Results

Phylogenomic analyses

Multidimensional scaling (MDS) plots based on a distance matrix generated in ngsdist of 2807 of the most informative SNPs reveal differentiation between global samples, especially along the first and section coordinate axes (Figure 3 ). Genetic differentiation among individuals in each of the 50 populations detected in our SNP data based on MDS plots was low as expected for a selfing plant. Six distinct genotype groups (groups I – VI) are clearly distinguishable. All individuals collected from the same population are genetically uniform such that they form part of the same genotype group except in the case of Denmark and Alaska, where individuals are split between two different groups and/or include a few outlying individuals. Some of the closest geographic populations sampled from the native range were resolved in different groups (i.e. France1 belongs to group I and France2 belongs to group II; Iran1 belongs to group IV and Iran2 belongs to group II). From left to right in Figure 3, the first genotype group (group I) includes populations from France1, Iceland, Ireland, Denmark (in part), Russia, Ukraine, North Dakota (USA), Washington (USA), New Brunswick (CAN), and New Zealand. The second (group II) includes France2, England, Netherlands, Denmark (majority of samples), Sweden, Norway, Finland, Estonia, Italy, Turkey, Iran2, Greenland, and Alberta (CAN), Colorado (USA), Ontario (CAN), Vancouver (CAN), Chicago (USA), Newfoundland (CAN), and Alaska (USA, majority of samples). These two genotype groups primarily consist of populations that originate above the 35–40o N latitude range (except for New Zealand;Figure 4b ). Individuals collected from Yukon (CAN) form their own genotype group (group III). Populations from Japan, South Korea, Iran1, California (USA), and the remaining individuals from Alaska form a fourth group (group IV). A fifth genotype group (group V) consists of populations from Spain (Spain1, Spain2), Greece, Gran Canaria, Morocco, Brazil. The final group (group VI) consists of populations from Portugal, Egypt, Tenerife, South Africa, Australia (Melbourne and Perth), Peru, Chile, Hawaiʻi (USA) and Florida (USA). Populations from groups V and VI all occur below the 35–40o N latitude range.

Demographic history & admixture

Shared ancestry was estimated on genotype likelihoods using 7594 SNPs in ngsadmix for values of K (ancestral populations) between two and 12, and results for value K =3 and K =8 are illustrated as a structure graph in Figure 4a . The colours represent estimated shared ancestry for all sampled populations for each value of K . Each population comprised 6 – 10 individuals for which genotyping was performed. Results for additional values ofK are shown in Supplementary Materials. At a K value of 3, populations from eastern Asia (Japan, South Korea), Yukon, Alaska (in part), California, and Iran1 comprise of similar ancestry components. Populations from North America (except for Florida), Iceland, Greenland, and New Zealand share ancestry with north-central and eastern Europe and western Asia (Russia, Turkey, Iran2). Populations from South America, Florida, Hawaiʻi, Africa, and Australia share ancestry with southern European populations (Figure 4a ). Individual populations from both introduced and native ranges are very homogenous such that very little admixture is seen within each population, except for populations from Alaska, California, and New Zealand. The population from New Zealand, even at the lowest K value (K = 2), shows more admixture than other populations, and although individuals from New Zealand share most of their ancestry with populations from Northern Europe and North America, there is evidence of admixture with lineages from Southern Europe, South America, Hawaiʻi, Africa, and Australia.
At a K value of 6, the groupings of shared ancestry resemble the groupings seen in the MDS plot (Figure 3 ) and therefore represent a likely ancestry scenario to base discussions on. Two notable differences are found here compared to the MDS plot. First, Japan and South Korea are inferred to have a different ancestry than do California and Iran1 further splitting group IV as shown in the MDS plot. Secondly, populations belonging to groups V and VI in the MDS plot are not split into unique ancestries. Only three of the 50 populations sampled (Denmark, Alaska and New Zealand) showed signs of admixture between two groups at this value of K .

Population splits and migration

The maximum likelihood tree resulting from the treemix analyses (Figure 5 ) shows a pattern similar to the genotype groups seen under MDS plots and ngsadmix results. The three most basal groups are similar to the groupings seen in admixture analyses forK =3, except for the positions of California and Iran which are shown to be sister to genotype groups V and VI (occurring below the 35–40o N latitude range) rather than more closely related to Japan, South Korea and Yukon. Our treemix results suggest that geneflow occurred between most of the genotype groups. Of the four migration events modelled, the strongest migration weight involves geneflow between populations from Alberta to populations in North Dakota (genotype group I to group II). The other migration events inferred include migration from Spain2 to New Zealand (genotype group V to group II), Yukon to Ukraine and Russia (genotype group III to group I), and Spain1 to Iran2 (genotype group V to group I).