Essential Role of Somatic Kv2 Channels in High-Frequency Firing in Cartwheel Cells of the Dorsal Cochlear Nucleus

Abstract Among all voltage-gated potassium (Kv) channels, Kv2 channels are the most widely expressed in the mammalian brain. However, studying Kv2 in neurons has been challenging because of a lack of high-selective blockers. Recently, a peptide toxin, guangxitoxin-1E (GxTX), has been identified as a specific inhibitor of Kv2, thus facilitating the study of Kv2 in neurons. The mammalian dorsal cochlear nucleus (DCN) integrates auditory and somatosensory information. In the DCN, cartwheel inhibitory interneurons receive excitatory synaptic inputs from parallel fibers conveying somatosensory information. The activation of parallel fibers drives action potentials in the cartwheel cells up to 130 Hz in vivo, and the excitation of cartwheel cells leads to the strong inhibition of principal cells. Therefore, cartwheel cells play crucial roles in monaural sound localization and cancelling detection of self-generated sounds. However, how Kv2 controls the high-frequency firing in cartwheel cells is unknown. In this study, we performed immunofluorescence labeling with anti-Kv2.1 and anti-Kv2.2 antibodies using fixed mouse brainstem slice preparations. The results revealed that Kv2.1 and Kv2.2 were largely present on the cartwheel cell body membrane but not on the axon initial segment (AIS) nor the proximal dendrite. Whole-cell patch-clamp recordings using mouse brainstem slice preparation and GxTX demonstrated that blockade of Kv2 induced failure of parallel fiber-induced action potentials when parallel fibers were stimulated at high frequencies (30–100 Hz). Thus, somatic Kv2 in cartwheel cells regulates the action potentials in a frequency-dependent manner and may play important roles in the DCN function.


Introduction
Neurons express a wide variety of voltage-gated potassium (Kv) channels that feature different voltage dependence and kinetics, thereby endowing neurons with a wide range of firing patterns (Hille, 2001;Vacher et al., 2008;Johnston et al., 2010). The Kv channel family can be divided into several subfamilies based on nucleotide sequence similarity and function. In the mammalian brain, messenger RNA expression studies and antibody-based protein detection indicate that Kv2 channels are the most widely expressed voltage-gated K channels in terms of tissue distribution (Trimmer, 1991;Hwang et al., 1993;Mandikian et al., 2014). Kv2 channels consist of Kv2.1 and Kv2.2, and are present not only in cortical, hippocampal, and a-motoneurons but also in retinal bipolar cells, cardiac myocytes, vascular and gastrointestinal smooth muscle, and pancreatic b cells (Johnson et al., 2019). Physiologically, Kv2 channels produce delayed-rectifier currents, and Kv2.1 channels have a high activation threshold of -15 mV (Johnston et al., 2010;Johnson et al., 2019). Indeed, delayed-rectifier currents in various neurons are generally because of Kv2 channels (Murakoshi and Trimmer, 1999;Malin and Nerbonne, 2002;Guan et al., 2007;Johnston et al., 2008;Tong et al., 2013). However, the study of Kv2-mediated current in neurons has been challenging because of the limited availability of high-selective blockers. In previous studies, Kv2 channels were blocked by an anti-Kv2 antibody, gene expression of dominant-negative Kv2 channels, or non-specific Kv2 channel blockers, or were eliminated by the knock-out mice. In 2006, a peptide toxin, guangxitoxin-1E (GxTX), extracted from the venom of the Chinese tarantula Plesiophrictus guangxiensis, has been identified as a potent and specific inhibitor of Kv2.1 and Kv2.2 channels, with a half-blocking concentration of 2-5 nM (Herrington et al., 2006;Herrington, 2007). Using this toxin, the physiological roles of Kv2 channels have been explored in the superior cervical ganglion neurons, CA1 pyramidal neurons, and the entorhinal cortex layer II stellate cells (Liu and Bean, 2014;Kimm et al., 2015;Hönigsperger et al., 2017). Interestingly, Kv2.1 expression in some types of cerebral neurons coincides with the presence of clustered ryanodine receptors in the cell body, suggesting that Kv2.1 expression can be observed in neurons expressing clustered ryanodine receptors in other brain regions such as the brainstem (Mandikian et al., 2014).
The mammalian brainstem contains several auditory nuclei, and among them, the dorsal cochlear nucleus (DCN) integrates auditory inputs and multimodal ones, including somatosensory, vestibular, and higher-level auditory information (Oertel and Young, 2004). Auditory inputs are conveyed by auditory nerve fibers and multimodal inputs by parallel fibers, which are axons of excitatory granule cells in the cochlear nuclei. The DCN plays several crucial roles in the monaural sound localization and cancelling detection of self-generated sounds (e.g., sound induced by licking behavior; May, 2000;Singla et al., 2017). Singla et al. (2017) reported that during licking behavior in mice, complex-spiking single units (i.e., firing brief high-frequency bursts of spikes) were activated in the DCN; these complex-spiking neurons likely correspond to cartwheel inhibitory interneurons (Manis et al., 1994;Davis and Young, 1997;Tzounopoulos et al., 2004;Roberts et al., 2008;Ma and Brenowitz, 2012). Cartwheel cells are medium-sized GABAergic/glycinergic neurons which fire mixtures of simple and complex action potentials (Manis et al., 1994;Golding and Oertel, 1996;Kim and Trussell, 2007;Roberts et al., 2008;Roberts and Trussell, 2010). Moreover, cartwheel cells have profuse spiny dendrites that receive excitatory synaptic inputs from parallel fibers (Wouterlood and Mugnaini, 1984). The activation of parallel fibers drives action potentials in cartwheel cells up to 130 Hz in vivo, and the excitation of cartwheel cells leads to the strong inhibition of principal cells (Davis and Young, 1997;Roberts and Trussell, 2010). Recently, it has been shown that cartwheel cells express clustered ryanodine receptors abundantly at putative subsurface cisterns immediately beneath the somatic membrane (Yang et al., 2016;Irie and Trussell, 2017). Considering that clustered ryanodine receptor-expressing neurons have Kv2.1 channels (Mandikian et al., 2014), it is possible that cartwheel cells also express Kv2.1 channels. However, where Kv2 channels are expressed in cartwheel cells in terms of subcellular localization and how Kv2 channels control high-frequency firing are unknown.
