Modulation of a voltage-gated Na+ channel bysevoflurane involves multiple sites anddistinct mechanismsAnnika F. Barbera, Vincenzo Carnevaleb, Michael L. Kleinb,1, Roderic G. Eckenhoffc, and Manuel Covarrubiasa

aDepartment of Neuroscience, Thomas Jefferson University, Philadelphia, PA 19107; bInstitute for Computational Molecular Science, College of Scienceand Technology, Temple University, Philadelphia, PA 19122; and cDepartment of Anesthesiology and Critical Care, Perelman School of Medicine, University ofPennsylvania, Philadelphia, PA 19104

Contributed by Michael L. Klein, March 28, 2014 (sent for review January 11, 2014; reviewed by Edward J. Bertaccini and Erik Lindahl)

Halogenated inhaled general anesthetic agents modulate voltage-gated ion channels, but the underlying molecular mechanisms arenot understood. Many general anesthetic agents regulate voltage-gated Na+ (NaV) channels, including the commonly used drug se-voflurane. Here, we investigated the putative binding sites andmolecular mechanisms of sevoflurane action on the bacterial NaVchannel NaChBac by using a combination of molecular dynamicssimulation, electrophysiology, and kinetic analysis. Structural mod-eling revealed multiple sevoflurane interaction sites possiblyassociated with NaChBac modulation. Electrophysiologically, se-voflurane favors activation and inactivation at low concentrations(0.2 mM), and additionally accelerates current decay at high con-centrations (2 mM). Explaining these observations, kinetic model-ing suggests concurrent destabilization of closed states andlow-affinity open channel block. We propose that the multipleeffects of sevoflurane on NaChBac result from simultaneous inter-actions at multiple sites with distinct affinities. This multiple-site,multiple-mode hypothesis offers a framework to study the struc-tural basis of general anesthetic action.

anesthesia | MD simulations | anesthetics | membrane proteins

General anesthetic agents have been in use for more than 160 y.However, we still understand relatively little about their

mechanisms of action, which greatly limits our ability to designsafer and more effective general anesthetic agents. Ion channelsof the central nervous system are known to be key targets ofgeneral anesthetic agents, as their modulation can account forthe endpoints and side effects of general anesthesia (1–4). Manyfamilies of ion channels are modulated by general anestheticagents, including ligand-gated, voltage-gated, and nongated ionchannels (2, 5–7). Mammalian voltage-gated Na+ (NaV) chan-nels, which mediate the upstroke of the action potential, areregulated by numerous inhaled general anesthetic agents (8–14),which generally cause inhibition. Previous work showed thatinhaled general anesthetic agents, including sevoflurane, iso-flurane, desflurane, and halothane, mediate inhibition by in-creasing the rate of Na+ channel inactivation, hyperpolarizingsteady-state inactivation, and slowing recovery from inactivation(11, 15–18). Inhibition of presynaptic NaV channels in the spinalcord is proposed to lead to inhibition of neurotransmitter release,facilitating immobilization—one of the endpoints of general an-esthesia (14, 19, 20). Despite the importance of NaV channels asgeneral anesthetic targets, little is known about interaction sitesor the mechanisms of action.What is known about anesthetic sites in NaV channels comes

primarily from the local anesthetic field. Local anesthetic agentbinding to NaV channels is well characterized. These amphi-philic drugs enter the channel pore from the intracellularside, causing open-channel block (21). Investigating molecularmechanisms of mammalian NaV channel modulation by generalanesthetic agents has been complicated by the lack of high-resolution structures of these channels as a result of their large

size and pseudotetrameric organization. However, the recent dis-covery of the smaller, tetrameric bacterial Na+ channel familyhas provided an invaluable tool to characterize the structuralfeatures of NaV channels and investigate their interactions withgeneral anesthetic agents at the molecular level (22, 23). Severalbacterial Na+ channels have been crystallized (24–27). Thesechannels have a classical domain structure in which helicesS1–S4 form the voltage sensor domain (VSD), S5 and S6 formthe pore, and the S4–S5 linker connects the voltage sensor tothe pore domain. One notable structural feature is the presenceof “fenestrations” or hydrophobic tunnels through the pore do-main (24).Although crystal structures are not yet reported, the bacterial

Na+ channel NaChBac has been extensively characterized byelectrophysiology (22, 28–36). Additionally NaChBac exhibitsconserved slow open channel block in response to local andgeneral anesthetic agents (15, 37). These anesthetic agents re-duce peak current and accelerate current decay, making it con-ceivable that local and general anesthetic agents could share asite of action in NaChBac. The local anesthetic binding siteidentified in the central cavity of the mammalian NaV1.2 chan-nel, which mediates open channel block, is partially conservedin NaChBac (37, 38). A recent molecular dynamics (MD) model-ing study found that isoflurane, which inhibits NaChBac (15),interacts with multiple regions of this channel, including thepore, the selectivity filter, and the S4–S5 linker/S6 interface (39).Although the importance of these interactions on the modula-tion of mammalian NaV channels remains to be determined,the available data indicate that NaChBac is currently one of thebest starting points to investigate the mechanisms of action ofsevoflurane.

