Animal handling and experiments complied with national animal care guidelines and were approved by the board for animal welfare of the University Medical Center Göttingen and the animal welfare office of the state of Lower Saxony, Germany.

Otof^–/– mice (Reisinger et al., 2011) were backcrossed onto C57BL/6N or CD1 background for at least 5 generations. Littermate Otof^+/+ mice were used as controls. For hair cell counts, we used C57BL/6N-Otof^–/– mice, Otof^–/– mice derived from homozygous matings of CD1 Otof^–/– females and C57BL/6N Otof^–/– males, and other Otof^+/+ mice of C57BL/6N or C57BL6/J background, as indicated by color labeling in Figures 1C,D. Animals of both genders were used in the experiments. Mice were housed in social groups in individually ventilated cages in a specific pathogen-free facility with free access to water and food and a 12/12 h light/dark-cycle.

Distortion product otoacoustic emissions (DPOAE) were recorded from mice anesthetized with ketamine (125 mg/kg) and xylazine (2.5 mg/kg) i.p. in a soundproof box as described previously (Jing et al., 2013). Briefly, a custom designed probe was inserted in the ear canal, containing a sensitive microphone (MKE-2, Sennheiser, Hannover, Germany) and silicone tubes connected to two speakers (MF1-S, Tucker Davis Technologies, Ft Lauderdale, FL, United States). Primary stimuli f1 and f2 were presented over 16 s at a frequency ratio f2 = 1.2xf1 and intensity ratio f1 = f2 + 10 with f1 frequencies of 6–23 kHz, controlled by a custom Matlab routine and TDT system III hardware (Tucker Davis Technologies, Alachua, FL, United States). Recorded emissions were fast-Fourier transformed and amplitudes of DPOAE at 2xf1-f2 measured and averaged over both ears.

Cochlear whole mounts of mice at the age of postnatal day 4 (P4) to 48 weeks were immunostained as previously described (Strenzke et al., 2016). The cochleae were perfused and fixed with 4% formaldehyde for 10–30 min or 45 min (Figure 3B) at 4°C. Cochleae of mice older than 4 weeks were decalcified in Morse’s solution (10% sodium citrate, 22.5% formic acid) for 15 min or in 0.5 M EDTA overnight before dissecting the whole organ of Corti including basal turns. For cryosections of the cochlea, samples were perfused and fixed in 4% formaldehyde for 30 min and decalcified in Morse’s solution for 2–24 h. Cochleae were kept in 20% sucrose until embedding and cutting into 20 μm sections (Cryotome 2800 Frigocut, Reichert-Jung, Austria).

The following antibodies were used: mouse anti-Ctbp2 (#612044, BD Bioscience, 1:50–1:200) for labelling synaptic ribbons, rabbit anti-Na/K-ATPase (SC28800, Santa-Cruz Biotech, 1:200) or rabbit anti-Shank1a (#RA19016, Neuromics, 1:300) for labelling postsynaptic boutons (Huang et al., 2012), guinea pig anti-parvalbumin (#195004, Synaptic Systems, 1:200) to label IHC and OHC as well as SGN cell bodies and afferent fibers, guinea pig anti-Vglut3 (#135204, Synaptic Systems, 1:300) to label IHCs, goat anti-calretinin (CG1, Swant, 1:300) to label IHC and OHC, mouse anti-neurofilament (N5389, Sigma, 1:400) to label afferent fibers and SGN cell bodies, mouse anti-MyoVI (#H00004646-M02, Novus Biological, 1:200) or rabbit anti-MyoVI (#25-6791, Proteus, 1:200) antibodies for labelling hair cells, and rabbit anti-oncomodulin (OMG4, Swant, 1:300–1:1,000) for labelling OHC.

The following secondary antibodies were used: DyLight 405 donkey anti-guinea pig (Jackson Immuno, 1:200), Alexa Fluor 568 IgG H + L goat anti-mouse, Alexa Fluor 633 IgG goat anti-rabbit (MoBiTec, 1:200), Alexa Fluor 488 goat anti-guinea pig, Alexa Fluor 488 goat anti-mouse IgG, IgA IgM (H + L), Alexa Fluor 647 donkey anti-rabbit, Alexa Fluor 568 donkey anti-goat, Alexa Fluor 488 donkey anti-mouse (Invitrogen, 1:200). Hoechst 34580 (1:1,000, H21486, Fisher scientific) or To-Pro-3 Iodide (642/661, 1:800, T3605, ThermoFisher) were used as nuclear stains.

