Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2023 Jan 1.
Published in final edited form as: Methods. 2021 Mar 13;197:89–96. doi: 10.1016/j.ymeth.2021.03.005

Hybrid resolution molecular dynamics simulations of amyloid proteins interacting with membranes

Mohtadin Hashemi 1, Yuri L Lyubchenko 1,*
PMCID: PMC8435541  NIHMSID: NIHMS1685644  PMID: 33727072

Abstract

A broad range of human diseases, including Alzheimer’s and Parkinson’s diseases, arise from or have as key players intrinsically disordered proteins. The aggregation of these amyloid proteins into fibrillar aggregates are the key events of such diseases. Characterizing the conformation dynamics of the proteins involved is crucial for understanding the molecular mechanisms of aggregation, which in turn is important for drug development efforts against these diseases. Computational approaches have provided extensive detail about some steps of the aggregation process, however the biologically relevant elements responsible for the aggregation and or aggregation propagation have not been fully characterized. Here we describe a hybrid resolution molecular dynamics simulation method that can be employed to investigate the interaction of amyloid proteins with lipid membranes, shown to dramatically accelerate the aggregation propensity of amyloid proteins. The hybrid resolution method enables routine and accurate simulation of multi-protein and complex membrane systems on microsecond time scales in the presence of large and biologically relevant lipid membranes mimicking brain lipid composition. The hybrid resolution method was applied to computer modeling of the interactions of α-synuclein protein with a mixed lipid bilayer.

Keywords: molecular dynamics, amyloid oligomer, hybrid resolution, coarse-grained, lipid bilayer, neurodegeneration, protein-lipid interaction

Introduction

The rapid growth of computational power has had a large impact on a variety of scientific fields, characterized by a growing number of available methods for simulations and data analysis. Computational modelling has benefitted from this rapid growth, which has enabled accurate prediction of protein structures1 and the characterization of their dynamics2–6. In particular, molecular dynamics (MD) simulations have proven critical, especially for the study of systems non-amenable to traditional structure determination methods, such as intrinsically disordered proteins.

A broad range of human diseases, known as protein conformational or protein misfolding diseases, arises from or have as key players intrinsically disordered proteins7. The most devastating diseases include neurodegenerative disorders Alzheimer’s, Huntington’s, and Parkinson’s disease. These diseases originate from the conversion of amyloid proteins from their soluble states into aggregates, known as amyloid fibrils, that form deposits throughout the brain8–13,14. In addition, recent advances provide firm evidence that soluble protein oligomers rather than fibrils are the pathogenic species15–24. These are transient states of proteins not amenable to traditional structural methods. Characterizing the conformation dynamics of amyloid proteins is crucial for understanding the molecular mechanisms of aggregation, which in turn is important for drug development efforts against these diseases.

While computational approaches have provided extensive atomistic detail about some steps of the aggregation process, the biologically relevant elements responsible for the aggregation and or aggregation propagation have not been fully characterized. In fact, interactions of amyloids with biomolecules and biologically relevant surfaces have only recently been systematically investigated. Lipid bilayers, shown to dramatically accelerate the aggregation propensity of amyloids25–27, are one such surface; being abundant in vivo and, due to the plethora of possible lipid combinations, presenting a diverse interaction partner. However, progress has been hampered by the fact that lipid bilayers are large and complex entities, and current computational characterization is limited to how large or for how long a system can be investigated. Either simulate a small system for the timescale of the molecular event or simulate a large system for a very short time.

Here we present a general hybrid resolution method that can be employed to investigate the interaction of amyloid proteins with lipid membranes. The hybrid resolution method enables routine and accurate simulation of multi-protein and complex membrane systems on microsecond time scales. Such computational investigations enable the characterization of conformational dynamics of amyloid peptides and proteins in large systems, e.g. 10’s to 100’s of protein molecules, in presence of large and biologically relevant lipid membranes, e.g. mimicking brain lipid composition, and furthers our understanding of their interactions and effects on the aggregation process.

2. Computational method to study the dynamics of complex amyloid systems

All-atom non-biased computational characterization of amyloid-bilayer systems are, unfortunately, hampered by insufficient sampling due to the complexity and large time scales involved in the molecular processes. To overcome this and similar problems, many approaches have been developed to accelerate the sampling of the protein conformational landscape. One such approach is the use of coarse grained (CG) methodology, e.g. utilizing MARTINI force field, one of the most well characterized CG force fields for proteins and lipids28, 29. MARTINI has successfully been used to conduct extensive simulation studies of the amyloid aggregation process30, 31. In particular very large systems, e.g. the intermediate processes involved in progression from early oligomers to mature fibrils can be investigated32. However, the CG approach comes with the trade-off of accuracy for computational efficiency, e.g. information about the secondary structure transitions in proteins is lost. Thus, CG only models have limited usability in studies of folding, aggregation, and large conformational change of proteins – all important for the study of amyloid oligomers. To overcome this limitation, several groups have proposed hybrid approaches, combining CG with atomistic force fields. One such approach is the combination of PACE protein force field33–35 with MARTINI solvent and lipid force fields. Briefly, proteins are represented by a united atom model, where heavy atoms and polar hydrogens are explicitly represented while solvents and lipids are modeled using MARTINI coarse-grained approach. Interactions between the PACE and MARTINI terms are handled using a Lennard-Jones potential.

