Long-term recovery of hearing in a mouse model of inherited deafness using targeted DNA editing

Within our deafness cohort (Cheng et al, 2025), we identified seven patients carrying KCNQ4 mutations linked to hearing impairment, including five pathogenic variants (Table EV1). Four of these variants have been previously reported: KCNQ4 c.887 G > A (p.G296D), c.853 G > A (p.G285S), c.961 G > A (p.G321S), and c.1251del (p.C418Afs*78). Furthermore, we discovered a novel pathogenic mutation, c.1107 T > A (p.Y369*, classified as PVS1 + PM2 + PP3). These patients presented a spectrum of hearing loss with varying ages of onset (Table EV1).

In vivo off-target profiling of the top ten Cas-OFFinder–predicted sites (≤3 mismatches) (Bae et al, 2014) together with four GUIDE-seq–identified loci (Malinin et al, 2021) revealed no detectable DNA off-target editing in the cochlear organ of Corti (Fig. EV4E–G). To assess systemic exposure following cochlear delivery, immunofluorescence and RT-qPCR analyses were performed after injection of AAV-ie–CMV-EGFP. Robust EGFP expression was detected in the hippocampus, with lower-level expression observed in the heart and liver (Fig. EV3D,E; Table EV2). Consistent with these findings, RT-qPCR quantification of ABE8e transcripts and corresponding target-site editing analyses across major organs—including the gonads—demonstrated detectable editor expression outside the cochlea, with the liver exhibiting the highest editing efficiency (~15%) (Appendix Fig. S5B,C). This peripheral activity is likely attributable to the immaturity of the blood–labyrinth barrier in neonatal mice (Xu et al, 2024), permitting transient vector leakage into the systemic circulation and preferential hepatic uptake; however, rapid hepatocyte turnover and immune-mediated clearance may contribute to the subsequent decline in editor expression at later time points (Kumar et al, 2024). Importantly, we did not observe overt hepatotoxicity (Appendix Fig. S7C). Given the potential safety implications of sustained editor expression in the central nervous system, we further interrogated DNA off-target activity at all twelve candidate sites in whole-brain tissue and again detected no significant off-target editing (Fig. EV4H). In addition, unbiased high-depth whole-genome sequencing (WGS; 100× coverage) of cochleae from high-dose–treated mice revealed no significant increase in either the frequency of A/T-to-G/C substitutions or the mutational spectrum compared with uninjected controls (Appendix Fig. S6). At the transcriptomic level, RNA-seq analysis identified an increase in A-to-I RNA editing events in the cochlea following AAV_ABE8e_N + C_322 administration (Fig. EV4I). Sequence-context analysis demonstrated a dose-dependent enrichment of these edits at endogenous ADAR-preferred motifs (Picardi et al, 2017). Notably, the majority of treatment-associated A-to-I events were intronic and functionally annotated as modifier variants (Appendix Fig. S5D–F), suggesting limited biological impact. Collectively, these data support a favorable genome- and transcriptome-wide safety profile of AAV-ie–mediated ABE8e delivery, while demonstrating efficient and precise in vivo correction of the pathogenic Kcnq4 p.G322S allele. These findings establish a robust preclinical foundation for the development of gene-editing–based therapeutic strategies for hearing restoration.

We quantified click ABR wave I amplitudes and latencies at 90 dB SPL to further characterize hearing recovery. At 4 weeks, no significant amplitude improvements were observed in either group (Fig. EV5L). However, partial frequency-specific latency reductions were observed in the treatment groups (Fig. EV5M). At 8 weeks, significant increases in amplitude were observed in both the low-dose (679.5–1124 nV) and high-dose groups (617.9–888.6 nV) across nearly all the tested frequencies, reflecting a significant recovery in the number of neurons firing (Fig. 4K). Latency reductions were also notable in both groups, with the low-dose group showing a range of 0.46–0.73 ms and the high-dose group 0.52–0.74 ms, indicating restored signal transmission velocity (Fig. 4L). In conclusion, our systematic evaluation highlights the efficacy of AAV_ABE8e_N + C_322 in restoring hearing in Kcnq4^+/G322S mice, with lower viral titers yielding superior long-term outcomes. These results underscore the importance of balancing therapeutic efficacy with cellular tolerance to optimize gene-editing interventions.

