Simulation in Sensory Neurons Reveals a Key Role for Delayed Na+ Current in Subthreshold Oscillations and Ectopic Discharge: Implications for Neuropathic Pain

Yifat Kovalsky, Ron Amir, Marshall Devor


Somata of primary sensory neurons are thought to contribute to the ectopic neural discharge that is implicated as a cause of some forms of neuropathic pain. Spiking is triggered by subthreshold membrane potential oscillations that reach threshold. Oscillations, in turn, appear to result from reciprocation of a fast active tetrodotoxin-sensitive Na+ current (INa+) and a passive outward IK+ current. We previously simulated oscillatory behavior using a transient Hodgkin–Huxley-type voltage-dependent INa+ and ohmic leak. This model, however, diverged from oscillatory parameters seen in live cells and failed to produce characteristic ectopic discharge patterns. Here we show that use of a more complete set of Na+ conductances—which includes several delayed components—enables simulation of the entire repertoire of oscillation-triggered electrogenic phenomena seen in live dorsal root ganglion (DRG) neurons. This includes a physiological window of induction and natural patterns of spike discharge. An INa+ component at 2–20 ms was particularly important, even though it represented only a tiny fraction of overall INa+ amplitude. With the addition of a delayed rectifier IK+ the singlet firing seen in some DRG neurons can also be simulated. The model reveals the key conductances that underlie afferent ectopia, conductances that are potentially attractive targets in the search for more effective treatments of neuropathic pain.


Repetitive discharge that develops ectopically in injured sensory neurons after nerve trauma is thought to contribute significantly to chronic neuropathic dysesthesias and pain (Devor 2006). The repetitive discharge, in turn, appears to be triggered by the depolarizing limb of intrinsic subthreshold membrane potential oscillations (Amir et al. 1999; Kovalsky et al. 2008; Yang et al. 2009). Here we explore the conductances that sustain the oscillatory behavior and resulting spike discharge. A simulation approach was taken in the absence of adequate pharmacological tools for controlling the relevant conductances in live cells.

Voltage-dependent Na+ conductance (gNa+) contributes to the depolarizing phase of the oscillations in live cells (Amir et al. 1999; Xing et al. 2001). Blocking voltage-sensitive K+ channels does not eliminate oscillations or spiking, although shifting the K+ reversal potential to neutralize all K+ channels does. This indicates a role for a passive gK+ in the repolarizing phase of oscillations. However, in contrast to live cells, in numerical simulations that included only a classical Hodgkin–Huxley-type voltage-dependent transient gNa+ and leak gK+, the oscillations formed never triggered spikes and the range of membrane potentials within which oscillation could be induced was very narrow (Amir et al. 2002a). This suggests that additional gNa+ components may be required. Indeed, dorsal root ganglion (DRG) cells are known to express many types of Na+ channels. These can be subdivided by their tetrodotoxin (TTX) sensitivity and their relative contribution to transient and more sustained Na+ currents (Amir et al. 2006). Our goal here was to find a set of Na+ conductances that, combined with a physiologically realistic leak current, is capable of inducing oscillations and triggering spikes within a “window” of membrane potentials compatible with live DRG neurons.

Our work was based largely on that of Baker and Bostock (Baker 2000; Baker and Bostock 1997, 1998) who presented evidence that in addition to the known fast inactivating transient gNa+, large-diameter DRG neurons in adult rats also generate a delayed current, including “late” and “persistent” components. The delayed current is TTX sensitive (TTX-S), it activates rapidly, but inactivates slowly. Baker and Bostock described the late current using the sum of two exponentials as it decays according to intermediate and slow time constants (τ). The component that inactivates at the intermediate rate was termed “late1” and the component that undergoes slow inactivation, “late2.” They also recorded a component that never fully decays to zero, termed the “persistent” current (INa+p). The decay rates τ of late1 and late2 currents are in the range of tens and hundreds of milliseconds, respectively. Delayed and, particularly, persistent Na+ currents are known to mediate subthreshold oscillations and to participate in burst generation in many brain cell types (e.g., Agrawal et al. 2001; Wu et al. 2005). The aim of the present study was to find whether the addition of late and persistent Na+ currents to the fast transient gNa+ used in the earlier DRG simulation would significantly enhance oscillatory behavior and ectopic electrogenesis.


We constructed a mathematical model of an isolated large-diameter DRG sensory cell soma based on quantitative voltage-clamp data recorded from dissociated neurons from the rat. The simulation used the NEURON programming environment (version 5.8; Hines and Carnevale 1997), running under Windows XP. We modeled a sensory cell soma that was designed to be simple, with as few unconstrained variables as possible. We used the Crank–Nicholson second-order accuracy method for integration (dt = 0.01 ms) with one computational compartment. Simulated impulses were monitored in the middle of the compartment. Prolonged step depolarizing stimuli (3–5 s) were given under current-clamp conditions. The stimulus amplitude was varied systematically with accuracy of 0.01 nA, increased to 0.001 nA for epochs of rapid transition. The range of amplitudes and durations examined are given in results.

We focused on somata of medium- to large-diameter neurons with myelinated Aβ afferent axons because these are the main contributors to ongoing spike discharge originating in DRG neurons in a variety of animal models of neuropathic pain, especially around the time of onset of mechanical hypersensitivity (Amir et al. 1999; Boucher and McMahon 2001; Devor 2006; Liu C-N et al. 2000a; Liu X et al. 2000b; Tal et al. 2006). Although these neurons normally signal touch and vibration sense, in the event of neuropathy they are widely believed to contribute to pain (tactile allodynia; Campbell et al. 1988; Devor 2009a; Koltzenburg et al. 1994). This modality change is thought to result from: 1) the presence of central sensitization and/or 2) a phenotypic switch that causes the neurons to begin expressing and releasing neurotransmitters that are normally exclusive to nociceptors and that are capable of driving postsynaptic pain signaling pathways and inducing spinal central sensitization (Devor 2006, 2009a; Malcangio et al. 2000; Noguchi et al. 1995; Weissner et al. 2006; Woolf and Salter 2000).

The parameters of interest were oscillatory behavior and the generation of spikes, spike bursts, and repetitive (sustained) firing. The oscillation amplitude was measured peak to peak. “Repetitive firing” was defined as firing that, once initiated, persisted at least until the end of the stimulation step. Spike frequency was computed as the inverse of the last interspike interval (ISI). Where possible oscillation frequency was computed by averaging over the entire trace. When firing interrupted oscillations, the average oscillation frequency was computed over the longest run of oscillations available between spike bursts. The presence of oscillations and spiking complicates the evaluation of membrane potential (Vm) during stimulus traces. When oscillations were present Vm was computed as the average of the peak and trough of the oscillations. Vm was not defined during periods of spiking.