We use α-synuclein (α-syn), a 140 amino acid amyloid protein associated with Parkinson’s disease, and describe the application of the hybrid PACE-MARTINI approach to elucidate the interaction of α-syn with lipid bilayers. The described approach is broadly applicable toward the study of protein interactions in the context of both simple membranes and complex mixtures of biologically relevant lipids.

2.1. Prerequisites

The basic requirements for performing molecular dynamics simulations is a software suite, such as the popular software packages AMBER36, CHARMM37, DESMOND38, GROMACS39, LAMMPS40, or NAMD5. Here, we will use the GROMACS package to illustrate how to setup and perform protein-membrane interaction simulations. Specific versions of GROMACS are required: for system preparation version 3.X, and for running GPU accelerated simulations, version 4.X. In addition, for compatibility with PACE force field, GROMACS v 3.X needs to be modified. The necessary steps are described in the PACE force field archives (http://www.ks.uiuc.edu/~whan/PACE/PACEnew/). Here we will use, GROMACS versions 3.3.4 (GMX3) and 4.6.7 (GMX4).

Several other software packages are also necessary: Linux operating system, terminal emulator (we will use BASH, as it is available in most Linux distributions), a compiler suite (any recent GNU Compiler Collection, GCC), molecular visualization program (VMD41 or PyMOL42), Python for system setup scripts, and finally the PACE force field parameters and scripts (in GROMACS format, http://www.ks.uiuc.edu/~whan/PACE/PACEnew/).

2.2. System preparation

Before setting up the simulation systems a few factors need to be decided. To characterize the behavior of an amyloid protein we need to closely mimic the experimental and or physiological conditions; the ion type (i.e. Na+, K+, Ca2+, etc.), the ionic strength, and pH must also be considered. In many cases, physiological pH is sufficient, however some proteins are known to have pH-dependent behavior. The pH of the system can be mimicked by changing the protonation states of the protein residues43.

Lipid species is another important consideration when investigating protein-lipid interactions. Not only due to the physiological significance of specific lipid types but also due to the availability of force field parameters. The PACE-MARTINI hybrid approach currently supports 15 lipid types with phosphatidyl-choline (PC), -ethanolamine (PE), - glycerol (PG), and -serine (PS) head groups44. Lipid mixtures are also supported; however, considerations must be taken to ensure the homogeneity and phase-coexistent behavior in the lipid bilayers. For example, a complex human brain plasma membrane mimic was recently simulated using MARTINI and showed the importance of lipid composition45.

The initial system files can be prepared using two different pathways, 1) using the CHARMM-GUI46, 2) local manual setup. The CHARMM-GUI approach is attractive as it increases the accessibility of the hybrid approach and the speed of system preparation. A general tutorial on how to use CHARMM-GUI is presented at http://www.charmm-gui.org/?doc=tutorial&project=membrane&chapter=membrane_intro; below only the important steps as they pertain to the current simulation system will be discussed, readers are advised to follow the tutorial to gain familiarity with CHARMM-GUI. Our test system is a 512 lipid bilayer from 1:1 mixture of 1-palmitoyl-2-oleoyl-glycero-3-phosphocholine (POPC) and 1-palmitoyl-2-oleoyl-glycero-3phosphochoserine (POPS) and we will use the available structure for α-syn protein (PDB ID: 1XQ8).

2.2.1. Preparation of initial parameters

To prepare a protein-membrane system we need to specify the initial structure of the protein. In CHARMM-GUI, on the PACE CG Bilayer Builder page, select Protein/Membrane System and use the PDB ID: 1XQ8 for α-syn and RCSB as the download source. It is also possible to upload the initial structure of the protein in PDB format, e.g. a previously published simulated structure for the protein of interest25.

CHARMM-GUI will then ask to select the protein chain, if the protein structure file contains multiple chains, and determine the N- and C-terminus modification, that is if they should be free or neutralized using caps such as acetyl or amide. For our purposes, chain A and no terminal modifications are selected.

The protein orientation is then set, either by using the principal axis along Z or by defining a custom vector. For the current system, align the protein principal axis to Z, translate the molecule 4 nm along Z, and rotate the protein by 45° and 90° with respect to the X- and Y-axis, respectively. This orients the protein diagonally on the bilayer patch, as illustrated in Fig. 1A.

Figure 1.

Figure 1.

Open in a new tab

Illustration of the protein orientation during CHARMM-GUI preparation. A) Shows the orientation of the protein, PDB ID: 1XQ8, with respect to a putative bilayer, shown as yellow squares. B) Shows the final orientation of the protein on a 512 lipid bilayer consisting of a 1:1 mixture of POPC and POPS.

The lipid and solvent parameters are then selected. The water thickness parameter determines the layer of water that is added on both sides of the bilayer; set this to 50 Angstrom. The bilayer composition can be specified using either molar ratios and size of bilayer patch or by specifying the number of lipids. We specify 128 POPC and 128 POPS molecules for each leaflet, for a total of 512 lipids as shown in Fig. 1B. Next, NaCl at 150 mM concentration is selected as the ion species for the system.

