Overloaded Adeno-Associated Virus as a Novel Gene Therapeutic Tool for Otoferlin-Related Deafness

Hearing impairment is the most common sensory disorder in humans. So far, rehabilitation of profoundly deaf subjects relies on direct stimulation of the auditory nerve through cochlear implants. However, in some forms of genetic hearing impairment, the organ of Corti is structurally intact and therapeutic replacement of the mutated gene could potentially restore near natural hearing. In the case of defects of the otoferlin gene (OTOF), such gene therapy is hindered by the size of the coding sequence (~6 kb) exceeding the cargo capacity (<5 kb) of the preferred viral vector, adeno-associated virus (AAV). Recently, a dual-AAV approach was used to partially restore hearing in deaf otoferlin knock-out (Otof-KO) mice. Here, we employed in vitro and in vivo approaches to assess the gene-therapeutic potential of naturally-occurring and newly-developed synthetic AAVs overloaded with the full-length Otof coding sequence. Upon early postnatal injection into the cochlea of Otof-KO mice, overloaded AAVs drove specific expression of otoferlin in ~30% of all IHCs, as demonstrated by immunofluorescence labeling and polymerase chain reaction. Recordings of auditory brainstem responses and a behavioral assay demonstrated partial restoration of hearing. Together, our results suggest that viral gene therapy of DFNB9—using a single overloaded AAV vector—is indeed feasible, reducing the complexity of gene transfer compared to dual-AAV approaches.


INTRODUCTION
According to the World Health Organization, ∼466 million people world-wide suffer from disabling hearing loss (WHO, 2019). In the majority of these cases, hearing impairment is caused by exposure to damaging noise levels as well as aging, likely with a contribution of a genetic predisposition. Defects of individual so-called "deafness genes" explain at least 50% of hereditary hearing impairment, which affects 1-2 in 1,000 newborns. Currently, more than 150 non-syndromic gene loci have been identified in humans (Willems, 2000;Bowl et al., 2017;Michalski and Petit, 2019). At present, no curative therapeutic approaches are available for these patients. Hence, hearing aids or cochlear implants represent the standard of rehabilitation, but fail to fully recover normal hearing function. While novel gene therapeutic approaches for this and other conditions are currently in development, these treatments are not yet clinically available. However, several recent preclinical studies demonstrate that AAV-mediated gene replacement can restore hearing in mouse models for various deafness genes (e.g., Akil et al., 2012, Askew et al., 2015Jung et al., 2015;György et al., 2017;Pan et al., 2017;Dulon et al., 2018;Akil et al., 2019;Al-Moyed et al., 2019;reviewed in Kleinlogel et al., 2020).
Among the targets for future gene therapy is the OTOF gene, which encodes the multi-C2-domain protein otoferlin. OTOF mutations lead to autosomal recessive non-syndromic prelingual deafness DFNB9 (Yasunaga et al., 1999) or hearing impairment that can be temperature sensitive (Varga et al., 2006). The prevalence of OTOF-related deafness accounts for up to 5-8% of autosomal recessive non-syndromic hearing loss cases in some Western populations (Rodríguez-Ballesteros et al., 2008) and hence, resides within the top five of genetic hearing disorders that require therapeutic intervention (Angeli et al., 2012). Otoferlin is abundantly expressed in sensory inner hair cells (IHCs) of the cochlea and serves a key role in the final steps of synaptic vesicle fusion at IHC ribbon synapses with afferent spiral ganglion neurons (Roux et al., 2006). Here, several functions of otoferlin have been proposed [reviewed in Moser and Starr (2016), Moser et al. (2019)], such as a role in (i) presynaptic Ca 2+ -sensing to trigger vesicular exocytosis upon IHC depolarization (Roux et al., 2006;Johnson and Chapman, 2010;Michalski et al., 2017) and (ii) efficient vesicular priming and replenishment of synaptic vesicles to the release site to ensure indefatigable and temporally-precise neurotransmitter release during periods of prolonged stimulation (Pangrsic et al., 2010;Jung et al., 2015;Vogl et al., 2015, Strenzke et al., 2016Vogl et al., 2016). Mouse mutants serve as a valuable model system to investigate synaptic disease mechanisms (Moser and Starr, 2016) and together with clinical data [reviewed in Santarelli et al. (2015), Moser and Starr (2016)] indicate that preservation of cochlear morphology and function make otoferlin-related auditory synaptopathy amenable to clinical gene therapy.
Medical need and permissive cochlear status have generated great interest in gene therapy approaches to otoferlin-related auditory synaptopathy. However, due to the large OTOF coding sequence of ∼6 kb and the restricted packaging size of standard adeno-associated viral vectors (AAV; <4.7 kb), (Grieger and Samulski, 2005) that are commonly employed for gene therapeutic applications based on their favorable biosafety profile, the delivery of full-length OTOF remains challenging. Recently, two independent studies have overcome this problem by using a dual-AAV approach and could show partial rescue of hearing in Otof-KO mice (Akil et al., 2019;Al-Moyed et al., 2019). Yet, the complexity of this approach makes clinical translation more challenging than in cases where a single AAV serves as the molecular therapeutic [reviewed in Kleinlogel et al. (2020)].
Here, we aimed to establish a simplified strategy for delivering the coding sequence for full-length Otof (fl-Otof -CDS) to IHCs of Otof-KO mice using a single AAV "overloading" strategy (Allocca et al., 2008;Grose et al., 2012;Hirsch et al., 2013;Pryadkina et al., 2015). Conceptually, this approach is based on attempting to package a CDS, which theoretically exceeds the predicted maximum insert capacity of the employed AAV-in this particular case, the fl-Otof -CDS (∼6 kb) into a standard AAV (maximum capacity: 4.7 kb)-thereby technically "overloading" the virus. In fact, to establish the proof-of-principle for this approach for future cochlear gene therapy, we packaged the fl-Otof -CDS into not only one, but rather several naturallyoccurring, as well as more recently developed and highly potent synthetic AAV serotypes. We then tested the potential of these overloaded AAVs for expressing functional otoferlin in IHCs of the mouse organ of Corti in organotypic cultures as well as in living mice upon early postnatal AAV injection. We show expression of fl-Otof in IHCs of Otof-KO mice that went along with partial hearing restoration, thereby demonstrating that single AAV overloading provides a feasible gene-therapeutic strategy for otoferlin-related auditory synaptopathy. Moreover, our data indicate that successful gene transfer by overloaded AAVs can be achieved independent of the employed AAV capsid composition.

