An adaptive interaction between cell type and metabolism drives ploidy evolution in a wild yeast

Ploidy is an evolutionarily labile trait, and its variation across the tree of life has profound impacts on evolutionary trajectories and life histories. The immediate consequences and molecular causes of ploidy variation on organismal fitness are frequently less clear, although extreme mating type skews in some fungi hint at links between cell type and adaptive traits. Here we report an unusual recurrent ploidy reduction in replicate populations of the budding yeast Saccharomyces eubayanus experimentally evolved for improvement of a key metabolic trait, the ability to use maltose as a carbon source. We find that haploids have a substantial, but conditional, fitness advantage in the absence of other genetic variation. Using engineered genotypes that decouple the effects of ploidy and cell type, we show that increased fitness is primarily due to the distinct transcriptional program deployed by haploid-like cell types, with a significant but smaller contribution from absolute ploidy. The link between cell-type specification and the carbon metabolism adaptation can be traced to the noncanonical regulation of a maltose transporter by a haploid-specific gene. This study provides novel mechanistic insight into the molecular basis of an environment-cell type fitness interaction and illustrates how selection on traits unexpectedly linked to ploidy states or cell types can drive karyotypic evolution in fungi.


ABSTRACT 23
Ploidy is an evolutionarily labile trait, and its variation across the tree of life has 24 profound impacts on evolutionary trajectories and life histories. The immediate consequences 25 and molecular causes of ploidy variation on organismal fitness are frequently less clear, 26 although extreme mating type skews in some fungi hint at links between cell type and adaptive 27 traits. Here we report an unusual recurrent ploidy reduction in replicate populations of the 28 budding yeast Saccharomyces eubayanus experimentally evolved for improvement of a key 29 metabolic trait, the ability to use maltose as a carbon source. We find that haploids have a 30 substantial, but conditional, fitness advantage in the absence of other genetic variation. Using 31 engineered genotypes that decouple the effects of ploidy and cell type, we show that increased 32 fitness is primarily due to the distinct transcriptional program deployed by haploid-like cell 33 types, with a significant but smaller contribution from absolute ploidy. The link between cell-34 type specification and the carbon metabolism adaptation can be traced to the noncanonical 35 regulation of a maltose transporter by a haploid-specific gene. This study provides novel 36 mechanistic insight into the molecular basis of an environment-cell type fitness interaction and 37 illustrates how selection on traits unexpectedly linked to ploidy states or cell types can drive 38 karyotypic evolution in fungi.

INTRODUCTION 40
pure species [47], Saccharomyces eubayanus has become a model for microbial population 89 genomics and ecology [48][49][50][51][52][53], as well as a key target for applied biotechnological research 90 [54][55][56][57][58][59]. A focal ecological and industrial trait in this wild species is the ability to consume and 91 metabolize the α-glucoside maltose, which is the most abundant sugar in the wort used to brew 92 beer [60,61]. This trait is nearly ubiquitous among isolates of Saccharomyces eubayanus and its 93 sister species Saccharomyces uvarum [62], but it has been lost [63] or severely curtailed [64] in 94 the Holarctic subpopulation of S. eubayanus [51]. Paradoxically, members of this group contain 95 functional structural maltose metabolism genes, and their inability to use this sugar as a carbon 96 source may be due to cis-or trans-regulatory evolution resulting in altered expression of 97 structural metabolism genes. 