Mechanotransduction is required for establishing and maintaining mature inner hair cells and regulating efferent innervation

In the adult auditory organ, mechanoelectrical transducer (MET) channels are essential for transducing acoustic stimuli into electrical signals. In the absence of incoming sound, a fraction of the MET channels on top of the sensory hair cells are open, resulting in a sustained depolarizing current. By genetically manipulating the in vivo expression of molecular components of the MET apparatus, we show that during pre-hearing stages the MET current is essential for establishing the electrophysiological properties of mature inner hair cells (IHCs). If the MET current is abolished in adult IHCs, they revert into cells showing electrical and morphological features characteristic of pre-hearing IHCs, including the re-establishment of cholinergic efferent innervation. The MET current is thus critical for the maintenance of the functional properties of adult IHCs, implying a degree of plasticity in the mature auditory system in response to the absence of normal transduction of acoustic signals.

T he sense of hearing relies on sound evoking the displacement of stereociliary bundles projecting from the upper surface of hair cells. Hair bundle displacement causes the opening of mechanoelectrical transducer (MET) channels located at the tips of the shorter and middle rows of stereocilia, and the transduction of acoustic information into a receptor potential inside the hair cells. Stereocilia are interconnected by several extracellular linkages 1 , including tip links that are directly involved in gating the MET channels. While tip links are formed by cadherin 23 (CDH23) and protocadherin 15 (PCDH15) 2 , several other proteins are involved in tensioning and anchoring these links 3 . In the absence of sound, a fraction of the MET channels are normally open, allowing a standing depolarizing current into the hair cells that is mainly carried by K + and Ca 2+ . Calcium entry through the MET channels causes adaptation and reduces their open probability 4 . However, the endolymph that surrounds the hair bundle has a low Ca 2+ concentration in both pre-hearing (second post-natal week:~300 μM) 5 and adult (~20-40 μM) 6 cochleae, resulting in a larger in vivo resting open probability of the MET channel. The driving force for the MET current is provided by an electrical gradient between the endocochlear potential (pre-hearing: ∼ + 20 mV; adult: ∼ + 90 mV) 7 and the hair-cell resting potential (~−60 mV) 8 .
Although the resting MET current contributes to setting the membrane potential of pre-hearing and adult inner hair cells (IHCs) 5,8 , its functional implication for IHC development is currently unknown.
Before the onset of hearing, IHCs fire spontaneous Ca 2+ action potentials (APs) 9,10 , which play a crucial role in promoting neuronal survival 11 and the refinement of the immature auditory system 12,13 . Functionally mature IHCs respond to sound stimulation with graded receptor potentials. This change in the receptor potential requires a completely new complement of basolateral ion channels that appear just before the onset of hearing 9,14 . We hypothesize that, in the immature cochlea, the functional MET channels together with the maturing endocochlear potential 7 and endolymphatic Ca 2+ concentration 5 , trigger the developmental switch from spiking IHCs to high-frequency signal transducers.
Using transgenic mice to manipulate the expression of molecular components of the MET apparatus in vivo, we demonstrate that the MET current is required for their developmental switch from immature to functional sensory receptors. Moreover, we found that the resting MET current in adult IHCs is required for maintaining their mature biophysical identity. The loss of the MET current in adult IHCs causes them to revert to an immature, pre-hearing phenotype and they re-acquire efferent innervation.

Results
MET channel failure prevents the maturation of IHCs. Protocadherin 15 (PCDH15) is localized to the lower part of the tip link where it contacts the top of the shorter rows of stereocilia. Mutations in Pcdh15 lead to hair-bundle disruption and loss of the MET current from early postnatal days 15 . Therefore, we investigated whether the absence of the MET current in PCDH15 deficient mice affected hair cell maturation. IHCs acquire most of their mature biophysical and morphological characteristics around the onset of hearing (~P12). This includes the expression of a rapidly activating large conductance Ca 2+ -activated K + current (I K,f ) 9 and a K + current (I K,n ) 14 carried by KCNQ4 channels. We found that I K,f was present in adult control IHCs (Pcdh15 +/av3J : Fig. 1a, 3.8 ± 0.5 nA at −25 mV, n = 10, mean ± SEM; see Methods for details) but absent in Pcdh15 av3J/av3J mutant mice (Fig. 1b). The absence of I K,f in Pcdh15 av3J/av3J unmasked the inward Ca 2+ current (−252 ± 23 pA, n = 12). The size of I K,n was significantly reduced in Pcdh15 av3J/av3J IHCs (−68 ± 11 pA at −124 mV, n = 8, P < 0.0001, t-test; see The adult-type currents (I K,f and I K,n ) were