Lattice Models for Modeling Sequence-Specific Phase Behavior of Low Complexity Domains

Abstract

Low complexity domains (LCDs), enriched in charged, polar, and aromatic residues are drivers of aggregation and phase separation in a variety of proteins including those that drive the formation of RNA-protein granules. Polyglutamine (polyQ) tracts are a distinct archetype of LCDs. Mutational expansion of polyQ tracts in specific proteins is associated with onset and progression of at least nine different neurodegenerative disorders including Huntington's disease. The latter is associated with an expanded polyQ tract in the protein huntingtin. Recent experiments have uncovered complex thermodynamic and kinetic phase diagrams for N-terminal fragments of huntingtin. These phase diagrams are modulated by polyQ length, sequence contexts, concentration, and solution conditions. The details of the molecular and supramolecular interactions that give rise to emergent phase behavior of polyQ containing at distinct length scales remains poorly understood. Computations play an important role in predicting the phase behavior of LCDs. These calculations require simulations encompassing several hundreds of thousands of protein molecules spanning several orders of magnitude in protein concentrations. Such simulations are beyond the scope of all atom models. Since phase behavior results from the synergistic interactions amongst small numbers of collective coordinates, simulations designed to uncover sequence-encoded phase behavior can be pursued using either systematic coarse-graining or phenomenological models. Recent work has demonstrated the utility of both approaches. However, these models lack the requisite molecular detail to enable a complete understanding of sequence-specific contributions that determine similarities and differences in sequence-encoded phase behavior of LCDs. Here we present a lattice model that is tailored for modeling sequence-specific phase behaviors of LCDs. The model serves as an intermediate description between ultra coarse-graining and all-atom extremes. In the lattice model, a single bead on a 3-dimensional lattice site describes each residue. Each bead has multiple faces and this helps us capture the directionality of backbone hydrogen bonding and sidechain-specific interactions. Monte Carlo simulations based on these lattice models help us uncover the interplay amongst different sequence-encoded energy scales and phase behaviors of LCDs with polyQ tracts. Specifically, we focus on explaining the role of spherical mesophases as mediators of fibril formation in polyQ containing proteins. We also show that the model can be deployed for understanding the interplay amongst liquids, fibrils, and gels for other LCDs.