Extended Data Fig. 3: SV2A-8783-UCB-2500 cryo-EM data and processing.

A representative micrograph is shown with a scale bar of 120 nm. Two datasets were collected, composed of 24,211 and 21,528 micrographs, respectively. After patch motion correction and CTF estimation, a blob picker was used to pick particles. Particles were classified in 2D and an ab-initio reconstruction was used to generate templates. The resulting templates were used to re-pick particles from each dataset, resulting in a total of ~10 million particles for each dataset, which were re-extracted and binned to a smaller size to facilitate data processing. Heterogenous refinement generated several ‘decoy’ volumes and one volume with SV2A features. Particles from each dataset were combined at this stage, analyzed by several rounds of heterogenous, non-uniform, and local refinement. The resulting particle stack was processed in Relion using Bayesian Polishing, then processed by further rounds of 2D classification and CTF refinement. Local refinement with masks focusing on either the transmembrane region (3.3 Å final resolution) or the luminal domain (3.8 Å final resolution) was used to produce the final maps, which were sharpened with DeepEMhancer to improve map interpretability.