Significance

Understanding of the mechanisms of modulation of voltage-gated Na+ (NaV) channels by general anesthetic agents such assevoflurane is critical to interpret their role in general anes-thesia. By using the bacterial NaV channel analogue NaChBacplus a combination of computational and electrophysiologicalanalyses, this work strongly suggests a multisite mechanismof sevoflurane action on NaV channels. Furthermore, computersimulations suggest specific putative anesthetic sites worthyof further investigation.

Author contributions: A.F.B., V.C., M.L.K., R.G.E., and M.C. designed research; A.F.B. andV.C. performed research; A.F.B., V.C., M.L.K., R.G.E., and M.C. analyzed data; and A.F.B., V.C.,M.L.K., R.G.E., and M.C. wrote the paper.

Reviewers: E.J.B., Stanford University School of Medicine; and E.L., Universityof Stockholm.

The authors declare no conflict of interest.1To whom correspondence should be addressed. E-mail: mlklein@temple.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1405768111/-/DCSupplemental.

6726–6731 | PNAS | May 6, 2014 | vol. 111 | no. 18 www.pnas.org/cgi/doi/10.1073/pnas.1405768111

Here, we investigated NaChBac to gain structural insight intothe mechanisms of inhaled anesthetic modulation of NaV chan-nels. The focus of this work is sevoflurane because this anestheticis commonly used in clinical settings and is a known inhibitor ofseveral mammalian NaV channels (NaV 1.4, 1.7, and 1.8) (11,13). A three-pronged approach incorporating MD simulation,whole-cell patch-clamp electrophysiology, and kinetic modelingsuggests that sevoflurane acts on multiple sites to alter gating andpermeation. Whereas the effect on gating results from modu-lating activation and inactivation gating at low concentrations(0.2 mM), the permeation effect is apparent at high concen-trations (2 mM) and results from open channel block (2 mM).Although the net inhibitory effect of these multisite interactionsis consistent with anesthetic-induced reduction of neuronal firing,general anesthesia does not simply result from a global reductionin firing. General anesthesia depends on complex mechanismsthroughout the brain, which include increases and decreasesin firing (3). Thus, precisely how Na+ channel activation bysevoflurane fits into the global effects of anesthesia remains to beseen. The present work helps elucidate the molecular mechanismof sevoflurane action on NaV channels.

Results and DiscussionMD Simulations. We used the MD flooding method (40) to ini-tially assess potential sevoflurane interactions on three structuralstates of NaChBac. The first one, the partially activated/closedstate, resembles closely the X-ray structure of NaVAb (24) andis an intermediate of activation with the pore is closed (41).Hereafter, we will use a nomenclature whereby the state of thepore domain follows that of the VSD. We have also investigatedthe resting/closed and activated/open states, which are the twoterminal states of the activation cycle (41). Worth noting is thatthe activated/open state undergoes pore closure during theflooding simulation as sevoflurane entry causes desolvation ofthe gate region. Overall, the structures were stable over thecourse of each simulation (Fig. S1). The initial model containedNaChBac, the membrane, a slab of water (25,310 molecules), and100 sevoflurane molecules initially distributed randomly in theaqueous phase (the initial aqueous concentration is ∼200 mM).This concentration is much higher than clinically relevant con-centrations, but is necessary to observe binding on the shorttime scales achievable in MD simulations. Sevoflurane thenpartitions almost completely into the lipid phase within the first200 ns of simulation (Fig. S2), and the final aqueous concen-tration of sevoflurane is ∼2 mM. To analyze putative bindingmodes and prioritize regions of the channel exhibiting continu-ous sevoflurane occupancy, we applied a protocol for clusteringdescribed elsewhere (39). This analysis identified two majorstate-dependent binding sites present in the activated conforma-tions (i.e., partially activated/closed and activated/open states):(i) a selectivity filter site and (ii) an S4–S5 linker site (Fig. 1).State-independent occupation of the central cavity and fenes-trations was found in all conformations (Figs. 1 and 2 A–C).