Following AAV_ABE8e_N + C_322 treatment, high-dose–treated Kcnq4^+/G322S mice exhibited a significant increase in I_K,n current density within the physiologically relevant voltage range (–40 to –10 mV) compared with untreated mutants, whereas the low-dose group showed a similar but statistically non-significant trend (Fig. 6B,C). Notably, both treatment regimens robustly restored the characteristic transient inward K⁺ currents, with current amplitudes at –140 mV increased by 113 and 87% in the high- and low-dose groups, respectively (Fig. 6B,D). Consistent with these improvements, AAV_ABE8e_N + C_322 treatment partially normalized the membrane polarization of mutant OHCs, as reflected by restoration of RMP by 17% (P = 0.0061) in the high-dose group and 12% (P = 0.0630) in the low-dose group relative to untreated mutants (Fig. 6E). Together, these electrophysiological restorations indicate that base editing–mediated correction of Kcnq4 partially rescues KCNQ4 channel function and re-establishes key biophysical properties of OHCs, thereby providing a mechanistic basis for the observed recovery of auditory function in vivo.

The systemic safety of AAV_ABE8e_N + C_322 was also assessed through comprehensive serum biochemical and immunological analyses of Kcnq4^+/+ mice four weeks after high-dose administration. No significant hepatic or renal function alterations were detected compared with those in uninjected controls (Appendix Fig. S7C). The evaluation of six inflammatory markers revealed no evidence of systemic immune activation (Appendix Fig. S7D). Survival analysis revealed comparable survival rates across high-dose, low-dose, and untreated Kcnq4^+/G322S mice (Appendix Fig. S7E), highlighting the tolerability of the treatment. These findings collectively underscore that, despite localized cochlear inflammation and immune activation, AAV_ABE8e_N + C_322 administration is well tolerated in mice, supporting its potential as a viable therapeutic strategy for KCNQ4-related hearing loss.

This study complied with NIH guidelines and was approved by the Institutional Animal Care and Use Committee (IACUC) of West China Hospital, Sichuan University (Approval No. 2020134A). Given prior reports on the use of gene editing to disrupt mutant alleles, the aim of this study is to investigate whether precise correction using base editors can restore hearing in mice more effectively. CRISPR/Cas9-mediated knock-in technology was employed to generate Kcnq4 c.964 G > A (p.G322S) mutant mice on an FVB/NJ background (MGP_FVBNJ_T0067200.1) by GemPharmatech Co., Ltd. (Nanjing, China). The mutation, located in exon 7, replaces GGC with AGC, and a synonymous mutation (CGG > AGA) at codon 339 is introduced to prevent re-cleavage (Fig. EV1). Briefly, targeted sgRNAs, Cas9 mRNA, and the donor template were injected into fertilized FVB/NJ eggs to generate F0 Kcnq4^G322S mice, which were then bred with wild-type mice to produce stable F1 progeny for further study. The mice were housed in a specific pathogen-free (SPF) facility at West China Hospital, with controlled temperature, humidity, and a 12-h light/dark cycle. For genotyping, tissue samples (e.g., tail or toe clips) were collected, and genomic DNA was extracted using the Animal Genomic DNA Quick Extraction Kit (Beyotime, China). PCR was performed with high-fidelity 2×KeyPo Master Mix (Vazyme, China) with forward (5′-CGTGAGCATCTGTGCAG-3′) and reverse (5′-GCTCCCTTTTCAGCTGTC-3′) primers. The PCR products were purified by 1% agarose gel electrophoresis and validated by Sanger sequencing.

