HCN channels in the mammalian cochlea: Expression pattern, subcellular location, and age‐dependent changes

Abstract Neuronal diversity in the cochlea is largely determined by ion channels. Among voltage‐gated channels, hyperpolarization‐activated cyclic nucleotide‐gated (HCN) channels open with hyperpolarization and depolarize the cell until the resting membrane potential. The functions for hearing are not well elucidated and knowledge about localization is controversial. We created a detailed map of subcellular location and co‐expression of all four HCN subunits across different mammalian species including CBA/J, C57Bl/6N, Ly5.1 mice, guinea pigs, cats, and human subjects. We correlated age‐related hearing deterioration in CBA/J and C57Bl/6N with expression levels of HCN1, −2, and −4 in individual auditory neurons from the same cohort. Spatiotemporal expression during murine postnatal development exposed HCN2 and HCN4 involvement in a critical phase of hair cell innervation. The huge diversity of subunit composition, but lack of relevant heteromeric pairing along the perisomatic membrane and axon initial segments, highlighted an active role for auditory neurons. Neuron clusters were found to be the hot spots of HCN1, −2, and −4 immunostaining. HCN channels were also located in afferent and efferent fibers of the sensory epithelium. Age‐related changes on HCN subtype expression were not uniform among mice and could not be directly correlated with audiometric data. The oldest mice groups revealed HCN channel up‐ or downregulation, depending on the mouse strain. The unexpected involvement of HCN channels in outer hair cell function where HCN3 overlaps prestin location emphasized the importance for auditory function. A better understanding may open up new possibilities to tune neuronal responses evoked through electrical stimulation by cochlear implants.


| INTRODUC TI ON
Auditory function essentially depends on the concerted interplay of the sensory cells and bipolar spiral ganglion neurons (SGNs) that convey receptor potential changes to ascending auditory pathways.
Although most computational processing of auditory information occurs in the central nervous system (CNS), SGNs are more than point-to-point wires that route sound information to the brainstem.
From eight to 20 of these bipolar neurons innervate a single inner hair cell (IHC) (Nayagam et al., 2011;Spoendlin, 1972) that are the perceptual cells of hearing. Multiple innervations of IHCs help to extract different qualities of sound through individual SGN subtypes.
Type I neurons have long been known to differ in sensitivity to sound and spontaneous neural activity (Kiang et al., 1965;Merchan-Perez & Liberman, 1996) that contributes considerably to the prodigious dynamic range of sound perception and temporal resolution. Many factors contribute to neural diversity, such as receptors that influence activity, ion channels, ribbon synapse variability, or efferent tuning of SGN activity (Davis & Crozier, 2015;Petitpre et al., 2018;Reijntjes & Pyott, 2016;. Expression patterns and localization of molecular components that shape these physiological characteristics remain elusive, and a meticulous analysis of all candidates is necessary to decipher the coding of sound. Intrinsic firing properties are mainly generated and formed by ion channels. Among voltage-gated channels, hyperpolarizationactivated cyclic nucleotide-gated (HCN) channels generate an inward current (I h ) through the entrance of sodium and potassium ions. Four different subunits (HCN1-4) assemble to form homo-or heterotetramers that broaden biophysical characteristics. HCN1 is the subunit with the fastest activation kinetics (τ = 25-300 ms, depending on the voltage) and lowest half activation voltage near to the resting membrane potential (RMP) (V 1/2 = −70 mV), while HCN4 is the slowest (τ presents a few hundred of ms, V 1/2 = −100 mV). HCN2 and HCN3 demonstrate gating characteristics in between (V 1/2 = −99 mV to V 1/2 = −77 mV, respectively; Wahl-Schott & Biel, 2009).
HCN channels are short-term regulated by cyclic nucleotides such as 3′,5′-cyclic (cAMP). When cAMP joins to HCN2 and HCN4, the opening of the channel is accelerated; therefore, the activation time is reduced. However, HCN1 and HCN3 are not modulated by cAMP binding, as noted in the comprehensive review by Sartiani et al. (2017). HCN channels are key regulators of neuronal excitability (Wahl-Schott & Biel, 2009) controlling the RMP, lowering activation thresholds, and shortening excitatory postsynaptic potentials (EPSPs), maintaining those signals briefly by avoiding summation. These channels act as pacemakers in heart muscle cells and have been hypothesized to generate spontaneous neural firing (He et al., 2014;Sartiani et al., 2017;Shah, 2018). This qualifies them as a potential component to determine neuronal subtypes in the cochlea.
Although several studies have examined physiological functions and the presence of HCN channels in the auditory (Horwitz et al., 2011;Kim & Holt, 2013;Liu, Lee, et al., 2014;Liu, Manis, et al., 2014;Ramakrishnan et al., 2012;Shen et al., 2018;Yi et al., 2010) and vestibular systems (Almanza et al., 2012;Horwitz et al., 2011Horwitz et al., , 2014, there are some controversies and a lack of information about their subcellular location (Horwitz et al., 2010;Ramakrishnan et al., 2009Ramakrishnan et al., , 2012. We performed an immunohistochemical analysis across different mammalian species including CBA/J, C57Bl/6N, Ly5.1 mice, guinea pigs, cats, and human subjects to determine subcellular location and co-expression of all four HCN subunits. Quantitative changes in key molecules in age-related hearing loss (ARHL) of C57Bl/6N and CBA/J mice potentially provide a tool to identify functional implications. HCN expression levels and hearing performance were therefore tested to identify any correlation and data were further used to identify common trends that these inbred strains share.
With our results, we aimed to address the following questions:

| Animals and human temporal bones
C57Bl/6NCrl (strain 027; Charles River, Germany), JAX™ CBA/J (JAX Nr. 000656; Charles River, France), and Ly5.1, a CD45.1 expressing B6.SJL-Ptprca Pepco/BoyCrl strain (strain 494; Charles River, Italy), were used to study murine species. The C57Bl/6N strain present an early onset and progression of ARHL and was compared with the CBA/J strain that shows a slower rate of ARHL (Kane et al., 2012). Ly5.1 mice share SGN characteristics of an unmyelinated neuron soma with humans (Jyothi et al., 2010) and served as a "human-like" animal model of SGNs. The animal studies conformed to the Austrian guidelines for the care and use of laboratory animals and were approved by the Austrian Federal Ministry of Education, Science and Research (reference number BMWFW-66.011/0120-WF/V/3b/2016). Animals were bred until different time points from 1 day to 19 months of age. Ears were fixed with 4% formaldehyde (PFA) overnight followed by decalcification in 20% ethylenediaminetetraacetic acid (EDTA). There were no gender differences in age-related hearing thresholds found in C57Bl/6N strains until 32 kHz (Hunter & Willott, 1987). In CBA/J, no difference in hearing deficits were previously reported until 18 months of age, when females showed faster deterioration of hearing than males (Ohlemiller et al., 2010). Because of these results, we did not account for gender effects. All 65 CBA/J (37 males and 28 females) and 84 C57Bl/6N (30 males and 54 females) were bred in the same room at the Innsbruck animal facility with a 12-hr dark-light cycle and unlimited access to food and water. Sections from cat and guinea pigs served only for comparative studies of subcellular HCN location. This archival material lacked data about the sex and was used in previous studies (Glueckert et al., 2008;Rattay et al., 2013).
Since we did not find any differences in immunolocalization of HCN channels between female and male murine species, we regarded sex information as not relevant. For our experiments, we tested puretone hearing thresholds in different age periods including young animals (1-4 months); 5-8 months when C57Bl/6N started losing hearing (Hunter & Willott, 1987); 9-12 months; 13-15 months; and 15-19 months which covers ARHL in CBA/J (Ohlemiller et al., 2010).
We added groups for immunolabeling analysis without functional hearing (P0-P14) and a young hearing animal group with an age that matched the final maturation period (P15-P31).
Human tissue from 10 different temporal bones was collected by the Division of Clinical and Functional Anatomy of the Medical University of Innsbruck. All the individuals were anonymized and had donated their bodies voluntarily to scientific research via their informed consent prior to death (McHanwell et al., 2008;Riederer et al., 2012). There was no evidence of any ear malformation. The time postmortem varied up to a maximum of 13.5 hr before they were fixed with 4% PFA for 48 hr. Sex and ages of individuals were not reported due to anonymization rules. Tissue was decalcified in 20% EDTA at 37°C. Human and animal specimens were cryoembedded as described elsewhere (Coleman et al., 2009) and cut at 10 µm.
Tone pips of 3-ms duration (1-ms rise and fall time) were presented at a rate of 60/s with alternating phases. Frequency-specific sound pressure levels were adjusted with a custom-made attenuator, and tone pips were amplified with a custom-made amplifier. Stimuli were delivered in free field to a Beyer DT911 loudspeaker calibrated in situ prior to each measurement with a KE4 microphone (Sennheiser, Wedemark, Germany). A software averager included an artifact rejection algorithm to facilitate the extraction of the evoked potentials from unwanted noise such as heart beating and muscle activity.
After amplification (100 dB) and band-pass filtering (high pass: cut off 220 Hz, 0 dB gain; low pass: cut off 5,000 Hz, 80 dB gain), electrical signals were averaged (64 repetitions each phase). Starting with 0 dB, the stimuli were increased in 5 dB steps up to 90 dB. Hearing thresholds were determined as the minimum stimulation level that produced a clearly recognizable potential. Recordings were evaluated by two independent researchers in a blinded manner. According to previous publications, we assigned 4 kHz to an apical-, 8 kHz to an apex-middle-, 16 kHz to a middle-, and 32 kHz to an upper basal turn (Liu et al., 2019) location.

| Manual immunohistochemistry
After blocking (1 hr, 30% normal donkey serum and 0.3% Triton X-100), primary antibodies were incubated overnight at 4°C, and then for 1 hr at RT, followed by secondary antibody exposure for 2 hr at RT. Mounting in Vectashield-DAPI (Cat# H-1200, RRID:AB_2336790; Vector Laboratories Inc, Burlingame, CA, USA) provided nuclear staining. In experiments with primary antibodies raised in the same species (rabbit anti-HCN2 and rabbit anti-HCN4), a heat denaturation step (8 min at 81°C) between immunostainings was applied.

| Quantification of neuronal staining and imaging
The membrane staining around the SGN somata was acquired with a PixeLink PL-B623 camera (Pixelink®, Canada) at 2,048 x 1,536 pixel resolution on a TisseFaxPlus® microscope system equipped with 40x/1,3 lens, and quantification was performed with the dedicated image analysis software Histoquest® 6, (RRID:SCR_014823; both from Tissuegnostics®, Vienna, Austria). A software internal color deconvolution strategy separated hematoxylin from DAB staining.
An algorithm was trained to segment specifically SGNs by their hematoxylin nuclei staining considering size, shape, mean staining intensity, and color shade until SGN recognition was appropriate. A growing mask around these nuclei served to label a ring-shaped area around the cell membrane in a reproducible way. Mean intensity of gray values for the DAB immunoreactivity was calculated for each

| Experimental design and statistical analysis
The number of animals based on sample size calculation studies is extensively described elsewhere (Arifin & Zahiruddin, 2017;Charan & Kantharia, 2013

| HCN localization
3.1.1 | HCN1 localization HCN1 was located at the plasma membrane of type I SGNs in mice  segment (pAIS) and a central axon initial segment (cAIS) located at the pre-and postsomatic segments, respectively (Hossain et al., 2005).
Pre-and postsomatic segments can be identified by anatomical position and the fact that the central axon is double the diameter of the peripheral axon (Potrusil et al., 2012). We could frequently observe HCN1-LI in the postsomatic region, but never observed staining at presomatic sites ( Figure 1e). Double stainings with contactin-associated protein 1 (CASPR 1) showed HCN1 staining along the entire cAIS ( Figure 1f). Previous studies in motor neurons localized CASPR 1 at the para-AIS compartment. This site marks the distal end of AIS Nav channel distribution and start of the axonal myelination in CNS neurons (Duflocq et al., 2011). Both of these segments and the soma are fully myelinated in rodents and cats, while most type I SGNs in humans completely lack myelination at these neuronal segments.
The length of postsomatic segments varies considerably in humans from more than 30 µm to not identifiable (Liu et al., 2015).
Tonotopical gradients of HCN1 expression were not static but F I G U R E 1 HCN1 expression in the adult mammalian cochlea. HCN1 channel subunits were highly expressed in neurons with intense cytoplasmatic and perisomatic membrane staining. a, CBA/J spiral ganglion neurons (SGNs) somatic staining with DAB. There was no visible gradient across SGNs of the middle (SGm) and basal (SGb) turns. The central cochlear nerve (CN) and axons in the osseous spiral lamina (OSL) were weakly stained. Confocal imaging in C57Bl/6N showed highest intensity of HCN1 at neuronal clusters (arrows) in the apical turn (b); neuron soma membrane staining in the basal turn (c). d, No specific staining was found in the sensory epithelia; inner hair cell (IHC). e, Intense staining at the axon hillock and central axon initial segment (cAIS, arrows). No staining was detected at the peripheral AIS. f, Double staining of HCN1 (f1) with contactin-associated protein 1 (CASPR, f2, arrow) identified HCN1 staining to overlap with the cAIS. g, Neurons in Ly5.1 presented staining with higher intensity. Cat (h) and human (i) shared equal HCN1 distribution at the neuron soma and cAIS (arrows). j, Double staining with peripherin (PRPH), a specific type II SGN marker does not co-express with HCN1 in C57Bl/6N (asterisk), but in cat (k) and human (l). A type II SGNs cluster in human presented different intensities of HCN1 (l1). f1, j1, k1, l1, DAB channel (HCN1), f2, j2, k2, l2 AEC channel (CASPR1 and PRPH) after color deconvolution. B6, C57Bl/6N; CBA, CBA/J; P, postnatal day [Color figure can be viewed at wileyonlinelibrary.com] Interestingly, guinea pigs and cats produced intense immunoreactivity at the lateral membrane of OHCs ( In summary, HCN4 located quite evenly across the perisomatic SGN membrane and was observed at both the pre-and postsomatic segments in all species. Membrane staining did not exceed most distal ends of the cAIS and pAIS. Neural clusters expressed considerably more HCN4. Type II neurons appeared negative in C57Bl/6N and CBA/J but showed reactivity in other mammals at varying intensities. The expression of HCN4 was highly variable at the peripheral nerve endings of type I and type II neurons among investigated mammalian species. Expression in the sensory epithelium was evident in Ly5.1 mice, guinea pig and cat but fell below detection levels with standard immunolabeling techniques in human.