Because GROMACS is used for this protocol, once Step 5 in CHARMM-GUI has been reached, and the step5_assembly.psf, step5_assembly.pdb, and step5_assembly.str files are available, the system preparation is complete. Completing the CHARMM-GUI process (i.e. reaching Step 6) results in generation of input files for a 6-step equilibration procedure that can be used with the modified NAMD v2.9 software package (http://www.ks.uiuc.edu/~whan/PACE/PACEnew/NAMD/paceForNAMD/NAMD_2.9_Source.tar.gz).

2.2.2. Preparation of GROMACS parameters

Once the initial protein-bilayer system has been obtained through CHARMM-GUI, that is step5_assembly.psf, step5_assembly.pdb, and step5_assembly.str files have been saved, the topology file in GROMACS format needs to be prepared. The topology file contains information about the molecule types (and their properties) and the composition of the system.

  1. Download the α-syn structure file (PDB ID: 1XQ8) from RCSB (www.rcsb.org).

  2. Load the GMX3 environment by issuing the following in a terminal,

    source /Software/gmx3/bin/GMXRC.bash

  3. Prepare coordinate and temporary PACE topology file,

    pdb2gmx - f 1XQ8.pdb - o 1XQ8.pace.pdb - p temp.top - ignh

    pdb2gmx - f 1XQ8.pdb - o 1XQ8.pace.gro - ignh
    1. pdb2gmx is run twice.
    2. pdb2gmx will ask for force field, select PACE.
    3. Including -glu, -asp, or -his in the commands will allow the user to set the protonation states of the respective residues.
    4. Select the exact same options for both runs of pdb2gmx.

  4. A special topology file is then be generated using the genPairPACExx script (xx denotes the PACE force field version) available in the PACE archive,

    /Software/PACE/genPairPACE15 #ATOM #RES 1XQ8.pace.pdb TER > 1XQ8.patch
    1. #ATOM denotes the number of atoms in the protein structure, 1166 for 1XQ8.pace.pdb.
    2. #RES denotes the number of residues in the protein, 140 for 1XQ8.pace.pdb.
    3. TER denotes the state of the protein termini: charged, 1, or capped, 0. 1 for 1xq8.pace.pdb

  5. The special topology and the temporary (from step 3) are then combined using another script found in the PACE archive, insert_param.py.

    python2.7 /Software/PACE/insert_param.py 1XQ8.patch temp.top > system.pace.top

  6. The topology file, system.pace.top, needs to be modified to point to the correct force field files for PACE and the MARTINI lipids, ions, and water. Furthermore, separating the protein topology from the system topology will greatly help organize the topology file.
    1. Separate the protein parameters by cutting everything from “[ moleculetype ]” to and including the section “; Include Position restraint file” into a new file named 1xq8.pace.itp.
    2. Include the main PACE force field: replace “ffPACE_1.5.itp” with the location of PACE force field file, e.g. “/Software/PACE/ffPACE_1.5.itp”
    3. Include MARTINI water by replacing “#include “spc.itp”” with “/Software/PACE/cgWater.itp”
    4. Include MARTINI ions: “#include “/Software/PACE/martini_v2.0_ions.itp””
    5. Include the MARTINI lipids, either for all supported lipids or only for the lipids used in the simulation: “#include “/Software/PACE/martini_v2.0_lipids.itp””
  7. The composition of the lipid bilayer is then added to the system topology.
    1. Using the step5_assembly.pdb file, re-arrange the protein and lipids into segments each containing only protein, similar lipids, water, or ions:
      head - 1169 step5_assembly.pdb > pro_only.pdb
      grep POPC step5_assembly.pdb > POPC_only.pdb
      grep POPS step5_assembly.pdb > POPS_only.pdb
      tail - n + 4 step5_assembly.pdb | grep WAT > wat_only.pdb
      grep NA step5_assembly.pdb > NA_only.pdb
      grep CL step5_assembly.pdb > CL_only.pdb
    2. Combine the segments in the desired order:
      cat pro_only.pdb POPC_only.pdb POPS_only.pdb wat_only.pdb NA_only.pdb CL_only.pdb > system_tmp.pdb
    3. Re-order the atom numbers and residue id’s for the re-arranged system:
      genconf - f system_tmp.pdb - o system_reordered.pdb
    4. Add the unit cell information, found in step5_assembly.str, to the re-ordered system file editconf - f system_reordered.pdb - o system_box.pdb - box 12.892 12.892 15.623
    5. Add the missing lipid, solvent, and ions information, found in step5_assembly.str, to the system topology file. Pay careful attention that the order in the topology file reflects the order of components in the system_box.pdb file.

  8. The topology, in GROMACS format, for the PDB file obtained from CHARMM-GUI is now complete. The final topology file with correct file pointers and composition is shown in Fig. 2.

  9. Prepare an index file that specifies which atomic indices belong to what molecular group, e.g. protein or lipid. This option can be used to give meaningful names to the molecular species in a system, which can then be used in the run parameter files to specify options, such as temperature or pressure.

    make_ndx - f system_box.pdb - o index.ndx

  10. Prepare position restraint files that specify atoms that are restrained to their initial position by a constant force, e.g. during the preparatory steps before production simulations such as energy minimization. Here the protein index group is restrained by 1,000 kJ*mol−1*nm−2:

    genpr - f system_box.pdb - n index.ndx - fc 1000 1000 1000 - o pro_1000.itp
    1. The index file is used to specify the group of atoms for which to make restraints.
    2. -fc specifies the force constant vector applied to restrain the atom.
    3. Position restraint files must be loaded in the topology file, e.g. in 1xq8.pace.itp,
      #ifdef POSRES_PRO
      #include “pro_1000.itp”
      #endif
Figure 2.

