The authors have declared that no competing interests exist.
Conceived and designed the experiments: SGR AFB DNL MLK VC. Performed the experiments: SGR AFB DNL VC. Analyzed the data: SGR AFB DNL VC. Wrote the paper: SGR AFB DNL MLK VC.
Despite the clinical ubiquity of anesthesia, the molecular basis of anesthetic action is poorly understood. Amongst the many molecular targets proposed to contribute to anesthetic effects, the voltage gated sodium channels (VGSCs) should also be considered relevant, as they have been shown to be sensitive to all general anesthetics tested thus far. However, binding sites for VGSCs have not been identified. Moreover, the mechanism of inhibition is still largely unknown. The recently reported atomic structures of several members of the bacterial VGSC family offer the opportunity to shed light on the mechanism of action of anesthetics on these important ion channels. To this end, we have performed a molecular dynamics “flooding” simulation on a membrane-bound structural model of the archetypal bacterial VGSC, NaChBac in a closed pore conformation. This computation allowed us to identify binding sites and access pathways for the commonly used volatile general anesthetic, isoflurane. Three sites have been characterized with binding affinities in a physiologically relevant range. Interestingly, one of the most favorable sites is in the pore of the channel, suggesting that the binding sites of local and general anesthetics may overlap. Surprisingly, even though the activation gate of the channel is closed, and therefore the pore and the aqueous compartment at the intracellular side are disconnected, we observe binding of isoflurane in the central cavity. Several sampled association and dissociation events in the central cavity provide consistent support to the hypothesis that the “fenestrations” present in the membrane-embedded region of the channel act as the long-hypothesized hydrophobic drug access pathway.
The molecular mechanisms mediating the pharmacologically induced state of general anesthesia are, in general, poorly understood. Modulation of voltage gated sodium channels is thought to play a major role in anesthesia, as several members of this class of channels show a significant response to general anesthetics. However, the detailed mechanism of inhibition or potentiation of these channels is completely unknown. Recently, the structures of several members of the bacterial family became available, thereby offering the opportunity to shed light on some of these issues. We have performed molecular dynamics simulations on one of these bacterial voltage gated sodium channels, NaChBac, to identify binding sites and access pathways for the volatile general anesthetic isoflurane. We found that isoflurane, at physiologically relevant concentrations, binds the channel at three distinct sites. One site is in the pore of the channel, suggesting that isoflurane may hinder the permeant sodium ions. Surprisingly, we found that this binding site is accessible to the drug even when the pore and the aqueous compartment at the intracellular side are disconnected. In our simulations, the “fenestrations” present in the membrane-embedded region of the channel act as the long-hypothesized hydrophobic drug access pathway.
Voltage gated sodium channels (VGSCs), which mediate the upstroke of the action potential in most excitable tissues, are key targets of anesthetics. The binding site and molecular mechanism of action for local anesthetics have been well characterized in the last few decades, while the role of VGSCs in general anesthetic action is less well understood and both mechanisms continue to be studied. Thus far, the VGSC binding sites for general anesthetics have not been identified. Identifying binding sites and access pathways for volatile general anesthetics is key to understanding their mechanism of action and to designing new drugs.
Sodium channels can be inhibited by a number of compounds, including toxins, quaternary ammonium compounds and local anesthetics
The classical studies of charged local anesthetics and their analogs showed that blocking and unblocking seemed to require open channels. This led to the description of the “hydrophilic pathway” for drug access. However, additional experiments showed that hydrophobic local anesthetics could bind and unbind even when channels are closed
Although large mammalian VGSCs have remained resistant to structural characterization, the discovery of the smaller bacterial VGSCs has provided a tool to characterize the structural features of these important channels
(A) Structural domains of NaChBac showing the VSD (purple), the S4–S5 linkers (red) and the pore domain (blue/green) with the S6 helices highlighted in green. (B) Isoflurane flooding simulation system initial setup showing the NaChBac protein (orange) in the POPC bilayer (white) with isoflurane molecules in the aqueous phase. (C) The three binding sites identified by clustering analysis: extracellular site (red), linker site (yellow) and cavity site (purple/green).
A striking feature of the bacterial VGSC crystallized thus far is the presence of the so-called fenestrations
The available atomic structure of a bacterial VGSC offers the first opportunity to address some of the most fundamental questions about the mechanism of volatile anesthetic action on VGSCs, namely what are the likely structural determinants of general anesthetic effects on VGSCs and could the fenestrations really provide access to the central cavity for small hydrophobic drugs? To investigate these questions, we present the results of a molecular dynamics (MD) simulation study of the bacterial VGSC NaChBac embedded in a lipid bilayer in presence of the inhaled general anesthetic isoflurane. Our unbiased “flooding” technique
To identify putative isoflurane binding sites, we performed MD “flooding” simulations
Due to the protein's four-fold symmetry, a functional channel has four equivalent copies of each of the three putative binding sites. To investigate whether the three putative sites are occupied symmetrically in the tetramer, and to determine which amino acids line each site, we analyzed the interactions between isoflurane and all the residues within the sites in each subunit. In particular, we monitored the contacts between any atom of the drug molecule and any atom of a given amino acid for each instantaneous configuration after isoflurane partitioning has reached equilibrium. Residues were classified as non-interacting, “possibly” interacting, and “likely” interacting according to the number of configurations in which the contact was detected. For the extracellular site, all amino acids in all four subunits are in contact with the drug. In the cavity site as well as through the fenestrations, most residues are possibly or likely interacting in all four subunits. Intriguingly, the linker site is asymmetrically occupied in two adjacent sites.
