XBP1 mitigates aminoglycoside-induced endoplasmic reticulum stress and neuronal cell death

Aminoglycosides alter translation fidelity

Drug-induced inhibition of translation was used to assess aminoglycoside activity on the eukaryotic ribosome. IC50 values were 0.3 μM for geneticin and 9.8 μM for gentamicin in the cell-free translation assays with rabbit reticulocyte lysate (RRL), and 4.4 μM for geneticin and 812 μM for gentamicin in assays with intact HEK293 cells (Supplementary Figures S1a and b). The ability of the drugs to induce mistranslation was analyzed using sensitive gain-of-function dual-luciferase assays to assess near-cognate misreading and stop codon readthrough. Near-cognate misreading was studied using constructs with substitution of amino-acid 245 in the active site of mutated firefly luciferase (wild-type His CAC→near-cognate Arg CGC), which results in loss of enzymatic activity with enzymatic function restored by misreading; stop codon readthrough was determined using constructs with in-frame stop codons abolishing firefly luciferase activity. Both geneticin and gentamicin decreased ribosomal accuracy in cell-free translation assays (RRL) and in HEK cells in a dose-dependent manner (Figure 1). Misreading was induced up to 25-fold in RRL and up to 8.5-fold in HEK cells compared with untreated controls; readthrough was induced up to 20-fold in RRL and up to 70-fold in HEK cells compared with untreated controls (Figure 1). In HEK cells transfected with the aminoglycoside phosphotransferase APH(3′), the geneticin-induced but not the gentamicin-induced translation inhibition and mistranslation were abrogated (Supplementary Figures S1c and d), consistent with the selectivity of the enzyme to inactivate geneticin but not gentamicin.^{38, 39} Aminoglycoside-treated and -untreated HEK wild-type cells showed similar metabolic activities and viability at both the 24-h and 48-h time points (Supplementary Figures S1e and f).

Aminoglycosides induce genome-wide upregulation of cellular folding capacity

In order to study the cellular response to aminoglycoside-induced mistranslation, we used whole-genome transcriptomic and proteomic analyses. A microarray analysis of geneticin-treated versus -untreated cells revealed a broad transcriptional response totaling 705 genes (selected for a fold change >1.2, Benjamini–Hochberg corrected P-value <0.05; Supplementary Figure S2a). Protein folding and transcription were among the most enriched functional ontologies (Supplementary Figure S2b), including the induction of the ER-specific chaperones BiP (HSPA5), glucose-regulated protein 94 (GRP94; HSP90B1), calreticulin (CALR), GRP110 (HYOU1), ERdj3 (DNAJB11) and ERdj6 (DNAJC3), the ER foldases PDIA3 (ERp57), PDIA4 (ERp70), ERp44 and FKBP7, and the N-linked glycosylation factor SDF2L1. Similarly, ER-associated degradation (ERAD) components such as VCP (p97), Derlin2 (DERL2) and Herp (HERPUD1) were significantly upregulated (Supplementary Figure S2c). This transcriptional response indicates a broad increased folding and degradation capacity in the ER. In addition, a large number of cytosolic chaperones⁴⁰ were upregulated, such as members of the Hsp40, Hsp70, Hsp90 and Hsp110 families and to a lesser extent foldases (peptidyl-prolyl cis/trans isomerases and protein disulfide isomerases; Supplementary Figures S2d and e), indicating an increased folding capacity in the cytosol. Supplementary Table S1 lists the genes included in the analysis. The microarray data have been deposited in NCBI's Gene Expression Omnibus and are accessible through GEO Series accession number GSE57198 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE57198).

Proteome analysis found 77 proteins to be regulated by geneticin (Bonferroni-corrected P-value <0.05). When applying a minimum fold induction of 0.3 (log2 scale) we identified 35 proteins that were upregulated. Grouping according to function revealed a predominance of proteins involved in protein folding (Figure 2a). Proteins associated with the ER and cytoplasmic UPR, such as BiP, GRP94, calreticulin, foldases, and members of the Hsp70, Hsp90, Hsp110 and Hsp40 families, were also upregulated (Figure 2b). Comparison with corresponding mRNA levels showed an upregulation of the folding machinery both at the transcriptomic and the proteomic level (Figure 2c). The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the data set identifier PXD000933 and DOI 10.6019/PXD000933.