Molecular Cloning
Full-length mouse otoferlin (fl-Otof ) was subcloned from a previously generated cDNA clone (pcDNA3-mOtof-IRES-EGFP) into a pAAV vector. The cloning procedure was done by in-fusion cloning according to manufacturer instructions (TaKaRa/Clontech, In-Fusion HD Cloning kit). Owing to the size of fl-Otof we amplified three overlapping otoferlin fragments which complement fl-Otof by high fidelity PCR using the HiFi-PCR-premix provided by the kit and fused into the linearized pAAV target vector to generate pAAV_fl-Otof. The corresponding vector map of pAAV_fl-Otof with ubiquitous CMV-β-Actin (CMV/hbA) hybrid promoter is shown in Supplementary Figure 1. Due to the lack of suitable restriction sites within the otoferlin coding region and the substantial length of the fl-Otof cDNA we refrained from using traditional T4 ligase-based cloning techniques. Instead, we applied in fusion cloning strategy to generate a fl-Otof plasmid. To this end we linearized the target vector for virus expression (pAAV) by digestion with the restriction enzymes NheI and HindIII (Fermentas). We amplified three otoferlin fragments complementing fl-Otof [fragment A (primer pair A_F2/R1), fragment B (B_F1/R1), and fragment C (C_F1/R2)], which contained overlapping regions with each other and with the linearized target vector (see Appendix A1). This step required fragment and primer optimization [e.g., by determining optimal fragment sizes and ratios and by using "split overlaps" to reduce primer lengths (primers see Appendix A1.1 and A1.2)]. Thus, the linearized vector and three fragments were simultaneously fused together using the TaKaRa/Clontech In-Fusion HD Cloning kit, following manufacturer's instructions. This approach yielded the final fl-Otof viral vector (pAAV_fl-Otof ) used for subsequent virus production (Supplementary Figure 1B). Target pAAV vector linearized by NheI/HindIII digest and PCR-amplified otoferlin fragments A, B, C using the primer pairs are shown in Supplementary Figure 1A and Appendix A1.2. Note that primer A_F2 contains a 15 bp overlap with the target vector at the 5'-end and encodes the START-codon for expression of otoferlin, while C_R2 encodes the STOP codon and a 15 bp overlap with the AAV vector at the 3'-end. Restriction enzyme digestions and Sanger sequencing (SeqLab, Germany) was employed to verify the correct fl-Otof insert. The length of the otoferlin coding region is 5,979 kb comprising the mouse transcript variant 1 which contains the alternative Exon5B (NCBI accession numbers NP_001093865.1, Appendix A4). This sequence shows 94% identity to the human otoferlin sequence (NP_001274418.1), as determined by using the BLASTP algorithm (NCBI). A protein alignment of mouse and human otoferlin is shown in the Appendix A4 [CLUSTAL O(1.2.4)]. The CDS size (5,979 kb) is beyond the packaging capacity of adeno-associated virus of 4.7 kb (according to (Grieger and Samulski, 2005). A truncated otoferlin, consisting only of the C2E and C2F including the transmembrane domain (TM)-named in this study as "miniOtoferlin" (Supplementary Figures 2, 3A)-was generated in a similar way. Briefly, two overlapping fragments covering the C-terminal part of Otof were amplified from the cDNA clone (pcDNA3-mOtof-IRES-EGFP) by high fidelity PCR using the HiFi-PCR-premix provided (TaKaRa/Clontech, In-Fusion HD Cloning kit) and overlapping primer pairs (G_F1/B_R1 and C_F1/R1 for fragments G and C2, respectively) (see Appendix A1). For linking to EGFP, both fragments were fused simultaneously into pEGFP-C2 linearized by SacI (Thermo Fisher Scientific) using the in-fusion reaction according to the manufacturer's instructions (TaKaRa/Clontech, In-Fusion HD Cloning kit). The resulting EGFP-miniOtof fusion construct was subcloned to the linearized pAAV vector. Insert and vector were restricted using NheI and HindIII (Thermo Fisher Scientific) and complementary ends were ligated using T4 ligase. This approach yielded the final EGFP-miniOtof viral vector (pAAV_EGFP-miniOtof ) used for subsequent virus production. Restriction enzyme digestion and Sanger sequencing (SeqLab, Germany) was employed to verify the correct EGFP-miniOtof insert.

