Simulations have played a key role in elucidating the dynamics of Ras proteins in the aqueous and membrane environments. We use classical and advanced molecular dynamics simulations to study the isolated catalytic domain in water, the lipid-anchor and the full-length Ras in bilayers of various lipid composition. Our current focus is on K-Ras, which is the most frequently mutated Ras isoform in human cancers and developmental disorders. Ongoing projects revolve around the question of how somatic and genetic mutations may alter the population of conformational states and oligomerization behavior of K-Ras, and how these may affect interaction with effectors or modulators.
Our work on membrane-organization of Ras proteins involves studying various lipid bilayer models using atomically detailed and coarse-grained simulations of Ras-membrane complexes. Our previous studies led to important insights into the physical basis for clustering and non-overlapping distribution of different Ras proteins in membrane domains. In particular, we found that the nature of lipid-modification dictates the distinct lateral organization of different Ras proteins on the plasma membrane, and that cholesterol modulates lipid domain stability and thereby the stability of Ras nanoclusters. We also found that surface-bound proteo-oligomers can be used to probe the mechanism by which membrane curvature might be generated and/or maintained. However, a number of technical challenges remain to be solved in order to accurately and fully model oligomers of surface-bound Ras and other lipid-modified proteins, which is the object of our current focus.
As part of a broader effort to developing anti-Ras therapeutics, we leverage insights emerging from the projects described above for the design of inhibitors that directly act on Ras. Our previous studies provided the initial clues about the allosteric nature of Ras and the role of conformational selection in its function. This led us to contemplate the potential druggability of Ras at the time when this was thought hopeless. Spurred in part by findings from large-scale genomic studies that the KRAS gene remains to be the main culprit in many forms of cancer, Ras is now back in the forefront of the search for new anti-cancer therapeutics. Our contribution to this effort includes the identification of novel allosteric ligand binding sites and prediction of small-molecule ligands that might bind to these sites. Moreover, working with cell biologists and pharmacologists, we showed for the first time that nucleotide exchange factors are required for oncogenic Ras signaling and inhibiting nucleotide exchange is a valid approach to abrogating the function of oncogenic mutant Ras. Our current effort in the search for isoform-selective Ras inhibitors includes developing methods to incorporate membrane into our dynamics-guided, ensembled-based drug-design scheme.
We would like to understand the physical basis for the ability of common non-steroidal anti-inflamatory drugs (NSAIDs), such as aspirin, to cause gastrointestinal injury through their detrimental effect on surface membranes. This effect appears to be exacerbated by bile salts and reduced or eliminated by conjugating NSAIDs with phosphatidylcholines (PCs). We study the surface behavior, morphology and size of micellar particles that appear during simulations of binary and ternary mixtures of NSAIDs and bile salts with different PCs. Initial results suggest that NSAIDs can form potentially cytotoxic aggregates with bile acids, and may alter the normal physiologic balance between bile acids and PCs .
We support all of our computational studies by the development of novel computational methods. These include techniques to expedite binding site identification and docking to ensembles of protein conformations both in solution and membrane environments. For instance, our method LIBSA analyzes multiple docked poses against a single or ensemble of receptor conformations and returns a metric for the relative binding to a specific region of interest. By using novel filtering algorithms and the signal-to-noise ratio, the relative ligand-binding frequency at different pockets can be calculated and compared quantitatively. Ligands can then be triaged by their tendency to bind to a site instead of ranking by affinity alone. Our most recent method, pMD-membrane, allows for the powerful probe-based molecular dynamics approach for the identification of druggable sites to be applicable to membrane-bound targets. pMD-membrane overcomes the negative impact of probes on bilayer structure by re-parameterizing selected pairwise interactions between probes and bilayer lipids. We have also developed a number of analysis tools that facilitate analysis of clusters of surface-bound proteins.