The Cl--channel TMEM16A is involved in the generation of cochlear Ca2+ waves and promotes the refinement of auditory brainstem networks in mice

Before hearing onset (postnatal day 12 in mice), inner hair cells (IHCs) spontaneously fire action potentials, thereby driving pre-sensory activity in the ascending auditory pathway. The rate of IHC action potential bursts is modulated by inner supporting cells (ISCs) of Kölliker’s organ through the activity of the Ca2+-activated Cl--channel TMEM16A (ANO1). Here, we show that conditional deletion of Ano1 (Tmem16a) in mice disrupts Ca2+ waves within Kölliker’s organ, reduces the burst-firing activity and the frequency selectivity of auditory brainstem neurons in the medial nucleus of the trapezoid body (MNTB), and also impairs the functional refinement of MNTB projections to the lateral superior olive. These results reveal the importance of the activity of Kölliker’s organ for the refinement of central auditory connectivity. In addition, our study suggests the involvement of TMEM16A in the propagation of Ca2+ waves, which may also apply to other tissues expressing TMEM16A.

In the developing inner ear, non-sensory inner supporting cells (ISCs) form a transient epithelial structure known as Kölliker's organ (Hinojosa, 1977;Hou et al., 2019). ATP released from ISCs through connexin hemichannels (Mazzarda et al., 2020) activates purinergic receptors in a paracellular manner, leading to cell volume decrease of ISCs and cochlear Ca 2+ transients (Babola et al., 2018;Tritsch et al., 2007). It was proposed that the Ca 2+ -activated Cl -channel TMEM16A, which is expressed in ISCs, might be the pacemaker for spontaneous cochlear activity (Yi et al., 2013). Indeed, spontaneous osmotic cell shrinkage was shown to be mediated by TMEM16A-dependent Clefflux, which forces K + efflux from ISCs and thus the transient depolarization of IHCs (Yi et al., 2013). Thereby, bursting activity of nearby IHCs, which will later respond to similar sound frequencies, becomes synchronized (Eckrich et al., 2018;Harrus et al., 2018;Wang et al., 2015), establishing a possible scenario for tonotopic map refinement in central auditory structures.
Using Ano1 conditional knockout mice, we show that TMEM16A not only modulates ISC volume but also drives the amplification of localized Ca 2+ transients to propagating Ca 2+ waves within the cochlea. Prior to hearing onset, knockout mice show reduced burst firing of neurons in the medial nucleus of the trapezoid body (MNTB), downstream of SGNs and neurons of the cochlear nucleus. Moreover, the frequency selectivity of individual MNTB neurons is diminished shortly after hearing onset (P14) pointing toward reduced refinement of auditory connections. Indeed, neurons from the lateral superior olive (LSO) received twice as many functional MNTB afferents in knockout mice compared to wildtype littermates. Taken together, these results suggest that the Ca 2+ -activated Cl -channel TMEM16A plays a significant role in the propagation of Ca 2+ waves and contributes to the refinement of auditory brainstem circuitries prior to hearing onset.

