonella. Note that on chromosomes 3 and 5, many of the SNPs in the high LD clusters in R. pomonella are also in high LD with one another in R. mendax. This suggests that the two species share large inversion polymorphisms on these two chromosomes.

Correlation coefficients (r) between SNP allele frequency differences in the hawthorn race eclosion time GWAS versus the differences between the apple and hawthorn host races (Ap-Haw) and R. mendax and the hawthorn race (Blue-Haw) at the Fennville site. Results are given for all variable SNPs in the hawthorn eclosion study, as well as for loci mapping to chromosomes 1–5 and for high (H), medium (M), and low LD classes of loci on each chromosome (Chrome), as determined in Egan et al. (2015) and Ragland et al. (2017). Probability (P) levels determined by Monte-Carlo, non-parametric computer simulations (n=10,000 replicates), with significant relationships bolded. n SNPs=number SNPs in comparison.

OSM 1: Additional information on Gambusia collection sites. (DOCX 2179Â kb)