Expression plasmids for ABEs and sgRNAs were obtained from Addgene (https://www.addgene.org/↗, #136300, #138489, #138491, #138492, #152989, #152990, and #47511). The blasticidin resistance gene was incorporated into the ABE plasmids using the NovoRec® Plus One-Step PCR Cloning Kit (Novoprotein, China). The sgRNA constructs were subsequently cloned and inserted into the pFYF1320 vector using the KLD Enzyme Mix (New England Biolabs, USA) according to the manufacturer’s instructions, with the cloning primers listed in Table EV3. The dual-AAV backbones Cbh_v5 AAV-ABE N-terminal and Cbh_v5 AAV-ABE C-terminal were also sourced from Addgene (#137177 and #137178) (Levy et al, 2020). The ABE8e protein was cleaved at the 574th amino acid of nSpCas9 to generate N- and C-terminal fragments, which were subsequently cloned and inserted into the respective vectors by homologous recombination (Novoprotein). Targeting and non-targeting sgRNAs were inserted into the C-terminal vectors using KLD Enzyme Mix. All the plasmids were verified by Sanger sequencing to confirm their correct construction.

For in vitro screening of ABE systems and sgRNAs, a partial mouse Kcnq4 sequence (Table EV3) harboring the p.G322S mutation (GGC to AGC), corresponding to the Kcnq4 p.G322S mouse model, was cloned and inserted into a lentiviral vector. Further, another HEK293T-KCNQ4 p.G321S cell line which carried the human p.G321S mutation (GGC to AGC) was constructed using the same strategy to validate its potential value on clinical translation. The lentivirus was packaged using a four-plasmid system. HEK293T cells were seeded in 10 cm dishes, and upon reaching ~70% confluence, they were transfected with the transfer plasmid containing Kcnq4 p.G322S, along with the packaging plasmids PLP1, PLP2, and VSV-G at mass ratios of 6.5 µg:5 µg:3.5 µg:5 µg. After 72 h, the culture supernatant was harvested, centrifuged at 500×g for 15 min to remove debris, and filtered through a 0.45 µm PVDF membrane. The filtered viral supernatant was used to infect HEK293T cells pre-seeded in 12-well plates, with polybrene (5 µg/ml) to enhance transduction. After 72 h, puromycin (2 µg/ml) was added to select transduced cells. A polyclonal cell line was established, and then serially diluted in 96-well plates to isolate single-cell wells. Monoclonal HEK293T cells carrying the Kcnq4 p.G322S target sequence were expanded for further use.

HEK293T cells were maintained in Dulbecco’s modified Eagle Medium (DMEM, Gibco) supplemented with 10% fetal bovine serum (FBS) and 1× penicillin/streptomycin under standard conditions (5% CO₂ at 37 °C). Transfections were performed using OriFect Transfection Reagent (Oriscience, China) following the manufacturer’s protocol. The cells were seeded in 24-well plates 12 h before transfection and transfected with 1 µg of plasmid DNA, maintaining the following ratios: ABE:sgRNA = 750:250 ng, ABE8e_N: ABE8e_C = 500:500 ng. After 8 h, the medium was replaced with fresh complete medium (DMEM + 10% FBS) to promote cell viability. At 24 h post-transfection, blasticidin (10 µg/mL) was added to select the transfected cells. At 96 h post-transfection, the supernatant was discarded, and the cells were lysed with cell lysis buffer (1 M Tris-HCl, pH 7.5, 10% SDS, 0.8 U/mL Protease K) and incubated at 37 °C for 1 h. The lysates were heat-inactivated at 80 °C for 20 min, vortexed, and allowed to return to room temperature for subsequent analysis.

The cochleae were dissected and placed in ice-cold PBS immediately. The cochlear capsule was carefully removed to expose the soft tissue, which was then separated from the modiolus. For DNA extraction, the cochlear organ of Corti was lysed in 30–50 µL of cell lysis buffer and incubated at 37 °C for 1 h, with vortexing every 15 min. After incubation, the samples were heat-inactivated at 80 °C for 20 min and stored at −80 °C. For RNA extraction, the RNA isolator Total RNA Extraction Reagent (Vazyme) was used. Fresh tissue was immersed in lysis buffer, and RNA was extracted according to the manufacturer’s instructions. The extracted RNA was reverse transcribed into complementary DNA (cDNA) using the PrimeScript™ FAST RT Reagent Kit with gDNA Eraser (Takara) and stored at −80 °C. The DNA of the mouse brain was extracted using a tissue genome extraction kit (TIANGEN, China) according to the manufacturer’s instructions.

