Presynaptic Resurgent Na⁺ Currents Sculpt the Action Potential Waveform and Increase Firing Reliability at a CNS Nerve Terminal

Presynaptic DAP and AP timing

Direct recordings from the rat calyx of Held revealed a prominent DAP following a single action potential spike (Fig. 1a) (see also Borst et al., 1995). This presynaptic DAP could be triggered by afferent fiber stimulation in the brainstem midline (Fig. 1a) or by a brief suprathreshold depolarizing current injection in a calyx terminal with a short axon stump (axon length in these cases was measured after filling the calyx with Alexa 555 and varied from 20 to 200 μm, n = 5 calyces) (Fig. 1b; supplemental Fig. 1, available at www.jneurosci.org as supplemental material). A long axon, extending back to the midline, was thus not necessary for generating the DAP, suggesting that the DAP was generated near the calyx nerve terminal.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

The presynaptic DAP at the calyx of Held. a, Afferent fiber stimulation evokes a single presynaptic AP spike followed by a depolarizing DAP in the calyx of Held (P13). Inset, Same trace on an expanded time and voltage scale showing the rise of the DAP and the fAHP. b, A brief suprathreshold depolarizing current injection (500 pA for 2 ms) evokes an AP followed by a DAP in a calyx terminal with a severed axon. c, A pair of APs in response to afferent fiber stimulation with 4 ms interval (black trace) and then combined with a short hyperpolarizing current injection (−100 pA) to the calyx to remove the DAP (red traces, arrowhead). Note the delay in the second AP (arrow) or failure to fire in the absence of the DAP.

The presynaptic DAP showed a clear rising phase that depolarized the calyx from the downward peak of the fast after-hyperpolarization (fAHP) to near −60 mV (Fig. 1a, inset). However, the membrane potential at the peak of the DAP rarely surpassed −60 mV and thus was always safely below spike threshold. The DAP amplitude measured from the resting potential to the peak of the DAP depended on the resting potential and was larger at more hyperpolarized potentials (supplemental Fig. 2, available at www.jneurosci.org as supplemental material). However, the DAP amplitude measured from the peak of fAHP to the peak of DAP had a fairly invariant amplitude of 7.7 ± 0.2 mV (n = 22 calyces) at adjusted resting membrane potentials ranging from −90 to −64 mV (supplemental Fig. 2, available at www.jneurosci.org as supplemental material). The 10 to 90% rise time of the DAP upstroke was dependent on the adjusted resting membrane potential (1.0 ± 0.1 ms at −90 mV, 1.7 ± 0.3 ms at −80 mV, 2.2 ± 0.2 ms at −70 mV, and 2.7 ± 0.3 ms at −65 mV; n = 3, 3, 6, and 5, respectively, for P13; p < 0.01, one-way ANOVA).

The presynaptic DAP of a first AP influenced the exact timing of a second AP spike. For a pair of APs with a 4 ms interval (corresponding to 250 Hz activity, well within the physiological range of the calyx of Held), a hyperpolarizing current injection of ∼−100 pA removed the DAP and increased the latency of the second AP (latency, measured from the stimulus artifact to the peak of the AP, increased from 0.8 ± 0.06 to 0.9 ± 0.08 ms, n = 5, p < 0.05) (Fig. 1c). In addition, suppressing the DAP also reduced reliability as the failure rate for the second AP was increased. The increase in AP failures may be mediated by the back-propagation of the hyperpolarization down the axon (Paradiso and Wu, 2009). However, we also performed the same experiment on calyces with severed axons of different lengths using a brief depolarizing current injection. Similar to the afferent fiber stimulation case, the complete elimination of the DAP via a step-like hyperpolarizing current injection significantly increased the second AP latency (n = 5; data not shown). These results thus indicate the DAP is generated near the calyx terminal, likely at the axonal heminode that contains a large density of Na⁺ channels (Leão et al., 2005), and also that the DAP influences the exact timing and reliability of subsequent APs.