As in live DRG neurons, sustained oscillations (i.e., oscillations that persisted throughout the prolonged stimulus pulse) occurred only within a defined “window” of membrane potentials. At potentials more hyperpolarized or more depolarized than this window, oscillations were either not present or they damped out. The boundaries of this window depended on parameters of the model cell. Under some conditions oscillations led to spiking within the range of membrane potentials that supported oscillations. The “threshold” for single spike generation was defined as the potential at which INa+ = IK + ILeak within a sharp voltage peak, where INa+ refers to the sum of all of the individual Na+ currents present. IK is the potassium K-current and ILeak is a voltage-insensitive linear leakage current. Spike height was measured from the resting membrane potential (Vr) to the depolarizing peak of the spike generated by a 1-ms pulse at threshold amplitude.

Membrane properties of the model neuron

An isopotential cylinder with 50-μm diameter and length was used to model a DRG neuron. Specific membrane capacitance was 1 μF/cm2 and temperature was 20°C. The cell contained fast, intermediate, and slow inactivating gNa+, each with transient and persistent components. These were meant to emulate experimentally recorded TTX-S Na+ currents in the cells of interest. We did not aim to model specific Na+ channel types. In particular, there was no Nav1.9-like current that yields INa+p without a fast INa+ component. Nav1.9 is selective to small-diameter DRG neurons and is not present in the cells modeled (Waxman et al. 1999).

There was also a voltage-insensitive linear leakage current, defined as ILeak = gLeak(VmELeak), where gLeak is the leak conductance and ELeak is the reversal potential of the leak current. ELeak was set at −65.5 mV, yielding Vr = −57.4. This approximates Vr of live DRG neurons treated with blockers of voltage-sensitive K+ channels (Amir et al. 2002a). The leak conductance was set at 1.42 mS/cm2 (calculated from Scroggs et al. 1994). To keep the model simple we generally excluded voltage-sensitive K+ conductances (gK+). We have previously shown in excised rat DRGs that K+ channel blockers do not prevent oscillatory behavior or spiking. In fact, they are facilitatory in the sense of allowing oscillations at membrane potentials closer to Vr and at lower frequencies (Amir et al. 2002a). Nonetheless, in some simulations we included a DRG-appropriate delayed rectifier gK+ to test whether this qualitatively changes the behavior of the system. Most of the parameters required by NEURON were provided by the literature. Some had to be estimated, as described in the following text.

Voltage-dependent Na+ currents

Our objective was to modify the model used by Amir et al. (2002a) in a manner that would generate oscillatory behavior over a broader, more physiological range of membrane potentials and that would also yield oscillation-triggered spikes and spike bursts. Amir et al. (2002a) used a single transient TTX-S-type voltage-sensitive Na+ conductance with kinetics of the squid giant axon Na+ channel (Hodgkin and Huxley 1952; hereafter H-H), but with maximal Na+ conductance (gNa+max) increased to more closely resemble that of mammalian DRG neurons. Here we replaced this with fast activating and fast inactivating Na+ currents based on DRG neurons (INa+fast; see following text). In addition, we included two types of fast activating and slow inactivating “delayed” Na+ conductances, known to play a role in large- to medium-diameter DRG neurons. Emulating live DRG neurons, both of these conductances (late1 and late2) were described by an H-H-type equation that generated a fast transient current as well as delayed and persistent currents.

The current that occupies the first 2 ms from the beginning of the stimulus pulse, referred to here as the “fast transient” INa+, derives almost exclusively from the fast and late1 conductances. From 2 to nearly 200 ms the current declines to a steady-state “persistent” value. For convenience, we call the declining current during the period 2–20 ms the “intermediate” INa+ and the declining current from 20 to 200 ms the “late” INa+. The intermediate and the late currents both derive from the two delayed conductances, late1 and late2. Following Baker and Bostock (1998) we term the “persistent” (steady-state) current remaining after 200 ms INa+p. It derives from all three conductances (fast, late1, and late2). In the model, the maximal value of INa+p is controlled exclusively by the maximal value of the three declining currents. It could not be adjusted independently (Fig. 1).

FIG. 1.

Characteristics of the simulated sodium currents. A: total current, summing all of the INa+ components, evoked by 200-ms voltage-clamp steps (−80 to 10 mV in 10-mV increments) from a prepulse potential of −110 mV. Inset: enlargement of the first 120 ms better shows the delayed currents. BD: current–voltage (IV) relation for the total INa+ shown in A. Curves plot the fast transient current peak (B), the current present 30 ms after the step depolarization (C), and the current present 200 ms after the step depolarization (D). The stimulation protocol for D differed from that of B and C, as indicated in the text. EG: the individual components of the total INa+: INa+fast (E), INa+late1 (F), and INa+late2 (G). Calibration in F is as in E. Note the different scale in G.

The modeling process began by systematically varying parameters (gNa+max and the values of infinity activation [m] and inactivation [h]) of the three Na+ currents within a range that produced an integrated peak INa+ and current–voltage dependence, consistent with measurements made in live DRG cells (peak INa+ up to −126 nA; see results). The objective was to find a combination of values that yielded oscillations and oscillation-evoked spike discharges resembling those of live cells as closely as possible. The values identified in this way were used for the “baseline model.” The effects on the baseline model of perturbations from these values are shown in results. The m and h variables were written as Boltzmann equations: m or h = 1/{1 + exp[(V1/2Vm)/slope factor]}, where V1/2 is the potential of half activation or inactivation. ENa was set at +62 mV (Baker and Bostock 1998). The three Na+ currents were as follows (summarized in Table 1).

View this table:

Parameters of the conductances used for the baseline model

1. Fast-inactivating Na+ current (INa+fast).

INa+fast was based on the conventional H-H model (with modified values): INa+fast = gNa+fastmax × mmmh(VmENa), where gNa+fastmax is the maximal fast conductance, which was set at 25 mS/cm2 in the baseline model. The values of m and h were modified to fall within the range present in large-diameter DRG neurons in rats. For m we set V1/2 to −34.12 mV and the slope factor to 9.14. In live cells V1/2 of m lies between −35 and −25 mV and the slope factor between 3.1 and 12 (Cummins et al. 1999; Hong and Wiley 2006; Yoshimura et al. 1996; Yu et al. 2005). The activation time constant (τm), chosen to simulate fast activation, was voltage sensitive, but always remained ≤0.11 ms (Kostyuk et al. 1981; Nowycky 1992). τm = 0.1092 exp{−0.5[(Vm + 28.71)/25.5]1.8}. For h we set V1/2 to −56.39 mV and the slope factor to 7.22. In live cells of small and large diameter V1/2 of h has been reported to lie between −69 and −56 mV and the slope factor between 7.3 and 16.5 (Baker and Bostock 1997, 1998; Cummins et al. 1999; Hong and Wiley 2006; Yoshimura et al. 1996). We acknowledge that the values chosen are low in the physiological range. The inactivation time constant (τh) was voltage sensitive and always remained ≤2 ms. τh = 0.246 + 1.63 exp{−0.5[(Vm + 61.87)/15.25]2}. Using these parameters INa+fast decayed to a small residual persistent current as in live cells (Chen et al. 2000); it did not fall to absolute zero (Fig. 1E).

