Research Overview

Simulations are important for interpreting experimental results, generating new hypotheses, and designing new experiments. However, the non-classical behavior of electrons makes chemical reactions difficult to simulate. To sidestep the limitations imposed by the computational burden of quantum calculations, we have developed an approach in which valence electrons are modeled as semi-classical particles (aka "Lewis dots"") that interact with each other and with atomic kernels via pairwise potentials that are modified from Coulombic to take wave properties into account.

For compounds with integral bond orders, we had some success with conventionally paired electrons (shown here in green). LEWIS, our heuristic sub-atomistic force field describes the polarizability, amphiproticity and hydrogen bonding of water in a balanced, intuitive and highly efficient manner, thereby facilitating large and long simulations of the solvation and dynamics of hydroxide and hydronium in the bulk, at surfaces, and in water chains (illustrated here by Grotthuss transport). The model provides an account of the negative surface charge of neat water, the kinetics of neutralization of aqueous hdyroxide and hydronium, and two-stages in the auto-ionization of water.

For transferability to other environments, our more recent work assigned the electron pairs a degree of freedom corresponding to spread, in addition to the three classical degrees of freedom corresponding to location. With potential forms guided by analytical results for electrons occupying floating spherical Gaussian orbitals (FSGOs), the spread-dependent pairwise potentials (dubbed LEWIS-B) are able to describe the behavior of valence electron pairs in hydrocarbons, including those in single, double, bridge and bent bonds of linear, branched, cyclic, cationic and anionic compounds. In addition, the new sub-atomistic force field efficiently simulates carbocation addition to a double bond and cation migration to a neighboring carbon.

Other work modeled the valence electrons singly, taking spin into account, in addition to spread. An initial, heuristic force field, LEWIS·, provided a good account of the non-monotonic variation of ionization energies and spin excitation energies across rows of the periodic table, while also predicting the monotonic variation of atomic polarizabilities. Extension to diatomic species described both bond orders and magnetic properties.

Most recently, a deconstruction of the above FSGO-inspired LEWIS-B force field allowed assignment of different potentials for interactions between electrons with like and unlike spins. Although trained exclusively on data from species with paired electrons, the electrons exhibit Linnett-like "double quartet" behavior when a sparser bonding environment allows them to separate. Under this force field (dubbed LINNETT), electrons in ethyne and benzene (compounds with only one H per C) pair up in the CH bonds, but are splayed in the CC bonds.

Gas vesicles present a particularly intriguing protein folding problem. Their thin shells must withstand considerable pressure while presenting a highly hydrophobic interior surface to prevent the condensation of permeating water vapor. But it is difficult to imagine how such a shell is created by the self-association of many small (7 kDa) subunits. Gas vesicles also have significant similarities with amyloid fibers: they are not dissolved by detergents and beta-sheet is a prominent, conserved feature. Yet, the host organisms can recycle the subunits, indicating that they are able to reversibly disassemble the gas vesicles. The potential for controlling amyloid pathologies might be enhanced by understanding how aquatic organisms manage their gas vesicle subunits. To determine the structure of gas vesicles, we employed multi-dimensional solid-state NMR methods. Resonances were assigned by homo- and hetero-nuclear correlations, and structural constraints were obtained by measuring dipolar interactions that reflect internuclear distances. Results pointed to an asymmetric dimer as the fundamental bulding block of the vesicles in the deep water organism Anabaena flos aquae. This polymorphism allows the beta strands to be oriented at the magic angle relative to the vesicle axis, a feature that is thought to account for the resistance of these vesicle to collapse under hydrostatic pressure. No investment in polymorphism was observed in the relatively weak gas vesicles of Halobacterium salinarum, an organism that inhabits the shallows of salt flats.

Many micro-organisms produce retinal containing membrane proteins similar to the mammalian visual pigments. These rhodopsins include energy transducers that use light to drive ion transport, as well as signal transducers that use light to stimulate phototaxis. The rhodopsins are thus light-driven analogues of the chemically-driven energy transducers (membrane ATPases) and signal transducers (hormone receptors) found in mammalian cells. The first rhodopsin to be discovered in a unicellular organism (known therefore as bacteriorhodopsin) undergoes a photocycle in which an ion is released on one side of the membrane and replaced from the other side of the membrane. Thus, light is used to create an electrochemical potential gradient across the membrane that the cell can use to drive other processes. To study bacteriorhodopsin in the native membrane, we employed solid-state NMR methods that achieve the high resolution of solution spectra while preserving the three-dimensional information of powder spectra. The results were interpreted empirically using data from model compounds and theoretically via quantum mechanical calculations. Specifically, we probed the detailed features of the photocycle intermediates that enforce unidirectional ion transport. Results pointed to the release of photon-induced torsion in the chromophore as responsible for the switch in hydrogen bond connections that prevents ion backflow. They also identified the primary proton acceptor in the photocycle and suggested that bacteriorhodopsin may be an inwardly-driven hydroxyl pump rather than an outwardly-driven proton pump.

Scenarios for the origins of life begin with the abiotic formation of biologically significant molecules from simple inorganic precursors. Much of our current understanding of this prebiotic chemistry relates to small molecules. However, polymers (as in the insoluble "gunk" accumulated in the famous Miller-Urey experiment) are also thought to have played significant roles. For example, HCN, which is thought to have been abundant in the prebiotic environment, spontaneously forms polymers that have been the focus of much speculation. Using the unique probes afforded by solid state NMR, we identified three different types of HCN polymers, all of which are unlike any of the structures that have been previously proposed. Sugars, another large class of prebiotically abundant molecules, also form polymers readily under conditions that form furan and pyrrole units. Here again, solid state NMR hshowed that the linkages between the heterocycles are not as previously proposed.

Cytoskeletal proteins have been the subject of intense experimental scrutiny. Of particular interest is the molecular basis for the spatial organization of cytoskeletal fibers in cells. Typically it is assumed that protein polymers should be randomly disposed in solution and that the non-random organization of cytoskeletal fibers must be due to the effects of various accessory binding proteins. However, the cytosol is a very crowded, and therefore highly non-ideal, solution. Under these circumstances, the theory of liquid crystals tells us that elongated particles may spontaneously align, coalesce into bundles, and form gels. We adapted these theories to heterogeneous systems with self-assembling fibers, in order to characterize the various mechanisms by which cells can control the spatial arrangement of cytoskeletal elements. We found that, under physiological conditions, long filaments are not only predicted to spontaneously form bundles, but these bundles will be spontaneously segregated according to the diameters and flexibilities of the filaments. The function of cross-linking by "bundling proteins" therefore appears to be only to fine tune the bundles (e.g., as to polarity and registration). The theory also predicted that cells can prevent crowding-induced bundling by using its capping proteins to reduce the lengths of the filaments and can frustrate bundling and form gels by using other accessory proteins that cross-link filaments in orthogonal configurations.