TMEM16A is required for the generation of cochlear Ca 2+ waves
The Ca 2+ activated Cl --channel TMEM16A is expressed in ISCs of Kölliker's organ (for a schematic representation of a part of the organ of Corti, see Figure 1A; for a differential interference contrast [DIC] image, see Figure 1B; Wang et al., 2015;Yi et al., 2013) and is activated by an ATP-induced increase in Ca 2+ concentration (Wang et al., 2015;Yi et al., 2013). The opening of TMEM16A triggers Clefflux, followed by K + efflux and cell shrinkage. The ensuing rise of extracellular K + drives electrical activity of immature IHCs (Wang et al., 2015). To assess the role of TMEM16A in the developing cochlea and the impact of TMEM16A-dependent cochlear signaling on the development of auditory brainstem nuclei, we disrupted Ano1 in the inner ear. This was achieved by mating our floxed Ano1 line (Ano1 fl/fl ) (Heinze et al., 2014) with a line expressing Cre-recombinase under the control of the Pax2 promoter (Ohyama and Groves, 2004), which is active in the otic placode (Lawoko-Kerali et al., 2002). This line is subsequently referred to as cKO mice. Ano1 deletion was confirmed by immunohistochemistry ( Figure 1-figure supplement 1A) and Western blot analysis (Figure 1-figure supple ment 1B). Importantly, organs of Corti of cKO mice showed no obvious morphological defects. The development of the tectorial membrane and the morphology of the hair cells appeared normal before hearing onset (P6), at hearing onset (P12), or in the weeks thereafter (3 weeks and 6 weeks after birth) ( Figure 1-figure supplement 1C and D), indicating that TMEM16A and TMEM16A-dependent activity of Kölliker's organ is not essential for the morphological development of the organ of Corti.
The online version of this article includes the following source data and figure supplement(s) for figure 1: Source data 1. Source data for Figure 1.    Video 1. TMEM16 A is required for the generation of spontaneous volume changes in Kölliker's organ (KÖ). Time-lapse imaging (one image per second) reveals spontaneous volume changes of inner supporting cells of in a wildtype mouse cochlea (P7), which propagate in a wave-like manner up and down the cochlea turn. In contrast, volume changes are almost absent in the cochlea isolated from a cKO littermate (the bottom of the video shows erythrocytes moving in a blood vessel). Images were processed using a custom-written ImageJ macro and the ImageJ software. Each frame was subtracted from an average of five preceding frames to highlight the changes in light scattering caused by the changes in cell volume. The video (seven images per second) shows the top view of an area from an isolated cochlea turn. KÖ, Kölliker's organ; IHC, inner hair cells.
https://elifesciences.org/articles/72251/figures#video1 Video 2. TMEM16A is required for the propagation of cochlear Ca 2+ waves. Time-lapse imaging (one image per second) reveals spontaneous Ca 2+ signals in the inner supporting cells that propagate up and down the cochlear turn in a wildtype mouse cochlea (P7). In contrast, Ca 2+ waves are reduced to local Ca 2+ events in the cochlea of a cKO littermate. Images were processed using a custom-written ImageJ macro and the ImageJ software. Each frame was subtracted from an average of five preceding frames to highlight the changes in Ca 2+ concentration. The video (seven images per second) shows the top view of an area from an isolated cochlea turn. KÖ, Kölliker's organ; IHC, inner hair cells.
TMEM16A-dependent cochlear activity modulates the burst-firing pattern of MNTB neurons By increasing the K + concentration, TMEM16A-dependent activity of Kölliker's organ leads to the generation of Ca 2+ action potentials in IHCs. This is followed by Ca 2+ -dependent exocytosis of glutamate at the IHC synapse, which drives burst firing of action potentials in SGNs (Wang et al., 2015). The bursting activity is then relayed to central auditory neurons (Babola et al., 2018;; Figure 2A shows a schematic representation of the auditory brainstem) and is believed to be important for the proper development of synaptic contacts and tonotopic maps (Clause et al., 2014;Clause et al., 2017).
Since TMEM16A is neither expressed in SGNs nor in CN, MNTB, and LSO neurons in wildtype mice and expression in the brainstem was limited to vascular smooth muscle cells (Figure 3-figure supplement 1), we primarily attribute the severely altered burst-firing activity to impaired Kölliker's organ activity in cKO mice.    from individual MNTB neurons recorded from mice before hearing onset (P8) in wildtype and cKO littermates. Dotplot graphs show respective interspike interval (ISI) distributions for 100 s of spontaneous discharge activity. On top of each dotplot raster is a 10 s period of original spike trains. Note that the wildtype MNTB neuron shows prominent burst firing, which is either strongly reduced (D) or absent in cKO mice (C, E). (F) Quantification of spike bursting patterns by calculating the coefficient of variation of ISIs yields significant differences between wildtype (n = 14) and cKO units (n = 15) (Mann-Whitney rank-sum test: p=0.006); also shown are boxplots indicating medians and 25% and 75% quartiles. Source data 2. Source data for Figure 2H.

