Coordinated Expression of Two Types of Low-Threshold K+ Channels Establishes Unique Single Spiking of Mauthner Cells among Segmentally Homologous Neurons in the Zebrafish Hindbrain

Abstract Expression of different ion channels permits homologously-generated neurons to acquire different types of excitability and thus code various kinds of input information. Mauthner (M) series neurons in the teleost hindbrain consist of M cells and their morphological homologs, which are repeated in adjacent segments and share auditory inputs. When excited, M cells generate a single spike at the onset of abrupt stimuli, while their homologs encode input intensity with firing frequency. Our previous study in zebrafish showed that immature M cells burst phasically at 2 d postfertilization (dpf) and acquire single spiking at 4 dpf by specific expression of auxiliary Kvβ2 subunits in M cells in association with common expression of Kv1.1 channels in the M series. Here, we further reveal the ionic mechanisms underlying this functional differentiation. Pharmacological blocking of Kv7/KCNQ in addition to Kv1 altered mature M cells to fire tonically, similar to the homologs. In contrast, blocking either channel alone caused M cells to burst phasically. M cells at 2 dpf fired tonically after blocking Kv7. In situ hybridization revealed specific Kv7.4/KCNQ4 expression in M cells at 2 dpf. Kv7.4 and Kv1.1 channels expressed in Xenopus oocytes exhibited low-threshold outward currents with slow and fast rise times, while coexpression of Kvβ2 accelerated and increased Kv1.1 currents, respectively. Computational models, modified from a mouse cochlear neuron model, demonstrated that Kv7.4 channels suppress repetitive firing to produce spike-frequency adaptation, while Kvβ2-associated Kv1.1 channels increase firing threshold and decrease the onset latency of spiking. Altogether, coordinated expression of these low-threshold K+ channels with Kvβ2 functionally differentiates M cells among homologous neurons.


Introduction
Neurons exhibit varied intrinsic excitability, as represented by temporal firing patterns in response to step depolarization. These are basically classified as either phasic-or tonic-firing patterns by the respective presence or absence of spike-frequency adaptation (Carr and Soares, 2002;Prescott and De Koninck, 2002;Maravall et al., 2004;Kalluri et al., 2010). Heterogeneity in firing threshold, duration, onset latency, and spike-frequency adaptation are also observed among phasic-firing neurons, even in specific neural populations (Locke and Nerbonne, 1997;Prescott and De Koninck, 2002;Soares et al., 2002;Reid et al., 2004). Although intrinsic firing properties are acquired developmentally from immature firing to the phasic or tonic type (Nakamura and Takahashi, 2007;Iwasaki et al., 2008;Hoffpauir et al., 2010;Marrs and Spirou, 2012), the underlying molecular mechanisms remain unclear.
Zebrafish and goldfish Mauthner (M) cells, a pair of giant reticulospinal neurons (RSNs) located in the fourth segment of the hindbrain, are known as a vertebrate model that generates only a single action potential (AP) at the onset of noxious sensory input, thereby triggering fast escape behavior (Zottoli, 1977;Nakayama and Oda, 2004;Kohashi and Oda, 2008). In contrast, two paired RSNs that are repeated in adjacent hindbrain segments (MiD2cm and MiD3cm) are morphologically homologous to M cells, and fire tonically with regular spiking at frequencies dependent on input intensity (Nakayama and Oda, 2004). Because M cells and M-cell homologs, collectively called the "M series" (Lee and Eaton, 1991), share direct projections from auditory afferents (Nakayama and Oda, 2004;Szabo et al., 2007), they are thought to play a role in initiating escape by encoding the onset and intensity of sound, respectively (Oertel, 1999). Because the M series is thought to be evolutionarily duplicated (Metcalfe et al., 1986), we hypothesized that functional differentiation of intrinsic excitability arises from segmental duplication of a common ancestral neuron accompanied by subsequent variation. Our previous zebrafish study indicated that single spiking of M cells is developmentally acquired at 4 d postfertilization (dpf) from immature phasic-firing by enhancing surface expression of voltage-gated K ϩ (Kv) channel ␣-subunits, Kv1.1a/ kcna1a (also known as zKv1.1a), via later developmental expression of auxiliary Kv␤2 subunits Kv␤2b/kcnab2b (also known as zKv␤2b; Watanabe et al., 2014). However, the phasic-firing pattern of 4-dpf M cells after pharmacologically blocking Kv1.1 channels substantially differs from repetitive firing at 2 dpf. These results raise the possibility that other ion channels contribute developmentally to distinct phasic-firing properties specific to M cells among the M series.