Presynaptic DAP: developmental changes

To study the role of DAP during synaptic maturation, we characterized its kinetics at three postnatal ages: shortly after formation of the calyx terminal at P5–P6 (Kandler and Friauf, 1993; Rodríguez-Contreras et al., 2008); just before the onset of hearing at P9–P10; and at P15–P16, when the synapse assumes an adult-like morphology and rat pups respond to sound stimulation (Kandler and Friauf, 1993; Sonntag et al., 2009). Measured from the downward peak of the fAHP to the peak of the DAP (Fig. 2a, inset), the amplitude of DAP rise was 2.4 ± 0.4 mV at P5–P6, 6.7 ± 0.7 mV at P9–P10, and 13.8 ± 0.6 mV at P15–P16 (n = 8, 9, and 4, respectively; p < 0.001, one-way ANOVA). For the same ages listed above, the single exponential decay time constants (τ_decay) of DAP were 79 ± 3.6, 30 ± 3.1, and 15 ± 2.4 ms, respectively (n = 8, 9, and 4, respectively; p < 0.001). This decrease in τ_decay closely paralleled a decrease in τ_ms, which were 50.5 ± 3.7, 22.4 ± 1.5, and 11.4 ± 1.6 ms, respectively, for the same age groups (n = 7, 5, and 5, respectively; p < 0.001) (Fig. 2b,c). The faster membrane time constant in more mature calyces was due in part to reduced input resistance, which was 375 ± 25.1, 322 ± 29.2, and 191 ± 16.2 MΩ, respectively, at P5–P6, P9–P10, and P15–P16 (n = 7, 5, and 5, respectively; p < 0.001). The 10 to 90% rise time of the DAP was also shorter in more mature rats (1.8 ± 0.13 ms in P15–P16 vs 3.0 ± 0.15 ms in P5–P6 calyces; resting potential, −80 mV; n = 4; p < 0.01). The DAP thus has a significantly faster rise and decay phase for more mature calyces.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

The DAP kinetics during postnatal development. a, A presynaptic AP spike and DAP in the calyx of Held from three developmental stages: P6 (red), P10 (blue), and P16 (black). The AP spike was elicited by afferent fiber stimulation. Resting potential was −80 mV (dashed line). Inset, Same traces on an expanded time and voltage scale. Note the changes in rise and decay times of the DAP at different developments ages. b, Membrane potential changes induced by a −50 pA hyperpolarizing current injection in calyces of different ages indicate that input resistance decreases with age. The P15 trace is also shown scaled to the P6 amplitude to facilitate comparison of decay times (gray trace). c, Summary of the single exponential τ_decay (left) of the DAP and the τ_m (right) for P5–P6, P9–P10, and P15–P16 calyces (p < 0.001, one-way ANOVA). d, A single AP at P10 calyx in the absence (control; black trace) or in the presence of TEA (1 mm; red trace); 1 mm TEA inhibited the fAHP in the calyx terminals. e, Action potential trains in response to 100 Hz afferent fiber stimulation in P5 (red) and P15 calyces (black). The P5 AP train adapted (black arrow). Right, Expanded figure of boxed area of AP trains superimposed. The P5 calyx train of APs showed a large depolarized plateau (large arrowhead) due to the summation of the slow DAPs, and its AP train had spike failures (small vertical arrowheads), whereas the P15 calyx AP train had no failures.

What is the mechanism that governs the fAHP peak? We first note that the fAHP becomes progressively larger (more hyperpolarized) with increasing postnatal age (−69 ± 0.7, −74 ± 0.4, and −78 ± 0.5 mV; n = 8, 9, and 8, respectively, for P5–P6, P9–P10, and P15–P16; p < 0.001) (Fig. 2a). Previous studies indicate that TEA-sensitive Kv3 channels undergo a developmental increase that contributes to shorter AP waveforms and more reliable spiking at high frequencies (Wang and Kaczmarek, 1998; Nakamura and Takahashi, 2007). To test the involvement of Kv3 channels in the generation of the fAHP, we applied 1 mm TEA and found the peak fAHP for P14 calyces was significantly decreased by 7.1 ± 0.3 mV (n = 3) from −75 to −67.7 mV (Fig. 2d), indicating that Kv3 currents are a major determinant of the fAHP peak (Wang and Kaczmarek, 1998).

How does the DAP affect AP firing during a high-frequency train of spikes? We examined this by recording AP trains in P5–P6 and P15–P16 calyces (Fig. 2e). At P5–P6, a long AP train elicited by afferent fiber stimulation at 100 Hz had a large depolarizing plateau, and AP spikes became progressively smaller in amplitude during the train (Fig. 2e, arrow). Action potentials eventually failed in the latter part of the train. However, at P15–P16 calyx AP trains had almost no depolarizing plateau and little spike attenuation during the 100 Hz train (Fig. 2e). Superposition of the two spike trains at higher time resolution shows that the slow P5–P6 DAP summates to produce a large depolarizing plateau (Fig. 2e). This plateau will cause a progressive inactivation of Na⁺ channels, which explains the attenuation of the spike amplitudes and the eventual failures in spiking (Klyachko et al., 2001; Leão and von Gersdorff, 2002). These developmental changes in the depolarizing plateau may be caused by an increased expression of Kv2-type channels in more mature calyces (Johnston et al., 2008). The fast decay of the DAP in more mature P15–P16 calyces is thus an important developmental change that helps to avoid DAP summation during a train and the subsequent failures in spiking seen in immature calyces.

