Resurgent current in context: Insights from the structure and function of Na and K channels

Abstract

Discovered just over 25 years ago in cerebellar Purkinje neurons, resurgent Na current was originally described operationally as a component of voltage-gated Na current that flows upon repolarization from relatively depolarized potentials and speeds recovery from inactivation, increasing excitability. Its presence in many excitable cells and absence from others has raised questions regarding its biophysical and molecular mechanisms. Early studies proposed that Na channels capable of generating resurgent current are subject to a rapid open-channel block by an endogenous blocking protein, which binds upon depolarization and unblocks upon repolarization. Since the time that this mechanism was suggested, many physiological and structural studies of both Na and K channels have revealed aspects of gating and conformational states that provide insights into resurgent current. These include descriptions of domain movements for activation and inactivation, solution of cryo-EM structures with pore-blocking compounds, and identification of native blocking domains, proteins, and modulatory subunits. Such results not only allow the open-channel block hypothesis to be refined but also link it more clearly to research that preceded it. This review considers possible mechanisms for resurgent Na current in the context of earlier and later studies of ion channels and suggests a framework for future research.

Introduction

The characteristic of rapid inactivation is among the most distinctive features of the voltage-gated Na currents that underlie the action potential upstroke in nearly all excitable cells. In the earliest voltage-clamp study, Hodgkin and Huxley (1) recognized the “dual effect” of voltage on the Na conductance: Depolarization leads to the activation of Na current as well as to its inactivation. In modern terminology, at hyperpolarized membrane potentials, voltage-gated Na channel proteins are in a non-conducting configuration, and step depolarizations induce conformational changes in each channel protein, which can open an ion-selective pore that conducts Na ions. Within milliseconds, conducting channels change conformation again into another non-conducting state. The open state, O, is thus a short-lived state between the more stable non-conducting states favored at more hyperpolarized voltages (closed states, C) and the non-conducting states favored at more depolarized voltages (inactivated states, I). When the voltage returns to more hyperpolarized values, open channels close rapidly, through the process called deactivation, whereas inactivated channels undergo a slower conversion back to closed states, through the process called recovery. Both the voltage dependence and time course of recovery from inactivation help set the refractory period for action potential firing.

The density and complement of all the ion channels in any cell contribute to shaping the interspike interval as well as the action potential waveform. In this respect, the many isoforms of K and Ca channels, with their widely ranging voltage dependences and kinetics, are well known to affect the distinct spike waveforms and firing patterns of different neurons and muscles. In contrast, nearly all voltage-gated Na channels share the basic attribute of rapid activation and fast inactivation, making these channels initially seem unlikely contributors to the diversity of excitability.

Just over a quarter of a century ago, however, an unusual component of Na current was identified in cerebellar Purkinje neurons (2). In these cells, depolarization evokes a brief or "transient" Na current that activates and decays rapidly and almost fully, in the usual manner (Fig. 1 A). Upon repolarization, however, Na current flows again, correlating with a rapid recovery from inactivation at moderately negative potentials (Fig. 1, B and C). The repolarization-evoked current cannot be attributed to the small equilibrium occupancy of open states responsible for steady-state or "persistent" Na current, since it is dynamically gating, with rise and decay times about 10-fold slower than transient current (4 ms and 20 ms, respectively at −30 mV). This surge of current, evoked upon repolarization even after maximal inactivation, was therefore dubbed "resurgent" Na current. (Hereafter, “transient,” “persistent,” and “resurgent” refer to current carried by Na ions unless otherwise specified.)

Figure 1.

Resurgent Na current in Purkinje cells. (A) Activation and inactivation of TTX-sensitive transient Na currents of Purkinje cells dissociated from rat pups. Left: currents evoked by 50-ms step depolarizations from −90 mV to potentials ranging from −80 to +50 mV (10-mV steps). Right: conductance-voltage curve for activation and availability curve for steady-state inactivation after 100-ms conditioning steps ((2), with permission; copyright 1997, Society for Neuroscience). (B) The first published recordings of resurgent current of Purkinje cells. Left: TTX-sensitive Na currents evoked by repolarizations from −60 to −20 mV (10-mV steps) after a 10-ms depolarization to +30 mV. Right: current-voltage relation for peak resurgent current ((2), with permission; copyright 1997, Society for Neuroscience). (C) Experimentally recorded resurgent current from a Purkinje cell from an adult mouse ((126), with permission) with superimposed traces simulated by a Markov model with resurgent current generated by permeation-dependent open-channel block and unblock ((84), with permission). (D) Resurgent current evoked by repolarization to −30 mV ((8), with permission), with C, O, X+, and X− states labeled.

Resurgent current, which initially appeared unique to Purkinje cells, has since been identified in more than 20 different cell types (reviewed in (3)). Many of these are in the cerebellum, brainstem, and basal ganglia, or in other neurons typified by rapid action potential firing or bursting, with short refractory periods. Additionally, changes in resurgent Na current in several cell types have been predicted to occur in disorders of excitability, including paroxysmal extreme pain disorder, epilepsy, paramyotonia congenita, long-QT syndrome, and neuropathy (4,5,6,7).

A primary question, relevant to understanding both Na channel gating and the resultant action potentials, relates to identifying the biophysical and molecular mechanisms of resurgent current. In vertebrate central neurons, resurgent current is fully blocked by tetrodotoxin (TTX), as are transient and persistent currents (2). All three components of current are thus unambiguously carried by TTX-sensitive voltage-gated Na channels, which have the capacity to inactivate. By analogy with the action of exogenous molecules whose interactions with ion channel pores are well described (discussed below), Raman and Bean (8) proposed that Na channels that generate resurgent current are subject to a rapid, voltage-dependent, open-channel block by an endogenous blocking particle. According to this hypothesis, depolarization would open Na channels, which would then rapidly become blocked by the native blocker instead of inactivating in the usual manner. Upon repolarization, the blocker would unbind, briefly leaving the pore open to pass resurgent current before channels inactivated normally (at moderately negative potentials) or deactivated (at more negative potentials).

Since the time of that proposal, many electrophysiological and structural discoveries about Na and K channels have been made, and studies of resurgent current have revealed links to classical physiological experiments. The purpose of this review, therefore, is to place the idea of open-channel block as a mechanism for resurgent Na current (with a focus on Purkinje neurons) in the context of earlier and later studies of ion channel biology. This topic is appropriate for the present issue, dedicated to Rick Aldrich, given his foundational studies of inactivation mechanisms of voltage-gated K channels (9,10), which have many parallels to resurgent current.

Electrophysiological and structural correlates of non-conducting states

A confounding factor in understanding the gating of Na channels that produce resurgent current is that electrophysiological measurements of voltage-clamped ion channels indicate only the presence or absence of current, and thus only conducting states (O) and non-conducting states (here denoted X) can be immediately distinguished. Since ion channel proteins can assume multiple non-conducting conformations, however, the challenge becomes to design protocols and read current traces to make inferences about non-conducting states. The characteristics of resurgent current hint at multiple non-conducting states, which can be operationally defined based on voltage protocols that elicit the current. A step depolarization (e.g., to +30 mV) leads channels to equilibrate into a non-conducting state that can be denoted X+, to indicate that it is favored at positive potentials; a step repolarization immediately afterward (e.g., to −30 mV) allows channels to reopen into state O before equilibrating into another non-conducting state that can be denoted X−, to indicate that it is favored at (moderately) negative potentials. Resurgent current therefore flows as channels make the transitions X+ → O → X− (Fig. 1 D). The puzzle becomes to define X+ and X−.