Aminoglycosides induce the UPR

To corroborate the results of the microarray analysis, mRNA levels of selected UPR genes were further analyzed by quantitative PCR and corresponding protein levels assessed by western blotting. Geneticin and gentamicin-induced mRNA expression of C/EBP homologous protein (CHOP), GRP94 and BiP in a time-dependent manner (Figures 3a–c). Increased protein levels of the two ER chaperones BiP and GRP94, as well as the transcription factor ATF4, which is regulated at the translational level,⁴¹ were observed in geneticin- and gentamicin-treated cells by western blotting (Figure 3e). As a further element of the UPR, we studied splicing of XBP1 mRNA, which is central for UPR activation.¹¹ Both geneticin and gentamicin induced XBP1 splicing (Figure 3d). In contrast, XBP1 splicing was induced neither by the non-misreading aminoglycoside hygromycin⁴² nor by cycloheximide, an inhibitor of ribosomal translocation,⁴³ indicating that XBP1 splicing depends on misreading and not on inhibition of translation. Furthermore, the presence of APH(3′) in HEK cells abrogated geneticin-induced but not gentamicin-induced XBP1 splicing.

The activity of transcription factors XBP1 and ATF6 was examined using reporter plasmids UPRE (p5xATF6-GL3-luc) and ERSE (pGL3-GRP78P(-132)-luc).^{44, 45} The UPRE reporter is specific for ATF6 activity, the ERSE reporter is regulated by both ATF6 and XBP1.^{44, 45} Both reporters showed a robust induction by geneticin and gentamicin (Figures 3f and g). Cycloheximide failed to induce any reporter activity consistent with the XBP1-splicing results (Figures 3d, f and g). The PERK signaling branch was investigated by assessing the formation of stress granules, cytosolic protein aggregates composed of 48 S preinitiation complexes and other factors. Stress granules are induced upon activation of PERK and phosphorylation of eIF2α.⁴⁶ Treatment of HEK wild-type cells with geneticin for 24 h increased immunostaining against phosphorylated eukaryotic initiation factor 2 alpha (p-eIF2α) in a dotted cytosolic distribution consistent with the formation of stress granules (Figure 3h). Arsenite treatment served as a positive control. A similar robust induction of UPR by aminoglycosides was observed in HeLa cells (Supplementary Figure S3).

Gentamicin induces ER stress in SGCs but not in auditory hair cells

To study the response of auditory hair cells to ER stress, we first used tunicamycin, an established ER stress inducer in early post-natal cochlear explants of the CBA/J mouse, a common strain for auditory research. Preliminary experiments (data not shown) with hair cell counts on surface preparations had established incubation with 0.07 μg/ml tunicamycin as a suitable treatment with hair cell death beginning at 48 h and encompassing 50% of cells by 72 h. The ER stress-associated pro-apoptotic factor CHOP already appeared after 8 h of incubation with tunicamycin and was expressed in the nuclei of most hair cells by 24 h (Figure 4a and Supplementary Figure S4). Staining mostly had disappeared at 48 h (Supplementary Figure S4) when loss of hair cells became apparent, implicating CHOP as an indicator of impending hair cell death. In the same explant model, treatment with gentamicin produced significant loss of hair cells with the pattern of loss showing the typical progression of aminoglycoside damage⁴⁷ causing most destruction in the base (Supplementary Figure S5), whereas inner hair cells (IHCs) were mostly spared. Despite continuing and increasing cell death, CHOP was not observed throughout the entire time course up to 72 h (Figure 4a).

The response of ganglion neurons to ER stress was studied in SGCs that were harvested from the base to the middle of the modiolus of cochlear explants and similarly treated with tunicamycin or gentamicin (Figure 4b). As expected from its activity as an ER stressor, tunicamycin induced CHOP in the nuclei of SGCs within 24 h. In contrast to its effect on hair cells, gentamicin increased the immunoreactivity to CHOP in SGCs, evident after 48 h of incubation.