Frequency selectivity of MNTB neurons is reduced in cKO mice
To test whether the changes in burst-firing patterns of cKO MNTB neurons have consequences on neuronal function after hearing onset, auditory brainstem responses (ABRs) were measured at P13-14. cKO mice had similar ABR thresholds in response to stimulation with clicks or tone bursts at 6, 12, and 24 kHz as wildtypes ( Figure 4A and B, Supplementary file 1b; n = 6 WT; n = 7 cKO). In both genotypes, ABRs to click stimuli of various intensities (40-100 dB) mainly consisted of three waves (labeled I-III), which were comparable in latency and amplitude (Figure 4-figure supplement 1, Supplementary file 1c and d). These data indicate that cKO mice have a normal sensitivity to sound stimulation and normal temporal precision of the spiking response to sound onset in the lower auditory pathway.
Next, we assessed whether the disruption of TMEM16A-dependent cochlear activity affects the frequency tuning properties of MNTB neurons. Therefore, the frequency response areas (FRAs) of single MNTB neurons were acquired in four cKO and four wildtype littermates using in vivo electrophysiology and tone burst stimulation. Juxtacellular recordings were performed at P14, that is, shortly after the onset of hearing to avoid possible compensatory effects of acoustically driven activity on neuron responsiveness (Werthat et al., 2008;Bogart et al., 2011). The characteristic frequencies (CFs) of the recorded MNTB neurons, that is, the frequency value at which the neuron is excited with the lowest intensity , ranged between 5.3 and 30.5 kHz and did not differ between the two groups (mean CF ± SEM: WT = 15.4 ± 1.1 kHz [n = 25]; cKO = 16.2 ± 1.3 kHz [n = 32]; p=0.51, Student's ttest). MNTB neurons recorded from wildtype mice showed the typical V-shaped FRAs with acoustically driven excitation sharply narrowing toward lower intensities ( Figure 4C). The filter characteristics of the FRAs were quantified by the Q n -value, a measure of the unit's sharpness of tuning, which is calculated as the ratio of CF to bandwidth at 10, 20, and 30 dB above threshold (e.g., Q 10 = CF/BW 10 with BW 10 = bandwidth at 10 dB above threshold). For wildtype mice, the median [25%, 75% quartiles] was Q 10 = 5.5 [4.7, 9.2], Q 20 = 4.6 [3.8, 6.8], and Q 30 = 3.6 [3.4, 5] ( Figure 4E). Neurons recorded from the cKO littermates (n = 32) had significantly broader excitatory response areas, that is, lower frequency selectivity as indicated by significantly lower Q-factors at all three above-threshold levels ( Figure 4D Figure 4F). Overall CF threshold levels tended to show a larger variability in knockout compared to wildtype mice (cKO: -7.3 dB SPL to 48.3 dB SPL; WT: -8.4 dB SPL to 18.7 dB SPL). Furthermore, the sound-evoked firing properties of the MNTB neurons in cKO mice were also affected. The rate-level functions at CF showed significantly lower firing rates for sound intensities at and above 10 dB SPL of tone-burst stimulation in comparison to wildtype littermates (effect of strain p<0.001, effect of intensity p<0.001, interaction strain × intensity p=0.006; two-way ANOVA) ( Figure 4G). The maximal firing rate of individual neurons in response to any CF/intensity combination was markedly diminished in cKO mice (mean FR ± SEM: WT = 263.6 ± 12.4 action potentials/s, n = 25; cKO = 223.9 ± 11.1 action potentials/s, n = 32; p=0.015, t-test) ( Figure 4H). Apparently, MNTB neurons in cKO mice achieve rates that are typically observed in wildtype littermates. Taken together, these data demonstrate that frequency selectivity and sensitivity to acoustic stimulation in single MNTB neurons are impaired upon disruption of Ano1 in the cochlea.
Since spontaneous burst activity of auditory neurons might promote the targeting and refinement of their projections as suggested for the developing visual system (Torborg et al., 2005), we assessed the synaptic and topographic refinement of MNTB and LSO neurons. MNTB neurons were activated via photolysis of caged glutamate. A 405 nm continuous diode laser was used for illumination. Laser flashes were delivered through a light guide of 20 µm diameter, which produced circular spots of 20 µm diameter in the focal plane. Laser pulses of 10 ms duration were delivered with 6 s delay time between uncaging sites. The number of functional inputs on individual LSO neurons was assessed by whole-cell current-clamp recordings. In a pilot experiment, the distance was measured at which glutamate uncaging produces action potentials in MNTB neurons. The mean distance (± SEM) was 19.2 ± 0.8 µm mediolaterally and 20.0 ± 1.3 µm dorsoventrally, indicating that glutamate uncaging was locally restricted to a surface area of 20 × 20 µm 2 and that light scattering that might influence the amount of uncaged glutamate in the tissue was negligible ( Figure 5-figure supplement 1A-C).
The online version of this article includes the following source data and figure supplement(s) for figure 4: Source data 1. Source data for Figure 4A and B.
Source data 2. Source data for Figure 4E-H.  defined as the maximal distance of responsive uncaging sites along the mediolateral axis, was twice as large ( Figure 5A, B and D; mean input width ± SEM: WT = 18% ± 3% of the MNTB's mediolateral length; cKO = 36% ± 5% of the MNTB's mediolateral length; p=0.0073, Student's t-test). The large increase in MNTB input width in cKO mice reveals that LSO neurons located in the medial highfrequency region of the nucleus not only received input from neurons of the high-frequency (medial) region of the MNTB, but also from neurons in the mid-nuclear region tuned to lower frequencies.
For additional examples comparing the size and width of MNTB input areas between wildtype and cKO mice, see Figure 5-figure supplement 1D and E. These data strongly point toward an impairment of the tonotopic refinement of MNTB-to-LSO projections in cKO mice.