In this study, we performed immunofluorescence labeling using anti-Kv2.1, anti-Kv2.2 channel, and other antibodies in fixed mouse brainstem slice preparations. We also performed electrophysiological recordings from cartwheel cells in acute brainstem slice preparations using the whole-cell patch-clamp method and GxTX, a specific Kv2 channel blocker. Immunofluorescence labeling revealed that Kv2.1 and Kv2.2 channels largely localized on the membrane of the cell body, not at the axon initial segment (AIS) nor the dendrite. Furthermore, blocking Kv2 channels by GxTX induced failures of parallel fiber-induced action potential generations only when parallel fibers were stimulated at high frequency (30-100 Hz). Thus, somatic Kv2 channels regulate action potentials in a frequency-dependent manner in cartwheel cells.

Animals
All animal care and handling procedures used in this study were approved by the Institutional Animal Care and Use Committee. Two male (age: 27 and 30 d) and one female (age: 34 d) ICR mice were used for fluorescence immunohistochemistry. ICR mice of both sexes (81 animals; age: 21-34 d) were used for electrophysiological recordings.

Fluorescence immunohistochemistry and confocal imaging
Mice were deeply anesthetized with isoflurane and fixed by transcardial perfusion of PBS, followed by 4% (wt/vol) ice-cold formaldehyde in PBS. The brains were removed from skulls, postfixed in formaldehyde at room temperature for 30 min, and cryoprotected in 30% (w/w) sucrose-containing PBS overnight at 4°C. Using a cryostat (CM3050S, Leica Microsystems), the cerebral hemispheres and brainstems were sliced into 30-mm-thick coronal sections. The sections were washed with PBS containing 0.3% (vol/vol) Triton X-100 (PBS-X), and then incubated overnight at 4°C in incubation buffer [1% (vol/ vol) goat and donkey serum, 0.25% (wt/vol) l -carrageenan, and 0.02% (wt/vol) sodium azide in PBS-X] containing the following four kinds of primary antibodies: mouse IgG2a monoclonal anti-ankyrin-G antibody (1 mg/ ml, clone N160/36, UC Davis/NIH NeuroMab Facility; King et al., 2014;Yang et al., 2016), mouse IgG3 monoclonal anti-Kv2.1 antibody (1 mg/ml, clone L80/21, UC Davis/NIH NeuroMab Facility; Bishop et al., 2015), mouse IgG2b monoclonal anti-Kv2.2 antibody (1 mg/ml, clone N372B/60, UC Davis/NIH NeuroMab Facility; Kirmiz et al., 2018b), and mouse IgG1 monoclonal anti-ryanodine receptor antibody (1 mg/ml, clone 34C, Thermo Fisher Scientific; this antibody detects all ryanodine receptor isoforms in mouse tissue; King et al., 2014;Yang et al., 2016;Irie and Trussell, 2017). The specificity of the anti-Kv2.1 antibody was validated by immunoblotting against wildtype and Kv2.1-KO mouse brain samples (datasheet). The anti-Kv2.2 antibody was validated by immunofluorescence labeling of wild-type and Kv2.2-KO mouse brain slices (datasheet). A detailed description of the primary antibodies is provided in Table 1. Furthermore, slices were washed with PBS-X, and then incubated for 4 h at room temperature in the incubation buffer containing the following secondary antibodies at a 1:500 dilution: CF405S goat anti-mouse IgG2a (20 381, Biotium), Alexa Fluor (AF) 488 goat anti-mouse IgG3 (A21151, Thermo Fisher Scientific), CF568 goat anti-mouse IgG2b (20 268, Biotium), and AF 647 goat anti-mouse IgG1 (A21240, Thermo Fisher Scientific). The slices were then washed with PBS-X, and finally coverslipped with Fluoromount-G (Southern Biotech). Intracellular labeling combined with fluorescence immunohistochemistry was performed as follows: cartwheel cells whose cell bodies exist in the surface of the acute brain slices (200-mm thickness) were whole cell patchclamped with the pipette containing 0.5% (w/v) biocytin (Sigma-Aldrich) for .10 min. Slices were then fixed in 4% formaldehyde for 30 min at room temperature. Slices were washed three times with PBS-X and incubated with mouse IgG3 monoclonal anti-Kv2.1 antibody, and mouse IgG2b monoclonal anti-Kv2.2 antibody overnight at 4°C. Slices were washed with PBS-X, and then incubated for 4 h at room temperature in the incubation buffer containing AF 488 goat anti-mouse IgG3, CF568 goat anti-mouse IgG2b, and AF 647 streptavidin (4 mg/ml, S21374, Thermo Fisher Scientific). The slices were then washed, and coverslipped. Considering that the thickness of the slices, images of the immunoreactivities were taken from the surface of the slices (within 10 mm in depth).
Immunofluorescence images were acquired using a confocal microscope (A1R, Nikon) with the following appropriate settings: CF405S (excitation, 405-nm laser; emission, 425-to 475-nm bandpass filter), AF 488 (excitation, 488 nm; emission, 500-550 nm), CF568 (excitation, 561 nm; emission, 570-620 nm), and AF 647 (excitation, 640 nm; emission, 662-737 nm). Images were obtained with a 60 Â/1.4 numerical aperture oil-immersion objective lens, and the confocal pinhole size was 1.0 airy unit. Then, Z-stack images of each dye were taken sequentially, and the image stacks were deconvoluted to remove out-of-focus signals with NIS Elements software (Nikon). The profiles of the signal intensities along cell membranes were measured by drawing 0.042 mm in width lines with the aid of the ryanodine signal, which overlaps with cell membranes of cartwheel cells (Irie and Trussell, 2017). Lastly, immunofluorescent data were analyzed using Fiji software (https://fiji.sc/).

