Machine learning-driven multiscale modeling reveals lipid-dependent dynamics of RAS signaling proteins.

RAS is a signaling protein associated with the cell membrane that is mutated in up to 30% of human cancers. RAS signaling has been proposed to be regulated by dynamic heterogeneity of the cell membrane. Investigating such a mechanism requires near-atomistic detail at macroscopic temporal and spatial scales, which is not possible with conventional computational or experimental techniques. We demonstrate here a multiscale simulation infrastructure that uses machine learning to create a scale-bridging ensemble of over 100,000 simulations of active wild-type KRAS on a complex, asymmetric membrane. Initialized and validated with experimental data (including a new structure of active wild-type KRAS), these simulations represent a substantial advance in the ability to characterize RAS-membrane biology. We report distinctive patterns of local lipid composition that correlate with interfacially promiscuous RAS multimerization. These lipid fingerprints are coupled to RAS dynamics, predicted to influence effector binding, and therefore may be a mechanism for regulating cell signaling cascades.

Multiscale simulations reveal TDP-43 molecular-level interactions driving condensation.

The RNA-binding protein TDP-43 is associated with mRNA processing and transport from the nucleus to the cytoplasm. TDP-43 localizes in the nucleus as well as accumulating in cytoplasmic condensates such as stress granules. Aggregation and formation of amyloid-like fibrils of cytoplasmic TDP-43 are hallmarks of numerous neurodegenerative diseases, most strikingly present in >90% of amyotrophic lateral sclerosis (ALS) patients. If excessive accumulation of cytoplasmic TDP-43 causes, or is caused by, neurodegeneration is presently not known. In this work, we use molecular dynamics simulations at multiple resolutions to explore TDP-43 self- and cross-interaction dynamics. A full-length molecular model of TDP-43, all 414 amino acids, was constructed from select structures of the protein functional domains (N-terminal domain, and two RNA recognition motifs, RRM1 and RRM2) and modeling of disordered connecting loops and the low complexity glycine-rich C-terminus domain. All-atom CHARMM36m simulations of single TDP-43 proteins served as guides to construct a coarse-grained Martini 3 model of TDP-43. The Martini model and a coarser implicit solvent C⍺ model, optimized for disordered proteins, were subsequently used to probe TDP-43 interactions; self-interactions from single-chain full-length TDP-43 simulations, cross-interactions from simulations with two proteins and simulations with assemblies of dozens to hundreds of proteins. Our findings illustrate the utility of different modeling scales for accessing TDP-43 molecular-level interactions and suggest that TDP-43 has numerous interaction preferences or patterns, exhibiting an overall strong, but dynamic, association and driving the formation of biomolecular condensates.

An anchor-tether 'hindered' HCN1 inhibitor is antihyperalgesic in a rat spared nerve injury neuropathic pain model.

BACKGROUND: Neuropathic pain impairs quality of life, is widely prevalent, and incurs significant costs. Current pharmacological therapies have poor/no efficacy and significant adverse effects; safe and effective alternatives are needed. Hyperpolarisation-activated cyclic nucleotide-regulated (HCN) channels are causally implicated in some forms of peripherally mediated neuropathic pain. Whilst 2,6-substituted phenols, such as 2,6-di-tert-butylphenol (26DTB-P), selectively inhibit HCN1 gating and are antihyperalgesic, the development of therapeutically tolerable, HCN-selective antihyperalgesics based on their inverse agonist activity requires that such drugs spare the cardiac isoforms and do not cross the blood-brain barrier. METHODS: In silico molecular dynamics simulation, in vitro electrophysiology, and in vivo rat spared nerve injury methods were used to test whether 'hindered' variants of 26DTB-P (wherein a hydrophilic 'anchor' is attached in the para-position of 26DTB-P via an acyl chain 'tether') had the desired properties. RESULTS: Molecular dynamics simulation showed that membrane penetration of hindered 26DTB-Ps is controlled by a tethered diol anchor without elimination of head group rotational freedom. In vitro and in vivo analysis showed that BP4L-18:1:1, a variant wherein a diol anchor is attached to 26DTB-P via an 18-carbon tether, is an HCN1 inverse agonist and an orally available antihyperalgesic. With a CNS multiparameter optimisation score of 2.25, a >100-fold lower drug load in the brain vs blood, and an absence of adverse cardiovascular or CNS effects, BP4L-18:1:1 was shown to be poorly CNS penetrant and cardiac sparing. CONCLUSIONS: These findings provide a proof-of-concept demonstration that anchor-tethered drugs are a new chemotype for treatment of disorders involving membrane targets.

