Understanding protein folding, evolution and function via molecular simulation

National Institute Of Diabetes And Digestive And Kidney Diseases

Investigators

Linked publications, trials & patents

Abstract

The project has addressed the following areas in the past year: 1. Association of highly charged intrinsically disordered proteins and their complex coacervation. Recent work in collaboration with Ben Schuler's single molecule FRET group in Zurich has shown that high affinity disordered complexes of proteins or proteins and nucleic acids of opposite charge may be ubiquitous in cell nuclei. We have previously shown that this mode of association allows a competitive substitution mechanism that speeds unbinding, e.g. allowing protein chaperones to release histones from nucleosome complexes (1). We are now seeking to develop a predictive model for the affinity and structure of these complexes, informed by molecular simulations of a variety of sequences, as well as experimental data. Using novel multiscale simulation methodology, are also performing all-atom simulations of a complex coacervate of two nuclear proteins in order to elucidate the interactions responsible for stabilizing these phases, and we are also using new experimental data from FRET experiments and osmotic pressure measurements to help refine the force fields for salt bridges in the proteins as well as interactions of the proteins with ions (M. Ivanovic). 2. Development of coarse-grained models for complex coacervation of intrinsically disordered proteins with single- and double-stranded nucleic acids. Going beyond the 1:1 complexes studied in project 1, it is also possible for oppositely charged macromolecules such as proteins and DNA or RNA to undergo complex coacervation, forming a separate phase with high macromolecular density, under the correct conditions. Such a phenomenon may provide a physical basis for the formation of some of the membraneless organelles observed in the cell nucleus. We have developed a coarse-grained simulation model of protein-nucleic acid interactions, and used it to study the ordering induced on formation of the condensed phase. We are currently extending this model to include sequence-specific effects as well as back-mapping to atomistic simulations in a multiscale approach (3). (K. Lebold) 3. Development of transferable sequence-specific models for liquid-liquid phase separation (LLPS) of intrinsically disordered proteins. We had previously shown that a comprehensive refitting of the parameters of our coarse-grained the energy function for proteins was able to better describe the both the properties of isolated disordered chains and also those of proteins which are known to phase separate; however that work relied on data from disparate sources and collected under different conditions. We have now used a FRET data set on a set of proteins with diverse sequences collected under identical conditions in the Schuler lab to further refine the potential, resulting in significant improvements. We are also extending this model to capture specific structure formation (e.g. amyloid) to model possible aging processes in droplets and developing models for cosolvent and Hofmeister effects which are sometimes of interest experimentally. (T. Dannenhoffer-Lafage) 4. Using sequence-based energy functions to describe protein fitness landscapes and for protein design. We have shown that it is possible to design novel foldable protein sequences using coevolutionary models. Most recently we have found that we can achieve increased thermostability via this route. We are also seeking to design sequences which can fold into two different structures as envisaged in our recent theoretical work, and we are collaborating with Susan Marqusee's group to test some of these ideas (P. Tian). We are also looking to develop similar ideas to identify proteins which naturally switch folds (such as RfaH), using sequence information. Separately, we are exploring the use of machine learning to predict fitness landscapes for proteins of any given structure (L. Frechette, D. Wang). 5. Modelling properties of the extracellular matrix using coarse-grained models. We have recently started developing bottom up coarse-grained simulation models to describe the extracellular matrix and how they are related to its underlying molecular structure. We have initially focused on developing accurate atomistic models for collagen, and validating them against the available data from NMR and small angle X-ray scattering; we are extending this to coarse-grained models that will allow the investigation of packing of tropocollagen in collagen fibers. (G. Pantelopulos). 6. Inferring interactions between proteins and cosolutes using NMR data. In collaboration with Yusuke Okuno and Marius Clore, we have been using simulations in conjunction with experiments to analyze association of denaturant molecules with proteins using spin-labelled denaturants. For the first time we have shown directly that the denaturant mechanism involves primarily interactions with the unfolded state. We are now extending this work and developing all-atom models for the spin-labelled cosolutes (T. Dannenhoffer-Lafage). 7. We are working in collaboration with Steve Vogel to understand the mechanisms in certain fluorescent proteins that allow coherent energy transfer via models parameterized from molecular simulations. (G. Taumoefolau). Group members or jointly supervised external collaborators involved in each project are listed at the end of each section.