African genomes illuminate the early history and transition to selfing in Arabidopsis thaliana

Abstract

Over the past 20 y, many studies have examined the history of the plant ecological and molecular model, Arabidopsis thaliana, in Europe and North America. Although these studies informed us about the recent history of the species, the early history has remained elusive. In a large-scale genomic analysis of African A. thaliana, we sequenced the genomes of 78 modern and herbarium samples from Africa and analyzed these together with over 1,000 previously sequenced Eurasian samples. In striking contrast to expectations, we find that all African individuals sampled are native to this continent, including those from sub-Saharan Africa. Moreover, we show that Africa harbors the greatest variation and represents the deepest history in the A. thaliana lineage. Our results also reveal evidence that selfing, a major defining characteristic of the species, evolved in a single geographic region, best represented today within Africa. Demographic inference supports a model in which the ancestral A. thaliana population began to split by 120-90 kya, during the last interglacial and Abbassia pluvial, and Eurasian populations subsequently separated from one another at around 40 kya. This bears striking similarities to the patterns observed for diverse species, including humans, implying a key role for climatic events during interglacial and pluvial periods in shaping the histories and current distributions of a wide range of species.

Keywords: climate; evolution; migration; population history; self-compatibility.

Fig. 1.

Sample map of accessions included in this study. Herbarium samples are shown as squares. Abbreviations are as follows: Algeria (DZ), Cape Verde (CV), Central Asia (C.AS), Central Europe (C.EU), Eurasian nonrelicts (ENR), Eurasian relicts (ER), Germany (DE), Italy, Balkans, and Caucasus (IBC), Iberian nonrelicts (INR), Iberian relicts (IR), Morocco (MA), North Sweden (N.SE), South Africa (ZA), South Sweden (S.SE), Tanzania (TZ), Western Europe (W.EU).

Fig. 2.

Global population structure. (A) Unrooted neighbor-joining tree, (B) PCA, (C) ADMIXTURE results for K = 4.

Fig. 3.

Patterns of diversity across geographic regions. Distributions of genome-wide pairwise differences per base pair in: (A) worldwide comparison and (B) within and between African populations, where overlap between distributions is shown as described in legends. (C) Numbers of private SNPs and haplotypes found in each cluster. Error bars denote 95% confidence intervals.

Fig. 4.

Map of S-locus haplogroup diversity.

Fig. 5.

Historical effective population size of A. thaliana inferred using MSMC. Although two-haplotype analysis provides more resolution in the distant past, eight-haplotype analysis provides better resolution in the recent past. (A) Inference using pairs of haplotypes, with lines representing medians and shading representing ±1 SD calculated across pairs. This analysis is expected to produce unbiased estimates between 40 kya and 1.6 Mya (SI Appendix). (B) Inference based on sets of eight haplotypes with lines representing medians. This analysis is expected to produce unbiased estimates as recently as 1.6 kya.

Fig. 6.

Inferred timing of population splits. (A) Relative CCR between populations. Decreasing values from 1.0 indicate population separation. The dashed line represents historical temperature (63). (B) A schematic model for the demographic history of A. thaliana based on CCR results, with hashes to represent uncertainty regarding possible timing of gene flow events.