Exploring CRD mobility during RAS/RAF engagement at the membrane.

During the activation of mitogen-activated protein kinase (MAPK) signaling, the RAS-binding domain (RBD) and cysteine-rich domain (CRD) of RAF bind to active RAS at the plasma membrane. The orientation of RAS at the membrane may be critical for formation of the RAS-RBDCRD complex and subsequent signaling. To explore how RAS membrane orientation relates to the protein dynamics within the RAS-RBDCRD complex, we perform multiscale coarse-grained and all-atom molecular dynamics (MD) simulations of KRAS4b bound to the RBD and CRD domains of RAF-1, both in solution and anchored to a model plasma membrane. Solution MD simulations describe dynamic KRAS4b-CRD conformations, suggesting that the CRD has sufficient flexibility in this environment to substantially change its binding interface with KRAS4b. In contrast, when the ternary complex is anchored to the membrane, the mobility of the CRD relative to KRAS4b is restricted, resulting in fewer distinct KRAS4b-CRD conformations. These simulations implicate membrane orientations of the ternary complex that are consistent with NMR measurements. While a crystal structure-like conformation is observed in both solution and membrane simulations, a particular intermolecular rearrangement of the ternary complex is observed only when it is anchored to the membrane. This configuration emerges when the CRD hydrophobic loops are inserted into the membrane and helices α3-5 of KRAS4b are solvent exposed. This membrane-specific configuration is stabilized by KRAS4b-CRD contacts that are not observed in the crystal structure. These results suggest modulatory interplay between the CRD and plasma membrane that correlate with RAS/RAF complex structure and dynamics, and potentially influence subsequent steps in the activation of MAPK signaling.

Computational Modeling of Realistic Cell Membranes.

Cell membranes contain a large variety of lipid types and are crowded with proteins, endowing them with the plasticity needed to fulfill their key roles in cell functioning. The compositional complexity of cellular membranes gives rise to a heterogeneous lateral organization, which is still poorly understood. Computational models, in particular molecular dynamics simulations and related techniques, have provided important insight into the organizational principles of cell membranes over the past decades. Now, we are witnessing a transition from simulations of simpler membrane models to multicomponent systems, culminating in realistic models of an increasing variety of cell types and organelles. Here, we review the state of the art in the field of realistic membrane simulations and discuss the current limitations and challenges ahead.

Predicting Small Molecule Transfer Free Energies by Combining Molecular Dynamics Simulations and Deep Learning.

Accurately predicting small molecule partitioning and hydrophobicity is critical in the drug discovery process. There are many heterogeneous chemical environments within a cell and entire human body. For example, drugs must be able to cross the hydrophobic cellular membrane to reach their intracellular targets, and hydrophobicity is an important driving force for drug-protein binding. Atomistic molecular dynamics (MD) simulations are routinely used to calculate free energies of small molecules binding to proteins, crossing lipid membranes, and solvation but are computationally expensive. Machine learning (ML) and empirical methods are also used throughout drug discovery but rely on experimental data, limiting the domain of applicability. We present atomistic MD simulations calculating 15,000 small molecule free energies of transfer from water to cyclohexane. This large data set is used to train ML models that predict the free energies of transfer. We show that a spatial graph neural network model achieves the highest accuracy, followed closely by a 3D-convolutional neural network, and shallow learning based on the chemical fingerprint is significantly less accurate. A mean absolute error of ∼4 kJ/mol compared to the MD calculations was achieved for our best ML model. We also show that including data from the MD simulation improves the predictions, tests the transferability of each model to a diverse set of molecules, and show multitask learning improves the predictions. This work provides insight into the hydrophobicity of small molecules and ML cheminformatics modeling, and our data set will be useful for designing and testing future ML cheminformatics methods.

ddcMD: A fully GPU-accelerated molecular dynamics program for the Martini force field.