Postnatal AAV-Injection
Postnatal AAV-injection into the scala tympani of the left ear via the round window was performed at p5-7 essentially as described in Akil et al. (2012). In this study, we used PHP.B-derived viral capsids in combination with a (CMV/hbA) promoter to drive transgenic expression of fl-Otof in cochlear IHCs. Both female and male animals were used equally during the study. Briefly, around 1-1.5 µl of PHP.B_CMV/hbA-fl-Otof (1.3 × 10 12 genome copies/ml) or PHP.eB_CMV/hbA-fl-Otof (5.0 × 10 12 genome copies/ml), AAV2/9_CMV/hbA-fl-Otof (2.4 × 10 12 genome copies/ml), Anc80L65_CMV/hbA-fl-Otof (6.0 × 10 12 genome copies/ml), and PHP.eB_CMV/hbA_eGFP-mini-Otof (4.9 × 10 12 genome copies/ml) were injected into the left ear via the round window. 4 weeks after injection, animals were used in auditory brainstem (ABR) recording experiments or Intellicage experiments. All experiments were done in compliance with the national animal care guidelines and were approved by the board for animal welfare of the University Medical Center Goettingen and the animal welfare office of the state of Lower Saxony (AZ33.9-42502-04-14/1391).

Immunohistochemistry and Confocal Microscopy
Acutely dissected or organotypically-cultured cochlear explants were fixed in 4% formaldehyde in PBS on ice for 1 h. Strongly ossified cochleae of adult mice were decalcified in 10% EDTA solution (prepared in PBS) for 10-11 days prior to dissection. After extensive washing, permeabilization (30 min; 0.5% Triton X-100 in PBS) and a blocking step with a goat serum dilution buffer (GSDB; 1 h; 16% normal goat serum, 450 mM NaCl, 0.3% Triton X-100 and 20 mM phosphate buffer at pH 7.4), the following primary antibodies were applied in GSDB overnight at 4 • C: mouse anti-otoferlin (Cat.-Nr. ab53233; Abcam) or rabbit anti-otoferlin (Vogl et  ; Thermo Fisher Scientific) were applied the following day (1 h; room temperature). After mounting the specimen in Mowiol mounting medium, images were acquired in confocal mode on an Abberior Instruments Expert Line STED microscope controlled by Imspector software, with excitation lasers at 485, 561, and 640 nm and either a UPlanSApo 20x 0.85 NA or UPlanSApo 100x 1.4 NA oil immersion objectives. Image stacks were acquired with voxel sizes of 350 × 350 × 5,000 nm (20x) or 80 × 80 × 200 nm (100x).
Transduction efficiency was manually quantified either (i) across the entire organotypically-cultured organ explant, or (ii) along the entire length of the acutely dissected apical to medium cochlear turn (the average length of the analyzed fragments as measured from the cochlear apex: Group 1: 1966.4 ± 113.3 µm, n = 459 IHCs, N = 2; Group 2: 1988.4 ± 53.7 µm, n = 704 IHCs, N = 3;). Ribbon counts were established per IHC in the parvalbumin channel and cells were then assigned either to the transduced (i.e., otoferlin-positive) or nontransduced (i.e., otoferlin-negative) group, based on the observed otoferlin immunoreactivity.
Synapse counts were taken from clearly identifiable positively or adjacent negatively transduced IHCs. Data normality was established using a Shapiro-Wilk test. Normally distributed data sets were then analyzed using a one-way ANOVA followed by a post-hoc Tukey test to assess individual statistical significances.

