Loss of Pex1 in Inner Ear Hair Cells Contributes to Cochlear Synaptopathy and Hearing Loss

Peroxisome Biogenesis Disorders (PBD) and Zellweger syndrome spectrum disorders (ZSD) are rare genetic multisystem disorders that include hearing impairment and are associated with defects in peroxisome assembly, function, or both. Mutations in 13 peroxin (PEX) genes have been found to cause PBD-ZSD with ~70% of patients harboring mutations in PEX1. Limited research has focused on the impact of peroxisomal disorders on auditory function. As sensory hair cells are particularly vulnerable to metabolic changes, we hypothesize that mutations in PEX1 lead to oxidative stress affecting hair cells of the inner ear, subsequently resulting in hair cell degeneration and hearing loss. Global deletion of the Pex1 gene is neonatal lethal in mice, impairing any postnatal studies. To overcome this limitation, we created conditional knockout mice (cKO) using Gfi1Cre or VGlut3Cre expressing mice crossed to floxed Pex1 mice to allow for selective deletion of Pex1 in the hair cells of the inner ear. We find that Pex1 excision in inner hair cells (IHCs) leads to progressive hearing loss associated with significant decrease in auditory brainstem responses (ABR), specifically ABR wave I amplitude, indicative of synaptic defects. Analysis of IHC synapses in cKO mice reveals a decrease in ribbon synapse volume and functional alterations in exocytosis. Concomitantly, we observe a decrease in peroxisomal number, indicative of oxidative stress imbalance. Taken together, these results suggest a critical function of Pex1 in development and maturation of IHC-spiral ganglion synapses and auditory function.


Introduction
Peroxisomal Biogenesis and Zellweger Spectrum Disorders (PBD-ZSD) are a group of rare autosomal recessive disorders caused by mutations in PEX genes, characterized by defective peroxisome assembly and function. Patients with PBD-ZSD display physiological, developmental, and neurological complications that include visual and sensorineural hearing loss (SNHL) [1][2][3]. Based on genotype-phenotype correlations, PBD-ZSD are classified as severe, intermediate, or mild within the PBD-ZSD spectrum [4]. Severe forms of the disease are associated with a complete loss of peroxisomal function. Infants with severe forms of PBD-ZSD display severe impairments at birth and are associated with a short life expectancy, typically under a year. Milder forms of the disease comprise missense mutations, are typically associated with milder slowly progressing phenotypes and lead to variable life expectancy up to adulthood [5]. While drugs have been developed that can partially alleviate the symptoms of the disease, there is currently no cure for PBD-ZSD.
In mammalian cells, peroxisomes play a key role in anabolic (biosynthetic) and catabolic (degradative) pathways [6]. These organelles are indispensable for lipid metabolism (i.e., biosynthesis of etherphospholipids, fatty acid alpha/beta-oxidation, bile acid and docosahexaenoic acid, DHA) but also may serve as protective organelles by playing a role in detoxification of reactive oxygen species (ROS) [7].
The biogenesis of peroxisomes depends on different peroxins, or PEX proteins, which play a role in the targeting of the peroxisomal membrane proteins (PMP) to the peroxisome membrane, import of peroxisomal matrix proteins and peroxisomal proliferation which occurs in response to external cues [8]. Mutations in thirteen PEX genes have been found to cause PBD-ZSD [4,[9][10][11]. Mutations in the PEX1 gene, encoding for the peroxisomal biogenesis factor 1, are the most common, found in nearly 70 percent of affected individuals [1,12,13]. The most common PEX1 mutation is hPEX1.G843D.
PBD-ZSD patients suffer from moderately severe to profound hearing loss [3]. Patients with the common G843D mutation in one of the PEX1 alleles along with a null allele (G843D/null) suffer from severe hearing loss with thresholds ranging 70 to 100 decibels (dB) sound-pressure level (SPL). Patients who possess homozygous mutations (G843D/G843D) suffer from moderate to profound hearing loss with threshold ranging 50 to 100 dB SPL [3].
SNHL most commonly arises from damage or loss of sensory hair cells, hair cellneuronal synapse or degeneration of neurons. Sensory hair cells and spiral ganglion neurons are particularly vulnerable to various stresses including oxidative stress. Even though a recent study demonstrated a non-canonical function of peroxisomes that leads to a better resistance of some cells to oxidative stress [14], we hypothesized that PEX1 mutations alter sensory hair cell stability and survival leading to progressive SNHL in patients partly due to the increase of oxidative stress, the severe decrease of plasmalogen synthesis [15] and defect in the import of antioxidant enzymes.
To assess how PEX1 mutations affect hair cells (HCs) we took advantage of different conditional knockout mouse models and assessed physiology and morphology of the auditory organ. Our work demonstrates that IHCs are vulnerable to the loss of Pex1 protein which leads to alteration of synapses and progressive hearing loss.