Here, we sought to determine the ionic basis for developmentally-acquired differentiation of firing properties among this neural population. We focused on lowthreshold K ϩ channels, including the Kv1 and Kv7/KCNQ families (Johnston et al., 2010), which are activated around the resting membrane potential and play a role in generating phasic-firing in sympathetic, auditory, and vestibular mammalian neurons (Brew and Forsythe, 1995;Wang and McKinnon, 1995;Kalluri et al., 2010). Although they have a similar low-threshold activation voltage, Kv7/ KCNQ channels exhibit slow activation-kinetics , whereas Kv1 channels show rapid activation (Hopkins et al., 1994). However, the differential contributions of these channels to phasic firing remains poorly understood.
In this study, we performed in vivo whole-cell recordings from M-series neurons in zebrafish embryos and larvae, and examined the effects of pharmacological manipulations of low-threshold K ϩ channels on firing patterns. We first demonstrate that Kv7/KCNQ is responsible for producing phasic behavior in M cells from 2 dpf, which is distinct from the later contribution of Kv1.1. Second, expression analysis of Kv7 ␣-subunits, electrophysiological assessment of Xenopus oocytes, and computational modeling revealed that early expression of Kv7.4/KCNQ4 in M cells mediates slowly-activating currents to produce phasic bursting. Third, later expression of Kv␤2 enhances inhibition of repetitive firing by Kv1.1, resulting in single spiking. Together, the contributions of two different lowthreshold Kv channels with an auxiliary subunit differentiate unique M-cell phasic-firing properties from a common tonic-firing property conserved among the M series.

In vivo whole-cell recording
Tol-026 or Tol-056 strain embryos and larvae at 56 -64 h postfertilization (hpf; 2 dpf) and 102-181 hpf (4 -7 dpf) were processed as described previously (Watanabe et al., 2014). In brief, to allow access to M cells or M-cell homologs, embryos or larvae were anesthetized with 0.02% tricaine mesylate (MS-222, Sigma-Aldrich), immobilized with 1 mM D-tubocurarine (Sigma-Aldrich), and then pinned and operated on in a silicon dish filled with extra-This work was supported by Grants-in-Aid for Scientific Research (KAKENHI 22300126, 23650164, 23115508, 25250002) and a fellowship from the Japan Society for the Promotion of Science (KAKENHI 09J09220).

Two-electrode voltage-clamp recordings in Xenopus oocytes
Xenopus laevis oocytes were collected and treated as previously described (Watanabe et al., 2014), and injected using a Drummond microinjector with 46 nl cRNA (250 ng/l) synthesized from full-length zebrafish Kv7.4/kcnq4, Kv1.1a/kcna1a, and Kv␤2b/kcnab2b cDNA using a mMESSAGE mMACHINE Kit (Life Technologies). The volume ratio of Kv1.1a and Kv␤2b cRNA solution was 1:5. K ϩ currents were recorded using two electrodes with a resistance of 0.5-1.0 M⍀ (when filled with 3 M KCl), and an Axoclamp 2B amplifier (Molecular Devices). During recordings, oocytes were perfused in a chamber with a bath solution containing: 96 mM NaMeSO 4 , 2 mM KCl, 1 mM MgCl 2 , 1 mM CaCl 2 , and 5 mM HEPES, adjusted to pH 7.6 with NaOH. Niflumic acid (0.3 mM) was added to block endogenous chloride currents (Hopkins et al., 1994). Transient capacitance and linear leakage currents were subtracted using a P/4 procedure. The conductancevoltage plots were fitted by a standard Boltzmann equation where V 1/2 is the voltage of half-maximal activation, and k is the slope factor). Data were analyzed and fitted using Clampfit 10 software (Molecular Devices).

