Directed evolution for cell separation in natural isolates of budding yeast reveals selection to deactivate AMN1 and the Rim101 pathway in haploids and selection in favor of Hawthorne's deletion in diploids

TL;DR

This study shows how natural yeast populations evolve to separate cells by deactivating specific genes and causing chromosome deletions.

Contribution

The paper reveals novel roles for the Rim101 pathway and CCH1 in cell separation and quantifies the rate of Hawthorne's deletion.

Findings

Haploid yeast populations show strong selection to deactivate AMN1 and the Rim101 pathway.

Diploid populations frequently undergo Hawthorne's deletion due to mating type locus fusions.

A nonsynonymous mutation in CCH1 was found in diploid populations without large deletions.

Abstract

Natural isolates of the yeast Saccharomyces cerevisiae were evolved under a transfer protocol that selected for cell separation and against clumpy growth. Whole-genome sequencing of haploid populations revealed strong selection to deactivate AMN1, a known regulator of postmitotic cell separation, as well as multiple instances of loss-of-function mutations to genes of the Rim101 pathway, pointing to a previously unknown role of the Rim101 pathway in regulating cell separation. In diploid populations, we observed repeated large partial deletions of chromosome III caused by fusions of the mating type loci MAT and HMR (Hawthorne's deletion) or MAT and HML (Strathern's circle). We measured the spontaneous rate of Hawthorne's deletion and found that it is within an order of magnitude of previously measured rates of whole-chromosome aneuploidy. A diploid population in which neither large deletion was detected instead fixed a heterozygous nonsynonymous mutation to the calcium channel CCH1, also pointing to a novel role for this gene in relation to cell separation.

Figures

a) In the evolution experiment, large cell clusters were allowed to settle out prior to transfer. b) The clumpy phenotype of the haploid ancestor compared to the separated phenotype of representative evolved clones. c) Evolved and knockout haploid strains have less sediment at the bottom of a culture after 2 h of settling. d) Evolved and knockout haploid strains have less downward progress of the cloudy area of culture after 2 h of settling. Curly braces indicate group comparisons. d and e share the same legend. e) Across clones isolated from 11 haploid populations, parallel changes at the gene (AMN1, STE11) or pathway (Rim101) level were observed.

a) Read depth in chromosome III showing MAT-HMR and MAT-HML deletions in clones isolated from evolved diploid populations. b) Schematic of the products of MAT-HMR and MAT-HML fusions. c) Clones isolated from diploid populations have reduced sedimentation compared to the ancestor. d) Dissection products of evolved diploid clones cured of their deletion have the ancestral sedimentation phenotype. e) Dissection products of the CCH1 heterozygote show that the reduced-sedimentation phenotype segregates with the F994L homozygote.

a) The empirical cumulative distribution of cell cluster size for evolved diploid genotypes. b) Selection scheme to enable estimation of the rate of Hawthorne's deletion: 1 copy of chrIII was doubly tagged with counter-selectable markers between MAT and HMR, while complete loss of chrIII was selected against by histidine auxotrophy. c) The spontaneous rate of MAT-HMR deletion of 1 copy of chrIII as estimated by the fluctuation assay.

Tables

Table 1.. Allele frequencies as revealed by whole-population sequencing of 2 d 57 haploid populations. Lineages with frequency <5% are not reported.

Haploid population H3		Haploid population H5a
Mutation	Freq	Mutation	Freq
AMN1 K315*	0.51	RIM20 N535fs	0.42
AMN1 R162*	0.13	AMN1 Q224*	0.34
AMN1 E476*	0.08
AMN1 L463F	0.07

Full text

Introduction

Yeasts are fungi that have re-evolved unicellularity as a derived trait from a multicellular fungal ancestor. Although they are unicellular, yeasts such as Saccharomyces cerevisiae are capable of phenotypic states in which individual cells are in sustained contact with one another. These various cell aggregation phenotypes include mating, flocculation, pseudohyphal growth, invasive growth, biofilms, and clumpy growth (Soares 2011). The latter phenomenon, clumpy growth (also called chain formation; Stewart 2009), occurs when growing cells do not fully separate after the mitotic process of budding, leading to clusters of related cells. The fundamental importance of the regulation of postmitotic separation to both pathogenicity (ie biofilms) and to organismal complexity and multicellularity has led to sustained interest in experimental selection for decreased postmitotic separation (Ratcliff et al. 2015; Hope et al. 2017; Tong et al. 2025).

