Report

A Yeast Hybrid Provides Insight into the Evolution of Gene Expression Regulation

See allHide authors and affiliations

Science  01 May 2009:
Vol. 324, Issue 5927, pp. 659-662
DOI: 10.1126/science.1169766

Abstract

During evolution, novel phenotypes emerge through changes in gene expression, but the genetic basis is poorly understood. We compared the allele-specific expression of two yeast species and their hybrid, which allowed us to distinguish changes in regulatory sequences of the gene itself (cis) from changes in upstream regulatory factors (trans). Expression divergence between species was generally due to changes in cis. Divergence in trans reflected a differential response to the environment and explained the tendency of certain genes to diverge rapidly. Hybrid-specific expression, deviating from the parental range, occurred through novel cis-trans interactions or, more often, through modified trans regulation associated with environmental sensing. These results provide insights on the regulatory changes in cis and trans during the divergence of species and upon hybridization.

Over the past years, extensive variations in gene expression were identified between closely related species (1, 2). The genetic basis of most differences, however, remains unknown. Expression divergence of a specific gene can result from mutations in its regulatory sequences, such as promoter elements (cis effects), or from mutations elsewhere in the genome that alter the abundance or activity of upstream regulators (trans effects). Distinguishing the relative contribution of cis and trans effects to the divergence of gene expression is an essential step in elucidating its genetic basis.

When comparing different strains of the same species, cis and trans effects can be approximated by using linkage analysis (35). Alternatively, for comparison of different species that produce viable hybrids, cis versus trans effects can be distinguished by using the interspecific hybrid (6). Within the hybrid, both alleles of each gene are exposed to the same nuclear environment. Thus, differences in the expression of the two hybrid alleles reflect cis effects, whereas expression differences between the parental genes that disappear in the hybrid reflect trans effects. This approach was previously used to analyze several dozen Drosophila (6, 7), yeast (8), and maize (9) genes and recently also a single mammalian chromosome (10), but has not yet been applied to a whole genome.

We designed a microarray that enables the measuring of allele-specific expression in a hybrid of Saccharomyces cerevisiae and S. paradoxus, two yeast species that diverged ~5 million years ago (fig. S1). Hybridization of genomic DNA from the two parental species verified the specificity of the microarray (Fig. 1A) and confirmed the lack of major variations in copy number. Biological repeats were highly correlated [Pearson correlation (r) ~ 0.98], significantly more so than the correlation in expression between species (r ~ 0.85) or between the corresponding alleles within the hybrid (r ~ 0.9) (Fig. 1B).

Fig. 1

Mapping cis and trans effects. (A) Genomic DNA of S. cerevisiae and S. paradoxus were differentially labeled and cohybridized to the array. Shown is the histogram of hybridization intensities (log2 ratio) for S. cerevisiae (black) or S. paradoxus (gray) probes. (B) Correlations between the genome-wide expression levels from the TSA experiment. All samples (including biological repeats) were processed in parallel (see fig. S2 for all experiments). (C) Clustered heat map of genes with significant cis and/or trans effects above 1.4-fold, and examples of genes with only cis [suppression of Gal11 expression 1 (SGE1)] or trans [zinc-regulated transporter 1 (ZRT1)] effects. Promoter analysis suggests a mechanistic basis for the cis effect in SGE1: an arsenicals resistance 1 (ARR1)–binding site at the S. cerevisiae promoter is mutated in S. paradoxus. (D) The number of genes with significant cis and trans effects above 1.4-fold in absolute expression levels (left) or in expression response (right). Rich, rich media; Gly, glycerol; HS, heat shock; TSA, TSA addition. (E) Genome-wide correlations between cis and trans effects across conditions.

Using the array, we measured the expression profiles of the two parental species and their hybrid under four different growth conditions and quantified the relative contribution of cis and trans effects to the interspecies divergence (Fig. 1C). As expected, cis (but not trans) effects were correlated with sequence divergence at promoters and regulatory elements (fig. S3), whereas trans (but not cis) effects were enriched with genes whose expression was altered upon deletion of transcription or chromatin regulators (fig. S4 and S5). A small portion of the trans effects (<1%) were enriched within contiguous chromosomal regions that display correlated expression divergence, possibly indicating epigenetic effects (fig. S6).

The relative contribution of cis and trans effects to the divergence of gene expression varied under the different environmental conditions (Fig. 1D). In three of the conditions [rich media, heat shock, and the addition of an Rpd3p inhibitor, trichostatin A (TSA)], cis effects dominated, which is consistent with previous reports of Drosophila (6, 7) and mammals (10). However, in glycerol, on which S. cerevisiae grows poorly (fig. S7), the number of cis and trans effects was comparable because of an increased number of trans effects.