Computational model
We simulated low-threshold K ϩ currents and M cell firing in a single-compartment model of a Xenopus oocyte and M cell, respectively, using NEURON simulator (version 7.3, RRID:SCR_005393; Hines and Carnevale, 2001). Our model channel generator and firing simulator are publicly available on Model DB repository (https:// senselab.med.yale.edu/ModelDB; accession # 232813) and as Extended Data 1. The simulations were based on a Hodgkin-Huxley (H-H) model of a mammalian ventral cochlear nucleus neuron, which like an M cell, possesses a single-spiking property (Rothman and Manis, 2003b). We constructed single cylindrical compartments for the Xenopus oocyte model (20-m length and 20-m diameter) and M-cell model (40-m length and 40-m diameter), possessing 12 pF and 50 pF, respectively (specific membrane capacitance: c m ϭ 1 F/cm 2 ). These compartments enabled us to directly observe the effects of ion conductances on the cell without cable properties derived from an axon and dendrites or non-uniform localization of ion channels. The parameters of the H-H model for lowthreshold K ϩ conductance of each zebrafish Kv7.4 (g Kv7.4 ), Kv1.1 (g Kv1.1 ), and Kv1.1 with Kv␤2 (g Kv1.1ϩKv␤2 ) were fitted to the data from the Xenopus oocyte system described above, by simulating voltage-clamp recordings at a virtual 20°C. The model M cell included the following maximal conductances: 6000 ns, voltage-gated Na ϩ conductance (g Nav : 120 ms/cm 2 ); 1500 ns, high-threshold K ϩ conductance (g high : 30 ms/cm 2 ); 300 ns, A-type K ϩ conductance (g A : 6 ms/cm 2 ); and 12 ns, leak conductance ( g leak : 0.24 ms/cm 2 ) and arbitrary low-threshold K ϩ conductance (g Kv7.4 , g Kv1.1 , g Kv1.1ϩKv␤2 ). Additionally, H-H models of g Nav , g high , and g A were modified from the study by Rothman and Manis (Rothman and Manis, 2003b) to fit M-cell firing and AP wave form. The reversal potentials of Na ϩ , K ϩ , and the leakage conductances were set to ϩ30 mV, Ϫ90 mV, and Ϫ85 mV, respectively, and the resting potential of the M-cell model was set to a default of Ϫ85 mV, as observed in vivo (Watanabe et al., 2014). Currentclamp simulation was conducted at a virtual temperature of 25°C, and the threshold current (1T) was found by 1 pA increments via current injection. The combination effects of Kv7.4 and Kv1.1 for 2-dpf M cells, or Kv7.4 and Kv1.1 with Kv␤2 for 4-dpf M cells were simulated by implementation of (g Kv7.4 and g Kv1.1 ) or (g Kv7.4 and g Kv1.1ϩKv␤2 ). Overall, 1071 firing patterns (21 ϫ 51 combinations) were produced by increments of 100 ns for g Kv7.4 and 20 ns for g Kv1.1 or g Kv1.1ϩKv␤2 , ranging from 0 to 2000 ns (40 ms/ cm 2 ) and from 0 to 1000 ns (20 ms/cm 2 ), respectively.

Statistical analysis
The pharmacological experiments in the M cells were analyzed using Mann-Whitney U tests for two independent data sets. Two-way repeated measures ANOVA was used for interspike interval (ISI) comparisons because ISI was repeatedly measured from the same cell at different time points from ISI 1st to ISI 12th . We used the Kruskal-Wallis test with the post hoc Bonferroni-corrected Mann-Whitney U test for analysis of the Xenopus oocyte system. Statistical analysis was conducted in EZR (Kanda, 2013). All data are represented as mean Ϯ SEM.

Coordinated contribution of Kv7/KCNQ and Kv1 channels to single spiking in zebrafish M cells
To identify the ion channels responsible for establishing the unique single-spiking property of M cells among M-series neurons, we first examined the effect of pharmacologically blocking low-threshold K ϩ channels on firing patterns in response to step depolarization. In vivo whole-cell recordings from zebrafish M cells at 4-7 dpf revealed typical single spiking at the onset of step depolarization, which contrasts the tonic firing of the M-cell homologs repeated in adjacent segments (i.e., MiD2cm and MiD3cm; Fig. 1). Bathing with either DTX or XE991, specific blockers for Kv1 or Kv7 channels, respectively, caused M cells to burst phasically (Fig. 1A). After treatment with both DTX and XE991, M cells showed tonic firing or regular spiking at frequencies that depended on the injected current intensity, similar to the segmental homologs ( Fig. 1A,B; Nakayama and Oda, 2004; Watanabe et al., 2014). We quantitatively illustrated the effect of K ϩ channel blockers by plotting the number of spikes elicited by depolarizing current pulses against the current intensity ( Fig. 1C). We found no significant difference in the ISI compared with M-cell homologs (Fig. 1D). These results demonstrate that both DTX-and XE991-sensitive Kv1 and Kv7, respectively, are responsible for acquisition of the single-spiking property by M cells. Hence, coordinated expression of both types of low-threshold K ϩ channels is likely a key step for functional differentiation of M cells as possessing a single-spiking activity pattern that distinguishes them from their homologous neurons.
It should be noted that DTX-and XE991-sensitive K ϩ channels may contribute differentially to M-cell excitability, because DTX treatment decreased threshold current intensity for spiking, and prolonged the time to spike initiation (spike latency) from stimulus onset at threshold intensity, whereas XE991 treatment alone did not affect these properties (Fig. 1E,F). Moreover, there was a tendency for a more prolonged spike latency with both DTX and XE991 treatment ( Fig. 1F; see Discussion).