| HCN subunit co-expression and correlative distribution in SGNs of mice and man
Since few neurons were completely void of any immunoreactivity against one of the HCN channels, a co-expression of several subunits within an individual neuron was obvious ( Figure 6). SGN clusters that contained the highest staining intensities for HCN1, −2, and −4 were excluded from evaluation and only the staining of the "normal" separated neuron somata was quantified. Resolution was not sufficient to detect a co-localization of the different HCN subunits in a heteromeric channel configuration. However, a lack of fluorescent overlap with conventional confocal imaging techniques provided information about the presence of non-paired HCN combinations that appeared as single-stained sites in double F I G U R E 2 HCN2 expression in the adult mammalian SGNs. HCN2 presented staining at the neuronal soma membrane and axons. a, Midmodiolar view exposed immunostaining in the organ of Corti (OC) and a rather homogeneous staining along the tonotopical axis in CBA/J across the spiral ganglion in the apical (SGa), middle (SGm), and basal (SGb) turn. b, Super-resolution confocal imaging with calretinin as neuronal counterstaining and DAPI nuclear stain. b1, direct-conjugated HCN2 antibody signal, b2, calretinin immunoreactivity. HCN2 was present only in perisomatic SGN membranes, axons (arrows) are void of any HCN2-LI. c, High intensity of the staining was visible in a Ly5.1 mouse in the apical turn within a neuron cluster (arrow), along the osseous spiral lamina (OSL) and in the organ of Corti (OC). d-i, Comparison of neuronal staining in three inbreed mouse strains identified similar expression levels of HCN2 in CBA/J (d, g) and C57Bl/6N (e, h) but higher intensity of staining in Ly5.1 mice (f, i). HCN2 in type I neurons of a guinea pig (j), type I and peripherin (PRPH)-positive type II neurons in cat (k) and human (l-p); k1 DAB channel (HCN2), k2, AEC channel (PRPH), after color deconvolution. l, Peripheral nerve fibers (PNF) appear weakly positive in human. m, Small human type II neurons were heavily stained for HCN2 (arrows). n, High magnification of human type I SGNs showed intense staining at the soma membrane, and weaker in the cytoplasm; postsomatic segment (asterisks) and the satellite glia cell (SGC) were void of reactivity. Double staining with PRPH was positive for HCN2 in human type II neurons (arrows) (o, p), but negative in CBA/J mice (q) and C57Bl/6N ( HCN1 intensity had limited dependence on HCN2 intensity within the same basal neuron. The majority of HCN1-rich cells (>60% relative HCN1 intensity) revealed the lowest HCN2 content (<60% relative HCN2 intensity) within a single neuron ( Figure 6v). This is consistent with previous reports that HCN1 is emphasized toward higher frequencies (Liu, Manis, et al., 2014). Colorimetric staining in human apical to middle SGNs showed some membrane spots were more positive for HCN1 and some more for HCN2 within the same type I neurons (Figure 6e Interestingly, we found type I neurons in the human middle turn that solely expressed HCN4. These neurons with high expression of HCN4 and low or negative staining for HCN1 were found in murine sections as well (data not shown), suggesting additional "specialized" subtypes.

| HCN2-HCN4 co-expression
The co-expression of these channels presented the strongest correlation coefficients for all turns, especially in the apex (Figure 6x: τb apical turn = 0.729, p < 0.001; τb middle.turn = 0.589, p < 0.001; τb basal turn = 0.608, p < 0.001). This distribution indicated a mutual relationship of HCN2 and HCN4 and could have accounted for concerted functions, such as both subunits being modulated by cAMP.
Fluorescent signals were less clearly separated in co-staining experiments (Figure 6o-r). Most immunoreactivity sites revealed clearly separated fluorescent signals, but we also observed co-localized pixels within spatial resolution limits. Hence, we cannot exclude some heteromeric HCN2/HCN4 channel combinations. In human, HCN2/ HCN4 co-staining resulted in most congruent color deconvoluted images in the apical turn (Figure 6s-u).
Overall, HCN1 dominated in high-frequency neurons over HCN2 and HCN4. HCN2 and HCN4 co-expression was mostly proportional. In both combinations of HCN4 with HCN1 and HCN2, we found single-stained HCN4-positive neurons in mouse and human.
In HCN1-HCN2 double stainings, we found single-stained neurons only in human sections indicating an even higher degree of specialized neurons.
F I G U R E 3 HCN2 expression in the adult mammalian sensory epithelia. HCN2 was present at afferent and efferent nerve fibers innervating the organ of Corti. a, Differential interference contrast image (DIC) of DAB visualized immunostaining showed presence in C57Bl/6N neurons, osseous spiral lamina (OSL), and underneath the inner hair cell (IHC). The inset with a magnified view of the organ of Corti exposed staining at the inner spiral plexus (isp) underneath the IHC, boutons and nerve fibers between outer hair cells (OHCs) and supporting cells (arrows) and the efferent fibers of the tunnel spiral bundle (tsb). b, Similar staining in CBA/J in the organ of Corti with expression underneath IHC and OHCs (arrows). Positive bigger caliber tunnel crossing fibers (tcf) account for the medial efferent innervation. c, DIC of a CBA/J horizontal orientation, inset: high magnification clearly showed staining in the large efferent synapses (asterisk), while smaller putative afferent type II terminals (arrow) were void of immunoreactivity. d, Ly5.1 mice HCN2 staining underneath the IHC, fibers traveling at the base of the tunnel (basilar fibers, bf) and thin caliber tcf (arrows) represent type II afferent fibers. Large boutons underneath OHCs correspond to the efferent innervation. Guinea pig (e) and cat (f) organ of Corti showed the same HCN2 staining pattern like Ly5.1 mice with type I and type II afferent and efferent fibers stained. In cat, the inset shows staining in the large synapses underneath OHCs from the medial efferent fibers (asterisk) and smaller putative type II terminals (arrow). Human sensory epithelia presented high levels of HCN2 expression at equal localization as in animals (g-m) emphasized in the massive staining in outer spiral bundles (osb) typical for human. g, Immunopositive isp and bf (type II afferents), as well as medial efferent fibers of the tsb. The prominent osb comprises type II afferent as well as efferent fibers shown in (h) in transmission electron microscopy.

