Cis- and trans-regulatory factors contributing to divergent activity of the TDH3 promoter in Saccharomyces yeast

Abstract

Changes in regulatory sequences controlling the timing and activity of gene products underlie much of natural phenotypic variation. Yet, identifying which divergent sites matter and how they impact gene expression remains challenging. Here, we investigate how transcriptional activity and homeostatic responsiveness of orthologous promoters of the metabolic gene TDH3 evolved among Saccharomyces yeast. We found that promoter expression level increased specifically in the Saccharomyces cerevisiae lineage and that a substantial part of this increase was caused by genetic variants located between the well-characterized, conserved binding sites for two direct transcriptional regulators. These nucleotide changes altered the promoters’ expression levels while leaving the expression dynamics conserved. Further, the effects of these nucleotide changes were only seen in the presence of a third transcription factor, TYE7p, which is known to be recruited by the other transcription factors through protein–protein interactions. These results suggest that the cis-regulatory changes act through their influence on the collective assembly/activation of a transcription factor complex and that changes acting through such a mechanism can allow distinct parts of gene expression, such as expression level and dynamics, to separately diverge.

Main text

Molecular processes that regulate the expression of genes in response to developmental and environmental cues are critical for determining the relationship between genetic and phenotypic variation. Genetic variants that affect gene regulatory mechanisms have become increasingly appreciated as the cause of phenotypic variation among populations and between species. This evolutionary importance of variation in gene regulation was presciently hypothesized 50 years ago (Britten and Davidson 1969; King and Wilson 1975), and investigations of phenotypic variation at various levels, from metabolism to morphology, have strengthened this hypothesis in the decades since (Shubin et al. 1997; Stern and Orgogozo 2008; Blount et al. 2012; Martin and Orgogozo 2013; Coyle and King 2025). Understanding the molecular and evolutionary processes that cause regulatory sequences to be the tools of evolutionary tinkering and the mechanisms that govern the evolution of regulatory sequences remains a pressing challenge for the field.

Expression of each gene is controlled by interactions between cis- and trans-regulatory factors, both of which contribute to the evolution of gene expression (Hill et al. 2021). Cis-regulatory factors include DNA sequences (eg promoters, enhancers, and untranslated regions (UTRs)) that affect the expression at a particular locus, and trans-regulatory factors include diffusible molecules (eg transcription factors and regulatory RNAs) present in a cell, which often regulate the expression of multiple genes from many loci. Despite progress in understanding the function of cis-regulatory sequences and predicting the trans-acting transcription factors that interact with them (de Boer et al. 2020; Avsec et al. 2021; Vaishnav et al. 2022; Mahendrawada et al. 2025; Xie et al. 2025), our ability to predict which genetic changes contribute to an evolutionary change in gene expression remains poor (Nora et al. 2023). Most often, mutations in transcription factor binding sites are assumed to be responsible for cis-regulatory divergence (Wray 2007), perhaps because mutations in transcription factor binding sites tend to have large effects on gene expression. Mutations with large effects, however, might not be the most likely to fix during evolution (Rockman 2012; Umans et al. 2021). To better understand the mechanisms of cis-regulatory divergence, specific nucleotide changes and the molecular causes of their contribution to differences in cis-regulatory activity between species need to be identified for more genes.