We have implemented the Martini force field within Lawrence Livermore National Laboratory's molecular dynamics program, ddcMD. The program is extended to a heterogeneous programming model so that it can exploit graphics processing unit (GPU) accelerators. In addition to the Martini force field being ported to the GPU, the entire integration step, including thermostat, barostat, and constraint solver, is ported as well, which speeds up the simulations to 278-fold using one GPU vs one central processing unit (CPU) core. A benchmark study is performed with several test cases, comparing ddcMD and GROMACS Martini simulations. The average performance of ddcMD for a protein-lipid simulation system of 136k particles achieves 1.04 µs/day on one NVIDIA V100 GPU and aggregates 6.19 µs/day on one Summit node with six GPUs. The GPU implementation in ddcMD offloads all computations to the GPU and only requires one CPU core per simulation to manage the inputs and outputs, freeing up remaining CPU resources on the compute node for alternative tasks often required in complex simulation campaigns. The ddcMD code has been made open source and is available on GitHub at https://github.com/LLNL/ddcMD.

Molecular Mechanism for Gramicidin Dimerization and Dissociation in Bilayers of Different Thickness.

Membrane protein functions can be altered by subtle changes in the host lipid bilayer physical properties. Gramicidin channels have emerged as a powerful system for elucidating the underlying mechanisms of membrane protein function regulation through changes in bilayer properties, which are reflected in the thermodynamic equilibrium distribution between nonconducting gramicidin monomers and conducting bilayer-spanning dimers. To improve our understanding of how subtle changes in bilayer thickness alter the gramicidin monomer and dimer distributions, we performed extensive atomistic molecular dynamics simulations and fluorescence-quenching experiments on gramicidin A (gA). The free-energy calculations predicted a nonlinear coupling between the bilayer thickness and channel formation. The energetic barrier inhibiting gA channel formation was sharply increased in the thickest bilayer (1,2-dierucoyl-sn-glycero-3-phosphocholine). This prediction was corroborated by experimental results on gramicidin channel activity in bilayers of different thickness. To further explore the mechanism of channel formation, we performed extensive unbiased molecular dynamics simulations, which allowed us to observe spontaneous gA dimer formation in lipid bilayers. The simulations revealed structural rearrangements in the gA subunits and changes in lipid packing, as well as water reorganization, that occur during the dimerization process. Together, the simulations and experiments provide new, to our knowledge, insights into the process and mechanism of gramicidin channel formation, as a prototypical example of the bilayer regulation of membrane protein function.

Folding and Lipid Composition Determine Membrane Interaction of the Disordered Protein COR15A.

Plants from temperate climates, such as the model plant Arabidopsis thaliana, are challenged with seasonal low temperatures that lead to increased freezing tolerance in fall in a process termed cold acclimation. Among other adaptations, this involves the accumulation of cold-regulated (COR) proteins, such as the intrinsically disordered chloroplast-localized protein COR15A. Together with its close homolog COR15B, it stabilizes chloroplast membranes during freezing. COR15A folds into amphipathic α-helices in the presence of high concentrations of low-molecular-mass crowders or upon dehydration. Under these conditions, the (partially) folded protein binds peripherally to membranes. In our study, we have used coarse-grained molecular dynamics simulations to elucidate the details of COR15A-membrane binding and its effects on membrane structure and dynamics. Simulation results indicate that at least partial folding of COR15A and the presence of highly unsaturated galactolipids in the membranes are necessary for efficient membrane binding. The bound protein is stabilized on the membrane by interactions of charged and polar amino acids with galactolipid headgroups and by interactions of hydrophobic amino acids with the upper part of the fatty acyl chains. Experimentally, the presence of liposomes made from a mixture of lipids mimicking chloroplast membranes induces additional folding in COR15A under conditions of partial dehydration, in agreement with the simulation results.

Fluorinated Alcohols' Effects on Lipid Bilayer Properties.

