Resurgent sodium current promotes action potential firing in the avian auditory brainstem

Key points Auditory brainstem neurons of all vertebrates fire phase‐locked action potentials (APs) at high rates with remarkable fidelity, a process controlled by specialized anatomical and biophysical properties. This is especially true in the avian nucleus magnocellularis (NM) – the analogue of the mammalian anteroventral cochlear nucleus. In addition to high voltage‐activated potassium (KHVA) channels, we report, using whole cell physiology and modelling, that resurgent sodium current (I NaR) of sodium channels (NaV) is equally important and operates synergistically with KHVA channels to enable rapid AP firing in NM. Anatomically, we detected strong NaV1.6 expression near hearing maturation, which was less distinct during hearing development despite functional evidence of I NaR, suggesting that multiple NaV channel subtypes may contribute to I NaR. We conclude that I NaR plays an important role in regulating rapid AP firing for NM neurons, a property that may be evolutionarily conserved for functions related to similar avian and mammalian hearing. Abstract Auditory brainstem neurons are functionally primed to fire action potentials (APs) at markedly high rates in order to rapidly encode the acoustic information of sound. This specialization is critical for survival and the comprehension of behaviourally relevant communication functions, including sound localization and distinguishing speech from noise. Here, we investigated underlying ion channel mechanisms essential for high‐rate AP firing in neurons of the chicken nucleus magnocellularis (NM) – the avian analogue of bushy cells of the mammalian anteroventral cochlear nucleus. In addition to the established function of high voltage‐activated potassium channels, we found that resurgent sodium current (I NaR) plays a role in regulating rapid firing activity of late‐developing (embryonic (E) days 19–21) NM neurons. I NaR of late‐developing NM neurons showed similar properties to mammalian neurons in that its unique mechanism of an ‘open channel block state’ facilitated the recovery and increased the availability of sodium (NaV) channels after depolarization. Using a computational model of NM neurons, we demonstrated that removal of I NaR reduced high‐rate AP firing. We found weak I NaR during a prehearing period (E11–12), which transformed to resemble late‐developing I NaR properties around hearing onset (E14–16). Anatomically, we detected strong NaV1.6 expression near maturation, which became increasingly less distinct at hearing onset and prehearing periods, suggesting that multiple NaV channel subtypes may contribute to I NaR during development. We conclude that I NaR plays an important role in regulating rapid AP firing for NM neurons, a property that may be evolutionarily conserved for functions related to similar avian and mammalian hearing.


Introduction
Voltage-dependent sodium (Na V ) channels play a critical role in generation of action potentials (APs), the firing pattern of which is fundamental for information processing in the nervous system (Eijkelkamp et al. 2012). A unique property of some Na V channels is a resurgent sodium current (I NaR ) (Raman & Bean, 1997). I NaR is the result of a voltage-dependent open channel block of the Na V α-subunit by an intracellular particle that competes with the classic inactivation gate (i.e. the cytoplasmic linker between the III and IV domains of the α-subunit) during depolarization. Channels that are in the 'blocked' state, but not the 'classic-inactivated' state, generate I NaR during AP repolarization because the open channel blocker loses affinity for the α-subunit at repolarized membrane potentials. As a result, I NaR provides a small depolarizing drive to the membrane near the AP threshold -a property that promotes repetitive neuronal firing. In addition, activation of the open channel blocker facilitates the recovery of Na V channels after depolarization and increases Na V channel availability by competing against classic inactivation (Raman & Bean, 2001). This specific blocker has been identified as the β4-subunit in cerebellar Purkinje cells (Grieco et al. 2005;Aman et al. 2009), whereas multiple α-subunits (e.g. Na V 1.2, 1.5, 1.6 and 1.7) have been shown to be capable of carrying I NaR (Rush et al. 2005;Jarecki et al. 2010). As a result of unique features of the open channel blocker, studies show that I NaR plays an important role in promoting high rates of AP firing in numerous mammalian neurons, underlying their highly specialized information processing patterns (Lewis & Raman, 2014).
In the auditory brainstem of all vertebrates, neurons are known for their remarkable ability to fire APs at high rates (Oertel, 1997). This specialization is important for encoding sound in an ultrafast and temporally precise manner, properties essential for survival and comprehension of behaviourally relevant communication functions, including sound localization and signal extraction in complex listening environments (Shannon et al. 1995;Anderson et al. 2010;Grothe et al. 2010). An exemplar of this ability is well studied in the avian auditory brainstem. Neurons in nucleus magnocellularis (NM)the avian analogue of bushy cells of the mammalian anteroventral cochlear nucleus -are able to phase-lock to inputs up to 1000 Hz with high fidelity, as demonstrated by single-unit recordings from barn owls (Koppl & Carr, 1997). Similarly, in the mammalian auditory brainstem, the calyx of Held is a pivotal synapse involved in the microsecond precision of sound localization computations. Here, fast Na V channel kinetics (Leao et al. 2005) and abundant expression of high voltage-activated potassium channels (K HVA ) ) along with I NaR present at both pre-and postsynaptic sites (Leao et al. 2006;Kim et al. 2010) promote high rates of AP firing. Based on the aforementioned properties of I NaR in mammals, as well as the established and shared functional phenotypes between species (Carr & Soares, 2002), we hypothesized that this unique current also plays an important role in shaping the fast firing pattern observed for avian NM neurons.
In this study we tested whether NM neurons present with I NaR . By investigating the function of I NaR in AP firing rates of chicken NM neurons both experimentally and computationally, we found that NM neurons have robust I NaR that significantly increases Na V availability immediately after depolarization and facilitates Na V channel recovery. Removal of I NaR in a model NM neuron undermines its ability to fire at high rates. We also examined the maturation of I NaR relative to hearing onset and the potential Na V channel subtype that carries this current in developing NM using immunocytochemistry. We conclude that I NaR plays an important role in regulating rapid AP firing for NM neurons, a property that may be evolutionarily conserved for functions related to similar avian and mammalian hearing.
Association and is appropriate for the species, stages of development and size of the embryos. For electrophysiological experiments, eggs were obtained from Sunnyside Farms, Inc. (Beaver Dam, WI, USA) and incubated in the central auditory physiology laboratory at Northwestern University. For immunocytochemical experiments, eggs were obtained from Charles River Laboratories (Wilmington, MA, USA) and incubated in a Florida State University vivarium. The authors understand and conform to the principles and regulations described by The Journal of Physiology (Grundy, 2015).