Chemical Tissue Clearing and Lightsheet Microscopy
The osseous labyrinth-i.e., cochlea and vestibular systemwas dissected from the mouse temporal bone and fixed in 4% formaldehyde (diluted in PBS) for 1 h on ice. Subsequently, samples were decalcified in 10% EDTA for ∼2 weeks at 4 • C. For chemical tissue clearing and immunostaining protocol, we followed an adapted version of the iDISCO+ approach (Renier et al., 2016). Here, all washing and incubation steps were performed at 37 • C with the samples continuously rotating. First, samples were incubated for 3 h in a permeabilization solution (0.5% Triton X-100 in PBS), which was followed by an overnight incubation in blocking solution (0.5% Triton X-100 + 10% v/v goat serum in PBS). Then, primary antibodies against Calretinin (Cat.-Nr. 214 106; Synaptic Systems) and Otoferlin (Cat.-Nr. 178 003; Synaptic Systems) were diluted in blocking solution and samples were incubated in the primary antibody solution for 48 h. After primary incubation, samples were washed in PBS for 24 h. The secondary antibodies goat anti-chicken 488 (Thermo Fisher, #A-11039) and goat anti-rabbit 647 (Thermo Fisher, #A-21244) were diluted in blocking solution and applied for 48 h. Finally, specimen were washed for 48 h and prepared for tissue clearing according to the iDISCO+ protocol (Renier et al., 2016).
Images were acquired with the LaVision UltraMicroscope II using the LaVision BioTech ImSpector Software. The numerical aperture was set to 0.148 which resulted in a sheet thickness of 5 µm. Stacks were acquired with a z step of 3 µm, a total zoom of 8x (2x objective plus 4x optic zoom) and an individual range for each cochlea that was covering all IHCs.
For analysis, images were imported into Imaris software (version 9.5.1, Oxford Instruments) to create a surface that contained all IHCs. This surface was as a mask to generate a new stack containing only the IHC row signal. Within this region of interest, we used a custom Matlab code that employed Greenwood's function (Greenwood, 1961) for tonotopic mapping of the IHC row, using parameters reported for the mouse cochlea in Müller et al. (2005). As an output of this code, we obtained 500 interpolated coordinates equally distributed along the tonotopic axis together with their corresponding frequency labels. By displaying the interpolated coordinates that most closely matched the frequency labels 4, 8, 12, 16, 24, and 32 kHz, we defined our frequency regions along the IHC row. IHCs within the respective frequency bands were finally counted in ImageJ/Fiji (Schneider et al., 2012) using the Cell Counter tool.

RNA Isolation and Reverse Transcription
Total RNA was isolated from 8-to 10-weeks-old mice organs of Corti with the Invitrogen TM TRIzol TM Plus RNA Purification Kit (#12183555, Thermo Fisher Scientific) following the manufacturer's instructions. The SuperScript R IV First-Strand Synthesis System (#18091050, Thermo Fisher Scientific) served to synthesize cDNA according to manufacturer's instructions (see Appendix A2 for specific primer sets).

Auditory Brainstem Responses and Otoacoustic Emissions
For recordings of ABR, mice were anesthetized with a combination of ketamine (125 mg/kg) and xylazine (2.5 mg/kg) i.p. The core temperature was maintained constant at 37 • C using a heat blanket (Hugo Sachs Elektronik-Harvard Apparatus). For stimulus generation, presentation, and data acquisition, we used the TDT II System and BioSig software (Tucker Davis Technologies). Tone bursts (4/6/8/12/16/24/32 kHz, 10 ms plateau, 1 ms cos 2 rise/fall) or clicks of 0.03 ms were presented at 40 Hz (tone bursts) or 20 Hz (clicks) in the free field ipsilaterally using a JBL 2402 speaker. The difference potential between vertex and mastoid subdermal needles was amplified 50,000 times, filtered (400-4,000 Hz), and sampled at a rate of 50 kHz for 20 ms, 1,300 times, to obtain two mean ABR traces for each sound intensity. Hearing thresholds were determined with 10 dB precision as the lowest stimulus intensity that evoked a reproducible response waveform in both traces by visual inspection. The injected ear of virus-injected Otof-KO was recorded first and then plugged with electrode gel and tissue paper to achieve a conductive hearing impairment of 30-40 dB (Pauli-Magnus et al., 2007). This limited acoustic crosstalk while recording from the non-injected contralateral ear, though some remaining activation via reflected sound waves and bone conduction is still possible. Tone burst thresholds exceeding the maximal loudspeaker output (80 dB for tone bursts, 100 dB for Clicks) were assigned a value of 90 and 110 dB, respectively, for the calculation of means and SEMs. Saturation of amplitude growth functions was defined as the intensity at which the amplitude (peak to trough of the entire waveform) reached 80% of the value obtained at 100 dB.