Figure 2.

Open in a new tab

Layout of the final topology file in GROMACS format for the CHARMM-GUI created protein-bilayer system.

2.3. Hybrid resolution simulations

2.3.1. Energy minimization

Once the simulation system has been solvated, neutralized, and is at the desired ionic strength it is necessary to ensure that no steric clashes or inappropriate geometry exists in the system. This is achieved by performing an energy minimization procedure that relaxes the molecules in the system.

According to the best practices for PACE-MARTINI to relax the system requires three steps: 1) restrained (protein and lipids are restrained using position restraint files) energy minimization, 2) Short restrained (again, proteins and lipids are restrained) MD, and 3) energy minimization. The GMX3 suit will be used to generate the binary run input files, which will then be run using GMX4. General run parameter files for all steps are available in the Supplementary-run_parameters.zip. The workflow is as follows:

  1. Load the GMX3 environment by issuing the following in a terminal:

    source /Software/gmx3/bin/GMXRC.bash

  2. Generate the binary run input file; all run input generation is performed using GMX3. Example run parameter file is presented in the supplementary Fig. S1.

    grompp - f emposres.mdp - p system.pace.top - n index.ndx - c system.box.pdb - o emposres.tpr

  3. Load the GMX4 environment; GMX4 will exclusively be used to run the simulation steps:

    source /Software/gmx4/bin/GMXRC.bash

  4. Run the simulation

    mdrun - v - rdd 1.9 - dds 0.9 - s emposres.tpr - deffnm emposres

  5. Next, using the minimized system, a short restrained MD is performed to further relax the solvent. Example parameter file is presented in the supplementary Fig. S2.

    grompp - f npt.mdp - p system.pace.real.top - n index.ndx - c emposres.gro - o emnpt.tpr

  6. Run the simulation:

    mdrun - v - rdd 1.9 - dds 0.9 - s emnpt.tpr - deffnm emnpt

  7. Following this short MD simulation another energy minimization step is performed. A generalpurpose run parameter file is presented in the supplementary Fig. S3.

    grompp - f em.mdp - p system.pace.top - n index.ndx - c emnpt.gro - o em.tpr

  8. Run the simulation:

    mdrun - v - rdd 1.9 - dds 0.9 - s em.tpr - deffnm em

To determine if the minimization procedure is successful, two factors can be evaluated. The first factor is the convergence of the potential energy, found in the .edr files after each of the simulation steps performed above. It is important that the potential energy is negative and that it converges to, preferably, a lower energy. Secondly, the maximum force on a given atom should be at or below the user-specified value. For example, in the above minimization steps the force was set to 1000 kJ/mol/nm. If the system does not reach a stable state during energy minimization, the next simulation steps may become unstable and crash.

To achieve better convergence the final minimization (steps 7–8) can be performed for longer duration and or in a stepwise fashion, with the restraint removed gradually in several steps.

2.3.2. Equilibration simulation

Energy minimization ensures that the starting states are optimized in terms of steric clashes. To investigate real dynamics, the solvent and ions must be equilibrated around the solute, i.e. protein and bilayer. Furthermore, the solvent and the solute need to be brought to the desired temperature. Once at the correct temperature (based on kinetic energies), the correct pressure must be achieved to reach proper density.

The first step of the equilibration is to bring the system to the desired temperature by performing an MD simulation using a canonical ensemble (NVT; constant Number of particles, Volume, and Temperature). The simulation length depends upon the system; typically, 50–100 ps should suffice. However, the simulation should reach a plateau at the desired temperature value before moving to the next phase of the equilibration. If the temperature does not stabilize, additional simulation time is necessary.

  1. Load the GMX3 environment by issuing the following in a terminal:

    source /Software/gmx3/bin/GMXRC.bash

  2. Generate the binary run input file using the NVT parameter file; example of parameter file is presented in the supplementary Fig. S4.

    grompp - f nvt.mdp - p system.pace.top - n index.ndx - c em.gro - o nvt.tpr

  3. Load the GMX4 environment:

    source /Software/gmx4/bin/GMXRC.bash

  4. Run the simulation

    mdrun - v - rdd 1.9 - dds 0.9 - s nvt.tpr - deffnm nvt

Once the temperature has stabilized at the desired value, the pressure of the system, and therefore the density, must be equilibrated. Equilibration of pressure is performed under an NPT ensemble (Number of particles, Pressure, and Temperature; also called isothermal-isobaric ensemble) and closely resembles experimental conditions. The time required to stabilize the pressure is typically tens of nanoseconds.

  1. Load the GMX3 environment by issuing the following in a terminal:

    source /Software/gmx3/bin/GMXRC.bash

  2. Generate the binary run input file using the NPT parameter file. A general NPT parameter file is presented in the supplementary Fig. S5.

    grompp - f npt.mdp - p system.pace.top - n index.ndx - c nvt.gro - o npt.tpr

  3. Load the GMX4 environment:

    source /Software/gmx4/bin/GMXRC.bash

  4. Run the simulation

    mdrun - v - rdd 1.9 - dds 0.9 - s npt.tpr - deffnm npt

The pressure equilibration step is repeated several times, with stepwise decrease of the solute restraint constant, e.g. initially at 1,000 kJ*mol−1*nm−2 then 500, 250, 125, 60, 30, 10 kJ*mol−1*nm−2, and finally with no restraints. Sharper drops of the restraints are also possible; however, jumps may introduce artifacts that can lead to instability in the simulations.

