Preferences in a Trait Decision Determined by Transcription Factor Variants

Preferences in a trait decision determined by transcription factor variants

Michael W. Dorritya,b, Josh T. Cuperusa,c, Jolie A. Carlislea, Stanley Fieldsa,c,d,1, and Christine Queitscha,1

aDepartment of Genome Sciences, University of Washington, Seattle, WA 98195; bDepartment of Biology, University of Washington, Seattle, WA 98195; cHoward Hughes Medical Institute, University of Washington, Seattle, WA 98195; and dDepartment of Medicine, University of Washington, Seattle, WA 98195

Contributed by Stanley Fields, June 27, 2018 (sent for review April 9, 2018; reviewed by Leah E. Cowen and Andrew W. Murray) Few mechanisms are known that explain how transcription factors mating, Ste12 can bind at pheromone-responsive genes as a can adjust phenotypic outputs to accommodate differing environ- homodimer or with the cofactors Mcm1 or Matα1 (10–13). The ments. In Saccharomyces cerevisiae, the decision to mate or invade consensus DNA-binding site of Ste12 is TGAAAC, known as the relies on environmental cues that converge on a shared transcription pheromone response element (PRE) (11). For invasion, Ste12 factor, Ste12. Specificity toward invasion occurs via Ste12 binding and its cofactor Tec1 are both required to activate genes that me- cooperatively with the cofactor Tec1. Here, we determine the range diate filamentation (8, 14, 15). Some of these genes contain a of phenotypic outputs (mating vs. invasion) of thousands of DNA- Ste12 binding site near a Tec1 consensus sequence of GAATGT, an binding domain variants in Ste12 to understand how preference for organization for heterodimeric binding known as a filamentation invasion may arise. We find that single amino acid changes in the response element (FRE). An alternative model, however, posits that DNA-binding domain can shift the preference of yeast toward either expression of invasion genes is driven by a complex of Ste12 and mating or invasion. These mutations define two distinct regions of Tec1, acting solely through Tec1 binding sites (15). An environmental this domain, suggesting alternative modes of DNA binding for each component of trait specificity has been shown in fungi that are animal trait. We characterize the DNA-binding specificity of wild-type pathogens; upon recognition of host body temperature (37 °C) (6), Ste12 to identify a strong preference for spacing and orientation Cryptococcus neoformans has decreased mating efficiency but in- of both homodimeric and heterodimeric sites. Ste12 mutants that creased ability to invade (16). Although trait preference in S. cerevisiae promote hyperinvasion in a Tec1-independent manner fail to bind cooperative sites with Tec1 and bind to unusual dimeric Ste12 sites depends on Ste12, the role of this protein in regulating mating and composed of one near-perfect and one highly degenerate site. We invasion in response to increased temperature is unknown. propose a model in which Ste12 alone may have evolved to activate Because the highly conserved Ste12 DNA-binding domain invasion genes, which could explain how preference for invasion ultimately enacts the choice between mating and invasion, here arose in the many fungal pathogens that lack Tec1. we sought to precisely define its DNA-binding specificity and ex- amine the consequences of mutations and increased temperature yeast | transcription factor | fungal mating | fungal invasion | on the balance of these traits. We reveal distinct organizational

evolutionary trade-off preferences for both homodimeric binding sites and cooperative GENETICS heterodimeric binding sites with Tec1. Furthermore, single amino ranscription factors interact with DNA, with cofactors, and acid changes suffice to shift the preference of yeast cells toward Twith signaling proteins to allow cells to respond to changes in one or the other trait; they also suffice to confer dependence on their environment. Despite the requirement to manage these multiple levels of regulation, most eukaryotic transcription factors Significance possess a single DNA-binding domain. Distinct responses must therefore be mediated by diversity in cofactors, organizations of Transcription factors have been intensively examined to de- binding sites, and conformational changes in the transcription cipher how they regulate cellular decisions, but there are few in- factor itself. For example, human GCM1 gains a novel recognition depth studies of these factors across traits, environments, and sequence when paired with the ETS family factor ELK1 (1); genetic backgrounds. Here, we analyze the Saccharomyces cer- auxin-responsive transcription factors regulate expression differ- evisiae Ste12 protein, a transcription factor essential for both entially depending on the arrangement of their binding sites (2); mating and invasion in many fungal species. Generating thou- and the heat-shock factor Hsf1 senses increased temperature by sands of variants in the Ste12 DNA-binding domain, we scored changing its conformation, which allows it to bind a unique rec- each variant for its activity in promoting both mating and in- ognition sequence (3). vasion. We found altered DNA-binding patterns of exceptional We sought to investigate the features of a single transcription variants that result in yeast that lose their mating efficiency, but factor that uses distinct cofactors, binding sites, and environmental gain increased competence in invasion. This surprising mallea- inputs to mediate a cellular decision. The Ste12 protein of the bility in transcription factor function has implications for un- yeast Saccharomyces cerevisiae governs the choice of mating or derstanding the evolution of pathogenicity in fungi. invasion. Each of these traits contributes to cellular fitness: mating Author contributions: M.W.D., J.T.C., S.F., and C.Q. designed research; M.W.D. and J.A.C. of haploid yeast cells is required for meiotic recombination, and performed research; M.W.D. contributed new reagents/analytic tools; M.W.D. analyzed invasion allows a cell to forage for nutrients or to penetrate tis- data; and M.W.D., S.F., and C.Q. wrote the paper. sues, a characteristic of pathogenic fungi. Mating is initiated by Reviewers: L.E.C., University of Toronto; and A.W.M., Harvard University. binding of the appropriate pheromone, which activates an evolu- The authors declare no conflict of interest. tionarily conserved G protein-coupled mitogen-activated protein Published under the PNAS license. kinase (MAPK) pathway (4, 5) (Fig. 1A). In contrast, invasion is Data deposition: The high-throughput sequence reads reported in this paper have been initiated in response to increased temperature and limited nutri- deposited in the National Center for Biotechnology Information Sequence Read Archive, ent availability (6–8). The shared protein kinase Ste11, a client of https://www.ncbi.nlm.nih.gov/sra (accession nos. SRR5408956–SRR5408965). the chaperone Hsp90 (9), likely contributes to the environmental 1To whom correspondence may be addressed. Email: [email protected] or [email protected]. sensitivity of both traits. This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. The two pathways converge on Ste12, which interacts differ- 1073/pnas.1805882115/-/DCSupplemental. entially with cofactors to activate either mating or invasion. For Published online August 1, 2018.

www.pnas.org/cgi/doi/10.1073/pnas.1805882115 PNAS | vol. 115 | no. 34 | E7997–E8006 Downloaded by guest on September 25, 2021 AD E

Fig. 1. Conservation of Ste12 and Tec1 and their DNA-binding sites. (A) The yeast mating and invasion pathways contain shared signaling components, and both depend on Ste12 for activation of distinct regulatory programs. (B) S. cerevisiae Ste12 protein has a nonconserved transcriptional activation domain, as well as a highly conserved DNA-binding domain. The secondary structure of this domain is predicted to contain a pattern of α-helices (gray boxes) interspersed with β-sheets (blue arrows) that are conserved among all fungal species. The region of the DNA-binding domain chosen for mutagenesis is shaded in orange. (C) A logo plot of the conserved portion of the DNA-binding domain of Ste12 generated from 1,229 fungal species. (D) A phylogeny of fungal species selected for those with functionally tested STE12 genes (except for the basal fungi Rhizophagus irregularis and Rozella allomycis, shown as outgroups). Pathogenic species are indicated with circles and colored according to plant (green) or animal (red) hosts. All fungal species were queried for the presence of a STE12 (blue) or TEC1 (orange) gene, and filled squares indicate presence of either gene; numbers inside boxes indicate species with multiple gene copies. Results for all species are shown below each gene’s column. (E) Proportional bar chart showing the number of S. cerevisiae genes that contain Ste12 sites (blue), Tec1 and Ste12 sites (yellow), and adjacent sites fewer than 24 bases from each other (green). (F) Histogram showing frequencies of spacing between all Ste12 and Tec1 binding sites in the S. cerevisiae genome. Negative values indicate that the Tec1 site is 5′ of the Ste12 site, and positive values the converse.

