Fitness Effects of Cis-Regulatory Variants in the Saccharomyces cerevisiae TDH3 Promoter

Abstract

Variation in gene expression is widespread within and between species, but fitness consequences of this variation are generally unknown. Here, we use mutations in the Saccharomyces cerevisiae TDH3 promoter to assess how changes in TDH3 expression affect cell growth. From these data, we predict the fitness consequences of de novo mutations and natural polymorphisms in the TDH3 promoter. Nearly all mutations and polymorphisms in the TDH3 promoter were found to have no significant effect on fitness in the environment assayed, suggesting that the wild-type allele of this promoter is robust to the effects of most new cis-regulatory mutations.

gene expression, distribution of mutational effects, robustness, fitness optimum, gene regulation

Genomic studies have identified extensive variation in gene expression within and between species (reviewed in Ranz and Machado 2006; Alvarez etal. 2015). Given the key role that gene expression plays in regulating development and physiology, many of these changes in expression are expected to affect higher order phenotypes, which may in turn affect fitness (reviewed in Gilad etal. 2006; Wray 2007; Carroll 2008; Fay and Wittkopp 2008; Wittkopp and Kalay 2012). Despite these often-stated expectations, only a handful of studies have directly measured the fitness effects of changes in gene expression (Dykhuizen etal. 1987; Perfeito etal. 2011; Rest etal. 2013; Keren etal. 2016; Rich etal. 2016; Bergen etal. 2016), and most of these studies have used heterologous or synthetic promoters to alter gene expression level. It therefore remains largely unknown how differences in gene expression induced by new mutations in regulatory sequences relate to changes in organismal fitness. Here we introduced mutations in the promoter of the TDH3 gene in Saccharomyces cerevisiae to alter TDH3 expression levels and determine the impact of these expression changes on fitness during clonal growth in a rich medium containing glucose, which is the preferred carbon source for S. cerevisiae (Gancedo 1998). We then use this information to predict the fitness effects of new mutations and natural polymorphisms within the TDH3 promoter that were previously characterized for their effects on expression (Metzger etal. 2015).

TDH3 encodes a glyceraldehyde-3-phosphate dehydrogenase best known for its role in central metabolism (McAlister and Holland 1985) but also implicated in silencing of Sir2-dependent genes in telomeric regions (Ringel etal. 2013) and observed among antimicrobial peptides secreted by S. cerevisiae during alcoholic fermentation (Branco etal. 2014). Prior work has shown that eliminating TDH3 decreased fitness in rich media (YPD) by ∼4% (Deutschbauer etal. 2005) and overexpressing TDH3 inhibited expression of telomeric genes (Ringel etal. 2013), suggesting that fitness (i.e., population growth rate) should be sensitive to TDH3 expression level. To determine the fitness effects of altered TDH3 expression levels, we selected eight G→A or C→T point mutations in the TDH3 promoter (P_TDH3) that caused a broad range of changes in expression of a yellow fluorescent protein (YFP) reporter gene (fig. 1A; Metzger etal. 2015) and introduced them into the native TDH3 locus (fig. 1B). To expand the range of expression changes assayed, we also constructed and analyzed three strains with duplications of the reporter gene separated by a URA3 marker as well as a strain with only the URA3 marker inserted downstream of the wild-type P_TDH3-YFP reporter gene to serve as a control for the duplication strains (fig. 1A). One of these duplications contained two copies of the wild-type P_TDH3allele and the other two duplications each carried a single mutation in both copies of the promoter (fig. 1A). A matching set of strains containing duplications of the TDH3 gene was also constructed and used to measure fitness (fig. 1B). We used the parent of the reporter gene strains (BY4724), which lacks the P_TDH3-YFP reporter gene (labeled “Deletion” in fig. 1A), to determine the level of autofluorescence in the absence of YFP, and we deleted the promoter and coding sequence of the native TDH3 gene (labeled “Deletion” in fig. 1B) to quantify fitness when TDH3 was not expressed. Prior work has shown that the effects of mutations in the P_TDH3-YFP reporter gene on fluorescence levels are nearly perfectly correlated (R² > 0.99) with the effects of the same mutations at the native TDH3 locus measured using a TDH3::YFP fusion protein (Metzger etal. 2016).

