Biological invasions in remote areas that experience low human activity provide unique opportunities to elucidate processes responsible for invasion success. Here we study the most widespread invasive plant species across the isolated islands of the Southern Ocean, the annual bluegrass, Poa annua. To analyze geographic variation in genome size, genetic diversity, and reproductive strategies, we sampled all major sub-Antarctic archipelagos in this region and generated microsatellite data for 470 individual plants representing 31 populations. We also estimated genome sizes for a subset of individuals using flow cytometry. Occasional events of island colonization are expected to result in high genetic structure among islands, overall low genetic diversity, and increased self-fertilization, but we show that this is not the case for Poa annua. Microsatellite data indicated low population genetic structure and lack of isolation-by-distance among the sub-Antarctic archipelagos we sampled, but high population structure within each archipelago. We identified high levels of genetic diversity, low clonality, and low selfing rates in sub-Antarctic P. annua populations (contrary to rates typical of continental populations). In turn, estimates of autogamy declined in populations as genetic diversity increased. Additionally, we found that most P. annua individuals are likely tetraploid and that only slight variation exists in genome size across the Southern Ocean. Our findings suggest multiple independent introductions of P. annua into the sub-Antarctic, which promoted the establishment of genetically diverse populations. Despite multiple introductions, the adoption of convergent reproductive strategies (outcrossing) happened independently in each major archipelago. The combination of polyploidy and a mixed reproductive strategy likely benefited P. annua in the Southern Ocean by increasing genetic diversity and its ability to cope with the novel environmental conditions.

All P. annua individuals (n = 470) were genotyped using the set of nine microsatellite loci previously developed by us (Mairal et al., 2022). PCR amplification of microsatellites was performed in two multiplex PCR reactions (see Mairal et al., 2022 for details). Given the sensitivity of multiplex reactions to genotyping errors (e.g., allele dropout), we adopted conditions recommended to minimize such errors, including the use of a standardized DNA concentrations (~100 ng/µL), special buffers for multiplexing, and selection of primers with similar annealing temperatures (Guichoux et al., 2011). All PCR reactions were carried out in 15 μL reaction volumes containing 1.5 μL diluted template DNA, 7.5 μL KAPA2G Fast Multiplex Mix (Kapa Biosystems, Cape Town, South Africa), 1.5 μL primer mix (2μM), and 4.5 μL distilled H2O. Samples were amplified using the following PCR conditions: 3 min of denaturation at 95°C, 30 cycles of 15 sec of denaturation at 95°C, 30 sec at multiplex-specific annealing temperature (see Mairal et al., 2022), and 25 sec of elongation at 72°C, and a final extension for 10 min at 72°C. Each 96‐well PCR plate contained 93 samples plus two randomly selected technical replicates and one negative control (H2O). All technical replicates (n=19) were used to count mismatches for scored genotypes and thus to estimate repeatability at each locus. Gel capillary electrophoretic separation of amplified fragments was carried out at the Central Analytical Facility, Stellenbosch University (Stellenbosch, South Africa). All loci were scored using GeneMarker software v 2.6.4 (SoftGenetics LLC, State College, Pennsylvania, USA) using the LIZ 500 size standard. We applied a semi-automatic genotype scoring for each allele, followed by manual correction when needed, to reduce scoring errors (Dewoody et al., 2006). Allele scoring was combined across microsatellite loci to generate a multi-locus phenotype (i.e., allele variation with unknown dosage) for each individual. In agreement with previous observations of tetraploid species (Nannfeldt, 1937; Frenot et al., 1999), all loci displayed up to four alleles per individual.

Unbiased analyses of microsatellite data in polyploids are hindered by factors such as unknown allele dosage, genotyping errors, and unknown mode of inheritance (Meirmans et al., 2018). Although some approaches have been developed to address the first two factors (e.g., Bruvo et al., 2004; Dufresne et al., 2014), not knowing modes of inheritance remains a major challenge when calculating population genetic metrics from allele frequency data (Meirmans et al., 2018). These challenges stem from the fact that the observed allelic patterns in allopolyploids can vary, even among microsatellite loci within species, as the genetic distance between parental taxa strongly determines the level of chromosome pairing in the resulting polyploid (Chester et al., 2012). As a result, a given polyploid species may not fit any of the extremes traditionally defined as allo- and autopolyploids, which may bias population genetic inferences based on allele-frequency calculations (see reviews in Dufresne et al., 2014; Meirmans et al., 2018).

Previous work suggests that P. annua originated from a relatively recent hybridization event between two closely related species, P. infirma and P. supina (Nannfeldt, 1937; Tutin, 1952). As such, P. annua would fit Stebbins’ (1947) definition of segmental allopolyploidy (i.e., some degree of homology may exist between the chromosomes of parental subgenomes), and cytological and phylogenetic evidence appear to support this view (Hovin, 1958; Soreng et al., 2010). We might therefore expect P. annua to follow an intermediate model of disomic inheritance at certain loci, typically expressed as fixed heterozygosity in allelic profiles (e.g., García-Verdugo et al., 2013), and tetrasomic inheritance for loci in chromosomes showing meiotic pairing or occasional recombination between the parental subgenomes (Hovin, 1958). To improve the accuracy of our analyses, we tested this expectation by examining how inbreeding coefficients (FIS) varied across loci. Under a model of complete disomic inheritance, all loci in an allotetraploid are expected to show fixed heterozygosity, leading to negative FIS values, while deviations from this expectation would suggest tetrasomic inheritance (Meirmans & Van Tienderen, 2013). To test this, we calculated inbreeding coefficients at the ‘locus × population’ level by running the approach implemented in GenoDive v.3.0 (Meirmans, 2020) for testing deviations of allele frequencies from those expected under Hardy-Weinberg equilibrium conditions.

Unlike our previous study that focused on extensive sampling of P. annua in the Prince Edward archipelago (Mairal et al., 2022), we found that only three loci in the present dataset (i.e., Poa3, Poa8 and Poa288) had significantly negative FIS values (Table S3, Supplementary Material), whereas the remaining loci tended to have positive values, as would be expected for a predominantly selfing species (Mengistu et al., 2000). To segregate the allelic composition of isoloci (i.e., loci Poa3, Poa8 and Poa288), we used the function ‘allele.correlations’ implemented in the package POLYSAT v 1.6 (Clark & Jasieniuk, 2011) using R 3.2.5 (R Core Team, 2019). However, this approach failed to confidently assign alleles to isoloci, likely because of population genetic structure (see Results) or because allele sizes between parental subgenomes were shared (Clark & Schreier, 2017). Considering these findings, we constructed two datasets to accommodate the heterogeneity among our microsatellite loci in all subsequent analyses, one containing the allelic information generated for all nine loci (hereafter “full dataset") and another set containing only the data for the six loci not following a pattern of disomic inheritance (hereafter “partial dataset”).