Confrontation of AlphaFold2 models with cryo-EM and crystal structures enlightens alternate geometries of the CYP102A1 multidomain protein

Abstract

Large range structural dynamics plays a critical role for the function of electron transfer proteins. This information is generally not available from crystallographic structures, while cryo-electron microscopy (cryo-EM) can provide some elements but frequently with a degraded spatial resolution. Recently, AlphaFold-based structural modelling was extended to the prediction of protein complexes. In this work, bacterial CYP102A1 from Priestia megaterium was used as a test case to evaluate the capability of AlphaFold2 to predict alternative structures critical for catalysis. CYP102A1 monooxygenase, a NADPH-supported fatty acid hydroxylase, works as a soluble homodimer, each monomer harboring two flavins (FAD and FMN) and one heme cofactors. Large conformational changes are required during catalytic cycle to allow successive electron transfers from FAD to FMN and finally heme iron. We used the recently released AlphaFold2_advanced notebook (AF2A), to predict the possible alternate conformations supporting electron transfers in CYP102A1 homodimer. Challenging AF2A-derived models with previously reported experimental data revealed an unforeseen domain connectivity of the diflavin reductase part of the enzyme. Intermolecular crossed complex constitutes a novel type of structural organization never previously described. The predicted formation within the dimer of a stable complex between the heme containing domains was challenged and found consistent with uninterpreted features of reported crystallographic structures and cryo-EM imaging. The particularly efficient CYP102A1 catalytic mechanism was revisited to the light of the new evidenced connectivity in which the FMN-binding domain of each monomer oscillates on themselves to alternatively receive and transfer electrons without needing large structural change in the dimer. Such model was found explanatory for previously contradictory reported biochemical data. Possibility to mimic CYP102A1 structural organization into bicomponent eukaryotic P450 systems was evaluated by designing and modeling in silico synthetic reductase domains built from composite sequence segments from P. megaterium and human origins. More generally, this work illustrates how the ability of AF2A to predict alternate complex structures can enlighten and explain conformational changes critical for bio-assemblies.