To study the stimulus frequency-firing pattern of NM neurons and the properties and function of I NaR , brainstem slices were taken from chickens at embryonic days (E) 19-21. Frequency-firing pattern is defined as the calculated firing probability (see below) of APs as a function of stimulus frequency. At this age range, near-mature hearing ability is established (Saunders et al. 1973;Rebillard & Rubel, 1981;Jones et al. 2006) and NM neurons have obtained mature-like morphology and physiology (Jhaveri & Morest, 1982b). To study the development of I NaR , along with the development of frequency-firing pattern, chickens at the age of E11-12 and E14-16 were included in the current study, corresponding to before and during hearing onset, respectively, while chickens at E19-21 were considered as after hearing onset (Jones et al. 2006). The brainstem was dissected and isolated in ice-cold (ß0°C) oxygenated low-Ca 2+ , high-Mg 2+ modified artificial cerebral spinal fluid (ACSF) containing the following (in mM): 130 NaCl, 2.5 KCl, 1.25 NaH 2 PO 4 , 26 NaHCO 3 , 3 MgCl 2 , 1 CaCl 2 and 10 glucose. ACSF was continuously bubbled with a mixture of 95% O 2 -5% CO 2 (pH 7.4, osmolarity 295-310 mosmol l −1 ). The brainstem was blocked coronally, affixed to the stage of a vibrating blade microtome slicing chamber (Ted Pella, Inc., Redding, CA) and submerged in ice-cold ACSF. Bilaterally symmetrical coronal slices were made (200-300 μm thick) and approximately three to seven slices (depending on age) containing NM were taken from caudal to rostral, roughly representing the low-to-high frequency regions, respectively. All neurons reported here were obtained from the rostral one-half of the entire nucleus, roughly representing the mid-to high-frequency regions of NM.
Slices were collected in a custom holding chamber and allowed to equilibrate for 1 h at room temperature in normal ACSF containing the following (in mM): 130 NaCl, 2.5 KCl, 1.25 NaH 2 PO 4 , 26 NaHCO 3 , 1 MgCl 2 , 3 CaCl 2 and 10 glucose. Normal ACSF was continuously bubbled with a mixture of 95% O 2 -5% CO 2 (pH 7.4, osmolarity 295-310 mosmol l −1 ). Slices were transferred to a recording chamber mounted on an Olympus BX51W1 (Center Valley, PA, USA) microscope for electrophysiological experiments. The microscope was equipped with a CCD camera, ×60 water-immersion objective and infrared differential interference contrast optics. The recording chamber was superfused continuously with a motorized pump (Welco, Tokyo, Japan) at room temperature (monitored continuously with a bath emerged thermometer at ß24°C, Warner Instruments, Hamden, CT, USA) in oxygenated normal ACSF at a rate of 1.5-2 ml min −1 .
Whole cell electrophysiology. Voltage-clamp and current-clamp experiments were performed using an Axon Multiclamp 700B amplifier (Molecular Devices, Sunnyvale, CA, USA). Patch pipettes were pulled to a tip diameter of 1-2 μm using a P-97 Flaming-Brown micropipette puller (Sutter Instrument, Novato, CA, USA) and had resistances ranging from 3 to 6 M . For voltage-clamp experiments of isolated Na V currents, the internal solution was caesium-based and contained the following (in mM): 150 CsCl, 10 NaCl, 0.2 EGTA and 10 HEPES, pH adjusted to 7.3-7.4 with CsOH. The Cs + -based internal solution was used to block K V currents and reduce space-clamp issues. The junction potential was ß−3 mV. Series resistance was compensated for by ß80% in all voltage-clamp recordings. For current-clamp experiments, the internal solution was potassium-based and contained the following (in mM): 105 potassium gluconate, 35 KCl, 1 MgCl 2 , 10 HEPES-K, 5 EGTA, 4 4-Mg 2 ATP, and 0.3 4-Tris 2 GTP, pH adjusted to 7.3-7.4 with KOH. The junction potential was ß−10 mV. Data in both voltage clamp and current clamp experiments were not corrected for junction potentials.
Pipettes were visually guided to NM and neurons were identified and distinguished from surrounding tissue based on cell morphology, known structure and location of the nucleus within the slice. After a gigaohm seal was attained, membrane patches were ruptured and neurons were first held in the voltage clamp mode of the wholecell configuration. A small hyperpolarizing (−1 mV, 30 ms) voltage command was presented to monitor whole-cell parameters (i.e. cell membrane capacitance, series resistance and input resistance). NM neurons were included in the data analysis only if they had series resistances <15 M . Raw data were low-pass filtered at 5 kHz and digitized at 50 kHz using a Digidata 1440A (Molecular Devices).
All experiments were conducted in the presence of the GABA A receptor antagonist picrotoxin (100 μM). Synaptic glutamate transmission was continuously blocked using DL-2-amino-5-phosphonopentanoic acid (DL-APV, 100 μM, an NMDA receptor antagonist) and 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX, 20 μM, an AMPA receptor antagonist). In current-clamp experiments, tetraethylammonium (TEA, 1 mM) was used to block K V 3-containing channels. Current commands of square pulse trains (duration 1 s) were injected into the soma of E19-21 NM neurons before and during TEA J Physiol 596.3 application. The current strength was 1 nA and individual square pulse width was 2 ms. Square pulse trains were applied at varying frequencies: 50,100,150,200,250 and 300 Hz. In order to profile the frequency-firing pattern for NM neurons, firing probability per square pulse (for simplicity, 'firing probability') was calculated as the number of APs divided by the total number of square pulses and plotted as a function of stimulus frequency. To study the development of frequency-firing pattern, the same current commands of square pulse trains were used for NM neurons at E14-16. For NM neurons at E11-12, however, individual square pulse width was extended to 5 ms, because neurons at this age are presented with significantly wider AP half-width (average = 4.6 ms) than the other two age groups (Hong et al. 2016). In addition, square pulse trains were applied at different frequencies: 10,30,50,70,100,120 and 150 Hz. In voltage-clamp experiments, isolated Na V currents were recorded with bath application of TEA (3 mM), 4-AP (30 μM) and CdCl 2 (0.2 mM) to block potassium and calcium channels. Classic voltage-clamp protocols to elicit I NaR were used for NM neurons (Raman & Bean, 1997;see Results). Briefly, repolarizations at membrane voltages from −70 to 0 mV (in a step of 5 mV) were applied after a depolarizing conditioning step. Three levels of conditioning step were used, +30, 0 and −30 mV, along with two durations, 10 and 100 ms. I NaR properties were characterized separately under these six experimental conditions. Afterwards, tetrodotoxin (TTX, 1 μM) was bath applied and the same voltage-clamp protocols were repeated during TTX application. Data reported in this study were obtained by subtracting the TTX-insensitive current traces from the control traces. Capacitive currents generated during voltage-clamp recordings were blanked or reduced offline. Data analysis. Recording protocols were written and run using Clampex acquisition and Clampfit analysis software (v. 10.3; Molecular Devices). Statistical analyses and graphing protocols were performed using Prism (v. 7.0b; GraphPad Software, La Jolla, CA, USA) and MATLAB (v. R2014b; The MathWorks, Natick, MA, USA) software. Student's t test or analysis of variance (ANOVA) with post hoc Bonferroni adjusted t test was used to determine significance. The standard for significant differences was defined as P < 0.05. Numerical values in the text are reported as means ± standard deviation (SD). Numerical values in Table 3 are reported as means ± standard error of the mean (SEM). Error bars in all figures represent SEM.