We compared the consistency of cis and trans effects among the conditions (Fig. 1E). Cis effects were highly correlated (r ~ 0.85), both in the identity of the divergent genes and the direction of the effect. Trans effects were more condition-dependent, and their number varied by approximately sixfold between conditions (Fig. 1D). Analysis of variance indicated that most (77%) of the cis effects were independent of condition, whereas most (67%) of the trans effects were condition-dependent. As a consequence of the consistency of cis effects, trans effects dominated the divergence of the transcriptional response to changing conditions (Fig. 1D), as has previously been observed for yeast strains (3).

Previous studies showed that the expression of certain genes diverges rapidly across different evolutionary lineages (1, 11), and that this propensity for expression divergence correlates with the presence of a TATA box and with the lack of a pronounced nucleosome-free region (NFR) adjacent to transcription start sites (12). In our data, we found that genes with trans (but not cis) effects were enriched with TATA boxes, lacked NFRs, and displayed a high expression divergence in seven existing data sets that compared the expression program of different yeast strains or species (Fig. 2 and fig. S8). Thus, the propensity to diverge seems to mostly reflect increased sensitivity to the divergence of trans regulation and not a higher sensitivity to mutations in cis.

Fig. 2

Propensity for expression divergence depends on trans effects. (A) Correlation between the extent of cis and trans effects in our data and the expression divergence defined by seven previous studies [all trans correlations are significant at P < 0.01 (20)]. (B) Genes were sorted according to the extent of cis (gray) or trans (black) effects averaged over the four conditions. Shown is the proportion of promoters containing a TATA box (top) and having an occupied proximal-nucleosome (OPN) pattern (12) (bottom) within a sliding window of 300 genes.

Trans effects could originate from functional divergence of the direct transcription and chromatin regulators (“regulator trans”) or from mutations in upstream components involved in transducing environmental or internal signals to these direct regulators (“sensory trans”) (Fig. 3A). The divergence (in sequence or in expression) of transcription regulators was not correlated with the trans divergence of their target genes (Fig. 3B and fig. S9). In contrast, the strength of trans divergence was strongly correlated with the variation in gene expression upon environmental changes [as estimated by a data set of hundreds of S. cerevisiae expression profiles (13)] (Fig. 3C), which suggests a dominant role of sensory trans effects.

Fig. 3

Trans effects are dominated by sensory trans. (A) Models of sensory trans versus regulator trans. (B) The average magnitude of trans effects for target genes of each transcription or chromatin regulator is compared with the expression divergence (cis + trans, left) or sequence divergence [rate of nonsynonymous substitutions (Ka), right] of that regulator. (C) The average magnitude of trans effects was compared with the normalized response to environmental changes, as determined from a database of >1500 experiments (13). (D) Correlations between the trans effects of different genes follow the similarity in their response to environmental perturbations rather than their association with the same transcription or chromatin regulators (20). (E) Distribution of trans effects for ESR genes among the four conditions.

To more rigorously distinguish between the regulator and sensory trans hypotheses, we examined pairs of genes that are controlled by different regulators but exhibit a similar response to environmental changes (Fig. 3D). If trans effects are due to regulator divergence, these pairs should diverge independently. Instead, we observed a correlated divergence for most such gene pairs (Fig. 3D), as was predicted by the sensory trans hypothesis. Similarly, divergence of most gene pairs that share a common regulator but respond differently to environmental changes was not correlated (Fig. 3D), which again is consistent with the sensory trans hypothesis. One example of the impact of differential sensing is activation of the environmental stress response (ESR). S. cerevisiae grows poorly on glycerol, whereas S. paradoxus grows poorly during heat shock. Accordingly, ESR genes were activated more strongly in S. cerevisiae on glycerol but more strongly in S. paradoxus during heat shock and were similarly expressed at the other two conditions (Fig. 3E and fig. S7). Furthermore, such patterns of growth-correlated expression differences were observed for 44% of all trans effects (as compared with less than 20% of the cis effects, as expected by chance). The idea that most trans effects reflect differences in sensory signals is supported also by previous analyses, in which variations in expression of multiple genes did not map to transcription regulators (4) but, in certain cases, to signal transduction genes (3, 4, 14).