Discussion
Kölliker's organ was identified as the origin of pre-sensory cochlear activity, which was suggested to serve the refinement of auditory circuits (Tritsch et al., 2007). The Ca 2+ -activated Cl --channel TMEM16A, which is expressed in Kölliker's organ shortly before hearing onset, mediates spontaneous osmotic cell shrinkage of ISCs, which forces K + efflux and thus the transient depolarization of IHCs (Wang et al., 2015;Yi et al., 2013). Thereby, bursting activity of nearby IHCs, which will later respond to similar sound frequencies, becomes synchronized (Eckrich et al., 2018;Harrus et al., 2018;Wang et al., 2015), establishing a possible scenario for tonotopic map refinement (Johnson et al., 2011;Tritsch et al., 2007). Here, we demonstrate that disruption of TMEM16A in the inner ear impairs prehearing cochlear activity and spontaneous burst firing as well as sensitivity and frequency selectivity of sound-evoked firing of MNTB neurons. Finally, we show that these alterations in prehearing auditory activity severely impair the refinement of the MNTB-LSO pathway.
Spontaneous activity of Kölliker's organ is characterized by recurrent ATP-dependent changes in cell volume and Ca 2+ transients of ISCs. The decreases in cell volume reflect the passive movement of water associated with Clefflux upon activation of the Ca 2+ -activated Cl --channel TMEM16A (Wang et al., 2015;Yi et al., 2013). In WT mice, we detected spontaneous Ca 2+ waves and volume changes of ISCs as reported previously (Tritsch et al., 2007). In cKO mice, however, volume changes were absent and Ca 2+ waves were reduced to local Ca 2+ transients. Changes in cell volume can serve as a mechanical stimulus that activates robust and large ATP release (Praetorius and Leipziger, 2009) and, in turn, ATP release through connexin hemichannels is attributed to the propagation of Ca 2+ waves through the cochlear epithelium (Anselmi et al., 2008;Schütz et al., 2010;Mammano, 2013;Jovanovic and Milenkovic, 2020;Mazzarda et al., 2020). Therefore, TMEM16A-dependent volume changes may amplify ATP release via connexin hemichannels and thus promote the propagation of local Ca 2+ transients as Ca 2+ waves. ATP-mediated amplification of the Ca 2+ signals may further activate Ca 2+ -activated TMEM16A Clchannels in a positive feedback mechanism. First support of this model comes from our observation that the purinergic receptor blocker suramin abolished Ca 2+ waves in WT cochlea (as also shown by Tritsch et al., 2007) but did not affect the size of Ca 2+ transients in cKO mice.
To analyze the impact of the disruption of TMEM16A on the auditory pathway, we performed in vivo juxtacellular recordings in the MNTB of cKO mice before and shortly after hearing onset. Burst-firing pattern of MNTB neurons was often completely absent or severely diminished in P9 cKO mice. MNTB neurons from P14 cKO mice also showed markedly diminished firing rates, indicating that they cannot respond to moderate-to-high-intensity stimulation with sufficient firing rates that are normally observed in wildtype littermates. As high sound intensities are coded by high firing rates, interaural-level differences might be affected and hence also sound source localization in cKO mice. Furthermore, individual MNTB neurons recorded from P14 cKO mice responded to sound stimuli from a much larger frequency range compared to wildtype mice. The reduced frequency selectivity could result from superfluous functional projections from globular bushy cells (GBCs) onto MNTB neurons. At P2, MNTB neurons receive on average 9.3 inputs from GBCs, which after a short period of intense competition and pruning end up in one calyx of Held, usually by P9 (Holcomb et al., 2013). Because the elimination of excessive inputs is probably activity-dependent (Holcomb et al., 2013), reduced bursting activity in cKO mice might result in MNTB neurons that receive more than one input and thus show broadened FRAs.
To test whether projection patterns between auditory nuclei in cKO mice are indeed altered, we examined MNTB projections to the LSO, a part of the sound localization pathway in which refinement has been well studied (Clause et al., 2014;Kim and Kandler, 2003;Müller et al., 2019). The necessary spatial resolution was achieved with a fast galvanometric photolysis system, which allowed a fivefold improvement in area resolution compared to fiber optic-based uncaging systems used in previous studies (Clause et al., 2014;Kim and Kandler, 2003;Kim and Kandler, 2010). In cKO mice, the number of functional connections of single LSO afferents (MNTB input area) and the respective mediolaterally covered area, that is, the area along the tonotopic axis (MNTB input width), was twice as large as in wildtype. Since it has been reported that LSO neurons normally lose about 50% of their afferents during the first two weeks of postnatal development of the mouse (Noh et al., 2010), the present results imply that silencing of synaptic connections was strongly diminished in cKO mice.
Processes that underlie sensory network refinement are well studied in the visual system: Before onset of vision, bursting activity of retinal ganglion cells leads to the refinement of their projections to lateral geniculate nucleus neurons (Wong et al., 1993;Penn et al., 1998). This process depends on the duration of bursts, inter-burst intervals, and the synchronization of bursts (Stellwagen and Shatz, 2002;Torborg et al., 2005;Butts et al., 2007), and was proposed to follow Hebbian plasticity rules (Hebb, 1949). Similar mechanisms also apply for the auditory system, where neighboring IHCs show synchronized bursting activity before the onset of hearing (Tritsch et al., 2007). Notably, the firing patterns are less synchronized in cKO mice since simultaneous K + release is not triggered in the absence of TMEM16A-mediated Clefflux (Wang et al., 2015). Because synchronized bursting activity of neighboring IHCs (Tritsch et al., 2007) is faithfully relayed to the brainstem , it is likely that LSO neurons receive desynchronized MNTB inputs in cKO mice. In WT mice, GABA spillover from MNTB axon terminals can lead to the excitation of nearby synapses via pre-and extrasynaptic GABA A receptors (Weisz et al., 2016;Fischer et al., 2019) and thus likely contributes to the synchronization of MNTB inputs. In cKO mice, however, the shorter bursting will diminish GABA spillover, thus further compromising the synchronization of MNTB inputs. Hence, only in WT mice tonotopically similar axons that overlap in the LSO can be simultaneously active and get strengthened, possibly via glycinergic LTP (Bach and Kandler, 2020), whereas tonotopically distinct axons that have less overlap in the LSO get silenced following associative plasticity rules (Hebb, 1949). This assumption is in agreement with the recent discovery that synchronization of Ca 2+ signals of neighboring outer hair cells (OHCs) is required for proper refinement of afferent projections onto OHCs as well as the maturation of OHCs (Ceriani et al., 2019). The synchronization of OHC activity is also mediated by ATP release from supporting cells, namely, Deiter's cells, during the first two postnatal weeks in the mouse (Jeng et al., 2020).
Besides spontaneous activity of ISCs, IHC-firing patterns are also modulated by cholinergic medial olivocochlear efferents (Glowatzki and Fuchs, 2000;Johnson et al., 2011;Sendin et al., 2014), which transiently innervate immature IHCs (Simmons et al., 1996a;Simmons et al., 1996b;Warr and Guinan, 1979). In mice that lack the α9 subunit of nicotinic acetylcholine receptors (α9 KO mice), this cholinergic input is disrupted (Clause et al., 2014). The fact that α9 KO mice show more subtle changes in the bursting behavior of MNTB neurons in vivo (Clause et al., 2014) than our cKO mice suggests a prevalent role of the activity of ISCs in shaping firing patterns of the early auditory pathway prior to onset of hearing. Still, our results showed reduced synaptic refinement before hearing onset similar to those of α9 KO mice (Clause et al., 2014) and otoferlin KO mice, which exhibit drastically diminished spontaneous activity  due to disrupted glutamate release from IHCs (Roux et al., 2006). While otoferlin KO mice have almost no discernible bursting activity, cKO mice showed a drastic reduction of the number of bursts. In contrast, the number of bursts was not changed in α9 KO mice. Firing rates within bursts did not differ between cKO and WT mice, but were 70-80% higher in α9 KO compared to WT mice. Overall firing rates, however, do not differ between both KO models. Consequently, we infer that the overall firing rate and firing rates within bursts, as well as the number of bursts, do not influence the physiological refinement of the MNTB-LSO pathway. Notably, the duration of bursts was markedly reduced in both KO mice (50% in α9 KO and 56% in cKO mice). Average ISIs, however, were significantly longer in cKO and shorter in α9 KO mice. Thus, it seems that both subtle and drastic changes in the temporal pattern and/or the duration of bursts can lead to a severe disruption of tonotopic maps. These changes have permanent consequences since tonotopic refinement of MNTB-LSO projections does not continue after hearing onset Walcher et al., 2011). Thus, α9 knockout mice display diminished sound localization and sound frequency processing after hearing onset as revealed by behavioral studies (Clause et al., 2017).
Whether TMEM16A-mediated prehearing activity has a similar impact on hearing after hearing onset remains to be determined. While ABR measurements in our study did not reveal gross alterations in signaling along the lower auditory pathway, our recordings from individual auditory brainstem neurons showed significantly elevated sound thresholds. We note that the discrepancy between significantly indistinguishable ABR thresholds and significantly elevated auditory thresholds of single MNTB neurons could arise for several reasons. First, by the nature of population vs. single-neuron responses, a direct correspondence is not warranted. Secondly, we speculate that the poorer tonotopic refinement of the auditory circuitry might have contributed to the higher thresholds of MNTB neurons involving cochlear excitation from IHCs off the peak of the traveling wave. However, other reasons cannot be excluded and a trend toward higher ABR thresholds was apparent for 6 and 24 kHz tone bursts.
The elevated sound thresholds and frequency selectivity that we observed in cKO mice at P14 could indicate impaired auditory processing after hearing onset. This is supported by findings showing that mutations in ion channels (hemichannel, gap junctions, and Ca 2+ channels) involved in prehearing activity of Kölliker's organ are linked to congenital deafness . Recent publications have established a causal connection between the bursting patterns of the ascending auditory system and the tonotopic refinement of MNTB projections to the LSO (Clause et al., 2017;Müller et al., 2019), and our work supports this notion. Moreover, our results link the propagation of Ca 2+ waves in developing supporting cells as a peripheral mechanism contributing to the refinement of ascending circuits.