Presynaptic Na⁺ currents generate the DAP

We next examined the conductances that underlie the DAP in calyx terminals. We found that a low concentration of TTX (20 nm) significantly decreased the amplitude of DAP from 9.0 ± 1.1 to 1.0 ± 0.7 mV with only a slight reduction of AP spike amplitude by 10% (n = 5) (Fig. 3a,f). For a pair of APs, TTX (20 nm) also had a stronger effect on the second AP latency than the first AP (Fig. 3a,c), indicating that the TTX-sensitive DAP is critical to the timing of subsequent AP. To further test that a Na⁺ current is the main conductance generating the DAP, we blocked Na⁺ channels with 1 μm TTX, which completely inhibits Na⁺ spikes, and then injected a large current (1 to 1.5 nA for 1.5 ms) to depolarize the calyx terminal up to ∼+30 mV, thus mimicking the peak potential reached during an AP. After the membrane potential reached +30 mV, we observed a fast repolarization of the membrane potential, but no DAP was generated (n = 3; P13 calyx) (Fig. 3b, black trace). This rapid decay is likely due to activation of high-threshold Kv3 currents (Dodson and Forsythe, 2004; Kaczmarek et al., 2005). The smaller current injections (0.5 nA), which depolarized the calyx to a peak value <0 mV, resulted in a much slower decay of membrane potential back to the resting potential (Fig. 3b, red trace). In this case, decay was fitted by a double exponential function consisting of a fast component (1.2 ± 0.2 ms) and a slow component (15 ± 0.7 ms; 17% of amplitude; n = 3). The fast component may be due to activation of low-threshold Kv1.1/1.2 channels (Dodson et al., 2003), and the slow component is probably due to the passive discharge of the membrane potential, since this slow τ_decay is comparable to the τ_m (11.4 ms at P15–P16) (Fig. 2a–c). Together, these results indicate that a TTX-sensitive Na⁺ current, and not depolarization per se, generates the DAP. We conclude that the calyx of Held AP waveform is generated by a primary Na⁺ influx event during the upstroke of the spike and a secondary Na⁺ influx event that causes the DAP. This secondary Na⁺ influx may also have important consequences for energy consumption in axons and nerve terminals (Alle et al., 2009; Carter and Bean, 2009).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Presynaptic Na⁺ current generates the rising phase of the DAP. a, Spikes evoked by afferent fiber stimulation in the absence (control; black trace) and presence of external TTX (20 nm, gray traces during onset of application and red traces end of application). For a pair of APs, with 4 ms stimulus interval, removing the DAP (arrowhead) delayed the latency to the peak of the second AP (red arrow). b, After the complete inhibition of voltage-dependent Na⁺ currents with 1 μm TTX, brief current injections [1.5 ms duration, 0.5 nA (red trace) or 1 nA (black trace)] depolarized the calyx to 0 mV (red) or +30 mV (black). Note the absence of a DAP after the larger spike, whereas the smaller spike has a monotonically decaying DAP with no hump. c, P13 calyx AP recorded with a pipette solution containing a low concentration of QX-314 (0.2 mm), a Na⁺ channel blocker. The black trace was obtained within a few seconds after break-in, and the red trace was recorded after 7 min of whole-cell recording. After 3 min of QX-314 dialysis, the rising phase of the DAP was significantly inhibited, although spike amplitude was little changed (gray trace). d, The calyx of Held was dialyzed for 5 min with QX-314 (0.3 mm), and the stunted rise of the DAP was recorded (red trace). The recording pipette was then removed, and the same calyx was patched again using normal internal pipette solution without QX-314. The normal rise and amplitude of the AP spike and DAP recovered after 7 min of dialysis, which removed QX-314 from the calyx through the patch pipette. Note the broader AP spike with QX-314 (inset). The removal of QX-314 also recovered the second AP, which failed to fire in the presence of QX-314. e, Veratridine (an activator of Na⁺ channels; 1 μm) increased significantly the amplitude of the DAP after 1 min (red trace) and elicited repetitive spontaneous spiking in response to a single afferent fiber shock (blue trace; recorded 10 s after red trace). f, Summary of DAP amplitude in the presence of TTX (20–50 nm, n = 5), QX-314 (0.2–0.3 mm, n = 7), and veratridine (1 μm, blue triangles, n = 3). Data were analyzed with paired t test (**p < 0.01; ***p < 0.001).

To test whether local Na⁺ spike after currents close to the calyx and its axonal heminode play a role in DAP generation, we recorded using a low concentration of the permanently charged lidocaine analog QX-314 in the pipette solution (0.2–0.3 mm). Shortly after whole-cell break-in, calyx spikes were followed by a DAP with a fast rising phase (Fig. 3c, black trace). During dialysis of QX-314, the amplitude of the DAP decreased markedly. After 5 min of QX-314 dialysis, the DAP, measured from the fAHP to the DAP peak, decreased from 8.0 ± 1.1 to 0.2 ± 0.1 mV (n = 8) (Fig. 3c,f), while the amplitude of the AP was only slightly decreased to 94 ± 6% (n = 8). In addition, the internal application of QX-314 significantly hyperpolarized the calyx of Held resting membrane potential by −4 mV (from −80 ± 0.6 to −84 ± 0.7 mV; n = 4; p < 0.01), suggesting a tonic Na⁺ channel activation at rest (Huang and Trussell, 2008). Moreover, this inhibition of the DAP with QX-314 was reversible. Patching the same calyx terminal a second time using a pipette without QX-314 restored the original DAP amplitude, and the adjusted resting membrane potential moved back to the original −80 mV. In addition, a second spike, which had failed to fire with QX-314 pretreatment, was revived after washing out of QX-314 (n = 3) (Fig. 3d). Na⁺ currents close to the calyx terminal thus generate the DAP.