2. Intermediate-inactivating Na+ current (INa+late1).

INa+late1 was calculated as: INa+late1 = gNa+late1max × mh(VmENa), where gNa+late1max is the maximal late1 conductance. The variable m was used to the first rather than the third power because this gave a better fit to the biological data. INa+late1 is known to be fast activating (Baker and Bostock 1998). However, since no direct measurements of its activation characteristics are available we set τm = 0 in the baseline model and explored the impact of increasing this parameter by simulation. For m we set V1/2 to −25.29 mV, the slope factor to 9.052, and m = m. Baker and Bostock (1998) could not estimate the inactivation steady state or kinetics of INa+late1, but they state that the kinetics (τh) appeared to be similar to that reported by Caffrey et al. (1992)) for INa+fast in excised patches from similar neurons. τh and gNa+late1max, which were extracted from Caffrey et al. (1992), were set to τh = 0.2218 exp(−0.06883Vm) (yielding 1–20 ms depending on Vm) and gNa+late1max = 27 mS/cm2. For h we set V1/2 to −72.5 mV and the slope factor to 8.0, slightly modified from Caffrey et al. (1992). Using these parameters INa+late1 decayed more slowly than INa+fast and, like INa+fast, it did not reach zero at steady state, but rather ended in a small residual persistent current (Fig. 1F).

3. Slow-inactivating Na+ current (INa+late2).

INa+late2 = gNa+late2max × mh(VmENa). The physiological value of gNa+late2max has not been measured directly. We set gNa+late2max in the baseline model to 0.128 mS/cm2, which resulted in a maximal total persistent current of −1.68 nA when all three contributions were summed (at 200 ms; Fig. 1D). This is near the upper end of the range (−0.4 to −1.7 nA) proposed by Baker and Bostock (1997). The values of m and h are taken from Baker (2000): m = 1/{1 + exp[(−51.8 − Vm)/4.6]}, h = 0.9827/〈1 + exp{−[(Vm + 55.67)/−6.552]}〉. Baker and Bostock (1997) had difficulty quantifying τm, although their data show that the current activates very quickly. Thus as for INa+late1, we set τm as instantaneous in the baseline model and explored this parameter space by simulation. The variable τh, derived from Baker and Bostock (1998), was set to τh = 1/[0.04 exp(Vm/25.5)] + 63.2 (yielding 100–200 ms depending on Vm). Using these parameters INa+late2 decayed to a small residual persistent current; it did not fall to absolute zero (Fig. 1G).

4. Persistent Na+ current (INa+p).

As noted, there was no independent INa+p. The persistent current constituted the persistent residues of the fast and two late Na+ currents. The time point of transition between the late (but still decaying) currents INa+late1 and INa+late2, and true INa+late-derived persistent current is arbitrary. As noted, the boundary was set at 200 ms (Fig. 1D). INa+p declined by only an additional 50% when tracked out to 5 s.

Delayed rectifier potassium current (IK+)

The potassium K-current was modeled as IK+ = gK+max × n(VmEK), where n is the activation variable. The first rather than the fourth power of n was used to improve the fit to the biological data. gK+max used was 1.5 mS/cm2, which is within the physiological range (0.7–2.6 mS/cm2), calculated from Everill et al. (1998) and Rola et al. (2003). V1/2 of activation and the slope factor were taken from small- and medium-sized DRG neurons (Rola et al. 2003) in the absence of data from large DRG neurons. We used these values to set n = 1/{1 + exp[(Vm + 9.2)/−16]}. τn = −23 + 69.46 exp(−0.0142Vm) following Fedulova et al. (1998). EK was set at −94 mV (Scroggs et al. 1994). Although derived from various sources, the voltage-clamp plot using these values fits nicely to that of large DRG cells (Everill et al. 1998).


Subthreshold oscillations and ectopic discharge in the baseline model cell

In current clamp we injected the cell with a series of 3-s depolarizing steps of gradually increasing amplitude. This evoked a characteristic range of oscillatory and spiking behavior quite similar to that observed in live DRG cells (see the following text). The model cell did not oscillate at rest (Vr = −57.4, Fig. 2A). However, a small depolarization evoked damped oscillations (Fig. 2B) and a slightly increased current evoked sustained oscillations (stimulus = 0.012 nA; Vm = −57.2 mV; oscillation frequency = 44.7 Hz; amplitude = 2.3 mV, Fig. 2C). Another increment in stimulus amplitude induced oscillations that triggered repetitive spike bursts (Fig. 2D) and a further increment yielded a spike burst at stimulus onset, without prior oscillations, that was followed by repetitive bursting (Fig. 2E). Subthreshold oscillations were present between the bursts. During the course of bursts spike amplitude declined slightly as did within-burst frequency. The duration of bursts and the interval between bursts were voltage sensitive; increasing stimulus current increased burst duration, whereas the interval between bursts decreased (Fig. 2, D and E). Finally, with a stimulus of 0.063 nA, the cell fired in a sustained manner for the duration of the 3-s pulse (at 36 Hz, Fig. 2F). Firing frequency proceeded to increase with further depolarization, reaching a maximum of 99.3 Hz with stimuli of 1.615 nA. With still greater depolarization, spikes (including spike bursts) ceased. In their place sustained oscillations emerged at about the frequency of the spikes. Further depolarization caused these oscillations to gradually decrease in amplitude and increase in frequency up to a maximum of 110.5 Hz at Vm = −36.67 mV (Fig. 2G). Oscillations ceased when the cell was still further depolarized.

FIG. 2.

The baseline model cell generates subthreshold oscillations, spike bursts, and tonic spike discharge. A: no oscillations were present at Vr (−57.42 mV). B: injection of a constant depolarizing current (top trace) sufficient to bring Vm to −57.29 mV (0.011 nA), evoked a brief run of damped oscillations (bottom trace). C: stable subthreshold oscillations were evoked when Vm = −57.21 mV ( 0.012 nA). Oscillation frequency averaged over the trace = 44.7 Hz. D: injection of 0.019 nA, which brought Vm to −57.14 mV, evoked repetitive burst discharge. Oscillation frequency during the interval between bursts averaged 44.7 Hz. The interspike interval (ISI) increased during the course of each burst. Firing frequency, based on the last ISI in the burst, was 28.7 Hz. E: further increasing stimulus strength (to 0.045 nA), which brought Vm to −56.74 mV, caused an increase in the duration of bursts and a decrease in the interval between bursts. Oscillation frequency during the interval between bursts did not change. Firing frequency was 28.0 Hz. F: stimulation at 0.063 nA evoked tonic firing at 36.0 Hz. G: intense stimulation (1.62 nA) strongly depolarized the cell (Vm stabilized at −36.67 mV) and evoked high-frequency oscillations (inset, 110.5 Hz) after a considerable delay, but no spiking.