the chaperone Hsp90 and responsiveness to higher temperature. Cooperative binding at adjacent (within 24 bp) Ste12 and Tec1 Some hyperinvasive separation-of-function mutations are in- binding sites (FREs) has been demonstrated in vivo and in vitro dependent of the cofactor Tec1, thought to be essential for in- (18), although most invasion genes do not contain FREs (15). For vasion. We show that these Ste12 variants bind to homodimeric all promoter sequences in S. cerevisiae, we assessed the frequency Ste12 binding sites in which one site is highly degenerate. Such a and spacing of sites, and identified putative FRE sites (<24 bp). − binding preference provides a plausible mechanism to activate the We found 792 unique promoters with matches (P < 1e 4)toa expression of invasion genes in a Tec1-independent manner, and Ste12 binding site, with 220 (27.8%) of these also having a Tec1 provides a model for the regulation of these genes in the many binding site (Fig. 1E). The median distance between Ste12 and fungal species that have no copy of TEC1. Tec1 sites is 360 bp, and only 12 pairs of sites (5%) are within 24 bp of each other (Fig. 1F). Among these 12 pairs, 4 have Results overlapping motifs consisting of a tail-to-tail arrangement, with STE12 Is Present in Nearly All Fungal Genomes, but Most Lack a TEC1 the Tec1 motif overlapping the 3′ end of the Ste12 motif (Fig. 1F, Ortholog. As the target of two different MAPK signaling cascades, Inset); this organization is more likely than nonoverlapping sites to Ste12 acts to define pathway specificity at the level of transcription occur by chance and may not be functional, although overlapping (Fig. 1A). This role in both mating and invasion appears to be sites have been observed in cooperative binding of mammalian conserved in other fungal species, and presumably derives from transcription factors (1). The small number of sites organized for properties of its DNA-binding domain, the most conserved segment cooperative binding with Tec1 contrasts with the hundreds of of the protein (Fig. 1B) (17). Although the Ste12 DNA-binding genes up-regulated under invasion conditions (19, 20), as well as domain has no match outside of the fungal kingdom, several resi- with the dozens of genes bound by Ste12 under invasion condi- dues as well as the overall predicted secondary structure are con- tions (57 genes unique to invasion, 100 overall) (14). Thus, the served in fungi (Fig. 1 B and C). We addressed the cooccurrence of model of Ste12 and Tec1 cooperatively binding to FREs within Ste12 and Tec1 by examining 1,229 fungal species for the presence invasion genes cannot alone account for the broad transcriptional or absence of these two genes. Nearly every fungal species contains a response during invasion observed in S. cerevisiae. Specificity may copy of STE12 (97.7% of species), while less than one-third (31.7%) derive from instances of long-range looping interactions, or appear to have a copy of TEC1 (SI Appendix,Fig.S1). Furthermore, binding of Ste12 to DNA indirectly through its interaction with even among fungal pathogens with a characterized role for Ste12 in Tec1 bound at Tec1 consensus sequences (15). However, given the invasion, we found several examples in which a TEC1 gene is not absence of Tec1 in many species, this model is also unlikely to be present (Fig. 1D). Therefore, fungal species must have evolved Tec1- applicable in fungi more broadly. independent strategies to regulate mating and invasion. We examined the extent to which Ste12 and Tec1 binding at Ste12 and Tec1 Preferentially Bind DNA in Defined Spacings and adjacent sites within filamentation response elements (FREs) Orientations. Because the yeast genome contains few FREs, we could explain invasion-specific activation by Ste12 in S. cerevisiae. sought to determine at high-resolution the in vitro DNA-binding

E7998 | www.pnas.org/cgi/doi/10.1073/pnas.1805882115 Dorrity et al. Downloaded by guest on September 25, 2021 preferences of Ste12 and Tec1. We used high-throughput sys- put sequences) for tail-to-tail binding sites with a 3-bp spacer (Fig. tematic evolution of ligands by exponential enrichment (HT- 2A, Top). While the most enriched sequences contained two SELEX), a method that determines a protein’s DNA-binding perfect TGAAAC sites with this spacing (Fig. 2B, Top), most of specificity by isolating and sequencing bound fragments present bound sequences contained one perfect TGAAAC site paired in a large, random pool of sequences (1, 21, 22). For the pool of with a mismatched site. This strong spacing preference had not DNA, we used fragments with a random sequence of 36 bp, which been identified in previous protein-binding microarray studies, allowed us to capture instances of Ste12 and Tec1 binding sites in which used shorter target sequences (23), but is consistent with the all orientations and at every possible spacing of 24 bp or fewer. In known binding of Ste12 as a homodimer (24). We found two in- the Ste12 sample, we captured monomeric or homodimeric instances of homodimeric sites with a perfectly spaced TGAAAC stances of the expected Ste12 binding site of TGAAAC, with di- pair in the S. cerevisiae genome: in the promoters of GPA1, meric sites found 20-fold more frequently than monomeric sites encoding the α-subunit of the G protein involved in pheromone (SI Appendix,Fig.S2C). Many transcription factors exhibit strong response, and of STE12 itself (Fig. 2C). Gpa1 is a component of preferences for spacing and orientation of sites, and this prefer- the initial signaling point after pheromone sensing, and Ste12 is ence is conserved within transcription factor families (1, 22). the ultimate transcription target of the MAPK signaling cascade, Ste12 showed a strong preference (33% of 2,229,168 bound out- and both are among the most highly induced genes in response to

AB3bp Ste12

0510 15 Tec1 orientation

051015 2bp Ste12-Tec1 1.0 0.5

spacing 0.0

frequency of frequency 051015 spacing between sites (bp) GENETICS CDpreferred homodimeric preferred heterodimeric two perfect sites GPA1 GTTTCAAATTGAAAC GPB1 GAATGGGATGAAAC STE12 GTTTCAAAATGAAAC CHS7 GTATGTTGTGAAAC GAATGTTATGAAAC 1 perfect + 1 mismatch SRL3 BST1 GAATGTGTTGAGAC GAATGATCTGAAAC FUS3 GTTTCACCCTGGAAC PRM6 GAATGCCTCGAAAC DUO1 GTTTCACAGCGAAAC BUD8 SIT4 GAATGGCTTGAAAC PMT6 GTTTCGCGTTGAAAC TIR1 GAATGCTATGAAAC RKI1 GTTGCAGTGTGAAAC SYS1 GAATGAGTTGAAAC GTTTCAGAATGCAAC FAR1 BCY1 GAACATCGTGAAAC PRM6 GTTTAAGTCTGAAAC STE13 GTTTCACGGTATAAC STE14 GATTGATAATGAAAC AGA1 GTTTCAAGTCAAAAC PRP39 GTTTCAGACTTAAAC KAR4 GTCGCAAGATGAAAC

Fig. 2. Identification of the DNA-binding preferences of Ste12 and Tec1 by HT-SELEX. (A) Heatmap showing relative frequencies of the possible orientations and spacings of the primary 6-mer selected in Ste12 (TGAAAC, Top) and Tec1 (GAATGT, Middle) binding reactions. The Bottom heatmap shows the frequency of each respective 6-mer in the cobinding sample containing both proteins. In the cobinding sample, we excluded sequences with homodimeric Ste12 sites, which were present at a 30-fold higher level than heterodimeric sites. The single dark green box in the Top and Bottom heatmaps indicates the most frequent orientation and spacing of sites; white boxes are at most 30% as frequent as the maximum. No frequent dimeric site organizations were observed for Tec1 (Middle). (B) Full motifs identified by Autoseed software (1). (C) Instances of the Ste12 homodimeric motif, with sites oriented tail-to-tail with a 3-bp spacer, were used to query native yeast promoters. Two pheromone-induced genes, GPA1 and STE12, contain two perfect sites, while most other genes contain a perfect site paired with a site containing one or two mismatches, with the 3-bp spacing intact. (D) A similar search of yeast promoters using the preferred Ste12 and Tec1 heterodimeric site identified a set of invasion-associated genes, as well as those not previously linked to invasion.