In the baker's yeast Saccharomyces cerevisiae, the TDH3 promoter has been used as a model system for understanding how genetic changes in both cis- and trans-acting sequences can introduce variation in gene expression and how selection shapes that variation within species (reviewed in Wittkopp 2023). The TDH3 promoter controls transcription of the most highly expressed of the three paralogous genes in S. cerevisiae encoding the core metabolic protein glyceraldehyde-3-phosphate dehydrogenase (McAlister and Holland 1985). It is regulated by four transcription factors: Rap1p, Gcr1p, Gcr2p, and Tye7p (Uemura and Jigami 1992; Deminoff and Santangelo 2001; Holland et al. 2019; Shively et al. 2019; Bergenholm et al. 2021). Rap1p and Gcr1p cooperatively recognize sequence elements in the promoter (Tornow et al. 1993; Deminoff and Santangelo 2001), Gcr2p forms a heteromeric complex with Gcr1p (Uemura and Jigami 1992; Deminoff and Santangelo 2001), and Tye7p is thought to be recruited through protein–protein interactions with the other transcription factors (Holland et al. 2019; Shively et al. 2019). Quantitative variation in TDH3 activity directly affects fitness during fermentative and respirative growth (Duveau et al. 2017a; Siddiq et al. 2024). Further, the activity of the S. cerevisiae TDH3 promoter is dynamically sensitive to internal and external cellular conditions: Promoter expression changes based on glucose availability during culture growth, and loss of TDH3 activity triggers a homeostatic feedback mechanism mediated by upregulation of Gcr1p (Vande Zande et al. 2023). Studies of mutations and polymorphisms in S. cerevisiae have demonstrated that the activity of the TDH3 promoter is strongly constrained, with mutations to the RAP1p and GCR1p binding sites being particularly costly (Metzger et al. 2015, 2016; Duveau et al. 2017a, 2017b, 2018, 2021; Metzger and Wittkopp 2019; Vande Zande et al. 2022; Siddiq et al. 2024). Yet, despite this constraint, comparative RNA-seq analyses show that the activity level of the TDH3 promoter has diverged through cis-regulatory differences between S. cerevisiae and its sister species Saccharomyces paradoxus (Krieger et al. 2020). We therefore sought to uncover the genetic changes through which the orthologous TDH3 promoters’ activity diverged and, in turn, better understand the mechanistic causes of the evolution of gene expression.

The genetic changes responsible for the divergent activity of the TDH3 promoter between S. cerevisiae and S. paradoxus arose since the species last shared a common ancestor, approximately 5 to 10 million years ago (Replansky et al. 2008). To infer the most likely direction of evolutionary change between these two species, we examined the activity of the TDH3 promoter from Saccharomyces mikatae and Saccharomyces kudriavzevii, which last shared a common ancestor with S. cerevisiae and S. paradoxus 10 to 15 and 15 to 20 million years ago, respectively (Replansky et al. 2008). We used each species’ promoter to drive expression of a yellow fluorescent protein (YFP) in a reporter gene that was integrated at the HO locus of the S. cerevisiae genome (Fig. 1a). These orthologous promoters ranged from 668 to 678 bp long and ranged from 73% to 85% sequence identity (Fig. 1b). YFP fluorescence was used as a proxy for TDH3 promoter activity, and fluorescence was measured using flow cytometry in at least 45,000 cells in each of three replicate populations for each of the four reporter genes in populations of cells grown in rich yeast extract peptone dextrose media (YPD). Cultures were grown at 30 °C, and yeast cells were sampled at the end of the fermentative growth stage (∼18 to 21 h). After correcting each individual cell for its estimated cell size, we took the median fluorescence of each sample, averaged these medians among the replicates of the same strain, and compared the expression levels driven by the orthologous promoters using a linear mixed model with a random effect to account for day-to-day variation. We found that the S. cerevisiae TDH3 promoter drove a significantly higher level of expression than the other three species, which all drove similar levels of expression (Fig. 1c; Table S1a). The ∼15% higher expression in S. cerevisiae was likely driven by selection, given that changes in TDH3 promoter activity of this magnitude cause measurable changes in fitness in S. cerevisiae (Duveau et al. 2017a; Siddiq et al. 2024) and none of the polymorphisms segregating among 85 strains of S. cerevisiae reduced TDH3 promoter activity to the levels seen in these other species (Metzger et al. 2015).

Figure 1.