Since the earliest study of Na current, a fundamental distinction has been made between non-conducting states from which opening can or cannot occur with further depolarization (1), now often termed available (closed) or unavailable (inactivated), respectively. After identifying at least two such states, Hodgkin and Huxley (11) used the non-exponential rise and exponential decay of Na current evoked by step depolarizations to infer the existence of three independent voltage-sensitive gating particles for activation, suggesting that the membrane could have one, two, or three particles in the resting position, each of which would represent a non-conducting but available state. A fourth independent particle was proposed to control inactivation.

While the question of strict independence of the activation and inactivation mechanisms persisted for some time (12,13,14,15), later studies gave structural meaning to these inferences: The voltage-gated Na channel was ultimately found to have four domains, DI, DII, DIII, and DIV, each with a voltage-sensing domain (VSD) including the charged transmembrane segment, S4, which could move outward to gate the channel (16,17,18). The inactivation “gate” was characterized as a proteinaceous intracellular particle that was susceptible to proteolytic cleavage (19). Consistent with these data, three amino acids (“IFM”) on the linker between DIII and DIV were shown to play a critical role in fast inactivation, the form of inactivation that terminates current flow within milliseconds (18,20,21,22,23,24).

Domain movement was meaningfully related to state by Chanda and Bezanilla (25), who monitored the position of fluorescently labeled VSDs upon depolarization. They demonstrated that the VSD₁, VSD_II, and VSD_III move outward before Na current flows, whereas VSD_IV moves after current flow (Fig. 2 A). Moreover, VSD_IV movement is rate limiting for inactivation of electrophysiologically measured currents (26) (Fig. 2 B). The movement of VSD_IV thus seems likely to expose the binding site for the DIII-DIV linker. This scenario has since been supported by cryogenic electron microscopy (cryo-EM) studies, which further reveal that, although the linker does not bind in the permeation pathway (27,28,29), its binding causes structural changes near the intracellular gate that limit the pore diameter (30). Recent electrophysiological studies of mutant channels provide evidence that the fast-inactivation process actually involves three components: 1) depolarization-dependent outward movement of VSD_IV, which 2) permits the DIII-DIV linker to bind to the channel, which 3) allosterically shifts leucine and isoleucine residues on DIII and DIV into the permeation pathway near the intracellular mouth of the channel. This shift creates a hydrophobic barrier to ionic flux, which is the physical manifestation of fast inactivation (31).

Figure 2.

Relation of VSD_IV to activation, inactivation, and recovery. (A) VSD_IV deployment is unnecessary for activation. Fluorescence signals with outward movement of VSD_IV superimposed on outward ionic current through Na_V1.4 channels evoked by depolarization. Arrows emphasize that current activation precedes outward movement of VSD_IV ((25), with permission). (B) VSD_IV deployment is sufficient for inactivation. Steady-state inactivation assayed by transient Na currents evoked by depolarizations after conditioning at different voltages in wild-type Na_V1.4 and channels mutated (CN) to stabilize each VSD (DI, DII, DIII, DIV) in the outward position. Top: sample traces. Bottom: steady-state inactivation curves (mean ± SEM), indicating that fast inactivation is unaffected by outward VSDs I, II, or III, but shifted to negative (arrow) with outward movement of VSD_IV ((26), with permission). (C and D) Recovery current does not flow upon repolarization. The complete decay of Na current (arrows) upon repolarization to −80 mV after a 1-ms or 15-ms depolarization to 0 mV in a squid axon (C) ((35), with permission), and upon repolarization to the voltages indicated, in rat CA1 hippocampal neurons (D) ((36), with permission).

Together, the body of electrophysiological, imaging, and structural studies largely agree that, at least to a first approximation, activation and inactivation can show coupling but not contingency: certain transitions follow others because of slower kinetics, but inactivation does not require prior channel opening, nor does any activation domain movement depend upon another domain moving first. In other words, depolarization favors voltage-gated Na channels to undergo the transitions C₀ → C₁ → C₂ → C₃ → O, where the subscripts on C denote the number of outwardly shifted VSDs of DI, DII, and DIII, and the final transition to an open state involves the opening of an intracellular gate but no VSD movements. From any C_n state (for simplicity, denoted C), however, channels can undergo a transition to a fast-inactivated I_n state (denoted I) upon movement of VSD_IV followed by linker binding and pore constriction. Weak depolarizations are relatively permissive for direct C → I transitions (likely via C₀, C₁, and C₂), while strong depolarization favors C → O → I transitions (via C₃).

Resurgent current and fast inactivation

The question underlying a mechanistic understanding of resurgent current is, do channels that have inactivated at positive voltages remain unavailable upon repolarization to negative voltages until the recovery process takes place, such that the primary recovery route is I → C? Or, alternatively, do inactivated Na channels pass from I → O → C with high probability? If the latter is true, resurgent current might be the consequence of fast-inactivated states recovering through open states (32). Stated structurally, if VSD_IV were to return to its resting position (with the associated linker unbinding and reversal of pore constriction) before VSD_I, VSD_II, and VSD_III did so, channels would indeed reopen upon repolarization. The resulting “recovery current” would have a rising phase that tracked the time course of VSD_IV inward movement and a falling phase related to the return to rest by VSD_I, VSD_II, VSD_III—whichever deactivated the fastest.

In Na channels, such recovery currents can be chemically induced by Cn2 toxin (33), which binds to the β-scorpion toxin binding site on the extracellular portion of DII, stabilizing VSD_II in the outward position (34). Cn2-modified Na_V1.6 channels open at unusually hyperpolarized potentials, with half-activation voltages about 30 mV more negative than for unmodified channels. Conversely, toxin-modified open channels fail to deactivate rapidly, even at hyperpolarized potentials at which VSD_IV returns to its inward, resting position (−50 to −80 mV). As a result, upon repolarization, recovery from inactivation (the return of VSD_IV to rest) precedes channel closure (the return of any other VSD to rest). The toxin therefore induces a recovery current that resembles a very slow resurgent-like current, with a rising phase of about 40 ms and an even slower decay (33).

Whether such current ever occurs naturally in native Na channels, in the absence of toxin modifications, was tested many years previously. Recording from squid axons, Armstrong and Croop (35) directly examined whether Na current flowed during the recovery interval between voltage steps that induced inactivation. They found no evidence of such recovery current (Fig. 2 C). Identical observations were made in mouse CA1 hippocampal neurons (36) (Fig. 2 D). These results, supported by simulations, indicated that fast-inactivated Na channels do not reopen as they recover. Instead, they reveal the existence of a closed-and-inactivated refractory state that can exist at negative potentials. In structural terms, VSD_IV is outward in such a state, with linker bound and pore constricted, but one or more of the three activation VSDs has returned to rest. A corollary of these results is that the hydrophobic pore constriction that is the ultimate mechanism of fast inactivation is relieved only after channels deactivate. The absence of recovery current indicates that, in TTX-sensitive neuronal Na channels under physiological conditions, VSD_IV returns to rest only after other VSDs do so.

Although most Na channels do not recover from fast inactivation through open states, it nevertheless remains possible that resurgent current arises from channels that are specifically modified to do so. It can therefore be helpful to assume that X+, the non-conducting state entered by Purkinje Na channels at positive potentials, is indeed the fast-inactivated configuration as in most Na channels, and proceed until this proposition is logically untenable. In the simplest scenario of this form, the X+ → O → X− transition that makes resurgent current would correspond to I → O → C, meaning that at voltages where resurgent current flows (more negative than approximately −10 mV), fast-inactivated channels would recover, reopen, and then deactivate.

