Ancient transcriptional regulators can easily evolve new pair-wise cooperativity.

Cells regulate gene expression by the specific binding of transcription regulators to cis-regulatory sequences. Pair-wise cooperativity between regulators-whereby two different regulators physically interact and bind DNA in a cooperative manner-is common and permits complex modes of gene regulation. Over evolutionary timescales, the formation of new combinations of regulators represents a major source of phenotypic novelty, facilitating new network structures. How functional, pair-wise cooperative interactions arise between regulators is poorly understood, despite the abundance of examples in extant species. Here, we explore a protein-protein interaction between two ancient transcriptional regulators-the homeodomain protein Matα2 and the MADS box protein Mcm1-that was gained approximately 200 million y ago in a clade of ascomycete yeasts that includes . By combining deep mutational scanning with a functional selection for cooperative gene expression, we tested millions of possible alternative evolutionary solutions to this interaction interface. The artificially evolved, functional solutions are highly degenerate, with diverse amino acid chemistries permitted at all positions but with widespread epistasis limiting success. Nonetheless, approximately ~45% of the random sequences sampled function as well or better in controlling gene expression than the naturally evolved sequence. From these variants (which are unconstrained by historical contingency), we discern structural rules and epistatic constraints governing the emergence of cooperativity between these two transcriptional regulators. This work provides a mechanistic basis for long-standing observations of transcription network plasticity and highlights the importance of epistasis in the evolution of new protein-protein interactions.