Identification of Key Residues in Allosteric Signaling of Photoactivated Adenylyl Cyclase

Photoactivated adenylyl cyclases (PACs) convert ATP to cyclic AMP (cAMP) through long-range photoinduced allosteric communication between a BLUF domain and a distant adenylyl cyclase (AC) domain. Although photoactivation of the BLUF domain induces only minimal structural changes, it activates a chemical reaction about 4-5 nm away. Here, we combine molecular dynamics simulations, electronic structure calculations, network theory, and machine-learning approaches to investigate photoinduced allostery in the PAC from Beggiatoa sp. (bPAC). We observed that the photoexcitation enables electron transfer from a conserved tyrosine (Tyr7) to the flavin isoalloxazine ring, while the free energy of the electron transfer remains similar across active and inactive mutants. Therefore, photoinduced allosteric activity arises from conformational effects rather than changes in the electronic parameters. Using network theory and eigenvector centrality analysis, we identified residues relevant to allosteric pathways linking the BLUF and AC domains. Furthermore, we used machine-learning (ML) models to distinguish active and inactive conformational states without prior knowledge of functional residues. Remarkably, the ML models identified key regions known from network analysis. Together, these results provide a generalizable framework for understanding allosteric pathways in blue-light-sensitive proteins.