2.3.3. Production simulation

Once the system has equilibrated at the desired temperature and pressure, production MD can be performed. The parameters for the production simulation are similar to the NPT simulation without constraints, with the exception of simulation length which is several to tens of microseconds.

3. Interaction of α-synuclein monomer with lipid bilayer

Interactions of α-syn monomer with POPC:POPS lipid bilayer was characterized using the above described method; the equilibrated α-syn-bilayer system was simulated for 5 μs. Snapshot from key points of the simulation are shown in Fig. 3. This figure demonstrates that α-Syn monomer rapidly interacts with the surface through the N-terminal segment, which leads to the insertion of a significant portion of the segment into the interfacial region after approximately 160 ns, depicted on Fig. 3A. Monomer insertion into the interfacial region extends over time to include the whole N-terminal segment, Fig. 3B. Following the insertion of the N-terminal segment, interactions between the NAC region and the lipid head groups become more frequent, resulting in insertion into the interfacial region. The monomer then undergoes a conformation change from the initially stretched conformation to an S-shaped conformation, Fig. 3C, before reverting back to the stretched conformation, Fig. 3D.

Figure 3.

Figure 3.

Open in a new tab

Snapshots of key events from hybrid resolution simulation of α-syn monomer with POPC:POPS bilayer. Proteins are depicted using the cartoon representation with α-helix in purple, Nterminal, NAC, and C-terminal backbones are colored blue, green, and red respectively. Lipids are represented as spheres in blue, red, and grey depicting PS, PC, and PO4 pseudo-atoms of the MARTINI lipids respectively.

Interactions between the monomer and the lipid head groups of the bilayer were analyzed by quantifying the residue-specific interactions with phosphate groups of the lipid heads. Fig. 4A depicts a kymograph of the normalized contacts between the protein residues and the phosphate groups of the bilayer. It is evident from the kymograph that the N-terminal plays an important role in the interaction and insertion of the monomer into the interfacial region. Moreover, it is clear that the C-terminal region, at all times, remains accessible to molecules in the bulk.

Figure 4.

Figure 4.

Open in a new tab

Hybrid resolution simulation of α-syn monomer with POPC:POPS bilayer. A) Kymograph of normalized contacts between protein residues and the PO4 groups of the lipids. B) Secondary structure map of the α-syn monomer obtained using STRIDE as implemented in VMD.

Structural transitions within α-Syn monomer have been obtained due to the hybrid resolution method. α-Syn monomer secondary structure as a function of time is presented in Fig. 4B. The monomer initially has high percentage of α-helical secondary structure in the segments identified by NMR47. As the monomer interacts with the lipid bilayer, is inserted into the interfacial region, and undergoes conformational change, the secondary structure of the protein changes. The highest degree of change in the secondary structure occurs in the NAC and C-terminal regions of the protein. The NAC region initially has two large helices, residue 61–77 and 81–93. The former helix is disrupted upon the monomer interacting with and inserting into the interfacial region. The latter helix is initially disrupted, but upon insertion of the NAC segment into the interfacial region the helix reforms and remains for the duration of the simulation. The large N-terminal helices remain stable throughout the simulation; however, they do experience small changes when the monomer conformation re-arranges. The C-terminal helix also is initially disrupted by interactions with the membrane, however following the insertion of the NAC region into the interfacial region it is reformed. Moreover, short β-strands appear in the C-terminal segment after approximately 2.5 μs.

Overall, the data show that interaction of α-syn with the lipid bilayer leads to dramatic changes in the conformation of the protein and to a smaller degree its structure. Structural change in α-syn when interacting with membranes has been shown experimentally48–52, and may be a part of its normal function53, 54. Additionally, structural change depending on the lipid composition may be important for the disease progression, as past reports have shown lipid-dependent aggregation propensity25 and CG simulations have suggested that the conformation of α-syn, when interacting with the membrane, is important for the aggregation and propagation of aggregates.

Conclusions and perspectives

The computer modeling of interactions between amyloid proteins and cell membranes is of great importance, as membranes play a critical role in oligomer assembly at physiologically low concentrations. A theoretical model, supported by experimental studies, provided an explanation for the catalytic property of membranes in oligomers self-assembly55. Computational modeling plays a critical role in elucidating the surface mediated conformational transition within monomers facilitating the aggregation process.

Many questions about the mechanism of aggregation and toxicity of oligomers on membrane surfaces remain to be addressed: Does formation of different size oligomers depend on membrane type or property? What property determines if an oligomer is able to disrupt a membrane and does it vary depending on amyloid species? These and more are yet to be investigated, partly due to complexity and partly due to the enormous computational resources required. Answering these questions are important for the development of preventive measures against the disease-prone aggregation process.