Additionally, the sevoflurane interaction pattern had two uniquefeatures in the closed state. First, an activation gate site notfound in other conformations and numerous but diffuse se-voflurane interactions with the VSD (Fig. 2D). NaChBac’scentral cavity was occupied by sevoflurane in all gating states,regardless of the state of the activation gate (Fig. 2 A–C). Aspreviously found for isoflurane (39), sevoflurane enters the cen-tral cavity via the fenestrations. To enter, sevoflurane displacesthe lipid tails that normally occupy the fenestrations, perhapssuggesting a basis for the well-known Meyer–Overton correla-tion. Despite being lipophilic, sevoflurane possesses a polar na-ture and can potentially interact electrostatically with its bindingpartners. In our simulations, sevoflurane is found often atcontact distance from polar moieties.All four fenestrations are traversed/occupied by sevoflurane in

each gating state. Over the equilibrated phase of trajectory for allgating states, at least two sevoflurane molecules were in thecavity; however, occasionally, three to four sevofluranes could beseen in this location (Fig. S3). Sevoflurane occupying the cavitydisplaces several water molecules, but not enough to preventNa+ solvation in the cavity. Although sevoflurane is highly mo-bile in the central cavity, it forms stable interactions in all states.Namely, it interacts with conserved residues (T220 and F227)known to play a role in local anesthetic action on mammalianNaV channels (F1764 and Y1771 in NaV1.2) (37, 38). Theobservations of consistent cavity occupation by sevoflurane andinteractions with residues known to be involved in open channelblock by local anesthetic agents supports the hypothesis thatsevoflurane may inhibit NaChBac function through an openchannel block mechanism.The sevoflurane extracellular site is located at the intersubunit

interface between the P-loops (Fig. 2 D and E), similar to apreviously identified isoflurane binding site on NaChBac (39).However, sevoflurane is found only in the extracellular site inthe partially activated/closed and activated/open conformations.Nonpolar (Y172, L178, L181, Q185 M197, F201) and charged(E171, R198) residues participate in the interactions. Se-voflurane enters the four equivalent extracellular sites quicklyduring the anesthetic partitioning phase, and remains there forthe duration of the simulations. The extracellular region ofNaChBac could influence the conformation of the selectivityfilter to regulate P/C type inactivation (28, 30). Therefore, sev-oflurane interaction at the extracellular site could alter the sta-bility of the filter region.The linker site is located at the interface between the distal S6

helix and intersubunit “corner” formed by the N terminus of onelinker and the C terminus of the linker from the adjacent subunit(Fig. 2 F and G). This linker site is found only in the partiallyactivated/closed and activated/open states. In the partially acti-vated/closed state, only one out of the four equivalent sites iscontinuously occupied by sevoflurane over the last half of thetrajectory (Fig. 2F). The activated/open state linker site is con-tinuously occupied in all four equivalent sites, sometimes bymore than one sevoflurane (Fig. 2G). In the partially activated/closed state, the linker site primarily involves nonpolar inter-actions with sevoflurane, with residue contributions from thelinker and S6 (L132, L142, I146, L225, I227, V234). Asparaginesalso participate (N145, N224), and could form polar interactionswith sevoflurane. More residues were found interacting withsevoflurane in the activated/open state, and many sevofluranemolecules occupied each site. Although we do not know whetherthis high level of occupation is physiological, it is an indicator ofa favorable interaction.An additional site at the activation gate is also found in the

resting state. Upon deactivation, this activation gate site emergesin the resting/closed state. This site involves residues of the S4–S5linker, but sevoflurane is primarily situated at the intersubunitinterface of all four S6 helices, below the hydrophobic seal of the

Fig. 1. Putative sevoflurane binding sites in various states of NaChBac: (A)Resting/closed, (B) partially activated/closed and (C) activated/open. Thethree binding sites identified by clustering analysis are extracellular site(red), linker site (yellow), activation gate site (orange), and cavity/fenestra-tion site (green/purple).

Barber et al. PNAS | May 6, 2014 | vol. 111 | no. 18 | 6727

MED

ICALSC

IENCE

S

activation gate (Fig. S4A). This activation gate site is also mainlynonpolar (L132, I230, V231, V234) and additionally contains anasparagine (N233).Mutation of residues in the lower half of S6 can affect the rate

of inactivation (29, 34), suggesting that this segment is implicatedin inactivation. Furthermore, the S4–S5/S6 interface has beenimplicated as an inhibitory general anesthetic interaction regionin K+ channels (41). Sevoflurane interactions in the S4–S5/S6interface or directly below the activation gate could thus affectinactivation and voltage-dependent activation by altering thecoupling between the VSD and the pore.

Electrophysiology. In light of the MD simulation results suggest-ing multiple sevoflurane sites in NaChBac, we anticipated mul-tiple distinguishable effects on the function of the channel. Totest this hypothesis, we investigated the gating properties ofNaChBac expressed in HEK-293 cells to establish a baseline(Fig. S5) and, subsequently, tested the effects of sevoflurane atvarious concentrations. Namely, we focused on the followingproperties: (i) conductance–voltage (Gp–V) relation, (ii) rate ofmacroscopic inactivation, (iii) prepulse inactivation, and (iv)recovery from inactivation.Given a significant cell-to-cell variability in the gating prop-

erties of NaChBac (Fig. S6), all whole-cell current recordingswere obtained in pairs from the same cell (before and aftersevoflurane). A preliminary assessment of the effects of se-voflurane on NaChBac currents revealed apparent potentia-tion at low concentrations (<0.5 mM) and inhibition at highconcentrations (>1 mM). To quantitatively characterize theseeffects, we investigated the channel’s gating properties at twosevoflurane concentrations differing by an order of magnitude,0.2 and 2 mM. Whereas 0.2 mM sevoflurane is within the ther-apeutic concentration range, 2 mM is outside this range. Wetested this concentration, however, to saturate putative relevantbinding sites. The low concentration of sevoflurane producedthe following effects on activation gating (Fig. 3, Fig. S6, and