Kv7/KCNQ channels are required for phasic bursting of immature M cells
Since immature M cells are known to burst ( Fig. 2; Watanabe et al., 2014), we next examined the contribution New Research of both types of low-threshold K ϩ channels on M-cell firing during early development. Blocking Kv1 channels with DTX treatment had little effect on immature zebrafish M cells at 2 dpf ( Fig. 2A), as we previously reported (Watanabe et al., 2014). In contrast, blocking Kv7 channels with XE991 treatment significantly enhanced excitability, especially at the later phase of step depolarization, resulting in tonic firing at regular intervals ( Fig. 2A). This quality was also exhibited by M-cell homologs in control preparations. We summarized the effects by plotting spike number against depolarizing current intensity (Fig. 2B). Immature M cells exhibited spike adaptation with increasing ISI at the later phase, with adaptation completely abolished by XE991 ( Fig. 2A,C). XE991 treatment also affected the spike latency of M cells (Fig.  2E). These results indicate that XE991-sensitive K ϩ channels mainly contribute to phasic bursting of M cells at the early stage of development. However, some contribution of DTX-sensitive channels cannot be excluded because threshold current and spike latency  were additionally affected by the combined application of DTX and XE991 (Fig. 2D,E). Overall, these results suggest that Kv7 channels are already expressed in 2-dpf M cells, and that they are required for the establishment of phasic bursting with strong spike-frequency adaptation. Further, this raises the possibility that Kv7 channels are specifically expressed in M cells among M-series neurons.

Figure 2.
Effect of DTX and/or XE991 on phasic bursting of immature M cells. A, Firing response from M cells at 2 dpf. XE991 treatment caused phasic bursting M cells (control) to fire tonically (ϩXE991, ϩDTX ϩ XE991), whereas DTX treatment did not (except for a slight delay of spike onset). B, Summary graph of the blocker effect at 2 dpf showing that the number of APs elicited during 100-ms current pulses significantly increased after XE991 treatment. C, In controls, M cells at 2 dpf showed spike accommodation with a gradual increase in ISI, whereas regular spiking with constant ISI was observed after XE991 treatment. Note that DTX application had no additional effect. D, The threshold current was significantly reduced after combined DTX and XE991 treatment. E, Spike latency greatly increased after XE991 treatment. n.s., not significant. ‫ء‬p Ͻ 0.05, ‫‪p‬ءء‬ Ͻ 0.01, Mann-Whitney U test (control, n ϭ 6; ϩDTX, n ϭ 6; ϩXE991, n ϭ 10; ϩDTX ϩ XE991, n ϭ 7). Two-way repeated measures ANOVA was used for ISI. et al., 2014). However, due to incomplete genomic sequencing, we were only able to amplify a single full-length Kv7.4/kcnq4 cDNA segment by RT-PCR from these two genomic regions, in which N-and C-terminal Kv7.4/kcnq4 are separately located (see Materials and Methods). In situ hybridization in 2-and 5-dpf zebrafish revealed that among these genes, only Kv7.4/kcnq4 mRNA was specif-ically expressed in M cells, but this expression did not take place in the MiD2cm or MiD3cm (Fig. 3B). In addition, Kv7.4/kcnq4 mRNA was also expressed in some vestibular neurons located in the lateral M-cell region (Fig. 3B). Other Kv7 subunits were not detected in M-series neurons (data not shown), demonstrating Kv7.4 probe specificity. These results suggest that homomeric Kv7.4 channels are already specifically formed in M cells at 2 dpf, in contrast to Kv1.1 ␣-subunits (kcna1a/Kv1.1a), which appear in all M-series neurons but exhibit enhanced expression via Kv␤2 (kcnab2b/Kv␤2b) auxiliary subunits after 2 dpf (Watanabe et al., 2014).