Fig. 1

Genomic constructs used to alter TDH3 expression and measure effects on fitness. Schematics show the 14 genomic constructs used to quantify the effects of different alleles of the TDH3 promoter (P_TDH3) on (A) expression level and (B) fitness. (A, B) Wild-type alleles of P_TDH3 are shown in brown with transcription factor binding sites for RAP1 (purple) and GCR1 (green) indicated. Arrows show transcription start sites, and thick black lines represent surrounding genomic sequence. Thinner black lines represent genomic constructs with point mutations (G→A or C→T) in P_TDH3 at sites indicated with red “X”s. Constructs shown in (A) have these alleles fused to YFP coding sequence and the CYC1 transcription terminator and were inserted into chromosome I at position 199,270. From top to bottom, these schematics represent the wild-type P_TDH3-YFP allele, eight alleles with single point mutation in P_TDH3, the parental BY4724 strain lacking P_TDH3-YFP, the control for duplication alleles with a URA3 marker inserted downstream of the wild-type P_TDH3-YFP allele, the duplication of the wild-type P_TDH3-YFP separated by the URA3 marker, and two duplication alleles that each have a point mutation in both copies of P_TDH3. Blue bars to the right of these 14 schematics show the average fluorescence level of each construct relative to the wild-type allele. Constructs shown in (B) have these P_TDH3 alleles inserted at the native TDH3 locus on chromosome VII. From top to bottom, they represent the wild-type TDH3 gene, eight alleles with single point mutations in P_TDH3, the null allele created by a deletion spanning the 678-bp promoter and the entire 999-bp TDH3 coding sequence, the control for duplication alleles with a URA3 marker inserted downstream of the native TDH3 gene, the duplication of the wild-type TDH3 gene (promoter, coding sequence, and transcription terminator), and two TDH3 duplication alleles that each contains a single point mutation in both copies of the P_TDH3 promoter.

The expression level driven by these 14 P_TDH3 alleles was estimated using the fluorescence level of the reporter gene strains measured by flow cytometry after growth in a rich medium (YPD). Relative median fluorescence of each genotype (shown in fig. 1A) was determined by averaging median fluorescence from at least four replicate populations and dividing it by the average fluorescence of the corresponding wild-type allele. To quantify the impact of changing TDH3 expression on fitness, each of the 14 TDH3 strains (fig. 1B) was cocultured with a common competitor strain in the same growth conditions (fig. 2A). Abundance of each strain relative to the competitor was quantified in at least seven replicates at four time points during ∼20 generations of growth (fig. 2B), and then relative fitness was calculated for each genotype by dividing the average competitive fitness across replicates by the average competitive fitness of the genotype carrying the corresponding wild-type allele of P_TDH3 (fig. 2A).