Reagents. All bath-applied drugs except for TTX were allowed to perfuse through the recording chamber for ß10 min before subsequent recordings. TTX application was allowed for 3-5 min before subsequent recordings. DL-APV, CNQX and all other salts and chemicals were

Computational modelling
Based on NM models previously described in detail elsewhere (Howard & Rubel, 2010;Lu et al. 2017), a singlecompartment computational model was constructed using NEURON 7.1 (Table 1) (Hines & Carnevale, 1997). This model contains currents mediated by low and high voltage-activated potassium channels (K LVA and K HVA , respectively), Na V and passive leak channels. Modelling schemes were identical to those described in Lu et al. (2017) for all membrane currents except for Na V currents ( Table 2). The Hodgkin-Huxley style formalism, previously employed to model Na V current, was replaced with a Markovian 13-state Na V -channel model, which generates the transient (I NaT ), persistent (I NaP ) and resurgent (I NaR ) current components simultaneously ( Fig. 1) (Khaliq et al. 2003). We adopted a modified version described previously (Akemann & Knopfel, 2006), which includes the Q 10 parameter to adjust the Na V current to match the experimental recording temperature of 24°C. Q 10 is a measure of the degree to which a biological process depends on temperature. It is defined as the ratio between the rate of a biological process at two temperatures separated by 10°C. In the context of ion channels, it can be applied to the temperature dependence of the rate of channel opening and closing and to the dependence of maximum channel conductance on temperature.
To remove the I NaR component, the rate constant for the O→OB transition, ε, was set to zero. This resulted in considerable slowing of I NaT decay (see Fig. 6B, blue trace). To modify I NaR without affecting I NaT and I NaP , we applied the method described previously (Magistretti et al. 2006). Kinetic constants were modified as follows: ε was set to 0; O on and O off were increased to 2.15 ms −1 and  0.01433 ms −1 to restore I NaT and I NaP , respectively (see Fig. 6B and C).

Frequency-firing pattern of NM neurons
Auditory brainstem neurons are able to fire APs at high rates (Oertel, 1997;Trussell, 1997Trussell, , 1999. Therefore, we first tested whether late-developing NM neurons (E19-21) follow varying frequency rates with high fidelity. Square pulse current trains at frequencies from 50 to 300 Hz were injected into the soma of NM neurons. Firing probability at each frequency was defined as the average number of APs per square pulse (see Methods). The majority of NM neurons (11 out of 12 neurons) were able to follow square pulse trains of 150 Hz in a one-to-one manner (i.e. high-fidelity, Fig. 2Aa and D). Firing probability dropped to ß0.6 at 200 Hz when obvious failures of spike generation are observed during the stimulus ( Fig. 2Ca and D, asterisks). When increasing the stimulus frequency  Table 2. J Physiol 596.3 beyond 200 Hz, firing probability was further reduced below 0.5 and in response to 300 Hz stimulation, NM neurons only fired a single onset AP (data not shown). Therefore, NM neurons are capable of firing APs at rates up to 200 Hz with good fidelity, and this ability of frequency firing is similar to other auditory brainstem neurons (Gao & Lu, 2008).
Our current-clamp recordings were conducted at room temperature (ß24°C). Chickens have a body temperature of ß42°C. Increasing recording temperature changes the physiology of neurons dramatically (Kushmerick et al. 2006). Our previous study recorded APs at near-physiological temperature (ß35°C) and we showed a significant improvement of AP kinetics for NM neurons (Hong et al. 2016). For example, AP half-width was reduced by ß30% and fall rate increased by ß40% when increasing the temperature to ß35°C. Based on these observations, one would expect that NM neurons recorded at higher temperatures could follow stimulations higher than 200 Hz. Indeed, a recent study using current-clamp recordings at 38-40°C reveals that NM neurons are able to follow square pulse trains up to 500 Hz with good fidelity, whereas the firing probability drops dramatically at 667 Hz (Kuba et al. 2015).
K V 3-containing channels, which belong to the family of K HVA channels, promote the AP repolarizing phase and have been shown to shape the frequency-firing pattern of many neurons Johnston et al. 2010;Hong et al. 2016). Based on previous findings, we hypothesized that K v 3-containing channels play a major role in regulating the rapid firing capability NM neurons. To test this hypothesis, we bath applied TEA (1 mM) to block K V 3-containing channels. Representative responses to 150 Hz square pulse trains before and during TEA application are shown in Fig. 2Aa and b. The enlargement of two overlaid APs from the same neuron is shown in Fig. 2B. During TEA application, the majority of NM neurons (10 out of 12 neurons) were still able to follow pulse trains of 150 Hz with 100% fidelity, despite a significant widening of APs (P = 0.0034, Fig. 2B and D). Correspondingly, no changes in firing probability at 50 or 100 Hz were observed with blockade of K V 3-containing channels (Fig. 2D). In contrast, we observed reduction in firing probability when increasing the stimulus frequency to 200 or 250 Hz ( Fig. 2Ca and b and D). In response to 200 Hz stimulation, firing probability was reduced by ß19% on average. Although this reduction is statistically significant (P = 0.0011), nearly half of the recorded neurons (5 out of 12 neurons) showed a reduction of only ß10% or less in firing probability. These observations with TEA application indicate that K V 3-containing channels contribute partially to higher frequency firing of NM neurons and thus gave rise to an important question: what other factor(s) regulate the ability of NM neurons to follow inputs up to 200 Hz with relatively high fidelity?

