Modeling of Intrinsically Discorded Protein Mimicking Hybrids for Membraneless Organelles Formation

Dynamic assembly and formation of membraneless organelles (MLOs) via liquid-liquid phase separation (LLPS) is a versatile mechanism for the spatiotemporal regulation of living cells. These MLOs actively regulate biochemical reactions and are enriched in various biomolecules to improve cell functions. Due to these functionalized abilities of MLOs, there has recently been an increasing trend to develop synthetic biomaterials that mimic natural MLOs. We have developed a multi-scale computational framework to predict the LLPS behavior of both Intrinsically Discorded Proteins (IDPs) and IDP mimicking hybrids developed in our lab extracting interactive motifs from FUS, a widely studied model for MLOs.<br/>Our computational model draws essential inspiration from the sticker-spacer representation described in associative polymer theory. The potential stickers are identified using the LARKS (low complexity aromatic rich kinked segments) identifier followed by molecular dynamics (MD) and Monte Carlo (MC) simulations to capture the interaction energies and represent the stochastic nature of MLOs, respectively.<br/>First, we identified 18 LARKS available in FUS protein using LARKSdb as potential stickers, and in-between peptide segments are considered spacers for modeling purposes. Molecular docking, followed by MD simulations for stickers, demonstrated that 13 stickers could form transient beta-sheet-like structures, and others can only form random coils. Further, their suitability for stickers was confirmed by the radius of gyration (6.18 ± 0.6 Å) and solvent-accessible surface area (7.3 to 12.0 nm2). Using the molecular mechanics Poisson−Boltzmann surface area (MMPBSA) method, the estimated pairwise interaction energies between stickers ranged from ~100 kJ/mol. All spacers were enriched in over 70% of disorder-promoting residues and had less than 30% of charged residues at physiological pH conditions by employing the CIDER webserver.<br/>Next, we generated lattice Monte Carlo simulations to describe the stochastic movements of IDPs under physiological conditions incorporating interaction energies found above. IDP chains were configurated as strings of coarse-grained beads and randomly placed in a three-dimensional simple cubic lattice as a canonical ensemble. We performed 106 orders of hypothetical Monte Carlo movements as self-avoiding random walks on the lattice to obtain an equilibrated system using the Metropolis-Hastings algorithm as movement acceptance/rejection criteria. We could qualitatively demonstrate the dynamic assembling and disassembling behavior of the system. We also used the radial distribution function to identify the system's phase boundaries and estimate the degree of LLPS propensity.<br/>FUS-full protein, FUS-prion-like domain, and FUS-RNA binding domain separately found that FUS-full protein has the highest propensity to undergo LLPS and FUS-RBD domain has the lowest propensity to undergo LLPS while FUS-PLD has an in-between propensity. These findings have a good agreement with available experimental observations. Hence the model was validated. Then we applied the computational framework to IDP mimicking hybrid having only two stickers from FUS protein incorporated with dextran as spacers. Our model could successfully track the synthetic hybrid system's dynamic assembly and disassembly behavior. These computational estimations were also verified by turbidity testing and confocal microscopic observations.<br/>In summary, our approach will provide a comprehensive theoretical framework to predict the LLPS behavior of IDPs and IDP-mimicking hybrids. It also provides a fast and economical means to guide the design of such systems to regulate the LLPS process efficiently.<br/><br/>The authors would like to thank the funding support from the Hong Kong Research Grants Council (GRF 16102520). M.K.S.Fernando receives financial support from the Hong Kong Ph.D.Fellowship Scheme.