only present in IHCs from control Pcdh15 +/av3j mice (a). IHCs from Pcdh15 av3j/av3j mice retained the currents characteristic of immature cells (b , I K and I K1 ). The absence of the rapidly activating I K,f in Pcdh15 av3j/av3j IHCs is also evident when comparing the activation time course of the total outward currents on an expanded time scale (see insets). a, b are representative recordings from 10 (3 mice) and 12 (3 mice) IHCs, respectively. c, d Voltage responses elicited by applying depolarizing current injections to control, Pcdh15 +/av3j c and mutant, Pcdh15 av3j/av3j d IHCs from their resting membrane potentials. Depolarizing current injections caused slow APs in the Pcdh15 avj3/av3j mutant IHC. c, d are representative recordings from 8 (3 mice) and 15 (3 mice) IHCs, respectively. Recordings were performed near body temperature Methods for details) compared to that of Pcdh15 +/av3J control cells (−216 ± 17 pA, n = 8). Pcdh15 av3J/av3J IHCs retained an immature phenotype by expressing the inward (I K1 ) and slowlyactivating outward (I K ) K + currents 14,16 . The size of I K , calculated as the difference between the steady-state and early (I K,f ) current was 2.4 ± 0.5 nA (Pcdh15 +/av3J ) and 1.6 ± 0.2 nA (Pcdh15 av3J/av3J ) (P = 0.159, t-test). The physiological consequence of the immature current profile in adult Pcdh15 av3J/av3J IHCs was that, instead of generating rapid and graded voltage responses characteristic of adult IHCs (Fig. 1c), they retained the ability to fire action potentials (APs) (Fig. 1d), a characteristic of pre-hearing IHCs 14 . A similar failure in the acquisition of the mature current profile has also been reported in IHCs from mice lacking other stereociliary bundle proteins, including the transmembrane channellike protein 1 (TMC1) (Deafness mouse) 17 , which is necessary for mechanoelectrical transduction 18 , and the unconventional Myo6 (Snell's waltzer mouse) 19 , which is required for the development of the MET apparatus 20 . The above findings indicate the existence of a link between a functional MET apparatus and IHC maturation. Mutations in or absence of stereociliary proteins early in development normally causes severe morphological defects of the stereociliary bundle (e.g. Myo6; 21 Myo7a; 22 harmonin 23 ). We used targeted knockouts obtained by crossing floxed mice (Ush1c fl/fl and Myo7a fl/fl ) with Myo15-cre +/mice 24 . Because the temporal expression of the cre driven by the hair cell specific Myo15 promoter occurs postnatally (∼P4 in the cochlear apical coil) 24 , the stereociliary bundle develops normally, allowing us to study the functional implications of the missing protein.
A determining factor for the presence of spontaneous AP activity in pre-hearing IHCs (second post-natal week) is the resting membrane potential (V m ), which has to be close to the activation threshold of the Ca 2+ current. In IHCs V m is determined by the balance between an inward, depolarizing resting MET current and outward I K 5 . Therefore, we investigated whether the substantially reduced resting MET current in immature IHCs from Ush1c fl/fl Myo15-cre +/− mice affected IHC physiology. Since the fraction of MET channels open at rest is determined by Ca 2+ -dependent adaptation 4,27 , the V m of IHCs in vitro strongly depends upon the extracellular Ca 2+ concentration. Using current clamp, we measured the depolarizing current required to elicit APs in IHCs at 35-37°C using 1.3 mM Ca 2+ , which is the Ca 2+ concentration present in the perilymphatic solution bathing the IHC basolateral membrane. Starting from the resting V m (Ush1c fl/fl : −72.2 ± 2 mV; Ush1c fl/fl Myo15-cre +/-: −74.9 ± 2 mV, P = 0.257, t-test), a similar amount of current was required to trigger APs in both control Ush1c fl/fl (75 ± 7 pA, n = 10, P9-P10) and Ush1c fl/fl Myo15-cre +/− (72 ± 9 pA, n = 13, P9-P11, P = 0.770, t-test) mice (Fig. 2g). We then measured the resting MET current near the V m of IHCs in 1.3 mM Ca 2+ and at room temperature, and found that it was much smaller (Fig. 2h) than that required to elicit APs (Fig. 2g), even when correcting for body temperature (Fig. 2h). However, the degree of adaptation of the MET channel in vivo should be much reduced as the estimated endolymphatic Ca 2+ concentration during the second postnatal week is 0.3 mM 5 . The application of 0.3 mM Ca 2+ increased the resting MET current in control Ush1c fl/fl IHCs (Fig. 2i, j) to a value that would depolarize them to their AP firing threshold (Fig. 2g: also see ref. 5 ), and even more so when values were adjusted for the temperature difference (Fig. 2i). This would also be the case if we were to subtract the resting MET current measured in 1.3 mM Ca 2+ (Fig. 2h), which is likely to contribute to the resting V m and thus to the current required for eliciting repetitive firing (Fig. 2g). The reduced resting MET current in Ush1c fl/fl Myo15-cre +/− IHCs bathed in 0.3 mM Ca 2+ (Fig. 2i) would not be sufficient to depolarize them to AP threshold. IHCs however would still be able to fire APs in response to ATP-induced Ca 2+ waves from surrounding nonsensory cells 10,28 .
