Extended Data Fig. 3: Single-particle cryo-EM processing pipeline for the Immature (PR-) MA lattice.

PR- virus particles consist of an uncleaved Gag polyprotein. The MA lattice is in an immature configuration. a, EM micrograph pre-processing, particle picking and particle extraction. Micrographs were motion corrected using MotionCor2 in Relion and CTF estimation was performed using patch CTF estimation in cryoSPARC. Picking was performed using crYOLO, where a model was trained to pick arbitrary positions across the entire virus surface. b, Two rounds of 2D classification were used to identify high quality particles containing the immature CA layer. After the first round, the particles were separated into two groups - top and side views. In the second 2D classification, these two groups were classified independently. The best classes were selected and passed to heterogeneous refinement. c, Heterogeneous refinement was performed to identify the highest-quality immature CA particles, which were selected for further processing. d, Several rounds of CTF refinements, local defocus refinements, Ewald sphere corrections, local refinements and one iteration of Bayesian polishing in Relion 4.0 were performed to obtain a high-resolution immature CA map (2.37 Å). Within the resulting reconstruction, diffuse but recognisable density corresponding to the immature MA layer above CA was observed. e, Refinement of the immature MA layer. To generate a reconstruction focused on the MA layer, the CA particles were symmetry expanded and the centre of the box was first shifted from the immature CA lattice to the three-fold symmetry centre of the diffuse MA layer. Particles were re-extracted with the particle centred on the MA layer while the strongly featured immature CA layer was subtracted. Then, three iterations of heterogeneous refinements of the MA lattice were performed to identify particles with ordered MA lattice, followed by non-uniform refinement. The fact that MA forms a 2D lattice was then exploited to add particles at the expected positions of symmetry related neighbours, thereby expanding the set of particles (lattice expansion). At this stage, the particles are centred on the three-fold symmetry axis. To perform lattice expansion, for each particle, three symmetry-related particles were created by symmetry expansion and were assigned positions and angles corresponding to the neighbouring three-fold symmetry axes (as determined from the 3D reconstruction), followed by centering of the box to the neighbouring 3-fold symmetry axes of MA (lattice expansion). We found that this improved the resolution of our reconstruction and the quality of the map. That was followed by another lattice expansion and by 3D classification without angular search, followed by another round of lattice expansion and a final 3D classification. The best-looking class showing secondary structure features was selected and was refined using a local refinement with imposed C3 symmetry, resulting in a map of immature MA lattice with a resolution of 5.83 Å.