Typical laboratory strains of S. cerevisiae are not clumpy. Much of the variation in clumpiness across strains is explained by different alleles of the cell separation regulator AMN1 (Li et al. 2013), which represses the cell separation program in haploids but not diploids (Fang et al. 2018). Laboratory strains such as S288C have a nonfunctional allele of AMN1. Since cultures of well-separated cells are easier to work with in many respects, this phenotype is advantageous to the experimenter and may have been semi-inadvertently selected for in the domestication history of common laboratory strains. In contrast to laboratory strains, natural isolates (“wild strains”) of S. cerevisiae are often clumpy growers, especially as haploids (Fig. 1b). Wild strains tend to be more robust than lab strains in many respects, including faster growth rate, higher sporulation efficiency, and lower rate of petite production (Dimitrov et al., 2009). These properties of wild strains are particularly interesting given the many industrial uses of S. cerevisiae. Wild strains are also interesting in light of the status of S. cerevisiae as an important model organism coupled with the well-known, but not necessarily well-characterized, importance of strain-specific background effects on genotype-phenotype mappings (eg Galardini et al. 2019). While some properties of wild strains render them more experimentally tractable than common laboratory strains, their clumpy growth presents technical difficulties for otherwise routine assays such as dilution plating to estimate population density and flow cytometry of single cells.

a) In the evolution experiment, large cell clusters were allowed to settle out prior to transfer. b) The clumpy phenotype of the haploid ancestor compared to the separated phenotype of representative evolved clones. c) Evolved and knockout haploid strains have less sediment at the bottom of a culture after 2 h of settling. d) Evolved and knockout haploid strains have less downward progress of the cloudy area of culture after 2 h of settling. Curly braces indicate group comparisons. d and e share the same legend. e) Across clones isolated from 11 haploid populations, parallel changes at the gene (AMN1, STE11) or pathway (Rim101) level were observed.

We were motivated to fill a gap in the literature by performing experimental selection for increased cell separation. Because they are clumpier to begin with, and because we were interested in creating less clumpy variants of wild strains, we carried out this experiment using wild strains of S. cerevisiae. In light of the role of AMN1 as a primary regulator of postmitotic separation in haploids, we founded populations with both haploid and diploid ancestors. We evolved replicate populations for 57 d (>380 generations) under a daily transfer protocol that selected against large clumps and for separated cells. We report here the genotypic and phenotypic evolution observed in this experiment.

Methods

Medium and growth conditions

The evolution experiment was carried out in glass 50 mL Erlenmeyer-type flasks containing 10 mL synthetic defined (SD) medium (6.7 g/L yeast nitrogen base, 2% glucose) supplemented with ampicillin and tetracycline, at 30 °C and shaken at 200 rpm. SD was chosen as the medium for the evolution experiment because cultures of the wild diploid strains used are already very well-separated in YPD after the exponential growth phase. For haploid clones isolated from the evolution experiment, some assays were carried out in YPD (2% peptone, 1% yeast extract, 2% glucose).

Strains

The diploid strain YPS602 is a monosporic isolate of YPS133, which was isolated from Tyler Arboretum, Pennsylvania, as described previously (Sniegowski et al. 2002). The closely related diploid strain YPS606 is a monosporic isolate of YPS142, also isolated from Tyler Arboretum. Both strains are related to the more widely known YPS128 and are members of the North American Oak clade (Peter et al. 2018). YPS2070 is a haploid MATα heterothallic ho::kanMX derivative of YPS602, and YPS2055 is a haploid MATa heterothallic ho::kanMX derivative of YPS606.

Evolution experiment

Four replicate populations of each of the 4 strains described above were founded from single colonies and transferred daily according to the following protocol (illustrated schematically in Fig. 1a). A 2 mL aliquot of well-mixed culture was removed to a small glass tube and allowed to settle undisturbed for 10 min (day 1 to day 8), 20 min (day 9 to day 14), or 30 min (day 15 to day 57). After the waiting period, 100 μL of culture was carefully removed from near the top of the aliquot and transferred to 9.9 mL fresh medium in a new flask.

Transfers were performed every ∼24 h for 57 d (implying at least 57 × log_2_100 = 380 generations). At regular intervals, 3 mL of culture was frozen in 15% glycerol and stored at −80 °C. Clonal isolates were obtained from frozen population samples by streaking to single colonies. For a control (no selection against clumping) evolution experiment, conditions were identical except that 50 μL of well-mixed culture was added to 9.95 mL fresh medium at each transfer, for 66 d.

Populations analyzed

The 16 populations founded from the haploids YPS2070 and YPS2055 and the diploids YPS602 and YPS606 were designated H1-H4, H5-H8, D1-D4, and D5-D8, respectively. After completion of the evolution experiment, it became apparent that at some point between days 29 and 35, the diploid populations D5 to D8 were inadvertently discontinued, while haploid populations H5 to H8 were duplicated. We subsequently adopted an a/b naming scheme to distinguish these duplicated populations. Population H4 was lost before the end of the experiment due to a separate technical error. Therefore, after 57 d of evolution, there were 11 haploid populations (H1, H2, H3, H5a, H5b, H6a, H6b, H7a, H7b, H8a, H8b) and 4 diploid populations (D1, D2, D3, D4) available for analysis, and additionally the diploid populations D5 to D8 completed 29 d of evolution.