We then investigated whether the absence of harmonin from pre-hearing Ush1c fl/fl Myo15-cre +/− IHCs would also affect their basolateral membrane properties. At P7, P9, and P10, the latter being a time when the MET current in mutant IHCs is almost absent (Fig. 2c, d), the biophysical properties of IHCs, including the size of K + currents, resting membrane potential and cell membrane capacitance, were not significantly different between Ush1c fl/fl and Ush1c fl/fl Myo15-cre +/− mice (Supplementary Table 1, Supplementary Fig. 2). The above data show that the loss of harmonin leads to a gradual disappearance of the MET current without affecting normal development of the basolateral membrane properties of pre-hearing IHCs, including the Ca 2+ channels 26 . After the onset of hearing, while young adult (P14-P15) and adult (P29) IHCs from control Ush1c fl/fl mice express both I K,f and I K,n (Fig. 3a) and exhibit graded voltage responses (Fig. 3c), agematched Ush1c fl/fl Myo15-cre +/-IHCs retained an immature current profile (Fig. 3b, Supplementary Table 2) and the ability to generate APs (Fig. 3d). Auditory brainstem responses (ABRs) showed that Ush1c fl/fl Myo15-cre +/− mice had severe hearing impairment at P16 (Fig. 3e) and were profoundly deaf at P37-45 (Fig. 3f). Adult mice also exhibit the classical head bobbing and hyperactive circling behavior characteristic of Ush1c mutant mice 29 . The above results indicate that the acquisition of the mature basolateral membrane profile in IHCs requires a functional MET current during prehearing stages of development.  Mature IHCs regress without the MET current. The unconventional myosin VIIa (Myo7a) is another hair cell protein that binds to harmonin 30 . In the absence of Myo7a no MET current is present at rest 31 , stereociliary bundles gradually degenerate (Myo7a sh6J/sh6J mice) 31 and the K + current profile of adult IHCs retains pre-hearing characteristics ( Supplementary Fig. 3). Different from pre-hearing Ush1c fl/fl Myo15-cre +/mice (Fig. 2), the size (−648 ± 40 pA at −121 mV, n = 6, 3 mice, P9-P10) and resting open probability (0.026 ± 0.004 pA, n = 6) of the MET current in IHCs of pre-hearing Myo7a fl/fl Myo15-cre +/− mice was similar to that measured in control Myo7a fl/fl IHCs (−725 ± 38 pA, P = 0.225, t-test; 0.038 ± 0.010 pA, P = 0.2318, t-test, n = 4, 3 mice, P10) ( Fig. 4a, b). The normal MET current in pre-hearing Myo7a fl/fl Myo15-cre +/− IHCs (Fig. 4a, b), but not in Ush1c fI/fI :Myo15-cre +/-(P15) Ush1c fI/fI ; Ush1c fI/fI ; Ush1c fI/fI Myo15-cre +/-Ush1c fI/fI Myo15-cre +/-, , , ,  Table 2. e, f Mean ABR thresholds ( ±SD ) for clicks and pure tone stimulation from 3 kHz to 24 kHz obtained from control Ush1c fl/fl and littermate Ush1c fl/fl Myo15-cre +/mice at P16 (e) and P37-45 (f). Ush1c fl/fl Myo15-cre +/mice are almost completely deaf already at P16. Open symbols represent data from single mice. Closed symbols represent averages (mean ± SEM) Ush1c fl/fl Myo15-cre +/cells (Fig. 2), suggests a possible slower turnover of Myo7a, compared to that of harmonin. This is supported by the normal appearance of Myo7a immunostaining in hair cells from one out of four P10 mice tested ( Fig. 4c; Supplementary Fig. 4). Alternatively, the late phenotype in Myo7a fl/fl Myo15-cre +/− IHCs could reflect a distinct, and as yet undefined, role for Myo7a in mechanoelectrical transduction.
The aforementioned experiments ( Fig. 2g-j) show that the resting MET current is required for the generation of spontaneous APs in IHCs during the last days preceding the onset of hearing. The normal MET current in IHCs of Myo7a fl/fl Myo15-cre +/mice meant that superfusion of the estimated endolymphatic Ca 2+ concentration (0.3 mM Ca 2+ ) 5 , caused them to depolarize to near the expected in vivo resting V m and elicited repetitive APs fI/fI (15) Myo7a fI/fI Myo7a fI/fI (P10)   Myo7a fl/fl Myo15-cre +/− 53 ± 3 n = 7; Fig. 4j). ABR thresholds at all frequencies tested were also similar between the two genotypes (two-way ANOVA, P = 0.174).
Despite normal ABR thresholds in P15 Myo7a fl/fl Myo15-cre +/− mice, the MET current became progressively affected from P14 compared to control Myo7a fl/fl IHCs ( Fig.  5a-c). The size of the MET current and the fractional resting open probability of the channel in control IHCs (Myo7a fl/fl : −974 ± 63 pA, 0.108 ± 0.028, n = 6, respectively, measured at −121 mV, P15) were significantly larger than those from Myo7a fl/fl Myo15-cre +/− cells at P14 (−669 ± 96 pA, n = 4, P < 0.05; 0.011 ± 0.014, P < 0.05) and P16 (−377 ± 64 pA, n = 6, P < 0.001; 0.031 ± 0.014, P < 0.05) (one-way ANOVA). The smaller MET current in Myo7a fl/fl Myo15-cre +/-IHCs after the onset of hearing occurred despite the normal appearance of the stereociliary bundles and the presence of tip links as well as prolate-shaped stereocilia tips (P30: Fig. 5d). Using TEM we found that the height and width of stereocilia in the tallest (s1) and second tallest (s2) rows in P37 IHCs were not significantly different between the two genotypes (Fig. 5e). The gradual loss of MET current was followed by a progressive disappearance of the adult-type K + currents (I K,f and I K,n ), such that by P59 Myo7a fl/fl Myo15-cre +/-IHCs exhibited an immature current profile ( Fig. 5f-i; Supplementary Table 3), including a slowly-activating outward (I K ) and an inward rectifier (I K1 ) K + current 14,16 . IHCs from control mice (Myo7a fl/fl :  Table 4) had the characteristic current profile of mature wild-type mice 9,14 .The size of IHCs, which is reported as cell membrane capacitance (C m : Fig. 5i), became smaller with age in both Myo7a fl/fl and Myo7a fl/fl Myo15-cre +/-IHCs, although it occurred earlier in the latter. The smaller C m is comparable to that measured in early postnatal IHCs 14 . The resting membrane potential (V m ) remained relatively constant as a function of age (V m : Fig. 5i). This is because the main K + current contributing to the resting V m in IHCs is I K,n in adult Myo7a fl/fl mice and I K1 in pre-hearing and "adult" Myo7a fl/ fl Myo15-cre +/− mice. The in vivo consequence of this progressive regression in IHC maturity was that while at P15 Myo7a fl/ fl Myo15-cre +/− mice had normal hearing thresholds (Fig. 4j), they were profoundly deaf at P62 (Supplementary Fig. 6). Adult Myo7a fl/fl Myo15-cre +/− mice also exhibit the classical head bobbing and hyperactive circling behavior characteristic of Myo7a mutant mice 32 . Both Myo7a fl/fl and Myo15-cre +/− mice showed normal hearing thresholds ( Supplementary Fig. 6).