Divergent activity of the TDH3 promoter activity in S. cerevisiae. a) Orthologous TDH3 promoters from S. cerevisiae (S. cer), S. paradoxus (S. par), S. mikatae (S. mik), and S. kudriavzevii (S. kud) were fused to the venusYFP coding sequence (YFP) to make a series of reporter genes. Each reporter gene was inserted into the HO locus of a common S. cerevisiae reference strain. b) Pairwise nucleotide sequence identity is shown for the different species’ TDH3 promoters. c) Mean activity of orthologous promoters as determined by flow cytometry (YFP expression, normalized for cell size) is shown with error bars indicating the standard error of the mean. Dots show the mean values for individual replicates, each consisting of >40,000 cells and collected on a different day. Asterisks designate statistically significant differences among genotypes, as determined using a linear model to estimate the effects of each genotype with the Tukey honestly significant difference (HSD) method for post hoc pairwise comparisons (S. cer vs. S. par: t = 5.27, P = 0.0003; S. cer vs. S. mik: t = 4.27, P = 0.002; S. cer vs. S. kud: t = 4.01, P = 0.004; all other comparisons had P > 0.5). d) Activity of the orthologous TDH3 promoter alleles in genomic backgrounds with (reference) and without a functional TDH3 gene (ΔTDH3); the different promoters were upregulated similarly in response to the deletion of TDH3 (F = 1.1988; P = 0.3394). Points and error bars show estimated means and standard errors, respectively. e) YFP fluorescence driven by TDH3 promoter alleles in the reference (gray) and TDH3 mutant (red) backgrounds during growth for 24 h on liquid YPD. Data is shown for four to eight replicates of each of the eight genotypes.

To determine whether the sequence divergence among these promoters also affected the homeostatic feedback mechanism reported for the TDH3 promoter in S. cerevisiae (Vande Zande et al. 2023), we used the same procedure to measure expression driven by each promoter in a strain of S. cerevisiae with the native TDH3 gene deleted (TDH3::ΔTDH3). In all cases, we found that deletion of the TDH3 gene increased the activity of the TDH3 promoter, showing that this compensatory response was conserved among species (Fig. 1d). This conserved dynamic behavior of the TDH3 promoter was not confined to homeostatic regulation—we also measured the activity of the species-specific TDH3 promoter alleles in S. cerevisiae, with and without a functional TDH3 gene, over 24 h of growth when cells are undergoing a diauxic shift on YPD and found that all four promoters, in both genetic backgrounds, showed a similar pattern of expression change over time, even though the S. cerevisiae promoter retained the highest overall expression (Fig. 1e; Table S1b). These data show that derived sequence changes in the S. cerevisiae TDH3 promoter increased the promoter's activity level without altering its dynamic homeostatic and metabolic regulation. This finding is consistent with genomic comparisons of S. cerevisiae and S. paradoxus, showing that expression level often evolves independently of expression dynamics (Krieger et al. 2020; Shih and Fay 2021) and that divergence in expression level tends to be caused by cis-acting genetic changes.

To try to identify specific nucleotide changes contributing to higher activity of the S. cerevisiae TDH3 promoter, we focused on a region containing an upstream activation sequence responsible for TDH3 expression (Bitter et al. 1991). We first looked for changes in experimentally validated binding sites that recruit the Rap1p (Repressor-activator protein) and the heteromer of Gcr1p and Gcr2p (Glycolysis regulator 1 and 2; hereafter Gcr1p/2p) transcription factors (Bitter et al. 1991; Chambers et al. 1995; Metzger et al. 2015). These proteins cooperatively activate the transcription of TDH3 as well as other glycolysis genes (Mizuno et al. 2004). We found that these transcription factor binding sites were highly conserved in sequence and position among all four orthologous promoters (Fig. 2a), suggesting that they are not the source of the divergent activity in S. cerevisiae. However, we noticed several differences in the 14-base-pair region between the Gcr1p/2p and Rap1p binding sites (Fig. 2a). We suspected that these changes may be functionally consequential due to their physical proximity to the binding motifs, potentially modifying the efficacy of TF–DNA interactions. Small differences in spacing between the RAP1 and GCR1 binding sites and from this region to the start codon also exist among the four species surveyed (Fig. 2a), but none of these changes correlate with expression divergence. They could, however, have functional effects that are counterbalanced by compensatory changes elsewhere in the promoter.

Figure 2.