The plasmids AAV_ABE8e_N, AAV_ABE8e_C_322, and AAV_ABE8e_C_nosgRNA were packaged into AAV vectors by co-transfection of HEK293T cells with Rep-Cap (AAV-ie) and Helper plasmids. Viral titers were quantified by quantitative PCR using primers specific to the ITRs. For the low- and high-dose groups, AAV_ABE8e_N and AAV_ABE8e_C_322 were co-administered at total doses of 1.30 × 10¹⁰ and 3.64 × 10¹⁰ genome copies per mouse, respectively, at an N-to-C terminal vector ratio of 1:1, in a total injection volume of 1 µL. All the AAVs were obtained from OBiO Technology (Shanghai, China).

Inner ear injections were performed on P2–P3 mice by the posterior semicircular canal (PSCC) and round window membrane (RWM) approach, as previously described (Kim et al, 2024). Briefly, AAV vectors were injected into the inner ears of Kcnq4^+/+, Kcnq4^+/G322S, and Kcnq4^G322S/G322S pups. Neonatal mice were anesthetized on ice, and a small incision was made below the right ear using sterilized ophthalmic scissors. The surgical site was exposed with an eyelid retractor (World Precision Instruments, USA), and the surrounding fat and connective tissue was carefully dissected to expose the sternocleidomastoid muscle, which was then incised to reveal the posterior semicircular canal or round window membrane. A glass micropipette (World Precision Instruments) attached to a nanoliter injection system (NL2020, World Precision Instruments) was used to inject 1 µL of the viral mixture containing AAV_ABE8e_N and AAV_ABE8e_C_322/nonsgRNA at a rate of 5 nL/s. After injection, the incision was sealed with 3 M Vetbond Tissue Adhesive (3 M Science, USA), and the pup was placed on a heated pad at 37 °C for recovery before being returned to its cage upon regaining full consciousness.

Mouse cochleae were fixed in 4% paraformaldehyde (PFA) at 4 °C overnight, followed by decalcification in 10% EDTA at room temperature for 48–72 h. Post-decalcification, the tissues were processed into cochlear whole mounts (apical, middle, and basal turns) and cochlear sections. The samples were permeabilized and blocked in 10% FBS with 0.1% Triton X-100 for 2 h and then incubated overnight at 4 °C with the following primary antibodies: Myosin-VIIa rabbit polyclonal antibody (1:1000, Proteus) for HCs, chicken anti-NeurofilamentH polyclonal antibody (1:1000, Sigma-Aldrich), rabbit anti-Iba1 antibody (1:500, abcam) for macrophages, mouse anti-Kcnq4 antibody (1:100, abcam) and mouse anti-Tubulin β3 (1:1000, BioLegend) for SGNs. Following primary antibody incubation, the samples were washed with PBS and incubated with secondary antibodies—Alexa Fluor Plus 647-conjugated donkey anti-rabbit IgG (1:500, Invitrogen) and Alexa Fluor 488-conjugated donkey anti-mouse IgG (1:500, Invitrogen), Alexa Fluor 488-conjugated Goat anti-Chicken IgY (1:500, Invitrogen) and Alexa Fluor 555-conjugated donkey anti-mouse IgG (1:500, Invitrogen)—for 2 h at room temperature, followed by nuclear staining with DAPI. Samples were mounted in Antifade Mounting Medium (Beyotime) and imaged on a Leica STELLARIS 5 confocal microscope (Leica Microsystems, Germany) using a ×63/1.40 oil-immersion objective, with or without digital zoom. Z-stack maximum projections were generated to compensate for tissue thickness in whole-mount preparations.

