Variation in Filamentous Growth and Response to Quorum-Sensing Compounds in Environmental Isolates of Saccharomyces cerevisiae

In fungi, filamentous growth is a major developmental transition that occurs in response to environmental cues. In diploid Saccharomyces cerevisiae, it is known as pseudohyphal growth and presumed to be a foraging mechanism. Rather than unicellular growth, multicellular filaments composed of elongated, attached cells spread over and into surfaces. This morphogenetic switch can be induced through quorum sensing with the aromatic alcohols phenylethanol and tryptophol. Most research investigating pseudohyphal growth has been conducted in a single lab background, Σ1278b. To investigate the natural variation in this phenotype and its induction, we assayed the diverse 100-genomes collection of environmental isolates. Using computational image analysis, we quantified the production of pseudohyphae and observed a large amount of variation. Population origin was significantly associated with pseudohyphal growth, with the West African population having the most. Surprisingly, most strains showed little or no response to exogenous phenylethanol or tryptophol. We also investigated the amount of natural genetic variation in pseudohyphal growth using a mapping population derived from a highly-heterozygous clinical isolate that contained as much phenotypic variation as the environmental panel. A bulk-segregant analysis uncovered five major peaks with candidate loci that have been implicated in the Σ1278b background. Our results indicate that the filamentous growth response is a generalized, highly variable phenotype in natural populations, while response to quorum sensing molecules is surprisingly rare. These findings highlight the importance of coupling studies in tractable lab strains with natural isolates in order to understand the relevance and distribution of well-studied traits.