Zebrafish Kv7.4 channels form low-threshold K ؉ currents with slow activation
As demonstrated, Kv7.4 and Kv1.1 channels appear to contribute differentially to the control of M-cell firing. To assess the electrophysiological kinetic differences between the zebrafish Kv7.4 and Kv1.1, we compared voltage-gated currents in Xenopus oocytes expressing either of these channels. Zebrafish Kv7.4 channels expressed in oocytes showed slowly-activating outward currents at command voltages above Ϫ60 mV (Fig. 4A). The threshold voltage was comparable to that of Kv1.1 (Fig. 4B,C; G/G max at Ϫ60 mV, p ϭ 0.357, Mann-Whitney U test; Kv7.4, n ϭ 20; Kv1.1a, n ϭ 32). However, the rise time of Kv7.4 channel activation was several ten-fold slower than that of Kv1.1 channels (Fig. 4D,E). Interestingly, when coexpressed with auxiliary Kv␤2 subunits, Kv1.1 channel activation was significantly accelerated, with a rise time that was ϳ2 ms faster above Ϫ40 mV (Fig.  4E) and an increased current amplitude ( Fig. 4B; Watanabe et al., 2014). That these channels have different kinetics is also demonstrated by the half activation voltage (V 1/2 ) and slope k parameters of the Boltzmann equation (Table 1). Therefore, the expression of slowlyactivating Kv7.4 channels in M cells appears to contribute to strong spike-frequency adaptation in phasic bursting at 2 dpf, and the expression of fast-activating Kv1.1 channels associated with Kv␤2 subunits establishes single spiking after 3 dpf.

Coordinated expression of Kv7.4 and Kv1.1 associated with Kv␤2 underlies developmental differentiation of M-cell excitability among homologous neurons
Next, we examined the role of coexpression of these two low-threshold K ϩ channels and Kv␤2 subunits on developmental acquisition of M-cell excitability by examining firing behavior in model cells. We focused on temporal patterns of firing in response to depolarization as well as the latency and threshold of firing. We represented the zebrafish Kv7.4 and Kv1.1 channels with and without Kv␤2 subunits via a basic H-H equation (Fig. 5) and integrated them into a control regular-spiking model (Fig.  6). Voltage-dependent open probability and timedependent kinetics of the channels were reproduced in the model (Fig. 5B-D, respectively) as observed in the Xenopus oocyte system (Fig. 4). It is nearly impossible to precisely determine all of the ionic conductances in M cells via voltage-clamp recording due to large leak cur-  (Figs. 1, 2), is based on that of mouse cochlear neurons (Rothman and Manis, 2003b). Mouse cochlear neurons express voltage-gated Na ϩ channels, high-threshold K ϩ channels, and A-type K ϩ channels, and fire tonically in response to step depolarization [ Fig    In addition to the typical examples above, we comprehensively analyzed the relationship between the amount of low-threshold K ϩ conductance and model cell excitability, as shown in Figure 7. While recruitment of Kv1.1 conductance in tonic-firing cells increased the threshold current and shortened the firing onset latency, especially when coexpressed with Kv␤2, the addition of Kv7.4 conductance had a relatively small effect (Fig. 7A,B). Increas-ing Kv7.4 conductance resulted in a gradual suppression of bursting (Fig. 7C,D) accompanied by strong spikefrequency adaptation ( Fig. 7E; ordinate), while increasing Kv1.1 conductance had a relatively small effect ( Fig. 7E; abscissa). Thus, Kv7.4 and Kv1.1 channels play distinct firing property roles, while Kv␤2 subunits enhance the effect of Kv1.1 channels. It should be noted that Kv␤2 coexpression had a significant impact on phasic firing, even at smaller conductances.
We represented behaviors of different types of voltagegated channels during phasic firing of M cells using the model cell (Fig. 8). During phasic bursting, Kv1.1 channels are periodically activated from the onset of current injection, whereas activation of Kv7.4 channels is gradually increased at the later phase of repetitive firing (Fig. 8A,B; red vs blue). Thus, fast activation of Kv1.1 channels with low density presumably increases firing frequency rather than suppressing later firing, whereas slow activation of Kv7.4 channels mainly affects spike-frequency adaptation. In the case of single spiking, activation of Kv1.1 with Kv␤2 rose rapidly immediately at onset, and maintained high levels until the end of spiking. In contrast, Kv7.4 activation increased gradually for phasic bursting (Fig.  8C,D). These data suggest that the fast activation and high conductance of Kv1.1 channels coexpressed with Kv␤2 strongly inhibits bursting after the first AP. Further, at juxta-threshold voltages, Kv1.1 with Kv␤2 also contributed to fast activation of voltage-gated Na ϩ conductance, whereas Kv7.4 did not (Fig. 8E-H). The differing contribution of Kv1.1 channels with Kv␤2 and Kv7.4 channels on Na ϩ channels may underlie the differences in spike latency of M cells during DTX and XE991 treatment.