The evidence provides an argument against this possibility, however, at least in Purkinje neurons. In those cells, the Na channel half-activation voltage is −35 mV and the slope factor is 5 mV (37). Consequently, the Na conductance is near maximal at −10 mV, such that <1% of channels are predicted to deactivate at −10 mV. An absorbing transition into closed states (O → C) is therefore unlikely at this voltage. Even assuming that the tiny amount of deactivation could account for resurgent current, a second line of evidence makes the idea still more improbable. Steady-state inactivation curves demonstrate that depolarizations from holding voltages more positive than about −30 mV fail to open more channels, indicating that the channels enter unavailable states within tens of milliseconds (8). Thus, the state X− into which channels equilibrate upon mild repolarizations, after resurgent current flows, cannot be C, an available state. X− must instead be an inactivated state.

Resurgent current and slow inactivation

One possibility is that X− is a state associated with slow inactivation, denoted S. Electrophysiologically, slow inactivation, which is favored by prolonged depolarization, is a non-conducting state that is not only entered slowly but also recovered from slowly (38). Such a slowly developing, slowly recovering inactivated state is also present in several voltage-gated K channels, some of which also show fast inactivation (39,40,41,42). In both Na and K channels, slow inactivation has been identified as a structurally distinct non-conducting state, since it is unaffected by disruption of the fast-inactivation gate by either intracellular proteolysis or mutation (9,38,43,44). Its onset, however, is antagonized by the occupancy of the external pore by extracellular cations (39,45,46,47). Structurally, slow inactivation seems to result from physical rearrangements of the pore that occur during prolonged periods with limited ionic flux, despite VSDs in the activated position (48,49). Since such conditions arise primarily when channels occupy the fast-inactivated state (45), channels apparently can be both fast- and slow-inactivated at once: like activation and inactivation, however, neither state is contingent upon the other.

These data suggest that both Na and inactivating K channels normally undergo slow inactivation from fast-inactivated states rather than from open states after recovery. It is nevertheless fair to consider the formal possibility that Na channels in Purkinje cells present a special case, and that the X+ → O → X− transition responsible for resurgent current at −30 mV might be I → O → S. If so, then the 20-ms time course of resurgent current decay at −30 mV would represent the onset of slow inactivation. The O → S transition does not proceed so quickly, however: even 50-ms steps to 0 mV do not yield detectable slow inactivation in Purkinje cells (50). Moreover, the non-conducting states entered within tens of milliseconds at −30 mV recover completely in 100 ms, a time course consistent with fast but not slow inactivation. Finally, slow inactivation is greater at 0 mV than at −30 mV, yet resurgent current is evoked with steps from 0 to −30 mV, invalidating the assumption that repolarizing steps would promote equilibration into S states (50). Therefore, X− cannot be S. Instead, in the voltage range of −10 to −40 mV, channels are absorbed into states with the characteristics of classical fast inactivation: an unavailable state favored at moderately negative potentials, which recovers, without reopening, in tens of milliseconds at more hyperpolarized voltages. Therefore, X− is likely to be I.

This conclusion forces a reconsideration of the supposition that X+ is I too. It is kinetically improbable for a voltage change to induce first a net exit from and then a re-entry back into the same voltage-dependent state; hence, X+ must be a state other than a fast-inactivated state. More broadly, X+ is a non-conducting, unavailable state that recovers through an open state. It cannot be a slow-inactivated state because it is entered on time scales too brief to induce slow inactivation (50). It is also apparently voltage dependent: it persists for tens to hundreds of milliseconds at positive voltages but is unstable enough at negative voltages to account for the rapid rise of resurgent current (X+ → O), which is ≈4 ms at −30 mV. It therefore is necessary to consider other possible conformations for X+.

Open-channel block as a distinct non-conducting state

Classic studies of voltage-gated K channels in squid axons raise the possibility that X+ might be an open-channel blocked state, here denoted B. Armstrong (51) discovered that intracellularly applied quaternary ammonium ions (QAs)—primarily derivatives of tetraethyl ammonium ion (TEA⁺) with one extended alkyl chain—can induce an “inactivation” phase of K currents that normally do not decay during sustained depolarization. He deduced that QAs bind in the internal mouth of the open permeation pathway and obstruct current flow. Evidence for this pore-blocking mechanism includes the quantitative observations 1) that the extent of inactivation is proportional to the extent of activation, suggesting that the transition to a QA-induced non-conducting state cannot proceed until channels have opened, 2) that the rate of inactivation is concentration dependent, suggesting a true binding event between QA ions and the pore, and 3) that recovery from the QA-induced non-conducting state is greatly facilitated by inward flux of K ions, suggesting that permeating ions electrostatically displace QA ions from their blocking sites. In fact, with high external K concentrations that generate inward K flux at negative potentials, repolarization evokes a current that rises in a few milliseconds, much like resurgent Na current (Fig. 3 A), rather than instantaneously, like normal tail currents (52). Simulations lent support to the conclusion that the delayed-onset “hooked tail” current at negative potentials results from inwardly permeating cations directly repelling the QAs, which must undergo non-synchronous unbinding transitions before exiting the pore and allowing current to flow.

Figure 3.

Open-channel block as a basis for hooked tail currents. (A) Resurgent current-like hooked tail currents (arrow) appear with inward flux of K⁺ ions through squid axon K channels inactivated by quaternary ammonium ion. Left: normal artificial sea water (ASW). Right: high external Na (440 mM) ((51), with permission). (B) Time course of hooked tail current carried by K⁺ flowing inward upon repolarization (diamonds) matches that of recovery from N-type inactivation (squares) in Shaker K channels ((53), with permission). (C) Quaternary ammonium ions and the inactivating N-terminus bind in the K channel pore. Representation of the KcsA channel (blue) crystallized with the inactivating quaternary ammonium ion tetrabutylammonium (red) ((57), with permission). (D) Intracellular exogenous blockers applied to squid axons accelerate inactivation of Na currents upon depolarization (top, arrows) and generate hooked tail currents upon repolarization (bottom, arrows). Left: Pancuronium ion ((63), with permission); middle: N-methyl-strychnine ((64), with permission); right: azure A ((35), with permission). (E) Cryo-EM structure of quinidine (blue) in the Na_V1.5 channel pore. Note the binding of the IFM sequence of the III-IV linker outside the pore (arrow) ((66,67), with permission). (F) Drug-binding sites in the pore domain of Na_V1.7. Site E, μ-conotoxin; site S, TTX; site C, quinidine, propafenone, and cannabidiol; site BIG (beneath the intracellular gate), carbamazepine, lacosamide, and bupivacaine. Site I indicates binding site of the fast-inactivation linker, outside the permeation pathway. F, fenestrations; G, intracellular gate ((66,67), with permission).

Subsequent studies provided further evidence that open-channel block was not limited to exogenous compounds but that some voltage-gated channels contained native, “endogenous” blockers. Aldrich and colleagues discovered that, in Drosophila Shaker K channels, the N-terminal of the protein serves to inactivate the channel—that is, to render it non-conducting—by open-channel block, in a process they called N-type inactivation (9,10,40). The N-terminus responds to manipulations as predicted by Armstrong et al. (19) for a “ball-and-chain” model of inactivation: Cleavage of the N-terminus abolishes rapid inactivation, while application of the N-terminal “ball”-like peptide alone without its tethering N-terminal chain is sufficient to restore inactivation in a concentration-dependent fashion (9,10). Unbinding of the N-terminal gate is facilitated when inward flux at negative potentials is increased by raising external K ions. The resulting hooked tail currents have time courses that match recovery from inactivation (Fig. 3 B), suggesting that N-terminal binding occludes the pore of the channel in the permeation pathway (53,54,55). Further, the flow of current during recovery from inactivation suggests that channels blocked by the N-terminus must recover through an open state (54). Finally, single-channel analysis confirms that inactivated potassium channels usually reopen to recover (53).