The MET current ensures that IHCs retain adult K + currents.
To confirm the hypothesis that the resting MET current is required for IHCs to maintain their mature biophysical characteristics, we set up organotypic cultures from the adult cochlea. For these experiments we used gerbils (P18-P24), instead of mice, because their adult IHCs are more robust and can survive in vitro for the long periods required to test the hypothesis. Initially, cochleae were cultured for 1 to 8 days in the DMEM/F12 medium supplemented with CaCl 2 (final concentration: 1.3 mM Ca 2+ , see Methods). Compared to IHCs recorded from acutely dissected gerbil cochleae (P21-P28: Fig. 6a, f), cells under the above culturing conditions gradually lost I K,f , the most distinctive current expressed in post-hearing IHCs (Fig. 6b, c, f). By the 5 th day in culture, I K,f was no longer present in IHCs while a large I K was evident (Fig. 6c, f). The main differences between the in vivo and our in vitro culture conditions, is that in the latter the IHC stereociliary bundles are not stimulated by sound, the endocochlear potential is absent and the resting open probability of the MET channel is very low because of the high extracellular Ca 2+ concentration (1.3 mM) 4 . Therefore, in vitro, only a very small or no depolarizing inward MET current is likely to be present 5,8 . By keeping IHCs depolarized using high K + in the culture medium (40 mM), as a proxy for the in vivo receptor potential caused by the resting MET current, I K,f was retained for longer periods (Fig. 6d, e, f). The direct contribution of the resting MET current in maintaning the adult IHC biophysical identity was tested by using a medium containing the expected in vivo mature endolymphatic Ca 2+ (40 μM), which substantially increases the resting MET current 4,27 . In 40 μM Ca 2+ (Fig. 6g, h, k), the size of I K,f was significantly higher compared to that measured in 1.3 mM Ca 2+ (Fig. 6b, c, k; P = 0.006, two-way ANOVA) or when 40 mM K + was used (Fig. 6d, e, f: P = 0.0166, two-way ANOVA). When cochlear cultures were incubated with 40 μM Ca 2+ together with 50 μM of the MET channel blocker d-tubocurarine 33,34 (Fig. 6i, j,  l), which has been shown to be non-toxic to IHCs in long-lasting cochlear cultures 34 , IHCs showed a strongly reduced I K,f (Fig. 6k) compared to when 40 μM Ca 2+ was used alone (Fig. 6k, P = 0.0001, two-way ANOVA), but produced similar results to when the open probability of the MET channel was very low (1.3 mM Ca 2+ , black circles: Fig. 6k, P = 0.6155, two-way ANOVA). Blocking the MET channel caused IHCs not only to rapidly lose I K,f but also to express the inward rectifier K + current I K1 , which is normally present in pre-hearing cells 16 (Fig. 6l). I K1 was already evident in 4-day cultured IHCs and became larger over time (253 ± 54 pA, n = 13, P18-P26 + 4-8 days). Although I K1 was also seen in IHCs with partially closed MET channels (1.3 mM Ca 2+ : 103 ± 22 pA, n = 7, P18 + 8 days) it was significantly smaller than that measured in the presence of 40 μM Ca 2+ and d-tubocurarine (P18 + 8 days: 393 ± 153 pA, n = 4, P = 0.0324, ttest).