Gentamicin reduces the number of SGCs and synaptic ribbons but not hair cells in XBP1^+/− mice in vivo

In wild-type strains such as the CBA/J mouse, the outer hair cells (OHCs) are the primary target of chronic aminoglycoside ototoxicity in vivo⁴⁸ and very little direct effect – if any – can be observed on SGCs. In view of the modest but significant gentamicin-induced CHOP expression in SGCs of cochlear explants, we investigated potential consequences of gentamicin-induced ER stress in a model of compromised UPR, an XBP1-haploinsufficient mouse. The local route of drug administration to the middle ear, chosen for this study, is able to isolate effects to the auditory periphery while avoiding adverse complications associated with systemic gentamicin treatment in the mouse.³⁰ Preliminary experiments with a series of gentamicin concentrations starting at 0.09 M established 0.56 M as a suitable dose that caused a moderate auditory threshold shift while avoiding major pathophysiology (data not shown).

Surface preparations from XBP1^+/− and wild-type littermates locally treated with gentamicin in vivo were examined from base-to-apex 3 weeks after drug injection. OHCs were present in all parts of the cochlea in both wild-type and XBP1^+/− mice except for some scattered loss at the very end of the basal turn (Supplementary Figure S5c). Quantitation of hair cell loss along the entire cochlea confirmed only minor damage at the extreme, the basal turn with no difference between wild-type and XBP1^+/− mice.

In the absence of any discernible defects on hair cell integrity and prompted by the in vitro results, we then analyzed spiral ganglion density and synaptic connections. Three weeks after gentamicin injection, the SGCs were counted on mid-modiolar cryosections stained for β-tubulin and nuclei. There was a significant reduction in spiral ganglion density in the basal turn of the cochlea in XBP1^+/− mice but not in wild-type littermates (Figures 5a and b). The innervation of hair cells by the spiral ganglion was assessed by staining synaptic ribbons with antibody to CtBP2, a constituent of the ribbon protein RIBEYE. Gentamicin reduced the number of synaptic ribbons per IHC by approximately 50% in the basal turn of the cochlea of the XBP1^+/− mice but not of corresponding wild-type littermates (Figures 5c and d).

Auditory physiology corroborates auditory pathology and ER stress

In order to assess the impact of the observed pathology on auditory function, we measured auditory brainstem responses (ABRs) and distortion product otoacoustic emissions (DPOAEs). ABR provides information on the ascending auditory pathway reflecting synaptic and neuronal activity, whereas DPOAE probes the functional integrity of OHCs. Deterioration of auditory thresholds was apparent 1 week after the injection of gentamicin and remained stable for up to 3 weeks, the latest time point studied (Figure 6a). Large threshold shifts were observed at 32 kHz in XBP1^+/− mice but not in wild-type littermates, which were little affected. Consistent with the morphological observations of intact OHCs, DPOAE remained unaffected by gentamicin treatment (Supplementary Figure S6).

Finally, in order to validate the potential contribution of protein misfolding to the gentamicin-induced changes in auditory thresholds, we treated animals with tauroursodeoxycholic acid (TUDCA), a clinically used chemical chaperone. Systemic TUDCA co-administration significantly attenuated gentamicin-induced ototoxicity in the XBP1^+/− mice (Figure 6b) as measured by ABR 3 weeks after the drug treatment.

Materials and sources

Mouse monoclonal anti-GADD 153 (B-3) antibody (Santa Cruz Biotechnology, Dallas, TX, USA); polyclonal antibody against neuronal class III β-tubulin (Covance, Princeton, NJ, USA); monoclonal anti-CtBP2 antibody (BD Biosciences, San Jose, CA, USA); polyclonal antibody against p-eIF2α (Cell Signaling, Danvers, MA, USA); polyclonal anti-myosin 7a antibody (Proteus Biosciences, Ramona, CA, USA); secondary goat anti-rabbit antibody conjugated with Texas Red (Abcam, Cambridge, MA, USA); rhodamine phalloidin, Invitrogen (Life Technologies, Carlsbad, CA, USA); HEK293 cells (Innoprot, Biscay, Spain); geneticin, gentamicin, tunicamycin, cycloheximide, arsenite, saponin and HEK aph(3′) cells (Sigma Aldrich, St. Louis, MO, USA); hygromycin (PAA Laboratories, Cansera, Canada); nucleotide primers (Microsynth, Balgach, Switzerland); cell culture media and trypsin (Life Technologies).