| HCN1 and HCN3
Before the onset of hearing, HCN1 and HCN3 were only detectable in neurons but not in the sensory epithelium (Figure 7). We were not able to find a distinct membrane staining.

| HCN2 and HCN4
We found HCN2 and HCN4 in SGNs and the sensory epithelium underneath the IHCs and OHCs as early as P1. Visually, no tonotopic gradient could be identified.
Newborn mice revealed intense immunostaining for all HCN channel types. HCN3 appeared to act more before birth and fade out after. HCN2 presented underneath the IHC at birth while HCN4 was additionally expressed at OHC afferents. HCN4 lost importance in the sensory epithelium until the onset of hearing function, and the faster kinetic HCN2 remained in IHC innervation and OHC efferents.

| eABR results
We confirmed the typical pattern of ARHL in our animals, with early elevated hearing thresholds of high frequencies in C57Bl/6N and of the low frequencies in CBA/J mice (Henry, 2002(Henry, , 2004Hunter & Willott, 1987;Sha et al., 2008). Significant differences are visualized in the tables below each graph (Figure 9). Hearing thresholds in CBA/J mice remained constant until 12 months before they in-

| HCN quantification
Semi-quantitative evaluation of HCN1, −2, and −4 staining intensities from newborn to 19-month-old C57Bl/6N and CBA/J mice F I G U R E 4 HCN3 expression in the adult mammalian cochlea. HCN3 was weakly expressed in spiral ganglion neuron (SGN) cytoplasm of mice (a-d), guinea pig (i), cat (k), and human (o). In guinea pig and cat HCN3 was detected at the level of outer hair cells (OHCs) (e-f and j). a, In a mid-modiolar view of C57Bl/6N SGNs presented weak staining. A tonotopic gradient was observed from apex to base. b, CBA/J showed weak staining in neuron cytoplasm. c-d, A more sensitive immunostaining protocol with alkaline phosphatase detection as marker enzyme was performed in Ly5.1; reactivity in apical neuron clusters (d, arrow) was more intense than in basal neurons (c). OHCs in the guinea pig showed specific staining at the lateral membrane (e-f). Staining pattern was similar to the motor protein prestin (g-h). i, HCN3 was also present in the guinea pig SGNs, in cat OHCs (j) as well as cat SGNs (k). l-n, Thin unmyelinated fibers at the peripheral (l, n) and central axon (m) were positive in cat. o, human tissue revealed similar staining as in animal tissue. Most, but not all neurons (arrows) appeared stained. Double staining with peripherin (PRPH) confirmed the presence of HCN3 in type II SGNs (arrows and asterisk) in cat (p), human apical turn SGNs (q) and in C57Bl/6N (r) with varying intensities. p1, q1, r1, DAB channel (HCN3), p2, q2, r2, AEC channel (PRPH) after color deconvolution. B6, C57Bl/6N; CBA, CBA/J; P, postnatal day [Color figure can be viewed at wileyonlinelibrary.com] of individual SGNs revealed only a few corresponding trends with eABR recordings (Figure 10). Most noticeably, the intensity of immunoreactivity for HCN1 and HCN2 were higher in the "good hearing" CBA/J (Figure 10a,b) compared with C57Bl/6N ( Figure 10d,e) mice, while HCN4 levels were similar. HCN1 was previously found to be expressed highest in the base and HCN4 in the apex in adult mice (Liu, Lee, et al., 2014). In contrast, our data showed that young adults up to 8 months of age present the highest levels of HCN1, −2, and −4 in low-frequency neurons. SGN clusters were identified as hot spots of HCN expression and were frequent in the apical turn in C57Bl/6N but rarely found in CBA/J (Jyothi et al., 2010). Therefore, excluding clusters from evaluation would probably not have changed this gradient. Statistically significant differences are presented among the turns within one age group ( Figure 11) and among the same turn at different ages ( Figure 12).

HCN1
Across all age groups, HCN1 varied most in both mouse strains  10d and 12b). The tonotopic gradient of HCN1 expression in C57Bl/6N mice was less evident and even reverted in animals older than 8 months (Figure 11b).

HCN2
CBA/J mice reached an early plateau of intensity after 4 months of age ( Figure 10b).
Immunostaining intensities in C57Bl/6N mice changed mainly in the middle and basal turns (Figures 10e and 11e). While there were no significant changes of HCN2 expression until 4 months, staining intensities tended to increase in mice older than 1 year of age. This coincided with an increase in hearing thresholds after 4 months of age. Diversity of individual expression levels of HCN2-LI represented by the spread of data (Figure 10e) was less pronounced.