Mouse strains
Mouse care and usage were in accordance with the German animal protection laws and were approved by the local authorities (license numbers: 33.9-42502-04-11/0439; TVV 06/09 and TLV UKJ-17-006). Ano1 fl/fl mice were crossed with a Pax2 Cre mouse line (a gift from A. Groves, House Ear Institute, Los Angeles) to obtain Ano1 ear conditional knockout mice. Throughout the article, these mice are referred to as cKO mice. Information about the mouse strains including genotyping primers can be found in Heinze et al., 2014 andOhyama andGroves, 2004. Ano2 knockout mice were a gift from T. Jentsch, Leibniz Institute for Molecular Pharmacology (Germany), and were described in Billig et al., 2011.

Western blot
Modioli including the SGNs were isolated from Ano2 knockout mice and their wildtype littermates (P5). For immunoblots, the tank blotting Mini-Protean Tetra system (Bio-Rad) and transfer buffer with methanol (25 mM Tris, 192 mM glycine) were used. 10 µg of proteins were separated in 5% acrylamide gels. Proteins were blotted on polyvinylidene membranes (Roti-PVDF, Roth). Protein transfer was controlled by staining with Ponceau S (0.2% Ponceau S, 3% acetic acid). Blocking was done in Tris buffered saline (TBS) supplemented with 0.05% Tween-20 and 5% dry milk powder. TMEM16B was detected using the same antibody as described in the immunohistochemistry section (1:1000). It was incubated overnight at 4°C in TBS-T with 1% dry milk powder. Washing was done with TBS/0.05% Tween-20. The secondary antibody (goat anti-guinea pig IgG antibody-HRP, Merck, 1:1000) was incubated at RT for 2 hr in TBS-T with 1% dry milk powder. Signals were visualized by chemiluminescence (Amersham ECL Prime Western Blotting Detection Reagent, GE Healthcare). Documentation was done with the camera system Stella 3200 (Raytest) and the Xstella software.
For DIC imaging, acutely isolated cochleae were transferred to an inverted microscope (Axio Observer.Z1, Zeiss). A time-lapse image series was taken for 20 min with a ×25 water-immersion objective. One image per second was acquired (1392 × 1040 px) using a CCD camera (Photometrics Cool Snap HQ 2 , Visitron) and the Metamorph imaging software (Visitron).
Ca 2+ signals were visualized by Ca 2+ imaging of cochleae that had been bath-loaded with 10 µM Fura-2 AM (Invitrogen) and 0.02% pluronic acid F-127 (Invitrogen) in ACSF at RT for 35-45 min. Loading was followed by another 30 min incubation in ACSF. Cochleae were imaged on an upright microscope (Axio Observer.Z1, Zeiss) using a ×25 water-immersion objective. Using the MetaFluor imaging software (Visitron), 400 time-lapse ratio images (1392 × 1040 px) were obtained at a rate of one image per second. Therefore, pairs of images were acquired at alternate excitation wavelengths (340/380 nm). The emission was filtered at 500-530nm.
Images acquired during DIC-and Ca 2+ imaging were processed using a custom-written ImageJ macro and the ImageJ software (NIH) deposited at https://github.com/alenameis/ANO1-hearingdevelopment.git (copy archived at swh:1:rev:4af9cb8bc927320653b556eab12000e675a61e12; Maul, 2022). Each frame (time point t n ) was subtracted from an average of five preceding frames: ] . The mean areas and frequencies of volume changes and Ca 2+ events were analyzed within a 10,000 µm 2 area of Kölliker's organ that was chosen from the center of the field of view. The frequencies of volume changes and Ca 2+ events were analyzed by counting the number of events per second. The area of each event was measured in square micrometer using the Metamorph or MetaFluor imaging software for volume changes or Ca 2+ signals, respectively.
To be regarded as an event, volume changes had to match the following two criteria: first, events had to show an increase in DIC contrast above 10% of the baseline level. Baseline contrast levels were measured from an area of the cochlea that did not show volume changes (i.e., an area outside of Kölliker's organ and the IHC region). Second, areas with changes in DIC contrast had to be larger than the surface area of one ISC.
To be regarded as an event, Ca 2+ signals had to match the following two criteria: first, events had to show an increase in fluorescence intensity greater than 10% of the baseline level. Baseline fluorescence levels were measured from an area of the cochlea that did not show Ca 2+ events (i.e., an area outside of Kölliker's organ and the IHC region). Second, areas with changes in fluorescence had to be larger than the surface area of one ISC. Videos were generated with the ImageJ software (seven images per second) and annotated using Premiere Pro (Adobe).