Fig. 2

Fitness effects of cis-regulatory mutations and polymorphisms in the TDH3 promoter. (A) An overview of the competition assay used to quantify relative fitness is shown. Each of the 14 TDH3 alleles shown in figure 1B was introduced into a strain carrying wild-type P_TDH3-YFP inserted at the HO locus, which was used to mark cells, not to measure expression. The resulting strains were competed individually against a strain wild-type for TDH3 that was marked with a green fluorescent protein (P_TDH3-GFP reporter gene at HO. Each of the 14 [YFP+] strains was mixed with the [GFP+] strain in at least seven replicates and grown at log-phase in rich medium (YPD) for 30 h (∼20 generations). The frequency of [YFP+] and [GFP+] cells was quantified by flow cytometry at four time points, once every 10 h, at which points cell cultures were diluted to fresh medium. Competitive fitness (W) was determined from changes in the frequency of [YFP+] and [GFP+] cells over time. The logarithm of the relative frequency of [YFP+] and [GFP+] cells was regressed on the number of generations of growth and the exponential of the regression coefficient corresponded to the competitive fitness of the [YFP+] strain. To account for the small difference in growth rate between the [YFP+] strain carrying the wild-type allele of TDH3 and the [GFP+] competitor strain, relative fitness was calculated by dividing the average competitive fitness measured for each mutant strain (W_mut) by the average competitive fitness measured for the strain with the wild-type allele of TDH3 (W_WT). (B) Changes in genotype frequency over time (number of generations) are shown for the eight alleles with single point mutations in P_TDH3 (top; plain gray lines), for the deletion of TDH3 (top; dotted line), for the three alleles with duplication of the entire TDH3 locus (bottom, plain gray lines), and for the corresponding control strains (black lines). For better visualization, variation in the starting frequency of [YFP+] and [GFP+] cells was removed by subtracting the logarithm of the ratio of [YFP+] and [GFP+] cells measured at the first time point from the ratio measured at each time point (y-axis). Dots indicate the average value of this corrected ratio across all replicate populations for each genotype and at each time point, whereas error bars show 95% confidence intervals across replicates. Note that the frequency of the wild-type [YFP+] strain slightly increased over time (top; black line) and that the control strain with the URA3 marker (bottom; black line) grew slightly faster than the wild-type strain without URA3 (top; black line), reflecting the small fitness advantage conferred by the URA3 marker in rich medium. (C) The relationship between relative P_TDH3 activity, measured by fluorescence of the YFP reporter genes shown in (A), and relative fitness is shown by the black points and the dotted curve. Each point represents the mean of median fluorescence levels (x-axis) among at least four replicate populations and the mean of relative fitness measurements (left y-axis) from at least seven replicate competition assays, with different symbols used for the wild-type allele (open circle), point mutations (closed circles), null allele (open triangle), and duplication alleles (open diamonds). Error bars show the 95% confidence intervals for both measurements. The dotted curve was defined using a LOESS regression of the fluorescence data on fitness, with the grayed interval around the curve representing the 95% confidence interval of this LOESS regression. The range of TDH3 expression levels with fitness comparable to the wild-type allele is shown in yellow. Histograms showing the distributions of effects on P_TDH3activity (as measured by fluorescence of the YFP reporter gene) for 235 mutations (red) and 30 polymorphisms (blue) are overlaid on the fitness curve. Each bar represents a bin of 0.8% on the x-axis, and the number of mutants in each bin is shown on the right y-axis. The vertical and horizontal dotted lines show the fluorescence level and fitness conferred by the wild-type TDH3 promoter allele.

These data allowed us to determine the relationship between TDH3 expression level and fitness upon growth in a rich medium (fig. 2C). We found that a decrease in TDH3 expression level as small as 6% relative to the wild-type level (as measured by YFP fluorescence) significantly reduced fitness by 0.3% (t-test, t = −2.08, P = 0.045), whereas a 36% increase in expression level significantly decreased fitness by 1.0% (t-test, t = −10.70, P = 2.11 × 10⁻⁷). The decrease in fitness associated with overexpression of TDH3, which is one of the most highly expressed gene in S. cerevisiae (Newman etal. 2006), could be due to the energetic cost of protein expression (Dekel and Alon 2005; Makanae etal. 2013; Chou etal. 2014; Kafri etal. 2016) or to specific functions of TDH3 such as an increase in the silencing of telomeric genes (Ringel etal. 2013). No significant changes in fitness were observed for P_TDH3 alleles driving expression levels between 96% (t-test, t = 1.91, P = 0.06) and 118% (t-test, t =0.08, P = 0.94) of the wild-type level, indicating a plateau of maximal fitness for expression levels in this range (fig. 2C). This plateau might extend to even higher expression levels given that it is defined by only one overexpression allele with a significant reduction in fitness, and this allele, a duplication of the native TDH3 gene, showed a 136% increase in fluorescence rather than the doubling reported previously for a duplication of P_TDH3 (Kafri etal. 2016), suggesting that our measurements underestimated YFP abundance at high fluorescence levels. Differences in expression noise (variability among genetically identical cells) among P_TDH3 alleles might also contribute to differences in fitness among these P_TDH3 alleles (Metzger etal. 2015), but the fitness effects of changing expression noise are expected to be much smaller than the fitness effects of changing mean expression level (Wang and Zhang 2011; Keren etal. 2016). The data presented here are insufficient to disentangle the fitness effects of changing expression noise and mean expression level because the P_TDH3 alleles examined show strongly correlated effects on these two measures (supplementary fig. 1, Supplementary Material online).