Behavioral Tests of Hearing Function
In the Intellicage/Audiobox (New Behavior), operant conditioning of transponder-injected mice was performed 24/7 under computer control for 15 days. Mice visited the corner on average 60 times per day. In 90% of these visits, no stimulus was presented and the mice were allowed free access to water. In 10% of visits, 20 Hz click trains (5 clicks over 0.2 s, 0.3 s pause) at 100 dB were presented for the whole duration of the visit. Drink attempts during these visits were punished by a mild air puff to the neck of the mouse (0.7 bar) and access to water was denied. Two cages were run in parallel: one with two male Otof -KO siblings, and one with five Otof -KO females (age 14.4 ± 2.4 weeks), all with a successful ABR rescue (click threshold 50 dB). Sex matched WT controls (C57Bl/6 background, 16.0 ± 2.6 weeks) were measured as a separate dataset.

Data Analysis and Figure Preparation
Data was analyzed and figures were prepared using Microsoft Excel, Igor Pro and custom written Matlab routines. Figures were assembled using Adobe Illustrator, Affinity Designer, SnapGene and BioRender.com software (Figures 1, 2

AAVs Can Efficiently be Overloaded With the Full-Length Otof Coding Sequence
Successful packaging of fragments larger than 4.7 kb into standard AAVs had been reported in the past (Allocca et al., 2008;Grose et al., 2012;Hirsch et al., 2013;Pryadkina et al., 2015). Encouraged by these results, we packaged the fl-Otof -CDS and regulatory elements into various AAVs. When designing the experimental strategy, we decided against attaching an additional fluorescent reporter tag considering the large size of the fl-Otof -CDS. Therefore, analysis of transduction efficiency relied on post-mortem otoferlin immunohistochemistry. The fl-Otof -CDS used for this study was generated in several steps by performing in-fusion cloning strategies ( Figure 1A, for details see Material and Methods section and Appendix A1). Integrity of the obtained construct was validated by Sanger sequencing. In the final construct, we decided to employ a hybrid CMV/hbA promoter that is known to drive strong transgene expression in mammalian tissues ( Figure 1B). Although this promoter does not offer IHC-specificity per se, we expected that the combination of capsid, promoter and target gene would favor expression in IHCs (Akil et al., 2012;Jung et al., 2015;Al-Moyed et al., 2019). Moreover, due to otoferlin's highly specific function in the final steps of IHC exocytosis, we reasoned that potential ectopic expression in adjacent supporting cells, etc. should not interfere with hearing restoration in Otof-KOs. In addition, for enhancement of expression the Woodchuck Hepatitis Virus Posttranslational Regulatory Element (WPRE) and for stability of the transcript the bovine growth hormone (bGH) polyadenylation sequence were included between the inverted terminal repeats (ITRs). The entire sequence between the two ITRs was around 7.5 kb ( Figure 1B and Supplementary Figure 1B) exceeding the regular packaging capacity of AAVs (4.7 kb). We successfully packaged the construct into the novel virus serotype PHP.eB with consistent titers (Figures 1C,D and Supplementary Table 1). Packaging was also successful in our hands with several other AAV serotypes with reproducible titers (Figure 1D). We checked all virus preparations for the purity by conventional silver staining and, in most cases, detected excellent purity with three most important viral capsid proteins VP1, VP2, and VP3 ( Figure 1E, one representative virus preparation). The virus preparations were then employed for transduction efficiency assessment on otoferlin-deficient hair cells-initially in organotypicallycultured organs of Corti and subsequently in cochleae of newborn Otof-KO mice ( Figure 1F). Animals typically recovered well following injection and assessment of hearing as well as expression analysis were performed 4-5 weeks and 5 months later ( Figure 1G).

fl-Otof-Overloaded AAVs Transduce Hair Cells in vitro
To test if the fl-Otof-CDS-overloaded AAVs are capable of transducing mammalian IHCs, we first tested our different virus preparations on organotypically-cultured cochlear explants of Otof-KO mice (Figure 2). In these experiments, we dissected the organs of Corti at postnatal day (p) 5, cultured them overnight in serum-free medium and subsequently incubated them for 48 h with AAV-containing growth medium. After additional 6-8 days in vitro (DIV), we fixed the explants and processed them for immunohistochemistry, finally probing them with an otoferlin-specific antibody (Figures 2A,B). Here, all AAV preparations successfully and reproducibly transduced cultured IHCs-though at varying degrees-as well as some OHCs and various types of supporting cells, with AAV2/8 consistently performing the worst and PHP.eB the best (Figure 2C; mean transduction rates: AAV2/8: 7.1 ± 1.6% [n = 4 organs from N = 4 animals]; AAV2/9: 27.3 ± 3.7% [n = 4 organs from N = 4 animals]; Anc80L65: 17.0 ± 3.6% [n = 4 organs from N = 4 animals]; PHP.B: 17.9 ± 6.7% [n = 3 organs from N = 3 animals] and PHP.eB: 31.9 ± 2.0% [n = 4 organs from N = 4 animals]). This finding shows that the principle of AAVoverloading is not limited to a single serotype/capsid variant and reproducible across trials.