Table S1): (i) increased the peak current by ∼15% (at 0 mV)and (ii) shifted the Gp–V curve to the left (half-activationvoltage ΔV1/2 ∼ −10 mV). Tail current kinetics were not sig-nificantly affected (Fig. S6 I and J). On inactivation gating, 0.2mM sevoflurane also had the following effects (Fig. 3 andTable S1): (i) shifted the steady-state inactivation curve to theleft (ΔV1/2 ∼ −14 mV), (ii) increased the rate constant ofcurrent decay by ∼20%, and (iii) increased the rate constant ofrecovery from inactivation by ∼36%.Generally, 2 mM sevoflurane produced qualitatively similar

results (Fig. 3, Fig. S6, and Table S1): (i) shifted the Gp–V curve tothe left (ΔV1/2 ∼ −6 mV) and slightly reduced the equivalent gatingcharge of activation, (ii) increased the rate constant of current decayby 70%, (iii) shifted the steady-state inactivation curve to the left(ΔV1/2 ∼ −19 mV), (iv) slightly reduced the equivalent gating chargeof inactivation, and (v) increased the rate constant of recovery frominactivation by ∼50%. However, instead of increasing the peakcurrent, as observed at low submillimolar concentrations, 2 mMsevoflurane slightly decreased peak current by ∼7%. Neither lownor high concentrations produced significant effects on the max-imum value of the current, Gmax (Fig. S6 C and D).The similarities in terms of the leftward shifts in the activation

and inactivation curves and accelerated recovery from inac-tivation at low and high doses of sevoflurane suggest that theanesthetic might have modulated voltage-dependent activationand the stability of the inactivated state. Because these effects donot display substantial concentration dependence between 0.2and 2 mM sevoflurane, gating modulation might be near satu-ration at submillimolar concentrations. Thus, it suggests rela-tively high-affinity interactions. Separately, however, acceleratedcurrent decay was dramatically enhanced at 2 mM sevoflurane.Such an effect suggests slow open-channel block involving a rel-atively low affinity interaction. This electrophysiological evidencefor two distinct functional effects of sevoflurane is consistentwith the simulation results suggesting multiple sevoflurane bind-ing sites, which might separately account for effects on gating

Fig. 2. Zoom views of putative sevoflurane binding sites. (A–C) Shown are top views of sevoflurane binding in the cavity and fenestrations of the resting/closed, partially activated/closed, and activated/open states, respectively. (D and E) Top view (Left) and zoom view (Right) of the extracellular site in thepartially activated/closed state (D) and the activated/open state (E). (F and G) Top view (Left) and zoom view (Right) of the linker site in the partially activated/closed state (F) and the activated/open state (G).

6728 | www.pnas.org/cgi/doi/10.1073/pnas.1405768111 Barber et al.

and permeation. A similar paradigm of distinct sites mediatingactivation and inhibition by anesthetic agents have been foundin pentameric ligand-gated ion channels (42, 43).These findings are also partly consistent with the effects of

inhaled anesthetic agents on mammalian NaV channels (10, 11,17, 18). An important difference, however, is in the recoveryfrom inactivation. Whereas sevoflurane accelerates recoveryfrom inactivation in NaChBac, inhaled anesthetic agents actingon other NaV channels typically slow recovery from inactivation(11, 15). The latter has been generally interpreted as resultingfrom anesthetic-induced stabilization of the channel’s inacti-vated state. Although leftward shifted prepulse inactivationcurves alone would be consistent with this interpretation, ac-celerated recovery and other results from our study require analternate explanation.

Kinetic Modeling. To explain the results more quantitatively andgain insights into the biophysical basis of the interactions ofsevoflurane with multiple sites, we used kinetic modeling ofNaChBac gating in the absence and presence of the anesthetic.We assumed a sequential five-state gating scheme based on thestudy by Kuzmenkin et al. (28). This model consists of fourclosed (C) states, one open (O) state, and one inactivated (I)state (Fig. 4A). In this scheme, C<>C and C<>O transitions are

governed by rate constants that strongly depend on membranepotential and the O<>I transition, which is weakly voltage-dependent (Table S1). This relatively simple scheme was ableto reproduce normal NaChBac gating and recapitulate theeffects of sevoflurane on NaChBac by making the followingminimal assumptions (Table S1): (i) accelerated forward rateconstant of voltage-dependent activation, (ii) slightly slowedforward rate constant of inactivation and accelerated back-ward rate constant of inactivation, and (iii) the presence of low-affinity open channel block. Parameter optimization yieldedchanges that semiquantitatively accounted for leftward shiftedGp–V and prepulse inactivation curves, accelerated recoveryfrom inactivation, and accelerated current decay at sub-millimolar and millimolar concentrations of sevoflurane(Table S2 and Fig. S7).The combination of MD simulations, electrophysiological in-

