a, Flow chart showing the image-processing pipeline for the cryo-EM data of RNAP–σA–CarD–RbpA–AP3 promoter complexes starting with 8,577 dose-fractionated movies collected on a 300-keV Titan Krios (FEI) equipped with a K2 Summit direct electron detector (Gatan). Movies were frame-aligned and summed using MotionCor229. CTF estimation for each micrograph was calculated with Gctf30. A representative micrograph is shown following processing by MotionCor2. Particles were autopicked from each micrograph with Gautomatch and then sorted by 2D classification using RELION33 to assess quality. The twelve highest-populated classes from the 2D classification are shown. The dataset contained 931,461 particles. A subset of particles was used to generate an ab initio map in cryoSPARC31. Using the low-pass-filtered (30 Å) ab initio map as a template, two rounds of 3D heterogeneous refinement were performed using cryoSPARC in a binomial-like fashion. Two major classes emerged, which were refined using cryoSPARC homogenous refinement and then sharpened for model building. b, Angular distribution calculated in cryoSPARC for particle projections. Heat map shows number of particles for each viewing angle. c, Gold-standard FSC41, calculated by comparing the two independently determined half maps from cryoSPARC. The dotted line represents the 0.143 FSC cut-off, which indicates a nominal resolution of 3.6 Å. d, FSC calculated between the refined structure and the half map used for refinement (work, red), the other half map (free, blue), and the full map (black). e, Top left, the 3.6 Å-resolution cryo-EM density map of RPo. Top right, cross-section of the structure, showing the DNA inside the RNAP cleft. Bottom, same views as above, but coloured by local resolution35.