|Home | About | Journals | Submit | Contact Us | Français|
Changes in gene regulatory networks are a major source of evolutionary novelty1–3. Here we describe a specific type of network rewiring event, one that intercalates a new level of transcriptional control into an ancient circuit. We deduce that, over evolutionary time, the direct ancestral connections between a regulator and its target genes were broken and replaced by indirect connections, preserving the overall logic of the ancestral circuit but producing a new behaviour. The example was uncovered through a series of experiments in three ascomycete yeasts: the bakers’ yeast Saccharomyces cerevisiae, the dairy yeast Kluyveromyces lactis and the human pathogen Candida albicans. All three species have three cell types: two mating-competent cell forms (a and α) and the product of their mating (a/α), which is mating-incompetent. In the ancestral mating circuit, two homeodomain proteins, Mata1 and Matα2, form a heterodimer that directly represses four genes that are expressed only in a and α cells and are required for mating4–6. In a relatively recent ancestor of K. lactis, a reorganization occurred. The Mata1–Matα2 heterodimer represses the same four genes (known as the core haploid-specific genes) but now does so indirectly through an intermediate regulatory protein, Rme1. The overall logic of the ancestral circuit is preserved (haploid-specific genes ON in a and α cells and OFF in a/α cells), but a new phenotype was produced by the rewiring: unlike S. cerevisiae and C. albicans, K. lactis integrates nutritional signals, by means of Rme1, into the decision of whether or not to mate.
In S. cerevisiae, K. lactis and C. albicans, three cell types (a, α and a/α) are specified by transcriptional regulators (sequence-specific DNA-binding proteins) encoded at the mating type locus. An important part of this cell-type-specific circuit is the regulation of the haploid-specific genes (hsgs), a group of genes that are expressed in a and α cells but not in a/α cells7. The full sets of hsgs were previously identified in S. cerevisiae4 and C. albicans (ref. 5, and B.B.T., Q. M. Mitrovich, F. M. De La Vega, C. K. Monighetti and A.D.J., unpublished observations) but not in the related species K. lactis. To examine the evolution of this portion of the mating circuit, we identified the genes in the K. lactis hsg regulon and compared them to those in S. cerevisiae and C. albicans. By profiling the expression patterns of wild-type a, α and a/α K. lactis cells genome-wide, we identified 12 genes that are clear hsgs under the conditions tested (Fig. 1a), two of which—RME1 (referred to previously as MTS1) and STE4—were previously identified as hsgs in K. lactis8. Comparison of all the hsgs in the three species revealed a substantial level of turnover in the regulon; in other words, an hsg in one species is not necessarily an hsg in the other two (Fig. 1b). However, an ancestral core of four hsgs (GPA1, STE4, STE18 and FAR1) share a common expression pattern in all three species. The first three genes encode the heterotrimeric G protein that, in the presence of mating pheromone, activates a downstream mitogen-activated protein kinase (MAPK) cascade required for mating9–11. Far1 lies further downstream in this pathway and mediates two responses needed as a prelude to mating, cell cycle arrest12 and the formation of mating projections13,14.
In S. cerevisiae and C. albicans a/α cells, all four genes of the hsg core regulon are directly repressed by the transcription regulator a1–α2 (refs 4, 6), a heterodimer encoded by one gene at the MATa locus and one gene at the MATα locus. To determine whether this was also true in K. lactis, genome-wide chromatin immunoprecipitations (ChIP-chip) of a1 and α2 were performed in K. lactis a/α cells. In total, the upstream regions of 14 genes were observed to be occupied by both a1 and α2, including the RME1 gene (Fig. 1c), which is also a1–α2 regulated in S. cerevisiae. a1 and α2 ChIP peaks were not observed at the promoters of any of the four core hsgs in K. lactis, indicating that, unlike in S. cerevisiae and C. albicans, these genes are not directly regulated by a1–α2.
To confirm the absence of direct a1–α2 regulation at K. lactis hsgs, we identified the a1–α2 recognition motif in K. lactis from the ChIP data, using a de novo motif-finding program15. The highest-scoring motif was similar to the a1–α2 motifs previously identified in S. cerevisiae16 and C. albicans17 (Fig. 1d). Indeed, the S. cerevisiae motif is efficiently recognized by the C. albicans a1–α2 protein17, confirming that key features of this sequence have remained largely unchanged in the three species. We searched the regions 2 kilobases upstream of each K. lactis core hsg for the K. lactis a1–α2 motif but did not find significant matches, confirming the absence of direct a1–α2 regulation of these genes. (Whereas the a1–α2 site upstream of RME1 had a log10-odds score of 4.98, the best matches at the core hsgs ranged from −0.70 to 0.93.) These results indicate that although the ancestral core hsg expression pattern is conserved in K. lactis, the mechanism of the regulation has changed.