Overall, oscillatory behavior began at Vm = −57.21 mV and ended at −36.67 mV, with a range of about 20 mV within which oscillations were generated. This span of potentials is the “window” for generating subthreshold oscillations. Spikes were generated within a part of this window. The breadth of the window proved to be dependent on parameters of the model cell as described in the following text. The patterns of spiking behavior and the breadth of the oscillatory window were similar to those previously described in live DRG neurons, including in neurons in which voltage-activated K+ channels had been blocked pharmacologically (Amir et al. 2002a). That is, oscillations transitioned into oscillation-triggered bursting and tonic firing and then back to oscillations of gradually declining amplitude, much as in live cells (Fig. 3). In these respects the baseline model was much more realistic than the H-H–based model reported by Amir et al. (2002a).

FIG. 3.

Output of the baseline model (B1B4) resembles that of live dorsal root ganglion (DRG) neurons (A1A4). Tracings of the live cell come from the study of Amir et al. (2002a). A1 and B1: Vr: −61 mV for A1, −57.4 mV for B1. A2 and B2: sustained subthreshold oscillations: Vm = −58 mV for A2, −57.21 mV for B2. Fourier analysis of the traces A2 (live) and B2 (model) are shown in the inset. A3 and B3: oscillation-triggered burst discharge: Vm = −57.0 mV for A3, −57.14 mV for B3. The onset of one burst is shown at higher gain below (spikes are truncated). A4 and B4: repetitive firing (spikes are truncated). Scale bars: 4 mV/100 ms for A1 and A2 and B1 and B2; 16 mV/2 s for A3, top trace, 10 mV/100 ms for A3, bottom trace; 20 mV/500 ms for B3, top trace, 20 mV/100 ms for B3, bottom trace; 10 mV/50 ms for A4; 10 mV/200 ms for B4.

Beyond general patterns of activity, many of the specific parameters of oscillatory and spiking behavior in the baseline model were lifelike (Fig. 3). This includes the general range of oscillation frequencies (45–110 Hz in the simulation and 32–62 Hz in live cells in the presence of gK+ block), the membrane potentials at which oscillations began and ceased (−57 to −37 mV in the simulation; −55 to −22 mV in live cells in the presence of gK+ block), the range of frequencies of spike discharge during bursts and tonic firing (36–99 Hz in the simulation and 25–50 Hz in live cells in the presence of gK+ block), and the fact that in both cases spike amplitude and frequency declined during the course of bursts. There was also some divergence. In the baseline model the transition from oscillations to tonic spiking occurred within a narrower range of membrane potentials than that in live DRG neurons and the model cell did not evoke singlet firing at irregular ISIs, a pattern common in live DRG neurons (Amir et al. 2002a,b). However, singlet discharge could be simulated by the addition of a voltage-sensitive K+ conductance that emulated IK (Fig. 4A). These singlet spikes transitioned to doublets, short bursts, and then tonic firing as the cell was further depolarized (Fig. 4, B and C). The addition of IK significantly reduced the amplitude of the postspike depolarizing afterpotential (DAP; Amir et al. 2002b) assessed using 1-ms pulses. This change reduced the duration of bursts and facilitated singlet spiking.

FIG. 4.

Addition of gK+ to the baseline model allowed singlet and doublet discharge. A: when depolarized from Vr = −59.5 to −54.6 mV (stimulus = 0.5 nA) the cell fired single spikes at low frequency with irregular ISI. B: depolarization to −54.2 mV (stimulus = 0.55 nA) evoked bursts of doublet spikes. C: still greater depolarization (stimulus = 0.9 nA) evoked tonic firing.

Validation of current parameters

Total Na+ current.

Since some of the kinetic parameters used were not based on explicit data, we compared the output of the baseline model with physiological values, particularly checking parameters of the total INa+. We applied to the model the same stimulation protocol that Baker and Bostock (1998) used to define the three INa+ components in live cells. Specifically, the model was held at a prepulse potential of −110 mV and stepped for 200 ms to potentials ranging from −80 to 10 mV in 5-mV increments. Overall, total INa+ behaved very much like that in live cells (Fig. 1, A and EG). The only prominent difference was that INa+late1 inactivated in about 20 ms, more rapidly than observed by Baker and Bostock (1998). With respect to the integrated fast transient Na+ current, INa+fast contributed about 25.5% of the current peak (Vm = −10 mV), INa+late1 about 74%, and INa+late2 only about 0.5%. With respect to the integrated INa+p, measured at 200 ms (Vm = −30 mV), INa+fast contributed about 69%, INa+late1 contributed 18.5%, and INa+late2 contributed the remainder (12.5%). The decaying component of INa+ (2–200 ms, excluding INa+p) was contributed entirely by INa+late1and INa+late2, with relative proportions varying with time.

Current–voltage (I–V) curves.

Also consistent with experimental data the integrated peak INa+ in the simulation started to activate at around −60 mV and reached a maximum at −10 mV (Fig. 1B). This is within the expected voltage range for large DRG cells (Abdulla and Smith 2002; Baker and Bostock 1998; Hong and Wiley 2006; Peng et al. 2002; Shah et al. 2004). The peak transient inward current in the model was −126 nA. Although high, this value is not inconsistent with physiological values for large DRG neurons, especially after axotomy. Baker and Bostock (1997, 1998) did not provide a direct measure of their peak INa+, although they stated that the current amplitude after 200 ms, as much as −1.7 nA, was about 1.1% of the transient peak. Extrapolating, the peak transient current was as much as −155 nA. Current measurements from excised membrane patches were consistent with this value (−80 to −120 nA, derived from Baker and Bostock 1998; Caffrey et al. 1992). Direct measurements from large-diameter DRG neurons in nerve-injured rats showed peak INa+ of up to about −100 nA, reflecting a 1.6-fold increase following axotomy (Abdulla and Smith 2002; also see Rizzo et al. 1995).

Baker and Bostock (1997) adopted an alternative voltage-clamp protocol for isolating INa+p. The membrane was held at a prepulse potential of +20 mV for 200 ms and then stepped to a series of values between −80 and 10 mV, in increments of 5 mV, for 100 ms. Applying this protocol to our baseline model yielded an IV plot, 200 ms after the beginning of the pulse, that represents the integrated INa+p (Fig. 1D). The peak inward current was −1.68 nA at −30 mV, a value similar to that recorded by Baker and Bostock (1997; −1.7 nA, their Fig. 1A).

Comparison to Amir et al. (2002a).

