Molecular Dynamics Simulations of Viral Capsids at Constant pH
Cowpea chlorotic mottle virus (CCMV) is a plant-based virus that specifically infects the cowpea plant. The capsid of CCMV is composed of a single ∼190 residue protein monomer which assembles into hexamers and pentameters to form an icosahedron. CCMV capsid assembly optimally occurs at low pH and the capsid disassembles at higher pH (>7) and salt concentrations. Interestingly, the capsid can also "swell" at more moderate pH, a response that is likely related to nucleic acid release during infection. Because its physical properties are increasingly well-characterized and tunable, CCMV has recently become a popular platform for macromolecular engineering, including use as a vector for vaccination. Nonetheless, the specific atomistic details giving rise to these phenomena are not completely understood and cannot be clearly inferred from simulations using simple phenomenological models. The simulation studied proposed here is a clear first step toward a molecular description of CCMV and will specifically address how to interpret the details of experimental results. These insights can eventually be used to address the broader challenge of capsid engineering.
While phenomenological models have been extremely successful in broadening our understanding of capsid structure, mechanics, and assembly, they inevitably ignore—or make assumptions regarding—chemical details. Nonetheless, even conventional all-atom models must make explicit assumptions regarding the protonation state of ionizable groups and how they are affected by pH. Generally, protonation states of a biomolecular system must be fixed at the outset, a fundamental limitation of classical simulations.
In order to address this issue, we have recently proposed a novel simulation algorithm designed to account for pH effects in a rigorous and efficient manner. The method, inspired by a scheme originally proposed by Stern, consists of carrying out short (∼50 ps) nonequilibrium molecular dynamics (neMD) switching trajectories to generate physically plausible configurations with altered protonation states. These are subsequently accepted or rejected according to a Metropolis Monte Carlo (MC) criterion. The overall approach was recently improved by including the introduction of a more efficient momentum reversal procedure, a generalization of the MC criterion to account for the nonequilibrium work, and a novel transition operator splitting that can significantly reduce the computational effort due to the switching procedure. The hybrid neMD constant pH method is well suited for explicit solvent systems with large numbers of ionizable sites. Scalability to a large number of ionizable sites and the ability to deal with explicit solvent are perhaps the two most critical features of a truly useful constant-pH simulation method for large-scale biomolecular systems such as virus capsids.
While our neMD constant pH algorithm has recently been implemented with the NAMD molecular dynamics package, it has not yet been deployed on large scale computing infrastructures and so its use has hitherto been limited to smaller systems (∼10,000-50,000 atoms). Use of the Blue Waters com- puting environment is critical in this regard, as it will allow the first proof-of-concept simulations at truly large scales. We can confidentially move forward in this task thanks to the exceptional track record of Blue Waters in supporting near record breaking speed and scaling with NAMD. The algorithmic, soft- ware, and hardware advances combined in this proposal are thus quite exceptional and represent a unique effort towards understanding the chemistry of viral capsids.