Acute brainstem slice preparation, electrophysiological recordings, and data analysis
Acute brainstem slices containing the DCN were prepared first by anesthetizing the mice deeply with isoflurane and decapitating them. Then, parasagittal slices of brain stems (200-mm thickness) were prepared using a micro slicer (PRO7, Dosaka) in ice-cold, cutting solution containing the following: 93 mM N-methyl-D-glucamine, 2.5 mM KCl, 1.2 mM NaH 2 PO 4 , 30 mM NaHCO 3 , 20 mM , 20 mM NaHCO 3 , 3 mM HEPES-Na, 10 mM glucose, 0.4 mM ascorbic acid, 3 mM myo-inositol, and 2 mM sodium pyruvate, and bubbled with 5% CO 2 /95%O 2 . The slices were then incubated for 40 min before use. In electrophysiological recordings, ascorbic acid was omitted from the ACSF. The ACSF was supplemented with fast synaptic blockers [10 mM NBQX (Alomone Labs), 5 mM MK-801, 1 mM strychnine, and 100 mM picrotoxin] unless otherwise stated. The brainstem slices were transferred to a recording chamber and continuously perfused at 3 ml/min with ACSF at 33-34°C using a peristaltic pump (Minipuls 3, Gilson) unless otherwise stated. Next, neurons were visualized with an upright microscope (BX51WI, Olympus) equipped with a 60 Â/1.0 numerical aperture water-immersion objective lens and near infrared-CCD camera (C3077-79; Hamamatsu Photonics). Cartwheel cells were identified based on their location within the DCN, somatic size and morphology, and characteristic responses to current injections (Golding and Oertel, 1997;Tzounopoulos et al., 2004;Kim and Trussell, 2007;Roberts et al., 2008). Lastly, data were collected with Molecular Devices hardware and software (Multiclamp 700B, Digidata 1440A, and Clampex 10.3). Signals were low-pass filtered at 6 kHz and digitized at 20-100 kHz. Whole-cell patch-clamp recordings were made using patch pipettes made from borosilicate glass capillaries (1B150F-4, WPI), which have a resistance of 2.5-3.5 MV when filled with a potassium gluconate-based internal solution containing the following: 125 mM K-gluconate, 10 mM KCl, 0.1 mM EGTA, 2 mM Mg-ATP, 3 mM Na 2 -ATP, 0.3 mM Na 2 -GTP, 13 mM Na 2 -phosphocreatine, and 10 mM HEPES. The pH was adjusted to 7.3 using KOH. In some experiments, 0.5% (w/v) biocytin was added to the pipette solution. In the current clamp recordings, series resistance was compensated using bridge balance, pipette capacitance was neutralized, and resting membrane potentials were adjusted to suppress spontaneous firing by injecting negative bias currents. When the voltage clamp recordings were made, series resistance was compensated by 60-80%, and the resistance was frequently monitored by turning off the compensation and applying 5-mV short step pulses (30-ms duration). If the series resistance changed by .10%, the records were rejected. Tail currents were measured at room temperature (23-24°C ) using the following protocol: outward currents were evoked by voltage steps from -80-mV holding potential up to 30 mV in 10-mV increments for 50 ms, followed by repolarization to -50 mV for 250 ms. GxTX-sensitive tail current amplitudes were normalized to the maximal current, and normalized conductance (G) was plotted as a function of voltage. G was obtained by dividing the peak tail current by the electrochemical driving force: The activation curves (G/G max ) were fitted with the Boltzmann function, G/G max = 1/[1 1 exp (V 1/2 -V)/k], where V 1/2 is the voltage at which 50% of the channels are activated, and k is a slope factor. The ACSF for outward current was further supplemented with 0.5 mM TTX (Alomone Labs), 100 nM apamin (a SK channel blocker, Alomone Labs), and 1 mM penitrem A (a BK channel blocker, Alomone Labs). To remove inward currents induced by voltage-gated calcium channels, CaCl 2 in the ACSF was excluded and replaced with equimolar MgCl 2 , and 0.25 mM EGTA-Na (pH adjusted to 7.4 with NaOH) was added. Activation and deactivation of the GxTX-sensitive current were fit with a single exponential function to obtain decay time constants. Current clamp was used to assess the effects of GxTX on intrinsic membrane properties. As cartwheel cells exhibit spontaneous firing in vitro (Manis et al., 1994;Kim and Trussell, 2007;Bender et al., 2012), the resting membrane potential was adjusted to around À80 mV at the beginning of the experiment to suppress spontaneous firing by injecting negative bias current. The amplitude of the bias current was kept constant throughout the experiment. Input resistance was measured by applying a small hyperpolarizing current pulse (À50 pA, 300-ms duration). For the measurement of action potential properties, simple or complex spikes were induced by brief, strong current injection (1-ms duration, 100-pA increment, up to 2500 pA). The threshold current was defined as the current amplitude that evoked action potential for the first time, while the fast afterhyperpolarization (fAHP) was defined as the most negative voltage between first and second action potentials in complex spikes. The half-width of action potential was measured at the potential between the threshold and action potential peak. In complex-spiking neurons, the threshold potential, action potential amplitude, and the maximum rate of spike rise and decay of action potentials were calculated from the waveform of the first action potential. EPSCs were recorded with a CH 3 CsO 3 S-based internal solution containing the following: 87 mM CH 3 CsO 3 S, 20 mM CsCl, 5 mM CsF, 10 mM TEA, 0.1 mM EGTA, 2 mM Mg-ATP, 3 mM Na 2 -ATP, 0.3 mM Na 2 -GTP, 13 mM Na 2 -phosphocreatine, 2 mM QX-314 Cl (Alomone Labs), and 10 mM HEPES. The pH of the solution was adjusted to 7.3 using CsOH. Moreover, EPSCs were recorded in the presence of 1 mM strychnine and 100 mM picrotoxin, and the membrane potentials were held at À80 mV under the voltage clamp configuration. EPSCs were evoked by electrical stimulation of parallel fibers with an ACSF-filled patch pipette (Tzounopoulos et al., 2004;Roberts and Trussell, 2010). The stimulus electrode was driven by the combination of an isolator (SS-203J, Nihon Kohden) and an electronic stimulator (SEN-7203, Nihon Kohden). The stimulus intensity was 34-89 V (100-ms duration). A train of EPSCs was evoked, and averages of five traces were used for the analysis. Synaptically evoked action potentials were also induced by electrical stimulation of parallel fibers. We defined that successful action potential has maximum rate of rise .30 V/s and peak amplitude higher than À15 mV. This is because in cartwheel cells, depolarization having 30-V/s maximum rate of rise elicits IPSCs at ;50% failure rate (Roberts et al., 2008). Moreover, in the cells, action potentials which have peak amplitude lower than approximately À15 mV cannot trigger transmitter release (Roberts et al., 2008). Maximum rate of rise was obtained by differentiating voltage response evoked by parallel fibers stimulation.