phenylethanol tryptophol pseudohyphal growth filamentous growth invasive growth The budding yeast, Saccharomyces cerevisiae, can respond to environmental cues with numerous morphological switches and developmental phenotypes that likely increase fitness in naturally occurring conditions (Zaman et al. 2008). One such phenotype, filamentous growth, is thought to be a foraging strategy in response to nutrient stress. It is characterized by elongated cell morphology, unipolar budding, incomplete separation of mother-daughter cells, and substrate invasion (Gimeno et al. 1992). In diploid cells it is primarily induced by nitrogen limitation and known as pseudohyphal growth (Figure 1), while a similar though distinct response is triggered by carbon source limitation in haploid cells and is known as haploid invasive growth Sprague 2000, Cullen andSprague 2012).
In a lab strain, Σ1278b, haploid and diploid filamentous growth were shown to occur in response to the aromatic alcohols phenylethanol and tryptophol (Chen and Fink 2006). Production of these compounds is dependent on cell density and regulated through positive feedback, suggesting they may function as auto-inducing, quorum-sensing (QS) molecules. The human commensal and opportunistic pathogen, Candida albicans, can also undergo a morphological switch to a form of filamentous growth in response to QS molecules, which may be related to its ability to be pathogenic (Hornby et al. 2001, Leberer et al. 2001, conserved (Lo et al. 1997, Cain et al. 2012; thus, filamentous growth may represent a general, social, yeast survival strategy in the natural environment (Wuster and Babu 2010).
In S. cerevisiae, filamentous growth is regulated by multiple evolutionarily conserved, pleiotropic signaling networks, including the glucose-sensing RAS/cAMP-PKA and SNF pathways, the nutrientsensing TOR pathway, and the filamentous growth MAPK pathway (reviewed in Granek et al. 2011, Cullen andSprague 2012). These signaling pathways converge to regulate the transcription of FLO11, which encodes a cell wall flocculin required for multiple S. cerevisiae developmental phenotypes, including filamentous growth (Lambrechts et al. 1996, Lo and Dranginis 1998, Pan and Heitman 1999, Rupp et al. 1999, Braus et al. 2003, Chen and Thorner 2010. In lab backgrounds, deletion (Jin et al. 2008, Ryan et al. 2012) and overexpression collections (Shively et al. 2013), as well as QTL mapping of genetic crosses (Song et al. 2014, Matsui et al. 2015, have identified hundreds of genes contributing to the phenotype. The extent of phenotypic and genetic variation in filamentous growth in natural populations is still under-explored, as studies of this phenotype are dominated by the lab strain Σ1278b. Previous work has shown that in comparison to natural isolates, lab strains are often genetically and phenotypically atypical (Warringer et al. 2011). Thus, incorporating environmental isolates into genetic research broadens the scope of our understanding, both in how genetic variation modulates traits and how phenotypic variation manifests in natural populations (Gasch et al. 2016). The present study makes use of the 100-genomes collection, a panel of yeasts from subpopulations around the world and from a broad diversity of ecological niches, including fermentation reactions, clinical patients, and soil, plant and insect samples (Strope, et al., 2015), to explore natural variation in filamentous growth and response to QS compounds.
Saccharomyces yeasts are presumed diploid in nature (Replansky et al. 2008); therefore, the focus of the present study is the diploid filamentous growth response, pseudohyphal growth (psh). Most studies of psh use agar invasion as a quantitative metric for the phenotype, while the appearance of "pseudohyphae" around a colony (Figure 1) is assessed qualitatively. Using these metrics, (Magwene et al. 2011a) found variation in a sample of environmental isolates, and Hope and Dunham (2014) found variation in the SGRP collection (Liti et al. 2009). Less is known about variation in psh response to QS molecules, and to our knowledge, no systematic surveys have been done.
Using a quantitative measure of the amount of pseudohyphae to estimate psh, which we call the "filamentous index", we found a large amount of variation in the 100-genomes collection. When the strains were classified by the broad ecological category from which they were isolated, fermentation isolates exhibited a slightly elevated filamentous index compared to other ecological niches; however, this result appeared driven by a single isolate. When strains were classified by their subpopulation membership (Figure 2), which took phylogenetic history into consideration, isolates from the West African subpopulation had an elevated filamentous index compared to other subpopulations. Surprisingly, we find that in most isolates, addition of either phenylethanol or tryptophol to the medium had a negligible effect on psh, with a few strain-specific exceptions.
As most genomic studies focus on Σ1278b, the present study also examined the amount of naturally occurring, segregating genetic variation for psh using a mapping population of segregants from YJM311 as a proxy. This strain is a highly-heterozygous clinical isolate (Granek et al. 2013) belonging to the "mosaic" subpopulation that contains genetic variation associated with each of the other major S. cerevisiae subpopulations ( Figure 2). As such, it represents an ideal representative genetic background to investigate. We find that this single background contains enough genetic variation to recapitulate the range of phenotypes found in the 100-genomes panel. Using a bulk-segregant analysis, we find 5 genomic regions with major peaks significantly associated with the traits. Numerous genes that have been shown to influence the trait in Σ1278b are located within the peaks, and could therefore plausibly harbor the causative alleles.
Overall, our results indicate that there is an extensive amount of phenotypic and genetic variation in a well-studied developmental phenotype in environmental isolates, and that the response to aromatic alcohols may be a more limited, strain-specific effect. The relevance of this phenotype in the natural environment remains unknown, as no single broad ecological category appeared to be strongly associated with the trait, while subpopulation membership did seem to be associated with psh ability. Our results highlight the importance of complementing studies in lab strains with numerous genetic backgrounds isolated from the environment.

MATERIALS AND METHODS
Strains 96 strains from the 100-genomes collection (Strope et al. 2015) were phenotyped for psh (Table S1); these diploid strains are derived from single spores from original environmental isolates. Three of the strains, wells H8, H9, and H10, are not S. cerevisiae and were not included the downstream analyses. YJM311, a homothallic, clinical isolate (Granek et al. 2013), was used to conduct a bulk segregant analysis (BSA). For a different study in our lab, the original diploid isolate was transformed to express a PGK1-mCherry-KanMX fusion (HMY7) and used to generate an F5 mapping population. This mapping population was used in the present study.