Fluorinated alcohols (fluoroalcohols) have physicochemical properties that make them excellent solvents of peptides, proteins, and other compounds. Like other alcohols, fluoroalcohols also alter membrane protein function and lipid bilayer properties and stability. Thus, the questions arise: how potent are fluoroalcohols as lipid-bilayer-perturbing compounds, could small residual amounts that remain after adding compounds dissolved in fluoroalcohols alter lipid bilayer properties sufficiently to affect membranes and membrane protein function, and do they behave like other alcohols? To address these questions, we used a gramicidin-based fluorescence assay to determine the bilayer-modifying potency of selected fluoroalcohols: trifluoroethanol (TFE), HFIP, and perfluoro-tert-butanol (PFTB). These fluoroalcohols alter bilayer properties in the low (PFTB) to high (TFE) mM range. Using the same assay, we determined the bilayer partitioning of the alcohols. When referenced to the aqueous concentrations, the fluoroalcohols are more bilayer perturbing than their nonfluorinated counterparts, with the largest fluoroalcohol, PFTB, being the most potent and the smallest, TFE, the least. When referenced to the mole fractions in the membrane, however, the fluoroalcohols have equal or lesser bilayer-perturbing potency than their nonfluorinated counterparts, with TFE being more bilayer perturbing than PFTB. We compared the fluoroalcohols' molecular level bilayer interactions using atomistic molecular dynamics simulations and showed how, at higher concentrations, they can cause bilayer breakdown using absorbance measurements and 31P nuclear magnetic resonance.

Synthetic Analogues of the Snail Toxin 6-Bromo-2-mercaptotryptamine Dimer (BrMT) Reveal That Lipid Bilayer Perturbation Does Not Underlie Its Modulation of Voltage-Gated Potassium Channels.

Drugs do not act solely by canonical ligand-receptor binding interactions. Amphiphilic drugs partition into membranes, thereby perturbing bulk lipid bilayer properties and possibly altering the function of membrane proteins. Distinguishing membrane perturbation from more direct protein-ligand interactions is an ongoing challenge in chemical biology. Herein, we present one strategy for doing so, using dimeric 6-bromo-2-mercaptotryptamine (BrMT) and synthetic analogues. BrMT is a chemically unstable marine snail toxin that has unique effects on voltage-gated K+ channel proteins, making it an attractive medicinal chemistry lead. BrMT is amphiphilic and perturbs lipid bilayers, raising the question of whether its action against K+ channels is merely a manifestation of membrane perturbation. To determine whether medicinal chemistry approaches to improve BrMT might be viable, we synthesized BrMT and 11 analogues and determined their activities in parallel assays measuring K+ channel activity and lipid bilayer properties. Structure-activity relationships were determined for modulation of the Kv1.4 channel, bilayer partitioning, and bilayer perturbation. Neither membrane partitioning nor bilayer perturbation correlates with K+ channel modulation. We conclude that BrMT's membrane interactions are not critical for its inhibition of Kv1.4 activation. Further, we found that alkyl or ether linkages can replace the chemically labile disulfide bond in the BrMT pharmacophore, and we identified additional regions of the scaffold that are amenable to chemical modification. Our work demonstrates a strategy for determining if drugs act by specific interactions or bilayer-dependent mechanisms, and chemically stable modulators of Kv1 channels are reported.

Computational 'microscopy' of cellular membranes.

Computational 'microscopy' refers to the use of computational resources to simulate the dynamics of a molecular system. Tuned to cell membranes, this computational 'microscopy' technique is able to capture the interplay between lipids and proteins at a spatio-temporal resolution that is unmatched by other methods. Recent advances allow us to zoom out from individual atoms and molecules to supramolecular complexes and subcellular compartments that contain tens of millions of particles, and to capture the complexity of the crowded environment of real cell membranes. This Commentary gives an overview of the main concepts of computational 'microscopy' and describes the state-of-the-art methods used to model cell membrane processes. We illustrate the power of computational modelling approaches by providing a few in-depth examples of large-scale simulations that move up from molecular descriptions into the subcellular arena. We end with an outlook towards modelling a complete cell in silico.

Exchange of Gramicidin between Lipid Bilayers: Implications for the Mechanism of Channel Formation.