Resurgent Na V current of NM neurons
A number of previous studies suggest that I NaR shapes the rapid firing and burst generation in many mammalian neurons (Lewis & Raman, 2014). Therefore, we speculated that NM neurons have I NaR , which is likely one of the factors that regulate the frequency-firing pattern of NM neurons. We first examined whether I NaR is present in NM by using classic voltage-clamp protocols (Raman & Bean, 1997  −90 mV before giving a depolarizing conditioning step to +30 mV (duration = 10 ms). A transient Na V current (I NaT ) was evoked by the conditioning step (Fig. 3Aa, arrowhead). Afterwards, neurons were repolarized to varying membrane voltages from −70 mV to 0 mV ( step = 5 mV). Inward-going I NaR , which could be eliminated by TTX application, was elicited by repolarization and its amplitude was plotted as a function of repolarizing membrane voltage (Fig. 3Aa, small arrow, and Ac). The I-V curve of I NaR for NM neurons peaked at −40 mV and presented with the typical 'V' shape that largely resembled that of I NaR reported in mammalian neurons (Lewis & Raman, 2014). The maximal amplitude of I NaR is usually less than 1 nA for NM neurons recorded from slice preparation, which is much smaller than the amplitude of I NaT (Fig. 3Ac). This property is also similar to that observed in other auditory brainstem neurons (Leao et al. 2006;Kim et al. 2010). Therefore, by using the conditioning step of +30 mV at 10 ms, NM neurons show robust generation of I NaR that is present with similar amplitude and voltage dependence properties to mammalian neurons. In mammalian neurons, the I NaR amplitude is dependent on the duration and level of the conditioning step (Raman & Bean, 2001). This is due to two competing inactivation mechanisms of Na V channels: open channel block induced by an intracellular particle and the classic inactivation induced by the 'gate' (i.e. cytoplasmic III-IV linker) within the α-subunit. As demonstrated in cerebellar Purkinje cells, the majority of Na V channels will be open channel blocked if neurons are exposed to a short and more positive depolarization (e.g. an AP). The unbinding of open channel block at moderately negative membrane voltage (e.g. −40 mV) under this condition generates large I NaR . With longer and less positive depolarization, Na V channels will mainly adopt the classic inactivation gate that requires hyperpolarization of membrane voltage to be released, and the generation of I NaR will be minimal. In order to investigate whether I NaR in NM has similar properties, we applied conditioning steps with different durations and levels and compared the I NaR amplitude among the conditions. For example, the duration of +30 mV depolarization was extended from 10 to 100 ms. Fig. 3Ab shows representative current traces in response to the prolonged conditioning step. The generation of I NaR can still be observed but with a smaller amplitude. Indeed, comparison between the two I-V curves revealed a significant reduction in current amplitude when the conditioning step was elongated to 100 ms (Fig. 3Ac). Interestingly, we also observed a shift in peak membrane voltage towards the positive direction by ß5 mV. Next, we changed the level of the conditioning step to 0 mV and −30 mV while maintaining the same duration (i.e. 10 ms). Fig. 3Ba and Ca shows representative current traces in response to conditioning steps of 0 mV and −30 mV, respectively. By comparing the I-V curves shown in Fig. 3Ac-Cc, we observed a decreasing trend (albeit not significant) in I NaR amplitude when the conditioning step was switched from positive to negative voltages. The voltage dependence did not change with different levels of conditioning steps (i.e. all peaked at −40 mV). Finally, we applied a long duration (i.e. 100 ms) conditioning step at non-positive levels (i.e. 0 and −30 mV), with the prediction that Na V channels will be unlikely to adopt the open channel block and thus elicit minimal I NaR . Indeed, a dramatic reduction in current amplitude was observed in both cases ( Fig. 3Bb and c and Cb and c) along with a shift in voltage dependence by ß10 mV. Under the condition of −30 mV, the reduction was so great that the typical 'V' shape of the I NaR I-V curve was lost (Fig. 3Cc), which can be explained by the following two factors. First, I NaR , if any, was hardly detectable in response to the conditioning step of −30 mV at 100 ms (Fig. 3Cb). Instead, steady-state current was evident following the repolarization. Second, low-level contaminant noise and/or other channel conductances can obscure small I NaR , which prevents the amplitude of I NaR from being reliably measured (Afshari et al. 2004;Aman et al. 2009).
We further characterized the kinetics of I NaR , which is considered to be another important property. Two variables of kinetics were used based on previous studies: time to peak and decay time constant (tau; Raman & Bean, 1997;Lewis & Raman, 2011). Fig. 4A shows the calculation of two variables in the current study. Time to peak is defined as the time interval between the onset of repolarization and the I NaR peak. Decay time constant was calculated by fitting a single exponential to the decay phase of I NaR . All calculations were conducted under the conditioning step of +30 mV at 10 ms. Across the population of NM neurons, the average time to peak was 3.97 ± 1.19 ms when measured at the membrane voltage that elicited the maximum current. The membrane voltage for the representative trace shown in Fig. 4A was −35 mV. The decay time constant was plotted as a function of repolarizing membrane voltage (Fig. 4B). Again, similar to mammalian neurons, the decay time constant of I NaR increased with more depolarized membrane voltage and was generally larger than the decay variable of I NaT (Ming & Wang, 2003;Rush et al. 2005), which confirmed the slower kinetics of I NaR . In summary, these observations suggest that the use of the open channel blocker, the inactivation mechanisms and the I NaR kinetics of NM neurons closely resemble those of mammalian neurons, properties likely conserved across various species and neural structures.

Function of resurgent Na V current for NM neurons: experimental results
We hypothesized that I NaR , along with the underlying open channel blocker, plays an important role in promoting the AP firing rates of NM neurons. We predicted that a neuron's activation of open channel blocker would promote Na V channel availability and recovery from depolarization, which is ultimately important for subsequent and rapid AP firing. Therefore, we applied two voltage-clamp protocols based on previous studies to test our prediction (Raman & Bean, 2001;Patel et al. 2015). In the first protocol, NM neurons were held at −90 mV before being depolarized to +30 mV for 5 ms (Fig. 5). According to our previous observations (see Fig. 3), this conditioned the majority of Na V channels to be occupied by the open channel blocker (referred here as the 'open channel block state'). The membrane voltage was then set at −65 mV for NM neurons to recover. The recovery time varied from 2 to 50 ms ( step = 2 ms). Finally, a depolarization to 0 mV was applied to evoke an I NaT . In the second protocol, the conditioning step was changed to −30 mV for 40 ms, in order to maximize the occupancy of the classic inactivation gate (referred to here as the 'inactivation state'). Fig. 5A and B shows representative current traces in response to the two protocols from the same neuron. When neurons were given a short and positive conditioning step, abrupt repolarization to the resting state resulted in an obvious generation of I NaR (arrow in Fig. 5A), while this was not observed when conditioned to the inactivation state (Fig. 5B). The generation of I NaR further supported the expectation that the occupancy of the open channel blocker was primary for the first protocol.