Animals
All animal experiments were performed in accordance with the NIH guidelines and were approved by the Institutional Animal Care and Use Committee (protocols #20-02-4149R and #00001240) at Boston Children's Hospital.
Male and female Pex1 fl/fl were obtained as a private strain from The Jackson Laboratory (JR 32722). These mice were generated from the Pex1-G844D mice (B6.Cg-Pex1 tm1.1Sjms /Mmjax; RRID:MMRRC_037405-JAXInfo) which include loxP sites flanking exons 12 and 13 ( Figure 1) and was edited to correct for the G844D mutation and bring it back to the wild type sequence. The Gfi1 Cre knock-in mice were generated by Dr. Lin Gan at the University of Rochester [16] and were provided for this study by Dr. Jian Zuo at St Jude Children's Research Hospital. These mice can only be bred with one copy of the Cre driver. Slc17a8-ires-Cre (referred to as VGlut3 Cre ) knock-in mice were obtained at Bradford Lowell/BIDMC [17]. Mice were maintained in a C57BL/6J background in our facility at Boston Children's Hospital. Mice of both sexes were used in similar numbers. In this case, the Cre insertion does not affect endogenous expression of VGlut3 and the mice can be bred to have two copies of the Cre allele. Breeding was carried out to obtain Cre expression under homozygous Pex1 floxed alleles with one copy of the Cre allele (Gfi1 Cre/+ ) for Gfi1 Cre expressing mice, and two copies (VGlut3 Cre/Cre ) for VGlut3 Cre expressing mice. Tissue collection was performed immediately post-mortem. Excision of exons 12 and 13, upon Cre recombination, leads to the production of a truncated protein (red vertical arrow). The Pex1 protein is composed of two N-terminal domains (N1 and N2) and two functional AAA domains (D1 and D2). The conserved Walker motifs domains into AAA domains are indicated (Walker A motifs: magenta and Walker B motifs: orange); (B) Breeding scheme for generating Gfi1 cre/+ Pex1 fl/fl mice line (Cre expressed in inner and outer hair cells around E15.5) and VGlut3 cre/cre Pex1 fl/fl mouse line (Cre expressed in inner hair cells around birth). Pex1 fl/fl mice were used as a control. Created with BioRender.com (Accessed date: 6 December 2022).

Genotyping
Genotyping was performed by Polymerase Chain Reaction (PCR) using GoTaq ® Master Mix (Promega #M7122) using primer sets included in Table 1. For the Pex1 fl/fl genotyping, we followed the protocol designed for the B6.Cg-Pex1 tm1.1Sjms /Mmja mice by The Jackson Laboratory. For the Gfi1 Cre genotyping, the PCR reaction was run in a thermocycler programmed for 94 • C for 3 min, [94 • C for 30 s, 61 • C for 1 min, 72 • C for 45 s] × 30 cycles, 72 • C for 5 min, and then maintained at 4 • C until further analysis. For the VGlut3 Cre genotyping, the PCR reaction was run in a thermocycler programmed for 94 • C for 5 min [94 • C for 1 min, 55 • C for 30 s, 72 • C for 1 min] × 30 cycles, 72 • C for 5 min, and then maintained at 4 • C until further analysis. A 1% agarose gel was used to analyze PCR products on iBright CL1500 imaging system.

Electrophysiology
ABRs (auditory brainstem recordings) and DPOAEs (distortion-product otoacoustic emissions) recordings were performed as previously described [18] from 5.6 to 32.0 kHz at 1 and 4 months on control Pex1 fl/fl , as well as Gfi1 Cre/+ Pex1 fl/fl ; and VGlut3 Cre/Cre Pex1 fl/fl mice. ABR wave I amplitude was determined from positive peak 1 (P1) to negative peak 1 (N1) and analyzed from 5.6 to 32.0 kHz at 1 and 4 months in control Pex1 fl/fl , as well as Gfi1 Cre/+ Pex1 fl/fl ; and VGlut3 Cre/Cre Pex1 fl/fl mice. ABR wave IV amplitude was determined from positive peak 4 (P4) to negative peak 4 (N4) and analyzed from 5.6 to 32.0 kHz at 1 and 4 months in control Pex1 fl/fl , and VGlut3 Cre/cre Pex1 fl/fl mice. ABR wave I and IV latency reflect the distance from the origin to positive peak 1 (P1) or positive peak 4 (P4), respectively.
Measurements of the resting membrane capacitance (cell size) of IHCs were obtained in whole-cell voltage-clamp configuration at −70 mV and after 2 min equilibrium of the internal patch-pipette recording solution with the IHC cytosol as described previously [19]. K + recordings were obtained in whole-cell voltage-clamp configuration with 20 ms and 500 ms depolarizing steps from −90 mV to +70 mV with a 10 mV increment. Membrane potentials were corrected offline for the residual series resistance Rs as described previ-ously [23]. Ca 2+ IV ramp curves were elicited in whole-cell voltage-clamp configuration with a depolarizing step from −80 mV to +10 mV increasing 1 mV every millisecond during 90 ms. Kinetics of the fast exocytosis of the readily releasable pool (RRP) of vesicles were evoked by depolarizing voltage-steps from −80 mV to −10 mV (corresponding to the maximum inward Ca 2+ current) with an increasing duration from 5 ms to 80 ms. Simple sustained exocytosis was obtained by 100 ms depolarizing voltage-steps from −80 mV to −10 mV. was used for visualization on a densitometric (iBright CL1500 imaging system). Each band was quantified using the 'Analyze > Gel' function on Fiji software. Each band was normalized to β-Actin level. Each experiment was carried out at least three times.