In vitro electrophysiological recording and functional mapping
The brain was removed from mice (P9-11) and the forebrain and parts of the cerebellum were dissected to isolate the brainstem. Coronal brainstem slices (300-400 µm) were cut on a vibrating microtome (Leica VT1200S) in ice-cold dissection solution containing (in mM) 50 NaCl, 2.5 KCl, 1.25 NaH 2 PO 4 , 3 MgSO 4 , 0.1 CaCl 2 , 75 sucrose, 25 D-glucose, 25 NaHCO 3 , and 2 Na pyruvate (280 mOsmol, pH 7.2). The solution was oxygenated with 95% O 2 and 5% CO 2 during dissection. A single brain slice per animal was obtained, which contained the MNTB and LSO to reach a maximum possible preservation of connections between these two nuclei. Slices recovered for 30 min at 36°C and for an additional 30 min at RT in oxygenated recording solution containing (in mM) 125 NaCl, 2.5 KCl, 1.25 NaH 2 PO 4 , 2 MgSO 4 , 2.5 CaCl 2 , 18 D-glucose, and 25 NaHCO 3 (290 mOsmol, pH 7.2). For electrophysiological recordings, brain slices were transferred to a recording chamber, mounted on an upright microscope (Axio Examiner A.1, Zeiss), and perfused with oxygenated recording solution at a speed of 2 ml/min using a pressure-driven perfusion system (ALA-VM8, ALA Instruments). The LSO was identified using a digital camera (C10600 Orca R 2 , Hamamatsu). Recordings were made from neurons with bipolar morphology from the medial (high frequency) region of the LSO in order to obtain comparable results and because the highest degree of refinement was reported for this frequency region (Sanes and Siverls, 1991). Electrodes had a resistance of 4-5 MΩ (Biomedical Instruments) when filled with intracellular solution containing (in mM) 54 potassium gluconate, 56 KCl, 1 MgCl 2 , 1 CaCl 2 , 5 sodium phosphocreatine, 10 HEPES, 11 EGTA, 0.3 NaGTP, and 2 MgATP (285 mOsmol, pH 7.2). Recordings were done in current-clamp in the whole-cell configuration at RT, and LSO neurons were held at 70 mV. Currents were acquired with an Axon CNS MultiClamp 700B amplifier (Molecular Devices) and lowpass filtered (3 kHz) and digitized (3 kHz) with a Digidata 1440A analog-to-digital converter (Molecular Devices) using the pClamp10 software (Molecular Devices).

Photolysis of caged glutamate
Presynaptic MNTB neurons were localized by local uncaging of 120 µM MNI-caged glutamate trifluoroacetate (Femtonics) that was added to the recording solution. Uncaging of glutamate was controlled by a fast galvanometric photolysis system (UGA40, Rapp OptoElectronic) that was triggered via a TTL impulse. A 405 nm continuous diode laser (DL405, Rapp OptoElectronic) was used as a light source. Laser flashes were delivered through a light guide of 20 µm diameter (Rapp OptoElectronic) with a ×10 water-immersion objective that produced circular spots of 20 µm diameter in the focal plane. For mapping, a virtual grid was defined over (and around) the MNTB harboring 250,300 grid points (spaced 20 µm apart) using the UGA-40 control 1.1 software (Rapp OptoElectronic). Each grid point represented one uncaging site. Laser pulses of 10 ms duration were delivered with 6 s delay time between uncaging sites. This resulted in reproducible LSO-MNTB input maps confirmed by rescanning in some cases. Only one MNTB map was obtained per mouse.