Sequence divergence between the Rap1p and Gcr1p binding sites affects activity of the TDH3 promoter. a) Sequence alignment shows a region of the TDH3 promoter from S. cerevisiae (S. cer), S. paradoxus (S. par), S. mikatae (S. mik), and S. kudriavzevii (S. kud) containing a previously characterized upstream activating sequence. This sequence includes binding sites for the transcription factors Rap1p and Gcr1p. High-affinity binding sites for both of these transcriptional regulators, shown in gray, are highly conserved across species. The position of this sequence region relative to the start codon is shown with subscripts. Asterisks highlight five positions between these binding sites that differentiate the S. cerevisiae allele from the S. paradoxus allele, and the parsimony-based inference of lineage on which the changes occurred are shown with dotted lines. The cladogram displays the timing of a one base-pair (bp) deletion, and the subsequent effect on intermotif spacing is displayed beside the alignment. b) The relative activity of the S. cerevisiae (S. cer) and S. paradoxus (S. par) TDH3 promoter alleles is shown alongside the activity of recombinant constructs in which the five divergent sites indicated with asterisks in (a) were swapped between the species-specific promoters (S.cer + par5 and S. par + cer5), as shown in the schematics. Each data point plotted represents a separate experimental replicate (>20,000 cells/measurement). P values displayed for pairwise differences were calculated from a linear model fit to the data with the Tukey HSD method for post hoc comparisons.

We tested whether changes in the region between the RAP1 and GCR1 binding sites contribute to divergence of the TDH3 promoter by swapping the five divergent sites in this 14 bp region between the S. cerevisiae and S. paradoxus TDH3 promoters and assaying activity of these chimeric promoters using the same YFP reporter gene inserted at the HO locus (Fig. 2b). We found that these five nucleotides did indeed have a significant effect on expression of the reporter genes: The S. paradoxus promoter with the S. cerevisiae alleles had higher expression than the wild-type S. paradoxus promoter (Fig. 2b; t = 3.97, P < 0.001, Tukey HSD), and the S. cerevisiae promoter with the S. paradoxus alleles had lower expression than the wild-type S. cerevisiae promoter (Fig. 2b; t = 2.91, P = 0.02, Tukey HSD). In neither case, however, was changing these five nucleotides sufficient to fully convert the expression level from one species to the other. When the S. cerevisiae nucleotides were introduced into the S. paradoxus promoters, they recovered 35% (95% CI: 22% to 41%) of the expression difference between the promoters. In the reciprocal experiment, the divergent nucleotides tested recovered 28% (95% CI: 14% to 41%) of the total expression difference. Note that the CIs of these two calculations overlap, suggesting that these divergent sites might have similar effects in both species’ promoters. The rest of the divergence in expression level between S. cerevisiae and S. paradoxus is presumably caused by one or more of the 87 other divergent sites between the promoters of these two species located outside this region.

The Rap1p and Gcr1p/2p transcription factors binding to sites that define the ends of this region form a complex and cooperatively regulate TDH3 expression (Mizuno et al. 2004). Consequently, sequence divergence in this intervening region might affect the formation and/or activity of this complex. The Tye7p transcription factor, a basic helix-loop-helix protein that binds to E-box motifs (Gordân et al 2013 ), also regulates TDH3 expression as a part of this complex (Holland et al. 2019; Liu et al. 2020; Bergenholm et al. 2021) (Fig. 3a). However, despite having a DNA recognition helix, Tye7p cannot directly bind the TDH3 promoter or other glycolytic promoters on its own (Shively et al. 2019). Instead, it is recruited to the promoter through interactions with Gcr2p in the Gcr1p/2p heteromer, after which it can make direct contacts with DNA as a part of this complex (Liu et al. 2020). Consistent with this observation, the 14 bp region does not contain any matches (strictly or more relaxed) to the E-box binding motif (CACGTG). Based on this information, we hypothesized that five sequence differences between the Rap1p and Gcr1p/2p binding sites might alter activity of the TDH3 promoter by altering the formation or activity of this complex—perhaps through changing the interactions of Rap1p or Gcr1p/2p with Tye7p.