HCN4
HCN4 is the most stable expressed subtype and little influenced by age or hearing performance. We observed a few marked changes mainly in the apical turn in CBA/J mice (Figure 10c). Immunoreactivity peaked in 5-to 8-and 13-to 15-month-old groups. The drop of HCN4-LI in the basal turn of 9-12-month-old-mice was statistically significant compared with the flanking age groups (Figure 12e). Especially the opposing behavior of up-and downregulation of HCN1 and HCN2 intensities at very old age appeared to be obscure since both mouse strains exhibited marked hearing loss in the 16-to 19-month-old group. The data did not allow us to extract a common trend in HCN expression in aging mice, since the changes were too varied and showed little correlation with the pure-tone threshold audiometry data. All immunostaining results were summarized in Table 3.
F I G U R E 5 HCN4 expression in the adult mammalian cochlea. a, Mid-modiolar view of a section from a C57Bl/6N mouse presented visually a tonotopical apex to base gradient. Neurons of the apical turn, especially in neuron clusters were intensely stained. b, Confocal microscopy confirmed the high expression of HCN4 at the neuronal membrane, especially between neurons of a cluster. c-e, HCN4 presented higher intensity of staining in Ly5.1 (e), compared to CBA/J (c) and C57Bl/6N (d). f-h, Confocal microscopy showed HCN4 to be present at the neuronal membrane of type I neurons as well as the peripheral axon initial segment (pAIS) and central axon initial segment (cAIS) in guinea pig (f). CASPR 1 was used as a marker for the distal end of the AISs. AISs lengths differed considerably (f, h). f1, h1, HCN4 staining, f2, h2, CASPR 1 staining, f3, nuclear staining (DAPI). Cat type II neurons presented immunoreactivity in single (i, arrow) and double-stained sections with additional peripherin (PRPH) staining (j). j1, DAB channel (HCN4), j2, AEC channel (PRPH). Human sections (k-m) presented staining at the perisomatic membrane as well as the cAIS and pAIS. Type II neurons with their delicate unmyelinated axons in humans are positively stained for HCN4 (k). l-m, cAIS length in human varied from very short (l) up to more than 30 µm here (m, arrows). SGN membranes were immunopositive especially in the neuron clusters (arrow, l). n, Double staining with PRPH in a CBA/J mouse showed no co-expression with HCN4 above background level, n1, DAB channel (HCN4), n2, AEC channel (PRPH) after color deconvolution. o-v, HCN4 expression in the sensory epithelia differed among species: in adult mice, no staining is visible in CBA/J and C57Bl/6N while in Ly5.1 showed staining in young (o) as well as old individuals (p, q). Guinea pig sections presented HCN4 mainly at type I afferents (r), while for cat mainly type II fibers (s, t) were positive. Standard immunohistochemistry methods were not sensitive enough to detect HCN4 in the human organ of Corti (u), but highly sensitive visualization methods using alkaline phosphatase enzymes exposed weak staining underneath the IHC (v). B6, C57Bl/6N; CBA, CBA/J; P, postnatal day [Color figure can be viewed at wileyonlinelibrary.com]

| Are HCN channels in all mammalians distributed the same?
We found that HCN channel subunits were distributed unevenly among mammalian auditory neurons and showed characteristic changes of expression levels during postnatal development and aging. Expression patterns were not uniform among species or even within mouse inbred strains. This variational expression may represent evolutionary adaptions or loss of functions. Despite largely overlapping expression patterns, individual neurons were distinct based on the subcellular location of the four ion channel subunits which significantly determine HCN function (Nusser, 2012).
Interspecies deviations of HCN localization may affect possible functions of HCN channels.

| HCN channels at the peripheral nerve endings
The combination of the highly specialized ribbon synapses with glutamatergic neurotransmission ensures quick and reliable sensory transduction to accurately preserve timing information of sound.
Dendritic HCN channels shape excitatory postsynaptic currents (EPSCs) by avoiding summation of signals . This keeps EPSCs brief and allows the maintenance of high EPSC frequencies but is bought by a potential reduction in EPSC amplitudes. That may be one reason for the very strong currents initiated at IHC synapses (Grant et al., 2010), which has been supported by findings that large EPSCs enhance the precision of spike timing . A fast instantaneous component of I h observed with pulsed hyperpolarization was found to grow maximal in HCN3 and HCN4 channels that presented the slowest activation kinetics. This pre-hyperpolarization enhancement was determined to be more significant at higher frequencies of hyperpolarization which may be caused by incomplete deactivation of HCN4 (Mistrik et al., 2006), representing one way to reduce AP rates. The loss of HCN4 in adult CBA/J and C57Bl/6N mice but not in Ly5.1 mice and other species likely represents a loss of this putative function.

| Somatic HCN channels
Most of the HCN channel protein content was located at the perisomatic membrane of SGNs. The main function of somatic HCN channels is to control the RMP. A fraction of HCN channels is tonically open at rest. In SGNs, 20% of I h current was estimated to be active at the RMP (Kim & Holt, 2013). This actively stabilizes the RMP leading to bigger voltage responses through a decreased input resistance (He et al., 2014;Kim & Holt, 2013). The bigger the size of a SGN soma and the more HCN channels that are open at rest, the lower this input resistance will be. This membrane shunting effect supports fast-rising and precisely timed synaptic events (Baumann et al., 2013;Khurana et al., 2012) and may facilitate auditory signal transmission across the big soma through a "depolarized RMP." The soma of bipolar neurons increases the capacitance that theoretically decreases amplitudes and delay signal propagation between peripheral and central axon as we modeled previously (Rattay et al., 2013). Acute block of I h resulted in RMP hyperpolarization (Kim & Holt, 2013). HCN channels generate an inward current and can even depolarize a neuron. Since there are no voltage-gated sodium (Na v ) channels located at the soma membrane to carry the depolarization current across the neuron cell body (Hossain et al., 2005),