Reconstruction and analysis of MNTB-LSO input maps
Events were detected with template search methods during a 20 ms window starting from the onset of the laser illumination. Peak amplitudes of the postsynaptic potentials (PSPs) were measured (Clampfit 10.2 software, Molecular Devices). Each uncaging site that evoked PSP amplitudes greater than 2 mV in the recorded LSO neuron was considered as a functional MNTB-LSO connection and marked as colored square according to the following color code: yellow = depolarizations > 2 mV, orange > 10 mV, red = action potential. Stimulation sites that evoked PSPs smaller than 2 mV or no PSPs were considered as unresponsive MNTB areas and were left blank. The total area covered by all colored squares was defined as the MNTB input area. The analysis of the MNTB input area was done in pixels (1 pixel represents 1 µm 2 ). The MNTB input area was calculated by multiplying the size of an uncaging site (~400 px) by the number of colored squares. The MNTB input area was normalized to the cross-sectional area of the MNTB. MNTB borders were determined by two investigators (one of them was blind to the mouse group) using images taken from the slices during the experiment. The result represents the MNTB input area in percent. The MNTB input width was calculated by measuring the maximal distance of stimulations sites that evoked depolarizations greater than 10 mV along the mediolateral axis. This distance was normalized to the mediolateral length of each MNTB. The result expresses the input width in percent along the tonotopic axis of the MNTB. On rare occasions, responses could be elicited from uncaging sites slightly ventral or dorsal to the defined MNTB boundaries in mice of both genotypes. These sites were included in the analysis.

Spatial resolution of glutamate uncaging
The spatial resolution was assessed by direct current-clamp whole-cell recordings from MNTB neurons. The holding potential was -70 mV and the electrode resistance was 34 MΩ when filled with intracellular solution containing (in mM) 54 potassium gluconate, 56 KCl, 1 MgCl 2 , 1 CaCl 2 , 5 sodium phosphocreatine, 10 HEPES, 11 EGTA, 0.3 NaGTP, and 2 MgATP (285 mOsmol, pH 7.2). A virtual 10 × 10 grid was defined over the recorded MNTB neuron with uncaging points spaced 10 μm apart. The 'action potential-eliciting distance' was defined as the maximal distance from the center of the uncaging site to the center of the cell body at which uncaging produced action potentials in the recorded MNTB neuron. The action potential-eliciting distance was measured for the mediolateral and dorsoventral direction of the MNTB. We used the determined action potential-eliciting distance to define the uncaging parameters for the mapping experiment.

In vivo recordings in the MNTB
Juxtacellular single-unit recordings were performed in mice before or right after hearing onset (P8, P14). Animals were anesthetized with an initial intraperitoneal injection of a mixture of ketamine hydrochloride (0.1 mg/g body weight; Ketavet, Pfizer) and xylazine hydrochloride (5 µg/g body weight; Rompun, Bayer). Throughout recording sessions, anesthesia was maintained by additional subcutaneous application of one-third of the initial dose, approximately every 90 min. The MNTB was approached dorsally and typically reached at penetrations depths of 5000-5500 μm.
The experimental protocol for acoustic stimulation and single-unit recording was described in detail previously (Dietz et al., 2012;Sonntag et al., 2009). Briefly, auditory stimuli were digitally generated using custom-written MATLAB software (The MathWorks Inc, Natick) at a sampling rate of 97.7 kHz. The stimuli were transferred to a real-time processor (RP2.1, Tucker-Davis Technologies), D/A converted and delivered through custom-made earphones (acoustic transducer: DT 770 pro, Beyer Dynamics). Two recording protocols were used: (i) pure tone pulses (100 ms duration, 5 ms rise-fall time, 100 ms interstimulus interval) were presented within a predefined matrix of frequency/ intensity pairs to determine the excitatory response areas of single units. CF (sound frequency causing a significant increase of unit's action potential spiking at the lowest intensity), the respective threshold, Q-values (Q n , a measure of the unit's sharpness of tuning calculated as the ratio of CF to bandwidth at n = 10, 20 and 30 dB above threshold), and maximum discharge rates were computed from the response area and used for the next protocol. (ii) Spontaneous neuronal discharge activity was assessed in the absence of acoustic stimulation to determine the average firing rate and CV of ISIs (ISIs = time that passes between two successive action potentials). Regularity of spontaneous discharge activity was quantified by CV, calculated as the ratio of standard deviation of ISIs and mean ISI. Additional analysis was done for MNTB units recorded in P8 mice, where spontaneous spiking discharges are grouped in bursts followed by periods of greatly reduced discharge activity (Sonntag et al., 2009). We used a statistical test based on gamma probability distribution of ISIs to determine the number and duration of bursts, and number of spikes per burst within each single-cell recording.
For the juxtacellular single-unit recordings, glass micropipettes (GB150F-10, Science Products) were pulled on a horizontal puller (DMZ universal puller, Zeitz) to have a resistance of 5-10 M when filled with 3 M KCl. The recorded voltage signals were amplified (Neuroprobe 1600, A-M Systems), bandpass filtered (0.3-7 kHz), digitized at a sampling rate of 97.7 kHz (RP2.1, Tucker-Davis Technologies), and stored for offline analysis using custom-written MATLAB software. Three criteria were used to classify single-unit recordings: (i) changes in the spike height did not exceed 20%, (ii) uniform waveforms, and (iii) signal-to-noise ratio at least 8:1. The principal neurons of the MNTB were identified by the complex waveform of the recorded discharges (Guinan and Li, 1990;Sonntag et al., 2009).
Histological verification of the recording site was done by iontophoretic injection of Fluorogold (4 µA for 7 min). The animal was perfused 4-6 hr after injection with 0.9% NaCl solution followed by 5% PFA. The brain was cut on a vibratome (Microm HM650), and the tissue sections (100 µm thick) were visualized under the fluorescent microscope (Axioskop, Zeiss). An example of a recording site is shown in Figure 2I.