By contrast, application of the Na⁺ channel activator veratridine (1 μm) significantly increased the DAP amplitude from 6 ± 1.8 to 14 ± 1.3 mV and slowed the decay of the DAP (Fig. 3e,f) (n = 3; p < 0.01). The progressive increase in the DAP amplitude with veratridine eventually resulted in repetitive spontaneous spiking in response to a single afferent fiber stimulation (Fig. 3e, blue trace). Veratridine increases the open time of Na⁺ channels, slows their inactivation rate (Garber and Miller, 1987), and shifts the activation curve to the left (Ulbricht, 1998). This suggests that the decay rate of the DAP may be influenced by the inactivation rate of the underlying Na⁺ channels that generate the DAP. Furthermore, an abnormally large DAP amplitude produces aberrant spiking and results in hyperexcitability at the calyx of Held (Dodson and Forsythe, 2004).

I_NaR and I_NaP at the calyx of Held

To identify the specific type of Na⁺ current responsible for the DAP, we first performed voltage-clamp recordings of calyx Na⁺ currents. A step-depolarization from the holding potential of −90 to +30 mV generated a large and rapidly inactivating I_NaT (Leão et al., 2005). Upon repolarization to potentials ranging from −20 to −70 mV, we observed a voltage- and time-dependent current with biophysical properties similar to the I_NaR present in Purkinje cells, cerebellar granule cells, vestibular neurons, and MNTB principal cells (Fig. 4a,b) (Raman and Bean, 1997; Leão et al., 2006; Magistretti et al., 2006; Gittis and du Lac, 2008). The current–voltage (I-V) relationship for I_NaR exhibited a peak of −408 ± 63 pA at ∼−30 mV in P9–P12 calyces (n = 16; no liquid-junction potential correction) (Fig. 4c). The decay kinetics of I_NaR was voltage dependent, with a single exponential τ_decay ranging from 5 to 40 ms (Fig. 4b,d). In addition, we also observed a I_NaP that remained after the decay of I_NaR (Fig. 4a) (Huang and Trussell, 2008). We also obtained outside-out patches from the calyx terminal that contained well clamped I_NaTs (n = 3; data not shown) (Leão et al., 2005), but these did not show I_NaR or I_NaP currents (see also Scott et al., 2010).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

I_NaR in the calyx of Held. a, I_NaT and I_NaR were elicited in a P10 calyx terminal by a 10 ms step depolarization to +30 mV, followed by step repolarizations to −70, −60, −50 (gray), −40 (green), −30 (blue), and −20 mV (red) as indicated in the voltage-clamp protocol (top). Note that after the decay of I_NaR, an I_NaP remains activated. b, Individual current traces for repolarizations to −70 and −30 mV are shown together with single exponential decay fits. The I_NaT peaks have been truncated. Note the fast deactivating tail currents (arrowheads) before the full activation of I_NaR. c, The current–voltage dependence (I-V curve) for peak I_NaR for P9–P12 calyces. The peak I_NaR was measured from the end of I_NaT to the peak of I_NaR. d, The τ_decay of I_NaR as a function of test membrane potentials. I_NaR decayed more slowly at more depolarized potentials.

Tetrodotoxin at high concentration (1 μm) blocked all three currents described above (I_NaT, I_NaR, and I_NaP), confirming their identity as Na⁺ currents (Fig. 5a,b). In contrast, internal QX-314 (0.3 mm) blocked the persistent and resurgent currents but spared 60% of the transient current after 10 min of dialysis of the calyx (supplemental Fig. 3a, available at www.jneurosci.org as supplemental material). Bath application of veratridine (5 μm) increased the peak amplitude of I_NaR from −273 ± 44 pA to −400 ± 62 pA (n = 4) and slowed the inactivation decay of I_NaR, whereas the peak of I_NaT was not significantly changed (−6.0 ± 0.5 nA versus −5.6 ± 0.8 nA; n = 4; p > 0.1; paired t test) (supplemental Fig. 3b, available at www.jneurosci.org as supplemental material). Low concentrations of TTX (50 nm) also blocked the I_NaR and I_NaP of more mature P15–P16 calyces (n = 3) (supplemental Fig. 4, available at www.jneurosci.org as supplemental material). Together, these results indicate that the calyx I_NaR and/or I_NaP, both of which are blocked by low doses of TTX and QX-314, and are enhanced by veratridine, underlie the upstroke phase of the DAP.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

