running and analyzing md simulation to study complex membrane systems

Simulating and Analyzing Lipid Bilayers Using Molecular Dynamics

Lipids are major constituents of membranes, which provide boundaries for cells and enable intracellular compartmentalization1,2,3. Lipids are amphiphilic, with a polar head group and two hydrophobic fatty acid tails; these self-assemble into a bilayer to minimize contact of the hydrophobic chains with water3,4. Various combinations of hydrophilic head groups and hydrophobic tails result in different classes of lipids in biological membranes, such as glycerophospholipids, sphingolipids, and sterols (Figure 1)1,5,6. Glycerophospholipids are primary building blocks of eukaryotic cell membranes composed of glycerophosphate, long-chain fatty acids, and head groups of low-molecular weight7. Lipid nomenclature is based on differences in head groups; examples include phosphatidyl-choline (PC), phosphatidyl-ethanolamine (PE), phosphatidyl-serine (PS), phosphatidyl-glycerol (PG), phosphatidyl-inositol (PI), or the unmodified phosphatidic acid (PA)5,6. As for hydrophobic tails, the length and degree of saturation vary, along with the backbone structure. The possible combinations are numerous, resulting in thousands of lipid species in mammalian cells6. Changes in membrane lipid composition lead to different mechanical and structural membrane properties that impact the activity of both integral membrane proteins and peripheral proteins2,6.
<img alt="Figure 1" class="xfigimg" src="/files/ftp_upload/65712/65712fig01.jpg" /> 
Figure 1. Representative lipid structures. Fatty acid tails are shown in blue boxes, common lipid head groups in orange, and sample backbones in purple. <a href="https://www.jove.com/files/ftp_upload/65712/65712fig01large.jpg" target="_blank">Please click here to view a larger version of this figure.</a>
Lipids are active players in cellular processes, protein activation in signaling cascades, and healthy cell homeostasis8,9. Altered lipid dynamics are the result of infection or can be markers of pathogenesis of disease10,11,12,13,14,15. As barriers for the cell, the study of membrane lipids and their role in permeation of small molecules is of relevance for drug delivery systems and membrane disruption mechanisms16,17. Chemical diversity and different ratios of lipid species across organelles, tissues, and organisms give rise to complex membrane dynamics2. It is therefore important to retain these characteristics in modelling studies of lipid bilayers, especially when the goal of a study is to examine interactions of other biomolecules with the membrane. The lipid species to consider in a model depends on the organism and cellular compartment of interest. For example, PG lipids are important for electron transfer in photosynthetic bateria18, while phosphorylated inositol lipids (PIPs) are major players in plasma membrane (PM) dynamics and signaling cascades in mammalian cells19,20. Inside the cell, the PM, endoplasmic reticulum (ER), Golgi, and mitochondrial membranes contain unique lipid abundances that influence their function. For example, the ER is the hub for lipid biogenesis and transports cholesterol out to the PM and Golgi; it contains a high lipid diversity with an abundance of PC and PE, but low sterol content, which promotes membrane fluidity21,22,23,24. In contrast, the PM incorporates hundreds and even thousands of lipid species depending on the organism25, it contains high levels of sphingolipids and cholesterol that give it a characteristic rigidity compared to other membranes in the cell24. Leaflet asymmetry should be considered for membranes like the PM, which has an outer leaflet rich in sphingomyelin, PC, and cholesterol, and an inner leaflet rich in PE, PI and PS that are important for signaling cascades24. Finally, lipid diversity also prompts the formation of micro-domains that differ in packing and internal order, known as lipid rafts24,26; these exhibit lateral asymmetry, are hypothesized to play important roles in cellular signaling26, and are hard to study due to their transient nature.
Experimental techniques such as fluoroscopy, spectroscopy, and model membrane systems like giant unilamellar vesicles (GUVs) have been used to investigate interactions of biomolecules with membranes. However, the complex and dynamic nature of the components involved is difficult to capture with experimental methods alone. For example, there are limitations on the imaging of transmembrane domains of proteins, the complexity of membranes used in such studies, and the identification of intermediate or transient states during the process of interest27,28,29. Since the advent of molecular simulation of lipid monolayers and bilayers in the 1980s29, lipid-protein systems and their interactions can now be quantified at the molecular level. Molecular dynamics (MD) simulation is a common computational technique that predicts the movement of particles based on their intermolecular forces. An additive interaction potential describes the bonded and non-bonded interactions between particles of the system30. The set of parameters used to model these interactions is termed the simulation forcefield (FF). These parameters are obtained from ab initio computations, semi-empirical, and quantum mechanical calculations, and optimized to reproduced data from X-ray and electron diffraction experiments, NMR, infrared, Raman and neutron spectroscopy, amongst other methods31.