Dorrity et al. PNAS | vol. 115 | no. 34 | E7999 Downloaded by guest on September 25, 2021 pheromone (25). Elsewhere in the genome, two sites spaced 3-bp vasion genes in the S. cerevisiae genome (GPB1, BST1, PRM6, apart, in which one site is perfect and one is mismatched, are BUD8, and SIT4) contain this type of heterodimeric site, al- found within several pheromone-regulated genes. This configu- though other genes with this binding site organization (CHS7, ration occurs in 29 genes in the S. cerevisiae genome, and 12 of the SRL3, TIR1, SYS1,andBCY1) have not been previously associated 50 most pheromone-induced genes contain these sites (examples with invasion (Fig. 2D). This arrangement is similar to homodi- in Fig. 2C). To analyze the ability of these sites to drive Ste12- meric Ste12 sites, except that one Ste12 site is replaced by a Tec1 dependent expression, we used a reporter assay in yeast. The in site, suggesting that a similar protein interaction surface may be vitro preferences for the dimeric sites, as well as for the base used for this heterodimeric binding mode of Ste12, as is used for flanking the TGAAAC motifs, correlated with in vivo activation the homodimeric binding mode. However, even in this sample, the (SI Appendix,Fig.S2A and B). pool of bound sequences was dominated by two Ste12 sites with HT-SELEX with Tec1 protein recapitulated the known bind- 3-bp spacing, suggesting that Ste12 has higher affinity to two PRE ing site of (A/G)GAATGT (Fig. 2B, Middle) (26). We found that sitesthantoanFRErequiringbindingwithTec1. the first base of this motif showed a strong preference for a purine base, and that the final T showed less specificity than the Mutations in the DNA-Binding Domain of Ste12 Separate Mating and rest of the motif. Unlike with Ste12, we did not observe any Invasion Functions. The organization of DNA-binding sites se- spacing and orientation preferences for two Tec1 sites, indicating lected by Ste12 alone or in combination with Tec1 suggested that that bound sequences with multiple Tec1 binding sites are likely Ste12 balances the expression of mating and invasion genes by its the result of independent binding events (Fig. 2A, Middle). mode of binding to DNA. We sought to determine whether this We next conducted HT-SELEX in the presence of both balance could be shifted by mutations within the Ste12 DNA- Ste12 and Tec1 to identify patterns of cooperative binding events binding domain. We conducted deep mutational scanning (28) of between the two proteins. The DNA-binding domain of Ste12 a segment of this domain by generating ∼20,000 protein variants (1–215) is sufficient for cooperative binding with Tec1 (1–280) in over 33 amino acids, including single-, double-, and higher-order vitro (26, 27). We detected bound sequences containing sites for mutants, and subjected yeast cells carrying this variant library to both proteins, with the highest abundance preference being a selection for either mating or invasion (Fig. 3 A and B and SI tail-to-head orientation of Tec1 and Ste12 sites separated by 2 bp Appendix, Figs. S4 and S5A). As a control, we employed selection (Fig. 2 A and B, Bottom, and SI Appendix, Fig. S3). Several in- for a third trait, response to osmotic stress, which shares upstream

ABposition in DNA binding domain 140 145 150 155 160 165 170 A A input C C input D D E E mating score F F log2 relative to wt α α G G 2 : 3 variants H H wash surface 1 : 2 10,000 protein wt I I K K 0 : 1 L L −1 : 0 M M −2 : −1 N N −3 : −2 output output P P Q Q −4 : −3 amino acid R R no data invasion mating S S T T wt aa V V W W Y Y * STOP 1 conservation 0 140 145 150 155 160 165 170 FRNMCLKTQ KKQ KVFFW FSVAHDK L F A DAL ERD wt sequence C α β α 2˚ structure prediction 01

invasion mating osmo control log2 95% conf. interval 4−2 − positional mean

Fig. 3. Deep mutational scanning identifies Ste12 DNA-binding domain mutants with altered mating and invasion function. (A) Two yeast populations were transformed with the same STE12 variant library. For assaying invasion, Σ1278b MATα cells (n = 400,000 per replicate, three biological replicates) were plated on selective media and grown. Selection was performed by washing colonies from plate surfaces and collecting cells embedded in the agar for sequencing. For assaying mating, BY4741 MATa ste12Δ cells were mated to MATα cells, and diploids (n = 500,000 per replicate, three biological replicates) were selected, scraped from the agar, and sequenced. In both cases, input variant frequencies were defined before selection. (B) The effects of single amino acid substitutions within Ste12’s DNA-binding domain on mating ability are shown. On the x axis, the wild-type Ste12 sequence is shown, along with its predicted secondary structure (helices shown as tubes) and conservation. Conservation was determined as a fraction of identity among 1,229 fungal species. On the y axis, amino acid substitutions are shown. Variants increasing mating efficiency are in shades of blue, and variants decreasing mating efficiency are in shades of

orange based on log2 enrichment scores relative to wild-type. Dark gray circles indicate the wild-type Ste12 residue, and crosses indicate missing data. Ste12 variants showed comparable expression levels (SI Appendix, Fig. S6). (C) Positional mean scores for single amino acid substitutions are shown for mating (black) and invasion (blue), excluding stop codons. Gray horizontal bar indicates confidence interval for experimental noise determined from selection for Ste12-independent high osmolarity growth.

E8000 | www.pnas.org/cgi/doi/10.1073/pnas.1805882115 Dorrity et al. Downloaded by guest on September 25, 2021 pathway components but does not involve Ste12. Mating selections effect is recapitulated in the effects of individual mutants tested in and osmotic stress selections were carried out in the BY4741 strain both trait selections at extreme positions on the front (SI Appen- background. However, as this strain contains a flo8 mutation that dix,Fig.S9). Positions well below the Pareto front, including most prevents invasion, we carried out invasion selections in a related prominently W156, had deleterious effects on both traits. These strain, Σ1278b (29), which has been used for large-scale invasion positions are significantly more conserved among fungi, consistent phenotyping (30). Although Σ1278b and BY4741 have many genetic with variation at these sites being disfavored given the constraint differences, Ste12 and Tec1 and their essential roles in balancing on Ste12 to maintain both mating and invasion function in the mating and invasion are conserved between the two strains. fungal lineage (Fig. 4C). That opposing preference for each trait is We expected that most STE12 mutations would affect mating better predicted by positional means than every individual muta- and invasion similarly, because of the conservation of the Ste12 tion suggests that trait preference is encoded in particular func- DNA-binding domain and its requirement for both traits. Indeed, tional regions of the Ste12 DNA-binding domain rather than we found positions in which almost any amino acid substitution was through overall features, like protein stability. highly deleterious to both mating and invasion (Fig. 3 B and C and Having established that trait preference can be modulated by SI Appendix,Fig.S5). We identified sites that were more sensitive mutations in Ste12, we asked whether temperature affected or less sensitive, on average, to mutation, by calculating a positional specificity toward mating and invasion, and if that specificity could mean score from all mutations tested at that site. Each mutation be altered by mutation. Furthermore, because chaperones main- was tested in triplicate, and the experimental error was calculated tain protein function at increased temperatures and Hsp90 mod- for each mutant individually such that the SE of the positional ulates pheromone signaling in S. cerevisiae (9), we also asked mean represents the variability among different amino acid sub- whether mating or invasion changed in the presence of radicicol, a stitutions, rather than experimental noise (Fig. 3C and SI Appendix, pharmacological inhibitor of Hsp90. We confirmed that mating is Fig. S4E). Conservation only partially explained these results, as modulated by temperature and by Hsp90 function by mating cells some of the most deleterious positions are invariant among fungi expressing wild-type Ste12 at increased temperature or in the (W156, C144), whereas others are not (K149, Q151, K152). Dif- presence of radicicol (Fig. 4D). We then subjected cells containing fering expression levels among STE12 variants were not predictive the Ste12 variant library to a mating selection at 37 °C or a mating for either mating or invasion phenotypes (SI Appendix,Fig.S6). selection in the presence of radicicol. Most Ste12 variants Because invasion was measured in a different strain background responded to increased temperature or Hsp90 inhibition as did than mating for each variant, we verified that strain background did wild-type Ste12 (SI Appendix,Fig.S10). However, there were two not account for the trait differences. We selected five variants that positions, K150 and K152, in which mutations resulted in highly conferred differential mating and invasion phenotypes [K149E, temperature-responsive and Hsp90-dependent mating (Fig. 4E). K150I, K150A, K152L, and (see below) the double-mutant S177K, This pair of lysines resides within the mutagenized segment that Q180R, denoted SKQR], and assayed their mating efficiencies in modulates mating and invasion specificity. We validated the the Σ1278b background. The mating efficiencies for each variant temperature and Hsp90 effect on a variant (K150I) that conferred correlated (r2 = 0.7) between strains, with the average difference mating at near wild-type levels in the absence of heat or radicicol for the five variants equal to 21% (SI Appendix,Fig.S6D). treatment (Fig. 4F). However, in the presence of heat or radicicol