Settling and sedimentation assays

When a well-mixed culture of yeast is placed in a clear vessel and allowed to settle, 2 phenomena are apparent: (i) sediment collects at the bottom of the vessel and (ii) the cloudy area of culture suspension moves visibly downwards, with clear medium above. We measured both manifestations of settling (for illustrative photographs, see Supplementary Fig. 2 in Supplementary File 1). Haploid assays were performed in YPD while diploid assays were performed in SD. For all assays, clones were started from a frozen chunk and grown overnight in YPD and then grown for two 24 h cycles in SD or one 24 h cycle in YPD, diluting 1/100 at each transfer, prior to the assay. In the sedimentation assays, 5 mL of well-mixed culture was placed into a 15 mL conical tube suspended vertically; after 2 h, the height of the sediment at the bottom was measured to the nearest tenth of a millimeter using a digital caliper tool. This height was subsequently converted to a volume. For haploids, the settling distance was measured concurrently with the sediment measurement, from the same 5 mL aliquot, estimated as the distance from the meniscus to the midpoint between the clear media and the cloudiest culture. For diploids, the settling distance was measured by a slightly different procedure: 900 μL of culture was briefly spun down (90 s at 10,000 rcf) and resuspended in 900 μL sterile water. Then 700 μL of this resuspension was pipetted into a narrow-profile cuvette and allowed to settle. We used cuvettes because we found that the flat surface aided in accurate measurement, and the resuspension step was necessary because cultures in SD do not settle evenly. After 2 h, the downward progression of the visible demarcation from clear to cloudy was measured to the nearest tenth of a millimeter using a digital caliper tool.

DNA extraction and whole-genome sequencing

We used the genome assembly of YPS128 as provided by Yue et al. (2017) as the reference for all genomic analyses described here. For clones, a culture was started from a frozen chunk, while for frozen populations, the entire frozen sample was thawed, and 10 μL was used to inoculate a new culture. After growth overnight in YPD, DNA was extracted from 1 mL of culture using kit D7005 from Zymo Research. Short-read whole-genome sequencing was performed on an Illumina NovaSeq X Plus sequencer in one or more multiplexed shared-flow-cell runs, producing 2 × 151 bp paired-end reads. Reads were mapped to the reference genome using bwa mem (Li 2013), and duplicate reads were removed by PicardTools MarkDuplicates. Mutations were called using freebayes (Garrison and Marth 2012) on the default settings, with all potential variants reviewed manually and examined in IGV (Robinson et al. 2011). Mutation calls from this pipeline were also spot-checked using breseq (Deatherage and Barrick 2014). A spreadsheet of all called mutations is available in Supplementary File 2. For the allele frequencies shown in Table 1, breseq was used in polymorphism mode, with a minimum allowed frequency of 5%. To locate chromosomal deletions, plots of read depth were created from the output of samtools -depth (Danecek et al. 2021). Additionally, long-read Nanopore sequencing was used to confirm the MAT-HML fusion in the clone isolated from population D2. To do so, we used blastn (Camacho et al. 2009) to map segments of genes bordering HML and MAT to all Nanopore reads ≥10 kb that mapped to chrIII, from which we detected single reads spanning both regions as evidence of fusion.

Table 1.: Allele frequencies as revealed by whole-population sequencing of 2 d 57 haploid populations. Lineages with frequency <5% are not reported.

Allele replacements

We replaced AMN1 alleles in evolved clones with the ancestral AMN1 sequence. To do so, we inserted the natMX cassette just downstream of AMN1 in the ancestral strain. We then amplified the entire AMN1-natMX cassette and transformed it into evolved strains. Successful transformation was confirmed by Sanger sequencing to ensure that the ancestral allele was introduced. Transformations were carried out by standard lithium acetate methods (Gietz and Schiestl 2007).

Cluster size analysis

While microscopic inspection of evolved haploid strains showed a clear and obvious cell separation phenotype (Fig. 1b), for diploids the phenotype was more subtle. To estimate the distribution of cluster sizes in diploid strains, we imaged and labeled growing cultures. A frozen chunk of the strain of interest was used to inoculate an overnight culture in YPD; the following day, this culture was diluted 1/100 into fresh SD and allowed to grow for 24 h; then another 1/100 dilution and growth cycle in SD was performed; and finally on the 4th day, 200 μL of culture was transferred to 12 mL fresh SD and grown until A600 = 0.6, at which point photographs of the culture were taken at 100× magnification. Subsequently, the number of cells in each cluster was manually labeled (example shown in Supplementary Fig. 4 in Supplementary File 1) and recorded for further analysis.