MD simulations can be used to study systems at various levels of resolution32,33,34. Systems that aim to characterize specific biomolecular interactions, hydrogen bonds, and other high-resolution details are studied with all-atom (AA) simulations. In contrast, coarse-grained (CG) simulations lump atoms into larger functional groups to reduce computational cost and examine larger scale dynamics33. Situated in between these two are united-atom (UA) simulations, where hydrogen atoms are combined with their respective heavy atoms to accelerate the computation33,35. MD simulations are a powerful tool for exploration of the dynamics of lipid membranes and their interactions with other molecules and can serve to provide molecular level mechanisms for processes of interest at the membrane interface. Additionally, MD simulations can serve to narrow down experimental targets and predict macromolecular properties of a given system based on microscopic interactions.
In brief, given a set of initial coordinates, velocities, and a set of conditions like constant temperature and pressure, positions and velocities of each particle are calculated through numerical integration of the interaction potential and Newton&#39;s Law of motion. This is repeated iteratively, thereby generating a simulation trajectory30. These calculations are performed with an MD engine; among several open-source packages, GROMACS36 is one of the most commonly used engines and the one we describe here. It also includes tools for analysis and constructing initial coordinates of systems to be simulated37. Other MD engines include NAMD38; CHARMM39, and AMBER40, which the user can select at their own discretion based on computational performance of a given system. It is critical to visualize the trajectories during the simulation as well as for analysis and interpretation of the results. A variety of tools are available; here we discuss visual molecular dynamics (VMD) that offers a wide range of features, including three dimensional (3-D) visualization with expansive drawing and coloring methods, volumetric data visualization, building, preparing, and analyzing trajectories of MD simulation systems, and trajectory-movie making with no limits on system size, if the memory is available41,42,43.
The accuracy of predicted dynamics between system components is directly influenced by the FF chosen for the propagation of the trajectory. Empirical FF parametrization efforts are pursued by few research groups. The most established and common FF for MD include CHARMM39, AMBER40, Martini44, OPLS45, and SIRAH46. The all-atom additive CHARMM36 (C36) force field47 is widely used for AA MD of membrane systems as it accurately reproduces experimental structural data. It was originally developed by the CHARMM community, and it is compatible with multiple MD engines like GROMACS and NAMD. Despite improvements across common FFs, there is a continuous effort to improve the parameter sets to allow predictions that closely reproduce experimental observables, driven by interests in particular systems of study48,49.
A challenge when simulating lipid membranes is determining the length of simulation trajectory. This is largely dependent on the metrics to be analyzed and the process that one aims to characterize. Typically, complex lipid mixtures require longer time to reach equilibrium, as more species must have enough time to diffuse on the membrane plane and reach a stable lateral organization. A simulation is said to be at equilibrium when the property of interest has reached a plateau and fluctuates about a constant value. It is common practice to obtain at least 100-200 ns of equilibrated trajectory to carry out appropriate statistical analysis on the properties and interactions of interest. It is common to run membrane-only simulations between 200-500 ns, depending on the complexity of the lipid mixture and research question. Protein-lipid interactions typically require longer simulation times, between 500-2000 ns. Some approaches to accelerate sampling and observable dynamics with membrane systems are: (i) the highly mobile membrane mimetic (HMMM) model, which replaces end carbons of lipids in the membrane with organic solvent to accelerate sampling50; and (ii) hydrogen mass repartitioning (HMR), which combines a fraction of the masses of heavy atoms within a system with those of hydrogen atoms to allow the use of a larger simulation timestep51.
The following protocol discusses a beginner-friendly approach to build, run, and analyze realistic membrane models using AA MD. Given the nature of MD simulations, multiple trajectories must be run to account for reproducibility and proper statistical analysis of the results. It is current practice to run at minimum three replicas per system of interest. Once the lipid species have been selected for the organism and process of interest, basic steps to build, run, and analyze a simulation trajectory of a membrane-only system are outlined and summarized in Figure 2.
<img alt="Figure 2" class="xfigimg" src="/files/ftp_upload/65712/65712fig02.jpg" /> 
Figure 2. Schematic to run MD simulations. Orange boxes correspond to the three main steps described in the protocol. Underneath is the workflow of the simulation process. During system setup, the system containing the initial coordinates of a solvated membrane system is built with a system input generator like CHARMM-GUI Membrane Builder. After transferring the input files to a high-performance computing cluster, the simulation trajectory is propagated using an MD engine, such as GROMACS. Trajectory analysis can be done on the computer cluster or a local working station along with visualization. Analysis is then carried out, using either packages with built-in analysis code such as GROMACS and VMD, or by using Bash scripts or various Python libraries. <a href="https://www.jove.com/files/ftp_upload/65712/65712fig02large.jpg" target="_blank">Please click here to view a larger version of this figure.</a>

