Peripheral membrane proteins (PMPs) can reversibly and specifically bind to biological membranes to carry out functions such as cell signalling, enzymatic activity, or membrane remodelling. Structures of these proteins and of their lipid-binding domains are typically solved in a soluble form, sometimes with a lipid or lipid headgroup at the binding site. To provide a detailed molecular view of PMP interactions with the membrane, computational methods such as molecular dynamics (MD) simulations can be applied. Here, we outline recent attempts to characterise these binding interactions, focusing on both intracellular proteins, such as phosphatidylinositol phosphate (PIP)-binding domains, and extracellular proteins such as glycolipid-binding bacterial exotoxins. We compare methods used to identify and analyse lipid-binding sites from simulation data and highlight recent work characterising the energetics of these interactions using free energy calculations. We describe how improvements in methodologies and computing power will help MD simulations to continue to contribute to this field in the future.

Peripheral membrane proteins (PMPs) bind reversibly to the surface of specific biological membranes, and thus can exist in both a soluble and membrane-bound state. Typically, membrane binding events are transient, and involve the PMP covalently (via a lipid anchor) and/or non-covalently interacting with the surface of the membrane. This ability to switch between a highly mobile soluble state and a specifically targeted membrane-bound state is often central to the PMP function. In particular, PMPs often interact with specific lipids in the membrane, making understating these interactions important to understanding the biology of PMPs.

PMPs are vital in a range of cellular processes, including signalling, trafficking, apoptosis and immunity. This makes PMPs attractive as drug targets, with approximately 30 human PMPs currently being targeted [1,2]. In particular, the lipid-binding domains of PMPs are promising targets for future therapeutics [3], as the membrane binding process is potentially open to modulation. Moreover, many PMPs are multidomain proteins, with one or more domains driving the binding to the membrane, and other domains performing the downstream function of the protein, e.g. acting as enzyme (Figure 1). This means that PMPs can potentially play important roles in biotechnology or medicine, such as drug delivery [4]. For all of these aspects of PMPs understanding the membrane binding process, which is usually driven by lipid interactions, is crucial.

Two examples of PMP action

Figure 1
Two examples of PMP action

(A) Phosphatase and tensin homologue (PTEN) is a PMP that consists of an enzymatic domain, a phosphatase domain (Ptase), and a lipid-binding C2 domain. PTEN binds to the inner leaflet of eukaryotic membranes, where Ptase dephosphorylates PI(3,4,5)P3 to form PI(4,5)P2. This process is part of a signalling pathway leading to apoptosis. (B) Cholera Toxin (Ctx) is also a PMP with an enzymatic domain, the A subunit, and a lipid-binding domain consisting of five B subunits. This B5 ring binds to the ganglioside GM1 and following the binding event, the complex is translocated into the cell by endocytosis. After translocation, the A subunit separates from the Ctx complex and carries out its cytotoxic mechanism.

Figure 1
Two examples of PMP action

(A) Phosphatase and tensin homologue (PTEN) is a PMP that consists of an enzymatic domain, a phosphatase domain (Ptase), and a lipid-binding C2 domain. PTEN binds to the inner leaflet of eukaryotic membranes, where Ptase dephosphorylates PI(3,4,5)P3 to form PI(4,5)P2. This process is part of a signalling pathway leading to apoptosis. (B) Cholera Toxin (Ctx) is also a PMP with an enzymatic domain, the A subunit, and a lipid-binding domain consisting of five B subunits. This B5 ring binds to the ganglioside GM1 and following the binding event, the complex is translocated into the cell by endocytosis. After translocation, the A subunit separates from the Ctx complex and carries out its cytotoxic mechanism.

Close modal

There are many ways in which PMPs can target specific lipids. In many cases lipid recognition is through direct electrostatic interactions, often between basic residues on the PMPs and lipids with anionic headgroups (Figure 2A). These interactions can be very strong and can involve lipids such as the monovalent phosphatidylserine (PS) or multivalent phosphatidylinositol phosphate (PIP) lipids, e.g. PIP2 and PIP3, both present in the inner leaflet of eukaryotic cell membranes. Alternatively, for PMPs which bind extracellularly, gangliosides containing negatively charged sialic acids are often important [5–7]. Electrostatic interactions can be mediated via positive ions (typically calcium) that bridge negatively charged patches on the PMPs with negative lipids [8,9] (Figure 2B). Alternative forms of interaction exist, such as insertion of hydrophobic loops [10–16] or hydrophobic helices [17] into the membrane (Figure 2C,D), cation–π interactions between aromatic residues and choline headgroups [18,19] (Figure 2E), or covalent attachment to a lipid whose hydrophobic tail is inserted into the lipid bilayer [20] (Figure 2F).

Different binding modes for PMPs

Figure 2
Different binding modes for PMPs

(A) Direct, electrostatic binding. The C2 domain of RIM2 has a cationic groove (blue) that binds to anionic PIP2 [21]. (B) Ion-promoted electrostatic binding. Annexin A2 binds to an anionic membrane bridged by calcium ions. Insert shows a Ca2+ ion bridging a negatively charged PS headgroup (red) and an aspartic acid side chain (D166) from Annexin A2 [9]. (C) Loop insertion. The FYVE domain of EEA1 (zinc ions from the four zinc fingers in blue) has a partly hydrophobic loop (V44 shown in inset) which plays a part in the membrane binding [16]. (D) Helix insertion. FakB1 inserts a helix in the membrane to expose hydrophobic residues of that helix to the hydrophobic membrane interior [22]. Note that highly mobile membrane mimetic (HMMM) was used in this study (see main text), so short lipids were used together with a hydrophobic solvent, DCLE (green), to represent the rest of the lipid tails. (E) Cation–π interactions. Membrane binding mediated by cation–π interaction between residues Y88 of PI-PLC (green) and a DMPC headgroup (red) [19]. (F) Binding by lipidation. Rab5 binds to membranes via its two geranylgeranyl lipid anchors [23]. Coordinates for the figure were kindly provided by the authors of the respective papers.

Figure 2
Different binding modes for PMPs

(A) Direct, electrostatic binding. The C2 domain of RIM2 has a cationic groove (blue) that binds to anionic PIP2 [21]. (B) Ion-promoted electrostatic binding. Annexin A2 binds to an anionic membrane bridged by calcium ions. Insert shows a Ca2+ ion bridging a negatively charged PS headgroup (red) and an aspartic acid side chain (D166) from Annexin A2 [9]. (C) Loop insertion. The FYVE domain of EEA1 (zinc ions from the four zinc fingers in blue) has a partly hydrophobic loop (V44 shown in inset) which plays a part in the membrane binding [16]. (D) Helix insertion. FakB1 inserts a helix in the membrane to expose hydrophobic residues of that helix to the hydrophobic membrane interior [22]. Note that highly mobile membrane mimetic (HMMM) was used in this study (see main text), so short lipids were used together with a hydrophobic solvent, DCLE (green), to represent the rest of the lipid tails. (E) Cation–π interactions. Membrane binding mediated by cation–π interaction between residues Y88 of PI-PLC (green) and a DMPC headgroup (red) [19]. (F) Binding by lipidation. Rab5 binds to membranes via its two geranylgeranyl lipid anchors [23]. Coordinates for the figure were kindly provided by the authors of the respective papers.

Close modal

The soluble states of lipid recognition domains of PMPs are generally straightforward to determine using X-ray crystallography or solution nuclear magnetic resonance (NMR). It can be challenging, however, to obtain a detailed atomic resolution description of the membrane-bound state of PMPs, in part because of the difficulty in stabilising membrane-bound states for structural analysis. In addition, there can be considerable complexity in the interaction which can make the exact state difficult to recreate experimentally. This includes such factors as the binding of ions [24], the macrodipole of the PMP [25], and the structural and chemical complexity of biological membranes [26].

In this review, we present how molecular dynamics (MD) simulations can be used to predict the membrane-bound state of PMPs. In the first part, we will focus on the identification of specific protein–lipid interactions, i.e. the binding of a particular lipid species to a residue or cluster of residues on the surface of a PMP. In the second part, we will review methods to calculate binding affinities and avidities for such binding events. Finally, we will outline future trends and discuss challenges in computational and experimental analyses of PMPs.

In this section, we will give a brief overview of the biological roles of some key PMPs that bind to the cytoplasmic plasma membrane. PMPs can be divided into two broad groups, intracellular and extracellular. These groups are split both in terms of their biological role and the type of lipids they bind, owing to the strongly asymmetric nature of the plasma membrane [27,28]. Here, we will mostly focus on intracellular PMPs which bind PIP lipids, and extracellular PMPs which bind to gangliosides. This is far from a comprehensive list, but rather focuses on examples which have been investigated with MD and demonstrate specific lipid-binding events.

Many cytoplasmic PMPs in eukaryotic cells use polybasic binding domains to bind negatively charged phospholipids, which are enriched in the inner cytoplasmic leaflet [29]. These include PSs, which are the most abundant anionic phospholipids in the inner leaflet. Although less abundant, the most important PMP-recruiting anionic lipids are probably phosphatidylinositols, PIs [30–32]. Importantly, PIs can be phosphorylated, which is utilised extensively in cell signalling processes, with the phosphorylation or dephosphorylation often carried out by PMPs. One example is the phosphatase and tension homologue (PTEN), which dephosphorylates PI(3,4,5)P3 to PI(4,5)P2. This is the first step in a pathway leading to apoptosis, so malfunctioning of PTEN can cause tumours and several mutations of PTEN are thus linked to cancer [33]. PTEN consists of a phosphatase domain, that dephosphorylates PIP3, and a lipid-binding C2 domain (Figure 1A). C2 is a large family of lipid-binding domains that exist in different proteins across species. Another abundant PI-binding family is that of PH domains, which is one of the largest families of lipid-binding domains, existing in more than 250 human proteins [34]. The PH domains vary in sequence, but the structure is conserved across protein families. Many PH domains show high selectivity to phosphorylated PI, thereby enabling spatial and temporal docking of their parent protein [35]. An example of a PH domain-containing protein is the General Receptor for Phosphoinositides 1 (Grp1), which binds to PIP3 with unusually high selectivity and affinity [36–38]. Other examples of lipid-binding domains are FYVE, PX, ENTH, CALM, PDZ and PTB domains [39], which also exist in many different PMPs.

As well as PMPs which bind to the plasma membrane from the cytoplasm, there are those which bind the extracellular face of the membrane. A notable category of these are bacterial exotoxins, which are secreted by certain pathogenic bacteria to target host membranes. One well-studied class is the AB5 exotoxins, which bind to specific glycolipids on the exterior of the cell membrane. A prominent example of this is cholera toxin (Ctx), secreted by the Vibrio cholerae bacterium, which has a B5 component, comprising five B subunits arranged in a ring shape, and an A subunit which sits on top of the ring [40] (Figure 1B). The B5 ring is responsible for recognising and binding specific receptors in the host membrane. For Ctx, the primary receptor is the ganglioside GM1 [41]. In addition, secondary binding sites are also present on the B5 ring for fucosylated Lewis antigens on the membrane of epithelial cells [42,43]. Another prominent example of AB5 exotoxins includes the structurally similar Escherichia coli heat-labile enterotoxin (LT), which also binds GM1 [44] as well as other gangliosides [45]. This too has secondary binding sites, this time for A-type blood antigens [42].

Identifying PMP orientations and lipid-binding sites

To understand the biological activity of a given PMP, it is essential to know how it interacts with the membrane, including its orientation relative to the lipid bilayer. For many PMPs, identifying specific interactions with lipids in the membrane is an important component of this. One way to investigate specific interactions is through computational analysis, such as with MD. In Figure 3A, we provide an overview of different MD-based computational methods commonly used to determine the orientation of PMPs, and the binding sites for specific lipids, and discuss each in turn below.

Strategies for simulating PMP binding with MD simulations

Figure 3
Strategies for simulating PMP binding with MD simulations

(A-i) All-atom (AA) simulation is performed, and binding sites are identified, e.g. with PyLipID or ProLint. (A-ii) Experiment or fragment docking simulations are used to identify binding sites of a PMP in solution. The fragments are, in this context, lipid headgroup mimics. The protein can then be solvated in silico, and fragments replaced by whole lipids. (A-iii) The Highly Mobile Membrane Mimetic (HMMM) model or coarse-grained (CG) simulation are used to sample binding events of PMP binding to lipid membrane, and binding sites are subsequently identified. To verify the simulations, the binding sites are converted into AA resolution and refined in an AA simulation. The panel shows the modular conversion of palmitoyl-oleoyl-glycero-phosphatidylcholine (POPC) from CG into AA resolution using the program CG2AT [46]; panel adapted from https://github.com/owenvickery/cg2at. (B) Overview of different simulation approaches and how they qualitatively compare with respect to resolution and computational cost. Implicit methods utilise mean force description of membrane and/or solvent to speed up the calculations. CG, HMMM, united-atom (UA) and AA are all described in the main text. Quantum mechanics/molecular mechanics (QM/MM) simulations are precise and can be used to model formation and breaking of covalent bonds. QM/MM simulations are (as of yet) too computationally expensive for simulations of PMPs binding to lipid membranes, but included for completeness.