98 In an effort to identify mechanisms by which maltose utilization might be refined or 99 regained after secondary loss, we previously subjected a Holarctic S. eubayanus strain to 100 adaptive laboratory evolution (ALE) under selection for improved growth on maltose [64]. Here 101 we map the genetic basis of adaptation in the evolved clones. We find that, surprisingly, 102 haploids emerged and rose to high frequency in replicate ALE populations founded with a 103 type engineering, and for estimating the frequency of haploids in adaptive laboratory evolution 155 populations after plating. The multiplex reaction gives rise to MATaand MATα-specific 156 amplicons of differing size, which were resolved on 2% agarose gels. All reaction conditions 157 were per the manufacturer's instructions and were carried out alongside controls (diploid 158 MATa/MATα; haploid MATa; haploid MATα; no input DNA). We discarded any experiment 159 where the controls did not produce the expected amplicons (or lack thereof). To estimate the 160 frequency of haploids in populations, we screened a total of 55-56 single colonies across four 161 independent platings of each population. We note that this approach cannot formally 162 distinguish between cells of different ploidies with rare aberrant MAT locus composition (e.g. 163 diploid MATa/MATa will generate the same amplicon pattern as haploid MATa; loss of MAT 164 locus heterozygosity in diploid S. cerevisiae has been estimated to occur at a rate of 2x10 -5 per 165 cell per generation [20]). In addition, this S. eubayanus background is homothallic, meaning 166 that any diploid colony recovered following plating might represent a haploid cell in the 167 experimental population maintained in liquid medium. The rate of mating type switching and 168 clone-mate selfing on solid medium is likely orders of magnitude higher than loss of MAT locus 169 heterozygosity [14,70]; thus, our PCR-based estimates of haploid frequency may be 170 conservative. 171

Mating type testing 172
In addition to molecular validation of engineered strains, we tested the expressed 173 mating type of strains with altered MAT locus composition using microbiological assays. To 174 assess MATα expression, a saturated liquid culture of S. cerevisiae bar1-Δ was diluted 100-fold 175 and spread-plated to YPD, and 10μL of overnight query strain culture was spotted on top. For 12 was extracted using the hot acid phenol/ethanol precipitation method, but we added glass 242 beads during vortexing to aid lysis efficiency. Genomic DNA was digested using Turbo DNAse 243 (Promega), and RNA yield and quality were assessed by Qubit BR RNA assay (Thermo Fisher 244 Scientific), agarose gel electrophoresis, and Qubit RNA IQ assay (Thermo Fisher Scientific further analysis. We removed from analysis a single library from an evolved isolate grown in 254 maltose, as manual inspection of normalized gene expression values revealed that this sample 255 had stochastically lost the ChrXV aneuploidy. This reduced our power to detect statistically 256 significant differences in expression for that specific evolved isolate. All other samples from 257 evolved isolates remained aneuploid in both conditions. We considered differentially expressed 258 genes between conditions and genotypes with expression changes of greater than or equal to 259 twofold in either direction and Benjamini-Hochberg adjusted p-values of less than or equal to 260 0.01 (false discovery rate of 1%). Full differential expression analysis results can be found in 261 were gently thawed, and approximately 50μL of slurry was inoculated directly to 1mL SC-2% 280 Maltose. The optical densities of these samples were monitored, and they were harvested and 281 fixed in early log phase after a minimum of two doublings. We sampled 10,000 cells for each 282 query on an Attune NxT flow cytometer (Thermo Fisher Scientific). Analysis was performed in 283 FlowJo v10.