fl-Otof-Overloaded AAVs Successfully Transduce IHCs of Neonatal Mice in vivo
Encouraged by these promising in vitro results, we now selected the best-performing PHP.eB virus construct for subsequent in vivo analyses. Initially, we evaluated the in vivo transduction efficiency of the rAAV PHP.eB vectors carrying fl-Otof via immunohistochemical analysis (Figure 3). Importantly, cochlear and vestibular anatomy appeared to remain largely intact in the injected ears, with no obvious signs of persistent inflammation, malformation or other morphological anomalies, suggesting that neither the surgical procedure nor AAV transduction itself did exert any major adverse effects. In an attempt to address stability of transgenic otoferlin expression, we dissected and analyzed the apical turns of two distinct age groups (Group 1: directly post-hearing onset, i.e., 3 weeks; Group 2 at ∼6 months old; Figures 3A,A') and found a comparable transduction rate in both age groups of ∼30% (Group 1: 31.3 ± 2.4%, N = 2; Group 2: 29.1 ± 5.0%, N = 3; Figure 3C). Interestingly, our data indicate that insert size is indeed a main determinant of transduction rate and/or transgenic protein expression, since a truncated and GFPtagged, C-terminal fragment of otoferlin ("miniOtof ") that was packaged using the identical capsid/promoter construct as the fl-Otof-CDS-and thus solely differs in insert size-achieved much higher transduction rates. Yet, consistent with a previous report using a similar approach (Tertrais et al., 2019) our mini-Otof construct failed to restore hearing in Otof-KO mice (Supplementary Figure 3).
Next, total RNA isolation from fl-Otof -injected and contralateral non-injected cochleae (n = 5, Figure 5) confirmed full-length wild-type otoferlin expression in the injected cochleae of these Otof -KO mice, as we were able to detect several otoferlin regions covering multiple exons from the N-to C-terminus using PCRs with region-specific primers (Appendix A2). As the KO-strategy deletes exon 14 ( Figure 5D, black square), the finding of residual otoferlin mRNA fragments from PCR1 and PCR4 in the non-injected ear is expected, while the faint bands of PCR2 and 3 likely relates to spread of virus during the early postnatal injections.

fl-Otof-Overloaded AAVs Partially Restore Hearing in Otof-KO Mice
Next, we injected Otof-KO mice with fl-Otof -PHP.eB at p5-7 and then measured the auditory brainstem response (ABR) at the age of 5.5 ± 0.2 weeks. In these experiments, we found partly recovered ABRs to click stimulation in 18/25 tested mice with an average threshold of 58 ± 3 dB. In response to tone burst stimulation, only 10 injected ears showed reproducible responses (Figures 6A,B). As expected for a partial rescue, the ABR wave amplitudes were smaller than in WT mice. Like in other mutant mice with deficits of the IHC ribbon synapse (Buran et al., 2010;Strenzke et al., 2016), waves II, IV, and V were preserved better than waves I and III (Figure 6C). Non-injected  (Greenwood, 1961). Parameters were set according to Müller et al. (2005) resulting in a frequency range from a minimum of 5.1 kHz to a maximum of 83.1 kHz. The blue dotted line indicates reference points for the frequency-mapping algorithm in which a total of 500 coordinates were interpolated to fit the cochlear spiral. Selected frequency labels indicating the projected boundaries of the defined frequency bands (white dots) are indicated in the figure panel. Please note that the 4-8 kHz region was structurally less well-preserved in several of the processed cochleae, thus preventing meaningful analysis of this particular frequency band using this approach. (B) Representative maximum projections of the indicated frequency regions at higher magnification. IHCs were immunohistochemically-labeled against calretinin and otoferlin to assess viral transduction rates. (C) Quantification of IHC transduction rates in the indicated frequency regions using this approach: mean ± SEM and estimate from individual mice (blue spheres). Scale bars. contralateral ears also showed reproducible responses to click stimulation with an average threshold of 93 ± 3 dB ( Figure 6D). We believe that this response mainly reflects viral spread to the non-injected ear during early postnatal development but cannot completely exclude acoustic cross-talk. The amplitude of clickevoked ABRs rose with intensity ( Figure 6E) and approached its maximum from 74 ± 2 dB (n = 14). This suggests that like in WT, the recruitment of IHC/SGNs and the activation of individual SGNs saturates at high intensities. Partial viral rescue of hearing function persisted for several months, as demonstrated by sizeable ABRs recorded at an age of 5.5 ± 0.2 months ( Figure 6F). Other overloaded AAVs also partially rescued hearing function, albeit with a lower success rate (Supplementary Table 1 and Supplementary Figure 4).
Finally, to assess the restoration of auditory perception in PHP.eB fl-Otof AAV-treated mice, we used operant conditioning under automated control in the "intellicage" system (de Hoz and Nelken, 2014;Strenzke et al., 2016) and trained mice to discriminate between click trains and silence (Figure 7). During 10% of visits to the experimental "corner, " 100 dB click trains were presented and drink attempts were punished by an air puff. In 90% of visits, no sound was presented and access to water was allowed. All six tested Otof -KO mice with partial rescue of cochlear function (ABR thresholds of 50 dB) successfully learned the task with an average discrimination performance of 27.6 ± 3.0%. Matched normal-hearing C57Bl/6 WT control mice had a slightly better discrimination performance of 49.1 ± 3.5%.