vestigation, and kinetic modeling strongly support a multisitemechanism of sevoflurane action. Sevoflurane may bind withhigh affinity to activation gating sites (e.g., sites involving S4–S5linker sites and the activation gate) and inactivation gate sites(e.g., extracellular sites that influence the selectivity filter).Consequently, NaChBac shifts to a gating mode exhibitingmore favorable voltage-dependent activation and inactivation,and a modestly destabilized inactivated state. Sevoflurane

Fig. 3. Electrophysiological effects of sevoflurane on NaChBac. (A and C) Representative currents evoked from a step from −100 to 0 mV, showing the effectof 0.2 mM (A) and 2 mM (C) sevoflurane on NaChBac current amplitude and rate of inactivation. (B and D) Scatter plot comparing paired measurements of thetime constants (τ) of current decay before and after application of 0.2 mM (B) and 2 mM (D) sevoflurane. (E and G) Representative peak Gp–V relationshipsfrom paired experiments before and after application of 0.2 mM (E) and 2 mM (G) sevoflurane. (F and H) Scatter plot from paired data showing the effect of0.2 mM (F) and 2 mM (H) sevoflurane on the V1/2 of activation. (I and K) Representative steady-state inactivation vs. voltage relationships from pairedexperiments before and after application of 0.2 mM (I) and 2 mM (K) sevoflurane. (J and L) Scatter plot from paired data showing the effect of 0.2 mM (J) and2 mM (L) sevoflurane on V1/2 of inactivation. (M and O) Representative recovery from inactivation curves from paired experiments before and after appli-cation of 0.2 mM (M) and 2 mM (O) sevoflurane. (N and P) Scatter plot from paired data showing the effect of 0.2 mM (N) and 2 mM (P) sevoflurane on thetime constant (τ) of recovery from inactivation.

Barber et al. PNAS | May 6, 2014 | vol. 111 | no. 18 | 6729

MED

ICALSC

IENCE

S

may also bind with low affinity to pore sites along the centralcavity. This binding is essentially responsible for the acceler-ation of current decay reflecting slow open channel block.Consistent with anesthetic action responsible for reducedneuronal firing, the combined net impact of these changesmight be inhibitory because sevoflurane shifts the prepulseinactivation curve to the left (decreasing channel availabilityat more negative membrane potentials) and induces poreblock. Further analysis by site-directed mutagenesis wouldhelp determine the precise contributions of each putativesevoflurane binding site to the effects of sevoflurane on NaChBacfunction. Also, the eventual direct structural analysis of mam-malian NaV channels would be necessary to help validate thesignificance of these binding sites in general anesthesia.

ConclusionUnderstanding of the molecular mechanisms of general anes-thetic modulation of NaV channels is critical to interpret theirrole in general anesthesia. In rat neurohypophysial nerve termi-nal preparations, isoflurane inhibits Na+ currents and dampensaction potentials (44). This inhibition reduces neurotransmit-ter release (45–48), which may play a critical role in mediatingkey physiological features of general anesthesia in vivo. In ad-dition, NaV channel modulation may explain anesthetic agent-induced immobilization (19, 20). This study paves the way tomap relevant general anesthetic binding regions in NaV channelsand helps understand how their modulation by sevofluranemight influence physiological processes implicated in generalanesthesia. Apparent activation at low concentrations of se-voflurane is particularly intriguing in light of the observationthat patients display excitation during the induction phase, be-fore experiencing the endpoints of anesthesia (49). Perhaps thisexcitatory phase is mediated in part by activation of NaV chan-nels. Our results also raise the fascinating possibility of a generalparadigm whereby modulation of ion channels by small molecules

relies on a complex interplay of several modes of action involvingdistinct binding sites. This speculative generalization is in partcorroborated by the fact that an analogous behavior is observedin Cys-loop receptors (50).

MethodsMD Simulations. Simulations were initialized by using theoretical modelsof NaChBac in the resting/closed, activated/open, and partially activated/closed conformations obtained previously (51). These homology modelswere built on the basis of the X-ray crystal structure of NaVAb (ProteinData Bank ID code 3RVY) (26). Each model, embedded in a fully hydratedbilayer of 1-palmytoyl-2-oleoyl-sn-glycero-3-phosphatidylcholine, was equili-brated by brief MD simulation runs. Sevoflurane molecules were initially placedin the aqueous phase with random positions and orientation. MD trajectorieswere collected for 0.4–0.5 μs using the CHARMM27 force field. Further detailsare given in SI Methods.