To determine Na V channel availability under the two protocols, we calculated the normalized ratio, which indicates the amount of available Na V channels after the recovery period. To do this, a reference pulse to 0 mV was applied to NM neurons prior to the implementation of the two protocols described above, and the amplitude of I NaT after the recovery was normalized to this 'reference amplitude' . The normalized ratio was plotted as a function of the recovery time in Fig. 5C. We observed a clear separation of recovery trajectories between the two conditions, which indicates that the presence of open channel block significantly increased Na V channel availability during the recovery. The recovery trajectory was fit with a single exponential in order to obtain a recovery time constant (tau). The open channel block significantly shortened the recovery time constant, facilitating the recovery of Na V channels (Fig. 5D). These observations provide supporting evidence for our prediction that the use of the open channel blocker can promote the availability and recovery of Na V channels.
In addition, we found that the distinction between the two states was larger when the amount of recovery time was short. For example, when NM neurons were only given 2 ms to recover, the availability of Na V channels was reduced by ß67% on average from the open channel block state to the inactivation state (Fig. 5E, left). But this difference reduced to ß20% when the recovery time was longer than 20 ms (Fig. 5E, middle and right). This suggests that the role of open channel block in recovering Na V channels is more critical when the recovery time is limited, which is reminiscent of a highly restricted interspike interval when NM neurons are performing rapid auditory tasks (Warchol & Dallos, 1990;Jones & Jones, 2000).

Function of resurgent Na V current for NM neurons: computational results
We further used a computational model to examine the function of I NaR in NM. This model is designed based on our previous study (Lu et al. 2017), in combination with the I NaR model from Khaliq et al. (2003). The model NM neuron was held at −90 mV before giving a depolarizing conditioning step to +30 mV (duration = 10 ms). An I NaT was evoked by the conditioning step (Fig. 6Aa,  arrowhead). The model NM neuron was then repolarized to varying membrane voltages from −70 mV to 0 mV ( step = 5 mV). The model NM neuron displayed slightly smaller I NaR amplitudes (Fig. 6Aa, small arrow) but comparable voltage dependence properties to our experimental data (Figs 6Ab and 3Ac). In order to remove I NaR , we first set the rate constant ε to zero (see Fig. 1  C, population data showing the Na V channel availability (%) as a function of recovery time. In order to calculate Na V channel availability, a reference pulse to 0 mV was applied to NM neurons (not shown in the figure), and the amplitude of I NaT after the recovery was normalized to this 'reference amplitude'. Before the normalization, the amplitude of I NaT was first adjusted by subtracting the steady-state current that remained at the end of the conditioning step. The recovery trajectory is fit by a single exponential, in order to obtain recovery time constant for reference). This modification, however, significantly slowed down the falling phase of I NaT (Fig. 6B, the '0-I NaR− ' condition, blue trace). The slower falling phase of I NaT is because the Markovian 13-state Na V -channel model sets the O→OB transition (with the rate constant ε, see Fig. 1) as a major exit path from the open state (Magistretti et al. 2006). Removing this path resulted in the slower speed of channels exiting the open state and thus led to a slower I NaT falling phase. Therefore, we next increased the rate constant O on to 2.15 ms −1 and the O off to 0.01433 ms −1 to restore the normal decay kinetics of I NaT and amplitude of I NaP , respectively ( Fig. 6B and C, the '0-I NaR+ ' condition, red trace). After these two modifications, I NaR was successfully eliminated (Fig. 6C). Both 0-I NaR− (blue) and 0-I NaR+ (red) conditions were used to characterize the spiking activity of model NM neuron without I NaR . The only difference between 0-I NaR+ and control conditions was the absence of I NaR , while the 0-I NaR− condition showed both the absence of I NaR and slower I NaT decay kinetics. Our model NM neuron generated similar voltage responses to square pulse current trains with varying frequencies, as compared to our experimental data. For example, the model NM neuron responded to a 200 Hz current injection with ß0.6 AP firing probability (Fig. 7Aa and c). It should be noted that an AP was identified only if the evoked voltage peak exceeded −30 mV. Voltages below this value were considered as failures (Fig. 7Ac, asterisks). Fig. 7Ab shows the total underlying Na V current required for AP generation in Fig. 7Aa. With the current and time scale expanded, we observed a clear but small inward Na V current immediately after a large I NaT (Fig. 7Ad, arrow). This inward Na V current is reminiscent of I NaR that occurs during the AP repolarizing phase and closely resembles the data reported by previous 'AP-clamp' studies in mammalian neurons (Raman & Bean, 1997, 1999. When we removed I NaR in the model NM neuron, the reduction in firing probability was identical in both 0-I NaR− and 0-I NaR+ conditions. Firing probability at 200 Hz was reduced by ß21% (Fig. 7Ba-c, asterisks; Fig. 7Ca-c, asterisks) and the small inward Na V current after I NaT was no longer visible ( Fig. 7Bd and Cd, arrow). This result further confirmed that the small inward current is I NaR and is induced by the mechanism of open channel block. In addition, when we removed both I NaR and K HVA currents, firing probability in response to 200 Hz stimulation was reduced by ß23% from the control -just slightly larger than the reduction we observed with I NaR removal only (data not shown).
It should be noted that I NaT generated in the 0-I NaR− condition was obviously wider than in the 0-I NaR+ condition, though the same amount of reduction in firing respectively. Ab, current-voltage relationship of simulated I NaR . B, simulated I NaT (upper panel) was evoked by depolarizing steps to 0 mV (lower panel). The current obtained under control condition is shown in black. Switching off I NaR by setting the rate constant for the O→OB transition, ε, to 0 resulted in considerable slowing of I NaT decay (0-I NaR− condition, blue trace). I NaT was restored under 0-I NaR+ condition (ε = 0, O on = 2.15 ms −1 , and O off = 0.01433 ms −1 ; red trace). C, simulated I NaR under control (black trace) and 0-I NaR+ condition (red trace). Removal of I NaR under the 0-I NaR+ condition has no effect on persistent Na V current (I NaP ). probability was observed for both conditions (Fig. 7Bd and Cd). This difference is due to the slower decay kinetics of I NaT in the 0-I NaR− condition that means Na V channels inactivate more slowly (see Fig. 6B). As a result, APs generated in the 0-I NaR− condition showed longer half-width and larger amplitude than in the 0-I NaR+ condition ( Fig. 7Bc and Cc). In addition, we also observed a difference between two conditions in a small I NaT that failed to generate an AP after a large I NaT (Fig. 7Bd and Cd, asterisks). In the 0-I NaR− condition, Na V currents after a large I NaT , the amplitude of which indicates the number of Na V channels recovered after an AP, were present with smaller amplitude than those in the 0-I NaR+ condition ( Fig. 7Bd and Cd, asterisks). This is because of the slower Na V channel kinetics in the 0-I NaR− condition and the highly restricted recovery time in response to 200 Hz square pulse trains. Nevertheless, the identical reduction in firing probability in both conditions confirmed that the reduced ability of the model NM neuron to follow inputs at high rates was not due to a slower I NaT but to the absence of I NaR . Together, our modelling results indicate that I NaR helps shape AP firing rate of NM neurons.