Additional challenge for computer simulations is the replication of the lipid composition of biological membranes. This characteristic of membranes varies depending on cell and organelle, and evidence suggest that the membrane composition may be an important factor triggering the aggregation process56–59. Importantly, the cell membrane composition changes with age, leading to the hypothesis regarding the role of membrane composition in triggering the disease. Recent coarse-grained efforts45 bring a certain level of optimism, but still miss the contribution of the cytoplasmic environment60. Which is another factor that needs to be incorporated in order to get detailed understanding of the aggregation process under physiologically important environmental conditions.

Supplementary Material

1

Figure S1. Preview of position restrained energy minimization run parameter file, emposres.mdp.

Figure S2. Preview of position restrained NPT simulation run parameter file, emnpt.mdp.

Figure S3. Preview of energy minimization parameter file, em.mdp

Figure S4. Preview of NVT ensemble run parameter file, nvt.mdp.

Figure S5. Preview of NPT ensemble run parameter file, promd.mdp, which can be used for pressure equilibration and production MD.

2
3
4
5
6

Acknowledgements

This research was funded by National Institutes of Health, grants GM096039 and GM118006 to Y.L.L. This work was completed utilizing the Holland Computing Center of the University of Nebraska, which receives support from the Nebraska Research Initiative.

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Supplementary-run_parameters.zip – compressed archive containing run parameter files described in the text and previewed in Figures S1–5.

