Direct cryo-ET detection of native SNARE and Munc13 protein bridges using AI classification and preprocessing

Synaptic transmission requires Munc13 and SNARE proteins for synaptic vesicle priming and fusion, and cryo-electron tomography detected multiple types of Munc13-or SNARE-dependent dependent molecular bridges that tether synaptic vesicles to the presynaptic active zone plasma membrane. To integrate the molecular scenario with structural observations, we obtained de novo, in situ cryo-electron tomography averages of native, mammalian SNARE-dependent and Munc13-dependent synaptic vesicle tethers. These provide direct evidence that both Munc13 and a complex comprising SNARE proteins link synaptic vesicles to the active zone plasma membrane. Furthermore, we determined the plausibility of different molecular compositions of tethers, placed constraints on their conformations and positioning, and propose the ex-istence of a complex downstream of Munc13 and upstream of SNARE complex formation. Because the detection and subtomogram averaging of membrane-bridging complexes is complicated by the presence of two lipid membranes and multiple protein species and conformations, we developed novel preprocessing methods and feature-based AI classifiers that outperformed the standard methods.