Cryo-EM data collection and processing.

Initially, 2000 micrographs were collected and manually inspected to remove any affected by drift, poor CTF fits, or low-quality ice, resulting in a total of 1856 micrographs. Particles were picked through Laplacian auto-picking, followed by 2D classification and 2D class average-based template picking on all 1856 micrographs. Consecutive 2D classification was used to eliminate picking contaminants and sort out suboptimal classes. An initial model was generated from the resulting 351,941 particles, and a 3D classification with image alignment was performed to find the most populated state. Particles contributing to this state underwent subsequent consensus refinement without symmetry, followed by post-processing, which yielded an 8.3 Å volume. To obtain an improved single-channel volume, focused refinement was performed on the two best-looking channels, leading to a 7.7 Å volume more clearly indicating a homo-heptameric assembly per MscS2 channel.

(TIF)