Sporulation and tetrad dissection

Sporulation of diploid clones was induced by transferring an aliquot of saturated culture to 1% potassium acetate for incubation at 30 °C for 3 d. Subsequently, 10 μL of sporulated culture was then added to 50 μL of a solution consisting of 0.5 mg/mL zymolyase (US Biological Z1005), 1 M sorbitol, and 0.1 M potassium phosphate buffer (pH 7.5) and incubated at 30 °C for 10 min, after which 800 μL of sterile water was added to stop the digestion. A portion of the resulting solution was evenly spread onto an area of an YPD agar plate, and a light microscope equipped with a micromanipulator actuating a glass needle (Singer Instruments) was used to separate the tetrad.

PCR verification of MAT-HMR and MAT-HML fusions

Primers 5′-TGGAAAGCGTAAACACGGAG-3′ and 5′-TCGGATTTGCGCTTGACAAT-3′ were used to verify the MAT-HMR deletion; these primers produce an amplicon of ∼3.5 kbp only in the case of a MAT-HMR fusion. Similarly, primers 5′-GTCCAGGGGCGGTTTATTTT-3′ and 5′-GGACTTGGAAGAAGCGTTGG-3′ were used to verify the MAT-HML fusion. Both primer sets were validated on known deletion strains as determined by whole-genome sequencing.

Estimation of the spontaneous rate of Hawthorne's deletion

We constructed the following heterothallic haploid strains in the YPS602 background. YPS4277 has the genotype MATalpha ho::hphMX ura3::natMX his4::kanMX THR4. YPS4263 has the genotype MATa  ho::hphMX ura3::natMX HIS4  thr4::URA3 amdSYM with the amdSYM marker (Solis-Escalante et al. 2013) inserted on chrIII between SED4 and ATG15. These strains were constructed by standard lithium acetate transformation methods. We then mated these strains to create the diploid YPS4283, which is sensitive to both 5-fluoroorotic acid (5-FOA) and fluoroacetamide (FAA). The markers responsible for the sensitivity were located on chrIII between MAT and HMR, so that a MAT-HMR deletion created a double 5-FOA/FAA resistant. Complete loss of the marked chrIII was selected against by the absence of histidine in the growth media.

To estimate the rate of MAT-HMR fusion, we conducted a fluctuation assay (Gerrish 2008) with 4 replicate cultures. These cultures were inoculated from single colonies of YPS4283 into 6 mL SD + 20 mg/L uracil and grown overnight at 30 °C with shaking, after which the density was estimated by hemocytometer and a target of 1,000 cells was transferred to 10 mL fresh SD + uracil. These subcultures were grown for 3 d at 30 °C with shaking, and then 40 μL (1/250th of the culture) was plated neat to SD + uracil + 1 g/L 5-FOA + 2.3 g/L FAA agar plates, and the culture density was estimated by hemocytometer counts. Plates were incubated for 2 d at 30 °C and then examined. Very small colonies were excluded from subsequent analysis. The resulting colonies, being 5-FOA^R^ FAA^R^ His+, were considered likely to have the MAT-HMR fusion on the marked chrIII (nonsporulating colonies were likely MAT-HML fusions, but we did not measure this rate). To confirm MAT-HMR fusions, we used a 2:2 live:dead segregation pattern of tetrads as a first check and a PCR test as a second check. From each plate, 32 to 38 colonies were picked to both 1 mL YPD and to 1 mL SPO + uracil, in glass tubes. The YPD cultures were grown overnight and then stored at 4 °C until needed. The SPO + uracil cultures were grown at 30 °C for 3 d and then examined via microscope. From any culture in which sporulation was observed, at least 4 tetrads were dissected. If the dissection yielded a consistent pattern of 2:2 live:dead, DNA was extracted from the corresponding YPD culture, and the presence of a MAT-HMR deletion was verified via PCR using the primers described above. Only 1 colony with a 2:2 live:dead segregation pattern did not give a positive PCR result (Supplementary Table 2 in Supplementary File 1). The total number of verified MAT-HMR deletions for each replicate was input into the package rSalvador (Zheng 2017), using the Lea–Coulson model with ε = 1/250, to generate a point estimate and confidence interval for the spontaneous rate of MAT-HMR deletion.

Protein alignment

The canonical amino acid sequence of human CACNA1A was downloaded from UniProt and aligned to S. cerevisiae CCH1 using both Clustal Omega (Sievers et al. 2011) and TM-Coffee (Floden et al. 2016), a specialized aligner for transmembrane proteins. Both aligners agreed on the local alignment around the site of interest, yeast F994.

Results

All evolved haploid clones had a mutation to either AMN1 or the Rim101 pathway

We isolated 1 clone from each final haploid population, except in the case of population H3, from which we isolated 2 clones with different AMN1 mutations. All 12 clones isolated from haploid day 57 populations were markedly less clumpy when examined microscopically (examples shown in Fig. 1b) and displayed changes of large and statistically significant magnitude in 2 assays measuring settling (Figs. 1c and 1d). Whole-genome sequencing revealed that all clones had either a mutation to AMN1, a known regulator of cell separation, or to a gene in the Rim101 pathway: RIM8, RIM13, RIM20, or RIM101. There were no AMN1/Rim101 pathway double mutants.