References

  • 1.Kryshtafovych A, et al. , Critical assessment of methods of protein structure prediction (CASP)— Round XIII. Proteins: Structure, Function, and Bioinformatics, 2019. 87(12): p. 1011–1020. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Hollingsworth SA and Dror RO, Molecular Dynamics Simulation for All. Neuron, 2018. 99(6): p. 1129–1143. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Kutzner C, et al. , More bang for your buck: Improved use of GPU nodes for GROMACS 2018. Journal of Computational Chemistry, 2019. 40(27): p. 2418–2431. [DOI] [PubMed] [Google Scholar]
  • 4.Le Grand S, Götz AW, and Walker RC, SPFP: Speed without compromise—A mixed precision model for GPU accelerated molecular dynamics simulations. Computer Physics Communications, 2013. 184(2): p. 374–380. [Google Scholar]
  • 5.Phillips JC, et al. , Scalable molecular dynamics on CPU and GPU architectures with NAMD. The Journal of Chemical Physics, 2020. 153(4). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Sedova A, et al. , High-Performance Molecular Dynamics Simulation for Biological and Materials Sciences: Challenges of Performance Portability, in 2018 IEEE/ACM International Workshop on Performance, Portability and Productivity in HPC (P3HPC). 2018. p. 1–13. [Google Scholar]
  • 7.Selkoe DJ, Cell biology of protein misfolding: the examples of Alzheimer’s and Parkinson’s diseases. Nat Cell Biol, 2004. 6(11): p. 1054–61. [DOI] [PubMed] [Google Scholar]
  • 8.Chiti F and Dobson CM, Protein Misfolding, Functional Amyloid, and Human Disease. Annu. Rev. Biochem, 2006. 75(1): p. 333–366. [DOI] [PubMed] [Google Scholar]
  • 9.Petkova AT, et al. , Self-Propagating, Molecular-Level Polymorphism in Alzheimer’s ß-Amyloid Fibrils. Science, 2005. 307(5707): p. 262–265. [DOI] [PubMed] [Google Scholar]
  • 10.Dobson CM, Principles of protein folding, misfolding and aggregation. Semin Cell Dev Biol, 2004. 15(1): p. 3–16. [DOI] [PubMed] [Google Scholar]
  • 11.Ross CA and Poirier MA, Protein aggregation and neurodegenerative disease. Nat Med, 2004. 10 Suppl: p. S10–7. [DOI] [PubMed] [Google Scholar]
  • 12.Uversky VN and Fink AL, Conformational constraints for amyloid fibrillation: the importance of being unfolded. Biochim. Biophys. Acta, 2004. 1698(2): p. 131–153. [DOI] [PubMed] [Google Scholar]
  • 13.Riek R and Eisenberg DS, The activities of amyloids from a structural perspective. Nature, 2016. 539(7628): p. 227–235. [DOI] [PubMed] [Google Scholar]
  • 14.Sunde M, et al. , Common core structure of amyloid fibrils by synchrotron X-ray diffraction 1 1Edited by F. E. Cohen. Journal of Molecular Biology, 1997. 273(3): p. 729–739. [DOI] [PubMed] [Google Scholar]
  • 15.Bieschke J, et al. , Small-molecule conversion of toxic oligomers to nontoxic β-sheet–rich amyloid fibrils. Nat. Chem. Biol, 2011. 8(1): p. 93–101. [DOI] [PubMed] [Google Scholar]
  • 16.Glabe CG, Structural classification of toxic amyloid oligomers. J Biol Chem, 2008. 283(44): p. 29639–43. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Ahmed M, et al. , Structural conversion of neurotoxic amyloid-beta(1–42) oligomers to fibrils. Nat. Struct. Mol. Biol, 2010. 17(5): p. 561–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Benilova I, Karran E, and De Strooper B, The toxic Abeta oligomer and Alzheimer’s disease: an emperor in need of clothes. Nat Neurosci, 2012. 15(3): p. 349–57. [DOI] [PubMed] [Google Scholar]
  • 19.Zhao LN, et al. , The toxicity of amyloid beta oligomers. Int. J. Mol. Sci, 2012. 13(6): p. 7303–27. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Glabe CG, Common mechanisms of amyloid oligomer pathogenesis in degenerative disease. Neurobiol. Aging, 2006. 27(4): p. 570–575. [DOI] [PubMed] [Google Scholar]
  • 21.Yu L, et al. , Structural Characterization of a Soluble Amyloid β-Peptide Oligomer. Biochemistry, 2009. 48(9): p. 1870–1877. [DOI] [PubMed] [Google Scholar]
  • 22.Laganowsky A, et al. , Atomic view of a toxic amyloid small oligomer. Science, 2012. 335(6073): p. 1228–31. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Liu P, et al. , Quaternary Structure Defines a Large Class of Amyloid-beta Oligomers Neutralized by Sequestration. Cell Rep, 2015. 11(11): p. 1760–71. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Zott B, et al. , A vicious cycle of β amyloid–dependent neuronal hyperactivation. Science, 2019. 365(6453): p. 559–565. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Lv Z, et al. , Assembly of α-synuclein aggregates on phospholipid bilayers. Biochim. Biophys. Acta, 2019. 1867(9): p. 802–812. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Banerjee S, et al. , Interaction of Aβ42 with Membranes Triggers the Self-Assembly into Oligomers. Int. J. Mol. Sci, 2020. 21(3): p. 1129. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Banerjee S, et al. , Cholesterol in membranes facilitates aggregation of amyloid β protein at physiologically low concentrations. 2020. [DOI] [PubMed]
  • 28.Monticelli L, et al. , The MARTINI Coarse-Grained Force Field: Extension to Proteins. J. Chem. Theory Comput, 2008. 4(5): p. 819–34. [DOI] [PubMed] [Google Scholar]
  • 29.Marrink SJ, et al. , The MARTINI force field: coarse grained model for biomolecular simulations. J Phys Chem B, 2007. 111(27): p. 7812–24. [DOI] [PubMed] [Google Scholar]
  • 30.Seo M, et al. , Improving Internal Peptide Dynamics in the Coarse-Grained MARTINI Model: Toward Large-Scale Simulations of Amyloid- and Elastin-like Peptides. J. Chem. Theory Comput, 2012. 8(5): p. 1774–1785. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Yu H, et al. , Transient beta-hairpin formation in alpha-synuclein monomer revealed by coarse-grained molecular dynamics simulation. J. Chem. Phys, 2015. 143(24): p. 243142. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Sørensen J, et al. , Protofibrillar Assembly Toward the Formation of Amyloid Fibrils. The Journal of Physical Chemistry Letters, 2011. 2(19): p. 2385–2390. [Google Scholar]
  • 33.Han W and Schulten K, Further optimization of a hybrid united-atom and coarse-grained force field for folding simulations: Improved backbone hydration and interactions between charged side chains. J. Chem. Theory Comput, 2012. 8(11): p. 4413–4424. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Han W, et al. , PACE Force Field for Protein Simulations. 1. Full Parameterization of Version 1 and Verification. J. Chem. Theory Comput, 2010. 6(11): p. 3373–3389. [DOI] [PubMed] [Google Scholar]