The mutational analysis revealed separation-of-function muta- treatment, mating of cells with the K150I variant was severely GENETICS tions that primarily reduced either mating or invasion. Substitu- decreased, indicating that this mutation led to buffering by Hsp90. tions with mostly deleterious effects on mating clustered in Thus, Hsp90 could facilitate a mutational path toward the pathogenic residues N-terminal to the conserved W156 (Fisher’s exact test lifestyle by minimizing mating costs at 30 °C and enhancing invasion P = 0.002, designated region I), while substitutions with increased at 37 °C. To test this idea, we conducted selection for invasion at effects on invasion appeared more frequently in this region (P = 37 °C, which yielded results comparable to Hsp90 inhibition for 0.0003). Substitutions that increased mating were rare. In contrast, mating. Indeed, the mean effect of all variants at K150 was to in- substitutions at some positions, almost all within region I, both crease invasion at high temperature, with a concomitant decrease in increased invasion and decreased mating. We verified separation- mating, and this result was also true of the individually validated of-function mutants, and further explored the apparent structure variant K150I (Fig. 4G), demonstrating that K150 variants were not of mutational effects, with a much larger pool of >20,000 double simply unstable at high temperature but gained a novel function. mutants. Double mutants involving two positions in region I were Thus, mutations in STE12 caninteractwithanenvironmental factor mostly defective for mating, while those involving two positions in to further bias cellular decision-making toward invasion over mating. region II were mostly defective for invasion, confirming the bi- Mutation of STE12 could allow S. cerevisiae to mimic the behavior partite arrangement of the mutagenized segment, split by the of fungal pathogens like C. neoformans, for which the sensing of central W156 (P = 1.3e-5) (Fig. 4A). Double mutants between the increased body temperature of their animal hosts facilitates regions I and II showed variability in mutational effects, with some the transition toward an invasive lifestyle (7). pairwise combinations increasing mating and others decreasing We examined whether natural variation in fungal Ste12 DNA- mating; the effects on invasion, however, for these same combi- binding domains includes amino acid residues found to affect S. nations did not follow the same pattern, suggesting epistatic in- cerevisiae trait preference, as fungal species have differential ca- teractions (examples highlighted in Fig. 4A and SI Appendix,Fig. pacities for mating and invasion (17, 32). We examined variation in S7). Thus, mutations in the Ste12 DNA-binding domain can im- the DNA-binding domain present in species with and without a pose preference for mating or invasion rather than similarly af- TEC1 gene (SI Appendix,Fig.S11) and detected variation outside fecting both traits, suggesting that the DNA-binding domain itself of the mutagenized region that might be expected to benefit in- contributes to trait specificity. vasion. For example, the Ste12 DNA-binding domain of the in- To explore the apparent trade-off between mating and invasion vasive pathogen Cryptococcus gattii (33) contains two additional in more detail, we used the Pareto front concept (31). This con- positive residues immediately C terminal to region II; such changes cept, rooted in engineering and economics, defines all feasible are found in many other species that lack Tec1 (SI Appendix,Fig. solutions for optimizing performance in two essential tasks. Here, S11). Introducing these C. gattii-specific residues (S177K, Q180R; we consider all feasible genotypes that affect mating or invasion. denoted SKQR) into the S. cerevisiae Ste12 DNA-binding domain We plotted the single mutation mean positional scores for both yielded a dominant invasion phenotype (Fig. 5A) with neutral effects traits, identifying positions close to the empirically determined on mating (SI Appendix,Fig.S6D)inS. cerevisiae. We introduced a Pareto front that show increased invasion and deleterious effects negatively charged residue (K175E) into this same region, which on mating (Fig. 4B and SI Appendix,Fig.S8). We found that this abolished invasion entirely (Fig. 5A). Similar to SKQR, the region

Dorrity et al. PNAS | vol. 115 | no. 34 | E8001 Downloaded by guest on September 25, 2021 AD

K150,D172 temp 37C mating increased 34C log2 score wt 30C - no inhibitor Hsp90 5uM - radicicol -6median 0 region II inhibited 25uM

region I 0 0.5 1 E relative mating efficiency FLSFLFRNMCLKTQKKQKVFFWFSVAHDKLFADALERDLKRES C ° 5 1.0

invasion log2 score K150 -3median 1 log2 shift in 37 K152 −1.0 0 0. −1.0 −0.5 0.5 1.0 −1.0−1 −0.5 0 0.5 1.0 log2 shift in -Hsp90

p = 3E-4 BCFK150I p = 8E-4 p = 8E-7 G mating 1 wt

F157 1.0 preferred

0 K150I 0 mating fficiency 012

-2 0 invasion −1 -1-2 among fungi below front conservation invasion 30°C 37°C −4 −2

log2 mating score preferred

W156 log2 mating e K152 0.3 30°C log2 positional mean K150 -Hsp90 −3 −2 −1 0 1 37°C log2 invasion score near below front front

Fig. 4. Mutations residing in a region of the Ste12 DNA-binding domain increase invasion at the cost of mating, with exceptional mutations doing so in a temperature- and Hsp90-dependent manner. (A) Mean effect of double mutations between all combinations of positions for mating (above wild-type se-

quence) and invasion (below wild-type sequence). A specific pair of positions is shown in bold. Effects are shown as log2-fold change relative to wild-type and color-coded as in Fig. 3, functional variants in shades of blue, nonfunctional variants are shown in shades of orange. Note difference in scale ranges between mating and invasion. Black lines emanating from W156 represent boundaries of region I, in which mutations primarily reduced mating, and region II, in which mutations primarily reduced invasion. (B) Scatterplot of positional mean scores for both mating and invasion showed inverse relationship, indicating a trade- off between both traits. A point is shown for each of the 33 mutagenized positions of STE12. The trade-off is visualized as a Pareto front (black line) determined empirically from the positional data; positions near the front maintain high values for one trait and minimize costs to the other. Arrows indicate preference for either mating (gray) or invasion (blue). Positions near the front were distinguished from those below the front (shaded in red) by calculating Euclidian distances. (C) Boxplots show conservation for positions near and below the front; positions below the front are significantly more conserved among fungi than those near the front (two-sided t test). (D) Mating efficiency of yeast cells with wild-type Ste12 at increased temperature (dark red) or in the presence of Hsp90 inhibitor radicicol (pale red) is reduced relative to an untreated control (black); error bars represent SEM. (E) The mating efficiency of Ste12 variants at high temperature or with Hsp90 inhibition is shown as the shift in mean effect at each mutated position. Two positions, K150 and K152 (pale red), showed greatest sensitivity to both treatments. (F) K150I was tested in a quantitative assay to validate its temperature-sensitive and Hsp90-dependent mating activity, shown relative to wild-type (left boxplot in each pair) in each treatment (n = 20 for each sample). (G) Ste12 variants at K150 (mean effect) decreased mating (Left) and invasion (Right) at standard temperature (gray bars). At high temperature (red bars), mating further decreased; however, invasion increased. Error bars represent SEM. The K150I variant is individually highlighted for comparison.