Except for experiments in rich medium shown in Fig. 2A, the conditions for fitness 286 assays were designed to mimic the original ALE conditions [64]. Briefly, this regime consisted of 287 culturing in 1mL SC medium with 2% maltose and 0.1% glucose (hereafter, "competition 288 medium") with semiweekly 1:10 dilutions into new competition medium. Query genotypes 289 were directly competed against a common competitor in co-culture. The competitor was a 290 haploid in the ancestral S. eubayanus strain background with the exception of a constitutively 291 expressed GFP using a TEF1 promoter and ADH1 terminator from Saccharomyces cerevisiae and 292 ste12 deletion (MATa hoΔ::PScTEF1-yEGFP-TScADH1-kanMX, ste12Δ::natMX). We chose a ste12 293 deletion to prevent any interaction with competitors expressing MATα. Strains were streaked 294 to single colonies on YPD containing antibiotic as appropriate, precultured in competition 295 medium for three days, mixed in approximately equal query-to-competitor ratios (except where 296 we reduced the competitor ratio against less-fit query genotypes), sampled into cold 1X PBST 297 for flow cytometry of timepoint 0, and inoculated into 1mL competition medium at an initial 298 OD600 of approximately 0.1. At each transfer, competitions were sampled into cold 1X PBST for 299 flow cytometry, and the optical density of each replicate was measured to calculate the number 300 of generations. Competitions in rich medium were carried out in the same manner, albeit that 301 preculturing and propagation were in sterile-filtered YPD in 2mL volume with daily dilutions of 302 1:100. For both competition regimes, we sampled 13,000 cells per replicate and timepoint on 303 an Attune NxT flow cytometer (Thermo Fisher Scientific) to quantify the abundance of 304 competitor (fluorescent) and query (non-fluorescent) cells, which always clearly formed distinct 305 populations. Analysis was performed in FlowJo v10. Fitness was calculated as the selection 306 coefficient, obtained by regressing the natural log ratio of query to competitor against the number of generations. To analyze the effects of ploidy, mating type, and cell type (diploid-like 308 and haploid-like) on the panel of engineered strains shown in Fig. 3, we used multiple linear 309 regression with measured fitness as the response and ploidy, mating type, and cell type as 310 categorical predictors with two levels each (for mating type, we grouped by whether genotypes 311 expressed any mating type-specific genes, or none). All statistical analyses and visualization 312 were performed in R. 313

PAGT1 reporter analysis 314
We generated single-copy genome integrations in haploids of yeast-optimized GFP 315 (yeGFP) expressed from both the native AGT1 promoter and a variant in which we abolished 316 the Tec1p consensus site (TCS) by making point mutations to each of its six nucleotides. 317 Sequence-verified GFP strains and controls were streaked to single colonies on YPD containing 318 antibiotic as appropriate, picked to SC-2% maltose and grown to saturation, back-diluted in 2mL 319 SC-2% maltose to an initial OD600 of 0.01, and grown to mid-log phase (OD600 ≈0.5-0.7). Cells 320 were collected by centrifugation, washed twice with cold PBST, and resuspended in PBST for 321 flow cytometry. We sampled 40,000 cells per replicate on an Attune NxT (Thermo Fisher 322 that displayed significantly increased growth (p=0.002, Mann-Whitney U tests) on maltose 330 compared to the ancestral strain (Fig. 1A). To map the genetic basis of improved growth on 331 maltose, we sequenced the genomes of each clone to a final average depth of 95-fold. We 332 mapped these reads to a re-sequenced and annotated assembly of the ancestral strain and 333 identified a total of four single nucleotide polymorphisms (SNPs) and three large-scale copy 334 number variants (CNVs) in the form of aneuploidies across the evolved isolates (Fig. 1B, Table  335 S3). We did not identify single-nucleotide variants in or near any genes with clear relationships 336 to α-glucoside metabolism, although one SNP introduced a premature stop codon in IRA1, a 337 common target of adaptive mutations in batch-style experimental evolution [7,22,97-99]. One 338 aneuploidy (ChrXV gain) was shared between evolved isolates and encompassed a homolog of 339 the S. cerevisiae generalist α-glucoside transporter AGT1/YGR289C, suggesting a potential 340 mechanism for adaptation ( Fig. 1B). 