  • 35.Wan C-K, Han W, and Wu Y-D, Parameterization of PACE Force Field for Membrane Environment and Simulation of Helical Peptides and Helix–Helix Association. J. Chem. Theory Comput, 2011. 8(1): p. 300–313. [DOI] [PubMed] [Google Scholar]
  • 36.Case DA, B. K, Ben-Shalom IY, Brozell SR, Cerutti DS, Cheatham TE III, Cruzeiro VWD, Darden TA, Duke RE, Giambasu G, Gilson MK, Gohlke H, Goetz AW,Harris R, Izadi S, Izmailov SA, Kasavajhala K, Kovalenko A, Krasny R, Kurtzman T, Lee TS, LeGrand S, Li P, Lin C, Liu J, Luchko T, Luo R, Man V, Merz KM, Miao Y, Mikhailovskii O, Monard G, Nguyen H, Onufriev A, Pan F, Pantano S, Qi R, Roe DR, Roitberg A, Sagui C, Schott-Verdugo S, Shen J, Simmerling CL, Skrynnikov NR, Smith J, Swails J, Walker RC, Wang J, Wilson L, Wolf RM, Wu X, Xiong Y, Xue Y, York DM and Kollman PA, AMBER 20. 2020: University of California, San Francisco. [Google Scholar]
  • 37.Brooks BR, et al. , CHARMM: The biomolecular simulation program. Journal of Computational Chemistry, 2009. 30(10): p. 1545–1614. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Bowers KJ, et al. , Scalable Algorithms for Molecular Dynamics Simulations on Commodity Clusters, in ACM/IEEE SC 2006 Conference (SC’06). 2006. p. 43–43. [Google Scholar]
  • 39.Abraham MJ, et al. , GROMACS: High performance molecular simulations through multi-level parallelism from laptops to supercomputers. SoftwareX, 2015. 1–2: p. 19–25. [Google Scholar]
  • 40.Plimpton S, Fast Parallel Algorithms for Short-Range Molecular Dynamics. Journal of Computational Physics, 1995. 117(1): p. 1–19. [Google Scholar]
  • 41.Humphrey W, Dalke A, and Schulten K, VMD: Visual molecular dynamics. Journal of Molecular Graphics, 1996. 14(1): p. 33–38. [DOI] [PubMed] [Google Scholar]
  • 42.Schrodinger LLC, The PyMOL Molecular Graphics System, Version 1.8 2015.
  • 43.Olsson MHM, et al. , PROPKA3: Consistent Treatment of Internal and Surface Residues in Empirical pKa Predictions. J. Chem. Theory Comput, 2011. 7(2): p. 525–537. [DOI] [PubMed] [Google Scholar]
  • 44.Qi Y, et al. , CHARMM-GUI PACE CG Builder for Solution, Micelle, and Bilayer Coarse-Grained Simulations. J. Chem. Inf. Model, 2014. 54(3): p. 1003–1009. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 45.Ingólfsson HI, et al. , Computational Lipidomics of the Neuronal Plasma Membrane. Biophysical Journal, 2017. 113(10): p. 2271–2280. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 46.Jo S, et al. , CHARMM-GUI: a web-based graphical user interface for CHARMM. J. Comput. Chem, 2008. 29(11): p. 1859–65. [DOI] [PubMed] [Google Scholar]
  • 47.Ulmer TS, et al. , Structure and Dynamics of Micelle-bound Human α-Synuclein. Journal of Biological Chemistry, 2005. 280(10): p. 9595–9603. [DOI] [PubMed] [Google Scholar]
  • 48.Davidson WS, et al. , Stabilization of alpha-synuclein secondary structure upon binding to synthetic membranes. J Biol Chem, 1998. 273(16): p. 9443–9. [DOI] [PubMed] [Google Scholar]
  • 49.Jo E, et al. , alpha-Synuclein membrane interactions and lipid specificity. J Biol Chem, 2000. 275(44): p. 34328–34. [DOI] [PubMed] [Google Scholar]
  • 50.Eliezer D, et al. , Conformational properties of alpha-synuclein in its free and lipid-associated states. J Mol Biol, 2001. 307(4): p. 1061–73. [DOI] [PubMed] [Google Scholar]
  • 51.Georgieva ER, et al. , Membrane-Bound α-Synuclein Forms an Extended Helix: Long-Distance Pulsed ESR Measurements Using Vesicles, Bicelles, and Rodlike Micelles. Journal of the American Chemical Society, 2008. 130(39): p. 12856–12857. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 52.Bodner CR, Dobson CM, and Bax A, Multiple tight phospholipid-binding modes of alpha-synuclein revealed by solution NMR spectroscopy. J Mol Biol, 2009. 390(4): p. 775–90. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 53.Diao J, et al. , Native alpha-synuclein induces clustering of synaptic-vesicle mimics via binding to phospholipids and synaptobrevin-2/VAMP2. Elife, 2013. 2: p. e00592. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 54.Venda LL, et al. , alpha-Synuclein and dopamine at the crossroads of Parkinson’s disease. Trends Neurosci, 2010. 33(12): p. 559–68. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 55.Pan Y, et al. , Molecular Model for the Surface-Catalyzed Protein Self-Assembly. The Journal of Physical Chemistry B, 2019. 124(2): p. 366–372. [DOI] [PubMed] [Google Scholar]
  • 56.Gellermann GP, et al. , Raft lipids as common components of human extracellular amyloid fibrils. Proceedings of the National Academy of Sciences of the United States of America, 2005. 102(18): p. 6297–6302. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 57.Wingo TS, et al. , Association of Early-Onset Alzheimer Disease With Elevated Low-Density Lipoprotein Cholesterol Levels and Rare Genetic Coding Variants of APOB. JAMA Neurol, 2019. 76(7): p. 809–817. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 58.Bucciantini M, Rigacci S, and Stefani M, Amyloid Aggregation: Role of Biological Membranes and the Aggregate-Membrane System. J Phys Chem Lett, 2014. 5(3): p. 517–27. [DOI] [PubMed] [Google Scholar]
  • 59.Roher AE, et al. , APP/Aβ structural diversity and Alzheimer’s disease pathogenesis. Neurochemistry International, 2017. 110(Supplement C): p. 1–13. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 60.Rickard MM, et al. , Crowding, Sticking, and Partial Folding of GTT WW Domain in a Small Cytoplasm Model. The Journal of Physical Chemistry B, 2020. 124(23): p. 4732–4740. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

1

Figure S1. Preview of position restrained energy minimization run parameter file, emposres.mdp.

Figure S2. Preview of position restrained NPT simulation run parameter file, emnpt.mdp.

Figure S3. Preview of energy minimization parameter file, em.mdp

Figure S4. Preview of NVT ensemble run parameter file, nvt.mdp.

Figure S5. Preview of NPT ensemble run parameter file, promd.mdp, which can be used for pressure equilibration and production MD.

NIHMS1685644-supplement-1.docx (339.5KB, docx)
2
NIHMS1685644-supplement-2.mdp (1.2KB, mdp)
3
NIHMS1685644-supplement-3.mdp (2.2KB, mdp)
4
NIHMS1685644-supplement-4.mdp (1.2KB, mdp)
5
NIHMS1685644-supplement-5.mdp (1.9KB, mdp)
6
NIHMS1685644-supplement-6.mdp (2KB, mdp)

RESOURCES