All observed AMN1 and Rim101 pathway mutations appeared to be loss-of-function

In total, we observed 10 independent AMN1 mutations (Fig. 1e and Table 1). Of these, 7 are nonsense or frameshift while 3 are nonsynonymous amino acid substitutions. All 3 substitutions to AMN1 occur between residues 452 and 472. Similarly, of the 4 mutations observed in the Rim101 pathway, 3 are frameshift or nonsense mutations, and 1 is a substitution. Our supposition was that all AMN1 and Rim101 pathway mutations, including the substitutions, were loss of function. Engineered Δamn1 and Δrim20 knockouts displayed similar phenotypes to the corresponding evolved clones (Figs. 1c and 1d). We also investigated 2 AMN1 substitutions more closely: in the respective evolved clones, we replaced AMN1^C452Y^ and AMN1^L472P^ with the ancestral allele. These replacements completely restored the clumpy phenotype (Supplementary Fig. 1 in Supplementary File 1), demonstrating that the substitutions alone were sufficient to cause the separated phenotype.

The phenotype of AMN1 and Rim101 pathway mutants differs

While both AMN1 and Rim101 pathway mutants are markedly less clumpy than the ancestor, settling assays reveal differences between the 2 genotypes. We allowed cultures to rest undisturbed for 2 h and measured 2 different aspects of the process of cells settling out of solution. Comparing group means, Rim101 pathway mutants accumulate more sediment at the bottom of the culture than do AMN1 mutants (168.3 mm^3^ vs 114.9 mm^3^, P < 1e-04; Fig. 1c). Conversely, in AMN1 mutants, more downwards progress of the visibly cloudy area of culture (5.6 mm vs 4.7 mm, P < 1e-06; Fig. 1d) is observed than in Rim101 pathway mutants.

Haploid populations were diverse and experienced soft sweeps

Pairs of clones isolated from populations that were originally a single population tended to have different AMN1 or Rim101 pathway mutations (Fig. 1e noting a and b labels). This suggested that populations were diverse, with many contending lineages. To confirm this, we performed whole-population sequencing of haploid populations H3 and H5a. The results (Table 1) confirmed our hypothesis: we detected at least 6 independent AMN1 or Rim101 pathway mutant lineages across the 2 populations. We suspect there are further low-frequency lineages that we were unable to resolve due to limitations of read depth.

Parallel loss-of-function mutations to STE11 occurred in haploids

In the sequenced haploid clones, we observed 3 independent mutations to STE11, each of which all appeared in conjunction with an AMN1 or Rim101 pathway mutation. We confirmed that these clones are sterile (do not mate), as expected for loss-of-function mutations. Because populations H7a and H7b were once a single population and share the mutation AMN1^D373*^, we can infer with relative confidence that the STE11^C430*^ mutation arose in a background that already had fixed AMN1^D373*^. We surmise, then, that all observed STE11 mutations were second mutations. Settling and sedimentation assays did not detect an effect of knocking out STE11; however, we did notice a subtle visual difference in settling appearance. It remains uncertain whether ste11 mutations rose in frequency in response to the selection for cell separation or whether they had a more general effect on fitness in the experiment.

Large deletions involving the mating loci evolved multiple times in diploid populations

In contrast to haploid populations, in which all mutations affecting the cell separation phenotype that we detected were single-base-pair point mutations, in diploid populations we found that a large deletion on chromosome III evolved repeatedly (Fig. 2a). This ∼90 kb deletion (about 30% of chrIII) spans the region from MAT to HMR or, in one case, the region from MAT to HML (∼125 kb, or about 40% of chrIII). Clones with this deletion segregated 2 live:2 dead spores upon dissection of tetrads, consistent with the presence of multiple essential genes present on the deleted segment.

a) Read depth in chromosome III showing MAT-HMR and MAT-HML deletions in clones isolated from evolved diploid populations. b) Schematic of the products of MAT-HMR and MAT-HML fusions. c) Clones isolated from diploid populations have reduced sedimentation compared to the ancestor. d) Dissection products of evolved diploid clones cured of their deletion have the ancestral sedimentation phenotype. e) Dissection products of the CCH1 heterozygote show that the reduced-sedimentation phenotype segregates with the F994L homozygote.

Whole-population sequencing of populations D1, D2, and D4 at day 57 revealed that the MAT-HMR or MAT-HML deletion was fixed or nearly fixed. We also found evidence for the presence of deletion lineages in 3 of 4 day 29 populations D5, D6, and D8 (Supplementary Fig. 5 in Supplementary File 1), so that in total, a MAT-HMR or MAT-HML deletion of observable frequency occurred in 6 of 8 diploid populations evolved under selection for increased cell separation.