The calyx I_NaR increases during postnatal development. a, I_NaT, I_NaP, and I_NaR were evoked at a P5 calyx by a short depolarization to +30 mV followed by a step repolarization to −30 mV. TTX (1 μm) completely inhibited I_NaT, I_NaP, and I_NaR in the calyx of Held. The bottom trace is the TTX-sensitive current obtained by subtraction. The large I_NaT peak has been truncated. b, I_NaT, I_NaP, and I_NaR at a P10 calyx. TTX (1 μm) completely inhibited I_NaP, I_NaR, and I_NaT. c, I-V curve for peak I_NaR at P5–P6 (black) and P10–P12 calyces (red). All sodium currents were leak subtracted (p/5) and TTX subtracted. There is a significant increase in I_NaR between the two age groups at the peak of I_NaP near −30 mV. d, I-V curve for the I_NaP for P5–P6 calyces (black) and P10–P12 calyces (red). There is a slight increase in I_NaP between the two different age groups at the peak of I_NaP near −10 mV, but this was not statistically significant (p = 0.38, Student's t test). Membrane potentials were not corrected for a 10 mV liquid junction potential. Persistent sodium current thus starts to activate at about −80 mV and may contribute to the resting membrane potential.

Developmental increase of calyx I_NaR

The rising phase of the DAP became significantly faster from P5 to P15 (Fig. 2a), a period when Na⁺ currents undergo developmental changes at the calyx of Held (Leão et al., 2005). To help distinguish the individual contributions of I_NaR and I_NaP in generating the DAP, we examined changes in I_NaR and I_NaP amplitudes during the first two postnatal weeks. We found that I_NaR increased in amplitude during the developmental period (Fig. 5a,b). The I-V curve of I_NaR showed an inverted-bell shape at P5–P6 and P10–P12, and a peak value at −30 mV (Fig. 5c). Immature calyces (P5–P6) exhibited significantly smaller inward peak I_NaR measured at −30 mV (−86 ± 10.6 pA, n = 3) than P10–P12 calyces (−313 ± 71 pA, n = 5; p < 0.02, Student's t test). An I_NaP remained as a noninactivating current after full decay of I_NaR, and it slightly increased from −335 ± 103 pA at P5–P6 (n = 4) to −533 ± 88 pA at P10–P12, but this increase was not significant (I_NaP measured at −10 mV; n = 5; p = 0.38, Student's t test) (Fig. 5c). I_NaR and I_NaP had different voltage dependences, since I_NaR and I_NaP displayed two distinct peaks in their I-V curves (at ∼−30 and −10 mV, respectively) (Fig. 5c,d). We thus conclude that I_NaR and I_NaP are distinct sodium currents. Furthermore, the faster rise of DAP amplitudes observed with more mature calyces correlates well with the large increase in I_NaR during development.

Computer modeling of the calyx DAP

The pharmacological compounds available to us are not selective for I_NaR or I_NaP. To obtain further insight into the Na⁺ conductance that generates the DAP, we thus performed computer simulations to test for the specific role of persistent and resurgent Na⁺ currents. Conductances included in our modeling were passive (leak) conductance; HT and LT potassium conductances; and transient, persistent, and resurgent sodium currents (Fig. 6, black lines) (Dodson and Forsythe, 2004; Kaczmarek et al., 2005). In this manner, we could investigate the effect of eliminating persistent current (Fig. 6, blue lines) or eliminating resurgent current (Fig. 6, red lines) on the DAP. Full details of the model and its parameters are given in Materials and Methods and in supplemental material (available at www.jneurosci.org as supplemental material).

• Download figure
• Open in new tab
• Download powerpoint

Figure 6.

Computer model of the role of I_NaR and I_NaP in the generation of the DAP. a, Model Na⁺ currents in four scenarios: control (black); after removal of persistent current (blue); after removal of resurgent current (red); and after removal of resurgent current with two times higher persistent current (green). b, Model action potentials in the same scenarios as described in a. Note that the DAP is normal without persistent current, but the DAP is absent without resurgent current. The action potential waveform is also broader without resurgent current. c, Effect of four different densities of persistent current on the DAP (relative to that measured in the calyx of Held). Black trace, 2.4 × I_NaP; blue trace, 3.2 × I_NaP; red trace, 4.2 × I_NaP; and green trace, 5.6 × I_NaP. Top, Persistent Na⁺ tail currents evoked by steps from 0 to −80 mV. Bottom, Resulting DAP waveforms. The deactivation time constant (τ) was fixed at τ = 0.24 ms. d, Same as in c except without resurgent Na⁺ current. No DAP hump was generated. e, The deactivation time constant of the persistent Na⁺ currents from 0 to −80 mV was varied (black, 0.2 ms; blue, 0.3 ms; red, 0.5 ms; green, 0.9 ms; and purple,1.6 ms). f, Model action potentials with only persistent Na⁺ currents in the same scenarios as described in e. g, h, Representative experimental data traces of deactivation of calyx Na⁺ currents from −30 to −80 mV in the P16 calyx (g) and P10 calyx (h). Single exponential fits are shown to the data (red). The more mature P16 calyx had a faster τ.