Figure 3.

Effects of divergent sites depend on Tye7p. a) Schematic shows a model of Tye7p recruitment by the Gcr1p/2p complex and Rap1p, as described in Shively et al. (2019). Binding motifs for Gcr1p and Rap1p are necessary for proper TDH3 promoter activation; Tye7p requires no specific binding motif to localize to the TDH3 promoter. However, we hypothesize that certain DNA sequences at the site of Tye7p localization might better stabilize the collective complex and lead to greater transcriptional activation. b) The relative activity of the S. cerevisiae (S.cer), S. paradoxus (S.par), and two chimeric TDH3 promoter alleles (S.cer + par5 and S. par + cer5) is shown in the presence (Tye7) and absence (Tye7 Deletion) of Tye7p. Opaque points and error bars display estimated mean values for each genotype with 95% CIs; transparent points represent observations from individual replicates. Density distributions are also shown, describing all replicates for each promoter allele. c to e) Same data as (b) but highlighting the Tye7-dependent effects of changing the five nucleotides between the Rap1p and Gcr1p binding sites in the S. paradoxus TDH3 promoter (c), the S. cerevisiae TDH3 promoter (d), and the overall effect of the combined 92 differences between the S. cerevisiae and S. paradoxus promoters (e). Statistical significance of pairwise comparisons between promoters is shown (values in Table S2). P values were calculated from a linear model with reporter, genetic background, and random effect of day with the Tukey HSD method. N.S. = P  > 0.10; * = P  < 0.05; *** = P  < 0.05.

To test whether the effects of the five differences were dependent on Tye7, we compared the expression of the YFP reporter genes driven by the S. cerevisiae, S. paradoxus, and two chimeric TDH3 promoters in a genomic background in which the TYE7 gene had been deleted (TYE7::ΔTYE7). After measuring YFP fluorescence for each of these eight strains, we used a linear model to estimate the individual and interaction effects of the promoter and TYE7 genotypes, accounting for day-to-day variation as a random effect. We found that all four promoters displayed lower activity when measured in the TYE7::ΔTYE7 strain (Fig. 3b; F = 563.3, P < 0.001, Type III analysis of variance (ANOVA)), indicating that Tye7p is required for normal TDH3 promoter activity in both S. cerevisiae and S. paradoxus. However, there was a significant interaction between promoter and TYE7 genotype (F = 4.06, P < 0.01, Type III ANOVA), indicating that removing Tye7p had different effects on different promoters. Specifically, in the absence of Tye7p, the S. cerevisiae alleles of the five divergent sites no longer caused a significant increase in expression level when introduced into the S. paradoxus TDH3 promoter (Fig. 3c; t = 2.29, P = 0.10, Tukey HSD). Similarly, introducing the S. paradoxus alleles for these five sites into the S. cerevisiae promoter no longer led to a statistically significant decrease in expression (Fig. 3d; t = 2.11, P = 0.16, Tukey HSD). These data show that the effects of the five divergent sites between the Rap1p and Gcr1p/2p binding sites are amplified by regulatory interactions involving Tye7p. Consistent with these findings, the difference in activity between the S. cerevisiae and S. paradoxus TDH3 promoters decreased by 49% in the absence of Tye7p, though the remaining difference was still statistically significant (Fig. 3e; t = 6.18, P < 0.001).

Prior work shows that Tye7p is first recruited to the TDH3 promoter through direct protein–protein interactions that require Gcr2p, after which specific Tye7–DNA interactions at the site of recruitment—through direct or indirect effects—can alter expression level even though they are not necessary for initial complex formation (Shively et al. 2019; Liu et al. 2020). The nucleotide divergence in this region may affect expression by altering the strength of direct contacts with the recognition helix of Tye7p or altering the shape of the DNA in this region. Our data cannot discern between alternative biochemical explanations for the observed effects. Nevertheless, genetic changes acting through either of these or other biochemical mechanisms may allow variation in expression level to evolve without altering the upstream regulators and signals to which the promoter responds, enabling different aspects of gene expression to evolve separately.

