The fact that changes in transcriptional logic bring about adaptive changes — as might occur when developmental modules are rearranged — is no longer front page news. Now, a study of mating type in two divergent yeast species shows the opposite principle: that phenotypes might stay the same during evolution, while the transcriptional logic is sometimes completely rewired.

The ascomycete yeasts Candida albicans and Saccharomyces cerevisiae each come in two varieties, a and α, depending on which allele they express at the mating-type (MAT) locus. Each mating type expresses a set of specific genes, which allows mating only between a and (types. In this respect the two yeast species are identical — however, they differ radically in how they regulate the expression of these antigens. In C. albicans, a-type genes are 'off' by default, and are turned 'on' in a-type cells; in S. cerevisiae, a-type genes are 'on' by default, and need to be repressed in α-type cells. The end result is the same — a-type genes are 'on' in a cells and 'off' in (cells — but the two yeasts use opposite means of achieving this pattern of expression. How did this transition take place, in the 200–800 million years since the two species shared a common ancestor? By combining microarrays (see image) and other bench experiments with comparative genomics in modern yeasts, Annie Tsong, Brian Tuch and colleagues have identified the cis and trans elements that were in place at the transition, and have traced the order in which the events took place.

A phylogenetic survey shows that the C. albicans style of activation is the ancestral state: in C. albicans and its ancestors, a-type genes are induced by an activator (a2) that is encoded by MATa. The challenge is to explain how this activator was lost in the S. cerevisiae lineage and replaced by a repressor — α2, which is encoded by MATα.

The authors started out by identifying the a-type genes that were specifically activated in the Candida lineage and then characterizing the cis-regulatory motifs that allowed a-type activation. By combining this information with knowledge of the mating-type circuitry in S. cerevisiae and the genome sequence of 16 ascomycete yeasts, the authors concluded that the transition from activator to repressor occurred as follows. First, regulation of a-type genes became independent of a2; in the ancestral state, Mcm1 (a protein in the MADS-box family) required a2 to activate a-type genes, but this dependence was lost owing to sequence changes in the cis-elements, which allowed Mcm1 to operate on its own. As Mcm1 is constitutively active, a-type genes would have been inappropriately expressed in (cells: this situation would have favoured a second step, in which α2 stepped in as an Mcm1 cofactor, thereby allowing the evolution of α2-mediated repression. This step would have required the conversion of a cis-sequence that recognizes a2 to one that recognizes α2, which the authors think could have occurred simply through changes both in cis and in trans.

Figure credit: Reproduced from Nature 443, 415–420 © (2006) Macmillan Publishers.

Beyond the immediate implications for the wiring of yeast mating systems, this study stands out because of the molecular resolution at which it has defined an evolutionary transition. Importantly, proper gene regulation seems to be maintained at all steps during the transition. Large evolutionary transitions can be ordered, stable affairs, as long as they involve coordinated interactions between proteins, and between proteins and DNA that do not compromise fitness.