Model sodium currents (Fig. 6a) were generated by turning off the other conductances and performing linear leak subtraction to remove capacitive currents. The simulated voltage-clamp protocol to elicit mixed resurgent and persistent Na⁺ current was the same as that used for the recordings shown in Figure 4a: a step depolarization to +30 mV for 5 ms that generated transient Na⁺ current followed by a repolarizing step from −20 to −70 mV in 10 mV steps. Simulated I_NaR showed a peak in the bell-shaped I-V relationship near −40 mV, and the decay of I_NaR also showed the same voltage dependency as the experimental data (simulation results not shown). Importantly, the rate of inactivation of the transient Na⁺ current was significantly faster with an I_NaR present in the simulations (Fig. 6a), as expected from the extra effect of the inactivation binding particle used to generate I_NaR (Raman and Bean, 2001). To generate an I-V relationship for the persistent Na⁺ current, we used a slow voltage ramp (20 mV/s) from a potential of −90 mV. This I-V relationship showed current activation at −70 mV (no liquid junction potential correction) and a peak near −20 to −10 mV, as we observed experimentally (Fig. 5d; simulation results not shown).

Action potentials were generated in the computer model by current injection (4 nA for 0.2 ms) into the distal end of the axon. Four Na⁺ current scenarios were studied: normal resurgent and persistent current (black traces), resurgent current without persistent current (blue traces), no resurgent current with normal persistent current (solid red traces), and no resurgent current with two times higher persistent current (broken green traces). Without I_NaR, we obtained a calyx action potential with no DAP (Fig. 6b). In addition, note that the half-width of the action potential was broader without I_NaR (as is also the case experimentally with 0.3 mm QX-314) (Fig. 3c,d). Removal of I_NaP had very little effect on the DAP (blue traces), and even increasing I_NaP two times over experimentally recorded values did not rescue the DAP in the absence of I_NaR. These simulations thus capture the main basic features of our previous experimental results, and they show that I_NaP by itself does not generate the DAP. They thus represent a proof of principle for our main hypothesis that I_NaR generates the DAP.

We next explored the effect of varying the amplitude of I_NaP (from 2.4 to 5.6 times the measured values) on the fAHP or DAP (Fig. 6c), while the time constant of deactivation was kept constant at 0.24 ms (close to the value measured in P16 calyces). In the presence of I_NaR, a 2.4-fold to 3.2-fold higher density of I_NaP increased the peak amplitude of the DAP (measured relative to the fAHP) only slightly, whereas a 4.2-fold to 5.6-fold increase in I_NaP was necessary to produce a significantly larger DAP with a prolonged time course (Fig. 6c). However, without I_NaR we observed that increasing persistent current removed the fAHP but did not produce a rebounding DAP “hump,” as observed experimentally (Fig. 6d). We conclude that persistent current in the calyx of Held is unlikely to generate the DAP, although it may modulate its amplitude and kinetics.

The persistent current used in the simulation shown in Figure 6a–d has fast deactivation kinetics (0.24 ms at −80 mV). However, we reasoned that an I_NaP with sufficiently slow deactivation kinetics could contribute to the DAP since during membrane repolarization following an AP the increased driving force for Na⁺ could result in a depolarizing drive similar to that caused by resurgent Na⁺ current. To explore this, we tested several different models in which the time constant for deactivation at −80 mV was varied from 0.2 to 1.6 ms (Fig. 6e). As the kinetics of persistent current gating were slowed, the fAHP was reduced, and for very slow time constants approaching 1 ms a DAP with a hump was indeed generated (Fig. 6f). However, in the calyx of Held, the experimentally measured persistent Na⁺ current deactivated with a fast time constant of 0.27 ± 0.07 ms (P16; n = 4) or 0.6 ± 0.1 ms (P10; n = 3) (Fig. 6g–h). We therefore conclude that the I_NaP in the calyx of Held deactivates too rapidly to generate a significant DAP.

β4 subunit peptides induce I_NaR in immature calyces

To further test for a causal relationship between I_NaR and the DAP, we next used a portion of the β4 sodium channel subunit to induce I_NaR in immature calyces. A leading model for how I_NaR can arise is through the voltage-dependent binding of a blocking particle (Raman and Bean, 2001), which may originate from the β4 subunit to the Na_V1.6 channel (Grieco et al., 2005). According to this model, when the β4 subunit particle binds it both blocks current flow and also prevents the normal “classical” inactivation process from occurring. On repolarization, and dissociation of the β4 subunit particle, the channel conducts again until normal deactivation or closure occurs. Interestingly, very high levels of β4 subunit mRNA are found throughout the adult rat brainstem (Yu et al., 2003), but the level of β4 subunit protein in the developing rat calyx of Held is unknown. If protein levels follow mRNA, an increased expression of β4 subunits during early postnatal development could explain the observed developmental increases in I_NaR and the DAP amplitude. Unfortunately, due to a lack of reliable antibodies, we have not been able to confirm a developmental increase in protein levels of the β4 subunit in rat MNTB.