To understand how the K. lactis hsgs are cell-type regulated we searched the upstream regions of the 12 genes identified as hsgs by expression array (Fig. 1a) for cis-regulatory motifs15. The second-highest ranking motif (the top-ranking motif was a repeat sequence) was found in 11 out of 12 of the promoters (Fig. 2a) and was similar to the S. cerevisiae Rme1 motif (K. lactis consensus, GAACCNMAA; S. cerevisiae consensus, GAACCTCAA18,19). This motif is also similar to, although longer than, the K. lactis Rme1 motif derived previously20. The Rme1 motif is absent from S. cerevisiae and C. albicans hsg promoters.
In S. cerevisiae, Rme1 was initially identified as a repressor of meiosis and sporulation21,22, and was later shown to act as a transcriptional activator of other genes19. In K. lactis, Rme1 was shown to regulate mating-type interconversion20 (the switching of a and α cells to the opposite cell-type by means of DNA rearrangement). We speculated that Rme1 was co-opted in the K. lactis lineage to positively regulate the core hsgs.
To test this hypothesis, we knocked out the RME1 gene in K. lactis a cells and examined the gene expression profile by microarray. We found that 20 genes were downregulated in the knockout strain (Fig. 2b), including all four of the core hsgs (P = 2 × 10−10, hypergeometric distribution). We also observed a set of genes that was upregulated in the absence of Rme1 (Supplementary Fig. 1), including a significant number of genes orthologous to S. cerevisiae sporulation genes (Gene Ontology (GO): 0043934, n = 14, P = 10−14, hypergeometric distribution) indicating that the function of Rme1 in regulating meiosis is shared with S. cerevisiae. Thus, whereas some of the targets of Rme1 have remained the same as in the common ancestor of S. cerevisiae and K. lactis, Rme1 has gained new targets in the K. lactis lineage, including the core hsgs.
To test whether Rme1 directly regulates the core hsgs in K. lactis, we performed a genome-wide ChIP of Rme1 in a cells. ChIP peaks were observed at the promoters of the four core hsgs (Fig. 2c) and were centred over the Rme1 motifs (Fig. 2c). Thus, in the K. lactis lineage, Rme1 was gained as a direct activator of the core hsgs by the acquisition of Rme1 cis-regulatory sequences at all four genes. We note that Rme1 is not the only regulator of the K. lactis hsgs; for example, STE18 is repressed by Sir2 (ref. 23).
We next tested the biological role of Rme1 in mating in K. lactis, S. cerevisiae and C. albicans by comparing wild-type and RME1 knockout a cells. In response to α pheromone, a cells form mating projections (polarized growth towards the source of pheromone). When S. cerevisiae and C. albicans wild-type and Δrme1 a cells were exposed to α mating pheromone, both strains formed mating projections normally (Fig. 2d). In contrast, whereas K. lactis wild-type a cells produced mating projections in response to pheromone, Δrme1 a cells did not, indicating that this biological response was dependent on Rme1 (Fig. 2d). As a second test of the role of Rme1, we examined mating directly using a quantitative mating assay. No difference was observed between the mating efficiencies of wild-type a cells and those of Δrme1 a cells for S. cerevisiae and C. albicans (Fig. 2e). In contrast, the K. lactis Δrme1 a cell mating efficiency was decreased, relative to the wild type, by a factor of at least 106 (Fig. 2e). Thus, the ability to mate is critically dependent on Rme1—but only in K. lactis.
Unlike S. cerevisiae and C. albicans, K. lactis requires a starvation signal to mate24 and to respond to pheromone25. Although several different types of starvation signal can prime K. lactis to respond to pheromone24,25, we found that phosphate starvation is particularly potent, and it was used in subsequent experiments. Our expression profiling experiments (Fig. 1a) revealed that K. lactis requires starvation to express most of its mating genes. RME1 was also highly induced (24-fold) by phosphate starvation (Fig. 1a). We note that S. cerevisiae RME1 transcript levels also increase tenfold under starvation conditions18, suggesting that regulation of RME1 by starvation may be ancestral to S. cerevisiae and K. lactis.
We next investigated in greater detail how the starvation signal is incorporated in the K. lactis mating regulatory circuit. The simplest model consistent with the data presented so far is that starvation upregulates RME1, which in turn activates transcription of the hsgs. A prediction of this model is that ectopic expression of RME1 in K. lactis should override the requirement for starvation in expressing the hsgs. We created an a strain overexpressing RME1 to levels that were within tenfold of the level in starved wild-type cells (using the Kl LAC4 promoter) and found that overexpression of RME1 (in rich YEP-galactose medium) was sufficient to induce expression of the heterotrimeric G protein subunits (Fig. 3a). Overexpression of RME1 is also sufficient to allow K. lactis to form mating projections in response to pheromone in rich medium (Fig. 3b). These results strongly support the model by showing that upregulation of RME1 is sufficient to cause biologically relevant upregulation of the heterotrimeric G proteins.