| SGN somata clusters
Tight association of two to eight neuron cell bodies into SGN clusters are very typical in humans and are frequently seen in some strains of mice. This is almost exclusively associated with a lack of myelin around the soma. Missing somatic myelination is the rule in humans, as only a few neurons contain a thin myelin layer (Arnold, 1987;Potrusil et al., 2012). In contrast, most other mammals reveal fully myelinated auditory neurons. Our immunostaining identified auditory neuron clusters as HCN hot spots containing the highest amounts of HCN1, HCN2, and HCN4, with the most intense staining at adjoining membranes. Recently, the presence of HCN channels was found to alter the phase but not the amplitude of local field potentials (Sinha F I G U R E 6 HCN channel co-expression in SGNs. Co-expression of HCN1, HCN2, and HCN4 shown in spiral ganglion neurons (SGNs) of 33-day-old CBA/J mice and in human. Confocal microscopy was used to unravel the distribution of these channels at the neuronal membrane in basal-middle turn (a-r). Colorimetric immunohistochemistry showed co-expression in human SGNs in the apical-middle turn (e-u). a-d, HCN1 and HCN2 co-expression in CBA/J. a, Co-expression of HCN1 and HCN2 channels at the neuronal membrane. b, c, Single channel expression for HCN1 (b) and HCN2 (c). d, High magnification of the neuronal membrane showed a spotted distribution without color overlapping suggesting homomeric channels. e-g, Co-expression of HCN1 and HCN2 in human showed a more balanced expression of these subunits, asterisks mark putative type II cells. HCN1 (DAB, f) and HCN2 (AEC, g) channels after color deconvolution. h-k, HCN1 and HCN4 co-expression revealed a patchy distribution of staining without color overlaps (h). i-j, Single channel staining for HCN1 (i) and HCN4 (j). k, High magnification of the membrane suggested predominant presence of homomeric channels. l-n, HCN1 and HCN4 co-expression in human (l). Single channel staining for HCN1 (DAB, m) and HCN4 (AEC, n) after color deconvolution showed single-stained neurons for HCN4 with a large central axon initial segment (asterisk). o-r, HCN2 and HCN4 co-expression in mice showed higher correlation of staining with partial overlapping fluorescent light (o). Single channel images for HCN2 (p) and HCN4 (q). r, High magnification of the membrane showed overlapping emission spectra (white color). The presence of heteromeric channels cannot be excluded here. s-u, HCN2 and HCN4 co-expression in human showed a very heterogeneous distribution pattern (s). Single channel staining for HCN2 (DAB, t) and HCN4 (AEC, u) after color deconvolution. v-x, HCN fluorescent staining intensities from different sections in each cochlear turn were plotted as XYdiagrams with its relative mean intensities of HCN1 versus HCN2 (v), HCN1 versus HCN4 (w) and HCN2 versus HCN4 (x) [Color figure can be viewed at wileyonlinelibrary.com] & Narayanan, 2015). Elevated amounts of these channels may boost ephaptic coupling between adjoining cells and are involved in synchronization of the AP firing in Purkinje neurons (Han et al., 2018).
This ephaptic coupling may be favored by the absence of an insulating myelin sheet. In SGNs, a preprocessing of phase characteristics may be especially valuable for the low-frequency phase-locking coding of sound. Why some animals or mouse strains present these neuron clusters frequently and others do not is unclear and will require further comparative studies. A complete lack of somatic myelination in humans favors the importance of ephaptic phase synchronization and may represent more of an evolutionary adaption than something "pathological" as suggested for the Ly5.1 mouse strain.

| HCN channels in pre-and postsomatic segments
The AIS located within the postsomatic segment is regarded as the AP initiation site for neurons, containing a tightly ordered cytoskeletal structure with highest densities of Na v and voltage-gated potassium F I G U R E 7 HCN1 and HCN3 postnatal expression until the onset of hearing. HCN1 and HCN3 were only visible at the neuron soma. Differences in the expression of these channels were evaluated during maturation from the first postnatal day until the onset of hearing at 16 days in C57Bl/6N. a-h, HCN1 expression in 1-, 7-, 9-, and 16-day-old mice. a, b, HCN1 in P1 C57Bl/6N mice showed an apical-basal gradient (a). Staining in neuron soma (b). c, d, Apical-basal gradient was present at P7 (c), staining intensity decreased at the soma and HCN1 was found mainly at the membrane (d). e, f, At P9, the gradient of staining between different turns faded (e), and HCN1 was mainly present at the neuronal membranes (f). g, h, At the onset of hearing, the apical turn presented high expression in neuronal clusters (g) and overall intensity increased in the perisomatic membranes and the cytoplasm (h). i-p, HCN3 expression at 1-, 7-, 9-, and 16-day-old C57Bl/6N. i-j, at P1 no gradient for HCN3 between turns was observed (i), but high expression of HCN3 in the neuronal soma (j). k-p, no visible gradient was detected at P7 (k, l) or P9 (m, n) until onset of hearing (o, p) and the cytoplasmic staining decreases (P7, l; P9, n; P16, p) [Color figure can be viewed at wileyonlinelibrary.com] (K v ) channels. It determines the lowest firing thresholds and improves the temporal precision of spike coding (Lazarov et al., 2018). Selective blocking of somatic HCN channels resulted in a decrease in spike probability while site-selective HCN channel blocking in the AIS increased spike probability. We have found distinct patches of HCN1 and HCN4 in the pre-and postsomatic segments that do not exceed cAIS and pAIS distal endings. The fast kinetic HCN1 may prevent hyperexcitability and control the RMP at the cAIS, avoiding continuous firing in response to small depolarizations (Ko et al., 2016). HCN4 in the cAIS and pAIS could favor strong depolarizations and suppress lower AP frequencies through the effect of prehyperpolarization enhancement at higher frequency of hyperpolarization (Mistrik et al., 2006). We

| HCN channels in type II SGNs
The small type II neurons with their thin nerve fibers differ considerably from the much bigger type I neurons in that they have a complete lack of myelination at the soma and axons. Type II afferent nerve endings of OHCs present the glutamatergic ribbon type synapses like in type I neurons, but OHCs typically release only single vesicles with low probability (Weisz et al., 2009). It is supposed that a single OHC does not trigger an AP but summation of at least six OHCs is necessary to reach the threshold (Weisz et al., 2014).
Peripherin-positive neurons showed different HCN subtype expression. We found all HCN subunits in human and cat, which is consistent with data in guinea pig (Bakondi et al., 2009). Adult C57Bl/6N and CBA/J mice only presented HCN3-LI in type II neurons. These inbred mice strains may have lost substantial expression levels of HCN channels in type II neurons and should not be used to study the influence of HCN channels in these types of neurons. Dendritic type II location may function similar to type I neurons but at a "slower" level. Type II peripheral axons were reported to contain Na v 1.6 within their course in the organ of Corti (Hossain et al., 2005) which corresponds to our HCN4 and HCN2 staining pattern. As a concerted stimulation of several OHCs follows a summation of EPSCs until threshold is reached, an AP would be triggered anywhere along the type II peripheral process within the sensory epithelium. HCN channels could open only at these AP initiation sites and avoid temporal summation. Also, modulatory factors such as cAMP could alter AP initiation and tune type II response.
Different subpopulations of peripherin-positive type II neurons were previously found in humans including a few myelinated (Arnold, 1987;Rattay et al., 2013) that could convey APs up to 4.5 faster than the classical unmyelinated type II neurons (Rattay et al., 2013). The function of these fast type II neurons remains unknown. Higher heterogeneity in human type II SGNs may be associated with higher diversity of HCN subtype expression.

