Combination of deep XLMS with deep learning reveals an ordered rearrangement and assembly of a major protein component of the vaccinia virion

Vaccinia virus, the prototypical poxvirus and smallpox/monkeypox vaccine, has proven a challenging entity for structural biology, defying many of the approaches leading to molecular and atomic models for other viruses. Via a combination of deep learning and crosslinking mass spectrometry (XLMS) we have developed an atomic-level model and an integrated processing/assembly pathway for a structural component of the vaccinia virion, protein P4a. Within the pathway, proteolytic separation of the C-terminal P4a-3 segment of P4a triggers a massive conformational rotation within the N-terminal P4a-1 segment that becomes fixed by disulfide-locking while removing a steric block to trimerization of the processing intermediate P4a-1+2. These events trigger the proteolytic separation of P4a-2, allowing the assembly of P4a-1 into a hexagonal lattice that encloses the nascent virion core.

Methods
Protein structure prediction, de novo: Monomer and multimer structure predictions for vaccinia protein P4a (precursor and processing products) were run on local installations of AlphaFold2 and AlphaFold-multimer, using a non-docker setup (https://github.com/kalininalab/alphafold_non_docker). Models uploaded here represent the top-ranked prediction for each protein or protein complex.

Predicted models were validated by XLMS detected for P4a, from intact and uncoated virions, which is described in detail in Mirzakhanyan and Gershon, PLOS Pathogens, 2019. Solvent-accessible surface distances between crosslinked residues were calculated using Topolink (A. J. R. Ferrari et al, Bioinformatics, 2019).