Our findings highlight the importance of sequences that may influence low-affinity interactions, cooperativity, and the assembly of multiprotein regulatory complexes. These factors, shown to be important in molecular and developmental studies of gene regulation (Junion et al. 2012; Crocker et al. 2016; Kribelbauer et al. 2019; Jindal and Farley 2021), can help explain not only how gene regulation works but also how variation in gene regulation evolves in populations and among species. As we expand our catalogs of evolutionarily significant regulatory mutations to include more genes and species, a more nuanced view of cis-regulatory evolution will likely emerge—one in which changes that fine-tune protein complex formation, rather than those that simply disrupt or create transcription factor binding sites, play a central role in shaping gene expression and ultimately contribute to phenotypic diversity.

Materials and methods

Yeast strains

The strains of S. cerevisiae used in this study were all derived from the standard S288c strain and had an alpha mating type. These strains had previously been modified to carry alleles of RM1, TAO3, CAT5, and MIP1 that increase sporulation efficiency and decrease petite frequency relative to the native alleles of the S288c, as described in Metzger et al. (2016). To quantify promoter expression, a reference nonfluorescent strain was engineered to carry different promoter–YFP constructs at a common position in the HO locus. The full cassette included the TDH3 promoter allele, a Venus YFP reporter, a Cyc1 terminator, and a KanMX resistance cassette. To swap out the different promoter alleles, an intermediate strain was created at the HO locus carrying the KanMX marker, a target site for CRISPR-Cas9 but lacking a promoter–YFP reporter construct; the different promoter–reporter constructs, including the reference S. cerevisiae TDH3 promoter allele, were then precisely introduced into this background using the same CRISPR-Cas9 target site using a modified version of the pML104 plasmid and accompanying protocol described in Laughery et al. (2015). The open reading frames for TDH3 and TYE7 were also removed using a similar CRISPR-Cas9-based strategy, albeit with 20 bp targets that were specific to the different loci (ACACACATAAACAAACAAAA for TDH3; TAGTCATATCAACGTCAACA for Tye7). For all strains generated using CRISPR-Cas9, the transformed cells were plated on SC-uracil and grown for 2 to 3 d at 30 °C. Colonies were screened for the correct genotype at the loci of interest with Sanger sequencing, and colonies with the desired changes were subsequently cured of the pML104 plasmid by selection on 5FOA media. Finally, the strains were grown in liquid YPD (10 g/L yeast extract, 20 g/L peptone, 20 g/L dextrose) until saturation, mixed with glycerol for a final concentration of 20% glycerol, and stored at −80 °C.

Annotated sequences of the constructs used at the HO locus are provided in File S1. Annotated sequences of the reference and deletion variants at the TDH3 and Tye7 locus, as well as a summary of yeast strains, are provided in File S2.

Plate reader assays

Strains with different genotypes were patched from glycerol stocks onto yeast extract–peptone–glycerol (YPG) agar media (10 g/L yeast extract, 20 g/L peptone, 20 g/L agar, 20 g/L glycerol) and grown at 30 °C for 2 to 3 d. Colonies were subsequently picked, placed onto 96-well 2 mL deep-well plates (randomized across wells) in 1 mL of liquid YPD, and grown with shaking for approximately 48 h at 30 °C. At this point, the strains had acclimated to growth on glucose and reached saturation. The strains were then diluted into 96-well plates, with 5 µL of saturated culture added to 195 µL of YPD. The plates were then incubated at 30 °C with shaking in a BioTek Synergy H1 plate reader (Agilent), with readings of OD660 and YFP fluorescence taken at 20-min intervals for 48 h. Each genotype was included in at least three separate wells per replicate, and three replicates were measured for each genotype.