Immunostaining and Synapse Analysis
Immunostaining and synapse analysis were performed as described previously [24]. In brief, cochlea of 4 months control Pex1 fl/fl , Gfi1 Cre/+ Pex1 fl/fl ; and VGlut3 Cre/Cre Pex1 fl/fl mice were fixed with 4% PFA for 1 h at room temperature, decalcified, and dissected for wholemount processing. Tissues were permeabilized and blocked with normal horse serum (NHS) (5%, Jackson ImmunoResearch, ref #008-000-121) and triton (0.3%, Sigma Aldrich, ref #X100-5ML) for 1 h at room temperature. Then, cochleas were immuno-stained with anti-CtBP2 (pre-synaptic marker, IgG1 After PBS wash, cochleas were mounted with Vectashield antifade mounting medium containing DAPI (Vector laboratories, ref #H-1200-10). Images were acquired using the Zeiss LSM800 confocal microscope. Z-stacks were obtained from 5.6 to 32.0 kHz regions with 0.3 µm step. Frequency maps were designed by measuring the apex-to-base length by using the 'measure line' plugin on Fiji software. The number and the volume of ribbon synapses per IHC were done using the function 'spots' and 'volumes' on Imaris Cell imaging software (Oxford Instruments).

Data Analysis and Statistics
ABRs and ribbon synapses data were analyzed with Prism-GraphPad version 9 software or OriginPro 2022b software (OriginLab, Northampton, MA, USA). For Western blot analysis, results were analyzed with Prism-GraphPad version 9 software. For K + and Ca 2+ currents recordings, capacitance measurement in IHCs, and exocytosis analysis results were analyzed with OriginPro 2022b software (OriginLab, Northampton, MA, USA). Statistical analyses with two data sets were made by two-tailed unpaired t-tests or for comparisons of more than two data sets, one-way ANOVA or two-way ANOVA followed by a Tukey mean comparison test. All results are expressed as mean ± S.E.M.

Resources
All resources used for this project are summarized in Table 2.

Generation and Characterization of Two New Organ-Specific Pex1 Knockout Mouse Lines
We generated two novel mouse models with conditional deletion of Pex1 in the inner ear to determine how deletion of Pex1 proteins affects the auditory organ and specifically sensory hair cells (HCs). For this purpose, we took advantage of Pex1 floxed (Pex1 fl/fl ) mouse line developed at The Jackson Laboratory that possess the wild type Pex1 allele with LoxP sites localized in introns 11 and 13 ( Figure 1A). The expression of the Cre-recombinase leads to the excision of the exons 12 and 13. Pex1 fl/fl mice were crossed to HC-specific Cre expressing mice to conditionally delete Pex1 in HCs of the inner ear. We used two Cre-driver strains, Gfi1 Cre and VGlut3 Cre mice. Gfi1 Cre mouse line was used to excise Pex1 gene in all HCs at early stages (~embryonic stage (E) 15.5), while VGlut3 Cre mouse line was used to excise Pex1 gene specifically in IHCs around birth ( Figure 1B). Pex1 fl/fl mice were crossed to Gfi1 Cre mice to produce Gfi1 Cre/+ Pex1 fl/+ mice. These mice were further crossed to obtain Gfi1 Cre/+ Pex1 fl/fl mice. In parallel, Pex1 fl/fl mice were crossed to VGlut3 Cre mice to obtain VGlut3 Cre/+ Pex1 fl/fl mice (with one copy of the Cre-expressing allele) and VGlut3 Cre/Cre Pex1 fl/fl mice (with two copies of the Cre-expressing allele). The homozygous VGlut3 Cre mice were generated to increase the efficiency of the recombination. Contrary to homozygous Gfi1 Cre mice, VGlut3 Cre mice can be bred to possess two copies of the Cre allele which does not disrupt VGlut3 expression [17]. All genotypes were analyzed but only results for Pex1 fl/fl , Gfi1 Cre/+ Pex1 fl/fl , and VGlut3 Cre/Cre Pex1 fl/fl are presented in this report. Cre recombination in Gfi1 Cre/+ Pex1 fl/fl and VGlut3 Cre/Cre Pex1 fl/fl mice was validated by RT-qPCR (data not shown) and Western blot in liver samples, which also express Gfi1 and VGlut3 (Figure S1A-D; Table S16). No obvious changes in gross tissue morphology were detected in the inner ear of the conditional mouse models, which displayed normal hair cell organization, stereocilia bundle shape and length as observed on whole mount cochleas from Pex1 fl/fl , Gfi1 Cre/+ Pex1 fl/fl and VGlut3 Cre/Cre Pex1 fl/fl mouse lines (data not shown). Breeding of the lines took three generations to obtain the experimental Cre expressing mice in presence of floxed allele and absence of wild type Pex1 allele. This model provides a unique opportunity to analyze Pex1 gene function in the inner ear.