Electrophysiology. Current recordings were obtained from calcium phos-phate transfected HEK-293 cells in the whole-cell patch-clamp configu-ration. The NaChBac-GFP bicistronic plasmid was a gift from D. Ren(University of Pennsylvania, Philadelphia, PA). All experiments wereperformed at room temperature (22–25 °C). After fire polishing, patchpipette resistance in bath solution was 1.5–2 MΩ. Passive leak current andtransients were subtracted online by using a P/4 procedure. Bath solutioncontained (in mM): 140 NaCl, 4 KCl, 1.5 CaCl2, 1.5 MgCl2, 10 Hepes, and 5D-glucose, pH 7.3 adjusted with NaOH. Pipette solution contained (inmM): 15 NaCl, 80 CsF, 40 CsCl, 10 EGTA, and 10 Hepes, pH 7.3 adjustedwith CsOH. Sevoflurane containing external solution was prepared bysonication and delivered by a gastight perfusion system as previouslydescribed (52). To eliminate inaccuracies caused by membrane lipid re-tention of anesthetic molecules, each cell was exposed to only one doseof anesthetic agent and washout data were not used. Additional detailsare provided in SI Methods.

Kinetic Modeling. Kinetic modeling was performed in IonChannelLab (53)based on the NaChBac kinetic model described previously by Kuzmenkinet al. (28). The goal of kinetic modeling was to approximate the followingobservations: leftward shifts in the V1/2 values of activation and inactivation,

Fig. 4. Kinetic modeling of NaChBac modulation by sevoflurane. (A and B) Kinetic schemes in the absence (A) and presence (B) of sevoflurane. Parametervalues are summarized in Table S1. Note that the rate constants α1 and β1 are strongly voltage-dependent, whereas the rate constants α2 and β2 are onlyweakly voltage-dependent. Only the binding rate constant kon depends on the concentration of sevoflurane. (C ) Simulated NaChBac currents in theabsence (black) and presence of 0.2 mM (blue) and 2 mM (red). The depolarizing step is from −100 to 0 mV. (D) Comparison of observed vs. simulatedtime constants of current decay. Observed data are shown with hatch marks: control (gray), 0.2 mM (blue), and 2 mM (red). (E ) Simulated Gp–V relationsbefore (black) and after application of 0.2 mM (blue) and 2 mM (red) sevoflurane. (F ) Comparison of observed vs. simulated V1/2 shifts estimated from thecorresponding Gp–V relations. Color scheme is as in C. (G) Simulated steady-state inactivation vs. voltage relations before (black) and after application of0.2 mM (blue) and 2 mM (red) sevoflurane. (H) Comparison of observed vs. simulated V1/2 shifts estimated from the corresponding steady-state in-activation curves. Color scheme is as in B. (I) Simulated trajectories of recovery from inactivation before (black) and after application of 0.2 mM (blue) and2 mM (red) sevoflurane. (J) Comparison of observed vs. simulated time constants of recovery from inactivation estimated from the correspondingtrajectories of recovery.

6730 | www.pnas.org/cgi/doi/10.1073/pnas.1405768111 Barber et al.

increased rate of recovery from inactivation, and increased rate of in-activation at high anesthetic agent concentrations. Individual rate con-stants were adjusted manually, and all observed gating properties wereevaluated. This process was repeated until best fits to all observationswere achieved. Initially, the model parameters were adjusted to ap-proximate the control data in the absence of sevoflurane, and subsequentlyrefined to create a model of sevoflurane modulation at 0.2 mM and 2 mMsevoflurane.

ACKNOWLEDGMENTS. This work was supported in part by NationalInstitutes of Health Grant National Institute of General Medical Sciences(NIGMS) P01 55876 (to R.G.E., M.C., and M.L.K.), National Institute of Neu-rological Disorders and Stroke (NINDS) Grant F31 077689 (to A.F.B.), and theCommonwealth of Pennsylvania. The computations were performed usingresources from Extreme Science and Engineering Discovery Environment(XSEDE, www.xsede.org/high-performance-computing) Grant MCA93S020(to M.L.K.) and the Temple University High-Performance Computing Systempurchased in part with NSF Grant MRI-R2 0958854 (to M.L.K.).

1. Hemmings HC, Jr., et al. (2005) Emerging molecular mechanisms of general anestheticaction. Trends Pharmacol Sci 26(10):503–510.

2. Franks NP (2006) Molecular targets underlying general anaesthesia. Br J Pharmacol147(Suppl 1):S72–S81.

3. Franks NP (2008) General anaesthesia: from molecular targets to neuronal pathwaysof sleep and arousal. Nat Rev Neurosci 9(5):370–386.

4. Urban BW (2008) The site of anesthetic action. Handbook Exp Pharmacol 183(182):3–29.

5. Krasowski MD, Harrison NL (1999) General anaesthetic actions on ligand-gated ionchannels. Cell Mol Life Sci 55(10):1278–1303.

6. Patel AJ, et al. (1999) Inhalational anesthetics activate two-pore-domain backgroundK+ channels. Nat Neurosci 2(5):422–426.