I mutation K150A conferred decreased mating and an increase in binding preference of this domain such that it can be recruited in the invasion. The dominant invasion phenotype for each variant suggest absence of Tec1 to binding sites sufficient for invasion. that these Ste12 variants engage in altered binding to DNA, altered To understand the relationship between the phenotypes con- homodimerization, or altered heterodimerization with Tec1. ferred by Ste12 variants and their DNA-binding specificity, we conducted HT-SELEX experiments. We chose the SKQR and Mutations in Ste12 Can Promote Tec1-Independent Invasion and Alter K150A variants, each of which showed an altered invasion pheno- DNA-Binding Specificity. Tec1 activates invasion genes, with Ste12 ei- type, and asked whether they could recognize the homodimeric ther binding directly to DNA cooperatively with Tec1 or indirectly as Ste12 sites favored by the wild-type protein. SKQR showed a greatly part of a complex with Tec1 (14, 15). However, in contrast to both reduced (by 45%) preference for two sites separated by 3 bp (Fig. established Ste12 binding modes, the invasion phenotype due to the 5B). Consistent with this reduced preference, SKQR conferred re- SKQR variant was independent of Tec1, as was the even stronger duced activation in vivo from a minimal promoter containing such invasion phenotype due to the K150A variant (Fig. 5A). Thus, a homodimeric sites (SI Appendix,Fig.S12A). Furthermore, an ex- single mutation in the Ste12 DNA-binding domain might change the amination of the most enriched sequences bound in vitro by the

E8002 | www.pnas.org/cgi/doi/10.1073/pnas.1805882115 Dorrity et al. Downloaded by guest on September 25, 2021 AB TEC1+ Ste12 K150A Ste12 SKQR 1.0 0.5 spacing

wt Ste12 equency of 0.0 orientation 0510 15 0510 15 homodimeric spacing homodimeric spacing fr S177K,Q180R 1.0 3bp 3bp (SKQR) 1bp 0. 0.8 1.0 equal preference for K150A 3bp dimer over others 6 in Ste12 SKQR K175E in Ste12 K150A dimer organization = preference (% of max) preference (% of max) most preferred (max) = 0.0.0 0.2 0. 0.4 0. 0. 0. 0. 0. 0. 0. 0. 1.0 0.0 0.2 0.4 0.6 0.8 1.0 0.00. 0.20. 0.40. 0.60. 0.80. 1.0 0.0.0 0.0.2 0.0.4 0.0.6 0.0.8 1.0 preference (% of max) preference (% of max) in Ste12 wild-type in Ste12 wild-type Ste12 wild-type CDgood match 3bp good match

Ste12 + Tec1 2.0

1.0 bits G 1.0 G A G C A A TT T A G C G AC CAT T TTA A T T 0510 15 0.0 0.5 Ste12 K150A +Tec1 spacing 0.0 K150A poor match requency of requency f 2.0 0510 15 1.0 Ste12 SKQR +Tec1 bits C TGAAAC GG C GG T A GG G AA T T G A T GT TCT ACC T CCG C T ACTT CC T 0.0 A A

SKQR GENETICS

0510 15 2.0 heterodimer spacing 1.0 bits A TG AC AA G C G A A T A T G T T T T T T CC A G C A G T G T GC C G G 0.0 T AT C T CCG

Fig. 5. Ste12 DNA-binding domain mutants that confer invasion independent of the cofactor Tec1 have altered binding specificity. (A) Wild-type Ste12 requires the cofactor Tec1 for invasion; a tec1Δ strain fails to invade. The region I mutant K150A showed a hyperinvasive phenotype with or without Tec1. Introduction of the two positively charged residues found in C. gattii (SKQR) also led to Tec1-independent invasion, whereas introduction of a negative charge (K174E) eliminated invasion. The hyperinvasive Ste12 mutants showed a dominant invasion phenotype, as they were tested in the presence of wild- type Ste12, in the Σ1278b background. (B) Spacing heatmaps for Ste12’s binding site are shown as in Fig. 2A. K150A retains the same preference for tail-to-tail sites with a 3-bp spacer as the wild-type Ste12, but SKQR’s preference is reduced. To visualize this difference, the scatterplots below show frequencies of site organizations relative to the most preferred site (% of max) for each variant (y axis) compared with wild-type Ste12 (x axis), where each point represents one orientation and spacing combination (example of SKQR’s reduced preference is shown: its preference for head-to-tail with 1-bp spacing is ∼78% that of the tail-to-tail with 3-bp spacing). (C) Spacing heatmaps are shown for the K150A and SKQR variants in cobinding reactions with Tec1; unlike wild-type Ste12 (repeated from Fig. 2A, Bottom for clarity), neither variant shows strong preference for any heterodimeric site organization. (D) Logo plots generated from the 50 most highly enriched sequences in HT-SELEX represent ideal sites for each variant alongside wild-type Ste12.

SKQR variant showed that it lost specificity for the sequence of the We also carried out cobinding SELEX experiments with Tec1, second site, which was also reflected in its in vivo sequence pref- as previously conducted with wild-type Ste12 (reproduced in erence (Fig. 5D and SI Appendix,Fig.S12B). In contrast, K150A Fig. 5C, Top). Consistent with their ability to promote Tec1- retained a strong preference for the sites spaced 3-bp apart, nearly independent invasion, both SKQR and K150A showed altered identical to wild-type Ste12 (Fig. 5B). However, analysis of its patterns of cooperative binding with Tec1. K150A showed no most enriched sequences revealed that, like SKQR, the K150A preference for heterodimeric sites with a 2-bp spacer, favored by variant showed reduced specificity for one site of the homodimeric the wild-type Ste12, in cobinding experiments with Tec1, and pair (Fig. 5D). The reduced specificity in these variants should SKQR had a greatly reduced preference for these sites (Fig. 5C). expand the number of their recognition sites. Because both of these variants recognized at least one perfect—or near perfect— Discussion Ste12 site, we conclude that the primary change to DNA-binding Transcription factors maintain their capacity to enact complex specificity in these variants derives not from loss of specific DNA regulatory programs over the course of evolution, despite many contacts, but rather from their diminished capacity for symmetric changes to the genomic locations of their binding sites and to the dimeric binding. repertoires of their available cofactors. Nonetheless, by analyzing