Assessment of mistranslation

Misreading and stop codon readthrough were assessed in gain-of-function dual-luciferase assays.^{60, 61} For translation in RRLs (Promega, Madison, WI, USA), luciferase mRNA was produced in vitro using T7 RNA polymerase (Thermo Scientific, Waltham, MA, USA) and plasmids pGL4.14 (humanized firefly luciferase, hFluc) and pGL4.75 (humanized renilla luciferase, hRluc; both from Promega), where the mammalian promoter was replaced by the T7 bacteriophage promoter. For misreading, we replaced residue His245 (CAC codon) with Arg245 (CGC near-cognate) in the hFluc protein by site-directed mutagenesis. Readthrough was assessed with a fusion construct in which hRluc and hFluc were fused by a 27-nucleotide linker encoding the polypeptide STCDQPFGF, using overlap PCR mutagenesis to result in the pT7 hRluc-hFluc vector; a UGA nonsense-codon was introduced at the glutamine residue (wild-type CAA) of the linker sequence by site-directed PCR mutagenesis. A cell-free luciferase translation assay was performed as described.⁶⁰

Mistranslation in HEK cells was determined using the pRM hRluc-hFluc H245R vector, where His245 (CAC codon) was replaced by Arg245 (CGC codon) in the pRM hRluc-hFluc vector. Readthrough was determined by pRM hRluc-hFluc D357X, where Asp357 (GAC codon) was replaced by a UGA nonsense-codon in the firefly luciferase transcript. Both constructs were designed by site-directed PCR mutagenesis. HEK wild-type cells were transfected with reporter plasmid using TurboFect (Fermentas, Vilnius, Lithuania) according to the manufacturer‘s protocol. After a 24-h incubation, medium was replaced by F10 with 15 μg/ml saponin. Aminoglycoside antibiotics were added and cells were incubated for another 24 h. Cells were lysed and luciferase activities determined; hRluc mRNA was used as an internal control and misreading and readthrough were quantified by calculating mutant firefly/renilla activities. The basal error frequency of the eukaryotic ribosome⁶² is 4 × 10^–4 to 10^–5. For each set of replicates, the hFluc/hRluc ratio of the untreated samples were set as 1, which reflects this basal error frequency. Luminescence was measured in a luminometer FLx800 (Bio-Tek Instruments, Winooski, VT, USA).

Viability assay

HEK cells were grown to 70% confluence and treated for the indicated time with 16 μM geneticin or 400 μM gentamicin in F10 medium with 15 μg/ml saponin. Ten percent Alamar Blue solution was added (v/v) for 3 h and fluorescence was monitored at 530 nm for excitation and 590 nm for emission. The fluorescence level of the control sample (untreated) was set as 100% after subtraction of background fluorescence, measured in cell-free wells.

Sytox dead cell stain

HEK cells were grown to 60% confluence in DMEM with 10% fetal bovine serum (FBS). Medium was changed to F10 with 15 μg/ml saponin and aminoglycoside antibiotics were added and cells were incubated for 24 or 48 h. Cells were detached by adding 100 μl accutase (Life Technologies) and were resuspended in 400 μl FACS buffer (1x phosphate-buffered saline (PBS), 2% FBS) and transferred to FACS tubes. Sytox Red (Life Technologies) was added to the cell suspension. The nucleic acid stain penetrates cells with compromised plasma membranes but will not cross uncompromised cell membranes. The samples were then analyzed with a BD FACS Canto II (BD Biosciences, Allschwil, Switzerland) and the FlowJo data analysis software (FlowJo, Ashland, OR, USA).

Microarray analysis

See the legend of Supplementary Figure S2.