We thus tested whether a peptide that mimics the putative action of β4 subunits on Na⁺ channels could generate I_NaR in immature P5–P6 calyces, which have little resurgent Na⁺ current (Fig. 7a,b). In recordings without peptide, the size of I_NaR in P5 calyces was relatively small and stable during whole-cell recording (control) (Fig. 7c). When 100 μm β4_peptide (KKLITFILKKTREK, Grieco et al., 2005) (Fig. 7b, red trace) was included in the pipette solution, I_NaR increased 10-fold, from 23 ± 8 pA at 2 min to 227 ± 53 pA after 8 min of intracellular dialysis of the peptide (n = 5, p < 0.001) (Fig. 7c); whereas, the increase of I_NaP with β4_peptide was not significant (from 32 ± 11 pA to 53 ± 20 pA after 8 min of dialysis; n = 5, p = 0.11). In the presence of the β4_peptide, the I-V relation of I_NaR was bell shaped and similar to that observed in more mature calyces (n = 8) (Fig. 7d).

• Download figure
• Open in new tab
• Download powerpoint

Figure 7.

Induction of I_NaR by dialysis of a β4 peptide into P5–P6 calyces of Held. a, An immature calyx terminal (P5) expresses a very small amplitude I_NaR, which was recorded with repolarizing step pulses from +30 to −20 mV. Inset, The P5 calyx of Held (arrowhead) and attached axon were filled with Alexa-555 via the patch pipette for morphological identification. b, I_NaR was recorded with control internal solution (black trace) or with an internal solution that contained β4_peptide (100 μm; amino acid sequence from the β4 subunit of Na⁺ channels, red trace) or with a peptide with a scrambled sequence of β4_peptide (blue trace). Inset, Expanded traces of the I_NaR shown from the shadowed box. c, The peak amplitude of I_NaR at −30 mV progressively increased during whole-cell dialysis of β4_peptide (100 μm; colored dots) in five individual P5–P6 calyx recordings. However, no significant change in peak amplitude of I_NaR was detected with control internal solution (black dots; n = 5). d, Current–voltage relationship of I_NaR recorded without β4_peptide (control, black; n = 5), with β4_peptide (red; n = 8), or with scrambled β4_peptide (blue; n = 5) for P5–P6 calyx terminals.

To confirm that the effects produced by the β4_peptide depended on the specific amino acid sequence, we tested the effect of a scrambled peptide on I_NaR in which charge was conserved but the clusters of hydrophobic and positively charged residues were disrupted (β4_scramble: KIKIRFKTKTLELK) (Grieco et al., 2005) (Fig. 7b,d, blue trace). In the presence of β4_scramble, the I_NaR did not significantly increase during 15 min of dialysis, and its I-V relationship remained similar to control (n = 5) (Fig. 7d). These data indicate that the β4_peptide can significantly increase I_NaR in immature P5–P6 calyces of Held.

β4 subunit peptides increase the DAP amplitude and firing reliability

Next, we investigated the effect of β4_peptide dialysis on the AP waveform of immature P5–P6 calyces. AP spikes were evoked once every 10 s by afferent fiber stimulation, and β4_peptide (100 μm) was dialyzed into the calyx via the patch pipette. The amplitude of the DAP in the presence of β4_peptide was significantly increased from 2.2 ± 0.2 to 4.0 ± 0.5 mV after 10–15 min of dialysis (P5–P6; n = 12, p < 0.01; DAP amplitude was measured from the fAHP) (Fig. 8a1). The potential reached at the peak of the DAP measured from the resting membrane potential was also increased by intracellular β4_peptide dialysis from −67 ± 0.4 to −63 ± 0.8 mV (n = 12, p < 0.01) (see supplemental Fig. 5a, available at www.jneurosci.org as supplemental material). The β4_peptide depolarized the average calyx adjusted resting membrane potential from −80 ± 0.4 to −76 ± 0.5 mV and significantly shortened the spike latency from 1.6 ± 0.2 to 1.3 ± 0.1 ms (n = 12, p < 0.001) (Fig. 8b). In addition, the 10 to 90% rise time of the spike was reduced from 328 ± 16.9 to 273 ± 14.3 μs (n = 9; p < 0.001) (Fig. 8b). Similar effects on AP spike latency and rise time are also observed in postsynaptic MNTB neurons after the dynamic clamp incorporation of I_NaR into normal MNTB neurons (Leão et al., 2006).