The quantification of cochlear HCs and SGNs was performed using LAS_X software on Kcnq4^+/+, Kcnq4^+/G322S, and Kcnq4^G322S/G322S mice post-injection. The number of Myosin-VIIa⁺ and DAPI⁺ IHCs and OHCs per 100 μm were counted at the apical, middle, and basal turns of cochleae. Tubulin β3⁺ and DAPI⁺ SGNs were quantified in cochlear sections according to tonotopic regions to assess the number of cells corresponding to auditory frequencies (Noh et al, 2022).

Mouse cochleae were harvested, punctured at the apex, and perfused with 2.5% glutaraldehyde for overnight fixation at 4 °C. After fixation, the tissues were rinsed in PBS, decalcified in 10% EDTA for 48–72 h, and dissected into apical, middle, and basal turns. The samples were subsequently fixed in 1% osmium tetroxide for 2 h, stained with 2% tannic acid for 1 h, and dehydrated through a graded ethanol series (50–100%). Following critical point drying (Quorum, UK), the samples were mounted on stubs, sputter-coated with gold (Smart Coater, JEOL, Japan), and imaged at ×2000 and ×8000 magnifications using a JSM-IT700HR scanning electron microscope (JEOL, USA).

ABR and DPOAE measurements were performed using the RZ6 acoustic system (Tucker-Davis Technologies, USA) (Cui et al, 2022). The mice were anesthetized by the intraperitoneal injection of xylazine (10 mg/kg) and ketamine (100 mg/kg) and were placed on a 37 °C heating pad to maintain thermoregulation. Following a 3–5 min stabilization, the animals were transferred to a soundproof chamber with continuous heating. ABR recordings were obtained via a closed-field setup. Reference, recording, and ground electrodes were placed subcutaneously at the cranial vault between the ears, the right pinna near the mastoid, and the dorsal rump, respectively. A sound delivery tube connected to an MF1 magnetic speaker (Tucker-Davis Technologies) was inserted into the external auditory canal. Tone burst stimuli (3 ms) were presented at frequencies from 4 to 32 kHz in half-octave intervals (4, 5.656, 8, 11.314, 16, 22.627, 32 kHz), with intensities ranging from 5 dB SPL to 90 dB SPL in 5 dB steps. Auditory signals were amplified by 10,000× and filtered from 300 Hz to 3 kHz. The ABR threshold was determined as the minimum sound intensity producing a detectable waveform, averaged from 400 tone bursts (ABR thresholds >90 dB SPL were capped at 100 dB SPL for analysis). For DPOAE, a custom probe containing two MF1 magnetic speakers and an ER 10B+ microphone (Etymotic Research) was inserted into the right external auditory canal. Primaries f1 and f2 (f2/f1 = 1.2) were presented with f2 spanning from 4 to 32 kHz in half-octave intervals, with L1−L2 set to 10 dB SPL. For each f2, L2 was incrementally varied from 20 to 80 dB SPL in 10 dB increments. DPOAE thresholds were defined as the f2 amplitude that generated a DPOAE response of 6 dB SPL above the noise floor, which was consistently below 0 dB SPL across all frequencies (DPOAE thresholds >80 dB SPL were capped at 90 dB SPL for analysis). The detailed parameters are shown in Table EV4. Mice were drawn from multiple litters and randomly assigned to experimental or control groups; to minimize litter effects, mice from different litters were distributed across groups. Right-ear (injected ear) hearing thresholds were measured for each mouse, and both sexes were included. All measurements were conducted by a single experimenter blinded to the group assignments to minimize potential bias. Following the experiments, the mice were placed on a 37 °C heating pad for recovery and returned to the breeding room once fully revived.

Putative gRNA-dependent off-target sites were predicted using Cas-OFFinder against the Mus musculus reference genome (GRCm38) (Bae et al, 2014). The top ten candidate loci (Fig. EV5E) were validated by HTS to assess A-to-G editing frequencies. Genome-wide off-target sites were also assessed on mouse-derived N2a cells using GUIDE-seq, as previously described (Malinin et al, 2021). Two candidate off-target sites were identified based on dsODN integration (Fig. EV5E) and validated by HTS to assess A-to-G editing frequencies.