Figure 3
Strategies for simulating PMP binding with MD simulations

(A-i) All-atom (AA) simulation is performed, and binding sites are identified, e.g. with PyLipID or ProLint. (A-ii) Experiment or fragment docking simulations are used to identify binding sites of a PMP in solution. The fragments are, in this context, lipid headgroup mimics. The protein can then be solvated in silico, and fragments replaced by whole lipids. (A-iii) The Highly Mobile Membrane Mimetic (HMMM) model or coarse-grained (CG) simulation are used to sample binding events of PMP binding to lipid membrane, and binding sites are subsequently identified. To verify the simulations, the binding sites are converted into AA resolution and refined in an AA simulation. The panel shows the modular conversion of palmitoyl-oleoyl-glycero-phosphatidylcholine (POPC) from CG into AA resolution using the program CG2AT [46]; panel adapted from https://github.com/owenvickery/cg2at. (B) Overview of different simulation approaches and how they qualitatively compare with respect to resolution and computational cost. Implicit methods utilise mean force description of membrane and/or solvent to speed up the calculations. CG, HMMM, united-atom (UA) and AA are all described in the main text. Quantum mechanics/molecular mechanics (QM/MM) simulations are precise and can be used to model formation and breaking of covalent bonds. QM/MM simulations are (as of yet) too computationally expensive for simulations of PMPs binding to lipid membranes, but included for completeness.

Close modal

Atomistic MD simulations

Atomistic simulations are a well-established and widely used flavour of MD. These simulations explicitly model all-atoms (AA), including hydrogens, as point charges with no electronic structure (Figure 3B). A variant is united-atom (UA) force fields, where certain hydrogens in the protein are incorporated into the particles of associated heavy atoms, so that e.g. –CH3 might be represented as a single particle. UA force fields are computationally less costly but with slightly lower resolution, although they are not necessarily less accurate than AA force fields [47]. There are many choices to make when setting up atomistic MD simulations of PMPs. A key decision is the choice of force field used to describe the protein, solvent and lipids, with many options available, including different UA and AA force fields. In addition, in certain force fields the protein and/or solvent can be polarised, for a potentially more accurate view of certain molecular phenomena [48]. Here, we will mostly focus on AA force fields, e.g. CHARMM36m [49], Amber [50] and OPLS-AA [51]. For a more detailed discussion on these choices, we recommend referring to [52–54].

The use of AA can in principle be very simple, as all that is needed is to place the PMP near a desired membrane and run an MD simulation, e.g. as done for the yeast oxysterol-binding protein (Osh4) [11]. Unfortunately, due to the long time scales needed to sample PMP binding to membranes, AA MD may be computationally too expensive even for small PMPs or isolated PMP domains. There are solutions available, such as using a massively parallel supercomputer, for example ANTON, which has been used to simulate Bruton’s tyrosine kinase binding to a PIP3 containing membrane [55]. Alternatively, crowd-sourcing computational resource is another solution to generate enough sampling data, such as shown using Folding@home to model protein kinase C binding to the membrane [56]. The high cost of these simulations is still prohibitive to many research groups, especially if multiple conformations or membrane lipid compositions are to be investigated, however, the initial hurdle of generating the bound state can be overcome through either interpretation of experimental data or by a more approximative computational simulation. AA can then be applied to provide a high-resolution view of binding with a potentially much lower computational cost (Figure 3A). These approaches will be described in the following sections.

Use of experimental information to seed AA MD simulations

A well-established method of generating an initial pose for AA MD is through integration of experimental data. A prominent example of this is with structural data, for instance where a fragment of a bound lipid has been co-crystallised with a PMP. This has been applied to AB5 exotoxins such as Ctx (Figure 1B), which has a structurally well-defined GM1-binding pocket. In one computational study, the authors placed the toxin above a membrane and manually replaced the structurally resolved sugars with full GM1 molecules from the membrane, to seed an MD simulation [57]. Alternatively, if the PMP binds to an integral membrane protein as well as forming specific lipid interactions, then structural information of the complex can be used to orient the PMP, as is seen for the structure of bacterial SecA bound to SecYEG [58]. From this input state, it is straightforward to setup subsequent MD simulations to identify lipid interactions [59]. Of course, this approach is reliant on being able to resolve a structure of the PMP bound to an integral protein.

Another technique often used is electron paramagnetic resonance (EPR). Early examples of this include work incorporating EPR to model the C2 domain of cytosolic phospholipase A2 on a PC membrane [60]. Docking information from EPR has also been applied in this manner to the C2 domain of Protein Kinase Cα to PS and PIP2 [61] and the Grp1 PH domain interactions with PIP3 [62].

In theory, a range of methods can be combined to get a prior idea of the binding mode. For instance, Ohkubo and Tajkhorshid used both X-ray crystallography, NMR and fluorescence data to orient the GLA domain on a membrane [63]. The group then used steered MD to pull the domain into the membrane and generate a bound state [63]. A similar approach was also used on protein kinase C, with the 200 ns steered MD trajectory used to seed subsequent Folding@home simulations [56]. Therefore, if the binding interface is already known, use of steered MD [64] can be a powerful way of sampling membrane binding in AA relatively quickly.

If experimental data are unavailable: docking of lipid head fragments

In the case that structural data are not available, or a different bound group is desired, a similar effect can be achieved using computational docking, where a soluble mimic of a lipid headgroup is docked on to the PMP-binding pocket. Whilst not necessarily as robust as structural analysis, this approach is able to form a solid basis for downstream analysis using MD. This approach has, e.g., been applied to investigate the PIP-binding sites of Smurf1-C2 [65], as well as interactions of the E. coli heat-labile enterotoxin with the blood group A pentasaccharide [42]. The high-throughput nature of the approach also makes docking of other molecules, such as potential therapeutic drugs, feasible, as shown with an expansive analysis of alkaloid binding to Ctx [66].

Accelerating lipid diffusion by the highly mobile membrane mimetic model

An appealing method to achieve more effective sampling in AA simulations is the highly mobile membrane mimetic (HMMM) model [67,68]. In this approach, most of the volume normally filled by the lipid tails is replaced by an organic solvent, whilst the lipids are represented by their short-chain homologues (Figure 2D). HMMM increases the diffusion rates of lipids in the membrane allowing relevant binding sites to be identified more efficiently. Examples of HMMM being applied successfully are in the identification of several orientations of KRas4b binding to membranes containing anionic lipids [69] and in the identification of canonical and alternative binding sites for PIP3 on the Grp1 PH domain [70].

Accelerating computational performance by coarse graining

In addition to AA MD and HMMM, coarse-grained (CG) MD methods are popular for looking at PMP interactions with membranes. In physics-based CG approaches, groups of atoms are represented by a single particle rather than explicitly. This considerably reduces the degrees of freedom needed to describe a specific system, as well as allowing a longer MD timestep to be used, both of which potentially increase simulation speeds by several orders of magnitude. Different levels of coarse graining, as well as the application of continuum models, are outlined in detail by Moquadam et al. [52], and will be briefly discussed below.

In terms of characterising specific protein–lipid interactions, considerable success has been achieved using the Martini 2 [71–73] and 3 [74] force fields, as well as variants of this: polarisable Martini [75,76], BMW-Martini with polarisable water [77,78] and Martini 2.3P for cation–π binding [79]. Martini 3 was recently tested on 12 different PMPs and lipid-binding peptides, and the authors found that the force field was in most cases able to identify experimentally known binding interfaces [80]. Some false-negative and false-positive results were, however, also reported, meaning that a validity check with AA simulations of the CG-binding poses is recommended. An important observation was that the elastic network which is often applied to conserve secondary and tertiary protein structure in Martini, prevented the PMPs from undergoing conformational changes upon binding. Aside from Martini, other CG force fields are also able to be used for looking at PMP–lipid interactions, including SIRAH [81–83], ELBA [84] and dissociative particle dynamics (DPD) [85,86]. Each have different advantages, for instance the SIRAH force field allows the secondary structure to change which may allow a more realistic view of the bound state. Of note, when choosing a force field it is worth considering what lipids are already parametrised in that model [87].

Implicit or continuum models are typically very fast for PMP surface identification [88], however specific PMP–lipid interactions can be more challenging to model this way. Poses generated from this can, however, be used to guide MD simulations. An early example of this is work done on prostaglandin H2 synthase-1 binding to a PC membrane [89]. In particular, this approach could potentially be extremely powerful for orienting PMPs prior to simulation, with several tools and databases available for this, including the DREAMM tool (http://dreamm.ni4os.eu/; [90]) and the Orientations of Proteins in Membranes database [15,91].

PMP self-assembly with CG MD can reveal intermediate states

Coarse graining makes it possible to simulate microsecond trajectories as a matter of course with basic computer hardware. This is very important for PMP/membrane systems, where the membrane and PMP need time to equilibrate and sample multiple configurations, often including PMP rotational dynamics, which can be relatively slow. The first CG simulations with PMPs encountering lipid membranes were performed 15 years ago, such as for SGTX1 binding a palmitoyl-oleoyl-glycero-phosphatidylcholine (POPC) or POPE/POPG membrane [92]. Since then, the method has been applied to many different systems, such as PIP5K1A [93] and several PH domains, including Grp1 [94], Dok7 [95] and kindlin-3 [96]. As computational power increases, even more comprehensive analyses are possible [97].

It is worth noting that many of these studies improve their sampling of PMP–membrane interactions through running several independently initiated simulations, typically between 5 and 25 repeats per system. The reported variation between repeats can be significant, reflecting the complicated energy landscape for a PMP with several membrane interaction sites binding a membrane with different lipid constituents, as well as the likely non-ergodic nature of single MD trajectories. We advise that the number of repeats be carefully set depending on how variable the simulations are from each other, but at the least five to ten repeats are generally advisable. Through these repeats it is often possible to see multiple minima reflecting different binding sites on the same PMP [21,93,95]. Some of these may reflect genuine metastable states which act as important intermediates before the final binding pose. Such intermediate binding poses are extremely difficult to determine experimentally, as any ensemble data will likely be dominated by the final and energetically most favourable binding pose. MD simulations can thus provide a unique insight into the binding (and unbinding) pathways of PMPs.

The best of both worlds: multiscale modelling

The speed-up obtained through coarse graining does come at a price. For the Martini force field this includes a simplified description of electrostatic interactions [98] and exaggerated protein stickiness [99], as well as other potential limitations. To ameliorate this, multiscaling can be utilised. Here, the configurational landscape of given system is rapidly sampled using CG, such that key states can be converted into AA description for analysis of these state in higher resolution [46,100]. An example of this being applied is to the voltage-sensitive phosphatase [101] where, following pose-generation with CG, subsequent AA simulations allowed the position of a bound PIP3 to be described in higher resolution than possible in CG. Of course, it is possible for any changes that occur in the AA to be converted back into CG for additional analysis. This might be important when looking at conformational changes, which can be difficult to accurately model using CG where an elastic network is often applied to stabilise the protein structure.

For HMMM, the increased diffusion rate likewise comes with a penalty. For instance, it is unclear how the exchange of lipid tails for organic solvent will impact the entropy of the membrane, and therefore the energetics of PMP binding. Similarly, the effects of different lipid tails and cholesterol presence are difficult to accurately account for. However, it is also possible to convert HMMM representations into an AA description once the desired pose has been sampled [68]. This was done, for instance, to describe the functional cycle of the fatty acid transfer protein FakB1 [22] or GRP1 PH binding to anionic lipid [70]. Similarly, systems can be converted from CG into HMMM description, as has recently been done for the binding of vinculin to PIP2 [102].

Utilising multiscale modelling for comparative studies of protein domain families

The speed-up provided by coarse-graining combined with hardware and software improvements (including improved GPU utilisation, e.g. [103,104]), has allowed MD studies of whole families of lipid-binding domains. This includes a comprehensive study of more than ten different PH domains and their interaction with PIP2- and PIP3-containing membranes [105], as well as a study of six different C2 domains [21] binding to anionic and zwitterionic model membranes. Such studies highlight the utility of multiscale simulations, where many systems can be concurrently analysed to provide a high-resolution comparison between domains from different proteins.

Selection/identification of binding sites from a trajectory

Whether a trajectory is generated by CG, HMMM or AA MD simulation, a critical step is the identification of lipid-binding sites from the trajectory data (Figure 3). Analysis by visual inspection is always a useful step, but this can be somewhat subjective and time-consuming, especially if using multiple repeats. Therefore, systematic analysis strategies are necessary.

As an initial analysis, different binding poses can be mapped out through simple geometric analysis of the trajectories. An example of this is in production of a 2D heatmap comparing PMP–membrane distance vs PMP orientation [21,93,95]. By quantifying the sampling density of the PMP along these two coordinates, it is possible to draw qualitative observations about stable and metastable binding modes, especially in combining multiple simulations into a single ensemble. Important residues and binding pockets which interact with membrane lipids can then be identified by, e.g., contact analysis. This is a somewhat complex exercise, but fortunately specialised tools have recently been developed for this purpose, including ProLint and PyLipID, which we will discuss below.

Dedicated analytical tools for analysis and identification of lipid binding sites

The study of PMP interactions with lipids is related to the more established field of ligand binding to soluble proteins. For this latter topic, a range of dedicated tools exists for identification and analysis of binding pockets [106]. The methods are, however, not directly transferable to PMPs even for specific lipid binding sites, as there are many differences between a soluble ligand and a membrane-bound lipid, including the limited rotation and vertical (membrane-perpendicular) translation of the lipid. Therefore, dedicated tools can be useful. We highlight two recent examples below.

The first is ProLint [107]. ProLint allows the user to visualise and analyse lipid–protein-binding sites and, e.g., identify contacts between specific lipid types and residues. A contact is defined as a distance below a user defined cut-off (between 3 and 8 Å). By examining, e.g., a heatmap of lipid contacts on the PMP, the user can identify binding sites. ProLint can be used as a Python package or via the web interface.