| HCN channels in efferent nerve fibers
We found only HCN2 in the efferent MOC fibers innervating the OHCs. I h is partially activated at rest, and together with low threshold K v (KLT) and high threshold K v (KHT) channels, HCN channels may act in repolarizing the axonal membrane thereby shorting APs.
This may enable a higher AP frequency for the very thin caliber efferent fibers. As HCN2 channels are sensitive to pro-inflammatory molecules following a cAMP rise (Emery et al., 2012), these mediators may influence the maximum AP frequency and tune MOC response at the level of the sensory epithelium. A cAMP rise may lead to a reduced cochlear amplification and could thereby help to protect the sensory cells from further damage over longer time periods.
All investigated species showed this expression pattern.

| HCN channels in hair cells
We located HCN channels mainly in neurons. HCN3 was solely present in the lateral membrane of OHCs in guinea pigs and cats and probably in humans. Membrane preservation may have been insufficient for a clear detection in human postmortal tissue. This location of an HCN channel was not reported previously. We could not confirm previous reports about the presence of HCN1 and HCN2 channels in auditory hair cell stereocilia (Ramakrishnan et al., 2009(Ramakrishnan et al., , 2012. However, other authors also did not find HCN subunits in hair bundles (Horwitz et al., 2010), but HCN1 was found at the basolateral membrane of utricular hair cells (Horwitz et al., 2011). Our findings identified HCN3 at the lateral cell membrane of OHCs overlapping with prestin location, the protein responsible for OHCs motility (Zheng et al., 2000).
HCN3 could add a voltage-gated cation conductance close to the prestin molecules. Chloride (Cl − ), which is important for the conformational change and function of prestin, is also known to regulate HCN channels extracellularly (Frace et al., 1992). As mentioned above, the fast component of I h grows maximally in slow HCN channels such as HCN3. Repeated hyperpolarization led to additive but quickly saturating enhancements of the fast component of I h . However, an increase of intracellular Cl − concentration suppressed this effect (Mistrik et al., 2006). A lowering of the Cl − concentration increased conductivity, thereby changing electrical properties at the lateral membrane of OHCs. How and if HCN3 directly interplays with prestin needs more functional studies, but the remarkable expression site overlap and common modulators suggested some involvement. Since a hyperpolarization is the trigger for HCN channels to open and increase conductivity across a cell membrane, an activation of HCN3 in the lateral membrane appears plausible. I h may also actively counteract deviations of the membrane potential to support fast repolarizations. We could not detect HCN3 in C57Bl/6N, Ly5.1, or CBA/J inbred mouse strains, so the animal model to study HCN involvement in OHC function may be crucial.

| HCN tonotopical gradients and co-expression in neurons
HCN channels are not uniformly dispersed along the tonotopic axis. Gradients depend on HCN subtype, mouse strain, and age.
In most situations, apical neurons displayed higher expression for HCN1 but also high levels of HCN2 and HCN4 channels, contrary to earlier research that indicated HCN1 was more expressed in basal neurons (Liu, Manis, et al., 2014 F I G U R E 8 HCN2 and HCN4 postnatal expression until the onset of hearing. HCN2 and HCN4 were expressed in spiral ganglion neurons (SGNs) and the organ of Corti of C57Bl/6N mice as early as P1 and changed expression during maturation until onset of hearing. a-l, HCN2 expression at 1, 7, 9, and 16 days after birth. a-c, HCN2 was highly expressed at P1 without any visible tonotopical gradient (a). Staining located at the soma (b) as well as in the organ of Corti at afferent fibers underneath the inner hair cell (IHC), (c). d-f, At P7 the staining decreases in neurons (d, e) as well as sensory epithelium (f). g-i, at P9, neurons presented similar staining compared to P7 (g, h) but the intensity underneath the IHC increased and a faint staining underneath the outer hair cells (OHCs) appeared (i). j-l, around the onset of hearing, SGNs presented a more intense staining at the perisomatic neuron membranes compared to P9 and P7 (j, k). The now functional organ of Corti expressed HCN2 in putative type I afferents as well as in the efferent fibers underneath the OHCs (l). m-x: HCN4 expression in 1-, 7-, 9-, and 16-day-old C57Bl/6N. m-o, High intensity staining of HCN4 was present at P1 at the soma membrane of SGNs in each turn (m-n). Prominent staining was visible underneath the IHCs and OHCs corresponding to afferent fibers at that developmental stage (o). p-r: at P7 a tonotopical gradient was visible with most intense reactivity in apical clusters (p). Staining intensity decreased at the neuron membrane (q) but prominent staining was still present in the afferent fibers of the organ of Corti (r). s-u, similar staining pattern was present in P9 neurons (s, t) and organ of Corti (u). v-w, after the onset of hearing, the tonotopical gradient faded with similar levels of staining in all turns (v), while overall staining in single neurons remained constant (w). No visible specific staining was detectable in the organ of Corti (x) [Color figure can be viewed at wileyonlinelibrary.com] strictly control timing precision to maintain their phase-locked sound coding that sets formidable demands. Moreover, binaural cues that utilize a submillisecond interaural time difference for sound localization largely depend on lower frequency neurons (Ison et al., 2017;Kim & Holt, 2013). Only HCN2 and HCN 4 showed tighter spatial association and quantitatively the best correlation, so some heteromeric HCN2/ HCN4 compositions may be present at auditory neuron somata.
More heterogeneity is seen in humans, as several individual neurons predominantly expressed one of the subunits. This phenomenon was not restricted to a certain frequency region and may F I G U R E 1 0 Differential expression of HCN channels in spiral ganglion neuron membrane with aging. HCN staining intensities at the perisomatic membrane of CBA/J and C57Bl/6N strains was semi-quantified for HCN1, HCN2, and HCN4. Age groups spanned 3 months starting from the onset of hearing until 19-month-old mice. a-c, C57Bl/6N semi-quantification of HCN1 (a), HCN2 (b), and HCN4 (c). d-g, CBA/J semi-quantification of HCN1 (d), HCN2 (e), and HCN4 (f). Evoked auditory brainstem responses results are plotted as a second y axis representing pure-tone hearing thresholds (compare to Figure 9). The nonparametric Kruskal-Wallis test and Dunn's post hoc test with Bonferroni correction was applied andp < 0.05 was considered statistically significant represent a higher grade of specialization of SGNs across the entire frequency spectrum of hearing.

