Migration without interbreeding: Evolutionary history of a highly selfing Mediterranean grass inferred from whole genomes

Abstract

Wild plant populations show extensive genetic subdivision and are far from the ideal of panmixia which permeates population genetic theory. Understanding the spatial and temporal scale of population structure is therefore fundamental for empirical population genetics - and of interest in itself, as it yields insights into the history and biology of a species. In this study we extend the genomic resources for the wild Mediterranean grass Brachypodium distachyon to investigate the scale of population structure and its underlying history at whole-genome resolution. A total of 86 accessions were sampled at local and regional scales in Italy and France, which closes a conspicuous gap in the collection for this model organism. The analysis of 196 accessions, spanning the Mediterranean from Spain to Iraq, suggests that the interplay of high selfing and seed dispersal rates has shaped genetic structure in B. distachyon. At the continental scale, the evolution in B. distachyon is characterized by the independent expansion of three lineages during the Upper Pleistocene. Today, these lineages may occur on the same meadow yet do not interbreed. At the regional scale, dispersal and selfing interact and maintain high genotypic diversity, thus challenging the textbook notion that selfing in finite populations implies reduced diversity. Our study extends the population genomic resources for B. distachyon and suggests that an important use of this wild plant model is to investigate how selfing and dispersal, two processes typically studied separately, interact in colonizing plant species.

Keywords: flowering time; heterozygosity; population structure; seed dispersal; selfing.

FIGURE 1

Islands of extended heterozygosity (IEH) among hard‐filtered SNPs (x‐axis) and among SNPs in high confidence regions (y‐axis), defined as regions where at least 90 percent of the accessions have a read coverage within 1.5 standard deviations of their mean coverage. (a) Overview of all 196 accessions, highlighting the six outlier accessions. (b) Zoom‐in on the small square in panel a

FIGURE 2

Geographic origin and genetic structure. (a) Map and pie charts showing shared ancestry. Only a subset of accessions is displayed for the dense local samples in Italy and France. (b) Cross‐entropy for increasing K in the ancestry analysis with sNMF. (c) Rooted phylogeny with divergence times, estimated under a multispecies coalescent model. (d) Principal component analysis. (e) Rooted phylogeny and coancestry matrix estimated with finestructure, as well as sNMF barplot for K = 5. The finestructure chunk length is the recombination map distance donated to individuals in rows from individuals in columns (log scale). Names in italics highlight regional populations, asterisks indicate small clusters: * Accessions from northern Italy, including a cluster of five accessions from Taro valley in Emilia‐Romagna. ** ABR9 collected in Albania and three accessions from the Gargano peninsula. *** Arn1 and Mon3. **** Three Turkish accessions clustering with Bd30‐1 from southern Spain. ***** Accessions from Sicily, Campania and Liguria. An interactive phylogenetic tree/map can be explored on http://github.com/cstritt/bdis‐phylogeo

FIGURE 3

Decay of linkage disequilibrium at different scales of population structure: (a) species‐wide, (b) in the three major genetic lineages, (c) in four geographic clades (C_Italia is not included as it is the sole representative of lineage C and thus already included at the lineage level), (d) four regional populations in Italy

FIGURE 4

Genetic diversity and differentiation. (a) Pairwise genetic differences (d _XY) within and between the five geographic clades. One panel is shown for each clade, the colours within the panel showing with which clade pairwise comparisons are made. (b) Correlation between geographic and genetic distance within the three major genetic lineages A, B and C. Colours within the panels indicate between which geographic regions pairwise comparisons are made. Correlation coefficients from Mantel test and their significance are displayed

FIGURE 5

Genetic differentiation across ecoregions in Italy. (a) Map showing the ecoregion classification of Blasi et al. (2014) and Brachypodium distachyon habitats in the four sampled regional populations. (b) Genetic diversity within the A_Italia clade, for which the most extensive sampling is available. Boxplots show the distribution of d _XY across the whole clade and within regional populations, while dotplots show how d _XY relates to geographic distance