We used the clustering data and proximity time analyses, in combination with analysis of hydrogen-bonding-like interactions and mobility of the drug molecule in the sites, to structurally characterize the sites and elucidate the key determinants of binding. Furthermore, to assess the pharmacological relevance of these sites, we estimated the binding free energy of isoflurane for each site using well-established FEP methods
The extracellular site sits at the intersubunit interface between the P-loops (
Extracellular (A and B) and linker site (C and D): residues interacting with isoflurane are highlighted in the insets B and D. For clarity, alternating subunits are shown in cyan/pink and labeled a through d. (E and F) Side and top views of isoflurane molecules occupying the cavity site and fenestrations.
The linker site is located at the “corner” formed by the N-terminus of one linker and the C-terminus of the adjacent linker (
Though the activation-gate is closed, we find that the central cavity is occupied by up to five isoflurane molecules at once (
(A) Absolute number of water molecules (red) and isoflurane molecules (blue) within the cavity as a function of time. (B) Number of cavity waters as a function of the number of bound isoflurane molecules. Note that each isoflurane molecule displaces approximately 6 waters.
Given that isoflurane molecules in the central cavity are in direct contact with water, we were surprised that diffusion translational constant is comparable to those observed in the other two binding sites.
Our homology model of NaChBac preserves the fenestrations found in the NavAb crystal structure
A and B show two views of an isoflurane molecule diffusing in (A) and out (B) of the central cavity through two adjacent fenestrations. Cavity and fenestrations are shown as grey surfaces.
Using an unbiased sampling of cavities within the bacterial channel NaChBac, we identified three putative binding sites for isoflurane. These binding sites are located at the extracellular side of the channel near the selectivity filter, at the linker between S4 and S5, and in the central cavity. Though the estimated binding free energies indicate high affinity, isoflurane molecules explore several orientations within each binding site. Due to the flexible nature of the Asn and Gln side-chains, isoflurane is able to form stable interactions with the protein at the extracellular and linker sites. Surprisingly, despite the fact that the pore is not accessible from the intracellular side, isoflurane molecules were able to reach the central cavity through the fenestrations. This observation suggests that the fenestrations could be the anesthetic hydrophobic pathways proposed more than three decades ago
All three identified sites are in regions of the protein that are predicted by mutagenesis to be critical to gating and conduction, and we hypothesize that some or all of them will play a role in inhibition of NaChBac by isoflurane. Since the time-scales involved in the molecular events relevant for gating are beyond normal MD simulation time-scales, we cannot directly probe the effect of drug interactions on protein conformation. However, based on knowledge of the mechanisms of gating, conduction, and inhibition by local anesthetics, we postulate possible mechanisms of drug action involving the binding sites suggested by our simulations. Isoflurane binding to the extracellular site positioned in the P-loops could affect the conformation of the selectivity filter, leading to inactivation through filter collapse
Though theories of anesthetic action focus primarily on anesthetic-protein interactions, the role of lipids cannot be discounted as partitioning into the bilayer changes lipid properties
This simulation of isoflurane binding sites and access pathways offers a number of experimentally testable hypotheses. While simulation shows that it is thermodynamically favorable for isoflurane to occupy these sites, it is possible that isoflurane occupation may have little or no physiological function. The functional relevance of each site can be ascertained by mutating each site individually to remove key determinants of isoflurane binding and evaluating whether NaChBac retains isoflurane sensitivity through electrophysiological assays
Though no one has previously performed molecular dynamics simulation on voltage-gated sodium channels to identify anesthetic sites, numerous simulations have been done on ligand-gated ion channels (LGICs) due to the availability of the structures of bacterial homologues
Experimental verification of the aforementioned predictions will prove crucial to deepen our understanding of sodium channels modulation by anesthetics and will likely prompt further computational investigations. Indeed, despite providing a relatively accurate and detailed description of drug-channel binding events, our computational model is characterized by several limitations: (i) simulations were performed on a homology-based model of NaChBac; (ii) only one of the metastable structural conformations of the channel was probed for drug binding; (iii) lipid dependence of drug action was not addressed. Though the high degree of sequence identity with NavAb gives us confidence on the overall architecture of NaChBac, inaccuracies in the structure of the binding sites or major reorganizations of these pockets along the activation pathway can both potentially affect our results. Increased availability of experimental structures
Simulations were initialized using the theoretical model of NaChBac in a closed conformation obtained previously