7. Cacheaux LP, et al. (2005) Impairment of hyperpolarization-activated, cyclic nucleotide-gated channel function by the intravenous general anesthetic propofol. J PharmacolExp Ther 315(2):517–525.

8. Ratnakumari L, Hemmings HC, Jr. (1998) Inhibition of presynaptic sodium channels byhalothane. Anesthesiology 88(4):1043–1054.

9. OuyangW, Wang G, Hemmings HC, Jr. (2003) Isoflurane and propofol inhibit voltage-gated sodium channels in isolated rat neurohypophysial nerve terminals. Mol Phar-macol 64(2):373–381.

10. OuYang W, Hemmings HC, Jr. (2007) Isoform-selective effects of isoflurane on volt-age-gated Na+ channels. Anesthesiology 107(1):91–98.

11. Ouyang W, Herold KF, Hemmings HC, Jr. (2009) Comparative effects of halogenatedinhaled anesthetics on voltage-gated Na+ channel function. Anesthesiology 110(3):582–590.

12. Herold KF, Nau C, Ouyang W, Hemmings HC, Jr. (2009) Isoflurane inhibits the tetro-dotoxin-resistant voltage-gated sodium channel Nav1.8. Anesthesiology 111(3):591–599.

13. Yokoyama T, et al. (2011) Effects of sevoflurane on voltage-gated sodium channelNa(v)1.8, Na(v)1.7, and Na(v)1.4 expressed in Xenopus oocytes. J Anesth 25(4):609–613.

14. Herold KF, Hemmings HC, Jr. (2012) Sodium channels as targets for volatile anes-thetics. Front Pharmacol 3:50.

15. OuyangW, Jih T-Y, Zhang T-T, Correa AM, Hemmings HC, Jr. (2007) Isoflurane inhibitsNaChBac, a prokaryotic voltage-gated sodium channel. J Pharmacol Exp Ther 322(3):1076–1083.

16. Horishita T, Eger EI, 2nd, Harris RA (2008) The effects of volatile aromatic anestheticson voltage-gated Na+ channels expressed in Xenopus oocytes. Anesth Analg 107(5):1579–1586.

17. Bean BP, Shrager P, Goldstein DA (1981) Modification of sodium and potassiumchannel gating kinetics by ether and halothane. J Gen Physiol 77(3):233–253.

18. Rehberg B, Xiao YH, Duch DS (1996) Central nervous system sodium channels aresignificantly suppressed at clinical concentrations of volatile anesthetics. Anesthesi-ology 84(5):1223–1233.

19. Zhang Y, et al. (2008) Intrathecal veratridine administration increases minimum al-veolar concentration in rats. Anesth Analg 107(3):875–878.

20. Zhang Y, et al. (2010) Bidirectional modulation of isoflurane potency by intrathecaltetrodotoxin and veratridine in rats. Br J Pharmacol 159(4):872–878.

21. Hille B (2001) Ion Channels of Excitable Membranes (Sinauer, Sunderland, MA).22. Ren D, et al. (2001) A prokaryotic voltage-gated sodium channel. Science 294(5550):

2372–2375.23. Koishi R, et al. (2004) A superfamily of voltage-gated sodium channels in bacteria.

J Biol Chem 279(10):9532–9538.24. Payandeh J, Scheuer T, Zheng N, Catterall WA (2011) The crystal structure of a volt-

age-gated sodium channel. Nature 475(7356):353–358.25. Zhang X, et al. (2012) Crystal structure of an orthologue of the NaChBac voltage-

gated sodium channel. Nature 486(7401):130–134.26. Payandeh J, Gamal El-Din TM, Scheuer T, Zheng N, Catterall WA (2012) Crystal

structure of a voltage-gated sodium channel in two potentially inactivated states.Nature 486(7401):135–139.

27. McCusker EC, et al. (2012) Structure of a bacterial voltage-gated sodium channel porereveals mechanisms of opening and closing. Nat Commun 3:1102.

28. Kuzmenkin A, Bezanilla F, Correa AM (2004) Gating of the bacterial sodium channel,NaChBac: Voltage-dependent charge movement and gating currents. J Gen Physiol124(4):349–356.

29. Zhao Y, Scheuer T, Catterall WA (2004) Reversed voltage-dependent gating of abacterial sodium channel with proline substitutions in the S6 transmembrane seg-ment. Proc Natl Acad Sci USA 101(51):17873–17878.

30. Pavlov E, et al. (2005) The pore, not cytoplasmic domains, underlies inactivation ina prokaryotic sodium channel. Biophys J 89(1):232–242.

31. Blanchet J, Pilote S, Chahine M (2007) Acidic residues on the voltage-sensor do-main determine the activation of the NaChBac sodium channel. Biophys J 92(10):3513–3523.