Flow cytometry experiments

Strains with different genotypes were patched from glycerol stocks onto YPG agar media and grown at 30 °C for 2 to 3 d and subsequently transferred to YPD agar media to acclimate to growth on glucose. Colonies were then picked and grown in 1 mL of liquid YPD at 30 °C on a shaker or a rotating wheel for approximately ∼17 to 21 h, which is the stage when growth slows down following the exhaustion of sugar and prior to the diauxic shift into aerobic metabolism. Cells at this stage were diluted to an OD660 between 0.5 and 0.8 in 1× phosphate buffered saline (PBS), and these samples were analyzed by flow cytometry using an Attune NXT4 flow cytometer. Forward scatter height, width, and area (FSC.A) were used to identify populations of singlets, and the 488 nm excitation laser and 530/30 emission filter were used to record their associated fluorescent values.

Cell size is positively correlated with fluorescence, and we adjusted for differences in cell size using a principal component analysis in R using the flowClust (Lo et al. 2009) and flowCore (Hahne et al. 2009) packages. We took the population of cells that fit our criteria for single cells (eg no evidence of cell division or more than one cell in a droplet), estimated the logarithm of the forward scatter (FSC.A) and fluorescence intensity (BL1.A), defined the vector (ν) between the origin and intersection of the two eigenvectors, and calculated the angle (θ) between the first eigenvector and ν. This rotation angle reflects the residual correlation between size and fluorescence that persists after linear scaling; we therefore transformed FSC.A and BL1.A by a rotation of angle θ centered on the intersection of the eigenvectors and then divided the transformed FL1.A by the transformed FSC.A to obtain cell-size-corrected fluorescence values. These corrected values were used for downstream statistical analyses.

Statistical analyses

We used mixed-effect linear models to estimate the effect of different genetic factors on expression levels and account for random day-to-day variation in R using the lme4 (Bates et al. 2015), lmerTest (Kuznetsova et al. 2017), and emmeans (Lenth and Piaskowski 2026) packages. To estimate the effects of the reporter genotype and the presence/absence of TDH3 in the genetic background, we fit the following linear model: Fluorescence ∼ Reporter × TDH3 Genotype + (1|Date). We used the emmeans package to estimate the marginal means and contrasts among reporter and TDH3 genotype levels. Contrasts with Tukey-adjusted P < 0.05 were deemed to be statistically significant.

We used a similar framework for testing for the effects of the five nucleotides that varied between S. paradoxus and S. cerevisiae promoters as well as their dependence on Tye7p. For these experiments, we had three technical replicates for each genotype on each day (biological replicate). We used this replicate data to first estimate the mean fluorescence level of our reference genotype—the unmutated S. cerevisiae TDH3 reporter construct in an unmutated background—and scaled the fluorescence level of each sample by this number. We then calculated the mean scaled fluorescence level for each reporter genotype for each day and removed background autofluorescence, which was estimated using a fluorescence-null strain. The mixed-effect linear model used to analyze these data was specified as Scaled Fluorescence ∼ Reporter × Tye7 Genotype + (1|Date). We estimated the marginal means and statistically significant differences among reporter and Tye7 genotypes as described above. To estimate the percentage of rescue caused by the five nucleotides, we scaled the difference between a reference and recombinant allele of interest (eg Spar_cer5nt - Spar) by the total difference between reference alleles (eg Spar - Scer). We then estimated the 95% CIs associated with these percentages using Monte Carlo simulations. Specifically, we examined 50,000 samples of each genotype from their joint sampling distribution, which consisted of the estimated marginal means and the covariance matrix from the mixed model. We recomputed the proportion for each draw and took the 2.5th and 97.5th percentiles.

Supplementary material

Supplementary material is available at Molecular Biology and Evolution online.

Funding

This work was supported by the National Institutes of Health (5F32CA261115 and T32HG000040 to MAS and 5R35GM118073 to PJW) as well as the National Science Foundation (DEB-1911322 and MCB-1929737 to PJW) In addition, MAS was supported by the Michigan Pioneer Fellows program at the University of Michigan, and NJB was supported by the EEB Summer Research and Travel Award and Program in Biology Director’s Award from the University of Michigan. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health or the National Science Foundation.

Data availability

The raw flow cytometry data and information about the yeast strains described in this manuscript are publicly available: https://doi.org/10.5281/zenodo.21709377.