| Prehearing postnatal HCN channel expression in mice
At perisomatic membranes, HCN channels control the RMP (He et al., 2014) which is directly related to the rate of intrinsic firing in postnatal neurons (Lin & Chen, 2000). Spontaneous firing activity of SGNs promotes their survival and maturation (Babola et al., 2018).
These APs are a result of spontaneous Ca 2+ spikes in immature hair cells initiated by the release of ATP from the glia-like inner supporting cells (ISCs) via the release of Cl − ions (Tritsch & Bergles, 2010;Tritsch et al., 2007) and start around P10 (Ehret, 1976). Interestingly, HCN2 and HCN4 showed increased I h conductance with increasing external Cl − concentration (Frace et al., 1992). Kim and Holt (2013) demonstrated that HCN1 and HCN2 produce I h at early postnatal stages. HCN4 could not be tested due to embryonic lethality of the HCN4 −/− mice, but was also suggested to be present. These data support the idea that HCN1, −2, and −4 are active components at neonatal stages. The most intense staining of HCN3 in P1 mice indicated a more relevant role of the channel prior to birth. Our data conform with previously reported expression ratios in early postnatal mice SGNs that identified HCN3 as the least expressed subtype (Kim & Holt, 2013). Recently, spontaneous activity was found to be related to HCN channels in neurons of the cochlear nucleus from prehearing mice (Yin et al., 2018). Spontaneous activity plays a major role in maturation and refinement of auditory circuit formation (Blankenship & Feller, 2010;Frank & Goodrich, 2018

| HCN channels in aging
The puzzling mixture of factors that are putatively causative for or associated with the aging process in neurons include a plethora of sources, such as elevated oxidative stress, homeostatic dysregulation,   DNA damage, protein aggregation, and impaired metabolic function as well as many other factors together with genetic predispositions and environmental influences (Mattson & Magnus, 2006). HCN channels were recently identified to be upregulated with aging (Shen et al., 2018). Our C57Bl/6N data on the upregulation of HCN1 and HCN2 and a few changes for HCN4 in ARHL largely confirm their observations. Ups and downs in HCN1 and HCN2 expression levels and considerable differences between mouse strains exposed agerelated HCN expression to be much more complex and that selection of age groups may be critical. In contrast to Shen et al. (Shen et al., 2018), we could not find relevant expression levels of HCN3.

F I G U R E 11
In their study, elevated expression levels of HCN1 in C57Bl/6N mice resulted in increased AP numbers in the basal turn. The spike probability enhancement by HCN upregulation may be involved in a feedback mechanism to combat sensorineural decline especially in the high-frequency region in C57Bl/6. There is a considerable loss of neurons at older ages in all turns (Shen et al., 2018), so HCN1/2 content is upregulated in the neurons that survive. Lowering the threshold of spiking through tonically open HCN channels and enabling fast RMP recovery after an AP may indicate that high sensitive SGNs survived better than low sensitive neurons and even increased their spiking rate.
The situation was very different for the "good hearing" CBA/J mice that retained better hearing performance with advanced age. The 15-month-old mice had a considerable threshold shift throughout all tested frequencies but showed a decrease in HCN1 and HCN2-LI. CBA/J mice started losing hair cells in the apex accompanied with only a modest loss of neurons (Sha et al., 2008).
Since CBA/J mice showed only a mild loss of cochlear neurons (Liu et al., 2019;Ohlemiller et al., 2010), the peripheral input may have been sufficient and an increase in hyperexcitability or enhanced spontaneous activity would not have been induced. HCN1 and HCN2 downregulation may result in a more hyperpolarized RMP, and that would decrease spike activity. This could break the vicious circle of upregulated activity with a loss of sensory input. Hair cell loss may account primarily for this reduced activity, but HCN downregulation could contribute. C57Bl/6N mice seemed to compensate for the loss of sensory cells with hyperexcitability of remaining SGNs that made these neurons likely to be more vulnerable to damage. Measurements of I h in old and young C57Bl/6N mice did not show any changes in V 1/2 or slope factors of I h (Shen et al., 2018) suggesting that the overall proportions of the HCN channel set was unchanged. In both mouse strains, HCN1 and HCN2 channels may be important regulators of SGNs excitability with aging, whereas the constantly expressed HCN4 subtype does not seem to be very involved.
Although not quantified, Ly5.1 presented the highest intensity of the staining for all the HCN subunits, suggesting that this strain is suitable for further physiological examinations. The non-myelinated SGNs also facilitated electrophysiological experiments, and the SGN anatomy was closer to the human situation.

| HCN channels and electrical stimulation with cochlear implants
Cochlear implants (CIs) rely on excitable neurons that can be stimulated through a pulsed electrical field. Electrical stimulation (ES) of SGNs differs considerably from acoustic stimulation, as CIs directly excite voltage-gated ion channels and bypass the IHCs. Under acoustic stimulation, SGNs reveal a wider dynamic range, show higher variability in firing rates, and undergo weaker phase locking (Boulet et al., 2016;Hartmann et al., 1984;Javel & Viemeister, 2000). Previous work highlighted the importance of considering temporal properties in order to improve CI performance (Boulet et al., 2016). Stimulusresponse phenomena produced in type I SGNs, such as refractoriness, facilitation, accommodation, and spike rate adaptation, cannot fully be explained with adapted Hodgkin-Huxley models (Mo & Davis, 1997;Negm & Bruce, 2014). Since HCN channels are considerably involved in shaping the timing of spiking, it seems likely that they also influence the neuronal responses evoked through ES, especially when the stimulation frequency is high. Recent modeling studies have suggested HCN channels to be actively involved in these phenomena, especially at high stimulation rates (Negm & Bruce, 2014). This was comprehensively reviewed recently (Boulet et al., 2016). However, the benefit of high stimulation rates of several thousand pulses/s to improve speech understanding is still under debate (Boulet et al., 2016;Kiefer et al., 2000;Loizou et al., 2000;Wilson et al., 1988). It is not yet fully understood why some CI recipients perform better with higher stimulation rates while others do not. Together with KLT channels, HCN activity can extend the neuron refractory period. Subthreshold pulse trains may reduce the number of open HCN channels and contribute to a reduced excitability (accommodation). HCN channels have also been proposed to be involved in spike rate adaptation through accumulating after hyperpolarization (Negm & Bruce, 2014). Pulsed hyperpolarization that grows maximal in HCN4 channels at higher frequencies (Mistrik et al., 2006) may account for spike rate adaptation with ES.

DECL AR ATION OF TR ANS PAREN C Y
The authors, reviewers and editors affirm that in accordance to the policies set by the Journal of Neuroscience Research, this manuscript presents an accurate and transparent account of the study being reported and that all critical details describing the methods and results are present. ing the free use of the pictograms used for the graphical abstract.

CO N FLI C T O F I NTE R E S T
There is no conflict of interest to declare.

PE E R R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1002/jnr.24754.

DATA AVA I L A B I L I T Y S TAT E M E N T
All data are accessible upon request. Human data are restricted to privacy policies so information is limited as all the subjects were anonymized.