The inference that both exogenous compounds and endogenous peptides can function as open-channel blockers received structural support from the first crystallization of a K channel, which directly showed that amino acids associated with TEA⁺ binding lie in the central cavity of the pore (56). Later crystallizations with the QA analog tetrabutylantimony revealed that the blocking ion binds in the pore, as do the last three amino acids of the Shaker N-terminus (57) (Fig. 3 C). N-termini extending into the permeation pathway and plugging the pore are likewise evident in structures of the inactivating bacterial Ca-sensitive K channel MtHK (58,59). Structural evidence for pore blockade by an N-terminus even extends to a tetrameric, protozoan Na channel, NaVEh, in which transient currents decay rapidly despite the absence of an inactivating IFM sequence; this channel also produces hooked tail currents (60,61).

Despite a variety of binding sites and obstruction mechanisms within the permeation pathway, open-channel blockers are functionally united by one key attribute. Unlike fast inactivation, which is independent of whether the channel is closed or open, relying only on the movement of VSD_IV and its sequelae as discussed above, pore blockade is by definition contingent, or dependent, on channel opening. As a consequence, pore blockade of Na channels cannot be modeled with Hodgkin-Huxley-type equations, which formalize independent activation and inactivation gating. The contingency can be expressed as serial rather than parallel transitions to reach a non-conducting blocked state: only O → B is possible while neither C → B nor I → B happens naturally. Conversely, channel closing (deactivation) is generally contingent upon blocker unbinding: B → O can occur, but not B → C. (Exceptions can occur in some cases at extreme voltages (51,62).) Thus, recovery from block happens only through an open state. For voltage-gated channels, which open at positive potentials and recover at negative potentials, pore block means that recovery—i.e., unblock—is necessarily associated with a flow of current at hyperpolarized potentials.

Regarding resurgent current, the question then becomes whether voltage-gated Na channels, which already have a native fast-inactivation gate, can be susceptible to open-channel block. The simple answer is yes: in work similar to Armstrong’s study of QAs, voltage-gated Na channels have been shown to be subject to pore blockade by several compounds, including pancuronium ions (63), N-methyl strychnine (64), thaizine dyes (35), and the synthetic peptide KIFMK (65). These open-channel blockers have two effects on inactivating Na current: 1) they accelerate current decay at positive potentials (Fig. 3 D, top); and 2) they unbind non-instantaneously at negative potentials before channels deactivate, generating hooked tail currents that resemble resurgent Na current (Fig. 3 D, bottom). Their action is effectively modeled as voltage-dependent pore blockade defined by the contingencies listed above. Indeed, cryo-EM structures (66,67) have allowed visualization of bound compounds in the Na channel pore, including the open-channel blocker quinidine (68), as well as several other drugs inferred to bind in the permeation pathway (Fig. 3, E and F). These results further support the idea that pore blockade of Na channels is structurally distinct from fast inactivation.

Resurgent current and open-channel block

The structural and physiological descriptions of identified open-channel blockers align well with the possibility that the X+ state from which resurgent current arises may be an open-channel blocked state. The kinetics of resurgent current have the hallmarks of recovery through an open state, and channel availability indeed increases and decays at −40 mV in parallel with the rise and fall of resurgent current (2,8). Additionally, even though the same α subunits are expressed by Purkinje cells, which have resurgent current, and hippocampal CA3 neurons, which do not have resurgent current, the fast decay phase of Na currents at positive potentials is faster in Purkinje cells, consistent with a cell-specific process that accelerates the rate of current decay (2). Finally, driving Purkinje Na channels into the X+ state antagonizes the action of lidocaine, which otherwise acts as an open-channel blocker (69).

Physically, a particle that acts as the X+ gate can be selectively destroyed by proteases such as trypsin or chymotrypsin (70). When protease is briefly applied to inside-out Purkinje cell patches held at −30 mV, which lets the fast-inactivation linker bind and thereby be protected from cleavage, the Na currents evoked after protease removal are altered to look like “classical” Na currents: First, the decay of transient currents evoked by depolarization becomes slower, resembling the time course of normal fast inactivation. Second, resurgent current is lost, indicating that the residual inactivated states do not recover through open states. The channels thus behave as though they have lost a proteinaceous, voltage-dependent pore blocker that is normally associated with Purkinje channels, even in excised patches (70,71) (Fig. 4 A). These results are all consistent with a scheme in which the X+ → O → X− responsible for resurgent current is B → O → I upon repolarization to moderately negative voltages; with repolarization to strongly hyperpolarized potentials, which favor deactivation, the transitions would be B → O → C.

Figure 4.

Interactions between fast inactivation and the open-channel blocked X+ state. (A) Intracellular proteolysis has the reverse effect of adding an exogenous open-channel blocker. Purkinje Na currents from an inside-out patch before (black) and after (gray) brief exposure to trypsin. With only classical fast inactivation remaining, transient currents are slowed (left) and resurgent current is removed (right) ((70), with permission). (B) Without expression of scn8a (Na_V1.6), Na currents from Purkinje cells show faster inactivation and smaller resurgent currents. Sample currents from a wild-type (black) and null mutant (red) Purkinje cell ((76), with permission). (C) Resurgent current is restored in Purkinje cells lacking scn8a expression by slowing fast inactivation with the site-3 toxin β-pompilidotoxin (β-PMTX). Na currents from an scn8a-null (med) Purkinje cells before and after application of the toxin. Note the slowing of transient current inactivation as well as the appearance of resurgent current ((75), with permission, copyright 2004, Society for Neuroscience). (D) Slowing inactivation with the site-3 toxin ATX-II lowers the affinity of the putative endogenous blocker. Resurgent current evoked upon repolarization to −30 mV averaged across several Purkinje cells and normalized to show the briefer rise time in the presence of toxin ((69), with permission).

Given the explanatory power of this model, it is reasonable to proceed under the hypothesis that X+ is an open-channel blocked state and X− is the fast-inactivated state until evidence emerges to the contrary. This scenario suggests that Purkinje cells (and many other cells that generate resurgent current) have two competing modes by which Na channels become non-conducting at positive voltages: 1) allosterically, through fast inactivation, which entails VSD_IV outward movement and binding of the DIII-DIV linker; and 2) directly, through open-channel block, which entails pore obstruction by an endogenous blocking particle. Structurally, fast inactivation can co-exist either with deactivated configurations (VSD_I, VSD_II, and/or VSD_III at rest) or with activated configurations (all three deployed, with or without an open intracellular gate), but open-channel block can occur only under the latter condition, with an open intracellular gate.

