Footprints of human migration in the population structure of wild baker’s yeast

Humans have a long history of fermenting food and beverages that led to domestication of the wine or baker’s yeast, Saccharomyces cerevisiae. Despite their tight companionship with humans, yeast species that are domesticated or pathogenic can also live on trees. Here we used over 300 genomes of S. cerevisiae from oaks and other trees to determine whether tree-associated populations are genetically distinct from domesticated lineages and estimate the timing of forest lineage divergence. We found populations on trees are highly structured within Europe, Japan, and North America. Approximate estimates of when forest lineages diverged out of Asia and into North America and Europe coincide with the end of the last ice age, the spread of agriculture, and the onset of fermentation by humans. It appears that migration from human-associated environments to trees is ongoing. Indeed, patterns of ancestry in the genomes of three recent migrants from the trees of North America to Europe could be explained by the human response to the Great French Wine Blight. Our results suggest that human-assisted migration affects forest populations, albeit rarely. Such migration events may even have shaped the global distribution of S. cerevisiae. Given the potential for lasting impacts due to yeast migration between human and natural environments, it seems important to understand the evolution of human commensals and pathogens in wild niches.

Methods
Data were collected as described in the Methods section of Peña et al on "Footprints of human migration in the population structure of wild baker’s yeast"

More specifically, whole-genome sequences for strains sampled from trees were compiled from publicly available data (N = 295; Table S1) (Almeida et al. 2015; Barbosa et al. 2016; Bergström et al. 2014; Duan et al. 2018; Fay et al. 2019; Gayevskiy et al. 2016; Han et al. 2021; Pontes et al. 2020; Skelly et al. 2013; Song et al. 2015; Strope et al. 2015; Yue et al. 2017). We defined S. cerevisiae tree-sampled strains as those isolated from tree bark, exudate and leaves from trees or litter, and we also included strains from any soil. New whole-genome sequence data was generated for strains from trees in Indiana and Kentucky (N = 7; Osburn et al. 2018), North Carolina (N = 9; Diezmann & Dietrich 2009), Europe (N = 3; Robinson et al. 2016) and for new S. cerevisiae strains from the bark of white oak (Quercus alba) and live oak (Q. virginiana) from Georgia, Florida, Pennsylvania, and North Carolina (N = 15; Bensasson lab).

Reads were mapped to the S. cerevisiae reference genome, S288c (SacCer_Apr2011/sacCer3 from UCSC), using Burrows-Wheeler Aligner (bwa-mem, version 0.7.17; Li & Durbin, 2009). We used SAMtools to sort, index, and compress bam files and generated a consensus sequence using the mpileup function with the -I option to exclude indels (version 1.6; Li et al. 2009). Next, we used the BCFtools call function with the -c option to generate a consensus sequence (version 1.9) (Li et al. 2009) and converted from vcf format to fastq format in SAMtools using the vcfutils.pl vcf2fq command. Lastly, base calls with a phred-scaled quality score of less than 40 were treated as missing data (calls were converted to “N”) using seqtk seq -q 40 in SAMtools. ### Files and variables