341 We noted two unexpected features of evolved mutations: all SNPs in the evolved 342 isolates were represented by a single, non-reference allele (Fig. S1), and all aneuploidies 343 identified were present at twofold relative copy number, while sequencing depth across the 344 genome of the ancestral strain indicated euploidy. Although mitotic recombination can 345 generate losses of heterozygosity at new or standing variation during adaptive evolution [100-346 105], our results differed significantly from two recent large-scale experimental evolution 347 studies in S. cerevisiae, which found approximately 5-10% of mutations to be homozygous in 348 diploid or autodiploid clones after 4,000 generations [7,9]. In comparison, our observed allele 349 frequencies at mutated sites are highly improbable under the null expectation of diploidy 350 (binomial tests: p=5.3x10 -6 , p=1x10 -4 , respectively). Similarly, while aneuploidies are common in both wild Saccharomyces isolates and as outcomes of experimental evolution or mutation 352 accumulation, tetrasomies-resulting in twofold relative copy number-are relatively rare in 353 diploids [8,15]. Thus, we reasoned that the observed patterns in copy number and allele 354 frequency might best be explained by an unexpected and atypical ploidy reduction to haploidy 355 during ALE. 356

Haploids emerged and rose to high frequency in diploid-founded populations 357
We directly determined the ploidy states of the evolved clones and the ancestral strain 358 using flow cytometry (Fig. 1C) and confirmed that the strain that was used to found the 359 experimental populations was diploid. Consistent with the results of genome sequencing, we 360 found that clones from both ALE replicates had become haploid. To test whether the clonal 361 isolates we analyzed were simply from a rare and non-representative subpopulation, we 362 assayed the ploidy states present at the population level in both replicates of the ALE 363 experiment (Fig. S2B). Haploids were clearly detectable in each replicate by generation 100 and 364 250, respectively. As an orthogonal approach, we plated cells from the terminal timepoint of 365 each population of the ALE experiment and used a PCR assay to genotype the MAT locus of 366 single colonies. By this method, haploids constituted 74-100% of the cells we genotyped in the 367 two ALE populations (Fig. S2C). Thus, although haploids did not sweep to fixation in both 368 experimental populations, they repeatedly emerged and rose to high frequency over the 369 duration of the ALE experiment. 370 Haploids exhibit a direct condition-dependent fitness advantage 371 The abundance of haploids in our experimental populations could be explained by two 372 alternative models: haploids might have a direct fitness advantage, or they might benefit 373 indirectly from increased adaptability in our ALE environment. Two well-documented lines of 374 evidence from previous studies seemed to strongly favor the latter hypothesis. First, S. 375 cerevisiae haploids have repeatedly been shown to adapt more rapidly than diploids during 376 experimental evolution, in part due to dominance effects at adaptive targets and ploidy-specific 377 mutation rates and spectra [7-12,106,107]. The common mutation shared by our adapted 378 clones, ChrXV aneuploidy, might be predicted to have a larger effect size in haploids than 379 diploids due to the difference in relative copy number conferred by the gain of a single 380 chromosome copy between ploidies and the general concordance between increased copy 381 number and gene expression in yeast [108-113]. Second, S. cerevisiae displays a strong trend of 382 converging on a diploid state during experimental evolution initiated with non-diploid strains 383 [18]. Although theory predicts that haploids may be better able to meet their metabolic needs 384 in nutrient-limiting conditions due to increased cell surface area-to-volume ratios, experimental 385 evidence in yeast has failed to find widespread support for such generalizable trends [23,114-386 117], and our experimental evolution conditions could not strictly be considered to be limited 387 in key nutrients. Given the relative simplicity of testing for differences in fitness between 388 ploidies, we first sought to support or refute the model of direct haploid advantage. 