Development of resurgent Na V current in NM
We confirmed the generation of I NaR in late-developing NM neurons (i.e. E19-21), when near-mature hearing ability is established (Jones et al. 2006). Next, we investigated whether NM neurons in the earlier periods of development have I NaR . Voltage-clamp experiments were first performed on NM neurons at E14-16, which corresponds to a developmental period when crude hearing ability is just established for chickens (i.e. during hearing onset). To our surprise, we observed robust generation of I NaR at E14-16 when using the conditioning step of +30 mV, 10 ms, with the peak amplitude just slightly smaller than that of late-developing NM neurons ( Fig. 8A and D). Also similar to late-developing neurons, the I NaR amplitude was reduced when we changed the level of the conditioning step to either less positive voltages (i.e. 0 and −30 mV) or extended the step duration (i.e. 10 and 100 ms) ( Fig. 8A and B and D-F). Additionally, we observed a shift in the peak of I-V curve in the positive direction by ß15 mV when we prolonged the duration of the step from 10 to 100 ms. These observations Despite these similarities, we also noticed that some properties of I NaR at E14-16 were different from latedeveloping neurons. It should be noted that the following differences are discussed using the conditioning step of +30 mV, 10 ms. Although the I NaR amplitude peaked at −40 mV for both age groups, it reduced more dramatically at older ages when membrane voltage became more positive or more negative (see Figs 3Ac and 8D). According to a previous study (Lewis & Raman, 2011), the relative amplitude of I NaR depends on two mechanisms. First, the rate of the open channel block unbinding from the Na V channel α-subunit, which indicates the affinity between these two particles, and second, after removal, the rate of the α-subunit entering the classic inactivation or closed states (depending on the voltage). Time to peak is a good index for the first mechanism (i.e. the rate of unbinding), while decay time constant is dependent on both mechanisms. We did not observe a significant difference of time to peak between two age groups (see Fig. 9D, P = 0.88). Furthermore, we plotted the decay time constant as a function of membrane voltage for E14-16 NM neurons and compared it with E19-21 NM neurons (Figs 8C and 4B, respectively). Indeed, E14-16 NM neurons showed a generally larger decay time constant than the older age group. Therefore, we suggest that Na V channels at E14-16 inactivate or close (at the membrane voltage more positive or more negative than −40 mV, respectively) significantly slower after the unbinding of the open channel blocker, resulting in a less dramatic fall-off of I NaR amplitude and thus a shallower slope of their I-V curve than Na V channels at E19-21.
Although additional experiments are required to test this suggestion, we speculate that longer I NaR decay kinetics are partially related to the type of Na V α-subunit(s) expressed at E14-16 compared to older ages (see below).
Next, we performed voltage-clamp experiments on NM neurons at E11-12, when chickens are not able to respond to sound (i.e. before hearing onset; Jones et al. 2006). In 10 out of 14 recorded neurons, we observed small I NaR when using the conditioning step of +30 mV, 10 ms (Fig. 9A, note scale). The maximal amplitude of I NaR was slightly above 100 pA, which is dramatically smaller than the current amplitude of the other two age groups (Fig. 9B). In addition, the voltage dependence of I NaR at E11-12 was very different from the other two age groups, showing an obvious shift in peak voltage (i.e. at −20 mV, Fig. 9B). Fig. 9C shows the decay time constant of I NaR for NM neurons before hearing onset, which presented with an increasing trend with more depolarized membrane voltage. Time to peak was 9.54 ± 2.87 ms when measured at the maximal I NaR . This value is significantly larger than those of the other two age groups, indicating a stronger affinity between the open channel blocker and the α-subunit of Na V channels, and thus slower kinetics of I NaR
at E11-12 (Fig. 9D). Taken together, during the prehearing period, I NaR was present in the majority of NM neurons with preliminary properties. With development, the I NaR amplitude increased, kinetics improved and its voltage dependence shifted towards negative direction. By the time of hearing onset, NM neurons obtained more mature-like I NaR properties. Table 3 summarizes the developmental changes in amplitude of I NaT , I NaR and I NaP , when measured at membrane voltage of −30 mV. The amplitude of I NaP was measured at the end of 100 ms repolarization after the conditioning step (+30 mV, 10 ms). We observed significant increases in all three current amplitudes for NM neurons as a function of development.

Development of frequency-firing pattern in NM
Based on the functional role of I NaR and its development in NM, we predicted that neurons at the age of E14-16 would show similar frequency-firing pattern with late-developing neurons (i.e. E19-21), whereas neurons at E11-12 would not be able to follow square pulse frequency as high as the other two age groups, due to their underdeveloped I NaR (see Fig. 9) and K V channels (Hong et al. 2016). To test this prediction, we first applied the aforementioned current commands of square pulse trains to NM neurons at E14-16. As expected, neurons at this age were able to follow square pulse trains of 150 Hz in a one-to-one manner ( Fig. 10Aa  and c). Firing probability dropped continuously when stimulus frequency was higher than 150 Hz (Fig. 10Ac). In response to square pulse trains of 200 and 250 Hz, firing probabilities were ß0.5 and ß0.25, respectively ( Fig. 10Ab and Ac), which are slightly lower than those of E19-21 NM neurons but these differences are not statistically significant (P = 0.49 and 0.32, respectively). When the stimulus frequency was 300 Hz, E14-16 NM neurons showed a single onset spike similar to late-developing neurons (data not shown).
For NM neurons at E11-12, individual square pulse width was extended to 5 ms because neurons at this age are present with average AP half-width of ß4.6 ms (Hong et al. 2016). We found that NM neurons at E11-12 were able to follow square pulse trains with good fidelity up to 70 Hz (Fig. 10Bc). Fig. 10Ba shows a representative E12 neuron that was able to reliably generate an AP in response to each pulse of the 70 Hz stimulation. When stimulus frequency was increased to 100 Hz or higher, however, firing probability dropped steeply below 0.5 and in response to 150 Hz stimulation, NM neurons at E11-12 only generated a single onset spike followed by membrane oscillations at a depolarized level ( Fig. 10Bb and c). This is in stark contrast to the other two age groups, which could follow 150 Hz square pulse trains in a reliable manner (see Figs 2D and 10Ac). Taken together, NM neurons at E14-16 showed a frequency-firing pattern closely resembling their late-developing counterparts, while the ability of NM neurons at E11-12 to follow high-frequency stimulations was markedly limited. These results are similar to our previous data, which show that E11-12 NM neurons are most responsive to low-frequency (<40 Hz) sinusoidal current injections (Hong et al. 2016).