Plotting voltage-clamp records from the model cell of Amir et al. (2002a) shows that the total INa+ also begins to activate at Vm = −60 mV. However, the peak inward current (−63 nA) was obtained at Vm = 0 mV, which is somewhat positive to the present value (−10 mV) and to the potential range seen in live cells (−30 to −10 mV; Abdulla and Smith 2002; Baker and Bostock 1998; Hong and Wiley 2006; Peng et al. 2002; Shah et al. 2004). Plotting the persistent current from the Amir et al. (2002a) model yields a peak of −1.93 nA at Vm = −30 mV. This value is slightly higher than that obtained in the current baseline model (−1.68 nA) and out of the range measured in live cells (−0.4 to −1.7 nA; Baker 2000; Baker and Bostock 1997). Overall, the fundamental difference between the present and the earlier model is the presence of the late currents.

Each Na+ conductance plays a distinctive role in generating ectopic activity

We assessed the functional contribution of the fast and the two late Na+ conductances individually by running simulations in which gNa+max of each was gradually reduced, whereas the others either maintained their original values or were augmented to compensate for the loss of specific INa+ components. Focus was on the ability of the system to generate oscillations and spikes, the membrane potential “window” within which oscillations and spikes were present, and their frequency.

Changing gNa+fast.

Reducing gNa+fastmax from its initial value of 25 mS/cm2 reduced the voltage window, within which sustained oscillations and spiking occurred, and the amplitude of both oscillation sinusoids and spikes at any given membrane potential. This is illustrated in the state diagram in Fig. 5A, which shows the pattern of activity observed over a range of membrane potentials as the value of gNa+fastmax was varied. In the diagrams in Fig. 5 the full repertoire of activity patterns noted earlier—oscillations, spike bursting, and sustained firing (Fig. 2)—is expressed as different shading fills. Observed by reading Fig. 5A horizontally for each value on the y-axis, all of the activity patterns were present until gNa+fastmax fell to <18 mS/cm2. For gNa+fastmax = 17 mS/cm2 oscillations and oscillation-evoked bursts continued to be present, but tonic firing was no longer evoked. When gNa+fastmax was reduced still further, into the range of 13–16 mS/cm2, repetitive bursting also ceased. Depolarizing steps now evoked a brief spike burst at the beginning of the pulse that then decayed into sustained subthreshold oscillations. After the initial spike burst no subsequent spikes were generated irrespective of the stimulation amplitude or duration.

FIG. 5.

Reducing each of the components of gNa+max from values in the baseline model alters the voltage window within which the various patterns of afferent activity are observed. The state diagrams in this figure show sustained subthreshold oscillations as white regions, oscillation-triggered repetitive spike bursts as black regions, and tonic firing as shaded regions. The upper horizontal line is identical in all of the diagrams and represents the behavior of the baseline model. As gNa+max was reduced, the window of membrane potential values that evoked spiking (tonic or burst) and oscillations narrowed. For example, in A, when gNa+max = 18 mS/cm2, oscillations were present for Vm ≃ −58 to −55 mV, bursts for Vm ≃ −55 to −51 mV, tonic firing for Vm ≃ −51 to −48 mV, and oscillations again for Vm ≃ −48 to −43 mV. A, B, and C show the effects of reducing each of the 3 Na+ conductances from their values in the baseline model independently, with the others left unchanged. Values of gNa+max in the baseline model were: 25 mS/cm2 for gNa+fastmax (A), 27 mS/cm2 for gNa+late1max (B), and 0.128 mS/cm2 for gNa+late2max (C). Reducing gNa+fastmax to < 13 mS/cm2 or gNa+late1max to < 17 mS/cm2 eliminated spiking and oscillations. However, reducing gNa+late2max to zero still permitted burst firing and oscillations, albeit within a narrow range of membrane potentials.

Interestingly, the frequency of subthreshold oscillations, about 45 Hz at liminal stimulus amplitude, and the threshold for evoking oscillations (about −57 mV) remained virtually unchanged as gNa+fastmax was reduced from 25 to 13 mS/cm2 (Figs. 5A and 6). When gNa+fastmax was reduced to <13 mS/cm2 the initial spike burst gave way to subthreshold oscillations whose amplitude damped out to zero during the course of the stimulus pulse. Sustained oscillations no longer occurred. Thus the window within which oscillations could be evoked narrowed progressively through the range 25–13 mS/cm2, with its center shifting progressively toward the threshold potential for evoking oscillations (−57 mV). The use of weak stimuli or further reducing gNa+fastmax yielded a progressively shorter initial burst until spiking ceased altogether, yielding damped oscillations at the beginning of the step in the absence of spiking (as in Fig. 2B). We conclude that gNa+fast is important for burst and sustained spike discharge because the most prominent effect of reducing it was to eliminate spiking. It is less critical for generating oscillations.

FIG. 6.

The frequency of subthreshold oscillations is affected by gNa+max . The minimum frequency at which oscillations can be generated increased as gNa+late1max or gNa+late2max was reduced. Reducing gNa+fastmax, in contrast, had no effect on the minimal oscillation frequency. Note that in each case, including gNa+fastmax, stimulus strength must be progressively increased in order to evoke oscillations.

In light of the suppressive effect of reducing gNa+fastmax on spike bursting, we evaluated how this change affects the postspike DAP, an element thought to be critical for maintaining spike bursts (Amir et al. 2002). As anticipated, the amplitude of the DAP that followed single spikes evoked by 1-ms depolarizing steps was highly sensitive to reduction of gNa+fastmax.

Changing gNa+late1.

All of the patterns of activity illustrated in Fig. 2 continued to be present as gNa+late1max was reduced from its baseline value (27 mS/cm2) to 22 mS/cm2, although the potential window in which oscillations and spiking occurred narrowed (Fig. 5B) and spike and oscillation amplitude decreased. Burst duration remained unchanged. Moving into the range 21–17 mS/cm2 spiking ceased, but subthreshold oscillations persisted albeit within a window that gradually narrowed, with its center shifting toward less negative potentials (Fig. 5B). Throughout the range 17–27 mS/cm2, lowering gNa+late1max also increased the oscillation frequency (Fig. 6, triangles). The loss of tonic and burst firing was accompanied by substantial reduction in the amplitude of the DAP that followed single evoked spikes. For gNa+late1max ≤ 16 mS/cm2 strong stimuli induced an initial spike or spike burst followed by damped oscillations, whereas weaker stimuli induced damped oscillations without an initial spike. With gNa+late1max < 10 mS/cm2 damped oscillations were no longer induced even following pulses strong enough to evoke an initial spike.

A substantial proportion of the INa+ at 2–20 ms, the time of the “intermediate” current, is carried by the INa+late1. Reducing INa+late1max by less than half was enough to dramatically reduce the ability of the system to generate both oscillations and spikes. To test whether this effect was due to the simultaneous reduction of the fast transient and persistent components of INa+, we eliminated gNa+late1 and restored the fast transient and persistent currents to baseline values by augmenting gNa+fastmax. This did not restore spiking or oscillations. We conclude from these results that the “intermediate” current is essential for generating oscillations.