389 We used a sensitive competition assay to measure the competitive fitness of isogenic 390 diploids and haploids in the wild strain background following HO deletion, sporulation, and 391 tetrad dissection. Consistent with observations in S. cerevisiae of direct or cryptic diploid 392 advantage [7,15,17,18,20,22,118], haploids in our strain background exhibited median fitness 393 defects of 1.5% (p=1.5x10 -5 ) to 2.7% (p=9.9x10 -5 ) relative to the isogenic diploid in rich medium 394 ( Fig. 2A). By contrast, in the ALE conditions, haploids displayed median fitness advantages of 24.8% (p=1.6x10 -9 ) to 28.8% (p=3.6x10 -9 ) per generation over diploids (Fig. 2B). Interestingly, we 396 observed a significant fitness difference between haploids of opposite mating types in both 397 environments tested (rich medium p=1.1x10 -9 , evolution conditions p=0.018), suggesting a 398 common underlying mechanism linked to mating type, rather than a specific mating type-by-399 environment interaction. Expression of the mating-type genes is known to be costly isogenic haploids [23,26], the subtle, but significant, differences we observed here may have 407 been below previous limits of detection. Irrespective of mating type, we find that haploids have 408 a large and unexpected advantage over diploids under the ALE conditions. 409

Haploid fitness advantage is primarily due to cell-type specification 410
In Saccharomyces, ploidy is intrinsically linked with cell-and mating-type specification, between DNA content and cell-type specification can serve to confound inferences of the of either absolute ploidy or MAT locus composition have been documented [7,23]. Here we 418 refer to "cell-type specification" as the distinction between genotypes with a full complement 419 of cell-type master regulators (e.g. wild-type diploids containing MATa1, MATα1, MATα2) and 420 those without. Cell types established by the absence of one or more cell-type regulators (e.g. 421 wild-type haploids) effect the de-repression of a handful of genes, commonly referred to as 422 "haploid-specific," but whose expression is technically independent of ploidy and mating type. 423 To dissect the contributions of DNA content and cell type to organismal fitness in our 424 system, we engineered a series of eight otherwise isogenic genotypes with unique 425 combinations of ploidy, mating type, and cell type-specific gene expression. Beginning with a 426 diploid ho deletion in the wild-type background and haploids of each mating type, we 427 manipulated MAT gene composition to generate diploids specifying a haploid-like cell type and 428 a single mating type (by deleting one copy of the MAT locus at random) or no mating type (by 429 deleting MATα1 in a MATa-Δ background), as well as in haploids specifying no mating type (by 430 deleting MATα1) or a diploid-like cell type (by integrating a copy of the other MAT allele in a 431 haploid of the opposite mating type). We measured the fitness of these strains in the ALE 432 condition and estimated the separable effects of ploidy, mating-type specification, and cell 433 type-specific gene expression patterns on fitness (Fig. 3). These three factors explained the 434 majority of the variance in measured fitness across genotypes (multiple R 2 =0.96, df=86, p < 435 2.2x10 -16 ), with each having a significant effect (p ≤ 2.56x10 -7 ). Remarkably, cell-type 436 specification had an impact on organismal fitness that was almost an order of magnitude 437 greater than either ploidy or mating type (Fig. 3B, fitness advantage 18.8%, 95% CI: 17.7, 19.9), 438 and explained far more of the variance (proportion sum squares: cell type, 0.93; ploidy, 0.016; mating type, 0.014). Absolute ploidy nonetheless impacted fitness across cell types, with 440 haploids experiencing a 2.3% advantage relative to diploids in the ALE condition (95% CI: 1.5, 441 3.1). Paradoxically, expression of mating type-specific genes appeared to modestly increase 442 fitness between haploid-like cell types in the ALE condition, in contrast to the documented cost 443 of their expression in other conditions [119]. While one possible interpretation is that both sets 444 of mating type-specific genes confer bone fide fitness advantages to cells growing in maltose 445 medium, a more likely explanation for this apparent discrepancy is that haploid-like, MAT-null 446 cells experience modest off-target fitness defects as a result of their extensive genetic 447 engineering in the stringent ALE condition. As such, our analyses may slightly underestimate the 448 fitness benefit attributable to haploid-like cell type in the ALE condition. We conclude that the 449 cell type specified by the MAT locus, rather than absolute ploidy per se, has the largest effect 450 on fitness in the ALE condition. 451