Expression of Na V 1.6 channels Na V 1.6 channels are extensively expressed in the mammalian central nervous system and act as a predominant carrier for I NaR (Eijkelkamp et al. 2012;Lewis & Raman, 2014). Therefore, we explored whether Na V 1.6 channels are expressed in developing NM, using an antibody specifically recognizing chicken Na V 1.6 (Kuba et al. 2006(Kuba et al. , 2010(Kuba et al. , 2014. In late-developing NM neurons (i.e. E21), strong Na V 1.6 immunoreactivity was observed as bright punctate segments ( Fig. 11Aa and b). Because E21 NM neurons are mostly adendritic (Jhaveri & Morest, 1982a), double staining of Na V 1.6 and neurofilament further demonstrated that Na V 1.6 is localized in neurofilament-stained NM axons that can be traced back to the cell bodies (Fig. 12). These Na V 1.6-containing segments closely resembled the characterized distribution pattern of Na V 1.6 in chicken auditory brainstem as reported by Kuba et al. (2010Kuba et al. ( , 2014 and possibly represented the axon initial segments (AIS) and nodes of Ranvier. In contrast, the Na V 1.6-containing segments were absent in NM at E15 (Fig. 11Ba and b) and E11 ( Fig. 11Ca and b). In addition, NM cell bodies contained a low level of Na V 1.6 immunoreactivity at E21, which became increasingly less distinct from E21 to E15 and from E15 to E11. J Physiol 596.3

Discussion
Auditory brainstem neurons of vertebrates fire phaselocked APs at high rates with remarkable fidelity. We show here that late-developing NM neurons fire reliable APs to square-pulse current injections up to 200 Hz. In addition to the recognized role of K HVA channels in this process, we also report that I NaR of Na V channels is an equally important component that operates synergistically with K HVA channels to enable rapid AP firing in NM, an evolutionarily conserved process between birds and mammals likely promoting similar hearing functions. To our surprise, small I NaR responses were present during a prehearing period (E11-12). At hearing onset (E14-16), I NaR properties closely resemble late-developing NM neurons (E19-21), despite developmental refinement in Na V channel protein expression patterns. In line with these results, NM neurons at E14-16 showed comparable frequency-firing ability to their late-developing counterparts, whereas NM neurons at E11-12 are most responsive to lower-frequency stimulations.

Factors regulating AP firing patterns in NM
K HVA channels are a common regulator of AP kinetics (Rudy & McBain, 2001;Johnston et al. 2010). In the auditory brainstem of both birds and mammals, K V 3containing channels are abundantly expressed Parameshwaran et al. 2001;Parameshwaran-Iyer et al. 2003). Blockade or knockout of K V 3 channels results in wider APs that undermine the ability to generate APs at high rates Klug & Trussell, 2006). We report a similar result here (Fig. 2). In addition, a recent study on Xenopus oocytes shows that K V 3.1 subunits are able to produce resurgent potassium current during repolarization, which provides an additional repolarizing drive to the membrane (Labro et al. 2015). This property is not due to the open channel blocker but to the unique gating kinetics of K V 3.1 subunits, and it facilitates the termination of APs and thus likely promotes high-frequency repetitive firing. Nevertheless, it should be noted that 1 mM TEA used in the current study can also block a small portion of K V 7-containing channels and calcium-activated BK channels (Johnston et al. 2010). K V 7-containing channels are one of the K LVA channels with slow activation kinetics (Johnston et al. 2010). They contribute minimally in NM and their immunoreactivity is minimal compared to K V 3 and K V 1 (Kuba et al. 2015). Due to their weak expression, blockade of K V 7-containing channels in NM neurons does not induce significant changes in AP properties (Kuba et al. 2015). In contrast, BK channels play an important role in regulating AP kinetics in many other neurons (Kimm et al. 2015). However, a previous study in NM reported no change in K V current when external calcium was replaced by cobalt ions, suggesting minimal contribution of calcium-activated potassium current to the total current (Koyano et al. 1996). Therefore, K V 3-containing channels are probably the primary targets of 1 mM TEA applied in our experiments. Another factor that regulates AP firing pattern is fast inactivation kinetics of Na V channels (i.e. the steep decay slope of I NaT ) observed in auditory brainstem neurons (Ming & Wang, 2003;Hong et al. 2016). As demonstrated by double-pulse experiments, a second I NaT can fully recover within 5 ms following the initial pulse (Leao et al. 2005; unpublished observation in NM), which explains in the current study why NM neurons are able to reliably follow the stimulus frequency at 150 Hz (interspike interval = 6.7 ms, Fig. 2D) after blockade of K HVA channels. Finally, K LVA channels with fast activation kinetics, such as K V 1-containing channels, work as another regulator of AP firing by controlling the time constant of passive membrane properties, shaping the speed of membrane voltage changes (Klug & Trussell, 2006). I NaR is an additional factor in regulating AP firing rates of NM neurons. The open channel block promotes Na V channel availability and recovery immediately after brief depolarization in NM, a result consistent with previous studies (Raman & Bean, 2001;Patel et al. 2015). However, we report two major differences. First, Na V channel availability is much higher in our study at any given recovery time period (Fig. 5C). This is likely because our recordings were made from intact neurons in brainstem slices that have significantly more neuron-specific Na V channels. Previously reported data were obtained from the soma of dissociated Purkinje cells or from HEK cells transfected with Na V channels. Second, the distinction in recovery time constant between the open channel block and classic inactivation state is smaller in our study (ß1 ms versus > 5 ms, Fig. 5D). This may be attributed to relatively fast kinetics of I NaT in auditory brainstem neurons. However, the extent to which the open channel block or the classic inactivation gate contributes to rapid I NaT kinetics in NM is unclear.