In the adult mammalian cochlea, the efferent fibers either form axo-somatic synapses with OHCs or axo-dendritic synapses with the afferent fibers that contact IHCs 37 . IHCs are directly innervated by cholinergic efferent endings only during early stages of development, but no longer respond to acetylcholine (ACh) from about P18 onwards 35,36,38 . Adult IHCs from P22 Myo7a fl/fl and Myo7a fl/fl Myo15-cre +/− mice did not respond to extracellular ACh (Fig. 7g; Supplementary Fig. 7a, b) and no longer showed axo-somatic efferent contacts at P20  Fig. 3a. i Size of I K,f , I K,n , membrane capacitance (C m ) and resting membrane potential (V m ) as a function of postnatal age. Number of IHC investigated is shown above or below the symbols (for more details see Supplementary Table 3). Statistical tests are performed with two-way ANOVA with Bonferroni correction. Average data is shown as mean ± SEM (c, i) or mean ± SD (e) ( Supplementary Fig. 7c, d), indicating that the re-wiring of maturing IHCs occurred normally. However, at older ages (P33-P49), while IHCs from control Myo7a fl/fl mice remained insensitive to ACh (Fig. 7e), those from Myo7a fl/fl Myo15-cre +/− mice responded with an instantaneous current (Fig. 7f) that increased the slope conductance at −84 mV (P = 0.0045, ttest, Fig. 7g). In pre-hearing IHCs the ACh-dependent current is mediated by Ca 2+ entering through α9α10nAChRs, which then activates SK2 channels 35,36 . Both currents components were observed when ACh was applied to Myo7a fl/fl Myo15-cre +/− IHCs (Fig. 7f), with the outward current mainly carried by SK2 channels and the inward relaxing current most likely due to cations entering the IHC through α9α10nAChRs from the holding potential of −84 mM 35 .   In order to establish whether re-appearance of ACh receptors in adult Myo7a fl/fl Myo15-cre +/-IHCs was associated with reestablishment of axosomatic efferent synapses, we perfused the cochlea with a solution containing 40 mM KCl. High K + is expected to depolarize the efferent synaptic terminals and trigger the release of ACh-containing vesicles, which generate synaptic currents in IHCs superimposed upon a sustained inward current due to the direct effect of KCl on the cell's basolateral membrane 38,39 . Synaptic inward currents were only seen in Myo7a fl/fl Myo15-cre +/-IHCs (Fig. 7h-j),with a frequency (5.7 ± 0.7 Hz, 715 events, n = 7, 4 mice) and amplitude (98 ± 19 pA, 594 events, n = 7) not significantly different from those previously reported in pre-hearing IHCs (P = 0.68 and P = 0.107, t-test, respectively) 39 . Reversible block of the ACh-mediated currents by 1 μM strychnine (Fig. 7k), a potent blocker of α9α10nAChRs 40 , confirmed the re-expression of these receptors in adult Myo7a fl/fl Myo15-cre +/-IHCs. TEM experiments on P37 mice ( ≥ 3 mice for each genotype) indicated that while IHCs from Myo7a fl/fl mice showed the characteristic axo-dendritic contacts between efferent and afferent fibers (Fig. 7l), Myo7a fl/fl Myo15-cre +/mice underwent major re-wiring with efferents re-forming direct axosomatic innervation with adult IHCs (Fig. 7m).

Discussion
Here we show that the MET current is essential to establish and maintain the key biophysical and morphological properties characteristic of adult IHCs. During the second post-natal week in mice, which is just before the onset of hearing at P12, open MET channels keep the IHC membrane potential at the threshold for Ca 2+ action potentials (APs), promoting their functional differentiation into mature graded sensory receptors able to encode sound stimuli. We also found that adult IHCs require functional MET channels to maintain their adult profile. The absence of the MET current during adult stages caused IHCs to lose their identity, regressing into an immature-like biophysical and morphological profile. Adult IHCs without functional MET channels became re-innervated by inhibitory cholinergic efferent neurons, which normally make transient axo-somatic contacts only with pre-hearing IHCs.
Pre-hearing IHCs of altricial rodents exhibit spontaneous Ca 2+ -dependent APs 5,10,14 that can be modulated by the inhibitory efferent neurotransmitter ACh 38 and by excitatory spontaneous ATP-induced Ca 2+ -waves in non-sensory cells, which synchronize the activity of several IHCs 10,28,41 . Indeed, patterned bursts of APs have been recorded from auditory afferent neurons prior to the onset of hearing in mice 42,43 , and both their survival 11 and the refinement of tonotopic maps in the auditory brainstem 13,44 rely on the firing activity of pre-hearing IHCs. However, the refinement of the auditory pathway mainly occurs during the first postnatal week 11,44,45 , a finding supported by evidence that surgical removal of the cochlea causes a substantial loss of cochlear nucleus neurons mainly during this period 46,47 . Despite the crucial role of APs in developing IHCs, spiking activity is not suitable to detect sound because it would modulate and degrade auditory stimuli in the mature cochlea. Therefore, just before hearing onset immature IHCs undergo major changes involving the acquisition and/or elimination of different ion channels and other membrane proteins 9,12,14 . Intracellular Ca 2+ transients associated with a precise frequency and/or pattern of AP activity can activate transcription factors controlling the expression of proteins linked to development 48,49 . In cochlear IHCs, repetitive Ca 2+ APs during the second postnatal week are driven by a standing depolarizing MET current 5 , resulting from the progressive reduction in endolymphatic Ca 2+ concentration 5 that, by reducing adaptation, increases the MET channel open probability. Therefore, AP activity driven by the greater standing MET current during the few days preceding hearing onset could provide the cue or trigger for the last step of IHC development by influencing gene expression. We demonstrate that genetic manipulations that abolish the contribution of the MET current in vivo during the second postnatal week prevented IHCs from becoming mature auditory receptors. We propose that, during pre-hearing stages of development, the primary function of IHC AP activity could switch from refining the auditory circuitry 13 to promoting the maturation of the sensory receptor itself.