Author Spotlight: Advancing Cell Membrane Biophysics - Exploring Interactions and Challenges Through Experimental and Computational Approaches 

Realistic Membrane Modeling Using Complex Lipid Mixtures in Simulation Studies

Cell membrane biophysics is rapidly growing. Common experimental approaches include fluorescence microscopy, neutron and X-ray scattering, mass spectroscopy, and atomic force microscopy. Computational approaches include phonological models and those based on statistical thermodynamics that help us understand atomistic or molecular level details of interactions occurring at the membrane interface and inside its hydrophobic core.When using molecular dynamic simulations, challenges include sampling events of interest, setting a simulations length, ensuring results converge and reproduce physical values, and accessing computational power. These challenges are particularly true for transient events at the membrane interface, such as the interaction of peripheral proteins with it or the aggregation of lipids and proteins that require large changes in confirmation. This protocol provides a beginner friendly step-by-step guide to start running molecular dynamic simulations of complex lipid membranes.There are many software alternatives to carry these simulations, and some packages have tutorials or manuals. We hope our protocol provides a concise foundation on realistic membrane modeling and tips on the considerations that influence the quality of results and these types of modeling studies. This protocol has allowed us to capture interactions between membrane lipids and other biomolecules that were not observed using pure or binary lipid mixtures.Many interactions at the membrane surface depend on the diversity of lipids on the membrane itself. Our models demonstrate the importance of incorporating appropriate lipid species to accurately explore biomolecular function within membranes.

Watch this Scientific Journal Video about Running and Analyzing MD Simulation to Study Complex Membrane Systems at JoVE.com

Running and Analyzing MD Simulation to Study Complex Membrane Systems  

After building the system coordinates for molecular dynamics simulations, create a submission script to minimize and relax the built coordinate system. Use the commands listed until hashtag production comment in the README file provided in the output from CHARMM-GUI to the submission script. Submit the relaxation script and ensure all output files from its steps have been produced before progressing to the production run.Upon completion, verify the presence of output files with different extensions generated from GROMACS during the six-step relaxation run. To perform benchmarking, generate short trajectories one to two nanoseconds long and estimate the computational cost using varying numbers of computing nodes. Compare the performance in nanoseconds per day for different numbers of computing nodes to determine the optimal resources for the run.Choose the number of nodes that result in a performance level between 75 to 80%of the maximum. Compress raw trajectory files denoted as trr files in GROMACS by changing the file format to xtc files or skipping frames to reduce file size to facilitate efficient transfer to the local station for visualization and analysis. Visualize the full trajectory before running any analysis to identify the molecules or atoms of interest and determine the trajectory portion intended for characterization.Then determine the area per lipid time series for membrane only simulations and identify the portion of the trajectory. Analysis of membrane structure revealed a significant difference in thickness between the two models for the endoplasmic reticulum, indicating an inverse relationship between the area per lipid and membrane thickness. Deuterium order parameters of each lipid species showed that there is little to no difference between the order of lipid tails between the models except for DPPE, which shows a slight increase for the SN1 tail in the PI model.Lipid composition and membrane models modulates interactions with other molecules. For example, simulations using different ratios of PC and PS lipids showed that electrostatic interactions drive the initial binding of D112, a delocalized lipophilic cation to anionic lipids and hydrophobic interactions pull the molecule into the membrane core.

running-and-analyzing-md-simulation-to-study-complex-membrane-systems

Research

JoVE Journal

Biology

142 次观看.  State University of New York at Buffalo. 构建用于分子动力学模拟的系统坐标后，创建一个提交脚本以最小化和放宽构建的坐标系。使用列出的命令，直到 CHARMM-GUI 输出到提交脚本的 README 文件中的 Hashtag production 注释。提交放宽脚本，并确保在进入生产运行之前已生成其步骤的所有输出文件。完成后，验证在六步放宽运行期间是否存在从 GROMACS 生成的具有不同扩展名的输出文件。要执行基准测试，请生成一到...