A fitness-modifying maltose metabolism gene has cell type-specific increases in expression 452
The conditional fitness advantage of haploids and increased fitness of haploid-like cell 453 types suggested an unexpected regulatory link between maltose metabolism and haploid-454 specific genes (i.e. those genes de-repressed in the absence of a heterozygous MAT locus). To 455 identify potential targets of this interaction, we analyzed mRNA-seq data collected from the 456 wild-type diploid and evolved haploids grown in conditions mimicking the evolution experiment 457 (SC-maltose), as well as a baseline for comparisons (SC-glucose). Although the haploid strains 458 had discrete polymorphisms, they shared a common cell type and aneuploidy of chromosome 459 genotypes. Transcriptomes of the evolved haploids were highly similar, as expected (ρ=0.85, 462 Fig. S4B), with genotype accounting for 18% of the total variance in gene expression across all 463 samples in principal component space (Fig. S4A). Differentially expressed genes (DEGs) between 464 the wild-type strain and evolved haploids were enriched for cell-and mating-type specific 465 transcripts and genes on aneuploid chromosomes; however, there was no clear functional 466 enrichment among DEGs to explain the maltose-specific haploid fitness advantage. The AGT1 467 transporter on ChrXV was the single maltose metabolism-associated gene upregulated in 468 maltose in both evolved haploids when compared to the wild-type strain, which was expected 469 given its twofold relative copy number in these isolates (Fig. 1B). Upon closer examination, 470 however, AGT1 expression was higher than the twofold increase expected commensurate with 471 its relative copy number [122,123]. Indeed, AGT1 expression in haploids exceeded null 472 expectations based on two distinct models (Fig. 4A): 1) we calculated the fold-change for AGT1 473 in the ancestral strain in maltose compared to glucose and applied this multiplier to the glucose 474 expression level in the evolved haploids; 2) we applied a twofold multiplier to the gene 475 expression levels in the wild-type strain in both glucose and maltose, which accounted for copy 476 number variation in the evolved haploids. While AGT1 expression in glucose in the evolved 477 haploids was in line with the naïve aneuploid expectation (p=0.81, one-sided Mann-Whitney U 478 test), its expression in maltose in the evolved haploids was an average of 76% higher than could 479 be modeled by accounting for copy number and native regulation (p=0.0005, one-sided Mann-480 Whitney U test). 481 We compared expression levels of two relevant classes of genes under which AGT1 falls and 484 which we reasoned might be subject to modest differential expression: maltose-induced genes 485 and subtelomeric genes (Fig. 4B, Fig. 4D). We also examined expression of genes on the 486 aneuploid ChrXV to test whether these broadly exceeded the expectation of a twofold 487 expression increase commensurate with copy number (Fig. 4C). In each case, expression in the 488 evolved haploids was not significantly greater than the null expectation (one-sided t-tests, We next considered whether increased expression of AGT1 alone contributes to overall 510 fitness. Agt1p is a homolog of the well-characterized S. cerevisiae α-glucoside transporter, but 511 in contrast to canonical MAL gene clusters that contain structural and regulatory maltose 512 metabolism genes, S. eubayanus AGT1 is isolated in the subtelomeric region of ChrXV. In our 513 genome assembly, no predicted genes intersperse the AGT1 start codon and the beginning of 514 telomeric repeats some 6,770bp upstream. We did not identify any homologs of genes 515 encoding MAL regulators, transporters, α-glucosidases, or isomaltases on ChrXV, nor was there 516 any significant enrichment of functional categories for genes on this chromosome. As transport 517 is generally regarded as the rate-limiting step for carbon metabolism in yeasts [124-127], even 518 modest increases in transporter expression level could influence phenotype sufficiently for 519 selection to act; indeed, regulatory variants with smaller effects on gene expression show 520 signatures of natural selection in other eukaryotic systems [128][129][130][131]. To test whether 521 increasing the expression of AGT1 affects fitness, we inserted an additional copy of AGT1 under 522 its native promoter and terminator into the genomes of diploids and haploids at a separate site, 523 and we measured the fitness of the resulting strains in the ALE conditions. Increased AGT1 copy 524 number conferred a substantial and significant fitness benefit across backgrounds (Fig. S3). 