Dorrity et al. PNAS | vol. 115 | no. 34 | E8003 Downloaded by guest on September 25, 2021 thousands of mutations in a yeast transcription factor, we found equivalent environmental stress results in a protein gaining a novel variants that can mediate only one of two alternative programs, function (34–39). and do so in the absence of a requisite cofactor. These variants of As the ultimate mediators of cellular decisions, transcription the S. cerevisiae Ste12 protein have single amino acid changes that factors occupy a predominant position in the signaling networks lead to DNA-binding preferences dramatically different from the that drive patterns of gene expression. For this reason, they are wild-type protein, as well to an increased ability of yeast to invade, susceptible to even subtle mutations that may result in conse- rather than mate, at high temperature. They thus reveal how quential phenotypes. Such effects will be less predictable than the natural variation in the DNA-binding domain of a transcription full loss-of-function phenotype resulting from, for example, a fail- factor might affect an environmentally sensitive decision to carry ure to bind to DNA. Thus, the type of detailed analysis of missense out one developmental process rather than another. variants as performed here will be useful in understanding the In S. cerevisiae, Ste12 drives expression of mating genes as a shifts in phenotype that enable adaptation to novel environments. homodimer and of invasion genes as a heterodimer with the in- Mutations in transcription factors that generate alternative vasion cofactor Tec1 (8, 18). The single mutations in Ste12 that DNA-binding modes have been previously identified. For exam- shift trait preference toward invasion in a dominant and Tec1- ple, rare coding variants in the homeodomain recognition helices independent manner provide a plausible model (Fig. 6) that can of several transcription factors alter DNA-binding specificity, explain how fungal species, including many pathogens, without which likely contributes to associated disease (40). These home- Tec1 or similar cofactors accomplish activation of invasion genes. odomain variants bind at sites not dramatically different from the In mating genes, two properly oriented binding sites facilitate the wild-type sites, yet this promiscuity is associated with a deleterious interaction of two Ste12 proteins; one site tends to be perfect and disease phenotype. The demonstration here of a surprising mal- the other slightly mismatched, thereby likely reducing coopera- leability of transcription factor binding stands in stark contrast to tivity. Under invasion conditions, the Ste12 dimer interface instead allows cooperative interaction between Ste12 and Tec1. We posit the fact that sequences of transcription factors tend to be highly that mutations in Ste12 that lead to hyperinvasiveness change this conserved. This study, together with those revealing only weak interface such that the conformation of the homodimer is altered purifying selection in functional regulatory regions, poses the and interaction with Tec1 is not possible. In this altered confor- challenge of reconciling flexible transcription factor activity and mation, which may be naturally present in species without Tec1, often highly variable regulatory regions with the longstanding Ste12 loses its preference for the perfect/near-perfect homodimeric observation that gene-expression patterns are robust to most binding sites and instead recognizes a degenerate version of its perturbations and conserved throughout evolution. binding site adjacent to a near-perfect site. The binding mode of Materials and Methods these Ste12 mutants exemplifies how transcription factors may – – access novel binding sites by changes in their recognition of pre- HT-SELEX. We purified fragments of Ste12 (1 215) and Tec1 (1 250) expressed ferred dimeric sites rather than of individual site specificity. from pGEX-4T-2 vectors. These protein fragments have been used previously Mutations at positions in Ste12 implicated in shifting trait and are sufficient for both individual and cooperative binding in vitro. Fragments of each protein, as well as protein variants, were purified using a preference toward invasion also conferred dependence on tem- GST tag and subsequently used for HT-SELEX. SELEX reactions with ho- perature and the chaperone Hsp90. For some fungal pathogens, mogenous and mixed protein populations were performed identically to increased temperatures associated with warm-blooded animals previous work (1, 21). Briefly, a 50-μL reaction containing purified Ste12 and promote invasive growth (7). In the nonpathogenic S. cerevisiae, Tec1 (1:25 molar ratio with DNA), 200-ng nonspecific competitor double- increased temperature decreases mating and promotes invasion stranded nucleic acid poly (dI/dC), and 100-ng selection ligand (36N) was

(7). The conserved role of Ste12 in invasion and mating, which incubated in binding buffer [140 mM KCl, 5 mM NaCl, 1 mM K2HPO4,2mM μ μ can be modified by mutations that affect the environmental MgSO4, 20 mM Hepes (pH 7.05), 100 M EGTA, 1 M ZnSO4]for2h.GST sensitivity of this trait choice, suggests that variation in Ste12 may Sepharose (GE) beads were then added to each reaction, incubated for potentiate pathogenic transformation in other species. The tem- 30 min, and unbound ligand was removed using seven buffer washes. perature- and Hsp90-dependent Ste12 variants do not resemble Output reactions were amplified by PCR after each round, and these prod- typical temperature-sensitive mutants in simply losing function ucts were subsequently used to prepare high-throughput sequencing li- under nonpermissive conditions. In response to high temperature or braries. SELEX motif enrichments were analyzed using Autoseed software Hsp90 inhibition, these variants failed to promote mating yet con- (1). Briefly, a pool of binding-selected output sequences was compared against a fully random input sequence to identify sequences, motifs, and ferred an increased ability to invade. These variants thus repre- orientations enriched relative to the unselected oligo pool. sent the unusual phenomenon whereby perturbation of Hsp90 or Generation of STE12 Mutant Libraries. The STE12 locus from S. cerevisiae strain BY4741, including the intergenic regions, was introduced into the mating gene invasion gene yeast vector pRS415 containing a LEU2 marker (41). Degenerate DNA sequence encoding a 33-amino acid (99 bp) segment of the DNA-binding do- Ste12 main of Ste12 was generated by 2.5% doped oligonucleotide synthesis Ste12 wild-type (Trilink Biotechnologies). Invariant 30-bp sequences were designed on either side of the mutagenized fragment; these flanking sequences contained NotI and ApaI cut sites found in the coding sequence of STE12, and were unique in the STE12 plasmid construct. Both fragment and plasmid were double- K150A mutant mismatch digested with NotI and ApaI, and the plasmid library was assembled by standard ligation. STE12 libraries were transformed into electrocompetent Fig. 6. A model for novel site recognition by Tec1-independent Ste12 vari- Escherichia coli (ElectroMAX DH10B; Invitrogen), and amplified overnight in ants. Wild-type Ste12 protein (blue) is compared with the K150A mutant (tan) selective media. Efficiency of ligation was verified by Sanger sequencing for binding at a mating gene with two perfect recognition sites (tail-to-tail across the mutagenized region of 96 transformants (42); no assembly errors with a 3-bp spacing, shown in blue) and an invasion gene with a single perfect were detected, and mutant proportions reflected those expected for doped site along with a degenerate site. Due to its altered dimer interface, the oligo synthesis at 2.5%. Plasmid libraries were used to transform yeast “kinked” mutant K150A is less capable of binding symmetrically at two perfect (BY4741 MATa or Σ1278b MATα) with a deleted endogenous copy of STE12 sites, consistent with its phenotype of reduced mating efficiency. However, the by high-efficiency lithium acetate transformation (43). The same plasmid conformation of K150A allows the variant protein to occupy novel sites within library was used to transform yeast (Σ1278b) for invasion selection. Indi- invasion genes that contain one perfect or nearly perfect site paired with a vidual point mutations were generated in wild-type STE12 plasmids by site- degenerate site; wild-type Ste12 is unable to occupy these sites. directed mutagenesis (Q5; New England Biolabs).