The second tool is PyLipID [108]. PyLipID is a Python-based package which reads in trajectory files and creates an interaction profile for each individual residue–lipid pair over the course of the simulation. It then clusters residues which simultaneously bind the same lipid, to allow identification of specific lipid sites. A similar approach was applied to study interactions of Kir2.2 with cholesterol [109] and with a more complex mixture of lipids [110]. The method has conceptual overlap with a recent study which measured the hydrogen bonds between PIP2 and vinculin at each simulation snapshot, and used this to generate binding sites [102]. One useful feature of PyLipID is that it gives statistics for identified binding sites, including site occupancy (% of time a lipid spends in the binding site) and a residence time and koff. Like ProLint, these measures rely on a user-provided cut-off, typically between 4 and 8 Å. Additionally, PyLipID produces representative poses of the bound lipid for either creation of figures or downstream analysis.

There are multiple additional programs designed to analyse membranes from MD simulations. These might not necessarily be optimal for looking at specific protein–lipid interactions but might offer useful tools for looking at the interaction of the PMP with the membrane more generally, including how the membrane changes upon PMP binding. Of particular note are FATSLiM [111], MemSurfer [112], LOOS [113,114] and LiPyphilic [115]. This range of programs, along with ProLint and PyLipID mentioned above, showcase the increasing desire for more rigorous and detailed analyses of membrane simulations, and the impressive commitment of the academic community to fulfil this demand. Going forward, we expect much more progress in both how people develop and use scientific software [116].

Complex binding sites

The structural resolution of MD also allows more complex interactions to be observed, including the clustering (leading to avidity effects) or perturbation of lipids upon binding. Here, we will briefly outline some examples of what type of more complex binding events have been investigated using MD.

Lipid variety

The type of lipid, including variation in both headgroup and lipid tails, can potentially make a big difference in how PMPs associate with the membrane. For example, a recent study used AA MD coupled with fluorescence microscopy to demonstrate that saturated and unsaturated GM1 have different properties when bound to Ctx (Figure 1B), including ability to co-cluster with GPI-anchored proteins [117]. The authors also showed that saturated GM1 was more able to couple trans-bilayer to PS than unsaturated GM1, which has implications for the ability of Ctx to interact with the actin cytoskeleton.

Multiple lipids and lipid clustering

Structural techniques such as X-ray crystallography and cryoelectron microscopy can, in some cases, resolve one or more tightly bound lipids (e.g. [118,119]). However, MD simulations on the same systems can often reveal additional lipid binding sites around the PMP. For instance, there are several MD studies which report on the importance of multivalent PIPs binding to PMPs, and suggest that PIP clustering can help stabilise the PMP on the membrane. Notably, Yamamoto et al. investigated PIP3 bound to the PH domain of GRP1 and showed that experimental binding values could only be reproduced in simulations after including multiple PIPs in the binding event [120] (Figure 4C). That is, avidity must be considered as affinities between single molecules are not trivially additive. Binding of multiple PIP2 were likewise observed for different membrane-binding C2 domains [21], for Ebola virus matrix protein VP40 [121] and for Brag2 [122], and multiple PI(4)P were found to bind the FYVE domain of early endosome antigen 1 [16].

Free energy calculation methods in MD
Figure 4
Free energy calculation methods in MD

(A) Schematic illustration of two enhanced sampling methods, that both alter the energy landscape to force the simulation out of energy minima; accelerated MD (aMD) and metadynamics (MetaD), see the main text. (B) Simple potential of mean force (PMF) calculation. The system is sampled along a reaction coordinate, typically the membrane–protein distance, and the force necessary to keep the protein in place along this coordinate is monitored. From these forces, a PMF can be calculated. Shown is C2 from PTEN bound to an anionic membrane [21]. (C) Complex PMF. PMFs calculated for binding of PH domains with, respectively, 1, 2, 3, 5 and 10 PIP3 bound [120]. (D) Thermodynamic cycle as used in FEP for switching between PIP3 and POPS for a PMP system. The difference between ΔGPIP3-POPS(bound) and ΔGPIP3-POPS(free) is the relative difference in PMP–lipid interaction for PIP3 and POPS. Below are the values from preliminary analyses of Grp1 PH-binding FEP calculations. Note that these data were generated for this review, with a full manuscript to follow.

Figure 4
Free energy calculation methods in MD

(A) Schematic illustration of two enhanced sampling methods, that both alter the energy landscape to force the simulation out of energy minima; accelerated MD (aMD) and metadynamics (MetaD), see the main text. (B) Simple potential of mean force (PMF) calculation. The system is sampled along a reaction coordinate, typically the membrane–protein distance, and the force necessary to keep the protein in place along this coordinate is monitored. From these forces, a PMF can be calculated. Shown is C2 from PTEN bound to an anionic membrane [21]. (C) Complex PMF. PMFs calculated for binding of PH domains with, respectively, 1, 2, 3, 5 and 10 PIP3 bound [120]. (D) Thermodynamic cycle as used in FEP for switching between PIP3 and POPS for a PMP system. The difference between ΔGPIP3-POPS(bound) and ΔGPIP3-POPS(free) is the relative difference in PMP–lipid interaction for PIP3 and POPS. Below are the values from preliminary analyses of Grp1 PH-binding FEP calculations. Note that these data were generated for this review, with a full manuscript to follow.

Close modal

MD can also investigate how the binding of multiple lipids affects the orientation of the PMP. As an example, several AB5 exotoxin structures have been identified bound to fragments of their specific glycolipid receptors. This binding is multivalent; there are five binding sites and five bound lipids. However, Sridhar et al. [123] used CG simulations to allow Ctx to freely bind GM1 in the membrane, and saw that GM1 typically bound three out of the five sites, which lead to a tilted conformation. Both tilting [124] and occupancy of three sites [57] have also been seen in AA, and the occupancy of three sites was shown to be the dominant state for Ctx using flow cytometry [125].

Another application of MD is the study of non-specific lipid clustering and raft formation upon binding of PMPs. One example is the raft-formation of dipalmitoyl-glycero-phosphocholine (DPPC) induced by influenza hemagglutinin [126]. Similarly, DPD simulations have demonstrated both that multiple copies of Ctx drive regions of lipid clustering [127], and that multiple copies of Shiga toxin can cluster on the membrane [128].

Lipid order

A third notable membrane variation that can be studied with MD is that of lipid order. A prominent example of this is a study of the matrix domain of HIV-1 gag, which has a myristoylated N-terminal domain that inserts into the bilayer upon binding. However, this insertion was shown only to take place in ordered membrane models, which contained cholesterol and only a small amount of unsaturated lipid tails [129].

Membrane curvature and very CG force fields

Several MD studies have investigated PMP-induced membrane curvature and some PMPs are localised to, or cause, regions of membrane curvature (see, e.g. [20,124,130–132]). We have, however, chosen not to cover this in depth in the present review, as we focus on more specific interactions between PMPs and lipids. For the same reason, we have chosen not to cover coarser membrane and protein representations than the Martini representation, such as with mesoscale simulations [133]. That said, incorporating mesoscale resolutions into studies along with CG and AA [134] can potentially be very powerful, especially for looking at such phenomena as membrane curvature [135]. Similarly, continuum models can be used to rapidly predict PMP association with regions of curvature via the PPM 3.0 webserver (https://opm.phar.umich.edu/ppm_server3; [136]).

Comparison with experimental data

Finally, although MD is a powerful tool in the identification and characterisation of protein–lipid interactions, integration with experimental data is still very powerful, both for validation and provision of a broader biological context to the data. Therefore, future MD work will continue to require additional experimental data. Promising steps are being made in structural experimental techniques to measure specific protein–lipid interactions.

X-ray crystallography has for many years been the dominating technique for determining specific lipids cites on PMPs, by co-crystallisation with lipid mimetic molecules, and has, e.g., been used to identify the PIP binding site of FYVE from EEA1 [118] and several PH domains [119]. NMR can also be used to resolve lipid binding sites of PMPs [137] and is complementary to crystallography as it provides the solution structure and contribute with dynamic information [22], such as how residue side chains rearrange upon membrane binding.

In recent years, cryoelectron microscopy (cryo-EM) has improved its resolution to a point where tightly bound lipids can be resolved from the electron density maps, as seen for integral membrane proteins, e.g., the P4-ATPase lipid flippase [138]. To our knowledge, however, there are still no cryo-EM structures of lipids bound to PMPs. One advantage of cryo-EM over crystallography, is that cryo-EM can solve protein structures bound to lipid nanodiscs or lipid nanotubes, and thus be used to determine the orientation with respect to that model membrane. Cryo-EM has, e.g., been used to determine the binding orientation of factor VIII [139].

Neutron and X-ray reflectometry is another promising technique for assessing PMP binding orientation. Here, the average electron or neutron scattering length densities can be measured with high accuracy perpendicular to the plane of the membrane. This has been utilised, e.g., to determine the orientation of tubulin [140], and to investigate membrane-induced conformational changes of alpha synuclein [141]. As mentioned previously, EPR is another important technique for determining orientation of membrane-bound PMPs [142].

Finally, small-angle neutron scattering (SANS) provides an interesting tool for investigating membrane-induced conformational changes of PMPs. By exchanging hydrogen with deuterium in solvent and/or lipids or detergents, the signal from the membrane mimic can be rendered effectively invisible. In that way, the membrane mimic is not providing any signal, and membrane-induced structural change of the PMP can be measured. This was done for myelin and matched-out micelles [143]. That method is limited in resolution by the difference between micelle and lipid head and tail-groups, as only the average scattering is matched out. Thus, the signal from the micelles/lipids is only matched out at the low q-values (low resolution). To improve this, special partly deuterated detergent and lipids have been developed to match out in the full-measured range [144]. This method has been shown feasible for bicelles [145], micelles [146] and nanodiscs [147]. Despite improvements, SANS is limited to approximately 10 Å resolution at best.

Use of molecular dynamics to estimate binding affinities of interactions

Increasingly, many researchers are not just interested in ‘if’ and ‘where’ a given lipid might bind to a PMP of interest, but also how strong the interaction is, i.e. its affinity. This can be important, as determining the affinity provides insight into how likely a given interaction is in the complex biological context of the cell. Research into lipid affinities for PMPs is arguably less developed than for integral membrane proteins [148], but many of the developments made there are now also being applied to PMPs.

There are different ways of estimating affinities from MD simulations, the simplest being to run an unbiased simulation and compare the likelihood of the PMP being bound to the lipids vs unbound. This method typically requires a prohibitive amount of sampling (i.e. several hundred microseconds) to achieve convergence, as several binding and unbinding events must take place during the simulation. Thus, enhanced sampling MD simulations are often run, in which the underlying potential of the force field is perturbed to drive the simulation to sample the landscape of PMP–lipid binding more efficiently.

Avoiding energy minima with enhanced sampling techniques

A range of enhanced sampling methods have been developed to drive simulations towards sampling of rare events. These methods can be split into reaction coordinate-based methods and reaction coordinate-free methods [149]. A reaction coordinate, also denoted a collective variable, defines a coordinate that promotes the reaction of interest. In this case, the reaction is likely to be binding/unbinding of the PMP to a lipid membrane, in other cases the reaction could be ligand binding, or reactions interpreted in a broader sense, such as conformational changes.

Metadynamics (MetaD) is an example of a method that needs a reaction coordinate. The MetaD algorithm stores the values of that variable while running the simulation and gives energetic penalties when states are revisited [150,151], preventing the simulation becoming trapped in energy minima (Figure 4A). The added potential can then be processed to produce a free energy landscape. A variation of MetaD is well-tempered MetaD, where the applied potential is gradually decreased to achieve convergence [152]. Well-tempered MetaD was recently applied to get 1D energy landscapes for the Grp1 PH domain binding to different membranes [70]. Notably, MetaD is not restricted to a single reaction coordinate (usually denoted ‘collective variable’ in the context of MetaD), so MetaD has also been applied to produce 2D energy landscapes for integral membrane proteins using CG simulations [153] and for a C2 domain using AA simulations [154], suggesting that this is a viable option for PMPs, where both distance from the membrane and rotation of the PMP might be of interest.

Accelerated MD (aMD) is an example of a reaction coordinate-free enhanced sampling method. The underlying potential is raised if it is below a defined threshold energy [155]. Just like MetaD, the purpose is to drive the simulation out of energy minima to sample more unlikely events (Figure 4A). In a recent study, aMD was used to investigate the flexibility of Lewis Y antigens when bound to Ctx, and MM/GBSA (see section below) was used to determine affinities between the antigens and Ctx [156].

It can be challenging to define a reaction coordinate, e.g. for conformational changes. Moreover, reaction coordinate-based simulations are biased towards a given reaction pathway and are therefore not explorative in nature. Thus, reaction coordinate-free methods are alluring. However, for most PMP/membrane systems, the dominant reaction coordinate is easy to define, namely the membrane–protein distance, and, more importantly, defining a reaction coordinate can prevent unwanted events; if a reaction coordinate is not defined for the system, there will inevitably be many ways out of a given energy minimum (Figure 4A), which can lead to, e.g. irrelevant conformational changes or unfolding of secondary structure etc.

Potential of mean force calculations

One prime example of a free energy method for PMP–membrane interactions is potential of mean force (PMF) calculations. There are different ways of running PMF calculations, with one of the simplest being applying a directional steering force (steered MD) to the system along a reaction coordinate of interest, in this case usually the distance between the centre of mass of the PMP and the centre of mass of the membrane. The work required to move the system along this reaction coordinate when steered using a constant velocity can be recorded and used to generate an estimate of the free energy of binding [157]. It is important to make sure that many repeats are run, as well as different velocities, to ensure that the computed forces are reproducible. This method has been applied previously to the PH domain from RRP, whereby binding energies were estimated for different membranes [158], and to the PMP dihydroorotate dehydrogenase [159]. In the latter, the PMP was pushed into a model membrane to estimate the PMP insertion depth, by monitoring the opposing reaction force.