Changing gNa+late2.

Reducing gNa+late2max from the baseline value of 0.128 mS/cm2 reduced the already narrow window within which bursting occurred and increased burst duration (Fig. 5C). When gNa+late2max reached 0.095 mS/cm2 bursting ceased entirely. As the cell was further depolarized, oscillations transitioned directly into tonic firing and then back to the oscillatory pattern. With continued reduction in gNa+late2max the window within which these patterns of activity occurred gradually narrowed with its center shifting toward less negative potentials. This behavior persisted until gNa+late2max was reduced to about 0.01 mS/cm2, at which point oscillations were no longer generated at potentials < −42 mV, although they continued to appear at potentials > −40 mV. Rather, suprathreshold stimuli directly evoked tonic spiking. This continued to be the case when gNa+late2max was set to zero. Throughout, lowering gNa+late2max increased the frequency of oscillations (Fig. 6, squares).

With gNa+late2max set to zero, returning the persistent current to baseline by augmenting gNa+fast had little effect, presumably because this maneuver did not restore the intermediate current. In contrast, when the persistent current was returned to baseline by augmenting gNa+late1, which does restore the intermediate current, oscillations could once again be generated by weak stimuli, although burst firing remained absent. We conclude that although the intermediate component of gNa+late2 contributes to oscillatory behavior, the main contribution of this conductance is to interrupt tonic firing and enable bursting.

To gain insight into the mechanism whereby gNa+late2 facilitates bursting, we plotted the status of the inactivation gates of all three conductances in the baseline model (Fig. 7, A and B) and in the absence of repetitive bursting (Fig. 7, C and D). This analysis revealed that, because of its prolonged τh, inactivation of the late and persistent components of INa+late2 summates during the course of a burst, closing slowly. This has the effect of damping the firing. During the subsequent interval inactivation of gNa+late2 slowly fades, permitting oscillations to augment in amplitude and eventually to trigger a new spike burst (Fig. 7, A and B). The effect of the slow inactivation of gNa+late2 is also evident in the presence of a single burst, accomplished by reducing gNa+fastmax (Fig. 7, C and D). Inactivation of gNa+late2 during the burst was more pronounced than that in the baseline model and its recovery was less complete. The reason is that stable oscillations appeared at less negative membrane potentials than those in the baseline model. As a result the membrane oscillates at relatively depolarized potentials, causing the inactivation (h-) gate of gNa+late2 to be closed and thus unable to initiate a subsequent burst. Studying the mesencephalic trigeminal nucleus (MesV), a brain stem DRG homolog, Enomoto et al. (2006) also showed that slowly inactivating components of INa+ contribute significantly to burst termination.

FIG. 7.

The status of inactivation of gNa+late2 controls the cycle of bursting. A: burst firing using parameters of the baseline model. B: the status of the inactivation (h) gates of the 3 Na+ conductances during the course of the bursts shown in A. Due to their relatively small τ, inactivation of gNa+fast and gNa+late1 reciprocate between relatively low (closed) and relatively high (open) values with each spike and oscillation sinusoid. Zero inactivation reflects closed h-gates (the maximally open value is 1.0). Inactivation of gNa+late2, on the other hand, is slow. For this reason its inactivation is integrated over time, rising and falling slowly. The rate of rise and fall determines the burst cycle. C and D: reducing gNa+fastmax to 16 mS/cm2 eliminated repetitive bursting. The step depolarization shown in C evoked a single spike burst followed by sustained subthreshold oscillations. The corresponding plot of h-gate inactivation of the 3 Na+ conductances (D) shows that, as in B, inactivation of gNa+fast and gNa+late1 track each spike and oscillation sinusoid. Inactivation of gNa+late2 falls slowly during the burst, terminating it, and then slowly stabilizes without triggering a second burst.

Role of the fast activating phase of INa+late1 and INa+late2

In our baseline model both gNa+late1 and gNa+late2 activated very rapidly, consistent with Baker and Bostock (1998). The fast activating component of these currents summed with the fast activating gNa+fast to generate the total fast transient INa+ peak. Since the specific parameters of activation of these conductances in vivo are unknown, we examined the effect of slowing activation rate on oscillations and spike electrogenesis. In the baseline model cell integrated INa+ activation was rapid (≤0.11 ms), in accord with measurements from live cells (Kostyuk et al. 1981; Nowycky 1992). Using stimuli of 5-s duration we gradually increased τm of gNa+late1 and gNa+late2 (from 0 in steps of 0.1 ms) while maintaining τm of gNa+fast unchanged. This narrowed the range of membrane potentials at which oscillations and tonic spiking were generated (Fig. 8). The window for burst firing expanded. At τm >0.5 ms spiking, tonic and burst firing ceased altogether and, as τm approached 1.0 ms, oscillations were no longer generated. At membrane potentials that supported oscillations, the oscillation frequency fell as τm increased.

FIG. 8.

The ability of depolarization to evoke subthreshold oscillations, oscillation-triggered bursts, and tonic firing also depends on the activation kinetics of INa+late1 and INa+late2. The 3 basic patterns of afferent activity are shown as state diagrams as in Fig. 5, with subthreshold oscillations indicated as white regions, oscillation-triggered spike bursts as black regions, and tonic firing as shaded regions. Relatively slow activation (τ >0.5 ms) sustains only oscillatory behavior. Accelerated activation (τ <0.5 ms) also permits bursting and tonic firing, over an increasingly wide span of membrane potentials.

Increasing τm of gNa+late1 had the additional effect of reducing the peak of the fast transient INa+, by about 20% when τm was set to 0.5 ms and by about 25% when it was set to 1.0 ms. Restoring the peak to baseline by increasing gNa+fast restored the ability of the system to generate spikes and oscillations. We conclude that in addition to the other contributions of the late Na+ currents, most notably the intermediate current at 2–20 ms, rapid activation of gNa+late1 is also important for oscillations and spike electrogenesis.


Nerve block using local anesthetics generally eliminates neuropathic pain caused by distal ectopic pacemaker sources such as neuromas, at least for the duration of the block. When the block is placed proximal to the DRG (foraminal or spinal block) the likelihood of pain relief is higher still. Ectopic discharge generated in the DRG appears to be triggered by subthreshold membrane potential oscillations generated in primary afferent neurons. Oscillations may also contribute to discharge generated in axons at the nerve injury site (Devor 2006; Kapoor et al. 1997). We investigated which characteristics of INa+ foster oscillations and spiking. In the absence of pharmacological tools capable of independently modulating the various kinetic components of INa+ we used a computational approach. In live DRG cells total INa+ is the integral of fast and delayed currents. Our model included a range of INa+ components: INa+fast, INa+late1, INa+late2, and INa+p, along with ohmic K+ leak. This system generated both oscillatory behavior and oscillation-driven spiking, which resembled recordings from live DRG neurons in experimental models of neuropathic pain.