Auditory brainstem response
ABR recordings were conducted as described previously (Jing et al., 2013). In brief, mice (P13-14) were anesthetized intraperitoneally with a combination of ketamine (125 mg/kg) and xylazine (2.5 mg/ kg). The ECG was constantly monitored to control the depth of anesthesia. The core temperature was maintained constant at 37°C using a temperature-controlled heat blanket (Hugo Sachs Elektronik-Harvard Apparatus). Note that in these small immature mice the temperature probe could not be placed rectally, contributing to the relatively long latencies and greater variability of the ABR waves. ABR peaks IV (~5.3 ms) and V (~7 ms) were very small and were thus excluded from analysis. For stimulus generation, presentation, and data acquisition, the TDT System II (Tucker-Davis Technologies) was used that was run by the BioSig32 software (Tucker-Davis Technologies). SPLs were provided in dB SPL root mean square (RMS) (tonal stimuli) or dB SPL peak equivalent (clicks) and were calibrated using a 1⁄4 inch Brüel and Kjaer microphone (model 4939). Tone bursts (6/12/24 kHz, 10 ms plateau, 1 ms cos 2 rise/fall) or clicks of 0.03 ms were presented at 40 Hz or 20 Hz, respectively, in the free field ipsilaterally using a JBL 2402 speaker. The difference potential between vertex and mastoid subdermal needles was amplified (50,000×) and filtered (400-4000 Hz, NeuroAmp) and sampled at a rate of 50 kHz for 20 ms, 1300× to obtain two mean ABRs for each sound intensity. Hearing threshold was determined with 10 dB precision as the lowest stimulus intensity that evoked a reproducible response waveform in both traces by visual inspection.

Statistics and statistical significance
For statistical analysis, the SigmaPlot 12.5 (Systat Software Inc) and GraphPad Prism software were used. Datasets with normal (Gaussian) distribution were assessed by unpaired Student's t-tests (twotailed distribution) if not said otherwise. For datasets with non-Gaussian distribution, the nonparametric Mann-Whitney rank-sum test was used. For analysis of the intensity dependence of ABR amplitudes and latencies and of ABR thresholds across frequencies, two-way ANOVA was used. If the data were normally distributed, the results are displayed as means ± standard error of means (s.e.m.) and otherwise as medians with the respective 25% and 75% quartiles. Mean cumulative distributions of interevent intervals and distribution of ISIs between wildtype and cKO groups were compared using the two-sample Kolmogorov-Smirnov test and the chi-square test, respectively (see Figure 2). To avoid assigning too much weight on high rate-spiking cells, distributions were created by pooling n random ISIs for each cell within the two groups (n = lowest number of events recorded in any of the cells). A statistical test based on gamma probability distribution of ISIs was used to determine the number and duration of bursts and number of spikes per burst within each single-cell recording. In brief, we assumed that action potential firing of MNTB neurons is a random Poisson-like process. In this way, the probability to encounter one ISI in time period τ is p=1-e -λτ , where λ stands for neuronal firing rate. Next, the probability to encounter k ISIs during the time period τ as k-fold convolution of exponential distribution density was calculated, thereby yielding gamma distribution of waiting times for k ISIs. The resulting probability is p =´τ 0 x k−1 λ k e −λs (k−1)! dx . We ran the statistical test through our recorded action potential times, and each spike train where probability stayed lower than 0.01 (p< 0.01) for at least 10 ISIs (k ≥ 10) was recognized as a burst. Significance levels < 0.05 are denoted by *, <0.01 **, and <0.001 ***. No statistical methods were used to predetermine sample sizes. the BMBF (01EW1706) and the DFG to CAH (HU 800/10-1) and grants of the DFG to NS, RR, and TM (priority program 1608). In addition, the work of TM was supported by Fondation Pour l'Audition (FPA RD-2020-10).

Ethics
This study was performed in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. All of the animals were handled according to our local authorities (license numbers: 33.9-42502-04-11/0439; TVV 06/09 and TLV UKJ-17-006).

Additional files
Supplementary files • Supplementary file 1. Distribution of interspike intervals and quantification of auditory brainstem response. (a) Comparison of the distribution of interspike intervals (ISIs) between wildtype and cKO mice. (b) Quantification of auditory brainstem response (ABR) thresholds (mean ± SEM) in response to stimulation with tone bursts at 6, 12, and 24 kHz, or click stimulation. (c) Quantification of peak amplitudes (mean ± SEM) of the first three ABR waves (I-III) in response to click stimuli of various intensities (40-100 dB). (d) Quantification of latencies (mean ± SEM) of the first three ABR waves (I-III) in response to click stimuli of various intensities (40-100 dB).
• Transparent reporting form

Data availability
All data generated or analysed during this study are included in the manuscript and supporting file; Source Data files have been provided and did not change for the revised manuscript.