• Download figure
• Open in new tab
• Download powerpoint

Figure 8.

Dialysis of β4_peptide into P5–P6 calyces increases the DAP amplitude and improves reliability of presynaptic firing. a1, Presynaptic AP spikes and DAP evoked by afferent fiber stimulation without β4_peptide (black) and with β4_peptide (100 μm, red) in the same P5 calyx terminal. The rising phase of the spikes are superimposed. The β4_peptide depolarized the calyx and increased the DAP. a2, Current injection (+15 pA) depolarized the calyx from −80 to −75 mV in the control, but this small depolarization had no effect on the amplitude of the DAP. b, Internal dialysis of β4_peptide depolarized the calyx and decreased spike latency, which was measured as the time from stimulus artifact (S.A.; arrow) to the peak of the spike. c, Spike trains were elicited by afferent fiber stimulation at 133 Hz (7.5 ms stimulus interval, 50 stimuli) in a P5 calyx without β4_peptide (black). The same calyx was then patch-clamped with a pipette containing β4_peptide (red). In control, the P5 calyx exhibited spike failures at 133 Hz (arrows), but dialysis of β4_peptide allowed the same calyx to fire spikes without any failures at 133 Hz (red). Inset, Spikes in the two shadowed boxes are superimposed and expanded in time scale. d, In control, the P5 calyx could not follow 200 Hz stimulus train (black trace). Depolarization from −80 to −70 mV by current injection (30 pA) had no effect on these failures (blue traces). The same calyx with β4_peptide had a reduced rate of failure at 200 Hz (red trace). Inset, Spikes in the shadowed boxes are superimposed. Note the shorter latency with β4_peptide (red arrowhead).

The increase in DAP amplitude by the β4_peptide was not due to membrane depolarization because depolarizing the calyx by 4 mV via current injection in control calyces lacking β4_peptide had no effect on the amplitude of DAP (n = 3) (Fig. 8a2). In three recordings, we also observed the diffusion of the β4_peptide using tetramethylrhodamine-tagged β4_peptide and found that it diffused along the axon for at least 100–200 μm in the 7–10 min after break-in to whole-cell recording mode (supplemental Fig. 5b, available at www.jneurosci.org as supplemental material). This indicates that the β4_peptide reaches the whole extent of the heminode region of the axon where Na_V1.6 channels are primarily located (Leão et al., 2005). To test whether the effects of the β4_peptide are due to changes in K⁺ current induced by the peptide, we recorded calyx K⁺ currents with pipettes containing β4_peptide. Insertion of β4_peptide into P5 calyces had no significant effect on the amplitude of voltage-dependent K⁺ currents after 10 min of β4_peptide dialysis (K⁺ currents at 0 mV were 1.9 ± 0.46 nA in control and 1.8 ± 0.59 nA with β4_peptide; n = 3, p > 0.05; data not shown). Thus, we conclude that an increase of I_NaR by β4_peptide dialysis enhances the amplitude of the DAP and reduces the AP latency.

We next examined the consequences of increased I_NaR and DAP for high-frequency firing by recording from P5–P6 calyces before and after β4_peptide dialysis. To reduce calyx-to-calyx variability (Sonntag et al., 2009), we first recorded spike trains (100–200 Hz) in control without β4_peptide. We then removed the patch pipette and repatched the same calyx with a pipette containing 100 μm β4_peptide. P5 calyces fired trains of 50 spikes without failure at 100 Hz but showed several failures at 200 Hz. The success rate (number of spikes per total of 50 stimuli) was smaller in control than after dialysis with β4_peptide. At 133 Hz (stimulus interval, 7.5 ms), several control P5 calyces (6 of 10) exhibited spike failures during the late phase of the stimulus train. After dialysis with β4_peptide, these same calyces showed no failures at 133 Hz (Fig. 8c, black and red traces, respectively). The success rate at 133 Hz increased from 92 ± 3.5% to 100% with β4_peptide (n = 10; p < 0.05).

To control for the possibility that the depolarization of ∼4 mV caused by the peptide could have effects on the calyx firing fidelity, we performed a control experiment without β4_peptide. Most P5 calyces could not follow 200 Hz stimulus trains for 250 ms, showing a large number of AP failures (e.g., 25 failures of 50 stimuli) (Fig. 8d). Constant current injection (∼30 pA) depolarized the calyx resting potential from −80 to −70 mV; nevertheless, the calyx still had the same number of AP failures at 200 Hz. In the same calyx, however, the dialysis of 100 μm β4_peptide reduced AP failures in the early phase of the train. The success rate increased from 59 ± 3.3% to 71 ± 3.9% with β4_peptide at 200 Hz (n = 10; p < 0.01) (Fig. 8d). Together, our results suggest that I_NaR induced by β4_peptide increases peak DAP amplitude, reduces the spike latency, and leads to an increase in firing reliability at high stimulation frequencies.