Confocal images were acquired using a laser scanning confocal microscope (Leica TCS SP5, Leica Microsystems GmbH, or QuadScan Super-Resolution Microsocope, Abberior Instruments, Göttingen, Germany) with a 10× air objective (low magnification), 20× or 40× oil immersion objective (medium magnification) or 63× oil (Figures 2 and 3A: 63× glycerol) immersion objectives for high magnification images. Z-stacks were acquired at 9, 15, 28, and 50 kHz frequency regions with high magnification for synaptic ribbon quantification and medium magnification for quantification of IHC, OHC, and SGN.

Images were processed and analyzed using ImageJ (NIH,¹). In samples from adult mice, synaptic ribbons (immunolabelled for Ctbp2) and postsynapses (immunolabelled for Na/K-ATPase) were manually counted in 5–15 μm z-stacks (0.2 μm step size, 3.3× digital zoom) using ImageJ. In P4-P14 organs of Corti, synapses were counted as Ctbp2 spots adjacent to Shank1a immunolabels using the “Spots” tool in Imaris 7.6.5 (Bitplane Scientific Software). For quantification of synapse size at P14, we segmented the confocal stacks for Ctbp2- and Shank1a-positive spots, respectively, using the surface function in Imaris. The expected spot sizes were set to 0.55 μm for Ctbp2 and 0.8 μm for Shank2. Background subtraction and the rejection thresholds in terms of intensity, size, sphericity and quality of spots were optimized for each image to detect as many true synapses as possible in which Ctbp2 and Shank1a were juxtaposed with a maximal distance of 600 nm. We clearly acknowledge that the results of the volume calculations can easily be influenced by the subjectivity of these settings. We focused on correctly segmenting the Shank1a channel and are thus less sure about an observed smaller increase in Ctbp2 spot size by 20% (Otof^+/+: 0.297 ± 0.006 μm³, n = 1,287; Otof^–/– 0.342 ± 0.006 μm³, n = 1,135; p < 0.0001).

Inner hair cell and OHC cell bodies were identified by nuclear and/or cell body staining (MyoVI/Parvalbumin and Hoechst) and manually counted along a defined length of basilar membrane measured along the row of IHC. Areas with dissection-induced damage to the preparation were carefully excluded from analysis. Hair cells with preserved cytoplasmic staining were counted as present even if they showed an irregular shape or position. For SGN quantification, we measured the area of the canalis spiralis cochleae and manually counted the density of SGN cell bodies (Hoechst staining co-immunolabeled with Parvalbumin).

All experiments were performed with at least three independent replicates. Data are plotted as mean ± standard error of the mean (s.e.m.) and graphics were designed using Igor Pro 6 (WaveMetrics). Statistical analysis was performed by means of GraphPad Prism 7.03 (GraphPad Software) or Excel (Microsoft Office). The Brown-Forsythe test or the F-test were used to test for equal variance. Statistical tests and number of replicates are indicated in the figures or figure legends.

In this study, we assessed the integrity of cochlear structures from birth to mid-aged Otof^–/– mice and wild-type C57BL/6 controls. We first quantified the number of inner and outer hair cells and of afferent synapses in four frequency regions, termed apical (best encoding frequencies around 9 kHz), middle (∼15 kHz), midbasal (∼28 kHz) and basal turns (∼50 kHz, Figures 1A,B; Müller et al., 2005). At the age of 8 weeks, Otof^–/– mice display a near-normal complement of a single row of IHCs and three rows of OHCs over the entire cochlea (Figures 1C,D). From 24 weeks of age on, a decay in numbers of IHCs and OHCs becomes apparent in Otof^–/– cochleae. While OHC numbers remain normal in the apical aspects of the cochlea in both genotypes, there is age-dependent loss of OHCs progressing from the base toward the middle of the cochlea, which is accelerated in Otof^–/–mice compared to Otof^+/+ controls. For example, in the midbasal and basal regions of Otof^–/–mice at age 24 weeks, there was a reduction to 84 and 28% of Otof^+/+ OHC numbers, respectively (Figure 1C). However, in the C57BL/6 mouse background strain, age-progressive hearing loss has previously been described and was attributed to the Cdh23^ahl allele (Kane et al., 2012; Mianné et al., 2016), explaining the almost complete loss of all OHCs in the 50 kHz regions by 48 weeks of age in both genotypes. As a result, the overall OHC numbers per age group were not significantly different between genotypes, despite the midbasal and basal OHC degeneration appeared more pronounced in Otof^–/–mice compared to Otof^+/+ controls.

