Computational Simulation Studies Of Large-scale Conformational Changes Of Proteins

Proteins play important roles in all kinds of biological process. They have specific 3-dimensional structures which are folded from peptides with certain amino acid sequences under physiological conditions. The structures get functional meaning in evolution. Nowadays with the development of protein structure determination methods, more and more high-resolution protein structures are obtained. They have revealed the molecular details of many important biological processes and are the theoretical bases of drug discoveries. However, new problems come out with the progress of structural biology and biophysical chemistry. People find that a 3-dimensional structure (such as a crystal structure) usually represents a static state, but proteins are flexible and always undergo conformational changes in physiological processes, producing a series of transient states. In other words, a protein in solution can be viewed as a soft material oscillating near its equilibrated state, or even going far from the equilibrated state, but a crystal structure is only a single point on the potential surface. To have a deeper understanding of protein function, we should not only determine the static structures, but also try to reveal the dynamics, including the structural flexibility and the transitions between different conformational states. So it becomes very necessary to introduce time dimension to conventional structural biological studies and extend the structure-function relationship to the structure-dynamics-function relationship. The dynamics of proteins cover a very broad time scale, ranging from fast and localized atomic fluctuations to slow and global motions such as protein folding. Large-scale motions are always related to protein functions. For example, active site conformation adaptation may be consisted of loop motions and secondary structure motions, spanning a characteristic time scale from nanoseconds to microseconds (10-9-10-6 s) and a range of amplitudes from 1 to 5 A. A more complicated conformational transition involves domain motions. It takes place in a time scale from microseconds to milliseconds (10-6-10-3 s) and falls into a range of amplitudes from 5 to 10 A or so. However, different types of motions from fast fluctuations to slow global motions are interdependent and coupled to one another. This complicates the studies on protein dynamics.Computational simulation has its own advantages over experiments in protein dynamics studies. It can start with a high-revolution structure and give a "full-scale" description of the dynamics of the protein (molecular dynamics, as an example). It traces the position and velocity of every atom at every moment and interprets every state on the conformational transition pathway with high accuracy, while experiments often ignore these states as they are of high energy, ephemeral and sparsely-populated. Computational simulation also involves all the particle interactions and monitors energy variations. When experiment tells us how protein moves, computational simulation can tell us why it moves. At present, molecular dynamics are limited in time scale from picoseconds to nanoseconds, still far from some slow motions falling in the range of microseconds to milliseconds. But there are different methods and strategies to overcome this difficulty, such as replacement of all-atom force field with a coarse-grained one or introduction of external force to accelerate the conformational transition.We select three systems, all of which undergoes large-scale motions under physiological conditions, to study the structure-dynamics-function relationship. These systems have very different structures, including ABC transporters such as importer BtuCD and exporter MsbA (with four domains), PDZ12 tandem of PSD-95 (witktwo domains) and SNARE protein Ykt6 (with one domain). We use conventional molecular dynamics simulation, together with targeted molecular dynamics simulation and normal mode analysis to study the large-scale conformational changes of these systems.ABC transporters constitute one of the largest superfamily in organism and they use the energy from nucleotide binding and hydrolysis to translocate substrates across cell membranes. They undergo large-scale conformational changes, switching between inward-facing and outward-facing conformation during the translocation process. ABC transporters are made up of at least four domains, two transmembrane domains (TMDs) and two nucleotide-binding domains (NBDs). ATP binding and hydrolysis drive the opening and closing motion of nucleotide-binding sites and the motions in NBDs are proposed to be coupled to changes of the translocation pathway between TMDs. Here we focus on import system BtuCD and export system MsbA to study the translocation mechanism.Vitamin B12 importer system BtuCD has twenty transmembrane (TM) helices and TMDs are connected to NBDs by L-loops in TMDs in crystal structures. We found that in the transitions between the inward-facing and the outward-facing conformation, NBD motions are always coupled with TMD motions at the cytoplasmic side through L-loops. In the outward-facing forward to inward-facing transition, the opening of nucleotide-binding sites stretches L-loops and expands the cytoplasmic gate. This coupling mode is similar in the backward transition. We also found that the forward and backward transitions are not the same. The cytoplasmic gate moves more symmetrically in the forward process, while its motion becomes more asymmetrical in the backward process. Our work reveals a detailed coupling mode of the domains in BtuCD and indicates an asymmetrical intermediate