Experimental Evolution Reveals Favored Adaptive Routes to Cell Aggregation in Yeast

Yeast flocculation is a community-building cell aggregation trait that is an important mechanism of stress resistance and a useful phenotype for brewers; however, it is also a nuisance in many industrial processes, in clinical settings, and in the laboratory. Chemostat-based evolution experiments are impaired by inadvertent selection for aggregation, which we observe in 35% of populations. These populations provide a testing ground for understanding the breadth of genetic mechanisms Saccharomyces cerevisiae uses to flocculate, and which of those mechanisms provide the biggest adaptive advantages. In this study, we employed experimental evolution as a tool to ask whether one or many routes to flocculation are favored, and to engineer a strain with reduced flocculation potential. Using a combination of whole genome sequencing and bulk segregant analysis, we identified causal mutations in 23 independent clones that had evolved cell aggregation during hundreds of generations of chemostat growth. In 12 of those clones, we identified a transposable element insertion in the promoter region of known flocculation gene FLO1, and, in an additional five clones, we recovered loss-of-function mutations in transcriptional repressor TUP1, which regulates FLO1 and other related genes. Other causal mutations were found in genes that have not been previously connected to flocculation. Evolving a flo1 deletion strain revealed that this single deletion reduces flocculation occurrences to 3%, and demonstrated the efficacy of using experimental evolution as a tool to identify and eliminate the primary adaptive routes for undesirable traits.


HOPE et al SUPPLEMENTARY TABLES:
Tab 1: Clone_data_rep1. Three measurement replicates from ImageJ (columns B-H) are provided for the 60-minute settling image for each evolved clone, with a calculation of the coordinate of half of the maximum grey value (I) as in (Hope and Dunham 2014), and a final calculation of the settling ratio (J).
Tab 2: Clone_data_rep2. A second biological replicate for each original clone, with three measurement replicates per image.
Tab 3: Segregant_data. A single measurement and ratio is provided for multiple progeny from the backcross of each original evolved clone (numbers in column A) to the laboratory strain FY4. , and copy number variants (CNVs) (Tab 4) for all evolved clones. Tabs 1-3 have the following column labels: Sample: sample number (note "EV" samples are from second evolution experiment); Chrom: chromosome; Pos: chromosome location; Ref: reference allele; Alt: alternative allele; Qual: quality score; Info: descriptive mutation information generated by bcftools, retroseq, or lumpy; Format: format of genotype information; Genotype: genotype information generated by bcftools, retroseq, or lumpy; Mutation type: coding-nonsynonynous, coding-synonymous, intergenic, or 5'upstream; Gene: systematic gene name; AA: amino acid annotation; Gene Alias: common gene name. Tab 4 has the following additional column labels: Start: start location of copy number segment; End: end location of copy number segment; Copy Number: average copy number for segment, calculated using DNAcopy.
Tab 1: SNPs_indels. SNPs/indels present in 7 or more samples for the original 23 evolved clones or 4 or more samples for the 5 new evolved clones were filtered out to remove common false positives or ancestral mutations, using bcftools isec (Li and Durbin 2009; Faust and Hall 2014). Mutations were then filtered for quality (QUAL>50, DP>=10), and mutations annotated as telomeric, mitochondrial, LTR_retrotransposon, intergenic, or coding-synonymous were removed. All remaining mutations were visually confirmed in IGV. See bctools documentation for genotype and quality annotation information.
Tab 2: retroseq. Ty insertions were called using the program retroseq (Keane, Wong, and Adams 2013) with non-default paramaters (discover: -q 28 -id 85 -len 25 -align, call: -depth 400), annotated (97), and verified by visual inspection in IGV. Intergenic mutations were ignored, excepting FLO gene promoters, which were manually re-annotated as 5'-upstream mutations. Additionally, known Ty insertions in the FLO1 promoter were not called by retroseq in Samples 2, 4 and 8, though PCR and visual inspection indicate otherwise. Note that retroseq gives inexact breakpoints so insertion positions are approximate. See retroseq documentation for genotype and quality annotation information.
Tab 3: lumpy. SVs were called using the program lumpy (Layer et al. 2014) with default paramaters. SVs with at least 10 supporting reads were confirmed using visual inspection in IGV. See lumpy documentation for information on SV type and quality scores.
Tab 4: CNV. CNVs were called as described in Material and Methods using the R program DNAcopy (Seshan and Olshen 2015), with additional visual inspection to validate the findings. To remove noise that would otherwise cause unnecessary splits, a standard deviation correction of 2 was implemented. Chromosomes or chromosome segments that had a copy number different from 1 are listed.