Media
Yeast were grown in liquid YPD (1% yeast extract, 2% peptone, and 2% dextrose). Psh was induced on 4X synthetic low-ammonium dextrose (SLAD; 0.68% yeast nitrogen base w/o amino acids or ammonium sulfate, 2% dextrose, 50 mmol ammonium sulfate, and 2% agar) (Chen and Fink Figure 1 Pseudohyphal growth. Images depict: a small colony with pseudohyphae surrounding it (A), a close-up of a colony perimeter (B), and an image of the same perimeter in a different focal plane showing the pseudohyphae growing into the agar (C). Scale bar represents 200 mm. To obtain images, strain YJM1439 was streaked on a SLAD plate and grown for 6 days.
Generation of an F5 Mapping Population HMY7 was cycled through 4 rounds of sporulation, digestion, mating and growth (described in FileS1). At the end of the last cycle, spores were plated to a density of $100 colonies per plate and 360 segregants were isolated and phenotyped. Each colony was presumed diploid due to self-mating.

Sequencing and Bulk Segregant Analysis
Segregants with the highest and lowest filamentous indices, as well as lowest variance among replicate measurements, were chosen for further analysis. After re-assaying to verify psh, 22 segregants were identified for each pool. Segregants were grown to saturation in YPD in a 96-well plate, then combined for total genomic DNA extraction with the MasterPure Yeast DNA Purification Kit. Bulk pools were sent to the University of Georgia Genomics and Bioinformatics core for KAPA library prep and paired-end 150bp sequencing on an Illumina MiSeq Micro platform for an average coverage of $55-fold. DNA from HMY7 was previously sequenced at the Duke Genome Sequencing Core on an Illumina HiSeq 2000 instrument with single-end 50bp reads to an average coverage of 110-fold.
Reads from the bulk pools were aligned to the HMY7 genome (Supplementary Text) using BWA (Li and Durbin 2009), and SNPs were called using Freebayes (Garrison and Marth 2012) with settings for a pooled population. SNPs were filtered for quality and coverage. Bulk pools were compared using the R-package QTLseqr (Mansfeld and Grumet 2018), which implements the smoothed-G statistics of (Magwene et al. 2011b).

YJM311 Subpopulation Membership
In order to assign YJM311 to an S. cerevisiae subpopulation, the fixed SNPs from its genome were included in the dataset from the 100genomes collection and analyzed using the program structure V 2.3.4 (Pritchard et al. 2000) following the specifications of Strope et al. (2015). Briefly, the large set of SNPs found across the complete strain panel was filtered and sampled to create four independent sets of $1,200 SNPs in low linkage disequilibrium and representing the distribution of minor allele frequencies (generously provided by D. Skelley). Once YJM311 was incorporated, each of the four data sets was run three times using the linkage model (Falush et al. 2003) with a burn-in of 200,000 iterations and 1,000,000 iterations of MCMC, and K = 6 groups. The results from the 12 independent runs were compared using CLUMMP V 1.1.2 (Jakobsson and Rosenberg 2007) and visualized using distruct V 1.1 (Rosenberg 2004).

Pseudohyphal Growth
For the 100-genomes strains, YPD cultures were grown to saturation ($24 hr) in a 96-well plate and $2 ml per well was transferred to OmniTrays (Nunc 264728) using a 96-pin multi-blot replicator (V&P Scientific no. VP408FP6). For a given assay, a single 96-well plate was pinned to four replicates of three different media types (SLAD, SLAD + PheOH, SLAD + TrpOH). OmniTrays were wrapped with parafilm to prevent drying and incubated at 30C for one week. After incubation, trays were scanned on an Epson Expression 11000 XL scanner, which produced RGB color images with 1200 dpi. For the F5 mapping population, the same procedure was implemented for the 360 segregrants, but only SLAD + PheOH medium was used and with only two replicates per assay. For both the 100-genomes panel and the mapping population, the entire assay was repeated three times.
Follow-up experiments required streaking freezer cultures onto YPD agar, then streaking isolated colonies onto SLAD agar (+ PheOH or TrpOH, when appropriate) and incubating at 30C for 5 days before imaging.

Image Analysis
The scanned images were processed using a custom Python script (referred to here as "Eclipse"; Supplementary Material) that utilized the skimage package (van der Walt et al. 2014) to read the color qualities of individual pixels. Eclipse discriminated between outercolony filamentous growth and the inner colony, and reported the ratio of the two, a metric inspired by Tronnolone et al. (2017).