E8004 | www.pnas.org/cgi/doi/10.1073/pnas.1805882115 Dorrity et al. Downloaded by guest on September 25, 2021 Large-Scale Trait Selections. The BY4742 MATα strain was used as the mating score indicates that the partner mutation decreases the deleterious effects of partner for the library-transformed BY4741 MATa in all selections and mating individual mutations (47). assays. Transformed yeast cells were grown to late log-phase in a single 500-mL culture, and cells were harvested to determine plasmid variant fre- Quantitative Mating Assay. Individual variants tested for mating efficiency quencies in the input population. The same culture was used to seed 36 in- were treated identically to the large-scale mating selection experiments, dependent mating selections for each treatment: 1 million MATa cells with except genotypes were scored individually on selective plates for either STE12 variants were mixed with 10-fold excess wild-type MATα cells and mated diploids (2N) or both diploids and unmated MATa haploids (2N, 1N). allowed 5 h to mate (44). Depending on treatment type, cell mixtures were left The proportion of mated individuals (mating efficiency) is taken as the ratio at 30 °C with DMSO, 30 °C with the Hsp90 inhibitor radicicol, or 37 °C with of colony counts on diploid to counts on diploid + haploid plates [2N/(2N + DMSO. A 5-μM concentration of the Hsp90 inhibitor radicicol (R2146; Sigma- 1N)]. Ste12 variants were tested alongside wild-type to determine relative Aldrich) was chosen due to its measureable effect on mating efficiency and changes in response to treatment. lack of pleiotropic growth defects. The temperature of 37 °C was chosen for the similarity of effects on mating between temperature and radicicol treat- STE12 Variant RNA-Seq. RNA was extracted from yeast cells harboring the ments. After mating was completed, cell mixtures were plated using auxo- STE12 mutant library grown under nonselective conditions using acid phenol trophic markers present only in mated diploids. Plasmids containing STE12 extraction, as previously described (48). STE12-specific cDNA was created variants were extracted from this output population, as well as from the using a gene-specific cDNA primer and SuperScript III (Life Technologies). premating input, for subsequent deep sequencing. Mating selections were cDNA was amplified in a manner similar to the plasmids, and prepared for Σ repeated in triplicate. For invasion, 1278b yeast cells transformed with the sequencing using Illumina Nextseq. Ste12 plasmid library were grown to late log-phase in a single 500-mL culture, diluted (10,000 cells per plate), and plated onto 40 plates of synthetic complete Large-Scale Analysis of Ste12 Binding Sites in Vivo. A HIS3 reporter gene was medium lacking leucine (2% agar). Plates were incubated at 30 °C or 37 °C for used for testing large populations of binding-site variants (SI Appendix,Fig. 72 h to allow for sufficient invasion, as previously described (30). After 3 d, cells S12) in media lacking histidine. Although Ste12 does bind single PREs as a were washed from the plate surfaces, enriching for cells embedded in the monomer, two sites are needed for signal detection in reporter assays (49). We “ ” agar. Agar pucks were removed from plates with a razor. Using the salsa designed oligonucleotides that maintain the central portion of the native PRE, blender setting (Hamilton Beach), a coarse slurry was generated and sub- but randomized six surrounding bases on either side (NNNNNNTTTCAAAAT- sequently poured over a vacuum apparatus lined with cheesecloth. The GAAANNNNNN). This library was cloned into a promoter from the same resulting liquid cell suspension was spun down at 5,000 rpm to collect cells for plasmid used in the luciferase assay, at the same position as the native PRE. subsequent deep sequencing. Individual invasion assays (as in Fig. 5A)were Strains containing one of four Ste12 protein variants were transformed with μ treated identically, but 10- L aliquots of OD-normalized cultures were plated the same binding-site library reporter population, and grown overnight in to image colonies for each strain. For high osmolarity growth, selections were synthetic media lacking histidine with 10 mM of the His3 competitive inhibitor conducted using the BY4741 MATa library-transformed population grown 3-amino-triazole. Three biological replicate selections were conducted for each overnight in media containing 1.5 M Sorbitol, as described previously (45). Ste12 protein variant tested against the binding-site library. Cells were col- Populations were sequenced before and after growth to determine enrich- lected and sequenced before and after selections to determine enrichment ment scores that defined the 95% confidence interval used in Fig. 3C.Forall scores for each binding-site variant. Computational analysis of binding-site trait selections, we chose sample sizes that were at least 10-fold higher than enrichment scores was identical to pipeline for Ste12 protein variants. En- the variant library size, ensuring that each variant would be adequately richment of all binding-site variants are shown relative to the enrichments of sampled in each of the three biological replicate selections. empty plasmids, which were spiked-in to the binding-site library as control. The 250 binding sites with the highest enrichments were grouped and GENETICS Sequencing and Determination of Trait Scores. Sequencing was completed on Weblogo (50) was used to generate base preference plots. Illumina’s MiSeq or NextSeq platforms. Sequencing libraries were prepared by extracting plasmids from yeast populations (Yeast Plasmid Miniprep II; Evolutionary Analysis of Ste12 DNA-Binding Domain Among Fungi. Using fungal Zymo Research) before and after selection. These plasmids were used as genomes deposited into the National Center for Biotechnology Information template for PCR amplification that added adaptor sequences and 8-bp and from the fungal sequencing project at the Joint Genome Institute, we sample indexes to the 99-bp mutagenized region for sequencing (all li- created a BLAST database of translated coding sequences and queried with the braries amplified <15 cycles). Paired-end reads spanning the mutagenized S. cerevisiae Ste12 protein sequence. Full protein sequences from BLAST hits region were filtered to obtain a median of 5 million reads per sample. Using were aligned using MUSCLE (51) and used to determine conservation at each ENRICH software (46), read counts for each variant before and after selec- site in the mutated segment. The secondary structure of S. cerevisiae’s Ste12 tion were used to determine mating efficiency and invasion ability of STE12 DNA-binding domain was determined using Psipred (52). variants. Briefly, counts for a particular variant in the input and output libraries were normalized by their respective read totals to determine fre- Data Availability. High-throughput sequencing reads have been submitted to quency in each, and a ratio of the output and input frequencies determine a National Center for Biotechnology Information Sequence Read Archive under variant’s functional score. Finally, enrichment scores were normalized by the accession nos. SRR5408956–SRR5408965. Datasets with calculated enrichment enrichment of wild-type Ste12 in each selection. Treatment scores were scores for each tested variant in each condition, along with positional mean scores, calculated identically, and in all cases where difference scores are shown, the and scores from binding site library selections are provided in Datasets S1–S5. score represents log2(treated) − log2(untreated).

ACKNOWLEDGMENTS. We thank J. Thomas for advice on analysis of fungal Calculating Intramolecular Epistasis Scores. Intramolecular epistasis scores variation, and D. Fowler and H. Malik for comments on the manuscript. The ’ were defined as the deviation of a double mutant s functional score (Wij) work was supported by National Institutes of Health Grant R01GM114166 × from the multiplied scores of its constituent single-mutants (wi wj). A (to C.Q. and S.F.). M.W.D. was supported by a National Science Foundation negative epistasis score indicates that the deleterious effect of one mutation Graduate Research Fellowship and a Washington Research Foundation-Hall is increased by the presence of the partner mutation, while a positive epistasis fellowship. S.F. is an investigator of the Howard Hughes Medical Institute.

1. Jolma A, et al. (2015) DNA-dependent formation of transcription factor pairs alters 7. Perfect JR (2006) Cryptococcus neoformans: The yeast that likes it hot. FEMS Yeast Res their binding specificity. Nature 527:384–388. 6:463–468. 2. Boer DR, et al. (2014) Structural basis for DNA binding specificity by the auxin- 8. Madhani HD, Fink GR (1998) The control of filamentous differentiation and virulence dependent ARF transcription factors. Cell 156:577–589. in fungi. Trends Cell Biol 8:348–353. 3. Hentze N, Le Breton L, Wiesner J, Kempf G, Mayer MP (2016) Molecular mechanism of 9. Louvion JF, Abbas-Terki T, Picard D (1998) Hsp90 is required for pheromone signaling thermosensory function of human heat shock transcription factor Hsf1. eLife 5: in yeast. Mol Biol Cell 9:3071–3083. e11576. 10. Zalatan JG, Coyle SM, Rajan S, Sidhu SS, Lim WA (2012) Conformational control of the 4. Mody A, Weiner J, Ramanathan S (2009) Modularity of MAP kinases allows de- Ste5 scaffold protein insulates against MAP kinase misactivation. Science 337: formation of their signalling pathways. Nat Cell Biol 11:484–491. 1218–1222. 5. Cook JG, Bardwell L, Thorner J (1997) Inhibitory and activating functions for MAPK 11. Dolan JW, Kirkman C, Fields S (1989) The yeast STE12 protein binds to the DNA se- Kss1 in the S. cerevisiae filamentous-growth signalling pathway. Nature 390:85–88. quence mediating pheromone induction. Proc Natl Acad Sci USA 86:5703–5707. 6. Beyhan S, Gutierrez M, Voorhies M, Sil A (2013) A temperature-responsive network 12. Errede B, Ammerer G (1989) STE12, a protein involved in cell-type-specific transcrip- links cell shape and virulence traits in a primary fungal pathogen. PLoS Biol 11: tion and signal transduction in yeast, is part of protein-DNA complexes. Genes Dev 3: e1001614. 1349–1361.