The normal AP activity in IHCs from pre-hearing Myo7a fl/ fl Myo15-cre +/mice meant that their auditory system was able, at least initially, to mature fully as judged by the presence of normal ABRs (P15), absence of axo-somatic efferent terminals on P20-P22 IHCs, and a normal IHC basolateral current profile at P20. However, the loss of IHC MET current after the onset of hearing caused a progressive down-regulation of the adult-type basolateral ion channels carrying I K,f and I K,n . By two months of age, IHCs from Myo7a fl/fl Myo15-cre +/mice have the same size and basolateral membrane currents as pre-hearing cells, including the re-expression of I K1 16 and I SK2 35,36 . A similar progressive loss of I K,f was also observed in cultured cochlear IHCs, in which cells are not stimulated by sound and the resting open probability of the MET current is either reduced or blocked. In vitro experimental manipulations that increased the MET channel open probability (endolymphatic-like Ca 2+ concentration) or mimicked the receptor potential due to sound stimulation (IHC depolarization) were able to maintain the adult current profile of cultured IHCs for longer. Calcium entering via the MET channel has been shown to be required for maintaining the morphological stability of transducing stereocilia in early postnatal hair cells 50 . Although in the presence of mature endolymphatic-like solution Ca 2+ entering through the MET channel seems to be cleared from the stereocilia by the plasma membrane Ca 2+ ATPase type 2 (PMCA2), some of it could diffuse into the cell body 51 and, together with the depolarization by the MET current, potentially regulate gene expression and as such the maintenance of the adult IHC current profile. The only adult-like function that was not affected by the loss of the MET channel was that associated with the synaptic machinery, indicating that once it has matured, a process that is controlled by a critical period of pre-hearing APs 12 , it becomes independent of IHC activity. Loss of ribbon synapses in the adult cochlea occurs following acoustic trauma or ototoxic antibiotics 52 , conditions that cause either a temporary or permanent IHC damage. IHCs from aging mice have also been shown to gradually lose ribbon synapses 53 . The fact that exocytosis in IHCs from Myo7a fl/fl Myo15-cre +/mice was normal at 7 weeks, despite IHCs starting to become smaller, may be due to ribbon loss being a slow process 53 .
While auditory ribbon synapses appeared to be resilient to post-maturational changes in MET function, the efferent fibers were more plastic. The efferent endings make transient cholinergic axo-somatic contacts with immature IHCs 54,55 . After the onset of hearing, efferent terminals only form axo-dendritic synapses with the auditory afferent fibers contacting IHCs 56 and expression of the postsynaptic α9nAChRs and SK2 channels is rapidly down-regulated 54 . Therefore, adult IHCs lose the ability to respond to ACh 38,54 . However, it has been shown that efferent endings can re-innervate IHCs following cochlear damage 57 or in aging mice 53,58 . We found that the re-emergence of axo-somatic efferent innervation with IHCs can also occur in mature mice with no functional MET current.
The above findings, including the reduced size of the IHCs, the direct re-innervation by efferent cholinergic fibers, the complete down-regulation of the adult-like basolateral membrane ion channels and the ability to fire APs, indicate that functional MET channels are required for mature IHCs to maintain their biophysical and morphological identity. Without functional MET channels, IHCs revert to an earlier, pre-hearing stage of development. It would be interesting in the future to determine whether the re-establishment of the immature-like biophysical and morphological profile in IHCs without working MET channels is the first stage of an intrinsic attempt to repair the faulty cochlea by "recapitulating" early development. In any case, this functional regression shows a surprising plasticity in the mature cochlea in response to events that abolish normal MET currents, which could occur due to either noise 59,60 , aging 61 or in forms of progressive hearing loss, such as Usher syndrome type III 62 .

Methods
Ethics statement. All animal studies were performed in the UK and licensed by the Home Office under the Animals (Scientific Procedures) Act 1986 and were approved by the University of Sheffield and University of Sussex Ethical Review Committees. ABR measurements (see below) were conducted under anesthesia using ketamine (100 mg/Kg, Fort Dodge Animal Health) and xylazine (10 mg/Kg, Rompun 2%, Bayer), which were administered with intraperitoneal injection as previously described 63 . At the end of the experiments, which lasted up to 40 minutes, mice were usually culled by a schedule 1 method. In a few animals, ABR measurements were followed by the recovery of the mice from anesthesia with intraperitoneal of atipamezole mix 63 .
Animal strains. For the conditional knockout mice, the targeted tm1a allele for both Myo7a (Myo7a tm1a(EUCOMM)Wtsi allele ID: 4431921) and Ush1C (Ush1c tm1a (KOMP)Wtsi allele ID: 4363497) were generated by the Mouse Genetics Programme at the Wellcome Trust Sanger Institute (Cambridge, UK). For Myo7a the critical exons 10 and 11 were floxed, while for Ush1c exons 5-8 were floxed. The tm1c allele was obtained by crossing the tm1a mouse to a FLPeR carrying mouse (Rosa26 Fki ). The tm1d alleles, which were used for the experiments, were obtained by crossing the tm1c mouse (Myo7a fl/fl and Ush1c fl/fl ) with the Myo15-cre mice and genotyped as previously described 24 . The genotyping protocols for mutant mice were also previously described: Myo7a (Myo7a sh6J/sh6J ) 31 and Pcdh15 (Pcdh15 av3J ) 15 . For the conditional knockout mice (tm1c alleles) genomic DNA was amplified under standard reaction conditions using Promega Taq DNA polymerase (Invitrogen, UK). The reaction mix was incubated at 94°C for 5 min followed by 40 cycles of 94°C for 30 s, 58°C for 45 s and 72°C for 90 s, followed by 72°C for 5 min. Primer pairs and product size are as follows: Myo7a: forward, 5′-GGGAGAGAAAGCGGGTGTG-3′ and reverse, 5′-AAGCTGGACTCTCTGGT GGC-3′, producing a 360-bp product for the WT and a 460-bp product for the tm1c homozygous mutant; Ush1C : forward, 5′-ACAGAGCCGTGGGTTCATTC-3′ and reverse, 5′-GTAATGGAGCTGAGGCAGGG-3′, producing a 326-bp product for the WT and a 430-bp product for the tm1c homozygous mutant.