Data were analyzed using Clampfit 10.3 software (Molecular Devices) and Igor Pro 6 software (Wavemetrics) with the added import functionality provided by ReadPclamp XOP of the NeuroMatic software package (http://www. neuromatic.thinkrandom.com/; Rothman and Silver, 2018). Liquid junction potentials (K-gluconate-based, À10 mV; CsCl-based, À5 mV) were corrected offline. All data are provided as mean 6 SEM unless otherwise stated. Numbers in parentheses in figures and n in the text and tables indicate the number of replications (cells). Statistical significance was tested using paired t tests unless otherwise stated (significance, p , 0.05). GraphPad Prism 5 (GraphPad Software) was used for the statistical analysis.

Peptide blocker application
When the peptide blocker GxTX (Alomone Labs) was used, 0.1 mg/ml bovine serum albumin was added to all ACSF to reduce non-specific binding. The final concentration of GxTX was 100 nM. GxTX was perfused for at least 5 min through recirculation with a peristaltic pump (Minipuls 3) before data recording.
Kv2.1 clusters are juxtaposed to clustered ryanodine receptors in specific neurons, including CA1 pyramidal neurons and striatal medium spiny neurons (Mandikian et al., 2014). To examine whether Kv2.1 clusters in cartwheel cells overlapped with ryanodine receptors and/or Kv2.2, cartwheel cells were immunolabeled with anti-Kv2.1, anti-Kv2.2, anti-ryanodine receptor, and anti-ankyrin-G antibodies (Figs. 1B,C, 2A,B). Figure 1B illustrates the immunofluorescence labeling of Kv2.1, Kv2.2, and ryanodine receptors. Cartwheel cells were easily identified in the DCN with the aid of the puncta of the ryanodine receptor along the cell membrane ( Fig. 1Biv; Yang et al., 2016;Irie and Trussell, 2017). Puncta of Kv2.2 signals were observed in cartwheel cells (Fig. 1Biii, arrowheads and arrows). On the other hand, those of Kv2.1 were detected most of cartwheel cells (Fig. 1Bii, arrowheads), but some of them were immunonegative to Kv2.1 (Fig. 1Bii, arrows). We measured the immunopositive rate of Kv2.1 and Kv2.2 in ryanodine-positive cartwheel cells (Table 2). Indeed, ;25% of cartwheel cells' bodies did not show clear Kv.2.1 immunoreactivity, but almost all of cartwheel cells' bodies were Kv2.2-positive. The profiles of signal intensity along the cell body's membrane were measured by drawing linear regions of interest along the cell membrane, guided by the ryanodine receptor signal (Fig. 1 Ci-Civ). The signal intensities of Kv2.1 and Kv2.2, or those of Kv2.1 and ryanodine receptors along the linear regions of interest are plotted in Figure 1Cv,Cvi, respectively. In Figure 1Cv, the profile had one peak that overlapped between the Kv2.1 and Kv2.2 signals (a hash mark in Fig.  1Cv). In Figure 1Cvi, the profile had three peaks that overlapped between the Kv2.1 and ryanodine receptor signals (hash marks in Fig. 1Cvi). We also labeled hippocampal slices with anti-Kv2.1 and anti-ryanodine receptor antibodies as the positive control, and confirmed that Kv2.1 clusters were juxtaposed to the puncta of ryanodine receptors on the membrane of the cell bodies' of CA1 pyramidal neurons (data not shown), as reported previously (Kirmiz et al., 2018a). These results indicate that cartwheel cell bodies express Kv2.2 and most of them also have Kv2.1, and that the puncta of Kv2.1 do not overlap completely with those of Kv2.2 or that of ryanodine receptors.
Using multiple immunofluorescence labeling, we explored the presence of Kv2.1 and Kv2.2 signals on AIS, which is a specialized membrane region in the axons of neurons where action potentials are initiated (Kole and Stuart, 2012). Because AISs express the cytoskeletal scaffolding protein ankyrin-G (Kole and Stuart, 2012), AISs were visualized by labeling ankyrin-G with anti-ankyrin-G antibody ( Fig. 2A,B, arrowheads). Cell bodies of cartwheel cells were labeled with anti-ryanodine receptor antibody. In Figure 2A, the ankyrin-G signal slightly overlapped with Kv2.2 ( Fig. 2Aii-Av, arrow), but not with Kv2.1. Another AIS is shown in Figure 2B, and no overlap was observed between ankyrin-G signal and Kv2.1 or Kv2.2. We quantify the immunolabeling in AISs in Table 2, showing that none out of 12 AISs exhibited the presence of Kv2.1 and/or Kv2.2 signals.