We found that the fast, delayed, and persistent Na+ currents make different and distinctive contributions to repetitive firing capability. gNa+fast is important for burst and sustained spike discharge. gNa+late1 appears to be particularly important for the generation of subthreshold oscillations. Since sustained burst spiking requires oscillations, the intermediate component of gNa+late1 (2–20 ms) also seems to be the key for enabling spiking in the presence of gNa+fast. gNa+late2 also provides some current in the 2- to 20-ms range, thus facilitating oscillatory behavior, although its contribution here is much smaller than that of gNa+late1. The main contribution of gNa+late2 is its later component (20–200 ms), which is important for burst firing. Our ability to test the contribution of INa+p was limited because this current could not be manipulated independently of the other currents.

The parameter space we studied most closely was channel inactivation, but we also looked at channel activation. In our baseline model τm of activation for the fast current was about 0.1 ms in the voltage range typical of subthreshold oscillations. Physiological values for activation τm of the late currents are not known with much certainty, but for gNa+late1, at least, τm appears to be very fast (Baker and Bostock 1998). In our simulations, oscillations and spiking proved to be fairly sensitive to the onset kinetics of the late Na+ conductances, failing for τm >1.0 ms. This observation implies that in vivo the late currents have a rapid onset (<0.5 ms). Thus if our model is correct they are unlikely to depend solely on Na+ channel types with slow onset kinetics (see following text).

Passive leak was the second type of conductance present. In contrast to the model of Amir et al. (2002a) the reversal potential of ILeak was not at the K+ reversal potential (EK+), but more depolarized. This was because setting ELeak at EK+ yielded an unrealistically hyperpolarized Vr. Voltage-sensitive K+ channels were not included because they are not required for oscillations or oscillation-evoked spiking in DRG neurons, although they do affect oscillation and spike discharge frequency (Amir et al. 2002a). However, just to be sure, we incorporated them in a few simulations. The effect on oscillatory behavior was minimal, although they did enable single-spike firing.

The basic features of in vivo ectopic spike patterning were well captured by the current simulation (Fig. 3). However, not all features were present. For example, in many live cells Vm shifts several millivolts in the hyperpolarizing direction during the course of a burst, contributing to burst termination. This behavior, at least partly attributable to Ca2+ entry during spiking and activation of a Ca2+-activated K+ conductance (Amir and Devor 1997), was not observed because neither conductance was included in the baseline model. Bursting was nonetheless present, indicating that other mechanisms may also contribute to the arrest of tonic firing (Fig. 7). In vivo recordings from neuroma afferents sometimes show exotic firing patterns also not observed in our simulations. Examples are regularly cycling modulations in spike frequency or repeated cycling between staccato and prolonged bursts (see Fig. 3.2 in Devor 1989). These patterns no doubt also reflect additional, idiosyncratic conductances present in particular cell types.

Role of delayed INa+ in repetitive firing

Delayed and persistent inward current carried by TTX-resistant (TTX-R) Na+ channel α-subunits is thought to contribute to hyperexcitability of afferent neurons, particularly in the event of tissue inflammation (Elliott 1997; Herzog et al. 2001; Roza et al. 2003; Waxman 2002). Delayed INa+ has also been described at nodes of Ranvier of myelinated peripheral nerve fibers (Dubois and Bergman 1975) and subthreshold oscillations have been observed in intraaxonal recordings from such fibers (Kapoor et al. 1997). Finally, large-diameter primary sensory neurons in MesV also show delayed TTX-S INa+. In these cells there are two components, one with τ of several seconds and another that is truly persistent (noninactivating). In vitro observations and simulations suggest that both contribute to cell resonance, subthreshold oscillations, and repetitive firing (Enomoto et al. 2006; Wu et al. 2001, 2005).

Delayed Na+ currents are also known to contribute to repetitive firing in other neuronal types (Crill 1996). For example, Pennartz et al. (1997) described in suprachiasmatic nucleus neurons a rapidly activating, slowly inactivating TTX-S current resembling gNa+late1. This contributes significantly to the slope of the depolarizing ramp following spikes and thus to spontaneous firing rate. Delayed INa+ may also facilitate electrogenesis in the CNS through other mechanisms, such as enhancement of oscillatory behavior (Schindler et al. 2006; Spampanato and Mody 2007). Using slow ramp depolarizations (100 mV/s) Kononenko et al. (2004a,b) revealed in suprachiasmatic neurons a TTX-S current with inactivation similar to that of gNa+late2 (τ ≃ 50–250 ms at its maximum). This current was a necessary supplement to the fast INa+ in the generation of sustained firing. Likewise, a delayed TTX-S current similar to that of gNa+late2 contributes to repetitive firing in neurons of the entorhinal (Agrawal et al. 2001) and frontal agranular cortices (Urbani and Belluzzi 2000). Finally, in spinal motorneurons depolarized with slow ramps Kuo et al. (2006) found that the persistent component of the total INa+ imparts an acceleration to the depolarization required to overcome membrane accommodation and to permit repetitive firing.

Source of delayed INa+

The identity of the ion channel(s) underlying late and persistent Na+ currents in large DRG neurons has not yet been determined. Indeed, these currents might well be generated by a repertoire of gating states of channels that also generate the fast transient INa+ (Baker and Bostock 1998; Crill 1996). The delayed currents are almost certainly not generated by the Nav1.9 Na+ channel because this channel is TTX-R, whereas the currents in question are TTX-S. In any event Nav1.9 is not expressed in large DRG neurons (Cummins et al. 1999). Likewise, the Nav1.8 channel generates delayed and persistent INa+, but these too are TTX-R (Renganathan et al. 2000). Both TTX-R channels have slow onset kinetics incompatible with oscillations and spiking (Fig. 8) and both are down-regulated following axotomy. Nav1.5 generates a delayed INa+ in cardiac myocytes (Noble and Noble 2006) and it has recently been identified in adult DRG neurons (Kerr et al. 2007). At both locations, however, Nav1.5 is TTX-R.