The model for the closed NaChBac channel, embedded in a fully hydrated lipid bilayer, was equilibrated by MD simulation. Specifically, the membrane is comprised of 1-palmytoyl-2-oleoyl-sn-glycero-3-phosphatidylcholine (POPC) lipid molecules. Isoflurane molecules were initially placed in the aqueous phase with random positions and orientation. The membrane protein complex contained a total of ∼120,000 atoms, including NaChBac, 434 lipid molecules, 25310 water molecules, 236 ions in solution and 145 isoflurane molecules. The resulting initial aqueous concentration of isoflurane is 300 mM; the equilibrium aqueous concentration 0.9 mM. Two Na+ ions were initially placed in the channel selectivity filter, in agreement with a previous computational study of NavAb showing double occupancy of the filter by Na+ ions
MD simulation used the CHARMM22-CMAP force field with torsional cross-terms for the protein and CHARMM27 for the phospholipids
This computational setup has several limitations. First, the time-scales achievable by our MD simulation are too short to observe channel gating. Second, the small size of the lipid bilayer in the simulation box results in an excessive drug concentration in the bilayer. The final limitation stems from the use of a oversimplified POPC lipid bilayer. Bacterial sodium channel function is strongly lipid dependent
To identify binding sites, we analyzed equilibrated configurations of the system and seeking regions characterized by high density of isoflurane. We first computed the center of mass (COM) of each isoflurane molecule in the MD trajectory frame. We then performed a cluster analysis on the resulting set of COM positions using the geometric distance between each pair of positions to build a proximity matrix. Partitioning of the set (clustering) was obtained using the Jarvis-Patrick algorithm
Free energy calculations were performed using the free energy perturbation (FEP) method. The binding free energies are calculated using the following scheme: DGbind = DGgas–prot−(DGsolv+DGrstr). Here, DGbind is the free energy of binding isoflurane to NaChBac, DGvac->prot is the free energy of transferring an isoflurane from the gas phase to the binding site, DGsolv is the isoflurane solvation free energy, and DGrstr is a measure of the entropy cost associated with the reduction in volume from a 1 M solution (V1M) to the volume available at the binding site (Vrstr), i.e., DGrstr = RT ln(Vrstr/V1M). For all reported binding energies, V1M is given by the volume associated with the flat-bottom spherical restraint applied to keep the isoflurane in the binding site during the interaction decoupling. Calculations were performed in NAMD 2.8 by varying the coupling parameter in steps of 0.025 at the ends and 0.05 in the middle.
This approach has been successfully applied to binding of anesthetics to proteins including the binding of R- and S- isoflurane enantiomers to apoferritin
(A) Left: displacement along the bilayer normal of the centers of mass of the isoflurane molecules as a function of time. Different colors are used to show the instantaneous positions of different isoflurane molecules. Right: density profile along the bilayer normal calculated for the centers of mass of all the isoflurane molecules averaged over the first 100 ns (black) and the subsequent 450 ns (red). Note that in the latter the values of the density outside the bilayer region are negligible. (B) Plot showing number of isoflurane molecules in aqueous phase over time. Inset show magnified scale for 100–550 ns, after isoflurane is fully partitioned. (C) Probability distribution function of the number of isoflurane molecules computed over the last 100 ns of simulation. The average number of isoflurane in water over this period is 0.4.
(TIFF)
Root mean square deviation (RMSD) of the backbone atoms from the initial structure employed in the MD simulation plotted as a function of time for the flooding simulation. The RMSD is shown separately for different regions of the channel: entire channel (black), voltage-sensor domains (blue), pore domain (red), and selectivity filter (green).
(TIFF)
Isoflurane flooding. The NaChBac pore domain (blue ribbons) sits in the bilayer (headgroups shown as grey spheres) while isoflurane (red/green molecules) begins in the aqueous compartment above and below the bilayer and partitions first into the bilayer and then into the protein structure. Voltage sensing domains, water molecules and lipid alkyl tails are not shown in this movie but are present in the simulation (
(MPG)
Representative isoflurane trajectory through fenestration to cavity site. Bottom view of NaChBac structure (grey ribbons) surrounded by lipid molecules (stick representations). Isoflurane (blue, space-filling representation) enters from the lipid phase into the cavity through a fenestration.
(MPG)
Proximity of isoflurane to residues within each site. The table reports the relative frequencies (as a percentage of frames) of contacts between isoflurane and the residues lining each binding site (a contact is detected if the distance between any atom of the isoflurane molecule and any atom of the given residue is smaller than 5 Å). Residues are classified as non-interacting (red), possibly interacting (orange), and likely interacting (green) depending on the probability of being engaged in a contact.
(DOCX)
Dynamics of Isoflurane in binding sites. Diffusion coefficients and rotational relaxation times of isoflurane computed, at each binding site, by averaging over a trajectory of approximately 500 ns. The water self-diffusion coefficient and the rotational relaxation time for the TIP3P model is reported for comparison.
(DOCX)
The authors thank Drs. Manuel Covarrubias and Roderic Eckenhoff for critical discussions and review of the manuscript.