In addition to cell bodies and AISs, the localization of Kv2.1 and Kv2.2 were explored in proximal dendrites of cartwheel cells (Fig. 2C). Cell bodies and dendrites were labeled using intracellular infusion of biocytin via patch pipette, and the dendrites were visualized by using streptavidin-conjugated AF 647 (Fig. 2Ci,Cv). Figure 2Cii-Cv show immunoreactivity of Kv2.1 and Kv2.2 in proximal dendrite, indicating the absence of Kv2.1 or Kv2.2 labeling. The immunoreactivities in dendrites were summarized in Table 2, demonstrating that none out of 10 proximal dendrites showed immunopositive to Kv2.1 and/ or Kv2.2. However, because Kv2.1 and Kv2.1 immunoreactivity in distal dendrites could not be evaluated, significant Kv2 expression may exist there (see Discussion). Taken together, these results suggest that AISs and proximal dendrites in cartwheel cells lack Kv2 channels, and cartwheel cell bodies largely express both Kv2.1 and Kv2.2 on the somatic membrane.

Biophysical property of Kv2 current in cartwheel cells
After confirming the presence of Kv2 channels in cartwheel cells by immunofluorescence labeling, we recorded Research Article: New Research  GxTX-sensitive Kv2 current using the whole-cell patchclamp method and GxTX, a specific Kv2 channel blocker (Herrington et al., 2006;Herrington, 2007). The recordings were performed at room temperature (23-24°C) to reduce the current amplitude, and 100 nM GxTX-containing ACSF was perfused for longer than 5 min. The ACSF was supplemented with fast synaptic blockers, 0.5 mM TTX, 100 nM apamin, and 1 mM penitrem A. To remove the current from voltage-gated calcium channels, CaCl 2 in the ACSF was replaced with equimolar MgCl 2 , and 0.25 mM EGTA-Na was added. Figure 3A shows the outward current in the absence or presence of GxTX recorded from the same cell and GxTX-sensitive current obtained by subtraction. Outward current was induced by depolarizing pulses up to 30 mV, followed by repolarization to -50 mV. The detailed voltage pulse protocol used is shown at the bottom of Figure 3A. In the absence of GxTX (Fig. 3A, control), the detectable outward current was first activated at voltage -40 mV, and grew with further depolarization, exhibiting a small amount of inactivation. Bath application of GxTX dramatically reduced the outward current (Fig. 3A, GxTX). The current-voltage relationships of outward current recorded under control conditions (Fig. 3A, control) and of that obtained in the presence of GxTX (Fig. 3A, GxTX) are plotted in Figure 3C. The current amplitude was measured at the end of depolarization pulses and then normalized to the average amplitude of the control at 30 mV. Outward current was significantly reduced by GxTX at 0-30 mV (**p , 0.01, ***p , 0.001 by two-way repeated measure ANOVA and Bonferroni post hoc tests, n = 13). To define the voltage dependence of steady-state activation, an activation curve was constructed from averaged data in 13 cartwheel cells using tail current after 50-ms depolarization pulses (Fig. 3C,D). The experimental data were fit well by a Boltzmann curve (Fig. 3D, black curve), with a V 1/2 of -11.84 6 0.86 mV (n = 13) and a k of 7.90 6 0.76 mV (n = 13). This V 1/2 value is very close to V 1/2 that was recorded in superior cervical ganglion neurons (V 1/2 = -12.9) and in entorhinal cortex layer II stellate cells (V 1/2 = -13.4), suggesting that the voltage clamp experiment was conducted accurately (Liu and Bean, 2014;Hönigsperger et al., 2017). Figure 3E illustrates the activation kinetics of the Kv2 current obtained by GxTX subtraction. The rising phase (20 ms starting from the depolarization onset) of Kv2 current was fitted by a single exponential function (Fig. 3Ei, red curves). Figure 3Eii summarizes the voltage dependence of the activation time constant. The value became smaller as the membrane potential became more depolarized. Figure 3F shows the deactivation kinetics of the Kv2 current obtained by GxTX subtraction. The deactivation phase (5-35 ms from the repolarization onset) was also fitted by a single exponential function (Fig. 3Fi, red curves). The time constant tended to increase as the repolarization became positive (Fig. 3Fii). Thus, Kv2 current in cartwheel cells is high-voltage activated and activates and deactivates slowly (Johnston et al., 2010).