PMFs can also be constructed through the application of umbrella sampling. Umbrella sampling is a reaction coordinate-based enhanced sampling method. In umbrella sampling, frames are typically extracted from a steered MD simulation at regular positions along the reaction coordinate. This time, the steered MD is not used directly for calculating the PMF, but the extracted frames are used as starting points, meaning that typically a single repeat will suffice. These frames are then used for a parallel series of simulations (umbrella windows), in which the protein is constrained to maintain the input value of the collective variable, i.e., stay at a specific distance from the membrane. The forces imposed by umbrella potential can then be used to construct a PMF using a technique such as the weighted histogram analysis method [160].

The umbrella sampling PMF method has been applied to a number of PMPs, including studies of PH domains [94,120], C2 domains [21], the FYVE domain of early endosome antigen 1 [16] and the related lipidated GTPase Rab5 [23], which are both involved in early endosome recruiting. An example for a C2 domain is given in Figure 4B. Umbrella sampling PMFs have also be constructed for different binding modes of the soluble domain from the mitochondrial translocase subunit TIM50 [161]. In all of those studies, the PMF was generated using CG MD. PMFs have also been calculated from AA simulations, e.g. to estimate the membrane penetration depth of the C2A domain of synaptotagmin [24] and to probe the binding energy of actin-binding proteins bound to PIP-containing membrane [162].

Free energy perturbation calculations

A different method for estimating the interaction energy between a PMP and a specific lipid is free energy perturbation (FEP) calculations. Here, alchemical transformations between lipid species are performed on an unbound lipid in the membrane and for the lipid bound to the protein, and the ΔG of each process computed using a technique such as MBAR [163]. The difference between the free and bound lipids, the ΔΔG, can then be calculated (Figure 4D). The calculations are normally run for the target lipid to a generic lipid, however the target lipid can also be converted into a different lipid of interest, such as switching from PIP3 to PI(4,5)P2 to PI(4)P to PI to compare the energetics of how different PIP lipids bind the same site [164]. We suggest references [153,165] for a more detailed run-down of this process.

Analysis done on integral membrane proteins suggests that the values obtained from FEP in CG are approximately equivalent to PMFs [153], which has also been shown for PMPs [21]. But FEPs generally require less sampling time, which can be very beneficial, especially when looking at multiple different lipids binding to the same protein. For PMPs in particular, FEPs are potentially very advantageous, as the length of a PMF reaction coordinate for a PMP can be up to 4–5 nm [16], and convergence can be difficult due to the slow rotation of the PMP. In addition, the ‘free’ and ‘bound’ FEP calculations can theoretically be done in the same simulation system, with the bound lipid in one membrane leaflet and the free in the other. This would serve to reduce computation time even further.

Certain technical advances would make FEP of PMP–lipid systems even more computationally efficient, in particular the use of non-equilibrium FEP [166]. Here, moderate (e.g. 100 ns) simulations are first run in both states A (target lipid) and B (generic lipid), and then very short (e.g. 200 ps) non-equilibrium simulations are run for several snapshots (e.g. 80) going from A to B and B to A. To enable statistical analysis, the whole cycle is repeated a number of times (e.g. 10 repeats). As for standard FEP, this is run for both the unbound lipid in the membrane, and the lipid bound to the protein. For our test simulations on the Grp1-PH domain, non-equilibrium FEP agrees well with equilibrium FEP (Figure 4D; ‘eq FEP’ vs ‘non-eq FEP’) but costing only ∼2 vs 10 µs. In addition, by combining both the free and bound lipid in the same system, this can be halved to only 1 µs per system, without impacting the final value (Figure 4D; ‘non-eq FEP dual’). Other potential savings in time might come from ways of increasing the sampling rate within each window, such as with the accelerated weight histogram method [167,168]. This will be particularly useful for any FEP calculations run using AA force fields.

Binding affinities from MM/GBSA and MM/PBSA

Binding affinities can also be estimated using the previously mentioned MM/GBSA, as well as the conceptually similar MM/PBSA. These approaches calculate binding energies by combining the covalent and non-covalent forces that are directly computed the by molecular mechanical force field (MM) with the solvation free energy change. This latter term is split into a polar term, computed using either the Generalised Born (GB) or Poisson–Boltzmann (PB) equation, and a non-polar term, related to formation of the solute cavity, approximated using the solvent-accessible surface area (SA) of the ligand. MM/GBSA and MM/PBSA are computationally fast methods for ranking ligand affinities, especially if precise values are not required [169]. However, there are several factors to account for when running these calculations [170], and there are only very few studies using these methods on lipids (see e.g. [171,172]), so it remains unclear how accurate they are for this purpose. In particular, special care should be taken when looking at charged lipids [173].

Advanced energy analysis methods from unbiased simulations

The free energy of a system can also be extracted from a series of trajectories using Markov State Modelling (MSM), where the transition states and corresponding transition rates are estimated post hoc from the trajectory [56], see reference [174] for a more detailed overview. As for many enhanced sampling methods, relevant reaction coordinates are needed for MSM to be run. For PMP/lipid simulations, such reaction coordinates could be membrane–protein distance, orientation of the PMP with respect to the membrane, or coordinates describing conformational changes in the PMP upon binding. As an MSM provides information about transitions, it constitutes a promising tool for precise descriptions of complex binding events involving intermediate states.

Another intriguing new method demonstrates that it may be possible to directly extract affinities in the form of apparent dissociation constants (Kd) from the simulation of binding saturation curves [175]. However, as the MSM, this method is still reliant on the unbinding of the PMP from the membrane, which might require very long (>100 µs) simulation times for certain PMP–lipid compositions. Therefore, it might be preferable to instead employ enhanced sampling techniques such as MetaD, aMD or umbrella sampling.

Comparison with experimental data

Comparison of predicted free energy values with those from experimental data is beneficial. Surface plasmon resonance is the most frequently used experimental technique assessing PMF/membrane-binding energies, and has been used for a range of protein domains, including PH domains [36,176], FYVE domains [177–179], ANTH [17], cytosolic phospholipase A2α [180], the adaptor protein Amot [181] and the Bcl2-associated agonist of cell death (BAD) [182]. Competitive FRET and isothermal titration calorimetry can likewise determine binding free energies [119,183].

A central question is whether the free energy values from simulations correspond well with those from experimental analysis. Good consistency with experimental values has been observed for AA FEP calculations on three different outer-membrane associating PMPs [19]. This study used a version of the CHARMM36 force field optimised for cation–π interactions [184,185] (Figure 2D). Energies determined by PMF from CG simulations with Martini 2 likewise showed good consistency for the PH domain from GRP1 [120] and the FYVE domain of EEA1 [16]. However, we note that poor agreements are less likely to be published than good agreements, so more systematic and comprehensive experimental validation of computational results will likely be essential in future.

Here, we have presented an overview of work done to identify and characterise PMP binding to specific lipids in biological membranes using MD. We have focused on PMPs, and point readers also interested in integral membrane proteins to other reviews [165,186,187]. We have chosen to highlight a number of key works where researchers have identified specific protein–lipid interactions, and several studies where the energetics are also considered.

Future studies may aim to incorporate more advanced data analysis methods to eventually reduce the dependence on lengthy MD simulations for site identification, e.g. machine learning approaches. Current work has demonstrated that certain structural features of PMPs can be used to rapidly predict membrane-binding properties, including hydrophobic protrusions [188]. This means that certain predictions can be made extremely cheaply, such as with the PePrMint webserver (https://reuter-group.github.io/peprmint/).

There exists a number of challenges facing development of MD for PMP analysis, which arise from the biological complexity behind the binding process. For instance, such biological processes as how protein complexes form at the surface of biological membrane are out of reach for all but the lengthiest simulations [55]. Additionally, more care should be taken in how systems are built for simulation. As an example, a recent study has demonstrated that the protonation states of charged PIP lipids might also play a considerable role in PMP binding [189], something which is often overlooked. Similarly, key features like how ions help stabilise the PMP binding increase complexity and hence may require costly simulations, e.g. with polarisable force fields (Figure 3B).

A key area of future work will be to see how membrane association affects the PMP itself. This is more challenging to address than simple site identification, in part because it typically requires far longer time scales for the effects to be observed, and the restrictive nature of coarse-graining on protein structure, including the elastic network in the Martini force field [74], means that AA simulations are likely necessary. In addition, there is far less experimental data to compare with, making interpretation of MD data more difficult. In a recent study, Gullett et al. describe the binding of a bacterial fatty acid protein in its open and closed states, allowing production of its complete exchange cycle [22]. Here, MD was performed separately on an open and closed conformation, both determined by X-ray crystallography. A current challenge is thus to directly simulate binding-induced conformational change of PMPs. As such, much of the current focus has been on smaller and more localised changes. For instance, several studies have looked at how loop dynamics are altered on membrane association, including for cytochrome P450 [190], or conformational changes of peptides. As an example of the latter, a study by Davis and Berkowitz reported increased helical content of an amyloid-β peptide upon binding to an anionic membrane [191].

Finally, more complex analysis methods are needed to shed light on transient binding modes, intermediate modes and binding pathways. An obvious tool for this would be Markov State Models as mentioned previously. This includes the challenge of directly combining simulations and experiments to achieve more accurate models for the transition states [192].

Going forward, we anticipate many advances in MD algorithms and force fields, improved hardware, better analysis methods and superior strategies for integration with experimental data. This will all result in fruitful insight into the binding of PMPs to lipid membranes. MD may also play a key role in development of a new class of drugs, targeting PMPs or membranes [3].

The authors declare that there are no competing interests associated with the manuscript.

This work was supported by the Carlsberg Foundation [grant number CF19-0288 (to A.H.L.)]; the Lundbeck Foundation [grant number R347-2020-2339 (to A.H.L.)]; the UKRI-BBSRC Interdisciplinary Bioscience Doctoral Training Partnership [grant number BB/M011224/1 (to L.H.J.)]; the ISIS Facility Development Studentship Program (to L.H.J.); the Wellcome [grant number 208361/Z/17/Z (to M.S.P.S. and R.A.C.)]; and the free energy calculations were run using the ARCHER/ARCHER2 U.K. National Supercomputing Service (http://www.archer.ac.uk), provided by HECBioSim, the U.K. High End Computing Consortium for Biomolecular Simulation (hecbiosim.ac.uk), which is supported by the EPSRC [grant number EP/L000253/1].

Open access for this article was enabled by the participation of University of Oxford in an all-inclusive Read & Publish agreement with Portland Press and the Biochemical Society under a transformative agreement with JISC.