Finally, we estimated a role for the expression of the two K ϩ channels on M cells in the firing response to auditory inputs by applying repetitive short-pulse depolarizing currents (0.5 ms, 0 -3 nA) at 500 Hz (Fig. 9). A B Figure 6. M-cell firing model integrating low-threshold K ϩ channels in response to long step currents. A, B, Representative current-clamp simulation of a model neuron integrating two types of low-threshold K ϩ channels associated with Kv␤2 subunits into a basic tonic-firing model. Long step currents of 1T, 1.5T, and 2T were injected. Maximum conductances (g Kv1.1 , g Kv7.4 ; A) or ( g Kv1.1ϩKv␤2 ,g Kv7.4 ; B) are indicated in parentheses. Bold italic numbers represent conductances associated with Kv␤2. Expression of low-threshold K ϩ channels alters the tonic-firing model to exhibit phasic bursting with long or short firing duration and single spiking, as observed with in vivo M-cell recordings (Figs. 1, 2).
The effects of integration of these channels on threshold current and latency of the first AP had the same tendency for both M cells in vivo (Fig. 1, 2) and those in the simulation with step current injection (Figs. 6, 9E,F ). Taken together, these results demonstrate that coordinated expression of slowly-activating Kv7.4 channels, fast-activating Kv1.1 channels, and enhancement by auxiliary Kv␤2 subunits produces heterogeneity of phasic firing in M cells with different developmental stages, thus further differentiating these neurons from homologous counterparts.

Discussion
Using the zebrafish as a model to understand how unique single spiking is produced based on a tonic-firing property, we have identified ionic mechanisms underlying
developmental differentiation of M-cell excitability among segmentally homologous hindbrain neurons. Singlespiking mature M cells begin to fire tonically, as observed in the M-cell homologs, when two types of low-threshold K ϩ currents are blocked (Fig. 1), indicating that the ionic basis for tonic firing is conserved in the M series. Conversely, these low-threshold K ϩ currents are responsible for differentiation of M-cell excitability from that of the M-cell homologs. In contrast to common Kv1.1 channel expression among the M series (Watanabe et al., 2014), slowly-activating Kv7.4 channels and auxiliary Kv␤2 subunits that accelerate and increase Kv1.1 currents are expressed specifically in M cells from embryonic and later developmental stages, respectively. The expression of Kv7.4 channels and Kv␤2 associated-Kv1.1 channels during development plays different roles in phasic firing such that the combination of these activities induces M cells to acquire their characteristic single-spiking pattern. Other mechanisms are known to generate phasic or single spiking properties. A change in intrinsic membrane properties from tonic to phasic firing has been well investigated in spiny lobster culture preparations of stomatogastric ganglion neurons, which can be accounted for by increased Ca 2ϩ currents and decreased K ϩ currents (Turrigiano et al., 1994;Turrigiano et al., 1995). In rat ventral horn neurons, pharmacological blocking of persistent Na ϩ currents causes tonic-firing neurons to fire a single spike (Theiss et al., 2007). Simulation analysis of RSNs in frog tadpoles suggests that single spiking evoked by step current injection is not a cellular property but the consequence of a shunting effect resulting from electrical cou-pling via gap-junction connections among a neural population (Hull et al., 2015). Because we did not test the contribution of these currents to M-cell firing, we cannot exclude the possibility that Ca 2ϩ currents, persistent Na ϩ currents, and gap-junction connections might underlie M-cell firing properties. However, even if they exist in M cells, their contribution will presumably only have a minimal effect on generating single spiking of M cells because of the necessity and sufficiency of coexpression of the two low-threshold K ϩ channels for single spiking.

Coexpression of Kv7.4 and Kv1.1 channels in vertebrate sensory neurons
