Advanced MD simulations for membrane proteins: conformational changes, aggregation and lipid interactions.

Proteins are biological macromolecules that consist of long chains of small building blocks, called amino acids. These long sequences of amino acids are unique for each protein, define a specific three-dimensional structure that allows the protein to carry out a specific function in a living organism. In fact, they can catalyse metabolic reactions, respond to stimuli, provide structure and transportation routes within the cell [1]. In a cell proteins are ubiquitous. They can be soluble in water and have usually a globular shape; they can be arranged in fibers, give structural integrity to their host, and provide the infrastucture upon which small molecules are transported where needed; they can be embedded, partially or totally, in the membrane, a wall of a lipidic bilayer, of the cell and mediate the exchange of matter with the environment. In particular, membrane proteins are categorised into three groups: permanently attached to the membrane, integral membrane proteins have several structural elements that span the width of the membrane; peripheral membrane proteins are temporarily attached to the lipid bilayer by hydrophobic and electrostatic interactions, usually following a post-translational modification of a soluble protein; water-soluble proteins, like toxins, that upon aggregation, attack the membrane and cause the disrupture of the cell. In the last decades, the availability of structural information on proteins and their three-dimensional conformation enabled the rapid development of a computational tool, molecular dynamics (MD), that allows to explore biological processes and systems at a sub-nanometer scale. The idea behind MD is to integrate Newton’s equations of motion to describe the evolution of a protein within its biological environment. The refinement of the empirical potentials, called force fields, that defines the interactions of the system of interest and the increase in the computational resources of modern computers have enhanced scientists to investigate and characterise dynamics and functions of protein with high predictive power. This methodology is nowadays widely established as an in silico technique and can be considered a real computational microscope [2, 3]. Despite its successes, the complexity and the timescale involved in realisation of a biological process required the development of new techniques that accelerate the dynamics of the system under scrutiny and the sampling of conformations of the macromolecule [4]. Enhanced sampling methods are, therefore, essential for the study of conformational transitions, key events that trigger the function of a protein. In this thesis I will focus mainly on three membrane proteins I studied in my research that span different functions and interactions with the lipid bilayer. The presence of the membrane slows down the dynamics of an em- bedded protein with respect to the water-soluble counterpart [5]. In addition, it requires a specific treatment of the system and the biological conditions necessary to mimic the experiments as close as possible. Therefore, the first chapter will be devoted to introduce molecular dynamics as a computational technique to shed light on proteins dynamics and the undelying mechanisms of the functions they perform. I will discuss the algorithms that allow a predictive use of molecular dynamics in the presence of the membrane, and a better approximation of the experimental conditions in which biological data are gathered [6]. In addition, I will briefly describe the enhanced sampling methods used to investigate large conformational changes, and the analysis techniques used to extract meaningful information from the simulations. The rest of the thesis will describe the systems that I studied in my research work. In the second chapter I will digress on the prestin protein. Prestin is a motor protein and it is present in arrays in the cochlear outer cells in the mammalian hearing mechanism. Due to its coordinated contraction and elongation in response to external stimuli, this protein changes the shape of the cell allowing the transduction of the signal. This mechanism is mediated by a ligand, but there is no evidence of the transport of the ligand across the membrane. The non-mammalian ortholog of this protein is highly similar in the amino acid sequence, but it does not perform the same function. In fact it is a transporter that allows the exchange of chloride ions, and oxalate molecules, from the intracellular to extracellular environment, and viceversa. To investigate this difference, first I performed the simulation of two proteins, the expression of prestin in the rat and in the zebrafish species, in two conformations, inward open and outward open, for 700 ns each starting from homology models, due to the absence of experimental crystal structures. I assessed the relaxation of the four structures toward a stationary state, and the equilibrated systems were simulated under the action of an external electric field to mimic the cellular environment. with this second step I was able to determine the different paths of chloride ions in the two homologs in the binding to a conserved residue, S398 in rat and S401 in the zebrafish. Finally, each expression of the protein underwent biased simulations to explore possible pathways in the change from the inward to the outward conformation. The data are not definitive to draw a conclusion, although the elevator mechanism seems to favour the elevator-like transport, a mechanism proper of other proteins in the same family of the prestin. In the third chapter I will discuss the insertion of the recoverin protein, a peripheral membrane protein, in a membrane patch. Recoverin is a calcium sensor protein expressed in the vertebrate retina. The binding of two calcium ions triggers the extrusion of a myristoyl group, a post-translational modification of the N-terminus of the protein that adds a hydrophobic chain. This extrusion gives the protein an anchor to bind the lipidic bilayer, and this insertion leads to the formation of a complex with rhodopsin kinase. In collaboration with a master student, I simulated the recoverin in two conditions, both isolated and in the complex with a peptide from the rhodopsin kinase, to investigate its unbiased anchoring. We found that the insertion of the myristoyl is highly enhanced by the electrostatic interaction of the lipidic charged group and arginines of the surface of the protein. The same pattern were found in both setups, and the abovementioned interactions were no longer required to keep the protein in contact with the membran after the myristoyl penetrated the lipidic patch. In addition we analysed the communication networks of the systems and how it was affected by the presence the peptide. This could shed a light on how the recoverin-rhodopsin kinase complex assemblies itself. The last chapter will be devoted to the conformational changes of aquaporin type 4 upon aggregation. This membrane protein is a water channel, assembled in tetramers. In the human species it is present in two isoforms, M1and M23, named after the starting residue of the N-terminus. Studies shows that in the isoform M23, AQP4 aggregates and is more likely to form large orthogonal array of particles (OAPs) that are target for the antibody AQP4-IgG. This leads to an inflammatory disease, neuromyelitis optica [7]. Although the AQP4 has already been studied as a pharmaceutical target, there is no in silico study of the protein in the isonform M23. In order to mimic the OAPs, I created an assembly of four tetramers and simulated it for 800 ns. I analysed the influence of the N-terminus after the aggregation, and no evidence of a significant difference in the global behaviour of the protein were found. New insights are instead evident in the arrangement of the transmembrane segments of the protein. Further developments are being studied to have a better understanding of the aggregation mechanism.