The MAT-HMR deletion is historically known as Hawthorne's deletion, and the MAT-HML deletion is known as Strathern's circle (Hawthorne 1963; Strathern et al. 1979; Herskowitz 1988). The latter deletion (or equivalently, fusion) is expected to produce a ring chromosome III (Fig. 2b). The arrangement of 1 linear and 1 circular chrIII is not expected to be deleterious under mitotic growth, but is expected to depress spore viability because crossovers during meiosis between the ring and the linear homolog generate a dicentric chromosome (Haber et al. 1980; Haber et al. 1984). While we did not observe any reduction in spore viability below 50% (Supplementary Table 1 in Supplementary File 1), the following evidence points to the existence of a true ring chromosome III: (i) a PCR product spanning MAT and HML is produced; (ii) we performed long-read sequencing and found approximately the expected number of long reads mapped to regions left of MAT and right of HML (Supplementary Fig. 3 in Supplementary File 1); and (iii) neither short- nor long-read sequencing suggested a new breakage site.

The association between large deletions and an increased cell separation phenotype

Evolved clones with either the MAT-HMR or MAT-HML deletion exhibited significantly less sedimentation than the ancestor (Fig. 2c), and an analysis of the distribution of cell cluster size, as quantified by microscopy, showed that these evolved strains were less clumpy than the ancestor (Fig. 3a). For example, the ancestral proportion of cells occurring in clusters of 4 cells or fewer was 43%; for the Hawthorne's deletion strains, this estimate ranged from 65% to 77%.

a) The empirical cumulative distribution of cell cluster size for evolved diploid genotypes. b) Selection scheme to enable estimation of the rate of Hawthorne's deletion: 1 copy of chrIII was doubly tagged with counter-selectable markers between MAT and HMR, while complete loss of chrIII was selected against by histidine auxotrophy. c) The spontaneous rate of MAT-HMR deletion of 1 copy of chrIII as estimated by the fluctuation assay.

We also quantified the budding pattern of the evolved clones. Diploid yeast tends to bud from alternating poles (Chant and Pringle 1995). We observed growing cells microscopically and scored whether 2 consecutive buddings occurred on alternate poles and found a significant depression in the frequency of alternate-pole budding for clones with the MAT-HMR or MAT-HML deletion isolated from the evolved populations, compared to the ancestor (Supplementary Table 3 in Supplementary File 1). This more disordered budding may be connected to the increase in cell separation.

Besides the parallelism observed in the evolution experiment, 3 additional lines of evidence support a causal relationship between the MAT-HML/HMR fusions and an increased cell separation phenotype. First, seeking to rule out that these large deletions were unconditionally beneficial in the experimental environment, we carried out a control evolution experiment under the same conditions as the main experiment, except that there was no selection against settling. In these populations, we observed no sign of Hawthorne's deletion (Supplementary Fig. 5 in Supplementary File 1). Second, we cured evolved clones of their deletions. The live products of dissected tetrads from the MAT-HML or MAT-HMR deletion clones, being homothallic, yielded diploid colonies cured of the deletion. These cured diploids no longer showed the reduced-sedimentation phenotype displayed in the strains with the deletion (Fig. 2d). Third, as described below, we measured the rate of the MAT-HMR fusion in the diploid ancestor. Spontaneous clones isolated from this experiment, possessing the MAT-HMR fusion, produced significantly less sediment during settling than the ancestor (Supplementary Fig. 7 in Supplementary File 1), akin to the evolved clones.

The rate of Hawthorne's deletion

The time to fixation of a mutation under positive selection depends on both the strength of selection and the rate of mutation. For rare mutations, there may be a substantial waiting time, and even when the mutation occurs, selection does not act deterministically until the lineage is of a certain size. The observed parallelism in this experiment suggested that the per-generation rate of Hawthorne's deletion could not be too small. To confirm this hypothesis, we estimated the spontaneous rate of Hawthorne's deletion via the classic fluctuation assay. We used genetic modifications (Fig. 3c) to facilitate selection for MAT-HMR deletions while counter-selecting against the complete loss of chrIII. We found a spontaneous rate of loss of MAT-HMR of 2.9e-07 (95% CI: 4.1e-08 to 3.4e-07). This rate represents the generation of MAT-HMR fusions from a single marked copy of chrIII, so the total rate of MAT-HMR fusions is twice as large, and the total rate of any Hawthorne's deletion including MAT-HML fusions is, by implication, up to 4 times the measured rate, although we did not measure the rate of MAT-HML fusion.

