The Myxococcus xanthus encapsulin is a self-assembling HK97-family protein cage whose high multivalency makes it an attractive scaffold for vaccine, drug-delivery, and biomanufacturing applications. Although protomers typically form T1 and T3 icosahedra, coat-protein interactions are plastic, and mutations can redirect assembly toward alternative geometries that are usually overlooked because of their low abundance. We revisit a richly heterogeneous cryo-EM dataset of an encapsulin construct carrying a 20-residue SARS-CoV-2 receptor-binding-domain insertion in the E-loop, previously reported to adopt eight assemblies beyond T1 and T3.
To characterize this landscape more completely, we developed an iterative classification pipeline using accessible tools in CryoSPARC. Starting from 6,742 micrographs and 2.6 million picked particles, we grouped cleaned particles into morphology-based super-classes and applied repeated rounds of heterogeneous refinement, 3D classification “cleanliness checks,” and artificial references until no new species emerged.
This near-exhaustive approach revealed 37 unique geometries, including enantiomeric pairs, all built from a single protomer. Across these geometries, hexamer number ranges from 2 to 20 while the pentamer count stays fixed at twelve per Euler’s formula, and symmetries span asymmetric, rotational, dihedral, tetrahedral, and icosahedral, accommodated by up to ~30° of flexibility at dimer interfaces that preserves the underlying interaction chemistry. Geometries differing by a single hexamer are related by Endo-Kroto insertions analogous to fullerene rearrangements, supporting a stepwise single-protomer addition model of assembly. We also find that the abundance of each geometry correlates roughly with the distribution of strain within the cage.
By combining this with calculated shared intermediates between observed cages, we aim to build a more complete stochastic growth model that accounts for the full range of geometries in our data. Together, our results demonstrate that a single protomer can access a near-continuous spectrum of quasi-equivalent geometries, providing a foundation for rationally tuning engineered protein nanoparticles.