Proteome analysis

Cell samples were incubated with lysis buffer (150 mM NaCl, 0.1% SDS, 0.5% Na-deoxycholate, 50 mM Tris pH 7.5 and 1 × complete protease inhibitor (Roche, Rotkreuz, Switzerland)) for 10 min at room temperature (RT) on a shaking mixer. The lysate was ultrasonicated for 10 min and centrifuged for 20 min at 16 000 × g at 4 °C. Eighty micrograms of protein of each sample were used for iTRAQ labeling (AB SCIEX, Framingham, MA, USA).

Each iTRAQ 4-plex experiment was carried out with two biological replicates of untreated HEK wild-type cells (114 and 116 label) and two biological replicates of cells treated with 16 μM geneticin for 32 h (115 and 117 label) following the manufacturer's protocol. iTRAQ samples were pooled, dried, reconstituted in solvent A (5% ACN, 8 mM KH₂PO₄ and pH 4.5), and fractionated by HILIC-HPLC (Pack Polyamine II, 250 × 4 mm, 120 Å S-5 μm, YMC). The column was equilibrated with solvent A. Peptides were eluted using solvent B (5% ACN, 100 mM KH₂PO₄ and pH 4.5) by a gradient of: 0–7.5 min, 0% B; 7.5–37.5 min, 0–50% B; 37.5–42.5 min, 50–100% B; 42.5–47.5 min, 100% B at a flow rate of 0.4 ml/min. The resulting 13 fractions were desalted using ZipTips (Millipore, Billerica, MA, USA) according to the manufacturer's protocol and reconstituted in solvent C (3% ACN and 0.1% formic acid) for LC-MS/MS analysis. Samples were auto-injected into an Eksigent-nano-HPLC system and separated on a custom reverse phase tip column (75 μm × 150 mm) packed with C₁₈ material (3 μm, 200 Å, AQ, Bischoff GmbH, Leonberg, Germany). The column was equilibrated with solvent C and 5% solvent D (0.2% FA in ACN). For elution, a flow rate of 300 nl/min was used and a gradient of 0–70 min, 5–25% D; 70–85 min, 25–50% D; 85–88 min, 50–98% D. High accuracy mass spectra were acquired with an AB SCIEX 5600 mass spectrometer (AB SCIEX) in the range of 385–1250 m/z. Up to 36 data-dependent MS/MS were recorded in high sensitivity mode of the most intense ions with charge states 2+, 3+ and 4+ using collision-induced dissociation. Target ions already selected for MS/MS were dynamically excluded for 90 s after three occurrences. MS/MS data were analyzed using Mascot 2.4 (Matrix Science, Boston, MA, USA) and searched against a decoyed human database from Swissprot (Lausanne, Switzerland; release December 2012) concatenated with an in-house build contaminant database. The search parameters were: precursor ion mass tolerance of 20 p.p.m., fragment ion mass tolerance of 0.05 Da, trypsin digestion, fixed modifications of MMTS-labeled cysteine, 4-plex iTRAQ modifications of free amines at the N-termini and of lysine and variable modification 4-plex iTRAQ of tyrosine. Peptides without 4-plex iTRAQ labeling at the N-terminus or at a lysine were excluded from the analysis. Scaffold_4.1 (Proteome Software Inc., Portland, OR, USA) was used to validate MS/MS-based peptide and protein identifications. We identified and quantified 1785 proteins (protein prophet probability 95%, minimum two peptides for identification of a protein and minimum Mascot Ionscore of 40). After the permutation test and further amendment of the P-value with the Bonferroni correction, 77 proteins were found to be regulated (P-value<0.05). Thirty-five proteins were upregulated based on a threshold of 0.3 (log₂ scale).

