Patterns of polymorphism and demographic history in natural populations of Arabidopsis lyrata

Abstract

Background: Many of the processes affecting genetic diversity act on local populations. However, studies of plant nucleotide diversity have largely ignored local sampling, making it difficult to infer the demographic history of populations and to assess the importance of local adaptation. Arabidopsis lyrata, a self-incompatible, perennial species with a circumpolar distribution, is an excellent model system in which to study the roles of demographic history and local adaptation in patterning genetic variation.

Principal findings: We studied nucleotide diversity in six natural populations of Arabidopsis lyrata, using 77 loci sampled from 140 chromosomes. The six populations were highly differentiated, with a median FST of 0.52, and structure analysis revealed no evidence of admixed individuals. Average within-population diversity varied among populations, with the highest diversity found in a German population; this population harbors 3-fold higher levels of silent diversity than worldwide samples of A. thaliana. All A. lyrata populations also yielded positive values of Tajima's D. We estimated a demographic model for these populations, finding evidence of population divergence over the past 19,000 to 47,000 years involving non-equilibrium demographic events that reduced the effective size of most populations. Finally, we used the inferred demographic model to perform an initial test for local adaptation and identified several genes, including the flowering time gene FCA and a disease resistance locus, as candidates for local adaptation events.

Conclusions: Our results underscore the importance of population-specific, non-equilibrium demographic processes in patterning diversity within A. lyrata. Moreover, our extensive dataset provides an important resource for future molecular population genetic studies of local adaptation in A. lyrata.

Figure 1. A map showing the locations and sample size (N) for the 60 long exons for each of the six populations studied: ICE = Iceland, GER = Germany, CAN = Canada; SWE = Sweden; RUS = Russia.

The results of structure analyses are shown below the map. The mostly likely number of clusters (k) is five. Considered separately, the USA/Canada individuals clearly differentiate into two geographically separate clusters.

Figure 2. Demographic models.

A) Schematic representation of the two-population bottleneck model used for parameter estimation. B) Schematic of the six-population model used for testing single-locus fit of F_ST. See text for parameter descriptions.

Figure 3. Silent site diversity within populations.

A) Shown are boxplots of θ_π, Tajima's D, the number of singletons η₁ and the population recombination rate ρ for each population. Bars represent the median, boxes the interquartile range, and whiskers extend to 1.5-times the interquartile range.

Figure 4. Summary statistics at silent sites among 77 loci in the German population.

Shown are θ_π per base pair, Tajima's D and range-wide F_ST. Loci are shown in the same order as in Supplementary Tables S1 and S2.

Figure 5. Pairwise population differentiation at silent sites.

A) shared and unique polymorphisms and fixed differences. B) F_ST. Boxplots show the median and interquartile distance of values across loci.

Figure 6. Posterior distributions for the parameters of the pairwise population divergence models.

Dashed curves represent the Bayesian prior for each parameter. Point estimates of the parameters for each population are shown in Table 1.