A dominant point mutation to CCH1 also produces less clumpy growth in diploids

Diploid population D3 did not have a high-frequency Hawthorne's deletion. The representative clone that we isolated had only 1 called mutation, a nonsynonymous heterozygous mutation (F994L) to the calcium channel gene CCH1, which is located on chrVII. Population sequencing revealed this mutation was fixed or at high frequency (Supplementary Fig. 6 in Supplementary File 1). The representative isolated clone sedimented significantly less than the ancestor (Fig. 2c) and was quantitatively less clumpy (Fig. 3a). We sporulated and dissected tetrads of this isolate to produce F994L/F994L and F994F/F994F homozygotes, the genotypes of which were confirmed by Sanger sequencing. The F994F/F994F homozygotes did not show a reduced-sedimentation phenotype, while the F994L/F994L homozygotes displayed a sedimentation phenotype very similar to the Hawthorne's deletion strains (Fig. 3e). The sedimentation phenotype of the F994L/F994L homozygote is not more extreme than that of the F994L/F994F heterozygote, implying that this mutation is dominant with respect to cell separation. We noticed when carrying out dissections that F994L colonies were smaller than F994F colonies, suggesting that under this experiment's particular selection regime, the heterozygote has the highest total fitness considering both growth rate and selection for cell separation.

CCH1

• is homologous to human CACNA1A (Paidhungat and Garrett 1997) and encodes the voltage-gated calcium channel subunit α1A. Variants of CACNA1A are implicated in several neurological diseases (Rajakulendran et al. 2012). Alignments of yeast CCH1 and human CACNA1A show that yeast F994 is located on a transmembrane helix and homologous to human F708. A search of ClinVar (Landrum et al. 2014) revealed 1 observation (accession: RCV000622937.4) of human F708L, reported in association with developmental and epileptic encephalopathy, a severe epilepsy syndrome.

Discussion

Selection on AMN1 in haploids was predictable

Amn1 causes clumpy growth in haploids by mediating the degradation of Ace2, the major transcription factor of the postmitotic cell separation program (Fang et al. 2018). It is likely that semi-inadvertent selection for AMN1 mutants is part of the evolutionary history of common lab strains, since cultures of well-separated cells are more experimentally tractable for various purposes. This experiment has in effect recapitulated that selection; thus, it is not surprising that we recovered many AMN1 mutants, although we did not observe the particular AMN1 allele, D368V, that is present in S288C and related strains. It is a demonstration of the usefulness of experimental evolution in genotype-to-phenotype mapping, if a selection gradient relevant to the phenotype of interest can be devised, that had AMN1's role in governing cell separation in haploids not been previously known this experiment would have clearly pointed toward it.

The Rim101 pathway is a novel regulator of cell separation during vegetative growth

Mutations to the Rim101 pathway were less expected. This pathway is known to have a role in pH sensing (Yan et al. 2020) and regulation of cell size (Shimasawa et al. 2023) and to connect to pathways regulating filamentous growth (Cullen and Sprague 2012). Arras et al. (2022) found that Rim101 pathway genes are required for sexual aggregation, but a role in the regulation of cell separation during typical vegetative growth conditions has not previously been reported. We suspect that the clear effect of Rim101 pathway genes on cell separation reported here has previously escaped notice because many laboratory strains such as S288C have a nonfunctional allele of AMN1 and are therefore already well-separated, which masks the increased cell separation resulting from disabling a Rim101 pathway gene in a background with functional AMN1.

The cell wall of yeasts requires specialized structures to effectuate cell division. At the mother–bud neck, septins provide a diffusion barrier and scaffold to help organize septum-building enzymes. The primary septum, made of chitin, acts as a temporary wall between the daughter and mother cell that is subsequently degraded from the daughter cell by the chitinase Cts1 (Weiss 2012). Ace2, which as noted above is degraded via Amn1, activates transcription of CTS1. Interestingly, Δrim101 and Δrim13 mutants have been shown to have increased transcription of CTS1, although there does not appear to be direct repression of CTS1 by Rim101, and Δrim101 mutants have an asymmetric distribution of the septin Cdc3 and an abnormal localization pattern of Chs4, a recruiter of chitin synthase III (Lamb and Mitchell 2003). These data suggest 2 possible nonexclusive mechanisms through which loss-of-function Rim101 pathway mutants might cause increased cell separation: increased production of chitinase and production of an altered septum more vulnerable to degradation. Rim101 pathway mutants and AMN1 mutants do not have identical phenotypes in our assays: Rim101 pathway mutants produce more sediment (Fig. 1c), but have a smaller settling distance (Fig. 1d), compared to AMN1 mutants. The latter finding is consistent with the notion that Rim101 pathway knockouts have decreased cell size (Shimasawa et al. 2023), given that smaller cells are expected to settle more slowly.