Gfi1
Cre/+ Pex1 fl/fl Mice Show Normal Hearing but Reduced Wave I Amplitude To determine how alteration in Pex1 expression affects HCs, we evaluated the auditory phenotype of the conditional Pex1 knockout model. To that end, we first assessed clickevoked auditory brainstem responses (ABRs) at two time points, 1 month and 4 months. The threshold corresponds to the lowest intensity of sound leading to a detectable wave form. At 1 month of age, we found similar thresholds in both Gfi1 Cre/+ Pex1 fl/fl and control Pex1 fl/fl mice, with an average of 44.58 ± 1.30 dB SPL (n = 12) and 44.17 ± 2.71 dB SPL (n = 6), respectively (Table S1; Figure 2A).
Similar results were observed at 4 months of age in Gfi1 Cre/+ Pex1 fl/fl and Pex1 fl/fl mice (Pex1 fl/fl : 45.63 ± 1.99 dB, n = 8; Gfi1 Cre/+ Pex1 fl/fl : 46.67 ± 1.86 dB, n = 9; Table S1; Figure 2B). ABRs were also assessed in response to pure tones, at frequencies ranging from 5.6 to 32.0 kHz. At 1 month of age, we observed a mild threshold elevation in the high frequency range (from 16.0 to 32.0 kHz) compared to control (Table S1; Figure 2C). At 4 months of age, no significant threshold difference was observed between Gfi1 Cre/+ Pex1 fl/fl and Pex1 fl/fl mice (Table S1; Figure 2D). Outer hair cell (OHCs) function was assessed by recording distortion products otoacoustic emissions (DPOAEs) in the control and conditional KO mice. Gfi1 Cre/+ Pex1 fl/fl mice had normal DPOAEs thresholds compared to Pex1 fl/fl mice at 1 and 4 months of age, suggesting there was no alteration of OHC function in the conditional mice (Table S2; Figure 2E,F). While these data suggest that Gfi1 Cre/+ Pex1 fl/fl mice have a normal auditory phenotype, analysis of the ABR waveforms revealed alterations of the response amplitude, affecting all peaks for the two time points tested ( Figure 2G,H). Wave I amplitude reflects sound evoked activity of the afferent fibers innervating the IHCs of the cochlea. Since wave I amplitude is a good indicator of the activity of the distal cochlear nerve, we measured wave I amplitude and latency in response to pure tone stimuli, at 1 and 4 months. To that end, we averaged the values for the three highest intensities tested (70, 80 and 90 dB SPL). We find that wave I amplitude (i.e., N1-P1) is significantly reduced at 1 and 4 months of age (Table S3; Figure 2I,J, p values between * p = 0.0293 and ** p = 0.0034).
We also performed wave I response growth analysis. For this analysis, we defined wave I slope by generating a linear regression curve of the wave I amplitude as a function of sound intensity. The data demonstrate significant change in slope values associated with the decrease in wave I amplitude in most of the frequencies tested in Gfi1 Cre/+ Pex1 fl/fl at 1 and 4 months (Figure S2A-L; Table S17; p value between * p = 0.0270 and *** p = 0.0005). In addition, wave I latency was slightly increased in Gfi1 Cre/+ Pex1 fl/fl mice at 1 and 4 months for all frequencies tested ( Figure Table S4; p value between * p = 0.0451 and ** p = 0.0024). Previous work has demonstrated that suprathreshold wave I amplitude is closely associated with synapse preservation in IHCs [25,26]. These results therefore suggest that IHCs might be further affected by the loss of Pex1 gene.  (Table S1; mean ± S.E.M. ns p = 0.9334, Mann-Whitney test) (A) and 4 months (Table S1; mean ± S.E.M. ns p = 0.9638, Mann-Whitney test) (B); (C,D) Pure-tone ABR recordings as a function of frequencies (kHz) on Pex1 fl/fl (grey) and Gfi1 cre/+ Pex1 fl/fl (green) mice at 1 month (Table S1;