     
  • AA

    all-atom

  •  
  • aMD

    accelerated MD

  •  
  • CG

    coarse-grained

  •  
  • cryo-EM

    cryoelectron microscopy

  •  
  • Ctx

    cholera toxin

  •  
  • DPD

    dissociative particle dynamics

  •  
  • EPR

    electron paramagnetic resonance

  •  
  • FEP

    free energy perturbation

  •  
  • GB

    Generalised Born

  •  
  • Grp1

    general receptor for phosphoinositides 1

  •  
  • HMMM

    highly mobile membrane mimetic

  •  
  • MD

    molecular dynamics

  •  
  • MetaD

    metadynamics

  •  
  • MM

    molecular mechanical force field

  •  
  • MSM

    Markov State Modelling

  •  
  • NMR

    nuclear magnetic resonance

  •  
  • PB

    Poisson–Boltzmann

  •  
  • PC

    phosphatidylcholine

  •  
  • PE

    phosphatidylethanolamine

  •  
  • PIP

    phosphatidylinositol phosphate

  •  
  • PMF

    potential of mean force

  •  
  • PMP

    peripheral membrane protein

  •  
  • PS

    phosphatidylserine

  •  
  • PTEN

    phosphatase and tensin homologue

  •  
  • SA

    surface area

  •  
  • SANS

    small-angle neutron scattering

  •  
  • UA

    united-atom

1.
Rask-Andersen
M.
,
Almén
M.S.
and
Schiöth
H.B.
(
2011
)
Trends in the exploitation of novel drug targets
.
Nat. Rev. Drug Discov.
10
,
579
590
[PubMed]
2.
Tiefenauer
L.
and
Demarche
S.
(
2012
)
Challenges in the development of functional assays of membrane proteins
.
Materials (Basel)
5
,
2205
2242
3.
Boes
D.M.
,
Godoy-Hernandez
A.
and
McMillan
D.G.G.
(
2021
)
Peripheral membrane proteins: promising therapeutic targets across domains of life
.
Membranes
11
,
346
[PubMed]
4.
Guan
J.
,
Qian
J.
and
Zhan
C.
(
2020
)
Preparation of cholera toxin subunit B functionalized nanoparticles for targeted therapy of glioblastoma
.
Drug Delivery Systems
, pp.
207
212
,
Springer
5.
Dong
M.
,
Liu
H.
,
Tepp
W.H.
,
Johnson
E.A.
,
Janz
R.
and
Chapman
E.R.
(
2008
)
Glycosylated SV2A and SV2B mediate the entry of botulinum neurotoxin E into neurons
.
Mol. Biol. Cell.
19
,
5226
5237
[PubMed]
6.
Zanchetti
G.
,
Colombi
P.
,
Manzoni
M.
,
Anastasia
L.
,
Caimi
L.
,
Borsani
G.
et al.
(
2007
)
Sialidase NEU3 is a peripheral membrane protein localized on the cell surface and in endosomal structures
.
Biochem. J.
408
,
211
219
[PubMed]
7.
Ledeen
R.W.
and
Wu
G.
(
2018
)
Chapter Fifteen - Gangliosides, α-Synuclein, and Parkinson’s disease
. In
Gangliosides in Health and Disease
(
Schnaar
R.L.
, ed.), pp.
435
454
,
Academic Press
,
Lopez PHHBT-P in MB and TS
8.
Corbalan-Garcia
S.
and
Gómez-Fernández
J.C.
(
2014
)
Signaling through C2 domains: more than one lipid target
.
Biochim. Biophys. Acta Biomembr.
1838
,
1536
1547
9.
Hakobyan
D.
,
Gerke
V.
and
Heuer
A.
(
2017
)
Modeling of annexin A2–membrane interactions by molecular dynamics simulations
.
PLoS ONE
12
,
e0185440
[PubMed]
10.
Psachoulia
E.
and
Sansom
M.S.P.
(
2009
)
PX- and FYVE-mediated interactions with membranes: simulation studies
.
Biochemistry
48
,
5090
5095
[PubMed]
11.
Rogaski
B.
and
Klauda
J.B.
(
2012
)
Membrane-Binding mechanism of a peripheral membrane protein through microsecond molecular dynamics simulations
.
J. Mol. Biol.
423
,
847
861
[PubMed]
12.
McDougle
D.R.
,
Baylon
J.L.
,
Meling
D.D.
,
Kambalyal
A.
,
Grinkova
Y.V.
,
Hammernik
J.
et al.
(
2015
)
Incorporation of charged residues in the CYP2J2 F-G loop disrupts CYP2J2-lipid bilayer interactions
.
Biochim. Biophys. Acta Biomembr.
1848
,
2460
2470
13.
Khan
H.M.
,
He
T.
,
Fuglebakk
E.
,
Grauffel
C.
,
Yang
B.
,
Roberts
M.F.
et al.
(
2016
)
A role for weak electrostatic interactions in peripheral membrane protein binding
.
Biophys. J.
110
,
1367
1378
[PubMed]
14.
Monje-Galvan
V.
and
Klauda
J.B.
(
2016
)
Peripheral membrane proteins: tying the knot between experiment and computation
.
Biochim. Biophys. Acta Biomembr.
1858
,
1584
1593
15.
Lomize
A.L.
,
Pogozheva
I.D.
,
Lomize
M.A.
and
Mosberg
H.I.
(
2007
)
The role of hydrophobic interactions in positioning of peripheral proteins in membranes
.
BMC Struct. Biol.
7
,
44
[PubMed]
16.
Larsen
A.H.
,
Tata
L.
,
John
L.H.
and
Sansom
M.S.P.
(
2021
)
Membrane-binding mechanism of the EEA1 FYVE domain revealed by multi-scale molecular dynamics simulations
.
PLoS Comput. Biol.
17
,
e1008807
[PubMed]
17.
Stahelin
R.V.
,
Long
F.
,
Peter
B.J.
,
Murray
D.
,
De Camilli
P.
,
McMahon
H.T.
et al.
(
2003
)
Contrasting membrane interaction mechanisms of AP180 N-terminal homology (ANTH) and Epsin N-terminal homology (ENTH) domains
.
J. Biol. Chem.
278
,
28993
28999
[PubMed]
18.
Grauffel
C.
,
Yang
B.
,
He
T.
,
Roberts
M.F.
,
Gershenson
A.
and
Reuter
N.
(
2013
)
Cation−π interactions as lipid-specific anchors for phosphatidylinositol-specific phospholipase C
.
J. Am. Chem. Soc.
135
,
5740
5750
[PubMed]
19.
Waheed
Q.
,
Khan
H.M.
,
He
T.
,
Roberts
M.
,
Gershenson
A.
and
Reuter
N.
(
2019
)
Interfacial aromatics mediating cation−π interactions with choline-containing lipids can contribute as much to peripheral protein affinity for membranes as aromatics inserted below the phosphates
.
J. Phys. Chem. Lett.
10
,
3972
3977
[PubMed]
20.
Masone
D.
,
Uhart
M.
and
Bustos
D.M.
(
2018
)
Bending lipid bilayers: a closed-form collective variable for effective free-energy landscapes in quantitative biology
.
J. Chem. Theor. Comput.
14
,
2240
2245
[PubMed]
21.
Larsen
A.H.
and
Sansom
M.S.P.
(
2021
)
Binding of Ca2+-independent C2 domains to lipid membranes: a multi-scale molecular dynamics study
.
Structure
29
,
1200
1213
[PubMed]
22.
Gullett
J.M.
,
Cuypers
M.G.
,
Grace
C.R.
,
Pant
S.
,
Subramanian
C.
,
Tajkhorshid
E.
et al.
(
2022
)
Identification of structural transitions in bacterial fatty acid binding proteins that permit ligand entry and exit at membranes
.
J. Biol. Chem.
298
,
[PubMed]
23.
Edler
E.
,
Schulze
E.
and
Stein
M.
(
2017
)
Membrane localization and dynamics of geranylgeranylated Rab5 hypervariable region
.
Biochim. Biophys. Acta Biomembr.
1859
,
1335
1349
[PubMed]
24.
Chon
N.L.
,
Osterberg
J.R.
,
Henderson
J.
,
Khan
H.M.
,
Reuter
N.
,
Knight
J.D.
et al.
(
2015
)
Membrane docking of the synaptotagmin 7 C2A domain: computation reveals interplay between electrostatic and hydrophobic contributions
.
Biochemistry
54
,
5696
5711
[PubMed]
25.
Galassi
V.V.
,
Villarreal
M.A.
and
Montich
G.G.
(
2018
)
Relevance of the protein macrodipole in the membrane-binding process. Interactions of fatty-acid binding proteins with cationic lipid membranes
.
PLoS ONE
13
,
e0194154
[PubMed]
26.
Coskun Ü
S.K.
(
2011
)
Cell membranes: the lipid perspective
.
Structure
19
,
1543
1548
[PubMed]
27.
Devaux
P.F.
(
1991
)
Static and dynamic lipid asymmetry in cell membranes
.
Biochemistry
30
,
1163
1173
[PubMed]
28.
Fadeel
B.
and
Xue
D.
(
2009
)
The ins and outs of phospholipid asymmetry in the plasma membrane: roles in health and disease
.
Crit. Rev. Biochem. Mol. Biol.
44
,
264
277
[PubMed]
29.
Clarke
R.J.
,
Hossain
K.R.
and
Cao
K.
(
2020
)
Physiological roles of transverse lipid asymmetry of animal membranes
.
Biochim. Biophys. Acta Biomembr.
1862
,
183382
[PubMed]
30.
Balla
T.
(
2013
)
Phosphoinositides: tiny lipids with giant impact on cell regulation
.
Physiol. Rev.
93
,
1019
1137
[PubMed]
31.
Hammond GR
V.
and
Balla
T.
(
2015
)
Polyphosphoinositide binding domains: key to inositol lipid biology
.
Biochim. Biophys. Acta Mol. Cell. Biol. Lipids
1851
,
746
758
32.
Di Paolo
G.
and
De Camilli
P.
(
2006
)
Phosphoinositides in cell regulation and membrane dynamics
.
Nature
443
,
651
657
[PubMed]
33.
Milella
M.
,
Falcone
I.
,
Conciatori
F.
,
Cesta Incani
U.
,
Del Curatolo
A.
,
Inzerilli
N.
et al.
(
2015
)
PTEN: multiple functions in human malignant tumors
.
Front. Oncol.
5
,
24
[PubMed]
34.
Singh
N.
,
Reyes-Ordoñez
A.
,
Compagnone
M.A.
,
Moreno
J.F.
,
Leslie
B.J.
,
Ha
T.
et al.
(
2021
)
Redefining the specificity of phosphoinositide-binding by human PH domain-containing proteins
.
Nat. Commun.
12
,
4339
[PubMed]
35.
Stahelin
R.V.
,
Scott
J.L.
and
Frick
C.T.
(
2014
)
Cellular and molecular interactions of phosphoinositides and peripheral proteins
.
Chem. Phys. Lipids
182
,
3
18
[PubMed]
36.
Manna
D.
,
Albanese
A.
,
Park
W.S.
and
Cho
W.
(
2007
)
Mechanistic basis of differential cellular responses of phosphatidylinositol 3,4-bisphosphate- and phosphatidylinositol 3,4,5-trisphosphate-binding pleckstrin homology domains
.
J. Biol. Chem.
282
,
32093
32105
[PubMed]
37.
Narayan
K.
and
Lemmon
M.A.
(
2006
)
Determining selectivity of phosphoinositide-binding domains
.
Methods
39
,
122
133
[PubMed]
38.
Naughton
F.B.
,
Kalli
A.C.
and
Sansom
M.S.P.
(
2016
)
Association of peripheral membrane proteins with membranes: free energy of binding of GRP1 PH domain with phosphatidylinositol phosphate-containing model bilayers
.
J. Phys. Chem. Lett.
7
,
1219
1224
[PubMed]
39.
Balla
T.
(
2005
)
Inositol-lipid binding motifs: signal integrators through protein-lipid and protein-protein interactions
.
J. Cell Sci.
118
,
2093
2104
[PubMed]
40.
Zhang
R.-G.
,
Scott
D.L.
,
Westbrook
M.L.
,
Nance
S.
,
Spangler
B.D.
,
Shipley
G.G.
et al.
(
1995
)
The three-dimensional crystal structure of cholera toxin
.
J. Mol. Biol.
251
,
563
573
[PubMed]
41.
Kuziemko
G.M.
,
Stroh
M.
and
Stevens
R.C.
(
1996
)
Cholera toxin binding affinity and specificity for gangliosides determined by surface plasmon resonance
.
Biochemistry
35
,
6375
6384
[PubMed]
42.
Ångström
J.
,
Bäckström
M.
,
Berntsson
A.
,
Karlsson
N.
,
Holmgren
J.
,
Karlsson
K.-A.
et al.
(
2000
)
Novel carbohydrate binding site recognizing blood group A and B determinants in a hybrid of cholera toxin and Escherichia coli heat-labile enterotoxin B-subunits
.
J. Biol. Chem.
275
,
3231
3238
[PubMed]
43.
Heim
J.B.
,
Hodnik
V.
,
Heggelund
J.E.
,
Anderluh
G.
and
Krengel
U.
(
2019
)
Crystal structures of cholera toxin in complex with fucosylated receptors point to importance of secondary binding site
.
Sci. Rep.
9
,
1
14
[PubMed]
44.
Moss
J.
,
Osborne
J.C.
Jr
et al.
(
1981
)
Escherichia coli heat-labile enterotoxin. Ganglioside specificity and ADP-ribosyltransferase activity
.
J. Biol. Chem.
256
,
12861
12865
45.
MacKenzie
C.R.
,
Hirama
T.
,
Lee
K.K.
,
Altman
E.
and
Young
N.M.
(
1997
)
Quantitative analysis of bacterial toxin affinity and specificity for glycolipid receptors by surface plasmon resonance
.
J. Biol. Chem.
272
,
5533
5538