Tissue preparation. Inner hair cells (IHCs) were studied in acutely dissected organs of Corti from postnatal day 7 (P7) to P218 mice and P21-P28 gerbils, where the day of birth is P0. In mice, experiments were performed on IHCs positioned in the apical coil of the cochlea (around 12 kHz) 64 . Gerbil IHCs were positioned at a frequency range of 250-420 Hz 65 . Animals of either sex were killed by cervical dislocation and the organ of Corti dissected in extracellular solution composed of (in mM): 135 NaCl, 5.8 KCl, 1.3 CaCl 2 , 0.9 MgCl 2 , 0.7 NaH 2 PO 4 , 5.6 D-glucose, 10 Hepes-NaOH. Sodium pyruvate (2 mM), amino acids and vitamins were added from concentrates (Fisher). The pH was adjusted to 7.5 (osmolality~308 mmol kg −1 ). The dissected apical coil of the organ of Corti was transferred to a microscope chamber, immobilized using a nylon mesh fixed to a stainless steel ring and viewed using an upright microscope (Leica DM LFS, Germany or Nikon FN1). Hair cells were observed with Nomarski differential interface contrast optics (×63 water immersion objectives). To expose the basolateral surface of the cells, a small tear was made in the epithelium with a suction pipette (tip diameter 3-4 μm) filled with extracellular solution.
When investigating basolateral membrane properties, the size of I K,f was measured near -25 mV and at either 0.7 ms or 1.0 ms after the start of the voltage step, while I K,n was measured as the difference between the peak and steady state of the deactivating inward current at −124 mV. Steady-state total currents were measured at 160 ms, at a potential of 0 mV (extrapolated from the current-voltage curves). The holding potentials used for these recordings were set at either −84 mV or −64 mV (normally specified in the figure legends).
The presence of inhibitory postsynaptic currents in IHCs, which are caused by the release of ACh-containing vesicles from the efferent terminals, was obtained by superfusing 40 mM extracellular K + in order to depolarize the efferent fibers. Event analysis was performed using the Mini Analysis Program (Synaptosoft Inc., USA). Note that the number of events used to calculate the frequency of inhibitory postsynaptic currents (715 events) was larger than that used for the amplitude (594 events: see Results linked to Fig. 7i-l) since the presence of partially overlapping events prevented the accurate assessment of the baseline.
Real-time changes in membrane capacitance (ΔC m ) were performed at body temperature as previously described 12 . Briefly, a 4 kHz sine wave of 13 mV RMS was applied to IHCs from −81 mV, unless otherwise stated, and was interrupted for the duration of the voltage step. The capacitance signal from the Optopatch was filtered at 250 Hz and sampled at 5 kHz. ΔC m was measured by averaging the C m trace over a 200 ms period following the voltage step and subtracting the pre-pulse baseline. Data were acquired using pClamp software and a Digidata 1440 A (Molecular Devices). ΔC m experiments were performed in the presence of 30 mM TEA and 15 mM 4-AP (Fluka) to block the BK current (I K,f ) 9 and delayed rectifier K + currents (I K , previously called I K,neo in the case of pre-hearing IHCs and I K,s in mature IHCs), and linopirdine (80 μM: Tocris) to block I K,n 14 .
For mechanoelectrical transducer (MET) current recordings, a fluid jet driven by a piezoelectric disc was used to elicit bundle displacement as previously described 4 . The tip of the fluid jet pipette had a diameter of 8-10 μm and was positioned at about 8 μm from the bundles. Saturating mechanical stimuli were applied using 50 Hz sinusoids (filtered at 1 kHz, 8-pole Bessel) with a driver voltage (V Piezo ) to the fluid jet of ± 20 V, with positive deflections moving the stereocilia in the excitatory direction. For some experiments a gravity fed local perfusion system was used to extracellularly apply the estimated endolymph-like solution (300 μM Ca 2+ ) 5 either alone or with 200 μM dihydrostreptomycin (DHS). The resting MET current becomes evident when the MET channels shut off in the inhibitory phase of the stimulus: the holding current minus the current present during inhibitory bundle deflection 4 .
Unless otherwise stated, electrophysiological data was analyzed with PCLAMP software (Molecular Devices) and ORIGIN (Origin Lab, USA).
Auditory brainstem responses. Auditory brainstem responses (ABRs) were recorded from either male or female knockout mice and their littermate controls between P15 and P62. Recordings were performed in a soundproof chamber (MAC-3 Acoustic Chamber, IAC Acoustic, UK) as previously described 63 . Briefly, stimuli were delivered to the ear by calibrated loudspeakers (MF1-S Multi Field Speaker, Tucker-Davis Technologies, USA) placed 10 cm from the animal's pinna. Sound pressure was calibrated with a low-noise microphone probe system (ER10B + , Etymotic, USA). Experiments were performed using BioSigRZ software driving an RZ6 auditory processor (Tucker-Davis Technologies). Response thresholds were estimated from the resulting ABR waveform and defined as the lowest sound level where any recognizable feature of the waveform was visible. Responses were measured for click and pure tone stimuli of frequencies between 3.0 and 24 kHz. Stimulus sound pressure levels were typically 0-95 dB SPL, presented in steps of 5 dB. The brainstem response signal was averaged over 256 repetitions. Tone bursts were 5 ms in duration with a 1 ms on/off ramp time, which was presented at a rate of 42.6/sec. Two-photon confocal Ca 2+ imaging. For calcium dye loading, acutely dissected cochleae were incubated for 40 min at 37°C in DMEM/F12, supplemented with fluo-4 AM (final concentration 10 μM; Thermo Fisher Scientific) as recently described 66 . Ca 2+ signals were recorded using a two-photon laser-scanning microscope (Bergamo II System B232, Thorlabs Inc., USA) based on a modelocked laser system operating at 800 nm, 80-MHz pulse repetition rate, < 100-fs pulse width (Mai Tai HP DeepSee, Spectra-Physics, USA). Images were captured with a 60x objective (LUMFLN60XW, Olympus, Japan) using a GaAsp PMT (Hamamatsu) coupled with a 525/40 bandbass filter (FF02-525/40-25, Semrock). Images were analyzed off-line using custom built software routines written in Python (Python 2.7, Python Software Foundation) and ImageJ (NIH). Ca 2+ signals were measured as relative changes of fluorescence emission intensity (ΔF/F 0 ).