VGlut3 Cre/Cre Pex1 fl/fl Mice Exhibit Mild Hearing Loss and Preservation of Neural Central Gain
Work with the Gfi1 Cre mouse line has previously revealed mosaic recombination in both IHCs and OHCs [27]. Furthermore, Gfi1 Cre mice have been shown to induce recombination in other cell types within the inner ear [28] and display progressive hearing loss. Thus, to specifically investigate the function of Pex1 in IHCs, we generated and analyzed auditory function of VGlut3 Cre/Cre Pex1 fl/fl mice. Similar to Gfi1 Cre/+ Pex1 fl/fl mice, on click-evoked recording, we noticed similar ABR thresholds at 1 and 4 months of age for VGlut3 Cre/Cre Pex1 fl/fl mice compared to control Pex1 fl/fl mice (at 1 month, Pex1 fl/fl : 45.29 ± 1.09 dB SPL, n = 17; VGlut3 Cre/Cre Pex1 fl/fl : 47.92 ± 1.56 dB SPL, n = 12-at 4 months, Pex1 fl/fl : 43.00 ± 1.53 dB SPL, n = 10; VGlut3 Cre/Cre Pex1 fl/fl : 46.11 ± 1.39 dB SPL, n = 10; Table S5; Figure 3A,B).
Interestingly, pure tones ABRs performed at frequencies ranging from 5.6 to 32.0 kHz show a progressive elevation of the threshold for all frequencies at both ages tested for VGlut3 Cre/Cre Pex1 fl/fl compared to control Pex1 fl/fl mice (Table S5; Figure 3C,D; p value between * p = 0.0382 and **** p = 0.00007). While DPOAEs were not affected (Table S6; Figure 3E,F), representative ABR waveforms were altered with significant changes in the amplitude of the response, affecting all peaks ( Figure 4A,B). We performed analysis of wave I amplitude and latency in response to pure tone stimuli, at 1 and 4 months in VGlut3 Cre/Cre Pex1 fl/fl compared to control Pex1 fl/fl mice, averaging values for the three highest intensities tested (70, 80 and 90 dB SPL). This analysis shows that wave I amplitude is strongly reduced for all frequencies tested at 1 and 4 months of age (Table S7; Table S18; p value between * p = 0.0283 and *** p = 0.0007). Wave I latency was significantly altered in VGlut3 Cre/Cre Pex1 fl/fl mice at 1 month and slightly increased at 4 months for all frequencies tested (Table S8, Figures 4E,F and S5A-L; p value between * p = 0.0390 and * p = 0.0146). Similarly, on clickevoked recording, the analysis of wave I in VGlut3 Cre/Cre Pex1 fl/fl mice revealed a progressive phenotype with a mild decrease (~25%) in ABR amplitude at 1 month of age (Pex1 fl/fl : 3.39 ± 0.31, n = 7; VGlut3 Cre/Cre Pex1 fl/fl : 2.55 ± 0.32, n = 7; Table S7; Figure 4G) and a severe reduc-tion (~60%) at 4 months (Pex1 fl/fl : 2.65 ± 0.17, n = 6; VGlut3 Cre/Cre Pex1 fl/fl : 1.05 ± 0.03, n = 7; *** p = 0.0002; Table S7; Figure 4H). Additionally, we observed a progressive increase of the wave I latency over age of VGlut3 Cre/Cre Pex1 fl/fl mice compared to the control mice (1 month: Pex1 fl/fl : 1.16 ± 0.01, n = 7; VGlut3 Cre/Cre Pex1 fl/fl : 1.17 ± 0.01, n = 7-4 months: Pex1 fl/fl : 1.12 ± 0.02, n = 6; VGlut3 Cre/Cre Pex1 fl/fl : 1.18 ± 0.04, n = 7; Table S8; Figure 4I,J). These findings suggest that deletion of Pex1 under VGlut3 promoter leads to mild hearing loss and has an impact on IHC function along the entire organ. Each ABR waveform is generated by the activation of anatomical sites along the auditory pathway in response to sound stimuli. While wave I corresponds to the activity of the distal part of the auditory nerve, wave IV is believed to mostly reflect a central component with activation of the superior olivary complex. To determine if the auditory abnormalities, previously noticed, were restricted to a dysfunction at the HC level, we analyzed wave IV amplitude and latency on click-evoked ABRs at 1 and 4 months of age in VGlut3 Cre/Cre Pex1 fl/fl mice. We found that wave IV amplitude was affected at both ages in VGlut3 Cre/Cre Pex1 fl/fl mice. However, the phenotype was more severe at 4 months in VGlut3 Cre/Cre Pex1 fl/fl mice with, respectively~22% and~38% wave IV amplitude decreased (at 1 month: Pex1 fl/fl : 2.31 ± 0.19, n = 7; VGlut3 Cre/Cre Pex1 fl/fl : 1.80 ± 0.21, n = 7-at 4 months: Pex1 fl/fl : 1.90 ± 0.22, n = 6; VGlut3 Cre/Cre Pex1 fl/fl : 1.18 ± 0.14, n = 7; * p = 0.0217; Table S9; Figure 4K,L). In contrast to 1 month, wave IV latency was slightly increased at 4 months of age in VGlut3 Cre/Cre Pex1 fl/fl mice (at 1 month: Pex1 fl/fl : 3.76 ± 0.04, n = 7; VGlut3 Cre/Cre Pex1 fl/fl : 3.74 ± 0.03, n = 7-at 4 months: Pex1 fl/fl : 3.47 ± 0.05, n = 6; VGlut3 Cre/Cre Pex1 fl/fl : 3.55 ± 0.07, n = 7; Table S10; Figure 4M,N). Central compensation for reduced cochlear input has previously been reported [29,30]. Compensation can be determined by analysis of wave IV/I ratio. To determine if compensation occurs in the VGlut3 Cre/Cre Pex1 fl/fl mice, we analyzed wave IV/I ratio at 1 and 4 months. At 1 month, a similar wave IV/I ratio was observed in VGlut3 Cre/Cre Pex1 fl/fl mice compared to control mice (Pex1 fl/fl : 0.73 ± 0.10, n = 7; VGlut3 Cre/Cre Pex1 fl/fl : 0.73 ± 0.09, n = 7 -Table S11, Figure 4O), meaning that no neural gain was observed at this stage, perhaps due to the limited reduction in wave I reported at 1 month (~25%). However, by 4 months of age, a significant increase (~50%, * p = 0.0288) of wave IV/I ratio was observed in VGlut3 Cre/Cre Pex1 fl/fl compared to control Pex1 fl/f mice (Pex1 fl/fl : 0.74 ± 0.10, n = 6; VGlut3 Cre/Cre Pex1 fl/fl : 1.11 ± 0.10, n = 7-Table S11, Figure 4P). Data show that VGlut3 Cre/Cre Pex1 fl/fl mice presented a disproportionally central response (larger wave IV amplitude) relative to the decreased wave I amplitude,~60% at this age. Collectively, these data show that VGlut3 Cre/Cre Pex1 fl/fl mice compensate centrally for the reduced cochlear input by generating central neural gain. This result suggests that negative impacts of the deletion of Pex1 under the VGlut3 promoter mainly occurs at the IHCs level.   (Table S5; mean ± S.E.M. ns p = 0.2243, Mann-Whitney test) (A) and 4 months (Table S5; mean ± S.E.M. ns p = 0.2962, Mann-Whitney test) (B); (C,D) Pure-tone ABR recordings as a function of frequencies (kHz) on Pex1 fl/fl (grey) and VGlut3 cre/cre Pex1 fl/fl (blue) mice at 1 month (Table S5;