Two open-channel blocked states

Many experiments suggest another contingency, namely, that fast inactivation requires channels to be unblocked. Stated another way, channels that undergo open-channel block do not undergo normal fast inactivation (B → I does not occur). The strongest evidence for this idea is the existence of resurgent current itself: After brief depolarizations that drive channels into the blocked state, channels readily reopen upon repolarization, which they would not do if they had also fast inactivated (2,8,72). Recovery from the blocked state also restores channel availability more rapidly than recovery from fast inactivation, suggesting that open-channel block prevents fast inactivation from happening normally. Indeed, this ability to terminate Na current flow rapidly during the action potential, yet prevent fast inactivation, may be one of the most relevant physiological consequences of the open-channel blocking mechanism, since such a cycle of block and rapid unblock shortens refractory periods and enables rapid firing (73).

Such a mutual exclusion of B and I states generates the prediction that conditions that slow the rate of entry into fast-inactivated states should augment resurgent current. Indeed, the wasp venom β-pompilidotoxin, which binds extracellularly to VSD_IV and slows its deployment (74), enlarges resurgent current in Purkinje cells. The toxin does not, however, detectably change the conductance or voltage dependence of Na channels, nor does it induce hooked tail currents in cells that usually lack resurgent current (75). These results suggest that, with inactivation slowed, more channels undergo open-channel block and thereby contribute to resurgent current upon repolarization.

Conversely, speeding fast inactivation reduces resurgent current. In mice lacking the TTX-sensitive voltage-gated Na channel α subunit Na_V1.6, encoded by the gene scn8a, resurgent current is reduced by ≈90% in Purkinje cells (76). This result was initially puzzling, since it indicated that Na_V1.6 was required for resurgent current to be normal, yet it was not necessary in an absolute sense. Also, Na_V1.6 did not itself produce resurgent current in oocytes or HEK cells (77,78). The explanation became clear with the recognition that the residual Na currents in Na_V1.6-null Purkinje cells—likely carried by Na_V1.1 and perhaps Na_V1.2 (79)—undergo fast inactivation more rapidly than does Na_V1.6 (Fig. 4 B). Slowing inactivation in Na_V1.6-null Purkinje cells with β-pompilidotoxin reveals large resurgent currents (Fig. 4 C), which are lost after proteolytic cleavage of the putative X+ gate (75). The simplest explanation for these results is that an open-channel blocker, endogenously expressed in Purkinje cells but physically distinct from α subunits, effectively binds to open Na_V1.6 pores before the channels undergo fast inactivation. Other Na_V channels, such as Na_V1.1 and/or Na_V1.2, have the capacity to bind this endogenous blocker, but they generally inactivate too rapidly to undergo pore blockade. Thus, ample evidence suggests that fast inactivation precludes open-channel block and vice versa.

Nevertheless, this conclusion has an additional layer of complexity. Anemone toxin (ATX-II), a Na channel site-3 toxin like β-pompilidotoxin, slows VSD_IV movement and delays fast inactivation (80). In addition to enlarging the peak current evoked upon repolarization, however, ATX-II reliably speeds the rise time of resurgent current (69) (Fig. 4 D), suggesting that the endogenous blocker binds less tightly to its site in the pore when VSD_IV fails to move outward. Consistent with this idea, in expressed Na channel subunits mutated to prevent fast inactivation through linker binding (Na_V1.4_CW), comparison of currents with and without ATX-II suggest that VSD_IV deployment increases the affinity of free blocking peptides for their binding site (81); in the absence of blockers, VSD_IV movement also reduces single-channel conductance in these channels, providing evidence that it alters the pore (82). The open-channel blocker QX-222 shows a similar shift from low- to high-affinity binding with VSD_IV at rest versus outward, in Na_V1.4 channels with the inactivation linker mutated (83). Thus, open-channel block probably does not fully prevent VSD_IV movement. Instead, these data suggest the existence of a second “block-locked” configuration of the open channel, i.e., blocked, but with VSD_IV in the outward position (84). These channels cannot be in a stable fast-inactivated state, however; the fact that resurgent current flows at all indicates that VSD_IV must return to rest and the fast-inactivation DIII-DIV linker must unbind—if it binds at all to open-blocked channels—before the blocker is expelled.

Recent electrophysiological studies of Na_V1.4 channels suggest a model that unifies these observations. Mutation of specific pairs of pore-facing, bulky isoleucine and leucine residues on DIII-S6 and DIV-S6 to small-volume alanine residues renders channels “leaky” in the inactivated state, in that the usually non-conducting conformation still conducts current. As mentioned above, these data support the interpretation that linker binding outside the pore, which is favored by VSD_IV outward movement, allosterically shifts the isoleucine and leucine residues into the permeation pathway, forming a hydrophobic barrier to current flow (31). In the context of resurgent Na current, it seems plausible that when the pore is constricted by this barrier, the endogenous blocker cannot bind (I precludes B). In contrast, when the blocker binds to the non-constricted pore, VSD_IV movement can still occur, trapping/binding the blocker more tightly. Pore constriction (and possibly linker binding) must be incomplete, however, since repolarization can readily induce recovery from open-channel block (84).

Linking permeation and gating in resurgent current

What factors favor recovery from open-channel block? Because resurgent current is observed upon repolarization, the simplest idea is that block and/or unblock are intrinsically voltage dependent. Several experiments, however, reveal limitations of this interpretation. After short (10-ms) pulses to +30 mV, repolarization evokes large resurgent currents (and concomitant recovery of transient currents), but after longer (100-ms) pulses to +30 mV, resurgent current is smaller. With conditioning pulses to the less-depolarized potential of −30 mV, repolarization after short pulses evokes small resurgent currents, and repolarization after long pulses evokes no resurgent current at all (8). Since the magnitude of resurgent current varies directly with the proportion of channels blocked just before repolarization, the blocker must gradually be displaced from the pore, and this displacement must occur more readily at less-depolarized potentials.

This idea was examined quantitatively by assessing the magnitude of resurgent current evoked after different conditioning durations at different potentials (50). Resurgent current flows upon repolarization after brief steps to any depolarized potential. Its peak magnitude decreases, however, as the depolarization is prolonged, and the less depolarized the conditioning step, the more rapidly the peak resurgent current magnitude falls off with conditioning duration. The data support the idea that channel activation is nearly always followed by open-channel block. With strong depolarizations channels tend to remain blocked, whereas with weak depolarizations the block is unstable and reopening occurs, ultimately allowing a transition to a more stable inactivated state. Indeed, such transitions are likely to account for the strongly biexponential decay of Purkinje Na currents at voltages near −30 mV: the fast time constant is block-dependent, and the slow time constant reflects reopening associated with unblock and fast inactivation (8).

Clues to the mechanisms underlying block and unblock are once more provided by studies of QA blockers of K channels (51), with parallels in peptide block of Na channels (65). These exogenous blockers bind to a site that is exposed by channel opening. Likewise, the resurgent current-inducing endogenous blocker binds before the fast-inactivation gate at all potentials, dependent only on channel opening. Moreover, in other Na and K channels, open-channel blockers can be “knocked off” by inward flux of ions (Fig. 5 A), and the fact that hooked tail K currents are seen only in high extracellular K reveal that unblock is highly dependent on inward permeation. Similar phenomena are seen in Purkinje cells and cerebellar nuclear neurons (85). Tripling the concentration of external Na ions (from 50 to 150 mM) disproportionately increases the resurgent conductance relative to the transient conductance, >2.5-fold more than predicted from the increase in charge-carrying ions (85) (Fig. 5 B). These results indicate that the reduced concentrations of external Na often used experimentally to maximize voltage control may artificially stabilize blocker binding, underestimating resurgent current.

Figure 5.