Image Processing
It was necessary to identify the color thresholds that designated the colony ring exclusively as white, the filamentous growth as gray, and the background as a separate entity. The image of the entire OmniTray was used to establish the values that best separated the parts of the colony; these values were then used to process the 96 cropped images representing individual colonies. Cracks, smudges, light reflection, and localized contamination interfered with image processing. In these cases, the individual colony images were examined and cropped to exclude trouble spots or dropped from analysis. For YJM311, only segregants that were consistently high-psh and low-psh were of interest for pooling in the bulk segregant analysis. We therefore manually inspected all images and dropped measurements that did not appear to accurately reflect the level of filamentation in the image (assessed qualitatively). This was not done for the panel of environmental isolates as we did not want to introduce bias, and because variation in the measurements was of interest for the downstream analysis. Figure 2 Population structure of the 100-genomes panel supplemented with strain YJM311, as inferred by the program structure. Following the analysis of Strope et al. (2015), each vertical line represents an individual strain with its fractional ancestry of K = 6 subpopulations represented by colors: green (North American), orange (Malaysian), red (West African), purple (Sake), blue (European/wine), and gray (human associated). Strains were assigned membership based on a threshold of .60% ancestry in a subpopulation, except for mosaic strains which had less than 60% ancestry in any other subpopulation.

Statistics
The data from the 100-genomes panel was analyzed in JMP 11.2.0 using an ANOVA framework with three different models. First, no group identity was assigned to the strains. The following model was fitted to the data: Y = m + Treatment + Strain + Strain x Treatment + Assay + Assay x Treatment + Strain x Assay + Strain x Assay x Treatment. Strain and treatment were considered fixed effects, while assay was considered a random effect. Next, strains were assigned to an ecological niche, which was considered a fixed effect, and the following model was fitted to the data: Y = m + Niche + Treatment + Strain[Niche] + Assay + Niche x Treatment + Strain x Treatment[Niche] + Assay x Treatment + Assay x Niche + Strain x Assay[Niche] + Strain x Assay x Treatment [Niche]. Brackets denote nested effects. Finally, strains were assigned to a subpopulation and the data were fitted to a model similar to the previous one. The data from the YJM311 segregants were transformed into z-scores for each plate; these values were used to help identify the strains with highest and lowest filamentous index.

Data Availability
Environmental strains from the 100-genomes collection are available upon request, as well as from the authors of the original study (Strope et al. 2015).

RESULTS
In order to investigate natural variation in psh and its induction by the QS compounds PheOH and TrpOH, the first 96 strains of the 100-genomes panel were plated on nitrogen-limiting medium using a pinning tool. Figure 3 shows the typical structure of a psh colony formed from pinning. Such colonies contain a white "ring" around the inner part where the pinner left cells; this ring separates the gray filamentous growth from the rest of the colony. Our image analysis pipeline located the highest and lowest values of the white ring on both the vertical and horizontal axes, which established a major and minor axis for the ring, and thus mapped out its location as an ellipse. This ellipse was used to "eclipse" all pixels inside of it, demarcating the inner colony. These eclipsed pixels were separated from the non-eclipsed pixels and the ratio of the two was calculated. The ratio, or "filamentous index", represents a rough quantitative measure of the filamentous growth of the colony; sample values are shown in Figure 3.
Across 29 agar trays in three independent assays, 2516 colonies were scored for psh. In the complete data set, the mean filamentous index was 13.01 and the median was 12. These data were analyzed using three linear models. The first did not assign any group identity to the strains and was used to investigate the variation among strains. The second model assigned a broad ecological category to the strains and tested for an effect of niche. The third and final model assigned subpopulation membership to the strains and tested for an effect of this phylogenetic history.