Synaptopathy Is Observed in VGlut3 Cre/Cre Pex1 fl/fl
We hypothesized that the decrease in wave I amplitude in the VGlut3 Cre/Cre Pex1 fl/fl resulted from defects in the synapse. To investigate whether the deletion of Pex1 under the VGlut3 promoter was affecting the synaptic region, we first examined the number of ribbons in IHCs of control and mutant mice. To that end, we labeled the pre-synaptic ribbon using anti-CtBP2 antibody and the post-synaptic ribbon using anti-GluR2 antibody. We labelled IHCs body using anti-Myo7a antibody and the cuticular plate using phalloidin toxins ( Figure 5A ; Table S12; Figure 5E,F). In addition, the same percentage of paired ribbon synapses, defined by colocalization of CtBP2 and GluR2 staining, were observed in control and mutant mice, meaning that there were no orphan ribbons present (Table S19; Figure S6A Table S19; Figure S6C). Nevertheless, changes in pre-synaptic ribbon volume were observed. To quantify ribbon volume, we measured the volume of CtBP2 puncta at 4 months of age at 8.0 and 22.6 kHz regions in VGlut3 Cre/Cre Pex1 fl/fl and control Pex1 fl/fl mice using Imaris software. A significant decrease (p < 0.0001) in ribbon volume was observed in VGlut3 Cre/Cre Pex1 fl/fl compared to control   Table S13; Figure 5I,J).
Taken together, these results demonstrate that Pex1 disruption in HCs leads to disruption of ribbon synapses and alteration in transmission of the sensory signal at the distal end of the auditory nerve.

IHCs from VGlut3
Cre/Cre Pex1 fl/fl Mice Present Altered Exocystosis BK (K Ca1.1) channels are known to be highly sensitive to oxidative stress. Changes in fast-repolarizing BK channels have been associated with aging and disruptions in the hair cell synapse [19]. To determine if Pex1 disruption affects the functional maturation of IHCs in VGlut3 Cre/Cre Pex1 fl/fl mice, we recorded K + voltage dependent currents in P18-P21 mature IHCs from Pex1 fl/fl and VGlut3 Cre/Cre Pex1 fl/fl mice. No significant changes in conductance, kinetic and voltage-dependent activation of the K + currents were observed in VGlut3 Cre/Cre Pex1 fl/fl mice (Figure 6A-C; Table S14; two-way ANOVA, ns p = 0.65). We also assessed membrane capacitance at rest to determine if changes in cell sizes were associated with the conditional deletion of Pex1 and we observed no significant changes ( Figure 6D; Table S14; Pex1 fl/fl : 10.82 ± 0.35 pF-VGlut3 cre/cre Pex1 fl/fl :10.48 ± 0.31 pF; unpaired t-test, ns p = 0.47).
Next, we investigated changes in exocytosis and Ca 2+ current in mature P21-P25 IHCs from Pex1 fl/fl and VGlut3 Cre/Cre Pex1 fl/fl mice to determine if Pex1 disruption affects ribbon synapse function. Ca 2+ currents are crucial to trigger fusion of the synaptic vesicles in the IHC synaptic active zone. Depolarization of IHC activates Ca V 1.3 Ca 2+ channels. These channels are present in the synaptic active zone, forming tight clusters with ribbons that aggregate synaptic vesicles [31,32]. Whole-cell patch clamp recordings of IHCs demonstrate significant decrease of inward Ca 2+ currents in VGlut3 Cre/Cre Pex1 fl/fl mice ( Figure 6D; Table  S14; Pex1 fl/fl : 138.8 ± 6.3 pA-VGlut3 cre/cre Pex1 fl/fl : −118.4 ± 5.89 pA; unpaired t-test, * p = 0.025), with, however, no shift in the activation curve ( Figure 6D; Table S14; Pex1 fl/fl : 21.36 ± 0.67 mV-VGlut3 cre/cre Pex1 fl/fl : −21.27 ± 0.68 mV; unpaired t-test, ns p = 0.93).  (Table S14, n = 14 and 15, respectively in P18-P21 Pex1 fl/fl and VGlut3 cre/cre Pex1 fl/fl ; ns p = 0.65). IV curve is obtained with 20 ms depolarizing steps from −90 mV to +70 mV with an increment of +10 mV for each depolarizing step with IK,f measured at 2 ms from the onset of the depolarizing step; (B,C) Potassium current responses from Pex1 fl/fl IHCs (grey) and VGlut3 cre/cre Pex1 fl/fl (blue). Currents were elicited using 500 ms depolarizing steps from −90 mV to +70 mV with an increment of +10 mV for each depolarizing step; (D) Comparative Ca 2+ current-voltage curve (Table S14, IV; n = 17 and 21, respectively for P21-P25 Pex1 fl/fl in grey and VGlut3 cre/cre Pex1 fl/fl in blue). IV curve is obtained with a ramp protocol starting at −80 mV to 10 mV with an increase of 1 mV each millisecond. Values in the right graphs indicate the max inward Ca 2+ current and the V1/2 of the IV curve. The peak of the Ca 2+ current was significantly higher in Pex1 fl/fl IHCs compared to VGlut3 cre/cre Pex1 fl/fl IHCs (Table S14; unpaired t-test, * p = 0.025). While the Ca 2+ current is decreased there is no shift in the activation of the Ca 2+ channels of VGlut3 cre/cre Pex1 fl/fl IHCs (Table S14; unpaired t-test, ns p = 0.93). Insert graph in left: comparative mean IHC resting membrane capacitance, measured at −70 mV, in whole-cell patch clamp configuration (Table S14; (Table S14; n = 14 and 17, respectively in Pex1 fl/fl and VGlut3 cre/cre Pex1 fl/fl ; unpaired t-test, ** p = 0.006); (H) Exocytosis Ca 2+ efficiency is also decreased in VGlut3 cre/cre Pex1 fl/fl IHCs showing that the decreased in exocytosis is not only due to a decrease in calcium current (Table S14; n = 14 and 17, respectively in Pex1 fl/fl and VGlut3 cre/cre Pex1 fl/fl ; unpaired t-test, * p = 0.015).
Exocytosis is elicited by the fusion of the synaptic vesicles to the cell membrane followed by the released of the neurotransmitter in the synaptic cleft. The ribbon structure permits synaptic vesicles aggregation to the IHC active zone that are ready to fuse at the membrane called Readily Releasable Pool of vesicles (RRP). RRP vesicles are crucial for the temporal precision of a fast exocytosis [33,34]. Aggregation of multiple vesicles by the ribbon facilitates vesicles recruitment and permit a sustained exocytosis [35]. Kinetics of brief stimulations (from 5 to 80 ms) addressing the released of RRP vesicles was impaired by Pex1 disruption ( Figure 6E; Table S14; two-way ANOVA, *** p = 7.6 × 10 −8 ). Sustained exocytosis during a 100 ms voltage-step stimulation from −80 mV to −10 mV was also significantly reduced in IHC lacking Pex1 (Table S14, Figure 6F,G; Pex1 fl/fl : 26.93 ± 3.25 fF-VGlut3 cre/cre Pex1 fl/fl : 15.93 ± 2.11 fF; unpaired t-test, ** p = 0.006).
Since Ca 2+ influx and exocytosis are tightly coupled, reduced exocytosis may result from the decrease in Ca 2+ currents. However, we also noticed a significant decrease in the Ca 2+ efficiency of exocytosis in IHCs of VGlut3 Cre/Cre Pex1 fl/fl mice ( Figure 6H; Table S14; Pex1 fl/fl : −0.194 ± 0.02 fF/pA-VGlut3 cre/cre Pex1 fl/fl : −0.132 ± 0.014 fF/pA; unpaired t-test, * p = 0.015). These results demonstrate that Pex1 disruption leads to a defect in Ca 2+ influx, a defect in fast and sustained exocytosis and is also paired with a defect in the coupling between Ca V 1.3 channels and synaptic vesicles.