The canonical mechanism of gramicidin (gA) channel formation is transmembrane dimerization of nonconducting subunits that reside in opposite bilayer leaflets. The channels do not open and close; they appear and disappear due to subunit association and dissociation. Many different types of experiments support this monomer ↔ dimer mechanism. Recently, however, this mechanism was challenged, based on experiments with lipid vesicle-incorporated gA under conditions where vesicle fusion could be controlled. In these experiments, sustained channel activity was observed long after fusion had been terminated, which led to the proposal that gA single-channel current transitions result from closed-open transitions in long-lived bilayer-spanning dimers. This proposal is at odds with 40 years of experiments, but involves the key assumption that gA monomers do not exchange between bilayers. We tested the possibility of peptide exchange between bilayers using three different types of experiments. First, we demonstrated the exchange of gA between 1,2-dierucoyl-sn-glycero-3-phosphocholine (DC22:1PC) or 1,2-dioleoyl-sn-glycero-3-phosphocholine (DC18:1PC) lipid vesicles using a fluorescence assay for gA channel activity. Second, we added gA-free DC22:1PC vesicles to both sides of planar DC18:1PC bilayers preincubated with gA, which reduced channel activity up to 10-fold. Third, we added gA-containing DC22:1PC vesicles to one or both sides of DC18:1PC planar bilayers, which produced much higher channel activity when the gA-containing vesicles were added to both sides of the bilayer, as compared to one side only. All three types of experiments show that gA subunits can exchange between lipid bilayers. The exchange of subunits between bilayers thus is firmly established, which becomes a crucial consideration with respect to the mechanism of channel formation.

Martini Coarse-Grained Force Field: Extension to RNA.

RNA has an important role not only as the messenger of genetic information but also as a regulator of gene expression. Given its central role in cell biology, there is significant interest in studying the structural and dynamic behavior of RNA in relation to other biomolecules. Coarse-grain molecular dynamics simulations are a key tool to that end. Here, we have extended the coarse-grain Martini force field to include RNA after our recent extension to DNA. In the same way DNA was modeled, the tertiary structure of RNA is constrained using an elastic network. This model, therefore, is not designed for applications involving RNA folding but rather offers a stable RNA structure for studying RNA interactions with other (bio)molecules. The RNA model is compatible with all other Martini models and opens the way to large-scale explicit-solvent molecular dynamics simulations of complex systems involving RNA.

Computational Lipidomics of the Neuronal Plasma Membrane.

Membrane lipid composition varies greatly within submembrane compartments, different organelle membranes, and also between cells of different cell stage, cell and tissue types, and organisms. Environmental factors (such as diet) also influence membrane composition. The membrane lipid composition is tightly regulated by the cell, maintaining a homeostasis that, if disrupted, can impair cell function and lead to disease. This is especially pronounced in the brain, where defects in lipid regulation are linked to various neurological diseases. The tightly regulated diversity raises questions on how complex changes in composition affect overall bilayer properties, dynamics, and lipid organization of cellular membranes. Here, we utilize recent advances in computational power and molecular dynamics force fields to develop and test a realistically complex human brain plasma membrane (PM) lipid model and extend previous work on an idealized, "average" mammalian PM. The PMs showed both striking similarities, despite significantly different lipid composition, and interesting differences. The main differences in composition (higher cholesterol concentration and increased tail unsaturation in brain PM) appear to have opposite, yet complementary, influences on many bilayer properties. Both mixtures exhibit a range of dynamic lipid lateral inhomogeneities ("domains"). The domains can be small and transient or larger and more persistent and can correlate between the leaflets depending on lipid mixture, Brain or Average, as well as on the extent of bilayer undulations.

High-Throughput Simulations Reveal Membrane-Mediated Effects of Alcohols on MscL Gating.

The mechanosensitive channels of large conductance (MscL) are bacterial membrane proteins that serve as last resort emergency release valves in case of severe osmotic downshock. Sensing bilayer tension, MscL channels are sensitive to changes in the bilayer environment and are, therefore, an ideal test case for exploring membrane protein coupling. Here, we use high-throughput coarse-grained molecular dynamics simulations to characterize MscL gating kinetics in different bilayer environments under the influence of alcohols. We performed over five hundred simulations to obtain sufficient statistics to reveal the subtle effects of changes in the membrane environment on MscL gating. MscL opening times were found to increase with the addition of the straight-chain alcohols ethanol, octanol, and to some extent dodecanol but not with hexadecanol. Increasing concentration of octanol increased the impeding effect, but only up to 10-20 mol %. Our in silico predictions were experimentally confirmed using reconstituted MscL in a liposomal fluorescent efflux assay. Our combined data reveal that the effect of alcohols on MscL gating arises not through specific binding sites but through a combination of the alcohol-induced changes to a number of bilayer properties and their alteration of the MscL-bilayer interface. Our work provides a key example of how extensive molecular simulations can be used to predict the functional modification of membrane proteins by subtle changes in their bilayer environment.