Variation in Pseudohyphal Growth
Of the $2500 colonies that were imaged and scored, 895 were grown without the addition of quorum sensing compounds and represent the base level of psh for the strains; the mean filamentous index was 13.29. Overall, there was a wide range of variation in the panel (black data points and distribution in Figure 4A,B) with a maximum average filamentous index of 33.3 for YJM1439 (derived from NCYC110), a ginger beer strain from West Africa, and a minimum average filamentous index of 7.5 for YJM1433 (derived from Yllc17_E5), a wine strain from France.
In all three linear models fitted to the data, the strain effect was significant (Table 1). Individual strains with significant parameter estimates (both above and below the mean) are listed in Table 2; strains that were significant in all three models are bolded. While the filamentous index only provides an approximate measure of psh, the behavior of individual strains appears to have been captured well, as the random effects in the model that were associated with replicate assays contributed little of the variation (a total of $15% among all the random effects).
In the second linear model, strains were divided into four ecological categories based on where they were isolated: fermentation, clinical, plant, and lab environments. The lab category represents strains that have been propagated in the lab environment for many years and may no longer represent the characteristics of the niche from which they were isolated, and includes the model strain Σ1278b. Each of the niche categories contained a wide range of variation in psh ( Figure  S2). The effect of niche was on the margin of significance in the linear model (P = 0.056); a post hoc Tukey's Honestly Significant Difference test found fermentation to be higher than the other categories (mean filamentous index of 14.29 compared to 13.01, 12.83, and 12.52 for lab, clinical, and plant, respectively). However, if the strain with the most abundant pseudohyphae, YJM1439, is removed from the analysis, the niche effect is no longer significant (P = 0.185; fermentation mean = 13.50), suggesting the effect is tenuous.
In the third and final linear model, strains were assigned membership to a subpopulation (based on the structure analysis) (Figures 4A,B). Most of the strains fell in the European/wine and mosaic categories, with the Malaysian subpopulation represented by a single strain; therefore the results for this analysis should be interpreted with caution. The effect of population was significant in the model (P = 0.0003). A post hoc Tukey's Honestly Significant Difference test found the West African subpopulation to have a higher filamentous index than the other categories ( Figure 4A, last panel). The West African subpopulation contained YJM1439, the strain with the highest filamentous index. When this strain was removed, the West African subpopulation Figure 3 Image processing pipeline. First column: three sample colonies from an Omnitray, which was blotted with 1.58 mm pins; in order, strains YJM984, YJM1336, and YJM1341, derived from 96-112, a clinical strain, M28s2, a European wine strain, and NRRL Y-12637, a South African wine strain, respectively. Second column: original images processed to differentiate white ring, filamentous growth, and background. Third column: inner part of the colony separated and pseudohyphal pixels counted to generate the filamentous index listed on the right. remained significantly higher than the others (mean = 14.77 compared to 18.47 with YJM1439). Thus, for at least one subpopulation, membership may be an important predictor for psh.

Variation in Response to Quorum Sensing Compounds
Of the colonies that were imaged and scored, 670 were grown in medium supplemented with PheOH and 951 were grown in medium supplemented with TrpOH; these produced a mean filamentous index of 14.20 and 11.89 respectively. Surprisingly, there was no overall effect of the addition of QS compounds (P = 0.2473 in the model with no group identity, P = 0.2478 in the niche model, and P = 0.2857 in the population model). Two strains significantly increased psh in response to PheOH in all three models and a different two strains increased in response to TrpOH. However, two strains also significantly decreased psh in the presence of one or both of the compounds (Table 3). Even for the strains that appeared to respond significantly, the effect sizes were small (on the order of 2-3 in the filamentous index). Mostly, PheOH and TrpOH appeared to have little to no effect on psh.

Comparison to Streaked Colonies
In order to verify the results from the high throughput assay, a selection of 10 strains that appeared to respond significantly to the QS molecules were streaked on SLAD, SLAD + PheOH, and SLAD + TrpOH agar plates ( Figure S3). We qualitatively assessed whether there appeared to be more filamentation in the different treatments (Table 3) in a manner similar to the study that originally reported the effects of the QS molecules on Σ1278b (Chen and Fink 2006). We found that 4 of the 10 strains appeared to respond in the direction predicted, as best as could be detected from visual inspection, but all responses were subtle.
The colonies arising from the streaks on SLAD agar were also compared to the images of the pinned colonies on the SLAD OmniTrays in order to verify that the high throughput method was correctly assessing the overall status of psh ability. While more psh was induced via streaking than pinning, there was clear agreement between the methods: strains with strong psh in one method exhibited strong psh in the other, while non-psh strains did not produce filamentation in either method ( Figure S3). However, it is also clear that filamentous index is a rough measurement, as strains that had similar psh induction did not have precisely the same index values. This is likely because all colonies stemming from one OmniTray were analyzed with the same color thresholds. This approach was taken in order to avoid bias, but future work analyzing each colony with its own optimized thresholds could potentially make the filamentous index more accurate. As it is currently being implemented, it appears to be appropriate for assessing general relative psh ability in a large panel.