Thus, the rewiring of the K. lactis hsg circuit (summarized in Fig. 4) resulted in a new network configuration and a novel phenotype, relative to the ancestor. Our results suggest a possible evolutionary path for this rewiring. In the ancestor of all three yeasts, the hsgs were directly repressed by a1–α2. Either in an ancestor to S. cerevisiae and K. lactis or independently in each lineage, RME1 was brought under nutritional regulation. Finally, in the K. lactis lineage alone, two steps occurred: the hsgs lost the cis-regulatory sequences for a1–α2 and gained the cis-regulatory sequences for Rme1. As described in Supplementary Fig. 2, it is possible to determine more precisely when the rewiring of the core hsgs occurred. We can infer that direct a1–α2 regulation of the core hsgs was probably lost several times in the ascomycete lineage, and that the K. lactis form of regulation of the hsgs probably arose after K. lactis and the closely related species L. kluyveri branched from their common ancestor.
Although we do not know whether acquisition of the K. lactis form of regulation was adaptive, this type of regulation makes logical sense given that the primary mode of growth of K. lactis is as a haploid26. The formation of spores is a strategy employed by many yeasts to survive harsh environments. For starvation to give rise to spores, K. lactis would first have to mate (to form the sporulation-competent a/α cell type), thus rationalizing the link between starvation and mating. In contrast, S. cerevisiae in the wild is typically at least diploid27 and forms spores directly in response to starvation. Thus, the coupling of mating and starvation makes conceptual sense for K. lactis in comparison with S. cerevisiae.
We have described a case in which a new tier of regulation has been intercalated into an ancient transcription circuit consisting of a regulator (a homeodomain heterodimer) and a set of target genes. This change involved breaking the original connections between the regulator and its target genes and replacing them with a more complex type of hierarchy (Fig. 4). Intercalation may be a common way in which regulatory circuits evolve. This type of ‘intercalary evolution’ was first proposed28 to account for a common origin of eyes. In a wide variety of species, the transcription regulator Pax6 lies at the top of the eye development hierarchy, and rhodopsins occupy the bottom. According to the proposal, different types of eye arose from evolutionary intercalation of a variety of regulatory and structural genes within this simple, deeply conserved, regulatory relationship. The change we describe here is less complex and provides a concrete example of evolutionary intercalation, one that is responsible for an important feature of modern mating behaviour in K. lactis. It has been known for decades that K. lactis (unlike its relatives S. cerevisiae and C. albicans) requires starvation to mate24, and we have shown that this behaviour is due to the new configuration of the K. lactis mating circuit.
RNA was isolated by hot phenol extraction and reverse transcribed, and the resulting complementary DNAs were coupled to Cy5. A pooled mixture of the cDNAs was coupled to Cy3 and used as a reference. Labelled cDNAs were hybridized to Agilent arrays for visualization.
ChIP experiments were performed as described previously29, with minor modifications.
Exponential-phase cultures were exposed to 13-mer α-mating pheromone, and the formation of mating projections was monitored by microscopy.
a and α cultures were grown independently to exponential phase and then mixed together with α cells in fivefold excess under mating conditions. The mating products were selected for on medium that either a and a/α cells or only a/α cells could grow on, and efficiencies were calculated as efficiency = (a/α colonies)/(a and a/α colonies).
Cultures were grown to exponential phase in YEP-galactose medium, and RNA was isolated by extraction with hot phenol. RNA was reverse transcribed and the cDNAs were quantified by quantitative PCR.
We thank Q. Mitrovich, O. Homann, A. Hernday, M. Miller, C. Cain, T. Sorrells and H. Madhani for helpful discussions and technical contributions; and S. Åström for generously providing the K. lactis strains used in this study. The S. cerevisiae strains were a gift from the H. Madhani and J. Li laboratories. The work was funded by grant RO1 GM037049 from the National Institutes of Health. L.N.B. is a National Science Foundation Graduate Research Fellow.
Full Methods and any associated references are available in the online version of the paper at www.nature.com/nature.
Supplementary Information is linked to the online version of the paper at www.nature.com/nature.
Author Contributions L.N.B. performed all experiments. L.N.B. and B.B.T. analysed data. L.N.B., B.B.T. and A.D.J. designed the study and wrote the paper.
Author Information The gene expression array data have been deposited in the NCBI Gene Expression Omnibus (GEO) under accession number GSE24874. For the ChIP-chip data the accession number is GSE25209. Reprints and permissions information is available at www.nature.com/reprints. The authors declare no competing financial interests. Readers are welcome to comment on the online version of this article at www.nature.com/nature.