[PubMed]
46.
Vickery
O.N.
and
Stansfeld
P.J.
(
2021
)
CG2AT2: an enhanced fragment-based approach for serial multi-scale molecular dynamics simulations
.
J. Chem. Theor. Comput.
17
,
6472
6482
[PubMed]
47.
da Silva
G.C.Q.
,
Silva
G.M.
,
Tavares
F.W.
,
Fleming
F.P.
and
Horta
B.A.C.
(
2020
)
Are all-atom any better than united-atom force fields for the description of liquid properties of alkanes?
J. Mol. Model.
26
,
296
[PubMed]
48.
Jing
Z.
,
Liu
C.
,
Cheng
S.Y.
,
Qi
R.
,
Walker
B.D.
,
Piquemal
J.-P.
et al.
(
2019
)
Polarizable force fields for biomolecular simulations: recent advances and applications
.
Annu. Rev. Biophys.
48
,
371
394
[PubMed]
49.
Huang
J.
,
Rauscher
S.
,
Nawrocki
G.
,
Ran
T.
,
Feig
M.
,
De Groot
B.L.
et al.
(
2017
)
CHARMM36m: an improved force field for folded and intrinsically disordered proteins
.
Nat. Methods
14
,
71
73
[PubMed]
50.
Wang
J.
,
Wolf
R.M.
,
Caldwell
J.W.
,
Kollman
P.A.
and
Case
D.A.
(
2004
)
Development and testing of a general amber force field
.
J. Comput. Chem.
25
,
1157
1174
[PubMed]
51.
Jorgensen
W.L.
,
Maxwell
D.S.
and
Tirado-Rives
J.
(
1996
)
Development and testing of the OPLS all-atom force field on conformational energetics and properties of organic liquids
.
J. Am. Chem. Soc.
118
,
11225
11236
52.
Moqadam
M.
,
Tubiana
T.
,
Moutoussamy
E.E.
and
Reuter
N.
(
2021
)
Membrane models for molecular simulations of peripheral membrane proteins
.
Adv. Phys. X
6
,
1932589
53.
Monje-Galvan
V.
,
Warburton
L.
and
Klauda
J.B.
(
2019
)
Setting up all-atom molecular dynamics simulations to study the interactions of peripheral membrane proteins with model lipid bilayers
.
Methods Mol. Biol.
1949
,
325
339
54.
Jefferies
D.
and
Khalid
S.
(
2021
)
Atomistic and coarse-grained simulations of membrane proteins: a practical guide
.
Methods
185
,
15
27
[PubMed]
55.
Wang
Q.
,
Pechersky
Y.
,
Sagawa
S.
,
Pan
A.C.
and
Shaw
D.E.
(
2019
)
Structural mechanism for Bruton’s tyrosine kinase activation at the cell membrane
.
Proc. Natl. Acad. Sci. U.S.A.
116
,
9390
9399
56.
Ryckbosch
S.M.
,
Wender
P.A.
and
Pande
V.S.
(
2017
)
Molecular dynamics simulations reveal ligand-controlled positioning of a peripheral protein complex in membranes
.
Nat. Commun.
8
,
6
[PubMed]
57.
Basu
I.
and
Mukhopadhyay
C.
(
2014
)
Insights into binding of cholera toxin to GM1 containing membrane
.
Langmuir
30
,
15244
15252
[PubMed]
58.
Zimmer
J.
,
Nam
Y.
and
Rapoport
T.A.
(
2008
)
Structure of a complex of the ATPase SecA and the protein-translocation channel
.
Nature
455
,
936
943
[PubMed]
59.
Corey
R.A.
,
Pyle
E.
,
Allen
W.J.
,
Watkins
D.W.
,
Casiraghi
M.
,
Miroux
B.
et al.
(
2018
)
Specific cardiolipin-SecY interactions are required for proton-motive force stimulation of protein secretion
.
Proc. Natl. Acad. Sci. U.S.A.
115
,
7967
7972
[PubMed]
60.
Jaud
S.
,
Tobias
D.J.
,
Falke
J.J.
and
White
S.H.
(
2007
)
Self-induced docking site of a deeply embedded peripheral membrane protein
.
Biophys. J.
92
,
517
524
[PubMed]
61.
Lai
C.-L.
,
Landgraf
K.E.
,
Voth
G.A.
and
Falke
J.J.
(
2010
)
Membrane docking geometry and target lipid stoichiometry of membrane-bound PKCα C2 domain: a combined molecular dynamics and experimental study
.
J. Mol. Biol.
402
,
301
310
[PubMed]
62.
Lai
C.-L.
,
Srivastava
A.
,
Pilling
C.
,
Chase
A.R.
,
Falke
J.J.
and
Voth
G.A.
(
2013
)
Molecular mechanism of membrane binding of the GRP1 PH domain
.
J. Mol. Biol.
425
,
3073
3090
[PubMed]
63.
Ohkubo
Y.Z.
and
Tajkhorshid
E.
(
2008
)
Distinct structural and adhesive roles of Ca2+ in membrane binding of blood coagulation factors
.
Structure
16
,
72
81
[PubMed]
64.
Isralewitz
B.
,
Gao
M.
and
Schulten
K.
(
2001
)
Steered molecular dynamics and mechanical functions of proteins
.
Curr. Opin. Struct. Biol.
11
,
224
230
[PubMed]
65.
Scott
J.L.
,
Frick
C.T.
,
Johnson
K.A.
,
Liu
H.
,
Yong
S.S.
,
Varney
A.G.
et al.
(
2020
)
Molecular analysis of membrane targeting by the C2 domain of the E3 ubiquitin ligase Smurf1
.
Biomolecules
10
,
229
66.
Gangopadhyay
A.
,
Chakraborty
H.J.
and
Datta
A.
(
2019
)
Employing virtual screening and molecular dynamics simulations for identifying hits against the active cholera toxin
.
Toxicon
170
,
1
9
[PubMed]
67.
Ohkubo
Y.Z.
,
Pogorelov
T.V.
,
Arcario
M.J.
,
Christensen
G.A.
and
Tajkhorshid
E.
(
2012
)
Accelerating membrane insertion of peripheral proteins with a novel membrane mimetic model
.
Biophys. J.
102
,
2130
2139
68.
Qi
Y.
,
Cheng
X.
,
Lee
J.
,
Vermaas
J.V.
,
Pogorelov
T.V.
,
Tajkhorshid
E.
et al.
(
2015
)
Builder for membrane simulations with the highly mobile membrane-mimetic model
.
Biophys. J.
109
,
2012
2022
[PubMed]
69.
Gregory
M.C.
,
McLean
M.A.
and
Sligar
S.G.
(
2017
)
Interaction of KRas4b with anionic membranes: a special role for PIP2
.
Biochem. Biophys. Res. Commun.
487
,
351
355
[PubMed]
70.
Pant
S.
and
Tajkhorshid
E.
(
2020
)
Microscopic characterization of GRP1 PH domain interaction with anionic membranes
.
J. Comput. Chem.
41
,
489
499
[PubMed]
71.
Marrink
S.J.
,
Risselada
H.J.
,
Yefimov
S.
,
Tieleman
D.P.
and
Vries
A.H.De.
(
2007
)
The MARTINI Force Field: coarse grained model for biomolecular simulations
.
J. Phys. Chem. B
111
,
7812
7824
[PubMed]
72.
López
C.A.
,
Sovova
Z.
,
Van Eerden
F.J.
,
De Vries
A.H.
and
Marrink
S.J.
(
2013
)
Martini force field parameters for glycolipids
.
J. Chem. Theor. Comput.
9
,
1694
1708
[PubMed]
73.
Monticelli
L.
,
Kandasamy
S.K.
,
Periole
X.
,
Larson
R.G.
,
Tieleman
D.P.
and
Marrink
S.J.
(
2008
)
The MARTINI coarse grained force field: extension to proteins
.
J. Chem. Theor. Comput.
4
,
819
834
[PubMed]
74.
Souza
P.C.T.
,
Alessandri
R.
,
Barnoud
J.
,
Thallmair
S.
,
Faustino
I.
,
Grünewald
F.
et al.
(
2021
)
Martini 3: a general purpose force field for coarse-grained molecular dynamics
.
Nat. Methods
18
,
382
388
[PubMed]
75.
Yesylevskyy
S.O.
,
Schäfer
L.V.
,
Sengupta
D.
and
Marrink
S.J.
(
2010
)
Polarizable water model for the coarse-grained MARTINI force field
.
PLoS Comput. Biol.
6
,
1
17
76.
Michalowsky
J.
,
Schäfer
L.V.
,
Holm
C
and
Smiatek
J.
(
2017
)
A refined polarizable water model for the coarse-grained MARTINI force field with long-range electrostatic interactions
.
J. Chem. Phys.
146
,
[PubMed]
77.
Wu
Z.
,
Cui
Q.
and
Yethiraj
A.
(
2011
)
A new coarse-grained force field for membrane-peptide simulations
.
J. Chem. Theor. Comput.
7
,
3793
3802
[PubMed]
78.
Miguel
V.
,
Perillo
M.A.
and
Villarreal
M.A.
(
2016
)
Improved prediction of bilayer and monolayer properties using a refined BMW-MARTINI force field
.
Biochim. Biophys. Acta Biomembr.
1858
,
2903
2910
79.
Khan
H.M.
,
Souza
P.C.T.
,
Thallmair
S.
,
Barnoud
J.
,
de Vries
A.H.
,
Marrink
S.J.
et al.
(
2020
)
Capturing choline-aromatics cation−π interactions in the MARTINI force field
.
J. Chem. Theor. Comput.
16
,
2550
2560
[PubMed]
80.
Srinivasan
S.
,
Zoni
V.
and
Vanni
S.
(
2021
)
Estimating the accuracy of the MARTINI model towards the investigation of peripheral protein-membrane interactions
.
Faraday Discuss.
232
,
131
148
[PubMed]
81.
Darré
L.
,
Machado
M.R.
,
Brandner
A.F.
,
González
H.C.
,
Ferreira
S.
and
Pantano
S.
(
2015
)
SIRAH: a structurally unbiased coarse-grained force field for proteins with aqueous solvation and long-range electrostatics
.
J. Chem. Theor. Comput.
11
,
723
739
[PubMed]
82.
Barrera
E.E.
,
Machado
M.R.
and
Pantano
S.
(
2019
)
Fat SIRAH: coarse-grained phospholipids to explore membrane-protein dynamics
.
J. Chem. Theor. Comput.
15
,
5674
5688
[PubMed]
83.
Barrera
E.E.
,
Frigini
E.N.
,
Porasso
R.D.
and
Pantano
S.
(
2017
)
Modeling DMPC lipid membranes with SIRAH force-field
.
J. Mol. Model.
23
,
259
[PubMed]
84.
Orsi
M.
and
Essex
J.W.
(
2011
)
The ELBA force field for coarse-grain modeling of lipid membranes
.
PLoS ONE
6
,
e28637
[PubMed]
85.
Groot
R.D.
and
Warren
P.B.
(
1997
)
Dissipative particle dynamics: bridging the gap between atomistic and mesoscopic simulation
.
J. Chem. Phys.
107
,
4423
4435
86.
Benjamini
A.
and
Smit
B.
(
2013
)
Lipid mediated packing of transmembrane helices - a dissipative particle dynamics study
.
Soft Matter
9
,
2673
2683
87.
Loschwitz
J.
,
Olubiyi
O.O.
,
Hub
J.S.
,
Strodel
B.
and
Poojari
C.S.
(
2020
)
Chapter Seven - Computer simulations of protein-membrane systems
. In
Computational Approaches for Understanding Dynamical Systems: Protein Folding and Assembly
(
Strodel
B.
, ed.), pp.
273
403
,
Academic Press
,
Barz BBT-P in MB and TS, editors
88.
Zhang
L.
,
Yethiraj
A.
and
Cui
Q.
(
2014
)
Free energy calculations for the peripheral binding of proteins/peptides to an anionic membrane. 1. Implicit membrane models
.
J. Chem. Theor. Comput.
10
,
2845
2859
[PubMed]
89.
Nina
M.
,
Bernèche
S.
and
Roux
B.
(
2000
)
Anchoring of a monotopic membrane protein: the binding of prostaglandin H 2 synthase-1 to the surface of a phospholipid bilayer
.
Eur. Biophys. J.
29
,
439
454
[PubMed]
90.
Chatzigoulas
A.
and
Cournia
Z.
(
2022
)
Predicting protein-membrane interfaces of peripheral membrane proteins using ensemble machine learning
.
Brief. Bioinform.
23
,
bbab518
[PubMed]
91.
Lomize
M.A.
,
Pogozheva
I.D.
,
Joo
H.
,
Mosberg
H.I.
and
Lomize
A.L.
(
2012
)
OPM database and PPM web server: resources for positioning of proteins in membranes
.
Nucleic Acids Res.
40
,
D370
D376
[PubMed]
92.
Wee
C.L.
,
Bemporad
D.
,
Sands
Z.A.
,
Gavaghan
D.
and
Sansom
M.S.P.
(
2007
)
SGTx1, a Kv channel gating-modifier toxin, binds to the interfacial region of lipid bilayers
.
Biophys. J.
92
,
L07
L09
[PubMed]
93.
Amos
S.-B.T.A.
,
Kalli
A.C.
,
Shi
J.
and
Sansom
M.S.P.
(
2019
)
Membrane recognition and binding by the phosphatidylinositol phosphate kinase pip5k1a: a multiscale simulation study
.
Structure
27
,
1336.e2
1346.e2
[PubMed]
94.
Naughton
F.B.
,
Kalli
A.C.
and
Sansom
M.S.P.
(
2018
)
Modes of interaction of pleckstrin homology domains with membranes: toward a computational biochemistry of membrane recognition
.
J. Mol. Biol.
430
,
372
388
[PubMed]
95.
Buyan
A.
,
Kalli
A.C.
and
Sansom
M.S.P.
(
2016
)
Multiscale simulations suggest a mechanism for the association of the Dok7 PH domain with PIP-containing membranes
.
PLoS Comput. Biol.
12
,
e1005028
[PubMed]
96.
Ni
T.
,
Kalli
A.C.
,
Naughton
F.B.
,
Yates
L.A.
,
Naneh
O.
,
Kozorog
M.
et al.
(
2017
)
Structure and lipid-binding properties of the kindlin-3 pleckstrin homology domain
.
Biochem. J.
474
,
539