The blockade of Kv2 channels depolarizes the potential of afterdepolarization (ADP)
In cartwheel cells, V 1/2 of Kv2 current was approximately -11.8 mV; therefore, Kv2 channels should be activated during action potentials. To confirm this hypothesis, we performed whole-cell current clamp recording from cartwheel cells at near-physiological temperature (33-34°C ). As cartwheel cells exhibit spontaneous firing in vitro (Manis et al., 1994;Kim and Trussell, 2007;Bender et al., 2012), resting membrane potentials were adjusted to suppress spontaneous firing by injecting negative bias currents (-50 to -150 pA). Cartwheel cells show two types of action potentials: simple spikes and complex spikes (Manis et al., 1994;Kim and Trussell, 2007). When a brief (1-ms duration) depolarizing current was injected to cartwheel cells, we observed either simple spikes (Fig. 4Ai,Bi, control) or complex ones (Fig. 4Ci, control). Here, we termed cells exhibiting simple spikes as simple-spiking cells, and complex ones as complex-spiking cells. Out of 39 cells, 24 corresponded to simple-spiking cells, and the rest were complex-spiking (15 cells). In simple-spiking cells, we observed three effects of GxTX on action potential behavior: depolarization of the ADP (Fig. 4A), broadening of the action potential (Fig. 4A), and conversion of simple-spiking to complex-spiking (Fig. 4B). The magnitude of the ADP was measured 10 ms after the beginning of depolarizing current injection (Fig. 4Ai, overlay, dashed line), as summarized in Figure 4Aii, and revealed that it was depolarized significantly by 4.25 mV in the presence of GxTX (-69.23 6 0.88 mV, GxTX, -65.01 6 0.90 mV, n = 17, ***p , 0.001; Fig. 4Aii, control). The half-width of the action potential was also significantly broadened by GxTX (0.452 6 0.013 ms, GxTX, 0.492 6 0.015 mV, n = 17, ***p , 0.001; Fig. 4Aiii, control). Out of 24 simple-spiking cells, 7 cells started showing complex spikes when Kv2 channels were blocked by GxTX (Fig. 4Bi,Bii). Presumably For the examination of immunoreactivity to cell bodies and AISs, ryanodine-positive, medium-sized cells were considered as cartwheel cells. When cartwheel cells were labeled intracellularly, the cell type was confirmed by recording the characteristic complex spiking behavior.  Ci, Complex spikes evoked by current injection. In GxTX, fAHP became less negative (inset) and ADP became more prominent. Cii-Civ, Summary of fAHP, ADP, and half-width of first action potential in complex spikes. ADP was measured at 10 ms after the beginning of depolarizing current injection (dashed line in Ci, overlay). Statistical significance was tested using paired t tests; *p , 0.05, **p , 0.01, ***p , 0.001. this change was because of the more depolarized ADP in GxTX which reached the threshold of the action potential, thus driving extra spikes. When complex-spiking cells were exposed to GxTX-containing ACSF, the fAHP between the first and second spikes became less negative (Fig. 4Ci, overlay, inset; -47.51 6 1.52 mV, GxTX, -44.85 6 1.67 mV, n = 15, ***p , 0.001; Fig. 4Cii, control). Moreover, the potential of ADP recorded after complex spikes became positive (Fig. 4Ci, overlay). The potential was measured at 10 ms after the beginning of depolarizing current injection (Fig. 3Ci, overlay, dashed line), and the mean of the shift was 3.13 mV (-62.51 6 2.70 mV, GxTX, -59.38 6 2.62 mV, n = 15, **p , 0.01; Fig. 4Ciii, control). In addition, the effects of GxTX on intrinsic membrane properties were also examined ( Table 3). The maximum rate of spike decay was obviously slowed, and threshold current was significantly increased by GxTX. The other intrinsic membrane properties, including resting membrane potential, input resistance, threshold potential, action potential amplitude, maximum rate of spike rise, were not affected by GxTX (Table 3). These findings demonstrate that Kv2 channels lower the membrane potential of ADP in both simple-spiking and complex-spiking cells and accelerate the repolarization of an action potential.
Sustained firing induced by depolarizing current injection are interrupted by the blockade of Kv2 channels Figure 4 and Table 3 illustrate that Kv2 channels contribute to lower the potential of ADP and accelerate the falling phase of an action potential. However, because of their slow activation kinetics, Kv2 channels may not fully open during the time course of a single action potential.
To determine what will happen if Kv2 channels are blocked during sustained depolarization in cartwheel cells, sustained firing was evoked by injecting a long (300ms duration) square-wave depolarizing current (Fig. 5). Simple-spiking cells exhibited sustained firing characterized by simple spikes when weak current was injected (Fig. 5A, 150 pA). With stronger current injection, these cells could fire both simple and complex spikes (Fig. 5A, 400 and 700 pA), consistent with previous reports (Kim and Trussell, 2007). In the presence of GxTX, cells often stopped firing altogether in the middle of a sustained current step (Fig. 5A, GxTX, 400, and 700 pA). Figure 5C summarizes the input-output relationship of simple-spiking cells. The firing frequency was reduced by GxTX between 250-and 700-pA current injection. A similar tendency was observed in complex-spiking cells (Fig. 5B,  D). In the control, depolarizing pulses produced either isolated complex spikes (Fig. 5B, control, 100 pA) or trains of complex spikes (Fig. 5B, control, 350, and 700 pA; Manis et al., 1994). In GxTX, weak current injection tended to induce more action potentials (Fig. 5B, GxTX, 100 pA), but the increase was not significant (Fig. 5D, GxTX, 50-150 pA, n.s.: not significant by two-way repeated measure ANOVA and Bonferroni post hoc tests). During stronger current injection, the cell ceased showing trains of complex spikes (Fig. 5B, GxTX, 350, and 700 pA). Figure 5D,E summarize the input-output relationship of complex-spiking cells and the combined data from both simple and complex-spiking cells. In both panels, the firing frequencies of the control data are seen to increase monotonically before reaching a plateau. On the other hand, the firing frequencies in GxTX increased in the range of 50-200 pA and then decreased as the injected current became stronger (*p , 0.05 and ***p , 0.001 by two-way repeated measure ANOVA and Bonferroni post hoc tests). These results clearly demonstrate that Kv2 channels are necessary not only for sustained tonic firing induced by current injection but also for preventing depolarization block in cartwheel cells.