Which neurons express both Kv7 and Kv1 channels, like M cells, in vertebrates? In zebrafish, vestibular nucleus neurons express Kv7.4 and Kv1.1 channels from 2 dpf, and Kv␤2 subunits from 5 dpf ( Fig. 3B; Watanabe et al., 2014). In rat vestibular ganglion neurons both Kv7 and Kv1 channels are required for single-spiking vestibular afferent neurons (Kalluri et al., 2010), which express K ϩ channels including Kv7.4 and/or Kv1.1 (Hurley et al., 2006;Lysakowski et al., 2011). Interestingly, phasic spiral ganglion neurons located at the base of the cochlea exhibit stronger expression of Kv7.4 and/or Kv1.1 than tonic neurons located at the apex (Beisel et al., 2000;Adamson et al., 2002). Furthermore, coexpression of Kv7 and Kv1 channels associated with Kv␤ subunits has been indirectly suggested not only in vestibular and auditory neurons but also phasic trigeminal ganglia neurons and dorsal root ganglia neurons, as well as their secondary sensory neurons, in mice and rats (Rhodes et al., 1996; A B C D E F G H Figure 8. Na ϩ and K ϩ conductance mechanisms for producing characteristic phasic properties. A-D, Representative dynamics of voltage-gated Na ϩ and K ϩ channel conductance for g Nav (black), g high (green), g A (gray), g Kv7.4 (blue), and g Kv1.1 or g Kv1.1ϩKv␤2 (red) during current-clamp stimulation. These conductances were normalized to maximum values of Na ϩ or among K ϩ conductance at 1.5T using a 2-dpf (A, B) and 4-dpf (C, D) M-cell model, shown in Figure 7A,  E). The corresponding Na ϩ and K ϩ conductance in each model during subthreshold firing (F-H). Na ϩ conductance became much larger after fast activation of K ϩ conductance (g Kv1.1ϩKv␤2 ; F, G) compared with slow activation of K ϩ conductance ( g Kv7.4 ; H). Kharkovets et al., 2000;Rasband et al., 2001;Caminos et al., 2005;Heidenreich et al., 2012). Thus, coordinated expression of two types of low-threshold K ϩ channels is observed in the vertebrate sensory system. Kv7.4 gene mutations in humans and mice lead to impairments of hearing and touch sensation caused by dysfunction and progressive degeneration of outer hair cells (Kharkovets et al., 2006) or disruption of phasic neurons in trigeminal and dorsal root ganglia (Heidenreich et al., 2012). Kv1.1 mutations in humans and knock-out in mice are known to cause episodic ataxia type 1 associated with epilepsy, which results from dysfunction of the cerebellum, hippocampus, cortex, and peripheral nervous system (Smart et al., 1998;D'Adamo et al., 2015). In the auditory system of Kv1.1 knock-out mice, phasic firing of the ventral cochlear nucleus and medial nucleus of trapezoid body neurons impairs temporal spiking precision and increases firing frequency (Brew et al., 2003;Kopp-Scheinpflug et al., 2003;Gittelman and Tempel, 2006), causing reduced performance of interaural intensity difference in the lateral superior olive (Karcz et al., 2011).
These findings demonstrate that losing either of two lowthreshold K ϩ channels in sensory neurons results in sensory dysfunction by altering phasic neurons to display hyperexcitability or loss of temporal precision. Therefore, coordinated expression in these neurons appears to play a crucial role in detecting phasic or onset timing and raising the firing threshold against sensory stimuli, as likely in the case of M cells. To reveal the physiologic role of these channels in M cell-initiated escape behavior, further behavioral experiments combined with M cellspecific gene manipulation of these channels is needed.
that the suppression of repetitive firing by Kv1.1 channels alone requires large conductance (Fig. 7C,D). Slowly-activating Kv7.4 channels work primarily via spike-frequency adaptation to produce shorter bursts at frequencies that increase with injected currents (Figs. 6,  7). This role for Kv7/KCNQ channels in spike-frequency adaptation has been emphasized in rat sympathetic neurons and mouse CA1 pyramidal neurons, in which heteromeric channels composed of Kv7.2/KCNQ2 and Kv7.3/ KCNQ3 subunits are expressed Chung et al., 2006;Soh et al., 2014). Phasic-bursting neurons in the mouse cerebral cortex also exhibit firing with frequencies proportional to input intensity and longer onset spike latency, i.e., over 10 ms (Soh et al., 2014). Thus, neurons expressing only Kv7 channels in addition to conventional voltage-gated Na ϩ and high-threshold K ϩ channels burst phasically with spike-frequency adaptation, as is the case with M cells at 2 dpf.
Here, for the first time, we have shown the necessity of integrated expression of Kv7.4 channels and Kv1.1 channels with Kv␤2 auxiliary subunits to produce single spiking as the most extreme pattern of phasic firing. Further, fast outward currents through the Kv1.1 channels substantially increased the threshold current for spiking, as reported before in rat vestibular neurons (Kalluri et al., 2010). Kv1.1 channels associated with Kv␤2 shortened the spike latency (Figs. 6,7,9). Therefore, neurons expressing Kv1 channels associated with Kv␤2 encode onset-timing information in response to relatively large inputs, but still burst at an input intensity-dependent frequency [Figs. 6B, 9B ; phasic busting (short)]. Together, neurons expressing both Kv7 and Kv␤2-associated Kv1 channels extract precise onset-timing information without duration and intensity, as is the case with M cells and rat vestibular neurons (Kohashi and Oda, 2008;Kalluri et al., 2010).