556
[PubMed]
97.
Le Huray
K.I.P.
,
Wang
H.
,
Sobott
F.
and
Kalli
A.C.
(
2021
)
Systematic simulation of the interactions of Pleckstrin homology domains with membranes
.
bioRxiv
,
98.
Alessandri
R.
,
Souza
P.C.T.
,
Thallmair
S.
,
Melo
M.N.
,
De Vries
A.H.
and
Marrink
S.J.
(
2019
)
Pitfalls of the martini model
.
J. Chem. Theor. Comput.
15
,
5448
5460
[PubMed]
99.
Javanainen
M.
,
Martinez-Seara
H.
and
Vattulainen
I.
(
2017
)
Excessive aggregation of membrane proteins in the Martini model
.
PLoS ONE
12
,
e0187936
[PubMed]
100.
Wassenaar
T.A.
,
Pluhackova
K.
,
Böckmann
R.A.
,
Marrink
S.J.
and
Tieleman
D.P.
(
2014
)
Going backward: a flexible geometric approach to reverse transformation from coarse grained to atomistic models
.
J. Chem. Theor. Comput.
10
,
676
690
[PubMed]
101.
Kalli
A.C.
,
Devaney
I.
and
Sansom
M.S.P.
(
2014
)
Interactions of phosphatase and tensin homologue (PTEN) proteins with phosphatidylinositol phosphates: insights from molecular dynamics simulations of PTEN and voltage sensitive phosphatase
.
Biochemistry
53
,
1724
1732
[PubMed]
102.
Braun
L.
,
Schoen
I.
and
Vogel
V.
(
2021
)
PIP2-induced membrane binding of the vinculin tail competes with its other binding partners
.
Biophys. J.
120
,
4608
4622
[PubMed]
103.
Abraham
M.J.
,
Murtola
T.
,
Schulz
R.
,
Páll
S.
,
Smith
J.C.
,
Hess
B.
et al.
(
2015
)
GROMACS: high performance molecular simulations through multi-level parallelism from laptops to supercomputers
.
SoftwareX
1-2
,
19
25
104.
Phillips
J.C.
,
Hardy
D.J.
,
Maia
J.D.C.
,
Stone
J.E.
,
Ribeiro
J.V.
,
Bernardi
R.C.
et al.
(
2020
)
Scalable molecular dynamics on CPU and GPU architectures with NAMD
.
J. Chem. Phys.
153
,
44130
105.
Yamamoto
E.
,
Kalli
A.C.
,
Yasuoka
K.
and
Sansom
M.S.P.
(
2016
)
Interactions of pleckstrin homology domains with membranes: adding back the bilayer via high-throughput molecular dynamics
.
Structure
24
,
1421
1431
[PubMed]
106.
Feng
T.
and
Barakat
K.
(
2018
)
Molecular Dynamics Simulation and Prediction of Druggable Binding Sites BT - Computational Drug Discovery and Design
. (
Gore
M.
and
Jagtap
U.B.
, eds), pp.
87
103
,
NY: Springer New York
,
New York
107.
Sejdiu
B.I.
and
Tieleman
D.P.
(
2021
)
ProLint: a web-based framework for the automated data analysis and visualization of lipid-protein interactions
.
Nucleic Acids Res.
49
,
W544
W550
[PubMed]
108.
Song
W.
,
Corey
R.A.
,
Ansell
T.B.
,
Cassidy
C.K.
,
Horrell
M.R.
,
Duncan
A.L.
et al.
(
2022
)
PyLipID: a python package for analysis of protein-lipid interactions from molecular dynamics simulations
.
J. Chem. Theor. Comput.
18
,
1188
1201
[PubMed]
109.
Barbera
N.
,
Ayee
M.A.A.
,
Akpa
B.S.
and
Levitan
I.
(
2018
)
Molecular dynamics simulations of Kir2. 2 interactions with an ensemble of cholesterol molecules
.
Biophys. J.
115
,
1264
1280
[PubMed]
110.
Duncan
A.L.
,
Corey
R.A.
and
Sansom
M.S.P.
(
2020
)
Defining how multiple lipid species interact with inward rectifier potassium (Kir2) channels
.
Proc. Natl. Acad. Sci. U.S.A.
117
,
7803
7813
111.
Buchoux
S.
(
2017
)
FATSLiM: a fast and robust software to analyze MD simulations of membranes
.
Bioinformatics
33
,
133
134
[PubMed]
112.
Bhatia
H.
,
Ingólfsson
H.I.
,
Carpenter
T.S.
,
Lightstone
F.C.
and
Bremer
P.-T.
(
2019
)
MemSurfer: a tool for robust computation and characterization of curved membranes
.
J. Chem. Theor. Comput.
15
,
6411
6421
[PubMed]
113.
Romo
T.D.
,
Leioatts
N.
and
Grossfield
A.
(
2014
)
Lightweight object oriented structure analysis: tools for building tools to analyze molecular dynamics simulations
.
J. Comput. Chem.
35
,
2305
2318
[PubMed]
114.
Grossfield
A.
and
Romo
T.D.
(
2021
)
Loos, a better tool to analyze molecular dynamics simulations
.
Biophys. J.
120
,
178a
115.
Smith
P.
and
Lorenz
C.D.
(
2021
)
LiPyphilic: a Python toolkit for the analysis of lipid membrane simulations
.
J. Chem. Theory Comput.
17
,
5907
5919
[PubMed]
116.
Grossfield
A.
(
2022
)
How to be a good member of a scientific software community
.
Living J. Comput. Mol. Sci.
3
,
1473
117.
Arumugam
S.
,
Schmieder
S.
,
Pezeshkian
W.
,
Becken
U.
,
Wunder
C.
,
Chinnapen
D.
et al.
(
2021
)
Ceramide structure dictates glycosphingolipid nanodomain assembly and function
.
Nat. Commun.
12
,
1
12
[PubMed]
118.
Dumas
J.J.
,
Merithew
E.
,
Sudharshan
E.
,
Rajamani
D.
,
Hayes
S.
,
Lawe
D.
et al.
(
2001
)
Multivalent endosome targeting by homodimeric EEA1
.
Mol. Cell
8
,
947
958
[PubMed]
119.
Ferguson
K.M.
,
Kavran
J.M.
,
Sankaran
V.G.
,
Fournier
E.
,
Isakoff
S.J.
,
Skolnik
E.Y.
et al.
(
2000
)
Structural basis for discrimination of 3-phosphoinositides by pleckstrin homology domains
.
Mol. Cell
6
,
373
384
[PubMed]
120.
Yamamoto
E.
,
Domański
J.
,
Naughton
F.B.
,
Best
R.B.
,
Kalli
A.C.
,
Stansfeld
P.J.
et al.
(
2020
)
Multiple lipid binding sites determine the affinity of PH domains for phosphoinositide-containing membranes
.
Sci. Adv.
6
,
eaay5736
[PubMed]
121.
GC
J.B.
,
Gerstman
B.S.
and
Chapagain
P.P.
(
2017
)
Membrane association and localization dynamics of the Ebola virus matrix protein VP40
.
Biochim. Biophys. Acta Biomembr.
1859
,
2012
2020
[PubMed]
122.
Karandur
D.
,
Nawrotek
A.
,
Kuriyan
J.
and
Cherfils
J.
(
2017
)
Multiple interactions between an Arf/GEF complex and charged lipids determine activation kinetics on the membrane
.
Proc. Natl. Acad. Sci. U.S.A.
114
,
11416
11421
123.
Sridhar
A.
,
Kumar
A.
and
Dasmahapatra
A.K.
(
2016
)
Multi-scale molecular dynamics study of cholera pentamer binding to a GM1-phospholipid membrane
.
J. Mol. Graph. Model.
68
,
236
251
[PubMed]
124.
Pezeshkian
W.
,
Nåbo
L.J.
and
Ipsen
J.H.
(
2017
)
Cholera toxin B subunit induces local curvature on lipid bilayers
.
FEBS Open Bio
7
,
1638
1645
[PubMed]
125.
Lauer
S.
,
Goldstein
B.
,
Nolan
R.L.
and
Nolan
J.P.
(
2002
)
Analysis of cholera toxin−ganglioside interactions by flow cytometry
.
Biochemistry
41
,
1742
1751
[PubMed]
126.
Parton
D.L.
,
Tek
A.
,
Baaden
M.
and
Sansom
M.S.P.
(
2013
)
Formation of raft-like assemblies within clusters of influenza hemagglutinin observed by MD simulations
.
PLoS Comput. Biol.
9
,
e1003034
[PubMed]
127.
Sun
H.
,
Chen
L.
,
Gao
L.
and
Fang
W.
(
2015
)
Nanodomain formation of ganglioside GM1 in lipid membrane: effects of cholera toxin-mediated cross-linking
.
Langmuir
31
,
9105
9114
[PubMed]
128.
Pezeshkian
W.
,
Gao
H.
,
Arumugam
S.
,
Becken
U.
,
Bassereau
P.
,
Florent
J.-C.
et al.
(
2017
)
Mechanism of Shiga toxin clustering on membranes
.
ACS Nano
11
,
314
324
[PubMed]
129.
Monje-Galvan
V.
and
Voth
G.A.
(
2020
)
Binding mechanism of the matrix domain of HIV-1 gag on lipid membranes
.
eLife
9
,
e58621
[PubMed]
130.
Nawrocki
G.
,
Im
W.
,
Sugita
Y.
and
Feig
M.
(
2019
)
Clustering and dynamics of crowded proteins near membranes and their influence on membrane bending
.
Proc. Natl. Acad. Sci. U.S.A.
116
,
24562LP
24567LP
131.
Yu
H.
and
Schulten
K.
(
2013
)
Membrane sculpting by F-BAR domains studied by molecular dynamics simulations
.
PLoS Comput. Biol.
9
,
e1002892
[PubMed]
132.
Thekke Veettil
S.K.
,
Lal Das
S.
and
Palakurissi Balagopal
S.K.
(
2018
)
Interplay of curvature sensing and generation mediated by peripheral membrane proteins
.
Biophys. J.
114
,
614a
133.
Chavent
M.
,
Duncan
A.L.
,
Rassam
P.
,
Birkholz
O.
,
Hélie
J.
,
Reddy
T.
et al.
(
2018
)
How nanoscale protein interactions determine the mesoscale dynamic organisation of bacterial outer membrane proteins
.
Nat. Commun.
9
,
2846
[PubMed]
134.
Pezeshkian
W.
,
König
M.
,
Wassenaar
T.A.
and
Marrink
S.J.
(
2020
)
Backmapping triangulated surfaces to coarse-grained membrane models
.
Nat. Commun.
11
,
2296
[PubMed]
135.
Pezeshkian
W.
and
Marrink
S.J.
(
2021
)
Simulating realistic membrane shapes
.
Curr. Opin. Cell Biol.
71
,
103
111
[PubMed]
136.
Lomize
A.L.
,
Todd
S.C.
and
Pogozheva
I.D.
(
2022
)
Spatial arrangement of proteins in planar and curved membranes by PPM 3.0
.
Protein Sci.
31
,
209
220
[PubMed]
137.
Wei
Y.
,
Stec
B.
,
Redfield
A.G.
,
Weerapana
E.
and
Roberts
M.F.
(
2015
)
Phospholipid-binding sites of phosphatase and tensin homolog (PTEN): exploring the mechanism of phosphatidylinositol 4,5-bisphosphate activation
.
J. Biol. Chem.
290
,
1592
1606
[PubMed]
138.
Timcenko
M.
,
Lyons
J.A.
,
Januliene
D.
,
Ulstrup
J.J.
,
Dieudonné
T.
,
Montigny
C.
et al.
(
2019
)
Structure and autoregulation of a P4-ATPase lipid flippase
.
Nature
571
,
366
370
[PubMed]
139.
Stoilova-McPhie
S.
(
2017
)
Lipid nanotechnologies for structural studies of membrane-associated clotting proteins by cryo-electron microscopy
.
Nanotechnol. Rev.
6
,
127
137
140.
Hoogerheide
D.P.
,
Noskov
S.Y.
,
Jacobs
D.
,
Bergdoll
L.
,
Silin
V.
,
Worcester
D.L.
et al.
(
2017
)
Structural features and lipid binding domain of tubulin on biomimetic mitochondrial membranes
.
Proc. Natl. Acad. Sci. U.S.A.
114
,
E3622LP
E3631LP
141.
Hoogerheide
D.P.
,
Rostovtseva
T.K.
and
Bezrukov
S.M.
(
2021
)
Exploring lipid-dependent conformations of membrane-bound α-synuclein with the VDAC nanopore
.
Biochim. Biophys. Acta Biomembr.
1863
,
183643
[PubMed]
142.
Nielsen
R.D.
,
Che
K.
,
Gelb
M.H.
and
Robinson
B.H.
(
2005
)
A ruler for determining the position of proteins in membranes
.
J. Am. Chem. Soc.
127
,
6430
6442
[PubMed]
143.
Raasakka
A.
,
Ruskamo
S.
,
Kowal
J.
,
Barker
R.
,
Baumann
A.
,
Martel
A.
et al.
(
2017
)
Membrane association landscape of myelin basic protein portrays formation of the myelin major dense line
.
Sci. Rep.
7
,
4974
[PubMed]
144.
Yepuri
N.R.
,
Darwish
T.A.
,
Krause-Heuer
A.M.
,
Leung
A.E.
,
Delhom
R.
,
Wacklin
H.P.
et al.
(
2016
)
Synthesis of perdeuterated 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine and characterisation of its lipid bilayer membrane structure by neutron reflectometry
.
Chempluschem
81
,
315
321
[PubMed]
145.
Dos Santos Morais
R
,
Delalande
O
,
Pérez
J
,
Mouret
L
,
Bondon
A
,
Martel
A
et al.
(
2017
)
Contrast-matched isotropic bicelles: a versatile tool to specifically probe the solution structure of peripheral membrane proteins using SANS
.
Langmuir
33
,
6572
6580
[PubMed]
146.
Midtgaard
S.R.
,
Darwish
T.A.
,
Pedersen
M.C.
,
Huda
P.
,
Larsen
A.H.
,
Jensen
G.V.
et al.
(
2018
)
Invisible detergents for structure determination of membrane proteins by small-angle neutron scattering
.
FEBS J.
285
,
357
371
[PubMed]
147.
Maric
S.
,
Skar-Gislinge
N.
,
Midtgaard
S.
,
Thygesen
M.B.
,