XBP1-splicing assay and qPCR

RNA samples from HEK cells were prepared using Trizol extraction (Life Technologies) and were reverse transcribed using a ThermoScript RT-PCR System (Life Technologies) according to the manufacturer's instructions. The XBP1-splicing assay used XBP1-specific primers that amplify spliced (−26 nt) and unspliced XBP1 mRNA (forward 5′-TTACGAGAGAAAACTCATGGCC-3′, reverse 5′-GGGTCCAAGTTGTCCAGAATGC-3′). PCR products were analyzed on a 2.7% agarose gel. Amplification of glyceraldehyde 3-phosphate dehydrogenase (GAPDH) cDNA served as loading control. For qPCR the Quantitect SYBR Green PCR Kit (Qiagen, Hilden, Germany) was used together with the 7500 Fast Real-Time PCR System (Applied Biosystems, Zug, Switzerland). The primers were CHOP: forward 5′-GCGCATGAAGGAGAAAGAAC-3′, reverse 5′-CCAATTGTTCATGCTTGGTG-3′ BiP: forward 5′-TTTCTGCCATGGTTCTCACTAAAA-3′, reverse 5′-AACATTTAGGCCAGCAATAGTTCC-3′ GAPDH: forward 5′-ACCCACTCCTCCACCTTTGA-3′, reverse 5′-CTGTTGCTGTAGCCAAATTCGT-3′ GRP94 forward 5′-TGGGAAGAGGTTCCAGAATG-3′, reverse 5′-GTTGCCAGACCATCCGTACT-3′. For relative quantification, GAPDH mRNA served as a reference. Measured quantification cycles were analyzed according to Pfaffl,⁶³ comparing treated with untreated samples. Three biological replicates were run in triplicates each and means and standard deviations were calculated.

Western blot

HEK cells were grown to 60% confluence in DMEM with 10% FBS. Medium was changed to F10 with 15 μg/ml saponin and aminoglycoside antibiotics were added and cells were incubated for 24 h. Cells were lysed in hypotonic buffer (20 mM HEPES pH 7.5, 10 mM KCl, 3 mM Mg acetate, 1 mM DTT and 10 μg/ml DNase I) and ultrasonicated. Lysates were centrifuged (13 000 r.p.m., 10 min) and protein concentration in the supernatant was measured by the Micro BCA Protein Assay Kit (Thermo Scientific). Ten micrograms of total protein were resolved on 10% SDS-polyacrylamide gels and blotted on nitrocellulose membranes, which were probed with specific antibodies. Amersham ECL Prime Western Blotting Detection Reagent (RPN2232; GE Healthcare, Glattbrugg, Switzerland) was used as a substrate for the horseradish peroxidase (HRP). The specific antibodies used in this study were: anti-BiP antibody (Abcam, ab21685); anti-GRP94 antibody (Abcam, ab87886); anti-ATF4 antibody (Abcam, ab23760); anti-β-actin antibody (A1978-200UL; Sigma-Aldrich); HRP-conjugated goat anti-rabbit (Invitrogen, G-21234) and goat anti-mouse antibodies (Invitrogen, A10551).

UPR reporter assay

Reporter plasmids UPRE (p5xATF6-GL3)⁴⁴ and ERSE (pGL3-GRP78P(-132)-luc)⁴⁵ carrying luciferase under the control of UPR-specific cis-acting elements were kind gifts from Kazutoshi Mori (Kyoto University, Japan). HEK cells were grown to 60% confluence and co-transfected with reporter constructs and pGL4.75 (Rluc) using TurboFect reagent (Fermentas) according to the manufacturer's protocol. After a 24-h incubation, medium was replaced by F10 with 15 μg/ml saponin. Aminoglycosides were added and cells were incubated for another 24 h. Cells were lysed and luciferase activities determined. Normalization of luciferase activities was performed as described above. Cycloheximide was used as a mistranslation negative control and tunicamycin was used as a positive control for UPR.

P-eIF2α immunofluorescence assay

HEK cells grown on poly-D-lysine (Sigma Aldrich)-coated cover slips (Thermo Scientific) were treated for 24 h with geneticin in F10 medium with 15 μg/ml saponin, or with arsenite for 1 h (positive control). Cells were then fixed with 4% paraformaldehyde and methanol and incubated with blocking solution (1 × PBS with 1% BSA and 0.5% saponin) for 1 h at RT. Immunostaining used a rabbit polyclonal antibody against p-eIF2α (1 : 250) and a secondary goat anti-rabbit antibody conjugated with Texas Red (1 : 250) diluted in blocking solution. Cover slips were mounted on glass slides (VWR) using Dapi fluoromount (Southern Biotech, Birmingham, AL, USA), and cells were imaged using a Lyca Sp2 confocal microscope and a 63 × objective. p-eIF2α and nuclear signals were quantified using Imaris software (Bitplane, Belfast, UK) and the dots-per-cell ratio was calculated.