Natural Genetic Variation in Pseudohyphal Growth
In order to investigate the amount of natural segregating genetic variation for psh, an F5 mapping population of YJM311 was phenotyped, and high and low segregants were pooled for sequencing and analysis. Across 24 agar trays, 360 segregants produced 1823 colonies that were scored. The range of phenotypic variation within the mapping population was comparable to that of the 100-genomes collection of environmental strains ( Figure 4B,D): the segregants had an overall mean filamentous index of 13.03 with a median of 11, and the maximum and minimum average filamentous index values were 49.8 and 5, respectively. The pools of segregants used in the sequencing analysis had distinct phenotypic distributions with a high pool mean of 30.32, (standard deviation = 6.45), and a low pool mean of 9.13 (standard deviation = 1.96) ( Figure 4D, insert).

Bulk Segregant Analysis
The allele frequencies of the bulk pools were compared using a smoothed-G statistic in order to find chromosomal regions that contain variation associated with psh (Magwene et al. 2011b, Mansfeld et al. 2018. Different window sizes for the smoothing function generated a variable number of significant mapping peaks, with windows of 60KB, 40KB, and 20KB producing 4, 26, and 29 significant peaks, respectively, at a false discovery rate of 0.01 ( Figure 5). While a smaller window size is likely to be appropriate for an F5 mapping population, we highlighted candidate genes in the four peaks stemming from the 60KB window, as well as one more on chromosome 14 which just missed the cut-off, as these peaks likely represent major effect loci (Table 4). In all five peaks, there were numerous genes that have been shown to either increase or decrease pseudohyphal or invasive growth in the Σ1278b background.

DISCUSSION
Microbes can engage in a myriad of social phenotypes that provide fitness benefits to individuals and genetic lineages (West et al. 2006). The model yeast, S. cerevisiae, exhibits multiple social phenotypes in the lab, including filamentous growth and quorum sensing. The filamentous growth phenotype appears to be conserved among other Saccharomyces spp (Kayikci and Magwene 2018) and among medically relevant yeasts, including Candida albicans (Cutler 1991), other Candida spp (Silva et al. 2011), Asperigillus fumigatus (Mowat et al. 2009) and Trichosporon asahii (Di Bonaventura et al. 2006), with filamentation ranging from pseudohyphae to true hyphae. Thus, filamentous growth is likely an important fungal response to environmental cues. This phenotype can be induced via QS in both S. cerevisiae (Chen and Fink 2006) and C. albicans (Hornby et al. 2001, Chen et al. 2004, though the QS molecules are not shared. The goal of this study was to assess the extent of variation in psh and response to external QS molecules in a range of isolates of S. cerevisiae in order to understand how the traits vary in natural populations. As such, we were interested in a strong response that is robust to slight environmental fluctuations. Our experimental protocol differed from those used in other studies of psh: we did not use Noble agar (highly purified) or wash cells before plating, and we attempted to quantify the amount of pseudohyphae rather than agar invasion. We also pinned from culture, transferring thousands of cells, rather than streaking to generate colonies from a single cell; our observation is that psh is more extensive when a colony is generated from a single cell ( Figure S3). The relative consistency among methods, assays, and replicate plates suggests our results captured the phenotype well, and represents an estimate of the general filamentous response in these environmental strains.