Total RNA sequencing libraries were prepared and subjected to paired-end 150 bp sequencing by BGI Genomics (Wuhan, China). Raw sequencing reads were filtered with fastp (Chen et al, 2018a) and quality-checked using FastQC. Clean reads were aligned to the Mus musculus reference genome (GRCm39) using STAR, and gene-level quantification was performed with RSEM (Dobin et al, 2013; Li and Dewey, 2011). To assess RNA off-target effects induced by base editors, variants were called with HaplotypeCaller (GATK v4.1.2.0) on deduplicated, post-alignment reads. For each SNV, only sites with read depth >20 and absent in untreated controls were considered high-confidence variant calls (Cui et al, 2024). To interrogate sequence-context preference, detected A-to-I variants were intersected with a curated set of endogenous ADAR hotspot loci (Picardi et al, 2017). Variant functional consequences were annotated using Ensembl variant effect predictor (VEP) to evaluate genomic distribution and predicted impact (McLaren et al, 2016).

Cochleae from high-dose–injected and uninjected control mice were collected 3 weeks post-administration. Genomic DNA extraction, WGS library preparation, and sequencing were performed by BGI Genomics. Libraries were sequenced on the DNBSEQ platform to generate 150-bp paired-end reads at an average depth of ~100×. Raw reads were processed using SOAPnuke (BGI) to remove adapter contamination and low-quality sequences (-n 0.01 -l 20 -q 0.5 --adaMR 0.25 --polyX 50 --minReadLen 150) (Chen et al, 2018b). Reads were discarded if they contained ≥25% adapter sequence (allowing ≤2 mismatches), were shorter than 150 bp, had ≥1% ambiguous bases (N), contained polyX tracts >50 bp, or had ≥50% of bases with Phred quality <20. The remaining reads were retained as clean reads. Clean reads were aligned to the FVB/NJ reference genome using BWA (v0.7.15) (Li and Durbin, 2009). Germline variants were identified using Sentieon (GATK HaplotypeCaller implementation) and functionally annotated with ANNOVAR (Van der Auwera et al, 2013; Wang et al, 2010). To assess global mutational burden in an unbiased manner, single-nucleotide variant (SNV) frequencies and base-substitution spectra were compared between high-dose–treated and uninjected control samples.

HTS was employed to assess gene editing efficiency and off-target effects. Specific primers, each associated with unique barcodes, were designed to amplify HEK293T DNA, cochlear organ of Corti DNA, or complementary DNA (cDNA) (Tables EV5 and EV6). These primers facilitated two-step PCR amplification to generate sequencing-ready libraries for evaluating editing outcomes. In the first PCR step, a 25 µl reaction mixture was prepared containing 0.5 µM each forward and reverse primer, 12.5 µl of 2 × Phanta Flash buffer, 0.5 µl of Phanta Flash Super-Fidelity DNA Polymerase (Vazyme), and 2 µl of sample DNA or cDNA. The PCR conditions were as follows: denaturation at 98 °C for 30 s, followed by 28 cycles of 98 °C for 10 s, 60 °C for 5 s, and 72 °C for 10 s, with a final extension at 72 °C for 2 min. The PCR products were purified using 2% agarose gel electrophoresis and extracted with the TIANgel Purification Kit (TIANGEN). In the second PCR step, indexed HTS adapters were added to the purified first-round PCR products using a 25 µl reaction mixture containing adapter-specific primers, 2 × Phanta Flash buffer, Phanta Flash Super-Fidelity DNA Polymerase, and 1 µl of the mixed purified first-round PCR product. The PCR program consisted of denaturation at 98 °C for 30 s, followed by ten cycles of 98 °C for 10 s, 60 °C for 5 s, and 72 °C for 10 s, with a final extension at 72 °C for 2 min. The final PCR products were again purified by gel electrophoresis and extracted with the TIANgel Purification Kit. The purified libraries were sequenced using the Illumina NovaSeq X Plus PE150 strategy at Novogene Co., Ltd. (Beijing, China). Sequencing data were demultiplexed based on barcodes and analyzed using CRISPResso2 software to quantify on-target editing efficiency and off-target modifications (Clement et al, 2019). On-target editing efficiency was calculated as the absolute correction rate, defined by subtracting the background G allele frequency in untreated controls from the G allele frequency in treated samples: (G/(A + G))%_Treated − (G/(A + G))%_Control.