Pex1 Deletion Affects Peroxisomal Number
Peroxin proteins are indispensable for peroxisomal biogenesis. At the same time, alterations in peroxins involved in the import of peroxisomal matrix proteins (PMP), such as Pex1, has been shown to lead to pexophagy [36,37]. We hypothesize that peroxisomal number is severally affected in absence of Pex1 protein. To assess peroxisomal number, we performed western blot analysis of PMP70 expression in VGlut3 Cre/Cre Pex1 fl/fl liver tissue, a tissue rich in peroxisomes also expressing VGlut3. PMP70 is a major component of peroxisomal membranes and disruptions in PMP70 can pinpoint to aberrant peroxisomal assembly. Here, we show significant decrease (~33%, **** p < 0.0001) of PMP70 protein (peroxisomal membrane protein) at 1.5 months of age in VGlut3 Cre/Cre Pex1 fl/fl compared to control mice (Pex1 fl/fl : n = 4; VGlut3 Cre/Cre Pex1 fl/fl : n = 8; Table S15; Figure 7A,B). This result confirms that loss of Pex1 protein affects the stability of peroxisomes. A similar phenotype is likely recapitulated in HCs of the inner ear. Unfortunately, specific changes, in IHCs only, could not be determined by global protein assessment of inner ear tissues from VGlut3 Cre/Cre Pex1 fl/fl , likely due to the low number of hair cells over the total cell population expressing peroxisomal proteins in our samples.