Animals

Male and female CBA/J mice were purchased from Harlan Sprague-Dawley Co. (http://www.harlan.com/products_and_services/research_models_and_services/research_models/cbaj_inbred_mice.hl) at an age of 6–8 weeks and bred in-house in order to obtain pups for organotypic cultures of the post-natal organ of Corti and SGCs (see next section). XBP1^+/− mice were from a stock kindly provided by Dr. Laurie H Glimcher and received via Dr. Randal J Kaufman, University of Michigan. Littermates served as wild-type (XBP1^+/+) controls. Animals were kept on a 12-h light/12-h dark cycle with free access to water and diet (Purina 5001, Lab Diet, St. Louis, MO, USA) and used in the in vivo studies at an age of 3–4 months. Experimental protocols were approved by the University of Michigan Committee on the Use and Care of Animals and animal care was under the supervision of the University of Michigan's Unit for Laboratory Animal Medicine.

Organotypic cultures of post-natal organ of Corti and SGCs

The procedures were as described previously.⁶⁴ Mice at post-natal day 2–3 (p2–3) were killed and cochleae dissected in cold Hank's Balanced Salt Solution to isolate the organ of Corti; SGCs were dissected from the base to the middle of the modiolus. Explants were placed onto a culture dish in 2 ml of medium consisting of basal medium Eagle, 1% serum-free supplement (Gibco #51500-056, Life Technologies, Grand Island, NY, USA), 1% bovine serum albumin, 5 mg/ml glucose and 10 U/ml penicillin G, allowed to settle for 4 h (37 °C, 5% CO₂) and incubated for 2 days to mitigate dissection stress. Medium was then exchanged for new media with or without drugs and incubation continued. For immunofluorescent labeling, explants were fixed with 4% paraformaldehyde overnight at 4 °C and permeabilized for 30 min with 3% Triton X-100 in PBS at RT. After three washes with PBS and blocking with 10% goat serum for 30 min at RT, incubation with the primary antibodies followed at 4 °C for 72 h. After three washes with PBS, secondary antibodies were applied (Alexa Fluor 488-conjugated goat anti-mouse and Alexa Fluor 546-conjugated goat anti-rabbit antibody; 1 : 200 in PBS) at 4 °C overnight in darkness. After several rinses, specimens were mounted on a slide with Prolong Gold anti-fade reagent (Life Technologies) and imaged with an Olympus Fluoview Confocal Laser Scanning Microscope-FV500 (Olympus America, Center Valley, PA, USA). For staining of hair cells, specimens were incubated at RT with rhodamine phalloidin (1 : 100) for 1 h; or for staining of nuclei with Hoechst 33342 (2 μg/ml in PBS) for 40 min. Presence or absence of OHCs and IHCs was determined on a Leitz Orthoplan microscope (Leica, Wetzlar, Germany) whose right objective had a 0.19-mm scale imposed on the field. Successive 0.19-mm fields were evaluated beginning at the apex by observers blinded to the experimental conditions. Cell counts were compared with a normative database (KHRI Cytocochleogram, version 3.0.6, Kresge Hearing Research Institute, University of Michigan, Ann Arbor, MI, USA).

Drug administration in vivo

Gentamicin was locally delivered as previously described.⁶⁵ Mice were anesthetized with an intraperitoneal injection of xylazine (7 mg/kg) and ketamine (90 mg/kg) and body temperature was maintained. The temporal bone was approached via a retro-auricular incision and a small hole was made in the thin part of the bulla with a 30-G needle. Surgical tubing was inserted through the hole, and 10 μl of 0.56 M gentamicin dissolved in saline was slowly injected. The skin incision was closed with tissue adhesive. TUDCA (Calbiochem, EMD Millipore) was dissolved in 0.15 M NaHCO₃ (adjusted to pH 7.4) and injected subcutaneously at 500 mg/kg body weight 6 days, 3 days and 3 h before gentamicin administration. Injections of 0.15 M NaHCO₃ served as vehicle controls. Injection of TUDCA did not cause any apparent side effects.

