Interactions of Proteins with Membranes: Insights from Molecular Dynamics Simulations

Membrane proteins play a crucial role in numerous biological processes, including signal transduction, molecular transport, and cell recognition. Due to their involve- ment in fundamental physiological functions and their relevance as pharmacological targets, understanding their structure and dynamics is of paramount importance. However, experimental techniques such as X-ray crystallography and cryo-electron microscopy often struggle to capture the full conformational landscape of membrane proteins due to their complex interactions with lipid bilayers and their dynamic nature. These limitations make computational approaches, particularly molecular dynamics (MD) simulations, invaluable tools for elucidating the mechanisms under- lying membrane protein function. MD simulations, indeed, by integrating physical models with statistical mechanics, provide a powerful computational framework to investigate biological systems at atomic or – when more suitable or computationally advantageous – at coarser resolution. Enhanced sampling techniques further ex- pand the ability of MD simulations to overcome energy barriers, enabling the study of slow and rare events, like large conformational changes or to estimate free energy variations. In this thesis, I applied atomistic MD simulations, along with protein docking and structure prediction methods, to investigate protein-membrane systems of relevant medical interest, bridging experimental results at the macro-scale with a micro-scale description offered by computer simulations. 1. Pore-forming toxins Pore-forming toxins (PFTs) are a class of proteins produced by various organisms, including bacteria, fungi, and marine animals, that disrupt cellular membranes by forming transmembrane pores. These toxins play a crucial role in pathogenesis and defense mechanisms by inducing cell lysis through membrane permeabilization. PFTs typically undergo a multi-step process, including binding to the membrane, conformational changes, oligomerization, and pore formation. The structure and stoichiometry of the final pore can vary depending on the toxin and the lipid com- position of the target membrane. Unveiling the steps underlying pore formation can be instrumental for the structure-based development of new antivirulence agents targeting these toxins. Moreover, due to their ability to selectively interact with specific membrane components and to form pores, PFTs have raised interest for potential biomedical applications, e.g. immunotoxin therapies and targeted drug delivery; as nanopores for protein/DNA sequencing. Lastly, despite the variety in dimension, stoichiometry, structure, or number of protomers involved in pore forma- tion, PFTs share features concerning the mechanism of pore formation. Therefore, PFTs pose as optimal targets for studying protein-membrane interactions and lipid reorganization. Despite extensive experimental studies, many aspects of their pore-formation mechanism remain unclear, making MD simulations a valuable tool for unravelling their structural and functional properties at the molecular level; for instance, the recognition of membrane receptors or specific lipids, the identification of the residues at the protomer interfaces that are crucial for oligomerization, and the analysis of the conformational rearrangements that are required for pore formation. In this work, I applied MD simulations, at atomistic representation, protein-protein docking and protein structure prediction methods to investigate the members of two different classes of PFTs, defined according to the 3D rearrangement of the transmembrane domain into either an α-helix or a β-sheet, after activation (one then refers to α-PFTs or β-PFTs). This classification reflects not only a structural feature but also distinct biological pathways leading to oligomerization and pore formation; in turn, the study of PFTs belonging to different classes results in the tackling of diverse open questions and challenges. 1.a γ-hemolysin The bi-component γ-hemolysin protein represents one of the most common β-PFTs expressed by the pathogenic bacterium Staphylococcus aureus. The toxin is used by the pathogen to escape the immune system of the host organism, by assembling into octameric transmembrane pores on the surface of the target immune cell and leading to its death by leakage or apoptosis. Despite the high potential risk associated with Staphylococcus aureus infections and the urgent need for new treatments, several aspects of the pore formation process from γ-hemolysin are still unclear. These include the identification of the interactions between the individual monomers that lead to the formation of a dimer on the cell membrane, which represents the unit for further oligomerization. In this work, I employed a combination of all-atoms explicit solvent MD simulations and protein-protein docking to determine the stabi- lizing contacts that guide the formation of a functional dimer. The simulations and the molecular modelling concur to reveal the importance of the flexibility of specific protein domains, in particular the N-terminus, to drive the formation of the correct dimerization interface through functional contacts between the monomers. The re- sults obtained are compared with the experimental data available in the literature. Further discussions on the difficulties underlying the simulation of conformational changes in this kind of PFTs – namely β-PFTs – and putative strategies to tackle this problem close the subchapter. 1.b Equinatoxin II Among eukaryotic PFTs, actinoporins (APs) from sea anemones have raised atten- tion in the last years, as models of protein-membrane interaction for α-PFTs and for their biomedical interest. Out of APs, one of the most characterized members is equinatoxin II (EqtII) from Actinia equina. However, although being extensively studied, many aspects of its self-assembly pathway remain elusive. Available struc- tural experimental data include the in-solution, inactive structure of EqtII, but not the membrane-inserted one nor the entire pore assembly. In particular, the architec- ture and the stoichiometry of the final pore and the role of membrane composition are the most interesting and controversial aspects I wanted to investigate. My approach aimed at directly tackling the problem of pore stoichiometry and architecture, starting with reconstructing the membrane-inserted eqtII monomer employing state-of-the-art protein structure prediction modelling and experimental knowledge of the 3D structure of other APs members. I have thus reconstructed a novel model for the membrane-inserted eqtII monomer and tested its stability in model membranes at different lipid compositions with all-atom MD simulations. The obtained MD trajectories reveal high stability of the reconstructed monomer and interesting insights which support the hypothesis that APs aggregation happens by sequential incorporation of already-inserted monomers. I then tested, through μs- long all-atom MD simulations, the hybrid octameric model for AP pore architecture, and assessed the role of lipids by monitoring both direct protein-lipid interactions and the membrane lipid phase, employing membranes at different lipid compositions. 2. Sigma-1 The sigma-1 receptor (S1R) is an endoplasmic reticulum (ER) transmembrane pro- tein involved in the regulation of calcium fluxes from the ER to mitochondria, lipid dynamics and ER stress response. Changes in S1R expression levels have been as- sociated with several pathological conditions; moreover, S1R’s ability to bind a host of structurally dissimilar pharmacologically active compounds with high affinity – e.g. benzomorphans and antipsychotics – makes the receptor a potential target in therapeutics. Recent studies revealed the presence of S1R in the SARS-Cov-2 inter- actome and identified several ligands targeting S1R with potent anti-viral activity in vitro, opening the path to new potential therapeutic applications of S1R. Crystallized as a homotrimer, S1R can exist in fact in numerous oligomerization states and experiments show that agonist-like binding promotes low molecular weigth species; experimental structural data and preliminary docking/MD data on the receptor in complex with some tested ligands, suggest that oligomerization state could be linked to a specific binding mode of agonist and antagonist-like compounds through an al- losteric mechanism. However, the exact way these ligands act on the receptor to alter its function is still unknown. In this work – a collaboration with the CNR Crystal- lography Institute – through a large set of atomistic molecular dynamics simulations at the μs scale, I investigate the membrane-bound S1R-ligands complexes starting from the available X-ray structures of the trimeric receptor in complex with classical antagonist haloperidol and classical agonist (+)pentazocine. Comparative analyses of the simulated systems provide insights into the S1R conformational changes induced by the binders, highlighting the role of α-helix 1 (transmembrane) and α-helix 4, in line with the previous experimental and in-silico results.