Population structure of CFM in the Canadian Prairies
We found little overall geographic structuring related to either canola production or soil zone in the 16 populations included in this study, although nuclear SNP analyses recovered substantially more population structure than COI haplotype analysis (Fig. 1). Both PCA and Structure analyses using SNPs recovered only four markedly divergent populations (Swan River and Portage la Prairie in Manitoba, and Athabasca and Sangudo in Alberta), located near the edges of canola production in those regions (Fig. 1, Fig. 2). While this may be indicative of an edge effect (sensu Cook 1961), other populations were also sampled near the edges of canola production but were not genetically distinct. The Portage la Prairie population is a possible exception to this, as these individuals were sampled from a research farm (Canada-Manitoba Crop Diversification Centre) located in the city of Portage la Prairie, and as a result may have reduced opportunities for gene flow with other CFM populations located on more rural farmland.
The remaining 12 central populations formed a large genetic cluster spanning eastern Alberta, Saskatchewan, and western Manitoba. Within this central cluster, Structure analysis indicated two distinct sources of genetic ancestry that were not clearly related to sampling geography (Fig. 1, Fig. 2D), and which was particularly pronounced in the Fairy Glen and Preeceville populations. Pairwise FSTwas also low between these central populations (Table 1), and IBD analysis was non-significant (Fig. 3C) suggesting few geographic barriers to gene flow. This is consistent with the homogenous landscape throughout much of the Canadian Prairies, and the high level of contemporary canola inventory in the sampling region of this study (Fig. 1A).
COI haplotype diversity was relatively low overall, and the four divergent populations in the SNP-based analyses were not distinct forCOI . These results are consistent with contemporary, widespread gene flow facilitated by large-scale canola production in the Canadian Prairies. It is possible that differences in recovered population structure between SNPs and COI are due to temporal differences in habitat connectivity resulting from year-over-year changes in canola inventory, and/or changes in effective population sizes of CFM due to regional and temporal differences in insecticide use. The COIgene represents only a single haploid marker, and if our sampling coincided with a period of greater effective population size and connectivity, we may expect to have less population structure in one or a few markers compared to thousands of diploid nuclear SNPs (Dussexet al. 2016; Liu et al. 2019). Data for historical year-over-year canola inventory production numbers or insecticide spray records are unavailable over this broad geographic range, so we cannot test this hypothesis.