Phenotypic Variation in Pseudohyphal Growth
There was a surprising amount of phenotypic variation in the 100genomes collection, with some strains exhibiting long, pronounced pseudohyphae and some strains having little to no pseudohyphal growth. The image analysis counted a small amount of the white ring of the colony; therefore, values below 10% represent no, or very little, psh. It is possible that if these low strains were assayed in a different manner (i.e., streaked on highly processed medium) more psh would be observed; however, our goal was to assay a general, robust psh response and these strains lacked one.
We hypothesized that clinical strains would exhibit a stronger phenotype due to the association of filamentous growth with biofilm formation and virulence in other yeasts (Fanning and Mitchell 2012). However, when the strains were divided into their ecological niche of origin, there did not appear to be a particular niche that had more psh than others. Filamentous growth is likely a more general response and the perceived association with virulence may simply be due to a bias in the organisms in which research is conducted. Another possibility, which is not mutually exclusive, is that S. cerevisiae is not adapted to specific ecological niches with regard to this phenotype. Rather, as was previously proposed by Goddard and Greig (2015), it is a "nomad" dispersed among many habitats due to its association with humans. It should be noted that due to the limited number of strains that we analyzed, we were only able to test very broad ecological categories. It is possible that categorizing strains using four general groups obscured the significance of more specific, specialized ecological niches. For example, the fermentation category, which included isolates derived from European wine, Asian sake, African ginger beer, and kefir, could mask important ecological differences associated with the production of these different beverages. Nonetheless, our analysis that divided strains by their population of origin suggested that certain subpopulations are associated with increased psh, n Table 1 Results of the 100 Genomes pseudohyphal and quorum sensing analyses. a-Degrees of freedom are estimates due to different numbers of samples in each category and to incomplete samples for some strains (i.e., images dropped from analysis) specifically, the West African subpopulation. This supports the idea that for this trait, there is a signal based on phylogeny rather than membership in a broad ecological group.

Phenotypic Variation in Response to QS Molecules
Based on research in Σ1278b, we hypothesized that some of the low-psh strains would be induced when PheOH and TrpOH were added exogenously. We also hypothesized that fermentation strains would be most sensitive to the QS compounds, as the molecules could disperse further in more viscous environments where QS could be beneficial in synchronizing populations. Furthermore, a recent study investigated the effect of compounds produced during aromatic amino acid metabolism on different wine yeast (González et al. 2018). Non-Saccharomyces yeast growth was negatively affected by the presence of TrpOH and PheOH, suggesting that these compounds could be particularly important in inter-species interactions in fermentation environments. Surprisingly, most strains in the 100-genomes collection did not respond to the addition of QS molecules to the medium. It is still possible that many of these strains use PheOH and TrpOH for communication, but that the response is too subtle to be detected in our assay (see below). However, if cells do indeed carefully regulate both the production of and the response to QS compounds, it is improbable that exogenous application would have so little detectable effect across a wide panel. At the very least, one would expect a slight change in the same direction in most strains, which is not what was observed. Instead, it is more likely that certain strains respond strongly to PheOH and TrpOH, but most simply do not.

Comparison to Σ1278b
The majority of research on filamentous growth and QS in S. cerevisiae has been done on strains derived from Σ1278b, which has proved an invaluable model for understanding the genetic basis of the trait and for generating a robust map of the genetic pathways controlling it (Cullen and Sprague 2012). Homologs of some of the genes implicated in the Σ1278b background have been shown to be important for filamentous growth in other yeast species (Lo et al. 1997, Cain et al. 2012. And in the present study, genes uncovered in Σ1278b potentially harbor causative allelic variation in the clinical isolate YJM311. Chen and Fink (2006) demonstrated changes in the amount of pseudohyphae produced when Σ1278b was exposed to dilute treatments of PheOH, TrpOH, and both in combination. This strain was included in our panel (well E4), and while it appeared to somewhat respond to one of the autoinducing chemicals (PheOH), our results were not as dramatic as theirs. This is likely because our phenotypic assay was not as sensitive: our analysis measures "fuzziness" around a large colony, so the difference between treatments has to be striking to be detected. When we streaked, rather than pinned Σ1278b, our results were similar to the previously published results ( Figure S4), but the amount of change induced is small compared to the range of variation found among environmental isolates. In our assay, we did find individual strains that significantly responded to both chemicals, as was expected. However, in the majority of strains, the results were not as anticipated, and in some cases, were actually the opposite of Σ1278b. It is possible that other strains in our panel could also have a subtle response to the QS compounds, but it is clear that in most strains, the molecules do not induce a dramatic phenotypic change. The difference in inducibility between Σ1278b and the majority of strains in the panel indicates a disparity in behavior between this popular model strain and environmental strains of Saccharomyces cerevisiae. Thus, when it comes to QS, the results from the model laboratory strain may not translate smoothly to the broader population of Saccharomyces yeasts and how they behave in the environment.
Our results beg the question of why some strains are able to respond to the QS molecules, while others are not. Given the complexity of the filamentous growth response, it is possible that genetic variants exist in all aspects of quorum sensing: some strains may unable to sense the QS molecules, some may sense them, but be unable to respond, and yet others may simply be unable to undergo filamentous growth under any circumstance. While our study did not investigate all of these possibilities, we did investigate natural genetic variation in psh.