Being upstream of the immediate regulators, sensory trans divergence is expected to coordinately influence a large number of genes. Mutations in cis could act to fine-tune these effects at particular genes (15, 16). The interaction between cis and trans effects could increase divergence if they act in the same direction (“enhancing”; for example, both effects increasing the expression of the S. cerevisiae ortholog) or could decrease divergence if they act in opposite directions (“compensating”) (fig. S10). Different evolutionary models make distinct predictions regarding the fraction of enhancing versus compensating interactions. Under complete neutrality, they should be equally favored. If natural selection acts to eliminate differences, compensating interactions would be favored. Finally, if expression changes are beneficial, then enhancing interactions would be preferred. We find a significant, albeit small, enrichment of compensating interactions (Fig. 4A), which suggests a role of purifying selection in the buffering of gene expression divergence. Consistent with this, both cis and trans effects are underrepresented among essential genes (fig. S11).

Fig. 4

Expression rewiring in the hybrid. (A) Enrichment of compensating over enhancing cis-trans interactions. (B and C) Examples of overexpressed genes with (B) or without (C) compensating cis-trans interactions, corresponding to allele-specific or common-allele overexpression, respectively. (D) Venn diagrams classifying over- or underexpressed genes according to their main contribution of hybrid-specific expression (allele-specific or common-allele changes) and the presence of compensating interactions.

Compensating cis-trans interactions can generate genetic variation that is invisible when comparing the two species but is revealed after interspecific hybridization (16) (fig. S10). Approximately 1 to 6% of the genes were overexpressed in the hybrid at each condition, displaying an expression that was higher than that of either parental species, and ~1 to 2% of the genes were underexpressed. Compensating cis-trans interactions were highly enriched among the genes displaying hybrid-specific expression (Fig. 4A); in most of these cases, only one of the hybrid alleles had changed in expression, whereas the other allele maintained an expression similar to that of the parental species (Fig. 4B). This is indeed expected if hybrid-specific expression results from interactions between a trans component of one genome and the respective cis element of the allele from the other genome (fig. S10) (17).

Although compensating cis-trans interactions were enriched in genes that displayed hybrid-specific expression, they accounted for only the minority (~20%) of such cases. More often, hybrid-specific expression was generated through a coordinated modulation of both alleles (Fig. 4, C and D), suggesting a modified trans-effect within the hybrid. These hybrid-specific trans-effects appear to be dominated by sensory trans (fig. S12), indicating that interactions between the two genomes alter the way in which the hybrid interprets its sensory signals (for example, compensating cis-trans interactions in sensory genes may propagate as hybrid-specific transeffects). Genes involved in respiration were highly enriched among the hybrid-specific trans effects (fig. S13). These genes were also subject to extensive divergence between the two parental yeasts, consistent with the idea that hybrid-specific trans-effects exaggerate differences already found between the parental species.

Taken together, our genome-wide mapping of cis and trans divergence points to three main conclusions. First, cis effects account for most of the expression divergence and are consistent across conditions, whereas trans effects are condition-specific and underlie the propensity for expression divergence. Second, trans effects are primarily attributable to differential interpretation of sensory signals and not to mutations in direct transcriptional regulators. Finally, parental genomes accumulate some compensating cis-trans effects, which is a signature of purifying selection, and their unleashing in the hybrid account for ~20% of over-dominance expression patterns. The majority of hybrid-specific expression, however, resulted from trans effects that were specific to the hybrid.

We have used the yeast hybrid as a tool, but hybridization and allopolyploidization are common in nature. Hybrids are often incompatible (18) but occasionally display beneficial phenotypes that are absent from the parents [hybrid vigor (19)]. Further studies are required to establish the possible connection between the expression rewiring described in our study and the emergence of novel hybrid phenotypes.

Supporting Online Material

www.sciencemag.org/cgi/content/full/324/5927/659/DC1

Materials and Methods

Figs. S1 to S14

Table S1

References

References and Notes

  1. Alternatively, it is possible that some of these allele-specific expression changes reflect epigenetic effects in the hybrid, although this possibility appears unlikely because cytosine methylation has not been found in budding yeast and the observed allele-specific changes are not clustered at specific chromosomal domains [as might be expected for epigenetic effects (fig. S6)].
  2. Materials and methods are available as supporting material on Science Online.
  3. We thank J. Berman, Y. Eshed, N. Sigal, A. Sheperberg, and members of our groups for discussions and comments. This work was supported by the Helen and Martin Kimmel Award for Innovative Investigations and grants from the Kahn Fund for Systems Biology at the Weizmann Institute of Science, the Israeli Ministry of Science, the Bi-national Science Foundation, and the European Research Council (Ideas) (to N.B). A.A.L. holds the Gilbert de Botton chair of Plant Science. Microarray data has been deposited at the Gene Expression Omnibus with accession number GSE14708.
View Abstract

Navigate This Article