525 Haploids received a more modest increase in fitness than diploids, likely attributable to 526 diminishing returns epistasis, but they were significantly more fit overall. There was no 527 interaction between the haploid mating type and the fitness effect of increased AGT1 528 expression (p=0.8, Mann-Whitney U test). The increase of AGT1 expression that we observed in 529 haploids is therefore likely to affect phenotype in an adaptive manner. 530 The AGT1 promoter integrates cell-type and sugar-responsive regulatory networks 531 We investigated potential regulators of AGT1 by scanning its promoter for putative 532 transcription factor-binding sites using high-confidence S. cerevisiae motifs. This analysis 533 identified binding motifs for two canonical regulators of maltose metabolism genes, Mal63p 534 and Mig1p (Fig. 5A) . We identified a predicted Tec1p-binding 550 site in the promoter of AGT1 with 100% identity to the TCS (Fig. 5A) and hypothesized that 551 Tec1p could mediate the cell type-specific increase in AGT1 expression we observed in haploids. 552 To test this hypothesis, we generated single-copy genomic integrations in haploids of a 553 GFP construct under the control of the wild-type AGT1 promoter (PAGT1), as well as a promoter 554 variant with point mutations in the predicted Tec1p-binding site (Pagt1-tcs). We then measured 555 single-cell fluorescence of the resulting strains grown in maltose by flow cytometry. Mutation 556 of the Tec1p-binding site significantly decreased fluorescence from the reporter construct 557 compared to the wild-type promoter (p < 2.2x10 -16 , two-sided t-test), but it did not abolish 558 expression completely (Fig. 5C). These results are consistent with the expression data and 559 collectively suggest that AGT1 receives regulatory input from both cell-type and sugar-560 responsive networks, with separable activation by Tec1p and induction in the presence of 561 maltose. In synthesis, the evidence for a direct fitness advantage by haploid-like cell types, 562 increased expression of fitness-modifying AGT1 in haploids, and the dependence of AGT1 563 expression on the motif for a haploid-specific transcription factor paints a clear picture of a 564 mechanistic relationship between ploidy evolution and adaptation in our system. 565

CONCLUSIONS 566
Resolving the genotype-to-phenotype map remains a central goal in genetics and 567 evolutionary biology, but it has frequently proven challenging, even in microbes. appears to be the focal maltose transporter is partially decoupled from such stringent 575 catabolite regulation: AGT1 is only induced ~2.3-fold in the wild-type strain in maltose (Table  576 S4). We can envision two potential explanations for the apparently unusual regulation of this 577 gene. 578 First, AGT1 is likely to encode a transporter with broad substrate affinity like its S. How generalizable might these principles be? Given the evolutionary lability of ploidy, 625 its link to cell type in many fungal species, and evidence for interactions between cell type and 626 conditionally adaptive traits in other fungal systems, we propose that environment-and 627 genotype-specific regulatory nuances might play a broad role in shaping both the extant 628 diversity of fungal ploidy states and the conflicting, and often cryptic, ploidy and cell type 629 evolution seen in systems experiencing intense selection. This view argues that interactions 630 between cell types, ploidy states, and conditionally adaptive traits may be common during 631 fungal evolution and may influence fungal life cycles more than is currently appreciated. 632

DATA AVAILABILITY STATEMENT 633
Strains and plasmids are available upon request. All raw sequencing data has been 634 deposited at NCBI SRA under BioProject PRJNA894214.

The contents of this manuscript have been disclosed to the Wisconsin Alumni Research 637
Foundation to evaluate whether a patent application should be filed. Their decision will not 638 affect data or strain availability for non-commercial, academic use.      Table S1. Strains and plasmids used in this study. 1262 Table S2. Oligonucleotides used in this study. 1263          U n t a g g e d P A G T 1 -G F P P a g t 1 -t c s -G F P p < 2.2x10 -16 Carbon metabolism-associated TFs Growth mode-associated TFs