In Otof^+/+ mice, the row of IHCs was generally completely preserved, except for limited age-dependent degeneration of IHC in the basal turn. In contrast, scattered loss of IHCs was noted in Otof^–/– mice over all frequency regions (Figure 1D) from 24 weeks of age on. Like for OHC, the loss of IHC in mutants was more pronounced in the midbasal and basal regions, where at 24 weeks of age IHC numbers were reduced to 56 and 64% of Otof^+/+ IHC counts, respectively, compared to 80 and 76% in the apical and middle aspects.

In summary, there is age-dependent loss of OHCs in control C57BL/6 Otof^+/+ animals which is mostly restricted to the basal aspects of the cochlea and accompanied by only minimal loss of IHCs. In Otof^–/–mice, IHC and OHC are initially present, but show a more pronounced age-dependent degeneration. While the loss of OHCs predominantly affects the basal and midbasal regions, IHCs degenerate also in the middle and apical turns to some extent, starting already at 6 months of age.

Most patients with pathogenic OTOF mutations lose DPOAE within their first two decades of life. Here, we first tested if the loss of DPOAE is reproduced in the mouse models for DFNB9 (Figures 1E–H). At the age of 8 and 12 weeks, we found DPOAE to be of comparable amplitude in Otof^+/+ and Otof^–/– mice. At the age of 24 weeks, DPOAEs were significantly reduced in Otof^–/– mice for stimulus frequencies (f2) of 12, 16 and 24 kHz compared to Otof^+/+ mice, indicating a role for otoferlin in preserving OHCs and active cochlear amplification. Thereafter, DPOAE amplitudes declined further between 24 and 48 weeks of age for both genotypes. In Otof^+/+ mice a mild reduction of DPOAE was observed for stimulus frequencies of 16 and 24 kHz, which can likely be attributed to the Cdh23^ahl allele in the C57BL/6 mouse background strain which is known to cause age-progressive high-frequency hearing loss (Kane et al., 2012; Mianné et al., 2016). A more severe DPOAE decline appeared in Otof^–/– mice between 12 and 24 kHz. Thus, the functional decline of cochlear amplification assessed as DPOAE was accelerated in absence of otoferlin and preceded the loss of OHCs. For both DPOAE loss and OHC degeneration, the high frequency regions were affected earlier and more severely than the low frequency regions of the cochlea.

A recent study indicated developmental deficits in Otof^–/– afferent synapses (Al-Moyed et al., 2019). This led us to count ribbon type synapses in frequent intervals during the maturation of IHCs shortly before the onset of hearing, which in rodents is at 2 weeks of age. In this period, the Shank1a-labeled postsynaptic boutons of wildtype IHC gradually become smaller and restricted to a small structure close to the ribbon, and extensive synaptic pruning occurs (Figures 2, 3A). Remarkably, postsynaptic boutons at Otof^–/– IHCs appeared morphologically different from age-matched wild-type boutons: The same immunostaining reveals that postsynaptic Shank1a-labeling remains in large patches throughout early postnatal development (Figure 2). At P14, Shank1a spot volume was increased by 50% (Otof^+/+: 0.532 ± 0.012 μm³; n = 1,306 spots from 4 animals; Otof^–/– 0.802 ± 0.019 μm³, n = 1,068/N = 4; p < 0.0001; mean ± s.e.m, Wilcoxon rank sum test).

In two-day intervals, we determined the number of synapses in IHCs of the apical region, defined as Ctpb2 immunoreactive spots juxtaposed to Shank1a labeled postsynapses. In Otof^–/– IHCs of the apical region, we found an initially higher number of synapses between postnatal day 4 (P4) and P12, but a faster decay of ribbon synapse numbers in the absence of otoferlin compared to wild-type IHCs (Figures 2, 3A). Previous studies from our and other labs indicate that in the third and fourth postnatal weeks, ribbon synapse numbers in Otof^–/– IHCs decrease to 60% of the numbers in wild-type IHCs (Roux et al., 2006; Al-Moyed et al., 2019).