Nav1.1 and Nav1.3, both TTX-S Na+ channels, generate a fast transient Na+ current when coexpressed with accessory β1 and β2 subunits. In addition, however, they can generate delayed INa+ (for certain Nav1.1 alleles τslow = 5–10 ms (Lossin et al. 2003; Montegazza et al. 2005); for human Nav1.3 expressed in HEK cells, τslow >150 ms (Chen et al. 2000)). Nav1.1 is expressed in normal adult DRG neurons, but Nav1.3 (and Nav1.4) is not and thus cannot contribute to the oscillations that occur in (a small fraction of) intact DRG neurons. However, Nav1.3 is up-regulated in DRG neurons after axotomy (Waxman et al. 1994). It is also up-regulated in dorsal horn neurons after spinal cord injury and in the latter case up-regulation has been shown to contribute to both enhanced firing during ramp depolarization, hyperexcitability, and pain behavior (Hains et al. 2003; Lampert et al. 2006). Nav1.2 and Nav1.6 are also TTX-S, but when expressed in isolation they generate little if any current with intermediate kinetics (Shirahata et al. 2006; Weisner et al. 2006). Nav1.7 is TTX-S and generates fast transient and delayed INa+, but essentially no INa+p, at least when expressed in isolation (Cummins et al. 2004; Herzog et al. 2003). Overall, these considerations point to Nav1.1, Nav1.3, and Nav1.7 as the most likely contributors of the key delayed TTX-S INa+ component in large-diameter DRG neurons.

Mitigating against Nav1.1 and Nav1.7 as essential carriers of delayed INa+ is the fact that these transcripts are down-regulated in DRG neurons following axotomy (Chung et al. 2003). Axotomy, however, is known to strongly enhance oscillatory behavior and spike electrogenesis in DRG neurons (Liu C-N et al. 2000a). This factor enhances the candidacy of Nav1.3, which is up-regulated following axotomy (Waxman et al. 1994). However, deletion of Nav1.3 does not apparently interfere with repetitive firing (Nassar et al. 2006) or oscillatory behavior in large-diameter DRG neurons (our unpublished observations). Thus, no single Nav type appears to be the key to ectopic electrogenesis in these neurons.

Gene regulation notwithstanding, little is known about the effect of axotomy on actual delayed and persistent Na+ currents, although Abdulla and Smith (2002) reported an increase in INa+ at intermediate latencies (10 ms) in large-diameter DRG neurons. Interestingly, this change occurred in animals that showed pain behavior (autotomy) after nerve injury, but not in nerve-injured animals without pain behavior. A variety of mediators present in injured nerves are able to alter whole cell INa+ kinetics, but it is uncertain whether this reflects a shift in the relative populations of the specific contributing Na+ channel types present or a change in the kinetics of individual channels (Bevan and Storey 2002).

Potential implications for pain and analgesic drugs

Our main conclusion is that the “intermediate” component of INa+ (2–20 ms) is a key contributor to sustained firing capability, despite the fact that the maximal amplitude of this current is only about 2% of the peak transient INa+ that generates the spike itself. In principle it should be possible to develop agents that act on the delayed component of the integrated whole cell INa+ by enhancing the inactivation kinetics of a variety of channel types rather than by blocking a particular channel. An example is riluzole (2-amino-6-trifluoromethoxy benzothiazole), a neuroprotective agent used for the treatment of amyotrophic lateral sclerosis. Riluzole blocks delayed INa+ including INa+p (Song et al. 1997; Urbani and Belluzzi 2000; Wu et al. 2005). As expected, it also suppresses subthreshold membrane potential oscillations and bursting in CNS neurons (Reboreda et al. 2003; Wu et al. 2005). Unfortunately, it is brain permeant and, at clinically tolerated doses, does not provide useful pain relief (Galer et al. 2000). Yang et al. (2009) reported recently that gabapentin suppresses subthreshold oscillations and repetitive firing in (live) medium to large DRG neurons by suppressing delayed Na+ current. Indeed, this may be the drug's primary mode of action in the relief of neuropathic pain (Devor 2009b). The effect was attributed to suppression of persistent INa+. True INa+p, however, was not adequately dissociated from other delayed Na+ currents. Suppression of an INa+ with the intermediate kinetics of INa+late1 might in fact be the effective target. An agent that selectively suppresses the intermediate current in afferent neurons, especially if it were excluded from the CNS, might have improved efficacy and specificity as a reliever of neuropathic pain.


This work was supported by the United States–Israel Binational Science Foundation, the National Institute for Psychobiology in Israel, and the Hebrew University Center for Research on Pain.


We thank C.-N. Liu for permission to use the live cell data in Fig. 3.


  1. Abdulla and Smith 2002.
  2. Agrawal et al. 2001.
  3. Amir et al. 2006.
  4. Amir and Devor 1997.
  5. Amir et al. 2002a.
  6. Amir et al. 1999.
  7. Amir et al. 2002b.
  8. Baker and Bostock 1997.
  9. Baker 2000.
  10. Baker and Bostock 1998.
  11. Bevan and Storey 2002.
  12. Boucher and McMahon 2001.
  13. Caffrey et al. 1992.
  14. Campbell et al. 1988.
  15. Chen et al. 2000.
  16. Chung et al. 2003.
  17. Crill 1996.
  18. Cummins et al. 1999.
  19. Cummins et al. 2004.
  20. Devor 1989.
  21. Devor 2006.
  22. Devor 2009a.
  23. Devor 1975.
  24. Dubois and Bergman 1975.
  25. Elliott 1997.
  26. Enomoto et al. 2006.
  27. Everill et al. 1998.
  28. Fedulova et al. 1998.
  29. Galer et al. 2000.
  30. Herzog et al. 2003.
  31. Herzog et al. 2001.
  32. Hines and Carnevale 1997.
  33. Hong and Wiley 2006.
  34. Kapoor et al. 1997.
  35. Kerr et al. 2007.
  36. Koltzenburg et al. 1994.
  37. Kononenko et al. 2004a.
  38. Kononenko et al. 2004b.
  39. Kostyuk et al. 1981.
  40. Kovalsky et al. 2008.
  41. Kuo et al. 2006.
  42. Liu et al. 2000a.
  43. Liu et al. 2000b.
  44. Lossin et al. 2003.
  45. Malcangio et al. 2000.
  46. Mantegazza et al. 2005.
  47. Nassar et al. 2006.
  48. Noble and Noble 2006.
  49. Noguchi et al. 1995.
  50. Nowycky 1992.
  51. Peng et al. 2002.
  52. Pennartz et al. 1997.
  53. Reboreda et al. 2003.
  54. Renganathan et al. 2000.
  55. Rizzo et al. 1995.
  56. Rola et al. 2003.
  57. Roza et al. 2003.
  58. Schindler et al. 2006.
  59. Scroggs et al. 1994.
  60. Shah et al. 2004.
  61. Shirahata et al. 2006.
  62. Song et al. 1997.
  63. Spampanato and Mody 2007.
  64. Tal et al. 2006.
  65. Urbani and Belluzzi 2000.
  66. Waxman 2002.
  67. Waxman et al. 1999.
  68. Waxman et al. 1994.
  69. Weissner et al. 2006.
  70. Woolf and Salter 2000.
  71. Wu et al. 2005.
  72. Wu et al. 2001.
  73. Xing et al. 2001.
  74. Yang et al. 2009.
  75. Yoshimura et al. 1996.
  76. Yu et al. 2005.
View Abstract