Permeation-dependence of unblocking and the resulting resurgent current. (A) Schematic of an open, blocked Na channel α subunit (left) and expulsion of the blocking particle by inward flux of Na⁺ ions (right). Labels: Sans-serif Roman numerals I–IV, the four domains; I with serifs, the fast-inactivation domain; B, a blocking particle ((84), with permission). (B) Resurgent current is disproportionately increased by raising external Na⁺ ions. Left: transient and resurgent currents recorded from a neuron isolated from the mouse cerebellar nuclei (CbN), in 50 mM Na (low Na/Ca) and 155 mM Na (high Na/Ca). The transient current at +30 mV increases 4.5-fold and the resurgent current at −30 mV increases 8-fold, despite a lower proportionate contribution of the change in driving force. Inset shows the resurgent current at higher gain. Right: current-voltage relation for resurgent current in low and high Na ((85), with permission). (C) Resurgent current depends on inward flux of Na ions. Left: simulations from a Markov model with resurgent current generated by permeation-dependent open-channel block and unblock (colored traces). Right: experimental currents recorded from different Purkinje cells with gradients as indicated (black). The same voltage protocol pertains to all sets of traces. Lower families of traces are horizontally offset for clarity, and upward and downward arrows indicate the onset and offset of the 10-ms depolarization to +30 mV. Resurgent current is evident upon repolarization from +30 mV regardless of the direction of the preceding transient current, but outward resurgent current is always very small or undetectable, even with the strongest reverse gradient (inset). Note similarity between Na currents in the near-symmetrical gradient and Shaker K currents with a similar gradient in Fig. 3B (data from (37), with permission, copyright 2010, Society for Neuroscience; simulations from (84), with permission).

The “knockoff” mechanism of unblocking further raises the question of whether inwardly permeating Na ions may be necessary for displacement of the native blocker. If so, resurgent current would not flow outward, because blockers that bind open channels upon depolarization would not be displaced. This prediction is largely fulfilled in experiments on Purkinje cells in Na gradients that are reversed (E_Na at −45 mV) or near symmetrical (E_Na at +20 mV) (37). In near-symmetrical gradients, depolarization to +30 mV evokes outward transient current, and repolarization evokes inward resurgent currents. The resurgent conductances (measured to control for the differences in driving force) are consistently larger than in normal gradients (E_Na = +73 mV). In contrast, in reverse gradients, repolarization evokes tiny or negligible outward resurgent currents, giving conductances much smaller than in control conditions (Fig. 5 C). Both results provide evidence that the endogenous open-channel blocker can bind with current flowing outward, but it cannot easily unbind unless current flows inward.

Importantly, after depolarizations above E_Na, which elicit outward transient current, resurgent current evoked by repolarization below E_Na has consistent amplitude regardless of the magnitude or duration of the conditioning step. Hence, gradual displacement of the blocker largely disappears when current is outward, supporting the idea that the macroscopically time-dependent and voltage-dependent unblocking transitions of C → O → B → O → I actually rely on inward flux. In contrast, with fixed-duration conditioning depolarizations below E_Na, which evoke inward transient current, resurgent current evoked upon repolarization is directly proportional to the driving force during conditioning. Thus, resurgent current is proportional to the amount of inward permeation and blocker displacement that occurred during the conditioning step (37).

These lines of evidence all suggest that the voltage dependence of resurgent current is largely secondary to inward flux and that the unblock is permeation dependent (84). The open-channel blocker responsible for resurgent current therefore is unlikely to detect the membrane field directly but is electrostatically repelled by Na ions. This scenario is consistent with a blocker binding site in the permeation pathway, possibly beneath the intracellular gate (67) (see Fig. 3 F). These data lend further support to the hypothesis that the X+ state from which resurgent current flows is structurally equivalent to a blocked but open channel. Indeed, incorporating permeation dependence into a Markov model successfully recapitulates all attributes of resurgent current recorded to date (Fig. 5 C) (84).

Molecular mechanisms of endogenous block of K and Na channels

While the basic mechanism of open-channel block and unblock is shared by resurgent current of Na channels and recovery from N-type inactivation of Shaker K channels, the nature of the endogenous pore-blocking molecule differs. In Na channels, the blocker seems to be distinct from the α subunit, whereas in Shaker and related channels, including K_V1.4, K_V3.3, and K_V3.4, it is intrinsic to the pore-forming subunit itself (9,10). K channels, however, can also be inactivated by a wide range of associated proteins that act through open-channel blocking mechanisms. The K_Vβ1 subunit, for example, has an N-terminus that occludes the pore of the associated voltage-gated channel, inducing current decay in intrinsically non-inactivating K_V1.1 channels and accelerating current decay in naturally inactivating K_V1.4 channels (86); through similar means, K_Vβ3 induces macroscopic inactivation of cardiac K_V1.5 channels (87).

Such a molecular mechanism extends to the family of K channels gated by both Ca and voltage (BK, or slo channels), which can also show an inactivation phase of macroscopic K current (88). This decay phase in BK current is a result of open-channel block by the accessory subunits β2 (89,90), β3a (91), β3b (92,93), or LINGO1 (94). Unlike the simple one-step pore block of Shaker K channels, the onset of BK channel block is a two-step process, whose reversal can generate distinct patterns of recovery. Unblocking by β3a subunits produces prolonged tail currents that are likely to contribute to slow afterhyperpolarizations (91,95), whereas unblock by β3b is ultra-fast and virtually indistinguishable from a tail current (93,96).

In Na channels as well, channel block need not be permissive of hooked tail or resurgent current upon repolarization. Na_V1.3 channels of chromaffin cells inactivate via two mechanisms: a classical fast inactivation requiring the DIII-DIV linker and a slowly recovering, “long-term inactivation” mechanism that is dependent on the expression of fibroblast growth factor 14 (FGF14) (FGF homologous factor 4 [FHF4]) (97,98). The action of FGF14 (FHF4) is thought to be mediated by a stable open-channel block by the N-terminus of the FGF14 protein, whose unbinding occurs less readily than recovery from fast inactivation (99) (see also below).

Na_V1.6 and Na_Vβ4

The molecular identity of the open-channel blocker responsible for resurgent current remains an area of active investigation. As mentioned above, early observations provided evidence in Purkinje cells that Na_V1.6 channels are well suited to carry resurgent current but that non-Na_V1.6 channels with toxin-slowed inactivation can also undergo block and unblock (75,76). Likewise, although Na_V1.6 appears to be the primary carrier of resurgent current in many other neurons (100,101,102), it is not the only isoform that can carry the current, even without toxin modification (50,103,104,105,106,107). These results suggest two possible scenarios. Either Na_V1.6 as well as non-Na_V1.6 channels each contain an endogenous domain that can act as a pore blocker, or open-channel block is achieved by a protein(s) separate from the α subunit.

Consistent with the former supposition, the blocker is retained in outside-out and inside-out patches, suggesting a physical association with the α subunit that is undisturbed by patch excision (70,71). Consistent with the latter idea, however, no known voltage-gated Na channel α subunit expressed in isolation intrinsically produces resurgent current (77,78,81,108,109). Together the data suggest that the resurgent current is a modular feature, resulting from protein-protein interactions or modifications of Na channel complexes (71,75,109). Indeed, dephosphorylation by a broad-spectrum phosphatase eliminates resurgent current in Purkinje cells (71), making it seem possible that post-translational modification, absent in expression systems, may transform a cytoplasmic domain into a blocker. Nevertheless, no compelling candidate has yet emerged for a blocking domain within any α subunit, particularly one that can be switched on and off, justifying an exploration of Na channel-associated proteins that may act as blockers.