'Real-time' I NaR appears to be a relatively small inward Na V current during repolarization from an initial I NaT , as shown by our model NM neuron (Fig. 7Ad). This observation closely resembles experimental data when AP waveforms are used as voltage commands and the dynamic changes in Na V current are documented (i.e. AP-clamp method; Raman & Bean, 1997, 1999. As seen from studies that utilized this technique, the I NaR provided a small depolarizing drive immediately after an AP, important for subsequent AP firing (Raman & Bean, 1997). However, AP-clamp is susceptible to various technical limitations (e.g. space clamp errors), and as a result the majority of these studies were conducted on dissociated neurons (Raman & Bean, 1997;Do & Bean, 2003). Nevertheless, there are successful slice recordings with AP clamp from adendritic mesencephalic trigeminal neurons (Enomoto et al. 2006) Given the fact that the majority of late-developing NM neurons are also adendritic, AP-clamp experiments could provide valuable insight into the biological relevance of I NaR . There are still J Physiol 596.3 technical limitations (e.g. sample rate, kinetics, immature dendritic NM neurons, etc.) and therefore we employed a computational model to better address its functional significance. Indeed, removal of I NaR reduced AP firing rate at 200 Hz for our model NM neuron (Fig. 7). Yet, the effects on Na V channels after the knockdown of open channel blocker remain to be determined, especially regarding the kinetics of I NaT under experimental condition. These effects are dependent on the types of β-subunits expressed in NM neurons and their specific interactions with Na V α-subunits (Qu et al. 2001;Aman et al. 2009;Bant & Raman, 2010). Therefore, our computational model of NM neuron is subject to future improvements when more information about molecular substrates for the open channel blocker in NM is obtained.

Development of resurgent Na V current in NM
The inner ear of chickens only responds to loud sound (>80 dB) and afferent ganglion neurons present with poor frequency selectivity at E14-16 (Jones et al. 2006). It is surprising at this age that I NaR properties are relatively established, except for decay kinetics, which are due in part to underdeveloped Na V α-subunit (Fig. 8, see below). I NaR properties at E11-12 -a developmental period when the auditory system does not respond to sound and is considered 'prehearing' -differ greatly from those of the other age groups (Fig. 9). I NaR development in NM parallels the development of frequency-firing pattern (Fig. 10), along with maturation of intrinsic AP properties and K V channel conductances as we previously reported (Hong et al. 2016). NM neurons at E14-16 generate APs as fast and reliably as late-developing neurons. Additionally, properties of K LVA and K HVA currents are underdeveloped at E11-12 but become more mature-like around hearing onset. Taken together, NM neurons have near-mature ion channel properties around hearing onset, suggesting a priming period during hearing development for establishing mature auditory functions.
I NaR at E11-12, though small, may be important for shaping NM's AP firing pattern. In response to low-frequency (5-10 Hz) sinusoidal current injections, NM neurons at this age generate short bursts of APs per cycle, despite the fact that their overall low K V channel conductances are not sufficient to work against the largely depolarized membrane voltage during bursts (Hong et al. 2016). I NaR also contributes to burst firing in other non-auditory neurons (Enomoto et al. 2006(Enomoto et al. , 2007. We speculate that I NaR at E11-12, albeit reduced, helps generate burst firing for immature NM neurons, but its functional significance at this age is unclear. Anatomically, we detected strong Na V 1.6 expression at a late-developing stage (E21), which was dramatically reduced at hearing onset (E14-16) and prehearing (E11-12) periods (Fig. 11). Additionally, the majority of Na V 1.6 channels in late-developing neurons are located along the axons, while this distribution pattern is not evident for the other age groups. The Na V 1.6-positive punctate segments in late-developing NM neurons are likely to be the AIS and the nodes of Ranvier, respectively, though future immunocytochemistry with double staining of ankyrin G, a scaffold protein that marks the AIS, is needed to confirm this speculation (Kuba et al. 2014). The presence of I NaR at E11-16 suggests that additional Na V subtypes may contribute to I NaR during development of NM neurons. The following evidence supports this suggestion. First, strong expression of Na V 1.2 channels is observed at E11-15 in NM's postsynaptic target nuclei (i.e. nucleus laminaris) (Kuba et al. 2014). Second, studies in spinal sensory neurons report the generation of I NaR from different Na V subtypes, including Na V 1.2 channels (Rush et al. 2005;Jarecki et al. 2010).

Resurgent Na V current is conserved across species and structures
I NaR has been reported as a conserved property in numerous neuronal types in the mammalian cerebellum, cortex, brainstem and spinal cord (Afshari et al. 2004;Cummins et al. 2009;Lewis & Raman, 2014). As for the mammalian auditory system, I NaR has been found at the calyx of Held and its postsynaptic target -the medial nucleus of trapezoid body (Leao et al. 2005;Kim et al. 2010). The only report of I NaR beyond the scope of mammals comes from chicken Purkinje cells (Lewis & Raman, 2011). Our results from the current study further confirm its contribution in the chicken auditory brainstem. Here, the AP firing pattern of NM neurons (and the analogous bushy cells of the mammalian anteroventral cochlear nucleus) is essential for the encoding of behaviourally relevant acoustic cues. Central to this are established and specialized structural and functional features shared across species (Carr & Soares, 2002;Koppl, 2009;Grothe et al. 2010). The report of I NaR in the current study strongly suggests yet another physiological feature shared across species within the central nervous system. Despite the distribution of I NaR in the central nervous system (Lewis & Raman, 2014), its underlying molecular substrates are unclear. The Na V β4-subunit has been identified as one important open channel blocker in cerebellar Purkinje cells, granule cells and dorsal root ganglia neurons (Grieco et al. 2005;Bant & Raman, 2010;Barbosa et al. 2015). Interestingly, the amino acid sequence of the β4-subunit is also conserved across species, from frogs to mammals, raising the possibility of its being the open channel blocker for avian NM neurons (Lewis & Raman, 2011). Nevertheless, questions are raised of whether the β4-subunit is the only possible open channel blocker. For example, both perirhinal and entorhinal pyramidal neurons show robust I NaR but not β4 expression at high levels, possibly indicating multiple molecular substrates of I NaR (Castelli et al. 2007;Nigro et al. 2012). In addition, FGF14, an intracellular protein and a member of the fibroblast growth factor homologous factors (FHFs), has been proposed as another key player in open channel block in Purkinje cells (Yan et al. 2014). A recent study on dorsal root ganglia neurons demonstrated that another member of the FHF family, FHF2, plays an important regulatory role for I NaR (Barbosa et al. 2017). Interestingly, two isoforms of FHF2 have opposite effects on the amplitude of I NaR . As for NM neurons, it should be noted that differences in I NaR across development, especially at E11-12, might be attributed not only to different expression of Na V α-subunits, but also to changes in the open channel blocker molecular substrates themselves. Therefore, what the underlying open channel blockers for NM neurons are and how the developmental expression of these blockers shapes I NaR maturation are interesting questions for future studies.