DISCUSSION
Here, we investigated the potential of single AAV overloading for the future gene therapy of otoferlin-related auditory synaptopathy. Packaging of fl-Otof-CDS into several AAV variants was feasible at high titers (10 12 genome copies/ml) and purities. While all tested overloaded AAV variants successfully transduced IHCs in vitro and led to otoferlin expression in IHCs of cultured cochlear explants, the synthetic PHP.eB fl-Otof AAV performed the best in our initial screen. Subsequent in vivo analyses revealed that (i) the efficiency of otoferlin expression, (ii) specificity of transgene expression in IHCs, and (iii) rescue of auditory brainstem responses in Otof -KO mice upon early postnatal injection of PHP.eB fl-Otof AAV appeared largely comparable to that previously reported for the dual-AAV approach using the same age of injection (Al-Moyed et al., 2019), rendering our simplified approach highly interesting for further clinical development. Finally, for the first time to our knowledge, using operant conditioning, we could further demonstrate auditory percepts in Otof -KO mice that had undergone viral gene therapy. The small size and consequently good tissue penetration, low immunogenicity, lack of pathogenicity and episomal (extrachromosomal) DNA organization in the host cell make AAVs attractive vectors for gene therapy of sensory disorders (recently reviewed in Kleinlogel et al., 2020). For example, the recent approval of AAV-RPE65 gene therapy (Luxturna) of Lebers' congenital amaurosis 2, AAV-mediated vision restoration has been hailed a major breakthrough in this field (Bennett, 2017;Ledford, 2017).
In the past, coding sequences exceeding the transgene capacity of AAVs (4.7 kb) have been placed into lentiviruses or adenoviruses, which are plagued by issues such as insertional mutagenesis, mobilization, and vector replication or immunogenicity (Kleinlogel et al., 2020). More recently, split-AAVs (Duan et al., 1998, Duan et al., 2001Maddalena et al., 2018) and "overloaded AAVs" (Allocca et al., 2008;Grose et al., 2012;Hirsch et al., 2013;Pryadkina et al., 2015) have been employed in preclinical studies to circumvent these issues. In split-AAVs, the expression cassette is separated into two (dual-AAV) or three (triple-AAV) pieces, containing highly recombinogenic sequences and then packaged into two or three AAVs, respectively. Upon co-transduction the contained DNAs recombine into one single CDS. Dual AAV vectors have been successfully employed for expressing full-length otoferlin in IHCs of Otof-KO mice (Akil et al., 2012;Al-Moyed et al., 2019) achieving expression in up to 50% [injection at p6-7, (Al-Moyed et al., 2019)] and above 80% [injection at p17 or p30, (Akil et al., 2019)] of IHCs. Although our AAV overloading approach appears to be inferior in regards to transduction efficiency at this stage, it is highly encouraging that we observed homogeneous transduction rates along the tonotopic axis-at least until the 24 kHz range. Here, the large insert size seems to be the main limiting factor of transgenic otoferlin expression, as a substantially smaller mini-Otoferlin construct-packaged The amplitude of rescued ABR waveforms [blue, same as in (A)] is much smaller than in a representative trace from a single non-injected C57Bl/6 WT mouse (black) but latencies are comparable. SP denotes the summating potential, the presumptive inner hair cell receptor potential. Roman numerals on top indicate ABR waves I-V (Jewett et al., 1970). (D) ABRs from successfully rescued [blue, same as in (A,C)] and non-injected contralateral ears of the same animals (red) and a are shown for comparison. (E) Amplitude growth function of rescued ears (grand averages of ABR waveforms to 50-100 dB click stimulation, n = 17). (F) Grand averages of ABR waveforms to 80 dB click stimulation from ears with successful viral rescue, measured at an age of 5.5 weeks (blue, n = 18) and 5.5 months (green, n = 8). All blue traces are from the same dataset of PHP.eB fl-Otof AAV injected ears, age 5.5 weeks. Shaded areas represent means ± SEM.
into the exact same promoter/capsid combination-achieved drastically improved AAV transduction rates. Here, we used the efficient CMV-hbA promoter which despite its broad scope for driving expression in various cells, led to relatively specific transgenic expression of otoferlin in IHCs. In future studies, higher viral titers, IHC-specific promoters-such as FIGURE 7 | Rescue of hearing function in Otof-KO mice: operant conditioning. Successful discrimination between silence (water access) and 100 dB click trains (no water access, air puff punishment) was achieved in all six tested Otof-KO mice with successful viral rescue (blue colors), albeit with lower discrimination performance than in five C57Bl/6 WT controls (black and gray colors).