Interestingly, whole-population sequencing revealed that at least some populations maintained competing AMN1 and Rim101 pathway lineages even after 57 d of evolution (Table 1). It is possible that certain uncontrolled aspects of our experimental design—eg the precise depth of the pipet tip used to transfer part of the culture after the settling period—influenced whether AMN1 or Rim101 pathway mutants predominated in any particular haploid population. Strain and mating type differences could also have played a role, although we do not have the statistical power to assign any such effects.

Implications of the rate of large deletions involving the mating type loci

In diploids, we observed repeated selection for MAT-HMR or MAT-HML deletions on chromosome III. These heterozygous large deletions (roughly 30% to 40% of chrIII) arise as a consequence of homology shared by the 3 mating type loci. They are historically known as Hawthorne's deletion and Strathern's circle, respectively, and were instrumental in the working out of the mating type switching system (Hawthorne 1963; Strathern et al. 1979; Haber et al. 1980; Herskowitz 1988; Haber 1998). Hawthorne's deletions were initially observed as a small proportion of rare mating type switching events in heterothallic (ho) strains. In this respect, it is interesting that we have observed these deletions in homothallic (HO) wild-type diploids under vegetative growth, that is, separately from any obvious context of mating type switching per se.

Our point estimate of the rate of MAT-HMR fusion on a single marked copy of chrIII, 3.5e-07 per generation, implies, if the rate of MAT-HML fusion is roughly equivalent, a total rate of either deletion on the order of 1e-06 per generation. This estimation was performed in ho diploids, by necessity of the process of strain construction; although HO is silenced in MATa/MATα diploids, it is possible that leaky expression of HO could drive a higher rate if some HO-initiated events resolve as deletions. The estimated rate is 1 to 2 orders of magnitude higher than the mutation rate to loss of function for a typical gene (Lang and Murray 2008). As such, it is conceivable that the observed parallelism of MAT-HMR/HML deletion is the result of selection for haploinsufficiency of a single gene, but we think it is more likely that gene dosage effects of at least 2 loci in the span of the deletion are involved.

As these large deletions may be considered segmental aneuploidies or monosomies, it is also illustrative to compare the rates of whole-chromosome aneuploidy. In that context, the measured rate is less than the average per-chromosome rate of aneuploidy observed in Zhu et al. (2014) and Sharp et al. (2018) (∼1e-04/16 = ∼6e-06), but higher than the average per-chromosome rate of reduction in copy number reported in those studies, the large majority of observed aneuploidies (∼90%) in both being gains in copy number.

In nature, MAT-HMR/HML deletions are presumably selected against as they are lethal in the haploid state. Therefore, the rate of such deletions may be considered to be a kind of load imposed by the 3-cassette mating type system. As this rate is roughly comparable to rates of aneuploidy, this load appears tolerable (although as discussed below, if aneuploidy often provides transient adaptive states, then aneuploidy rates are not true loads). Assuming that the rate of MAT-HMR/HML fusion is indeed a load in natural populations, the long-term benefits of the mating type system must necessarily outweigh this and other loads associated with sexual reproduction.

Aneuploidy as an adaptive mechanism

Several previous evolution experiments have observed aneuploidy in S. cerevisiae and other yeasts in response to various selective pressures (Gresham et al. 2008; Rancati et al. 2008; Selmecki et al. 2009; Chen et al. 2012; Gilchrist and Stelkens 2019). Most often, an increase in chromosomal copy number is observed, but there are some reports of selection for monosomy (Yang et al. 2013; Barney et al. 2021). It is hypothesized that adaptation via aneuploidy might be important in natural populations by providing a transient adaptive state, especially as natural isolates of S. cerevisiae tolerate aneuploidy more easily than lab strains (Muenzner et al. 2024). It would be interesting to extend the evolution reported here to observe how and if this segmental aneuploidy is resolved into, or replaced by, some euploid adaptive genotype, or whether it persists indefinitely.

Conclusions and future directions

In this experiment, we selected for increased cell separation in haploid and diploid populations of wild yeast strains for over 380 generations. In haploids, we found parallel selection for loss-of-function mutations to AMN1 and for loss-of-function mutations to genes of the Rim101 pathway, pointing to a previously unrecognized role for the Rim101 pathway in regulating cell separation. In diploids, we found parallel selection for large heterozygous deletions spanning MAT-HMR or MAT-HML on chromosome III. These large deletions caused less sedimentation, smaller clusters of cells, and an altered budding pattern. We also found 1 diploid population fixed for a heterozygous mutation to CCH1, a calcium channel known to play a role in emergence from mating pheromone-induced arrest (Fischer et al. 1997) and other stress responses that depend on calcium influx. Further work is required to elucidate the roles of CCH1 and the Rim101 pathway in influencing cell separation and to determine how heterozygous MAT-HMR/HML deletions affect this phenotype. Continued evolution in this system could lead to insights regarding the long-term maintenance of large heterozygous deletions in diploids and/or the role of transient states of aneuploidy in adaptation.

Supplementary Material

jkag011_Supplementary_Data