Molecular-level observation of the self-assembly of a virus-like particle

Biomolecular assembly is a cornerstone of cellular organisation and function. Revealing its underlying principles is essential for understanding biological processes, and their malfunction in disease. Viral capsid assembly is the archetypal self-assembly system that has been central in establishing the fundamental principles of biological self-assembly, providing a conceptual and geometric framework that underpins the current understanding of supramolecular biomolecular systems, the development of new biomaterials, and advancing therapeutic design. Yet, despite decades of experimental efforts, observation and quantification of virus self-assembly pathways and dynamics have remained elusive. Here, we combine mass photometry with a non-perturbative single molecule trapping method, enabling direct, real-time monitoring of the self-assembly of individual virus-like particles (VLPs) with molecular resolution. We show that weak, diffusion-limited, and reversible multivalent interactions control the assembly process by facilitating stochastic selection of a limited set of on-path, topologically closed intermediates. Assembly is finely tuned by the transition rates between these topologically closed configurations and proceeds through a sequence of effectively irreversible first-passage events. The intrinsic first-passage times are a consequence of VLP symmetry, creating a separation of timescales between the formation of the first closed intermediate and subsequent elongation. This separation results in a nucleation and growth mechanism that yields an equilibrium distribution consistent with the law of mass action, despite the overall irreversibility of assembly. Our approach enables direct and complete characterisation of both the thermodynamics and the kinetics governing VLP assembly and reveals how the system achieves specific assembly of one final structure with high fidelity despite the availability of thousands of assembly intermediates. More broadly, our approach provides a general framework for visualising and quantifying the dynamics of multimeric biological machines at the molecular level.