The blockade of Kv2 channels induces failure of sustained firing evoked by parallel fiber stimulation in a frequency-dependent manner The previous experiment explored sustained firing in response to a constant current injection. However, the responsiveness of the neurons might be quite different in response to trains of discrete EPSPs. To study the role of  4.09*** As cartwheel cells exhibit spontaneous firing in vitro, at the beginning of the experiment, the resting membrane potential was adjusted to around À80 mV to suppress spontaneous firing by injecting a negative bias current (À50 to À150 pA). The amplitude of the bias current was kept constant throughout the experiment. Input resistance was measured by applying a small hyperpolarizing current pulse (À50 pA, 300-ms duration). For the measurement of action potential properties, simple or complex spikes were induced by brief, strong current injection (1-ms duration, 100-pA increment, up to 2500 pA). Threshold current was defined as the current amplitude that evoked action potential for the first time. The half-width of action potential was measured at the potential between the threshold and action potential peak. In complex-spiking neurons, the threshold potential, action potential amplitude, and the maximum rate of spike rise and decay of action potentials were calculated from the waveform of the first action potential. Statistical significance was tested using paired t tests; n.s.: not significant, *p , 0.05, **p , 0.01, ***p , 0.001.
Kv2 channels under more physiological stimuli, we activated excitatory inputs to cartwheel cells. Cartwheel cell dendrites receive glutamatergic excitatory synaptic inputs from parallel fibers (Wouterlood and Mugnaini, 1984). Electrical stimulation of parallel fibers in vitro evokes EPSCs in cartwheel cells (Roberts and Trussell, 2010;Vogler et al., 2020), and activation of parallel fibers in vivo can drive action potentials in cartwheel cells (Davis and Young, 1997). Thus, parallel fiber input is considered to be the principal excitatory input in cartwheel cells (Wouterlood and Mugnaini, 1984). To examine how parallel fiber-induced action potentials are affected by the blockade of Kv2 channels, we first evaluated the effect of GxTX on parallel fiber-EPSC (Fig. 6), and then blocked Kv2 channels of cartwheel cells during repetitive action potentials (Fig. 7). Cartwheel cells were held at -80 mV under voltage clamp conditions, and a train of stimuli (20 times) was applied on parallel fibers in the ACSF supplemented with 1 mM strychnine and 100 mM picrotoxin. The stimulus frequency was varied from 10 to 100 Hz. At 10 and 30.3 Hz, a train of EPSCs showed slight facilitation at the beginning of stimulation, and then sustained responses (Fig.  6A, 10 and 30.3 Hz). At 71.4 Hz, a train of EPSCs showed facilitation in the beginning, followed by a slight depression. Figure 6B-D summarize a series of experiments, revealing that parallel fiber-cartwheel cell synapses do not show clear synaptic depression, and that the amplitude of the 20th EPSC is larger than that of the first (10-71.4 Hz) or similar to that of the first (100 Hz). The effect of GxTX on a train of EPSCs (100 Hz) was also tested (Fig. 6E). GxTX had little effect on the EPSC amplitude, and the differences were not significant (two-way repeated measure ANOVA and Bonferroni post hoc tests; Fig. 6Eii), indicating that GxTX does not affect the ability of parallel fiber axons to conduct high-frequency spikes or the transmitter release from parallel fiber.
To explore how a train of action potentials evoked by parallel fiber stimulation is affected by the blockade of Kv2 channels, recordings were made under currentclamp conditions (Fig. 7). Because the input-output relationship of simple-spiking cells was quite similar to that of complex-spiking cells (Fig. 5C-E), the data from both cell types were combined and pooled for the following quantitative analysis (Fig. 7E). Spike probability was calculated from five trials, and we defined that successful action Figure 6. A train of parallel fiber-EPSCs evoked by various stimulus frequencies. A, A train of EPSCs induced by repetitive stimulation of parallel fibers. The membrane potential was held at -80 mV in the presence of strychnine and picrotoxin. Averages of five traces were used. Note that the amplitude of the last EPSCs is still large compared with the first EPSC in every stimulus frequency. B-D, Summary of relative amplitude of EPSCs plotted against stimulus numbers. Ei, A train of EPSCs in the absence (control) or presence (GxTX) of GxTX. Eii, Summary of the effects of GxTX on a train of EPSCs evoked at 100 Hz. Statistical significance was tested using two-way repeated measure ANOVA and Bonferroni post hoc tests. n.s.: not significant. Figure 7. The effects of GxTX on sustained firing evoked by repetitive parallel fiber stimulation. A-C, Tonic firing evoked by repetitive parallel fiber stimulation. Traces in A, B were recorded from the same cells. Traces in C were from another cell. Top traces in each panel were used for the overlays. A part of overlayed traces is expanded as insets. Note that in the presence of GxTX (red traces), some of action potentials were not evoked by the stimulation. Failures of action potentials are indicated by hash marks (#). Stimulus artifacts were truncated for clarity. D, The relationship between maximum rate of rise of spikes and peak amplitude of potential has maximum rate of rise .30 V/s and peak amplitude higher than À15 mV (Fig. 7D). Without GxTX, cartwheel cells showed trains of action potentials with high fidelity at 20-71.4 Hz (Fig. 7A,B, control). At 100 Hz, a few failures were observed in each trial (Fig. 7C, control), but most of the stimuli succeeded in inducing action potentials. At 20 Hz, GxTX had little or no effect on the spike probability when parallel fibers were stimulated (Fig. 7E, 20 Hz). At higher stimulus rates, however, the blockade of Kv2 channels induced the failure of action potentials in the middle and late of the train stimulation ( Fig. 7A-C,E, 30.3-100 Hz). Taken together, these data demonstrate that blocking Kv2 channels induces failure of sustained firing evoked by a parallel fiber stimulation frequency dependent manner.

Discussion
To the best of our knowledge, the present study is the first to demonstrate that Kv2 channels play a crucial role in regulating synaptically induced, repetitive firing in a frequency-dependent manner in neurons. Largely, Kv2.1 and Kv2.2 channels were present on the cell body of cartwheel cells, not on AISs nor dendrites (Figs. 1, 2), and Kv2 current accounted for about one-third of the delayed rectifier K current (Fig. 3). Kv2 contributed to accelerating the repolarizing phase of action potential and hyperpolarizing the potential of ADP (Fig. 4). Furthermore, blocking Kv2 channels induced the failure of action potentials in the middle and late sustained firing evoked by current injection (Fig. 5) or parallel fiber stimulation (Fig. 7).