state in the transition from the inward-facing to outward-facing conformation.MsbA is a bacterial lipid efflux and it is homology to P-glycoprotein, the multidrug resistance ABC exporter in human. Studies on the translocation mechanism of MsbA would be very instructive to cancer drug design. We again found the coupling between TMDs and NBDs by simulating the outward-facing to inward-facing transition, but the details in MsbA are totally different from BtuCD. The conformational change follows a clear spatio-temporal order. The opening of NBD dimer interface is the first event and the highly conserved X-loop transmits the changes to the cytoplasmic tetra-helix bundle. Then the breaking of tetra-helix bundle network induces large-scale rearrangement of the cytoplasmic side and TM6 helix brings these changes to the periplasmic side. Different parts of the structure, such as NBD, X-loop, tetra-helix bundle and TM6 helix closely cooperate and together buildup the signal transition pathway. The sequential transition process points out the functional importance of the highly conserved X-loop and approves the hypothesis that the large-scale conformational change of the transporter is triggered by the motion of nucleotide-binding sites.PDZ12 tandem contains two N-terminal PDZ domains of PSD-95 and the domains are closely connected by a conserved peptide linker of five amino acids. Both PDZ domains can bind specifically to a short peptide at the extreme C terminus of target proteins. This is crucial for PSD-95 to organize signal transduction complexes, cluster membrane receptors, and maintain cell polarities. In PDZ12 tandem, the two PDZ domains have limited freedom of rotation relative to each other, but this restrained interdomain orientation disappears when the tandem is in complex with its binding peptide. This means ligand binding increases the inter-PDZ mobility remarkably. We performed two independent molecular dynamics simulations of 12 ns for the ligand-free and-bound forms of PDZ12 and found their dynamic properties are remarkably different. In the peptide-bound system, the interdomain orientation is not restrained and the protein samples a much larger conformational space than the free form. This means the conformational flexibility of PDZ12 tandem increases dramatically upon peptide binding by losing the relative interdomain orientation. The case of PDZ12 represents a new mode of "induced-fit" effect. The dynamics variation upon ligand binding is attributed to the changes of interdomain mobility in addition to the local induced fit within an individual domain. By utilizing this domain cooperativity, PDZ12 gains extra compensatory conformational entropy favoring its target binding. We anticipate that this may be one of the general strategies adopted by multidomain scaffold proteins to facilitate its target recognition.SNARE Ykt6 is an essential protein involved in multiple membrane fusion reactions. It is very flexible and can change between open state and closed state. It is consisted of an N-terminal longin domain, a conserved central 60-70 amino acid "SNARE core" and a C-terminal "CCAIM" motif. The SNARE core mediates the specific targeting and fusion of different classes of transport vesicles to their distinct membrane destinations, and the longin domain is capable of regulating the activity of SNARE core. The unlipidated Ykt6 shows multiple interconverting conformational states in solution and the states can be shifted into one homogenous conformation by addition of a long acyl chain fatty acid, DPC. This homogenous conformation is attributed to a closed state similar to farnesylated Ykt6 under physiological conditions. In this structure, the SNARE core folds around the longin domain and forms a hydrophobic groove at their interface to accommodate the entire aliphatic tail of DPC. We characterized the unlipidated state, the DPC-binding state and the farnesylated state of Ykt6 by molecular dynamics simulations. The unlipidated state shows a collapsed hydrophobic groove and is wandering among different conformational states as experiments revealed. DPC interacts with Ykt6 by both its hydrophobic and hydrophilic end and localizes the protein in one of the unlipidated conformational state. Farnesylation shows a similar effect with DPC by stabilizing the closed state and also traps the protein in a local area in the conformational space. Thus, the population of conformational states is changed upon lipidation and one of the states is selected to take the dominance. We propose this process fits the "conformational selection" mechanism. Lipidation is often used to anchor proteins to membranes, but Ykt6 represents a novel case as lipidation can regulate the conformation of proteins.Our work shows the dynamics-function relationship is of great importance to proteins. ABC transporter BtuCD and MsbA closely couple the motions of their TMDs and NBDs to translocate the substrates across cell membrane. PDZ12 tandem is endowed with cooperativity between PDZ domains. The interdomain mobility gets increased to create conformational entropy upon ligand binding. SNARE Ykt6 can response to lipid binding or lipidation with its hydrophobic groove and shift the population of conformational states. Our work reveals that molecular dynamics simulation is a very powerful tool for the studies on large-scale motion of proteins and targeted molecular dynamics simulation and normal mode analysis are also very useful methods. We anticipate that computational simulation will be more powerful with the development of cpu speed and the optimization algorithms and plays a more important role in the exploration of structure-dynamics-function relationship of proteins.