Schiller
J.
,
Frielinghaus
H.
et al.
(
2014
)
Stealth carriers for low-resolution structure determination of membrane proteins in solution
.
Acta Crystallogr. D Biol. Crystallogr.
70
,
317
328
[PubMed]
148.
Corey
R.A.
,
Stansfeld
P.J.
and
Sansom
M.S.P.
(
2020
)
The energetics of protein-lipid interactions as viewed by molecular simulations
.
Biochem. Soc. Trans.
48
,
25
37
[PubMed]
149.
Yang
Y.I.
,
Shao
Q.
,
Zhang
J.
,
Yang
L.
and
Gao
Y.Q.
(
2019
)
Enhanced sampling in molecular dynamics
.
J. Chem. Phys.
151
,
70902
150.
Laio
A.
and
Parrinello
M.
(
2002
)
Escaping free-energy minima
.
Proc. Natl. Acad. Sci. U.S.A.
99
,
12562
12566
151.
Bussi
G.
and
Laio
A.
(
2020
)
Using metadynamics to explore complex free-energy landscapes
.
Nat. Rev. Phys.
2
,
200
212
152.
Barducci
A.
,
Bussi
G.
and
Parrinello
M.
(
2008
)
Well-tempered metadynamics: a smoothly converging and tunable free-energy method
.
Phys. Rev. Lett.
100
,
20603
153.
Corey
R.A.
,
Vickery
O.N.
,
Sansom
M.S.P.
and
Stansfeld
P.J.
(
2019
)
Insights into membrane protein-lipid interactions from free energy calculations
.
J. Chem. Theor. Comput.
15
,
5727
5736
[PubMed]
154.
Enkavi
G.
,
Mikkolainen
H.
,
Güngör
B.
,
Ikonen
E.
and
Vattulainen
I.
(
2017
)
Concerted regulation of npc2 binding to endosomal/lysosomal membranes by bis(monoacylglycero)phosphate and sphingomyelin
.
PLoS Comput. Biol.
13
,
e1005831
[PubMed]
155.
Hamelberg
D.
,
Mongan
J.
and
McCammon
J.A.
(
2004
)
Accelerated molecular dynamics: a promising and efficient simulation method for biomolecules
.
J. Chem. Phys.
120
,
11919
11929
[PubMed]
156.
Roy
R.
,
Ghosh
B.
and
Kar
P.
(
2020
)
Investigating conformational dynamics of Lewis Y oligosaccharides and elucidating blood group dependency of cholera using molecular dynamics
.
ACS Omega
5
,
3932
3942
[PubMed]
157.
Park
S.
and
Schulten
K.
(
2004
)
Calculating potentials of mean force from steered molecular dynamics simulations
.
J. Chem. Phys.
120
,
5946
5961
[PubMed]
158.
Lumb
C.N.
,
He
J.
,
Xue
Y.
,
Stansfeld
P.J.
,
Stahelin
R.V.
,
Kutateladze
T.G.
et al.
(
2011
)
Biophysical and computational studies of membrane penetration by the GRP1 pleckstrin homology domain
.
Structure
19
,
1338
1346
[PubMed]
159.
Costeira-Paulo
J.
,
Gault
J.
,
Popova
G.
,
Ladds
M.J.G.W.
,
van Leeuwen
I.M.M.
,
Sarr
M.
et al.
(
2018
)
Lipids shape the electron acceptor-binding site of the peripheral membrane protein dihydroorotate dehydrogenase
.
Cell Chem Biol.
25
,
309.e4
317.e4
[PubMed]
160.
Kumar
S.
,
Rosenberg
J.M.
,
Bouzida
D.
,
Swendsen
R.H.
and
Kollman
P.A.
(
1992
)
The weighted histogram analysis method for free‐energy calculations on biomolecules
.
J. Comput. Chem.
13
,
1011
1021
161.
Malhotra
K.
,
Modak
A.
,
Nangia
S.
,
Daman
T.H.
,
Gunsel
U.
,
Robinson
V.L.
et al.
(
2017
)
Cardiolipin mediates membrane and channel interactions of the mitochondrial TIM23 protein import complex receptor Tim50
.
Sci. Adv.
3
,
e1700532
[PubMed]
162.
Senju
Y.
,
Kalimeri
M.
,
Koskela
E.V.
,
Somerharju
P.
,
Zhao
H.
,
Vattulainen
I.
et al.
(
2017
)
Mechanistic principles underlying regulation of the actin cytoskeleton by phosphoinositides
.
Proc. Natl. Acad. Sci. U.S.A.
114
,
8977
8986
163.
Shirts
M.R.
and
Chodera
J.D.
(
2008
)
Statistically optimal analysis of samples from multiple equilibrium states
.
J. Chem. Phys.
129
,
124105
[PubMed]
164.
Pipatpolkai
T.
,
Corey
R.A.
,
Proks
P.
,
Ashcroft
F.M.
and
Stansfeld
P.J.
(
2020
)
Evaluating inositol phospholipid interactions with inward rectifier potassium channels and characterising their role in disease
.
Commun. Chem.
3
,
147
165.
Corey
R.A.
,
Sansom
M.S.P.
and
Stansfeld
P.J.
(
2021
)
Identification and characterization of specific protein-lipid interactions using molecular simulation
. In
Computational Design of Membrane Proteins
(
Moreira
I.S.
,
Machuqueiro
M.
and
Mourão
J.
, eds), pp.
121
139
,
NY: Springer U.S.A.
,
New York
166.
Gapsys
V.
,
Pérez-Benito
L.
,
Aldeghi
M.
,
Seeliger
D.
,
Van Vlijmen
H.
,
Tresadern
G.
et al.
(
2020
)
Large scale relative protein ligand binding affinities using non-equilibrium alchemy
.
Chem. Sci.
11
,
1140
1152
167.
Lidmar
J.
(
2012
)
Improving the efficiency of extended ensemble simulations: the accelerated weight histogram method
.
Phys. Rev. E.
85
,
56708
168.
Lindahl
V.
,
Lidmar
J.
and
Hess
B.
(
2014
)
Accelerated weight histogram method for exploring free energy landscapes
.
J. Chem. Phys.
141
,
44110
169.
Aldeghi
M.
,
Bodkin
M.J.
,
Knapp
S.
and
Biggin
P.C.
(
2017
)
Statistical analysis on the performance of molecular mechanics Poisson-Boltzmann surface area versus absolute binding free energy calculations: bromodomains as a case study
.
J. Chem. Inf. Model.
57
,
2203
2221
[PubMed]
170.
Wang
C.
,
Greene
D.
,
Xiao
L.
,
Qi
R.
and
Luo
R.
(
2018
)
Recent developments and applications of the MMPBSA method
.
Front. Mol. Biosci.
4
,
[PubMed]
171.
Rosenhouse-Dantsker
A.
,
Noskov
S.
,
Durdagi
S.
,
Logothetis
D.E.
and
Levitan
I.
(
2013
)
Identification of novel cholesterol-binding regions in Kir2 channels *
.
J. Biol. Chem.
288
,
31154
31164
[PubMed]
172.
Chirasani
V.R.
and
Senapati
S.
(
2017
)
How cholesteryl ester transfer protein can also be a potential triglyceride transporter
.
Sci. Rep.
7
,
6159
[PubMed]
173.
Genheden
S.
and
Ryde
U.
(
2015
)
The MM/PBSA and MM/GBSA methods to estimate ligand-binding affinities
.
Exp. Opin Drug Discov.
10
,
449
461
,
2015/04/02
[PubMed]
174.
Pande
V.S.
,
Beauchamp
K.
and
Bowman
G.R.
(
2010
)
Everything you wanted to know about Markov State Models but were afraid to ask
.
Methods
52
,
99
105
[PubMed]
175.
Ansell
T.B.
,
Curran
L.
,
Horrell
M.R.
,
Pipatpolkai
T.
,
Letham
S.C.
,
Song
W.
et al.
(
2021
)
Relative affinities of protein-cholesterol interactions from equilibrium molecular dynamics simulations
.
J. Chem. Theor. Comput.
17
,
6548
6558
[PubMed]
176.
He
J.
,
Haney
R.M.
,
Vora
M.
,
Verkhusha
V.V.
,
Stahelin
R.V.
and
Kutateladze
T.G.
(
2008
)
Molecular mechanism of membrane targeting by the GRP1 PH domain
.
J. Lipid Res.
49
,
1807
1815
,
[PubMed]
177.
Gaullier
J.-M.
,
Rønning
E.
,
Gillooly
D.J.
and
Stenmark
H.
(
2000
)
Interaction of the EEA1 FYVE finger with phosphatidylinositol 3-phosphate and early endosomes: role of conserved residues
.
J. Biol. Chem.
275
,
24595
24600
[PubMed]
178.
He
J.
,
Vora
M.
,
Haney
R.M.
,
Filonov
G.S.
,
Musselman
C.A.
,
Burd
C.G.
et al.
(
2009
)
Membrane insertion of the FYVE domain is modulated by pH
.
Proteins Struct. Funct. Bioinformatics
76
,
852
860
179.
Gil
J.-E.
,
Kim
E.
,
Kim
I.-S.
,
Ku
B.
,
Park
W.S.
,
Oh
B.-H.
et al.
(
2012
)
Phosphoinositides differentially regulate protrudin localization through the FYVE Domain
.
J. Biol. Chem.
287
,
41268
41276
[PubMed]
180.
Lamour
N.F.
,
Subramanian
P.
,
Wijesinghe
D.S.
,
Stahelin
R.V.
,
Bonventre
J.V.
and
Chalfant
C.E.
(
2009
)
Ceramide 1-phosphate is required for the translocation of Group IVA cytosolic phospholipase A2 and prostaglandin synthesis
.
J. Biol. Chem.
284
,
26897
26907
[PubMed]
181.
Heller
B.
,
Adu-Gyamfi
E.
,
Smith-Kinnaman
W.
,
Babbey
C.
,
Vora
M.
,
Xue
Y.
et al.
(
2010
)
Amot recognizes a juxtanuclear endocytic recycling compartment via a novel lipid binding domain
.
J. Biol. Chem.
285
,
12308
12320
[PubMed]
182.
Hekman
M.
,
Albert
S.
,
Galmiche
A.
,
Rennefahrt
U.E.E.
,
Fueller
J.
,
Fischer
A.
et al.
(
2006
)
Reversible membrane interaction of BAD requires two C-terminal lipid binding domains in conjunction with 14-3-3 protein binding
.
J. Biol. Chem.
281
,
17321
17336
[PubMed]
183.
Klarlund
J.K.
,
Tsiaras
W.
,
Holik
J.J.
,
Chawla
A.
and
Czech
M.P.
(
2000
)
Distinct polyphosphoinositide binding selectivities for pleckstrin homology domains of GRP1-like proteins based on diglycineversus triglycine motifs
.
J. Biol. Chem.
275
,
32816
32821
[PubMed]
184.
Khan
H.M.
,
Grauffel
C.
,
Broer
R.
,
MacKerell
A.D.
,
Havenith
R.W.A.
and
Reuter
N.
(
2016
)
Improving the force field description of tyrosine-choline cation−π interactions: QM investigation of phenol-N(Me)4+ interactions
.
J. Chem. Theor. Comput.
12
,
5585
5595
[PubMed]
185.
Khan
H.M.
,
MacKerell
A.D.
and
Reuter
N.
(
2019
)
Cation-π interactions between methylated ammonium groups and tryptophan in the CHARMM36 additive force field
.
J. Chem. Theor. Comput.
15
,
7
12
[PubMed]
186.
Manna
M.
,
Nieminen
T.
and
Vattulainen
I.
(
2019
)
Understanding the role of lipids in signaling through atomistic and multiscale simulations of cell membranes
.
Annu. Rev. Biophys.
48
,
421
439
[PubMed]
187.
Corradi
V.
,
Sejdiu
B.I.
,
Mesa-Galloso
H.
,
Abdizadeh
H.
,
Noskov
S.Y.
,
Marrink
S.J.
et al.
(
2019
)
Emerging diversity in lipid-protein interactions
.
Chem. Rev.
119
,
5775
5848
[PubMed]
188.
Fuglebakk
E.
and
Reuter
N.
(
2018
)
A model for hydrophobic protrusions on peripheral membrane proteins
.
PLoS Comput. Biol.
14
,
e1006325
[PubMed]
189.
Palmere
R.D.
,
Case
D.A.
and
Nieuwkoop
A.J.
(
2021
)
Simulations of Kindlin-2 PIP binding domains reveal protonation-dependent membrane binding modes
.
Biophys. J.
120
,
5504
5512
[PubMed]
190.
Mustafa
G.
,
Nandekar
P.P.
,
Mukherjee
G.
,
Bruce
N.J.
and
Wade
R.C.
(
2020
)
The effect of force-field parameters on cytochrome P450-membrane interactions: structure and dynamics
.
Sci. Rep.
10
,
7284
[PubMed]
191.
Davis
C.H.
and
Berkowitz
M.L.
(
2009
)
Interaction between amyloid-β (1-42) peptide and phospholipid bilayers: a molecular dynamics study
.
Biophys. J.
96
,
785
797
[PubMed]
192.
Orioli
S.
,
Larsen
A.H.
,
Bottaro
S.
and
Lindorff-Larsen
K.
(
2020
)
Chapter Three - How to learn from inconsistencies: Integrating molecular simulations with experimental data
. In
Computational Approaches for Understanding Dynamical Systems: Protein Folding and Assembly
(
Strodel
B.
, ed.), pp.
123
176
,
Academic Press
,
Barz BBT-P in MB and TS, editors

Author notes

*

Present address: Department of Neuroscience, Faculty of Health and Medical Sciences (SUND), University of Copenhagen

This is an open access article published by Portland Press Limited on behalf of the Biochemical Society and distributed under the Creative Commons Attribution License 4.0 (CC BY). Open access for this article was enabled by the participation of University of Oxford in an all-inclusive Read & Publish agreement with Portland Press and the Biochemical Society under a transformative agreement with JISC.