Bacterial microcompartments (BMC) are nanoporous shell structures composed of self-assembled proteins that encapsulate enzymes. These proteinaceous organelles facilitate specific metabolic processes by sequestering reaction substrates, intermediates, and products to enhance metabolic efficiency and protect the bacterial cell from toxic intermediates. While there is interest in using BMCs as a platform for metabolic engineering, BMC engineering remains difficult because the molecular factors that control macroscopic assembly outcomes are presently unknown. Studies have shown that the expression of different BMC proteins alters the final assembled structure. When the hexamer protein (BMC-H) of Haliangium ochraceum is solely expressed, rosette-like structures are formed. Conversely, the pseudo-hexamer protein (BMC-H2), a tandem copy of BMC-H connected by a peptide linker, assembles into 25 nm icosahedral shells. Herein, we use molecular dynamics to elucidate the molecular factors of BMC-H and BMC-H2 that dictate observed differences in self-assembly behavior. We employ enhanced sampling methods with atomistic molecular dynamics to compute potentials of mean force describing protein-protein association along different collective variables, including those learned using data-driven techniques. We also develop and apply bottom-up coarse-grained models to simulate BMC-H and BMC-H2 assembly and isolate the molecular determinants that lead to their respective assembled morphologies, which we determine is due to the subtle impact of the linker on intramolecular conformations. Finally, we demonstrate how engineering perturbations to the linker allows greater exploration of possible morphological states spanning hexameric sheets, nanotubes, and icosahedral shells. We anticipate that our approach can be expanded to investigate the growing family of known BMC shell proteins.