Whole-cell voltage‒clamp recordings of K⁺ currents in OHCs were conducted using an EPC10 amplifier (HEKA Elektronik, Lambrecht/Pfalz, Germany). Patch pipettes were pulled from borosilicate glass capillaries and heat-polished to achieve tip resistances of 3–4 MΩ. The cochleae were acutely isolated and maintained in an ice-cold dissection solution containing (in mM): 144.6 NaCl, 5.5 KCl, 1 MgCl₂, 0.1 CaCl₂, 0.5 MgSO₄, 10.2 HEPES, and 3.5 L-glutamine, pH 7.2 (adjusted with NaOH). The apical turn of the cochlea was dissected, mounted onto a coverslip with a nylon net, and perfused with an external bath solution containing (in mM): 145 NaCl, 5.8 KCl, 0.9 MgCl₂, 1.3 CaCl₂, 0.7 NaH₂PO₄, 10 HEPES, and 5.6 D-glucose, pH 7.4 (adjusted with NaOH). The pipette solution consisted of (in mM): 135 KCl, 3.5 MgCl₂, 0.1 CaCl₂, 2.5 Na₂ATP, 5 HEPES, and 5 EGTA, pH 7.4 (adjusted with KOH). To record K⁺ currents in OHCs in the whole-cell configuration, the cells were held at −80 mV, with step voltages ranging from −140 to +30 mV in 10-mV increments. After breaking through the cell membrane to establish a giga-ohm seal, the series resistance (R_s) and membrane capacitance (C_m) were compensated. The current density‒voltage relationship was derived by normalizing the current to the cell capacitance. The cell capacitance was estimated using capacitive transients as an indirect measure of the cell size. Liquid junction potentials were measured and appropriately corrected.

Ten inflammatory parameters, including IFN-γ, IL-10, IL-12p70, IL-1β, IL-2, IL-4, IL-5, IL-6, KC/GRO, and TNF-α, were measured using electrochemiluminescence V-PLEX proinflammatory panel 1 mouse kit (MSD, USA). Briefly, Cochlear samples were lysed using tissue cell lysis buffer (Absin, China), yielding ~25 μL of supernatant. This supernatant was subsequently diluted threefold and subjected to further processing following the manufacturer’s instructions. Final concentrations were calculated using the MSD Discovery Workbench 4.0 software (MSD, USA).

Serum biochemical parameters in high-dose-injected and control Kcnq4^+/+ mice were quantified by Lilai Biotechnology Co., Ltd. (Chengdu, China) using a BS-460 biochemical analyzer (Mindray, China). The parameters measured included ALT, AST, ALP, CREA-S, TP, ALB, LDH, UREA, and Glo. Serum immune markers (IFN-γ, IL-1α, IL-1β, IL-6, IL-10, and TNF-α) in these mice were assessed using the Multiplex ELISA Kit for Mouse Cytokine Panel 1 (6-Plex) (MEK1011, BOSTER, USA) according to the manufacturer’s instructions. Survival analysis was performed on high-dose, low-dose, and untreated Kcnq4^+/G322S mice over a 32-week period.

Statistical analyses and data visualization were performed using R v4.4.1. The sample sizes, exact p values and the specific statistical tests employed are detailed in the figure legends. All replicates are biological unless stated otherwise. Randomization and blinding were used whenever applicable to minimize the effects of subjective bias.

The schematics in Figs. 1, 3, 6 and EV1; Appendix Fig. S1 and synopsis graphics were created with BioRender.com