Hair cell counts in adult mice

Three weeks after injections, cochleae were fixed as described above. The apical bony capsule was removed and the cochlea decalcified in 4% sodium EDTA (adjusted to pH 7.4) for 7 days at 4 °C. Subsequently, cochleae were dissected into apical, middle and basal segments. Segments were permeabilized in 3% Triton X-100 for 30 min at RT, washed three times with PBS and incubated with rhodamine phalloidin (1 : 100) at RT for 1 h. The procedures for cell counting were the same as for explants.

Quantification of SGCs and synaptic ribbons

Following decalcification with 4% EDTA, cochleae were cryo-sectioned. Sections of 8 μm thickness were incubated in 0.3% Triton X-100 in PBS for 30 min at RT, blocked with 10% goat serum for 30 min, followed by incubation with anti-neuronal class III β-tubulin antibody (1 : 2000) for 48 h at 4 °C. After three rinses in PBS, sections were incubated with a secondary antibody (Alexa Fluor 546-conjugated; 1 : 500) at 4 °C overnight in darkness. After three washes with PBS, sections were stained with Hoechst 33342 (2 μg/ml in PBS) for 40 min at RT. After a final wash, the slides were mounted with Prolong Gold anti-fade reagent. Controls were processed without primary antibody. The number of SGCs in Rosenthal's canal was quantified using ImageJ software (National Institutes of Health, Bethesda, MD, USA) by counting the β-tubulin- and Hoechst-positive cells on images taken with an Olympus laser confocal microscope. Two mid-modiolar sections, separated by 40–50 μm, were used to obtain the average for each animal. For synaptic ribbon counts, cryosections were incubated for 30 min at RT with 5% donkey serum in PBS with 0.3% Triton X-100 and overnight in darkness at 4 °C with antibodies against CtBP2 (1 : 200) and Myo7a (1 : 200). After three washes in PBS (15 min each), tissues were incubated with secondary antibodies (Alexa Fluor 488- and Alexa Fluor 546-conjugated; 1 : 1000) at RT for 1 h. After three washes, the epithelia were mounted and images taken on an Olympus laser confocal microscope. Images were reconstructed three-dimensionally using Imaris software (Bitplane). The number of synaptic ribbons was quantified per IHC based on an average of 14 IHCs per sample.

Auditory function measurements

For ABRs, animals were anesthetized with an intraperitoneal injection of xylazine (7 mg/kg), ketamine (65 mg/kg) and acepromazine (2 mg/kg), and placed in a sound-isolated and electrically shielded booth (Acoustic Systems, Austin, TX, USA). Body temperature was maintained near 37 °C with a heating pad. Acoustic stimuli were delivered monaurally to a Beyer earphone attached to a speculum inserted into the left ear canal. Subdermal electrodes were inserted at the vertex of the skull, under the left ear and under the right ear (ground). ABRs were measured at 12 and 32 kHz using Tucker Davis Technology (TDT) System III hardware and SigGen/Biosig software (Tucker Davis Technology, Alachua, FL, USA) to present the stimuli (15 ms tone bursts, 1 ms rise-fall time) and record up to 1024 responses for each stimulus level. Thresholds were determined by reducing the intensity in 10-dB increments and then in 5-dB steps until no organized responses were detected. Threshold shifts were calculated for individual animals as the difference in auditory thresholds between ABR measurements before and at the end of the studies. For the DPOAE procedure, see the legend of Supplementary Figure S6.

Statistical analysis

Data were evaluated by one-way ANOVA followed by Tukey's honestly significant difference tests using JMP version 8.0.1 (SAS Institute Inc., Cary, NC, USA) or Student's t-test. All tests were two-sided with significance set at P<0.05.