The first candidate for an open-channel blocker responsible for resurgent Na current in Purkinje cells was the Na channel β subunit Na_Vβ4 (70). This single-transmembrane subunit, encoded by the gene scn4b, attaches covalently to Na channel α subunits, much like Na_Vβ2, the subunit that it most resembles (110). The distinguishing feature of Na_Vβ4 is its 9-amino-acid insertion at the beginning of the cytoplasmic tail that includes a series of positively charged lysines (K) surrounding a phenylalanine (F); it is also highly expressed in Purkinje cells. As described above, when the native open-channel blocker is proteolytically removed from Purkinje cell inside-out patches, transient current is slowed and resurgent current disappears. In the same patches, a peptide including the Na_Vβ4 insertion (KKLITFILKKTREK), dubbed the “β4 peptide,” is sufficient to restore both transient current and resurgent current, with amplitude and kinetics that precisely match the natural condition (70) (Fig. 6 A). The onset of block, reflected by the time course of transient current decay, varies with the concentration of free peptide and matches the native currents at 100 μM, a concentration consistent with that estimated for K channel N-type inactivation gates (10). The time course of unblock is set by the dissociation rate of the blocking agent for the channel, which is reflected by the rising phase of blocked-induced hooked tail or resurgent currents (51,111). The similarity of unbinding rates of the β4 peptide and the native blocker, at likely comparable concentrations, suggests that the two molecules have a similar affinity for the Purkinje Na channel.

Figure 6.

Candidate proteins involved in resurgent current in different cells. (A) The cytoplasmic tail of Na_Vβ4 (scn4b) reproduces the magnitude and kinetics of resurgent Na current in Purkinje cells. Representative trace of transient currents evoked by steps from −90 to 0 mV (left) and resurgent currents evoked by steps to −30 mV after 10 ms at +30 mV (right) in control conditions (black), after proteolytic cleavage of the putative blocker (gray), and after application of 200 μM of the β4 peptide (red) ((70), with permission). (B) Deletion of scn4b reduces resurgent current in medium spiny neurons. Sample traces from neurons from wild-type (left) and knockout (right) mice ((116), with permission). (C) Knockdown of Na_Vβ4 reduces resurgent current in cultured cerebellar granule cells. Sample traces of resurgent currents in three representative cells, transfected with either control non-targeted siRNA (black), on-target siRNA to knock down Na_Vβ4 expression (red), or on-target siRNA with the β4 peptide added to the intracellular solution (blue) ((117), with permission) (D) Expression of wild-type but not mutant Na_Vβ4 induces resurgent current in expressed Na_V1.6 channels engineered to be TTX-resistant (Na_V1.6_r) in dorsal root ganglion neurons. Left: schematic of Na_Vβ4 with the lysine residues expected to be sensitive to knockoff highlighted in the wild-type and substituted with alanine residues in the mutant. Right: A representative family of negligible currents (gray) evoked by repolarization with co-transfection of DRG cells with Na_V1.6_r and the alanine mutant Na_Vβ4, with the largest current in red, with the corresponding resurgent currents from Na_V1.6_r alone (black) or co-transfected with wild-type Na_Vβ4 (blue) ((118), with permission). (E) Expression of FGF14-1a (FHF4A) induces resurgent current in expressed Na_V1.8 channels but not Na_V1.6 channels. Representative families of Na currents evoked by repolarization following brief depolarizations in Na_V1.8 alone (left) or with co-transfected FGF14-1a (FHF4A) (middle), illustrating the first successful reconstitution of resurgent Na current in an expression system. Right: expressed Na_V1.6 generates no resurgent current with or without co-transfection of FGF14-1a (FHF4A) or either the A or B isoform of the related protein FGF13 (FHF2) ((109), with permission).

Moreover, mutating the phenylalanine or the lysines, or altering their sequence, yields peptides that remain capable of open-channel block and unblock, but with affinities that differ greatly from the control condition. The free β4 peptide can also induce a resurgent-like current in cells lacking an endogenous blocking particle. Thus, the native blocker of Purkinje Na channels seems likely to have a sequence and/or structure highly similar to that of the β4 peptide. From these results, Na_Vβ4 has emerged as a plausible candidate for the native open-channel blocker (3,70).

Subsequent studies have offered evidence both for and against this idea. On the “con” side, the straightforward test of sufficiency fails, as expressing Na_Vβ4 with any of several α subunits does not yield Na channels with a resurgent component (108,112,113). The simplest test of necessity also fails, as knocking out Na_Vβ4 either constitutively (114) or embryonically, or deleting only the insertion in the cytoplasmic tail (115), does not eliminate resurgent current in Purkinje cells.

On the “pro” side, deletion of Na_Vβ4 indeed abolishes or greatly diminishes resurgent current in striatal medium spiny neurons (116) (Fig. 6 B). Likewise, acute small interfering RNA (siRNA)-mediated knockdown of Na_Vβ4 reduces resurgent current and the associated rapid repetitive firing in cultured cerebellar granule cells (117) (Fig. 6 C) and dorsal root ganglion neurons (118); it also slows firing in adult Purkinje cells (114). Most interestingly, co-expression of a TTX-resistant Na_V1.6 with Na_Vβ4 in dorsal root ganglion (DRG) neurons yields TTX-resistant Na currents with a large resurgent component in 100% of cells tested, whereas Na_Vβ4 with the cytoplasmic lysines mutated to alanines does not generate this increase in resurgent current (118) (Fig. 6 D).

Together, the data show that expression of Na_Vβ4 alone does not guarantee that a cell produces resurgent current, yet Na_Vβ4 indeed has the capacity to influence and possibly generate resurgent current under some circumstances. While such observations can be initially bewildering, they have precedent in K channels, in which both pore-forming subunits and blocking proteins can be modified to alter susceptibility to inactivation. For instance, pore blockade can be prevented in K_V1 channels by modulatory domains within the pore-forming subunits, such as the N-type inactivation-prevention domain of K_V1.6 channels (119), or by additional proteins within ion channel complexes, such as Lgi1 (120). Conversely, in mammalian K_V1 channels, the redox state of pore-blocking β subunits can affect binding interactions, which either restrain them to prevent inactivation or free them to inactivate channels (121). Thus, it seems possible that either Na channel α subunits or Na_Vβ4 itself must undergo cell-specific transformations that are permissive for functional open-channel block.

Other molecular participants in resurgent current

Nevertheless, the resistance of resurgent current to deletion of scn4b, at least in Purkinje cells, suggests the existence of resurgent current-generating blocking particles besides Na_Vβ4. To begin the search for such proteins, Lewis and Raman (111) identified common attributes of putative blocking peptides. By testing β4-like peptides that were either species variants or targeted mutants, they found that a phenylalanine and lysines, in the fixed structural relationship found in the wild-type mouse β4 peptide, are critical for generating resurgent-like currents in CA3 neurons, which lack a native blocker. Using this template, White et al. (115) identified several proteins expressed in Purkinje cells, including FGF14-1a (FHF4A), GPR158, FGFR3, and MCTP1, all of which share the sequence KxxxFxI/LK. When peptides corresponding to the putative blocking sequence of each protein were applied intracellularly to CA3 cells, the FGF14-1a (FHF4A) peptide produced a large resurgent-like current. The other peptides did not evoke clear resurgent-like currents, although only the FGFR3 peptide was absolutely without effect; likewise, deleting the FILKKTR sequence from the β4 peptide (and adding residues to maintain a 14mer) abolished its blocking/unblocking action. These results suggest that other residues, possibly the R shared by the FGF14-1a and β4 peptide that is absent from the other candidates, are also crucial for block. Moreover, mutation of the leucine residue to a serine on the FGF14-1a (FHF4) peptide, to match the largely identical region on FGF13 (FHF2), also eliminated its blocking action. Thus, key residues for Purkinje-cell-like open-channel block may include FxI/LKKxR (115).