Together, our data reveal that the presence of otoferlin is required for synaptic maturation before the onset of hearing.

We next assessed the long-term effect of otoferlin-deficiency on synapse numbers in the four regions of the cochlea described in Figure 1B. Along the length of the cochlea, at 8 weeks of age, we found ∼7 synapses per Otof^–/– IHC. Between 12 and 48 weeks of age, we counted on average ∼6 synapses per Otof^–/– IHC, with no obvious decay in synapse numbers over time (Figure 3B). While synapse numbers in Otof^+/+ IHCs for different frequency regions declined rather constantly over time (Figure 3C), no gross difference was apparent for the different turns of the cochlea regarding synapse numbers in Otof^–/– mice (Figure 3D).

We next counted the number of spiral ganglion neurons (SGNs) at the age of 8 and 48 weeks (Figure 4). At 8 weeks of age, despite afferent boutons from these SGNs already being reduced to about half of wild-type numbers, the numbers of SGNs were comparable for Otof^–/– and Otof^+/+ mice. Since studies of age-dependent and noise-induced synaptopathy revealed that a loss of SGN bodies occurs with a delay of several months or even years (Kujawa and Liberman, 2009, 2015; Sergeyenko et al., 2013), we counted SGN cell bodies again at the age of 48 weeks. Here, we found a slight but significant loss of SGN cell bodies at the age of 48 weeks in Otof^–/– mice to an overall of 79% of Otof^+/+ SGN counts (p = 0.038, 2way ANOVA genotype difference across tonotopic regions). In terms of cochlear implantation and gene therapy, the long-term presence of SGNs is promising, since the remaining SGNs can be electrically stimulated and a re-growth of dendrites could potentially be initiated by endogenous or supplemented nerve growth factors.

Synaptic transmission is independent of otoferlin in immature IHCs up to postnatal day 3 (Beurg et al., 2010). From P4-5 on, hardly any Ca²⁺-triggered synaptic transmission can be recorded in Otof^–/– IHCs (Beurg et al., 2010). Up to postnatal day 6, spontaneous action potentials can be recorded from IHCs, giving rise to synaptic activation (Kros et al., 1998; Marcotti et al.,2003a,b). Thus, during early postnatal development, transsynaptic activity is supposed to be wild-type like in absence of otoferlin, but will be greatly diminished from P5 on in Otof^–/– mice. We found that synaptic maturation differs in absence and presence otoferlin from P6 on. Synaptic pruning in otoferlin-deficient IHCs started between P6 and P8, with a delay of 2 days compared to wild-type controls (Figure 3A). In Otof^+/+ IHCs, synapse numbers remained mostly constant from P12 on, while they declined further in Otof^–/– IHCs. From the third and fourth postnatal weeks, ∼40–50% fewer synaptic contacts are present in Otof^–/– IHCs compared to littermate controls (Roux et al., 2006; Al-Moyed et al., 2019). A similar loss of approximately 40% of the ribbon synapses and delayed loss of SGNs has been described in Vglut3^–/– mice, in which IHC synaptic vesicles are not loaded with glutamate (Seal et al., 2008). Later, mRNA sequencing suggested that the low spontaneous rate/high threshold SGNs were missing in this mouse line (Shrestha et al., 2018). If glutamatergic signaling was the key factor for retaining low spontaneous rate/high threshold synaptic contacts, then presumably also in Otof^–/– cochleae this type of synapses would degenerate during development. Remarkably, 6 out of 7 synapses remaining at 8 weeks of age in each Otof^–/– IHC persist for at least 40 weeks. Possibly, the rare spontaneous synaptic events present in Otof^–/– mice, recorded as excitatory postsynaptic potentials in postsynaptic boutons (Pangrsic et al., 2010; Takago et al., 2019) support their maintenance, either by glutamatergic or by neurotrophic signaling. In contrast, a 90% reduction in voltage gated Ca²⁺ current in Ca_v1.3 knockout mice not only reduced the number of synapses to ∼60% within the first 4 weeks of life, but also to a further decay down to ∼2 synapses per IHC at the age of 29 weeks (Nemzou et al., 2006).