Although there are some reports demonstrating that Kv2 channels are necessary for sustained tonic firing induced by current injection and that Kv2 plays a role in the prevention of depolarization block (Tong et al., 2013;Liu and Bean, 2014;Hönigsperger et al., 2017), this study is the first one that shows the role of Kv2 channels in action potential generation in response to excitatory synaptic inputs, which are more physiological stimuli compared with current injection. In medial nucleus of the trapezoid body auditory neurons, Kv2.2 channels are highly expressed in AIS, and Kv2.2 is necessary for high-frequency firing (Johnston et al., 2008;Tong et al., 2013). Indeed, cartwheel cells belong to auditory neurons, but Kv2.2 distribution in cartwheel cells we observed was strikingly different from that in medial nucleus of the trapezoid body neurons.

Limitation of immunofluorescent study in dendrites
The cartwheel cells whose cell bodies exist in the surface of the acute slices (200-mm thickness) were wholecell patched, and the dendrites were labeled by intracellular infusion of biocytin via patch pipette. After that, the dendrites were visualized by streptavidin-conjugated AF 647. When the dendrites were examined under a confocal microscope, we often observed the dendrites elongate to deep direction, where immunoreactivity of Kv2 were not detected. Therefore, considering that the thickness of the slices and the penetration of antibodies, we had to acquire images of proximal dendrites, which exist close to the surface of the slices. Clear immunoreactivities of Kv2 channels were not observed in the proximal dendrites, but we did not slice again the thick slices for patch-clamp recordings to label distal dendrites with antibodies. Therefore, we cannot rule out the possibility that distal dendrites do express Kv2 channels.

Localization of Kv2.1, Kv2.2, and ryanodine receptors in the cell bodies
Immunofluorescence studies revealed that in most case both Kv2.1 and Kv2.2 existed on the cell body of cartwheel cells ( Fig. 1B; Table 2), but the overlap between Kv2.1 and Kv2.2 was not observed frequently (Fig. 1Cv), suggesting that not all Kv2.1 channels form heterotetramers with Kv2.2 channels (Kihira et al., 2010), most Kv2.1 or Kv2.2 channels might exist as homotetramers in cartwheel cells. In specific neurons, including CA1 pyramidal neurons and striatal medium spiny neurons, Kv2.1 clusters are juxtaposed to clustered ryanodine receptors, a marker of endoplasmic reticulum (ER; Mandikian et al., 2014). Kv2.1 channels organize ER and plasma membrane (PM) junctions (ER-PM junctions) through an interaction with vesicle-associated membrane proteinassociated protein isoform A (VAPA) and VAPB (Johnson et al., 2018;Kirmiz et al., 2018a). We expected that Kv2.1 clusters overlapped with the ryanodine receptor; however, such overlap was not found frequently in cartwheel cells (Fig. 1Cvi), suggesting that Kv2.1 channels are not involved in the formation of ER-PM junctions in the cells. Cartwheel cells might lack VAPA and/or VAPB.
Putative mechanism underlying interrupting of sustained firing by GxTX The activation kinetics of Kv2 channels reported in superior cervical ganglion neurons and in entorhinal cortex layer II stellate cells are relatively slow, and those observed in this study were similar to those in previous studies ( Fig. 3E; Liu and Bean, 2014;Hönigsperger et al., 2017). During the sustained depolarization induced by current injection or a train of excitatory synaptic input, Kv2 channels would gradually open and contribute to the repolarization of action potentials. Once Kv2 channels continued action potentials plotted on a semi-logarithmic scale. The plot was made by differentiating six traces showing 20 synaptic responses at 100 Hz in the presence or absence of GxTX. Action potentials which have the rate of rises lower than 30 V/s or have peak amplitude lower than -15 mV are colored in blue (failures). E, Summary of spike probabilities plotted against stimulus numbers. For the analysis, five traces in control and five in GxTX were used from each recording for the calculation of the probabilities. Statistical significance was tested using two-way repeated measure ANOVA and Bonferroni post hoc tests. n.s.: not significant; *p , 0.05, **p , 0.01, ***p , 0.001. were blocked by GxTX, the membrane potential between action potentials becomes more positive, leading to the accumulation of Na channel inactivation, to the interruption of sustained firing, and to produce depolarization block. However, to confirm this hypothesis, further studies using a model simulation are necessary.

Implications for auditory function
Currently, there have been few studies reporting the excitability of cartwheel cells in vivo (Davis and Young, 1997;Ma and Brenowitz, 2012). In anesthetized cats, the maximum rate of firing in response to the best frequency tone is 74 Hz, and the maximum rate reaches 50-130 Hz in response to somatosensory stimulation conveyed by parallel fibers. This firing frequency is within the range of firing frequency examined in this study. The maximum firing frequency induced by the current injection was ;100 Hz, and the blockade of Kv2 channels by GxTX suppressed the firing frequency to ;70 Hz ( Fig. 5E). Cartwheel cells fired reliably up to 100 Hz in response to parallel fiber stimulation, but reliable firing was not observed with GxTX even at 30-Hz stimulation (Fig. 7E). Thus, Kv2 channels in cartwheel cells must be activated in vivo, and the activity of Kv2 channels could play an important role in the high-frequency firing of cartwheel cells in vivo. As cartwheel cells strongly inhibit principal cells of the DCN (Roberts and Trussell, 2010), the regulation of excitability of cartwheel cells by Kv2 channels might be important for DCN functions such as monaural sound localization and cancelling detection of self-generated sounds (May, 2000;Singla et al., 2017).