32. DeCaen PG, Yarov-Yarovoy V, Zhao Y, Scheuer T, Catterall WA (2008) Disulfidelocking a sodium channel voltage sensor reveals ion pair formation during activation.Proc Natl Acad Sci USA 105(39):15142–15147.

33. Paldi T, Gurevitz M (2010) Coupling between residues on S4 and S1 defines thevoltage-sensor resting conformation in NaChBac. Biophys J 99(2):456–463.

34. Irie K, et al. (2010) Comparative study of the gating motif and C-type inactivation inprokaryotic voltage-gated sodium channels. J Biol Chem 285(6):3685–3694.

35. Yarov-Yarovoy V, et al. (2012) Structural basis for gating charge movement in thevoltage sensor of a sodium channel. Proc Natl Acad Sci USA 109(2):E93–E102.

36. DeCaen PG, Yarov-Yarovoy V, Scheuer T, Catterall WA (2011) Gating charge inter-actions with the S1 segment during activation of a Na+ channel voltage sensor. ProcNatl Acad Sci USA 108(46):18825–18830.

37. Lee S, Goodchild SJ, Ahern CA (2012) Local anesthetic inhibition of a bacterial sodiumchannel. J Gen Physiol 139(6):507–516.

38. Ragsdale DS, McPhee JC, Scheuer T, Catterall WA (1994) Molecular determinants ofstate-dependent block of Na+ channels by local anesthetics. Science 265(5179):1724–1728.

39. Raju SG, Barber AF, LeBard DN, Klein ML, Carnevale V (2013) Exploring volatilegeneral anesthetic binding to a closed membrane-bound bacterial voltage-gatedsodium channel via computation. PLOS Comput Biol 9(6):e1003090.

40. Vemparala S, Domene C, Klein ML (2008) Interaction of anesthetics with open andclosed conformations of a potassium channel studied via molecular dynamics andnormal mode analysis. Biophys J 94(11):4260–4269.

41. Barber AF, Liang Q, Amaral C, Treptow W, Covarrubias M (2011) Molecular mappingof general anesthetic sites in a voltage-gated ion channel. Biophys J 101(7):1613–1622.

42. Sauguet L, et al. (2013) Structural basis for potentiation by alcohols and anaestheticsin a ligand-gated ion channel. Nat Commun 4:1697.

43. Brömstrup T, Howard RJ, Trudell JR, Harris RA, Lindahl E (2013) Inhibition versuspotentiation of ligand-gated ion channels can be altered by a single mutation thatmoves ligands between intra- and intersubunit sites. Structure 21(8):1307–1316.

44. Ouyang W, Hemmings HC, Jr. (2005) Depression by isoflurane of the action potentialand underlying voltage-gated ion currents in isolated rat neurohypophysial nerveterminals. J Pharmacol Exp Ther 312(2):801–808.

45. Westphalen RI, Hemmings HC, Jr. (2006) Volatile anesthetic effects on glutamateversus GABA release from isolated rat cortical nerve terminals: 4-aminopyridine-evoked release. J Pharmacol Exp Ther 316(1):216–223.

46. Westphalen RI, Yu J, Krivitski M, Jih T-Y, Hemmings HC, Jr. (2010) Regional differencesin nerve terminal Na+ channel subtype expression and Na+ channel-dependent glu-tamate and GABA release in rat CNS. J Neurochem 113(6):1611–1620.

47. Westphalen RI, Kwak N-B, Daniels K, Hemmings HC, Jr. (2011) Regional differences inthe effects of isoflurane on neurotransmitter release. Neuropharmacology 61(4):699–706.

48. Westphalen RI, Desai KM, Hemmings HC, Jr. (2013) Presynaptic inhibition of the re-lease of multiple major central nervous system neurotransmitter types by the inhaledanaesthetic isoflurane. Br J Anaesth 110(4):592–599.

49. Vlajkovic GP, Sindjelic RP (2007) Emergence delirium in children: Many questions, fewanswers. Anesth Analg 104(1):84–91.

50. Brannigan G, LeBard DN, Hénin J, Eckenhoff RG, Klein ML (2010) Multiple bindingsites for the general anesthetic isoflurane identified in the nicotinic acetylcholinereceptor transmembrane domain. Proc Natl Acad Sci USA 107(32):14122–14127.

51. Barber AF, et al. (2012) Hinge-bending motions in the pore domain of a bacterialvoltage-gated sodium channel. Biochim Biophys Acta 1818(9):2120–2125.

52. Barber AF, Liang Q, Covarrubias M (2012) Novel activation of voltage-gated K(+)channels by sevoflurane. J Biol Chem 287(48):40425–40432.

53. Santiago-Castillo JAD, Covarrubias M, Sánchez-Rodríguez JE, Perez-Cornejo P,Arreola J (2010) Simulating complex ion channel kinetics with IonChannelLab.Channels (Austin) 4(5):422–428.

Barber et al. PNAS | May 6, 2014 | vol. 111 | no. 18 | 6731

MED

ICALSC

IENCE

S