Population structure analyses
For STRUCTURE analysis, a thinned version of the VCF file was generated with vcftools 0.1.15 (–thin 500) (Danecek et al., 2011), containing 11,447 similarly-spaced SNPs, while including only L. cidristrains. Structure was run on this dataset five times for K values ranging from 1 to 7, with 10,000 burn-in and 100,000 replications for each run, using admixture model, infer alpha, lambda =1, fpriormean =1, unifprioralpha 1, and alpha max 10. The structure-selector website was employed to obtain the optimal K values (http://lmme.qdio.ac.cn/StructureSelector/) (Li and Liu, 2018) according to the Evanno method (Evanno et al., 2005). The results for each K were plotted using CLUMPAK (Kopelman et al., 2015) and visualized using a structure plot (http://omicsspeaks.com/strplot2/ ) (Ramasamy et al., 2014). In addition, we performed clustering analyses of the same samples by using SMARTPCA without outlier removal (Patterson et al., 2006). For fineSTRUCTURE analysis (Lawson et al., 2012), a VCF file that included all SNPs among L. cidri strains was phased using BEAGLE 3.0.4 (Browning and Browning, 2007). Since we lacked a L. cidrirecombination map, we used a constant recombination rate between consecutive SNPs based on the average recombination rate of L. kluyveri (0.4 cM kbp-1, (Brion et al., 2017)). All versus all chromosomal painting was performed with Chromopainter V2, and the output was further analyzed with fineSTRUCTURE (-x 100000 -y 100000 -z 1000). Plotting of the ancestry matrix was done using fineSTRUCTURE R scripts.