Discussion
In this study we demonstrate the essential role of Pex1 protein in inner hair cells. While progressive sensory-neural hearing loss has been reported to be associated with PBD-ZSD [3], the etiology of the disease, in the ear, has remained unknown. Mutations in PEX1 gene are more commonly associated with PBD-ZSD. While prior studies have highlighted the function of Pex1 in the retina and liver [38][39][40], no studies have explored how disruption in peroxisomes affects the auditory system and leads to hearing dysfunction [37]. As Pex1 is ubiquitously expressed, total deletion of this gene is neonatal-lethal in mice, thereby preventing functional studies in young and adult animals. To overcome this limitation, we sought to develop two new conditional Pex1 mouse models. Given the high similarity of murine and human Pex1 genes (~80%), we hypothesized that early deletion of Pex1, specifically in inner ear tissue, would lead to progressive hearing loss as reported in PEX1 patients. We used two mouse models with deletions of Pex1 specifically in hair cells, Gfi1 Cre/+ Pex1 fl/fl (both IHCs and OHCs) and VGlut3 Cre/Cre Pex1 fl/fl (IHCs only). While DPOAEs were not altered, analysis of ABR thresholds and waveforms revealed changes in ABR thresholds from one to four months in VGlut3 Cre/Cre Pex1 fl/fl , and ABR wave I in both models. Our results demonstrate that IHCs are especially vulnerable to disruption on Pex1 expression. IHC are the primary sensory receptors of the auditory organ. These cells have been shown to be especially vulnerable to noise exposure and aging [25,[41][42][43]. Cochlear damage associated with noise and aging, however, typically starts with alterations in the IHC synapse with no change in ABR threshold but decrease in wave I amplitude [44]. Wave I is generated by the distal portion of the auditory nerve at the IHCs synapse. Decreases in wave I amplitude have been associated with defects in the synapse, so called synaptopathy. Here, we show that, while the number of ribbon synapses was unchanged, their volume was severely affected with a worsened phenotype in VGlut3 Cre/Cre Pex1 fl/fl . A similar result has been previously documented in a rodent model after noise exposure [45]. Electrophysiological recordings from IHCs of conditional KO mice also revealed functional alteration of the synapse with reduced exocytosis. Such phenotype, if also present in PEX1 patients, would predict high sensitivity to noise induced hearing loss along with progressive hearing loss. Progression while mild, was indeed observed in PEX1 patients [3].
Peroxisomes are small ubiquitous organelles involved in a variety of metabolic reactions, such as lipid biogenesis and reactive oxygen species (ROS) products detoxification. Previous work has demonstrated the essential function of peroxisomes against noise exposure, specifically through Pejvakin-mediated pexophagy (selective degradation of damaged peroxisomes) which protects auditory hair cells from oxidative stress [46]. Moreover, a previous study has shown that exposure to loud sounds leads to peroxisomal proliferation in cochlear hair cells [47]. This proliferation has been hypothesized to be a physiologically protecting response associated with increase in ROS exacerbated by loud sound exposure [47]. As peroxisomal proliferation cannot occur in absence of Pex1 protein, we suspect that this protective mechanism is severely impacted in Pex1 mutant mice. The balance between peroxisome biogenesis and degradation is crucial for redox cell homeostasis. Our work shows that IHCs and IHC/Spiral ganglion synapses are particularly vulnerable to alteration in Pex1 protein expression and peroxisomal biogenesis. As hair cells do not regenerate, protective mechanisms are crucial for their survival. As such, we predict that Pex1 mutant mice will display additional vulnerability to damaging sound exposure and aging.
Mechanistically, the loss of Pex1 protein is likely to impair recycling of Pex5 receptor, indispensable for the import of antioxidant enzymes, such as catalase. As such, Pex1 impairment is expected to lead to rapid accumulation of ROS products leading to activation of the pexophagy pathway [8,48,49]. Interestingly, 65% of PDB-ZSD phenotypes are associated with pexophagy [36]. As such, we propose that hearing loss in PEX1 patients is the result of oxidative stress imbalance, increase in pexophagy and, at least initially, IHCs dysfunction.
Some caveats in this study must be noted. The phenotype observed in the conditional mice, used for our study, does not completely phenocopy the hearing phenotype observed in PEX1 patients. This may be due to incomplete deletion of Pex1 gene as even low expression levels of Pex1 protein (~5-15%) can prevent severe forms of PBD-ZSD [50]. In our study,~15% of the full-length Pex1 protein remains in liver tissues which suggests incomplete penetrance of the Cre excision. It should also be noted that while previous studies have demonstrated strong expression of VGlut3 in IHCs, VGlut3 is also expressed in other cell types, including OHCs, albeit at a lower level [51][52][53]. Accordingly, we observed loss of OHCs in 4 months old VGlut3 Cre/Cre Pex1 fl/fl (data not shown). Furthermore, as peroxisomes are ubiquitous, we cannot exclude the role of other cell types/regions in the disease phenotype. The stria vascularis is a highly metabolic region of the inner ear that plays a role in generation of the endolymphatic potential and oxidative metabolism [54,55]. Intermediate cells of the stria vascularis derived from neural crest melanocytes are rich in peroxisomes [56] and largely implicated in the generation of the endocochlear potential. Disruption in stria function would have dire consequences to the function and survival of the sensory cells and would lead to severe hearing loss. Similarly, disruption in spiral ganglion neurons would also lead to auditory neuropathy and hearing loss.
In addition, a recent study investigating the role of another Pex gene in the ear, Pex3 (Pex3 tm1a mutant mice), similarly demonstrates mild progressive hearing loss in the high frequency range that is associated with a decrease in ABR wave I amplitude and synaptopathy with presence of orphan ribbon synapses and no change in ribbon number [57]. However, Sox10 Cre excision (targeting all inner ear cell type), in Pex3 tm1d mice, lead to worsened hearing phenotype across all frequencies, decrease of the ABR wave I amplitude along with synaptic defects that included presence of orphan ribbons and decrease of ribbon number. These results highlight the role of Pex3 in the inner ear and reinforce the notion that Pex1, as well as Pex3 genes, are indispensable for the maintenance and proper function of sensory hair cells of the inner ear.
In summary, this study demonstrates the essential role of Pex1 in HC development and function. The use of conditional Pex1 mouse models can be further explored in other cell types in the auditory organ and can also be used to explore pathophysiology associated with Pex1 disruption in other organs such as liver, kidney, or brain.
Supplementary Materials: The following supporting information can be downloaded at: https://www. mdpi.com/article/10.3390/cells11243982/s1. Figure S1: Characterization of new organ specific Pex1 knockout mice; Figure S2: Decrease of the wave I amplitude observed over age during the recording of auditory brainstem responses (ABRs) in Gfi1 cre/+ Pex1 fl/fl mice compared to control (Pex1 fl/fl ); Figure S3: Slight increase of the wave I latency measured on ABR recordings in Gfi1 cre/+ Pex1 fl/fl mice compared to control (Pex1 fl/fl ); Figure S4: Decrease of the wave I amplitude observed over age during ABR recordings in VGlut3 cre/cre Pex1 fl/fl mice compared to control (Pex1 fl/fl ); Figure S5: Slight elevation of the wave I latency measured on ABR recordings in VGlut3 cre/cre Pex1 fl/fl mice compared to control (Pex1 fl/fl ); Figure S6: Unchanged distribution of ribbon synapses along Modiolar/Pillar axis in VGlut3 cre/cre Pex1 fl/fl compared to control mice (Pex1 fl/fl ). Table S1 to S19 are also included.