Computational study of Liquid-Liquid Phase Separation in Biomolecular

Liquid-liquid phase separation (LLPS) is a critical process in cellular organization, forming biomolecular condensates that regulate essential cellular functions and are implicated in diseases such as neurodegeneration, cancer, and viral infections. Because of its crucial role, it has recently attracted the attention of researchers, especially in computational science.

1. Development of new method to study LLPS

Despite significant efforts by the scientific community, constructing accurate temperature-density phase diagrams for LLPS remains a challenging task. One of the goals of this project is to develop a new, efficient and accessible protocol for constructing LLPS phase diagrams in protein and protein-RNA/DNA systems using molecular dynamics simulations. By integrating temperature replica exchange molecular dynamics with advanced reweighting techniques, the protocol will allow us to obtain spinodal and binodal curves, including temperatures close to the critical point.
To construct the spinodal and binodal curves of the phase diagram for LLPS, we will calculate the heat capacity Cp. If the E curve is smooth enough, we can directly obtain heat capacity from derivative of E with respect to temperature. To determine temperature-density (T-ρ) phase diagram, we need to calculate Tspi and Tbin at fixed ρ. Peaks in derivative of Cp withs respect to T indicate phase transitions, corresponding to the spinodal Tspi and binodal Tbin temperatures at a fixed density. We will employ temperature replica exchange molecular dynamics to enhance conformational sampling, ensuring a smooth Cp curve by simulating multiple temperature replicas, enabling accurate differentiation. To achieve continuous thermodynamic properties across a temperature range, we will combine T-REMD with reweighting technique. We will test our method on the well-studied van der Waals liquid system and then apply it to other biomolecular systems.
The source code implementing our method will be made publicly available under an open-source license, enabling free use and adaptation by the scientific community for future research.

2. Influence of lasso entanglement on LLPS

A protein entanglement involves the intertwining of protein backbone loops and segments leading to one of two broad outcomes: a knot – in which pulling on both termini of the protein tighten the knot preventing the full extension of the protein; or a lasso entanglement – in which pulling on both termini disentangle the loops and the backbone can be fully extended. Knots are rare in the protein database, accounting for less than 2% of the total number of deposited structures, while around 70% of globular proteins are recognized to contain non-covalent lasso entanglements. There is, therefore, the potential for fundamental discoveries in this understudied structural space.
Our knowledge about the biological consequences of lasso entanglements of native proteins is very limited. The presence of lasso entanglements in the native state of proteins influences their folding times and co-translational folding. Near-native misfolded states with non-native entanglements interact with chaperones in a similar way to the native state, and that helps these misfolded states bypass the chaperones. Non-native entanglement that very slowly interconvert to the functional native state, can ultimately affect the specific activity of enzymes. It has recently been discovered that this non-covalent entanglement may influence the human aging process. Thus, it would be interesting to study its possible influence on LLPS.
In this project we will study the effect of native entanglement on LLPS. Since intrinsically disordered proteins (IDPs) undergo liquid-liquid phase separation (LLPS) unlike native proteins, we will consider a class of proteins that do not fold into a well-defined three-dimensional shape but adopt an ensemble of interconvertible conformations. More precisely, we will study the influence of entanglement arising in the most populated state of IPD on LLPS (structures of this state can be found in various databases). Sequences of the same length, but with and without native-like entanglement, will be selected for the study using the new approach proposed in this project. We predict that lasso entanglement will reduce propensity to LLPS.

3. Influence of mutations on LLPS

The abundance of IDPs in the cell is tightly regulated to ensure precise functioning, mutations in IDPs, or changes in their cellular abundance associated with diseases, such as Alzheimer’s, Parkinson’s, type II diabetes, cancer, and so on. Despite their cellular importance and clinical significance, the role of mutations on LLPS have not been explored. This project will address this issue with a focus on the assembly of human islet amyloid polypeptide (hIAPP), which is one of the pathological hallmarks of type 2 diabetes. By calculating phase diagrams and analyzing changes in molecular interactions among monomers within condensates, we aim to elucidate the molecular mechanisms governing LLPS dynamics in these systems.

4. LLPS in protein-RNA complexes

The emergence of SARS-CoV-2 caused one of the most disaster pandemics in human history. The nucleocapsid (N) protein of SARS-CoV-2 is observed that its complex with RNA in the virus core undergoes LLPS. A segment of designed RNA which interferes the LLPS of nucleocapsid protein and RNA of virus can inhibit the viral replication. Interestingly, the phosphorylation of N protein regulates two dual roles of N protein. Wildtype N protein binds strongly to RNA in gel like droplets when LLPS, these droplets form the nuclear of virus. When N protein is phosphorylated, the complex of N protein-RNA is loosen facilitating RNA is freely from N protein binding and moves into double membrane vesicle, in which the replication of RNA is performed. All-atoms simulation study suggests that the phosphorylation of N protein weakens interaction with RNA. However, the molecular mechanism of phosphorylation on LLPS of N protein-RNA and dual role of N protein is still ambiguous. We will address this problem using coarse-grained Martini models, which allow the study of liquid-liquid phase separation with a large number of proteins. It should be noted that although our study concerns the N-protein-RNA complex of the SARS-CoV-2 virus, its qualitative results should be applicable to other viruses as well.
The computational results obtained through this project will not only shed light on existing experimental data but will also promote future experiments.