The effect of combined DTX and XE991 treatment was greater than either treatment at 2 and 4 dpf in M cells (Figs. 1F,2E), and this phenomenon was replicated in model cells (Figs. 7B,9F ). Thus, in addition to Kv1, Kv7 currents also shorten spike latency but the effect is much smaller than Kv1 currents. This difference may be mainly due to the distinct impact on Na ϩ conductance activation (Fig. 8E-H). Faster activation of Na ϩ conductance may be attributed to reduced membrane resistance and a shorter time constant of the depolarizing phase before spiking, which is caused by rapidly increasing K ϩ currents from Kv1.1 channels associated with Kv␤2 rather than slowly-activating Kv7.4 channels. Conversely, without low-threshold K ϩ currents, the membrane depolarization time constant becomes much longer (as passive membranes demonstrate), and then a slower depolarization only weakly activates Na ϩ conductance, resulting in a considerably longer spike latency of tonic firing.

Developmental increases in Kv7.4 and Kv1.1 conductance
Our comprehensive simulation produced pairs of conductances for low-threshold K ϩ channels that best explain our experimental data, with a relatively higher conductance for Kv7.4 versus Kv1.1 (Fig. 6). The higher Kv7.4 conductance presumably originated from incomplete modification of basic H-H models of g Nav , g high , and g A based on cochlear neurons (Rothman and Manis, 2003b) along with the differences in voltage range and rise time for activation of Kv7.4 versus Kv1.1 channels (Fig. 4). Our modifications of g Nav , g high , and g A were arbitrarily adjusted by fitting model firing and AP wave form to experimental data of 4-dpf M cells, which somehow affected our model results throughout development. A comparison between the zebrafish Kv7.4 and the other Kv7 channels in the mouse indicated that zebrafish Kv7.4 channels have more depolarized V 1/2 parameters compared with mouse KCNQ1, KCNQ2 and KCNQ4 channels (Nakajo and Kubo, 2005). This suggests that these channels play different roles in zebrafish M cells and mouse neurons.
Our M-cell model indicated that increasing the Kv7.4 and Kv1.1 currents by 2 and 4.7 times, respectively, sufficiently reproduced M-cell development and data from previous pharmacological studies. Increases in Kv1.1 currents can be explained by enhanced surface expression associated with Kv␤2, as observed in the Xenopus oocyte system (Fig. 5), and/or possibly by developmental accumulation of gene expression in M cells. Indeed, this increase in gene expression corresponds to the development of auditory and vestibular neurons in mammals and birds, and these neurons can alter intrinsic firing properties in response to increases in Kv1 currents and expression levels (Gamkrelidze et al., 1998;Nakamura and Takahashi, 2007;Iwasaki et al., 2008;Iwasaki et al., 2012). Alternatively, the molecular basis for the developmental increases in Kv7.4 conductance may not simply be explained by an increase in gene expression, because strong expression was already observed at 2 dpf (Fig. 3B).
What possible mechanisms underlie the increase of Kv7.4 currents? Alternative splicing variants in mouse Kv7.4/KCNQ4 (variants 1-4) exhibit distinct current amplitudes when expressed in Chinese hamster ovarian cells (Xu et al., 2007). These isoforms are differentially expressed in inner hair cells, afferent spiral ganglion neurons (Beisel et al., 2005), and vestibular ganglion neurons (Rocha-Sanchez et al., 2007), suggesting nervous system regulation of conductance by Kv7.4 isoforms. Moreover, mammalian Kv7 conductance is modulated and inhibited by phosphorylation of C-terminal sites via protein kinase C (PKC; Hoshi et al., 2003;Nakajo and Kubo, 2005). Further, membrane surface expression and kinetics are regulated by association with potassium voltage-gated channel subfamily E (KCNE) ␤-subunits (KCNE1-5; Strutz-Seebohm et al., 2006;Pongs and Schwarz, 2010). In this study, we found that the zebrafish Kv7.4 is comparable to a short-type variant of the mouse homolog, with conserved PKC-phosphorylated residues in the C terminus, shown by amino acid sequence alignment across vertebrates (Fig. 3A;Nakajo and Kubo, 2005). M cells express PKC␥ protein, at least from 2 dpf (Patten et al., 2007). Zebrafish KCNE1 and KCNE4 subunits are present in the zebrafish genome database (Zv9), but not all homologs have been identified. Further studies are required to examine possible changes in splicing isoform expression, and to determine if regulation of Kv7.4 channels via PKC or KCNE ␤-subunits is involved in M-cell excitability during development.
In conclusion, M-series neurons may represent the simplest model for explaining the developmental establishment of different phasic-firing properties from a tonicfiring property in a neural population via coordinated expression of low-threshold K ϩ channel subunits. Because Kv1, Kv␤, and Kv7 gene families are widely expressed in the vertebrate nervous system, it is likely that distinct activation-kinetics and expression or conductance levels of these low-threshold K ϩ channels play important roles in producing heterogeneity of phasicfiring properties.