Genetic Variation
The present study aimed to determine whether and how much natural allelic variation existed in psh by using in a heterozygous clinical isolate from the mosaic subpopulation as a proxy. The phenotypic variation in the mapping population recapitulated the variation in the environmental panel, and five major and many minor peaks were associated with the trait, suggesting an abundance of segregating variation for psh in the environment. Complex phenotypes can be strongly influenced by SNPs at non-synonymous, synonymous, and regulatory locations (She and Jarosz 2018); all these types of genetic variation were identified in the major mapping peaks of YJM311. We mostly highlighted candidate genes in the peaks that have been implicated in psh in the Σ1278b background, but it is not clear whether or not they contain the causative alleles. These loci influence numerous cellular processes such as cell wall biosynthesis, mitochondrial function (Kang and Jiang 2005), cell polarity (Song et al. 2014), progression through the cell cycle (Zhu et al. 2000), and ammonium uptake (Lorenz and Heitman 1998). While investigating the functional effect of various alleles was beyond the scope a-Strains that were significant in all three models are in bold. b-The parameter estimates are from a linear model and indicate the amount a given strain is above or below the estimate of the intercept. The intercepts for the models are as follows: no group membership-13.16, niche-13.17, population-12.59. In the models with either a niche or population classification, the strains were nested within their group. The estimate for a strain is therefore the combination of the intercept, the strain parameter, and the group parameter. c-A p value of less than 0.05 indicates a significant difference from 0.
of this study, we anticipate future studies could harness the power of this approach. It is worth noting that a recent study comparing psh in S. cerevisiae and S. bayanus found that the cyclic AMP-Protein Kinase A pathway plays an important regulatory role in both. However, the manner in which the genetic network regulates the phenotype has diverged: increasing levels of cAMP has the opposite effect on the induction of the phenotype in the two species (Kayikci and Magwene 2018). This suggests selection to maintain filamentous growth over a long time scale, but also the ability of the complex genetic network underlying the trait n a-Strains that were significant in all three models are in bold. b-Strains whose phenotypic response was investigated via streaking have a yes or no to indicate whether the predicted response was detectable; "-" indicates the strain was not streaked. c-A p value of less than 0.05 indicates a significant difference from 0. Figure 5 Genome-wide comparison of the allele frequencies in the high and low psh pools of YJM311 F5 segregants. The G-prime statistic was calculated with a sliding window size of 60,000 bp (A) and 20,000 bp (B). Red line represents the cut-off for significance at a false discovery rate of 0.01. Candidate loci are listed above major, significant peaks.
to adapt and change. Future work identifying the genetic basis of some of the phenotypic variation observed in this study could shed light on the components of the genetic network that currently harbor segregating allelic variation, and upon which selection could ultimately act.

ACKNOWLEDGMENTS
We thank Paul Magwene for strains, Daniel Skelley for genomic data files, Rachel Rambadt for computational assistance, and the University of Georgia Genomics and Bioinformatics core for sequencing the samples. The research was funded by National Institutes of Health grant R15 GM122032 to H.A.M. and a William & Mary Biology DeFontes Fellowship to B.A.L.