Given the early maturational stage where an effect of otoferlin-deficiency on synapse development becomes apparent, a gene replacement would need to be applied even prenatally to ensure proper synaptic development. After successful gene substitution in rodents at P6-P7, the number of ∼7 synaptic contacts was unchanged, which was sufficient to elicit auditory responses after gene substitution (Al-Moyed et al., 2019). As a result, although in humans OTOF-gene therapy will take place at a time point when synapse numbers are presumably already reduced, a rescue of hearing function is likely to be successful.

In addition to the alteration in synapse numbers, postsynaptic boutons appear to be larger in absence of otoferlin. While Shank1a immunolabeling in P14 wild-type controls was confined to small postsynaptic patches, Shank1a immunofluorescence appeared expanded in Otof^–/– cochleae. Similarly, Kim et al. (2019) demonstrated enlarged postsynapses labeled for GluA3 in Vglut3^–/– mice, in which inner hair cells exocytose vesicles lacking glutamate. We presume that neurotransmitter release from presynaptic ribbon synapses regulates the size of the postsynaptic density.

Despite a reduction in synapse numbers in the third and fourth postnatal week, the number of SGNs was wild-type like in 8 week old Otof^–/– cochleae. Even at 48 weeks of age, only about half of the animals displayed a loss in SGN density (Figure 4D). Despite mRNA of otoferlin could be detected in (immature) SGNs (Yasunaga et al., 1999) and in brain (Yasunaga et al., 2000; Schug et al., 2006), our immunostainings of whole mount organs of Corti do not indicate that SGNs express significant amounts of otoferlin. Moreover, since cochlear implants that bridge only the IHC-SGN synapse have good long-term outcomes in DFNB9 patients, SGNs and neurons of the central auditory pathways seem not to functionally depend on otoferlin. We hypothesize that SGN cell bodies in Otof^–/– cochleae persist despite the loss of the IHC ribbon synapse for several months, similar to the long delay between a noise trauma leading to a 40% reduction in synapse numbers and the death of SGN cell bodies (Kujawa and Liberman, 2009, 2015; Sergeyenko et al., 2013). Due to the shorter life span, mice age much faster than humans do, such that we presume that SGNs might persist in DFNB9 patients for years even without therapeutic intervention. A re-growth of the dendrites, induced by application of nerve growth factors, is experimentally addressed in mouse models to initiate recovery of synaptic contacts (Chen et al., 2018; Hashimoto et al., 2019). More studies will be required to test if a re-establishment of afferent synapses could be achieved in Otof^–/– IHCs by a similar approach in the future.

Previous studies, some of which with longitudinal studies, found OAEs to degenerate in DFNB9 patients (Rodríguez-Ballesteros et al., 2003, 2008; Kitao et al., 2019). This is the first study addressing long-term effects in mouse models for DFNB9. Similar as in human patients, DPOAEs were wild-type like in the first weeks of life. In Otof^–/– mice, first signs of OAE decay were found at 24 weeks of age. At 48 weeks of age, a major deterioration of DPOAEs was observed for the 12 and 16 kHz region. In the basal turn, OAEs were strongly decreased also in Otof^+/+ controls, most likely due to the Cdh23^Ahl allele leading to progressive hair cell degeneration (Kane et al., 2012; Mianné et al., 2016). While a more detailed study using mouse strains with a corrected Cdh23^Ahl would be useful to analyze OHC degeneration solely due to otoferlin deficiency, we demonstrate here that the loss of OHCs in Otof^–/– clearly precedes the one in Otof^+/+ mice of the same background. At 24 weeks of age, wild-type Otof^+/+ littermates and non-littermate controls of C57BL/6 background displayed consistently more OHCs around 28 and 50 kHz than Otof^–/– mice (Figure 1C). In addition, at 48 weeks of age, no loss of OHCs was apparent in the 28 kHz region of Otof^+/+ mice, but ∼70% of OHCs were lost in Otof^–/– mice.

Importantly, our study is the first to elucidate the loss of IHCs in Otof^–/– mice. This loss was found along the length of the cochlea, with a shallow basal to apical gradient. In contrast to Otof^–/– cochleae, IHCs were mostly preserved in Otof^+/+ controls even at 48 weeks of age, despite the presence of the Cdh23^Ahl allele. To date, it is unclear why Otof^–/– IHCs and OHCs degenerate, and whether the loss of hair cells can be prevented by a gene replacement approach. Long-term observational studies with morphological analysis of treated and untreated Otof^–/– organs of Corti will be required to predict a long-term outcome of gene therapy for DFNB9 patients.