Myosin 15-or combinatorial virus application approaches (Lee et al., 2020) may help improve the efficiency and specificity of transgenic expression of otoferlin in IHCs. In addition, some other approaches coming into the focus of gene therapy studies-such as non-integrating lentiviruses (Suwanmanee et al., 2014;Snowball et al., 2019) or novel oversized AAVs, namely XL-AAVs (Ding and Gradinaru, 2020), with much larger packaging capacity-might overcome this limitation and finally lead to the full functional rescue of otoferlin-related hearing deficits.
The need to combine multiple distinct rAAVs presents an additional biosafety burden that may hinder clinical translation of such a treatment. This issue motivated us to explore the potential use of "overloaded AAVs" for future gene therapy of otoferlin-related auditory synaptopathy. Here, the omission of a fluorescent marker from the transgene is beneficial from a translational point of view, but presented a drawback for this preclinical study as it hampered the easy identification and more detailed analysis of IHC exocytosis by cell physiology. This was particularly critical given that otoferlin expression was post-hoc detectable in, on average, only a third of the IHCs with the present protocol. Future studies should compare the efficiency of both, AAV-overloading and split-AAV approaches using the same serotype/capsid and promoter combination as well as equal titers and virus administration protocols. Cell-physiological analysis, an important objective for future work, will be facilitated by employing a small tag for easy identification of transduced IHCs. Moreover, next to assessing otoferlin protein expression at various time points, single IHC sequencing should ideally be used to reveal the Otof DNAs generated by both approaches.
Finally, the extent of hearing restoration should be scrutinized using several complementary approaches. Partial restoration of synaptic function achieved with AAV-mediated Otof gene transfer into IHCs of Otof-KO mice has been demonstrated by recordings of ABRs (Akil et al., 2012;Al-Moyed et al., 2019 and this study), patch-clamp recordings from IHCs (Al-Moyed et al., 2019) and, finally, behavioral experiments (this study). Clearly, these results are very promising and indicate that clinical translation might eventually enable unaided hearing in people affected by OTOF mutations. Given the auditory synaptopathy phenotype (Moser and Starr, 2016) that is sensitively reported by lack or impairment of the population responses of SGNs [ABRs and electrocochleography (Moser and Starr, 2016)], the remaining ABR deficit might lead to an overestimation of the hearing impairment. Indeed, normal acoustic sensitivity has been reported for patients with some OTOF missense mutations (for review see: Santarelli et al., 2015) and the respective mouse model (Strenzke et al., 2016). Obviously, the more functional otoferlin remains present in case of missense mutations affecting the abundance of otoferlin at the plasma membrane-such as those causing the Ile515Thr substitution (Varga et al., 2006) or the Ile1967 deletion (Vogl et al., 2016)-the more likely it is that the transgenic supplementation of otoferlin expression in IHCs is going to restore normal hearing. Future preclinical work should aim to increase the certainty of predicting the outcome of hearing restoration in the clinical application of gene-therapeutic approaches to otoferlin-related auditory synaptopathy. For example, studies of temporal processing and of loudness adaptation will be helpful to assess potential remaining deficits (Strenzke et al., 2016). Eventually, work with species more closely related to humans, such as non-human primates, will help to prepare and accompany clinical trials.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

ETHICS STATEMENT
The animal study was reviewed and approved by board for animal welfare of the University Medical Center Goettingen and the animal welfare office of the state of Lower Saxony (Animal protocol Nr. 14_1391).

AUTHOR CONTRIBUTIONS
VR, CV, NS, and TM designed the study and wrote the manuscript with contributions of all other authors. VR, CV, IB, ND, CD-A, AT, TW, and KK performed experiments and/or analyzed data. All authors contributed to the article and approved the submitted version.

FUNDING
This work was supported by funding from the Deutsche Forschungsgemeinschaft via the collaborative research center 889 (to CV, NS, and TM) and under Germany's Excellence Strategy-EXC 2067/1-390729940 (to TM) and a grant of a University Medical Center Göttingen (UMG)-based donation of Dr. Grun (to TM). CV is an Otto Creutzfeldt Fellow of the Helmut and Elisabeth Uhl Foundation and ND was supported by a scholarship of the Göttingen Promotionskolleg für Medizinstudierende, funded by the Jacob-Henle-Programm or Else-Kröner-Fresenius-Stiftung (Promotionskolleg für Epigenomik und Genomdynamik, 2017_Promotionskolleg.04).