Using the relationship we observed between changes in TDH3 expression level and relative fitness, we inferred the fitness effects of 235 G→A or C→T point mutations in P_TDH3 previously characterized for their effects on YFP fluorescence in the same environment (Metzger etal. 2015). We found that 216 (92%) of these 235 mutations caused changes in expression that are not predicted to affect fitness in this environment (red distribution in fig. 2C). None of these mutations increased expression more than 4% relative to the wild-type allele, and the distribution of mutational effects was concentrated at the lower expression end of the fitness plateau (fig. 2C), suggesting that a mutational bias limits increases in TDH3 expression from arising in similar environments. To understand how selection has filtered this mutational variation in natural populations, we also examined the effects of 30 unique polymorphisms on P_TDH3 described in Metzger etal. (2015). We found that 29 (97%) of these 30 polymorphisms caused expression levels within the observed plateau of maximal fitness (blue distribution in fig. 2C), consistent with the absence of evidence for selection acting on P_TDH3 activity level in a glucose-based medium reported previously (Metzger etal. 2015). The effects of polymorphisms were also concentrated near the lower expression end of the fitness plateau (maximum level = 103%), again suggesting that natural variation in the activity of the TDH3 promoter is constrained by mutational biases. Growth in other environments might change the effects of P_TDH3 mutations on gene expression and/or the relationship between TDH3 expression level and fitness (Keren etal. 2016) and should be investigated in future work. Effects of different types of mutations (e.g., indels, duplications, aneuploidies) should also be examined to determine how they might change the distribution of mutational effects.

Finding that most cis-regulatory mutations and polymorphisms cause changes in TDH3 expression level within the observed fitness plateau indicates that their effects are largely buffered at the fitness level, conferring robustness in the activity of the wild-type P_TDH3 allele against new mutations, at least in the environment assayed. Nonlinear fitness functions with plateaus located around the wild-type level of gene expression have previously been described in S. cerevisiae for the LCB2 gene using an inducible promoter (Rest etal. 2013) as well as for other yeast genes using synthetic promoters (Keren etal. 2016) or interspecific comparisons (Bergen etal. 2016), and are consistent with predictions from metabolic flux control theory (Kacser and Burns 1973). These observations suggest that fitness plateaus might be common for expression–fitness functions and might play an important role in shaping the diversity of gene expression levels observed in natural populations.

Supplementary Material

Supplementary data are available at Molecular Biology and Evolution online.

Acknowledgments

We thank Chetna Gopinath for technical assistance, Brian Metzger, David Yuan, Bing Yang, and Andrea Hodgins-Davis for helpful discussions, and Brian Metzger, Jennifer Lachowiec, and Andrea Hodgins-Davis for comments on the manuscript. This work was supported by a European Molecular Biology Organization postdoctoral fellowship (EMBO ALTF 1114-2012) to F.D., a PhD fellowship from Ecole Doctorale BMIC de Lyon to W.T., and grants from the National Science Foundation (MCB-1021398) and National Institutes of Health (R01GM108826 and 1R35GM118073) to P.J.W.

Author notes

Associate editor: Ilya Ruvinsky