Interestingly, FGF14 (FHF4) had previously been identified as a modulator of high-frequency firing in cerebellar neurons as well as of resurgent current. Despite the apparent association of FGF14 (FHF4) with induction of long-term inactivation, deletion of the protein reduces rather than raises excitability in cerebellar neurons (122,123). Likewise, knockdown of FGF14 (FHF4) in cultured Purkinje cells stabilizes fast inactivation and reduces resurgent current (124). Similar changes are seen in Purkinje cells acutely isolated from global FHF14 (FHF4) knockout mice (115). Yan et al. (124), however, found that the wild-type gating properties of FGF14 (FHF4)-knockdown cultured Purkinje cells can be recovered by expression of the 1b isoform, which lacks the putative blocking sequence of 1a. This observation raises the possibility that FGF14 (FHF4) deletion reduces resurgent current by accelerating fast inactivation rather than by eliminating an open-channel blocker. Slowing fast inactivation with ATX-II in FGF14 (FHF4) knockout mice, however, does not restore resurgent current to control values (115), leaving the protein’s mechanism of action ambiguous.

Recent work helps clarify this ambiguity. In DRG neurons, knockdown of FGF14 (FHF4) 1a isoform greatly decreases resurgent current through the naturally TTX-resistant Na channel of the peripheral nervous system Na_V1.8 (109), much like Na_Vβ4 knockdown in granule cells (117). Xiao et al. (109), however, further demonstrated that resurgent current is generated in HEK293 cells by co-expression of Na_V1.8 or Na_V1.9 with FGF14-1a (FHF4A) (Fig. 6 E, left and middle). This result is the first successful reconstitution of resurgent current in a heterologous system by co-expression of a Na channel α subunit and one other full-length protein! Moreover, the related protein FGF13-1a (FHF2A) induces a resurgent current in the (cardiac) subunit Na_V1.5 or the (peripheral) subunit Na_V1.7. To date, these are the only reconstitutions of resurgent current in expression systems.

Two puzzles regarding the role of FGF proteins in producing resurgent current remain, however. The first relates to the domain that interacts with the channel, and the second relates to its mechanism of action. Both White et al. (115) studying CA3 cells, which primarily express Na_V1.6 and Na_V1.1, and Xiao et al. (109), studying expressed Na_V1.8, found that a peptide from the N-terminus of FGF14-1a (FHF4A) could induce resurgent-like current in Na channels lacking it. The peptides, however, were different in the two studies. (Neither study tested the peptide used in the other paper.) The FHF14-1a (FHF4A) peptide that induced resurgent-like current was AAAIASGLIRQKRQAREQHW (residues 2–21, here the AAA peptide) in the peripheral channels, but KVRIFGLKKRRLRR (residues 50–63) in CA3 neurons. Notably, co-expression in HEK cells of full-length FGF14-1a (FHF4A) with Na_V1.6 does not produce resurgent current (Fig. 6 E, right) (109), and the AAA peptide applied to expressed Na_V1.6 channels produces long-term inactivation rather than resurgent current (99).

A straightforward interpretation might be that the AAA peptide sequence in the native FGF14 (FHF4) protein blocks the pore in a manner that reverses rapidly in peripheral (DRG) α subunits but only slowly in central (Purkinje) α subunits. Despite high FGF14 (FHF4) expression, however, repeated depolarizations do not induce long-term inactivation in wild-type Purkinje cells (69). Moreover, deletion of the FGF14 (FHF4) protein promotes rather than relieves inactivation in Purkinje cells, despite loss of the putative long-term inactivation particle (115). These data indicate that Purkinje channels are somehow protected from long-term inactivation—possibly by a different, resurgent-current-inducing pore blocker.

The data also raise the possibility that the reconstituted resurgent current in TTX-resistant channels might occur through a means other than (or in addition to) pore blockade. Co-expression of FGF14-1a (FHF4A) with Na_V1.8 depolarizes the midpoint of the steady-state inactivation curve by 20 mV while hyperpolarizing the activation curve by 6–7 mV; similar changes, though not as extreme, are present in Na_V1.9. These large shifts broaden the window in which both fast inactivation and deactivation are unfavored, which are precisely the conditions that could induce a blocker-independent resurgent current through recovery currents. Studies of the permeation dependence of the current may provide insight into whether resurgent Na currents induced by FGF (FHF) isoforms rely on a pore-blocking mechanism. Regardless of mechanism, however, these auxiliary proteins clearly generate a resurgent current in peripheral neurons, with potentially significant physiological consequences.

Conclusions

Originally identified as an idiosyncrasy of Purkinje neurons, resurgent current has emerged as a characteristic component of Na channels in a wide range of neurons, as well as a factor in diseases of hyperexcitability. The term “resurgent” was initially applied to Purkinje Na currents with a definition of channel reopening associated with recovery from inactivation upon repolarization. By analogy, any macroscopically voltage-dependent reopening from an unavailable, non-conducting state might be described as resurgent (125). Mechanistically, however, it is now evident that resurgent Na current in Purkinje cells—and likely in other central neurons with TTX-sensitive Na channels—results from the relief of open-channel block by a protein closely associated with, if not part of, the pore-forming α subunit. This cycle of blocker binding upon channel opening (with depolarization) and blocker displacement by inward permeation (upon repolarization) is analogous to N-type inactivation and recovery in K_V1 channels. Since the physiological reversal potential for K ions is quite hyperpolarized, however, the driving force on K current at negative voltages favoring recovery from block is low, predicting only subtle effects on excitability. Conversely, for sodium channels, in which the driving force on Na ions is high at hyperpolarized potentials, the consequence of inward flux associated with recovery may be more substantial. Functionally, both the inward current and restoration of availability at least must counteract the tendency of K efflux to repolarize the cell strongly, and at most may trigger another action potential soon after the downstroke of the preceding action potential, facilitating rapid or burst firing.

The body of data on resurgent current, along with its relation to functional and structural studies of other ion channels, also allows a more precise formulation of questions remaining to be addressed, which include but are not limited to the following. 1) What is/are the native blocker(s) of Purkinje cells and other central neurons? The evidence suggests the existence a family of proteins with functional redundancy with identifiable shared characteristics. 2) Are Na channels or associated subunits modified to regulate the efficacy of block, and, if so, how? The data from K channels present tantalizing precedents for complex regulation of resurgent current. 3) Does open-channel block underlie resurgent current in all Na channel isoforms, or do some recover from fast inactivation through an open state? Experiments with altered gradients offer a route to an answer. 4) Where do resurgent current-inducing blockers bind? The significant advances in structural studies may soon shed light on this question.

From a physiological standpoint, the magnitude and time course of current are the only important factors with regard to shaping excitability. From a biophysical standpoint, however, the mechanism is likely to matter, both for accurately defining channel-gating properties and from the perspective of drug development and therapeutic advances, where modulating resurgent current may be desirable. Keeping the accumulated mechanistic knowledge in the foreground of subsequent studies promises to facilitate progress in understanding resurgent current in both health and disease.

Author contributions

T.K.A. and I.M.R. drafted, edited, and wrote the manuscript.

Editor: Meyer Jackson.