Cochlear culture preparation. Cochlear cultures from P18-P24 gerbils were prepared as described previously 1 . Briefly, cochleae were dissected in HEPES buffered (10 mM, pH 7.2) Hanks' balanced salt solution (HBHBSS), placed onto collagen-coated glass coverslips, fed with 100-150 μl of medium containing 98% or 93% standard Dulbecco's modified Eagle's DMEM/F12 with additional 10 mM HEPES buffer, 2% or 7% foetal bovine serum (Biosera, UK) and 10 μg/ml ampicillin (Sigma) and maintained at 37°C for up to 11 days. The standard DMEM/ F12 solution containing 4.18 mM KCl and 1.09 mM CaCl 2 (Sigma, UK) was used, additionally supplemented with KCl or CaCl 2 depending on the experiments (see Results and Fig. 6). For the experiments designed to test the effect of the adult endolymphatic-like solution (20-40 μM Ca 2+ ) 6,67 the above medium was buffered with 2.2 mM HEDTA. For the recordings aimed at testing the effect of high extracellular K + , the above medium was supplemented with 1 M KCl to reach the final desired concentrations (15 mM or 40 mM KCl). In order to block the MET channels in the presence of the endolymphatic-like solution (40 μM Ca 2+ : see above), we used 50 μM of the MET channel blocker d-tubocurarine (Tocris, UK).
Scanning electron microscopy (SEM). The auditory bullae were isolated from the heads of the mice and opened to expose the cochleae which were fixed, by intralabyrinthine perfusion using a fine hypodermic needle through the round window. The fixative was 2.5% glutaraldehyde in 0.1 M cacodylate buffer containing 3 mM calcium chloride (pH 7.3). Following perfusion the intact cochleae were immersed in the fixative for 2 h and then decalcified by immersion in 4.0% EDTA (pH 7.3) for 2 days at 4°C. Subsequently the cochleae were dissected under cacodylate buffer to isolate approximately half turn segments of the organ of Corti from apex to base. These organ of Corti segments were post-fixed (2 h) in cacodylate buffered 1% OsO 4 , before processing through the thiocarbohydrazide-Os-repeated procedure 69 , followed by dehydration in an ethanol series and critical point drying. The organ of Corti samples were mounted on SEM support stubs with silver paint and sputter coated with a thin (ca 5 nm) layer of platinum before examination in a JEOL 6700 cold field-emission SEM operating at 3 or 5 kV. Images are a sample from 3 or more mice for each genotype.
Transmission electron microscopy (TEM). For TEM cochleae were fixed as for SEM but postfixed by immersion for 1 h in 1% osmium tetroxide in 0.1 M cacodylate buffer, dehydrated and embedded in Spurr resin 70 . For ultrastructural evaluation of the nerve terminal region of IHCs, ultrathin (80-90 nm) sections were cut in radial planes from the apical coil, mounted on 200 mesh thin bar copper grids (Agar Scientific, Stansted, UK) and stained in 2% uranyl acetate in 70% ethanol for 20 min, followed by 0.4% lead citrate dissolved at high pH in distilled water for 5 min. They were examined in a JEOL 1230 TEM operated at 100 kV accelerating voltage. Digital images were acquired in using a Megaview III (SIS systems, Olympus Microscopes Ltd). The height and width of stereocilia were measured as previously described 71 . Semi-thin sections (250-270 nm) of the organ of Corti were obtained in a radial plane and examined unstained in a JEOL 1230 TEM operated at 100 kV. The height and width was measured using imageJ (NIH). The heights of first row IHC stereocilia (s1) were measured only when 90-100% of their length from the distal tip lay within the section (confirmed by examination of the adjacent serial section). The full length of second row (s2) stereocilia was readily captured in these sections and was measured to the maximum distal tip including the bevelled portion where present. The width was measured at half maximum height of each stereocilium. For TEM, ≥ 3 mice were processed for each set of experiments and for each genotypes. Images were taken from the same region (around 12 kHz) used for the electrophysiological recordings.
Statistical analysis. Statistical comparisons of means were made by Student's twotailed t test or for multiple comparisons, analysis of variance (one-way and twoway ANOVA followed by Bonferroni's test) were applied. P < 0.05 was selected as the criterion for statistical significance. Mean values are quoted in text and figures as means ± SEM. apart those referring to ABR and TEM measurements, which are reported as means ± SD. Only mean values with a similar variance between groups were compared. Electrophysiological recordings in which the IHC leak conductance, normally calculated between −84 and −94 mV, was significantly larger than that previous reported for both immature and adult cells 14 or with a residual series resistance (R s ) after compensation > 5 MOhm were excluded from the analysis. Animals of either sex were randomly assigned to the different experimental groups. No statistical methods were used to define sample size, which was defined based on previous published similar work from our laboratory. The majority of the experiments were performed blindly to animal genotyping.

Data availability
The data supporting the findings of this study are available within the paper and its Supplementary Information files and from the corresponding author upon reasonable request.