Dorrity et al. PNAS | vol. 115 | no. 34 | E8005 Downloaded by guest on September 25, 2021 13. Yuan YO, Stroke IL, Fields S (1993) Coupling of cell identity to signal response in yeast: 34. Sangster TA, Lindquist S, Queitsch C (2004) Under cover: Causes, effects and impli- Interaction between the alpha 1 and STE12 proteins. Genes Dev 7:1584–1597. cations of Hsp90-mediated genetic capacitance. BioEssays 26:348–362. 14. Zeitlinger J, et al. (2003) Program-specific distribution of a transcription factor de- 35. Jarosz DF, Taipale M, Lindquist S (2010) Protein homeostasis and the phenotypic pendent on partner transcription factor and MAPK signaling. Cell 113:395–404. manifestation of genetic diversity: Principles and mechanisms. Annu Rev Genet 44: 15. Chou S, Lane S, Liu H (2006) Regulation of mating and filamentation genes by two 189–216. – distinct Ste12 complexes in Saccharomyces cerevisiae. Mol Cell Biol 26:4794 4805. 36. Jarosz DF, Lindquist S (2010) Hsp90 and environmental stress transform the adaptive 16. Dong H, Courchesne W (1998) A novel quantitative mating assay for the fungal value of natural genetic variation. Science 330:1820–1824. pathogen Cryptococcus neoformans provides insight into signalling pathways re- 37. Sangster TA, et al. (2008) HSP90 affects the expression of genetic variation and de- sponding to nutrients and temperature. Microbiology 144:1691–1697. velopmental stability in quantitative traits. Proc Natl Acad Sci USA 105:2963–2968. 17. Wong Sak Hoi J, Dumas B (2010) Ste12 and Ste12-like proteins, fungal transcription 38. Karras GI, et al. (2017) HSP90 shapes the consequences of human genetic variation. factors regulating development and pathogenicity. Eukaryot Cell 9:480–485. Cell 168:856–866.e12. 18. Madhani HD, Fink GR (1997) Combinatorial control required for the specificity of 39. Sahni N, et al. (2015) Widespread macromolecular interaction perturbations in human yeast MAPK signaling. Science 275:1314–1317. genetic disorders. Cell 161:647–660. 19. Boer VM, de Winde JH, Pronk JT, Piper MDW (2003) The genome-wide transcriptional 40. Barrera LA, et al. (2016) Survey of variation in human transcription factors reveals responses of Saccharomyces cerevisiae grown on glucose in aerobic chemostat cul- – tures limited for carbon, nitrogen, phosphorus, or sulfur. J Biol Chem 278:3265–3274. prevalent DNA binding changes. Science 351:1450 1454. 20. Prinz S, et al. (2004) Control of yeast filamentous-form growth by modules in an in- 41. Mumberg D, Müller R, Funk M (1995) Yeast vectors for the controlled expression of – tegrated molecular network. Genome Res 14:380–390. heterologous proteins in different genetic backgrounds. Gene 156:119 122. 21. Jolma A, et al. (2010) Multiplexed massively parallel SELEX for characterization of 42. Sanger F, Nicklen S, Coulson AR (1977) DNA sequencing with chain-terminating in- human transcription factor binding specificities. Genome Res 20:861–873. hibitors. Proc Natl Acad Sci USA 74:5463–5467. 22. Jolma A, et al. (2013) DNA-binding specificities of human transcription factors. Cell 43. Gietz RD, Woods RA (2002) Transformation of yeast by lithium acetate/single- 152:327–339. stranded Carrier DNA/polyethylene glycol method. Methods Enzymol 350:87–96. 23. Gordân R, et al. (2011) Curated collection of yeast transcription factor DNA binding 44. Leu JY, Murray AW (2006) Experimental evolution of mating discrimination in bud- specificity data reveals novel structural and gene regulatory insights. Genome Biol 12: ding yeast. Curr Biol 16:280–286. R125. 45. Posas F, Witten EA, Saito H (1998) Requirement of STE50 for osmostress-induced ac- 24. Yuan YL, Fields S (1991) Properties of the DNA-binding domain of the Saccharomyces tivation of the STE11 mitogen-activated protein kinase kinase kinase in the high- cerevisiae STE12 protein. Mol Cell Biol 11:5910–5918. osmolarity glycerol response pathway. Mol Cell Biol 18:5788–5796. 25. Roberts CJ, et al. (2000) Signaling and circuitry of multiple MAPK pathways revealed 46. Rubin AF, et al. (2016) Enrich2: A statistical framework for analyzing deep mutational – by a matrix of global gene expression profiles. Science 287:873 880. scanning data. bioRxiv, 10.1101/075150. 26. Heise B, et al. (2010) The TEA transcription factor Tec1 confers promoter-specific gene 47. Araya CL, et al. (2012) A fundamental protein property, thermodynamic stability, regulation by Ste12-dependent and -independent mechanisms. Eukaryot Cell 9: revealed solely from large-scale measurements of protein function. Proc Natl Acad Sci 514–531. USA 109:16858–16863. 27. Olson KA, et al. (2000) Two regulators of Ste12p inhibit pheromone-responsive 48. Cuperus JT, Lo RS, Shumaker L, Proctor J, Fields S (2015) A tetO toolkit to alter ex- transcription by separate mechanisms. Mol Cell Biol 20:4199–4209. pression of genes in Saccharomyces cerevisiae. ACS Synth Biol 4:842–852. 28. Fowler DM, Fields S (2014) Deep mutational scanning: A new style of protein science. 49. Su TC, Tamarkina E, Sadowski I (2010) Organizational constraints on Ste12 cis- Nat Methods 11:801–807. elements for a pheromone response in Saccharomyces cerevisiae. FEBS J 277: 29. Winzeler EA, et al. (1999) Functional characterization of the S. cerevisiae genome by – gene deletion and parallel analysis. Science 285:901–906. 3235 3248. 30. Ryan O, et al. (2012) Global gene deletion analysis exploring yeast filamentous 50. Crooks GE, Hon G, Chandonia JM, Brenner SE (2004) WebLogo: A sequence logo – growth. Science 337:1353–1356. generator. Genome Res 14:1188 1190. 31. Shoval O, et al. (2012) Evolutionary trade-offs, Pareto optimality, and the geometry of 51. Edgar RC (2004) MUSCLE: Multiple sequence alignment with high accuracy and high phenotype space. Science 336:1157–1160. throughput. Nucleic Acids Res 32:1792–1797. 32. Jones SK, Jr, Bennett RJ (2011) Fungal mating pheromones: Choreographing the 52. McGuffin LJ, Bryson K, Jones DT (2000) The PSIPRED protein structure prediction dating game. Fungal Genet Biol 48:668–676. server. Bioinformatics 16:404–405. 33. Fraser JA, et al. (2005) Same-sex mating and the origin of the Vancouver Island 53. Grant CE, Bailey TL, Noble WS (2011) FIMO: Scanning for occurrences of a given motif. Cryptococcus gattii outbreak. Nature 437:1360–1364. Bioinformatics 27:1017–1018.

E8006 | www.pnas.org/cgi/doi/10.1073/pnas.1805882115 Dorrity et al. Downloaded by guest on September 25, 2021