Global structural changes of an ion channel during its gating are followed by ion mobility mass spectrometry.

Mechanosensitive ion channels are sensors probing membrane tension in all species; despite their importance and vital role in many cell functions, their gating mechanism remains to be elucidated. Here, we determined the conditions for releasing intact mechanosensitive channel of large conductance (MscL) proteins from their detergents in the gas phase using native ion mobility-mass spectrometry (IM-MS). By using IM-MS, we could detect the native mass of MscL from Escherichia coli, determine various global structural changes during its gating by measuring the rotationally averaged collision cross-sections, and show that it can function in the absence of a lipid bilayer. We could detect global conformational changes during MscL gating as small as 3%. Our findings will allow studying native structure of many other membrane proteins.

The activation mode of the mechanosensitive ion channel, MscL, by lysophosphatidylcholine differs from tension-induced gating.

One of the best-studied mechanosensitive channels is the mechanosensitive channel of large conductance (MscL). MscL senses tension in the membrane evoked by an osmotic down shock and directly couples it to large conformational changes leading to the opening of the channel. Spectroscopic techniques offer unique possibilities to monitor these conformational changes if it were possible to generate tension in the lipid bilayer, the native environment of MscL, during the measurements. To this end, asymmetric insertion of l-α-lysophosphatidylcholine (LPC) into the lipid bilayer has been effective; however, how LPC activates MscL is not fully understood. Here, the effects of LPC on tension-sensitive mutants of a bacterial MscL and on MscL homologs with different tension sensitivities are reported, leading to the conclusion that the mode of action of LPC is different from that of applied tension. Our results imply that LPC shifts the free energy of gating by interfering with MscL-membrane coupling. Furthermore, we demonstrate that the fine-tuned addition of LPC can be used for controlled activation of MscL in spectroscopic studies.

Lipid organization of the plasma membrane.

The detailed organization of cellular membranes remains rather elusive. Based on large-scale molecular dynamics simulations, we provide a high-resolution view of the lipid organization of a plasma membrane at an unprecedented level of complexity. Our plasma membrane model consists of 63 different lipid species, combining 14 types of headgroups and 11 types of tails asymmetrically distributed across the two leaflets, closely mimicking an idealized mammalian plasma membrane. We observe an enrichment of cholesterol in the outer leaflet and a general non-ideal lateral mixing of the different lipid species. Transient domains with liquid-ordered character form and disappear on the microsecond time scale. These domains are coupled across the two membrane leaflets. In the outer leaflet, distinct nanodomains consisting of gangliosides are observed. Phosphoinositides show preferential clustering in the inner leaflet. Our data provide a key view on the lateral organization of lipids in one of life's fundamental structures, the cell membrane.

Membrane lipids drive formation of KRAS4b-RAF1 RBDCRD nanoclusters on the membrane.

The oncogene RAS, extensively studied for decades, presents persistent gaps in understanding, hindering the development of effective therapeutic strategies due to a lack of precise details on how RAS initiates MAPK signaling with RAF effector proteins at the plasma membrane. Recent advances in X-ray crystallography, cryo-EM, and super-resolution fluorescence microscopy offer structural and spatial insights, yet the molecular mechanisms involving protein-protein and protein-lipid interactions in RAS-mediated signaling require further characterization. This study utilizes single-molecule experimental techniques, nuclear magnetic resonance spectroscopy, and the computational Machine-Learned Modeling Infrastructure (MuMMI) to examine KRAS4b and RAF1 on a biologically relevant lipid bilayer. MuMMI captures long-timescale events while preserving detailed atomic descriptions, providing testable models for experimental validation. Both in vitro and computational studies reveal that RBDCRD binding alters KRAS lateral diffusion on the lipid bilayer, increasing cluster size and decreasing diffusion. RAS and membrane binding cause hydrophobic residues in the CRD region to penetrate the bilayer, stabilizing complexes through ß-strand elongation. These cooperative interactions among lipids, KRAS4b, and RAF1 are proposed as essential for forming nanoclusters, potentially a critical step in MAP kinase signal activation.