Atoh1-CreER™ (Chow et al., 2006) and Prestin^CreERT2(Fang et al., 2012) mice were obtained from Dr. Suzanne Baker and Dr. Jian Zuo at St. Jude Children's Research Hospital (Memphis, TN), respectively. Pou4f3^loxP/loxP (stock # 10560) (Badea and Nathans, 2011) and ROSA26^{CAG−loxP−stop−loxP−tdTomato} (ROSA26^tdTomato) mice (stock # 7914; also called Ai14) (Madisen et al., 2010) were purchased from The Jackson Laboratory (Bar Harbor, ME). All genotyping was performed by Transnetyx, Inc. (Cordova, TN) and mice of both genders were used in the study. All animal work was performed in accordance with approved animal protocols from the Institutional Animal Care and Use Committee at Southern Illinois University School of Medicine.

CreER recombination was induced at different postnatal ages by intraperitoneal (IP) injections of tamoxifen (Sigma-Aldrich, St. Louis, MO) dissolved in corn oil. Injections of 3 mg/40 g body weight (b.w.) were given 20–24 h apart at postnatal day (P)0 and P1 or at P12 and P13. Adult mice were injected with 9 mg/40 g b.w. tamoxifen for 2 consecutive days (20–24 h between injections) at either 4 or 8 weeks of age. Cre-negative littermates injected with tamoxifen served as controls.

Temporal bones were post-fixed in 4% paraformaldehyde (Polysciences Inc, Warrington, PA) in 10 mM PBS for 2 h at room temperature. Samples from mice that were older than 1 week were decalcified in 120 mM EDTA (Sigma-Aldrich, St. Louis, MO) for 1–5 overnights (depending on the age of mouse at tissue collection) at room temperature (RT) using an end-over-end rotator and the EDTA solution was changed daily. Whole mount dissection and immunostaining was performed as previously described (Montgomery and Cox, 2016) with the one exception that the samples were incubated with signal enhancer (cat #I36933; Life Technologies, Waltham, MA) for 30 min at RT prior to the blocking/permeabilization step. The following primary antibodies were used: mouse anti-beta III tubulin (Tuj1, 1:500, Biolegend, cat # 801201), rabbit anti-myosin VIIa (1:200; cat. #25-6790; Proteus Biosciences, Ramona, CA), mouse IgG1 anti-Pou4f3 (1:300; cat. #sc-81980; Santa Cruz Biotechnology, Dallas, TX), and goat anti-Sox2 (1:400, cat. #sc-17320; Santa Cruz Biotechnology, Dallas, TX). Alexa-conjugated secondary antibodies (Thermo Fisher Scientific, Hampton, NH) were used at 1:1,000 and nuclei were labeled with Hoechst 33342 (1:2,000 in 10 mM PBS, cat. #H3570 Thermo Fisher Scientific, Hampton, NH). tdTomato was visualized using endogenous fluorescence. TUNEL staining was performed using the In Situ Cell Death Detection Kit, TMR Red (cat. #12156792910; Roche Applied Science, Indianapolis, IN) following the manufacturer's instructions.

Post-fixed temporal bones were washed three times in 10 mM PBS and decalcified in 120 mM EDTA (Sigma-Aldrich, St. Louis, MO) for 1–2 weeks at room temperature (RT) using an end-over-end rotator. EDTA solution was changed daily. The decalcified cochleae were embedded in optimal cutting temperature (OCT) media (cat #4585, Fisher Healthcare, Houston, TX) in cryo-molds placed on a slurry of dry ice in 70% ethanol. The embedded cochleae were stored at −80°C until sectioning. For analyses of SGNs, mid-modiolar cryo-sections (12 μm thickness) were obtained using an Epredia Microm HM525 NX cryostat (Thermo Fisher Scientific, Hampton, NH) as previously described (Coleman et al., 2009).

Samples were imaged using a Zeiss LSM800 (Oberkochen, Germany) confocal microscope and processed using Zen 2.5 lite software (Oberkochen, Germany). Myosin VIIa-positive HCs and Sox2-positive SCs were manually quantified from two representative 150 μm regions from each cochlear turn (apex, middle, and base) from cochlear whole-mounts. Tuj1-positive neuronal cell bodies from cochlear sections were quantified in a 10,000 μm² region in the middle turn of cochleae from three alternate sections per sample. For quantification, cochlear whole mounts or sections were imaged using a 40× oil immersion objective with a numerical aperture (NA) 1.3 and a resolution of 1,024 × 1,024 pixels. The low magnification imaging of cochlear sections was done using a 10× objective with NA 0.3 and a resolution of 1,024 × 1,024 pixels. The N values represent number of mice as only one cochlea per mouse were analyzed for each immunostaining experiment type.

Mice aged 4 weeks or older were anesthetized with avertin (250–500 mg/kg, IP) and kept on a heating pad at 37°C in a sound attenuated chamber. Auditory brainstem response (ABR) measurements were performed using an Intelligent Hearing System (IHS, Miami, FL) 4964 high frequency system using the left ear of each mouse. Therefore, N values represent the number of mice. Subdermal stainless-steel electrodes were inserted at the vertex of the skull, below the pinna of the left ear, and a ground electrode was located at the base of the tail. ABR waveforms were obtained in response to 8, 12, 16, and 22 kHz tones (5 ms tone bursts presented at a rate of 19/s and averaged over 512 presentations) given in 5 dB SPL steps decrements between 80 and 5 dB SPL. ABR thresholds were determined by the lowest sound intensity that produced a visually distinct response in wave I and II as assessed by a researcher who was blinded to the genotypes.

The cochlear portion of temporal bones were collected and snap frozen on dry ice and kept at −80°C until further processed for RNA isolation. Both cochleae from each mouse were pooled for RNA isolation hence N values for qPCR analysis represent the number of mice. RNA isolation and real-time quantitative polymerase chain reaction (qPCR) was performed as previously described (McGovern et al., 2018) using SYBR Green (cat #K0391, Thermo Fisher Scientific, Hampton, NH) and a CFX Connect Optics Module (Bio-Rad, Hercules, CA). The primers used for each gene are listed in Supplementary Table 1. Comparison of gene expression levels was determined using the Pfaffl method which includes the primer efficiency in the delta-delta Ct equation (Pfaffl, 2001).

All data are presented as mean ± SEM. Comparisons of ABR thresholds between genotypes and across frequencies were done using a two-way ANOVA followed by Bonferroni-corrected post-hoc testing at each frequency. HC counts from control samples across time for each dataset were tested for similarity using a one-way ANOVA which revealed no significant differences in HC numbers across controls. Thus, we used the mean number of HCs from the latest timepoint assessed for each age of Pou4f3 deletion as the control samples in Figures 1–4 and to calculate the percentage of HC loss. Percent HC loss was calculated by subtracting the HC counts from an individual sample (control or Pou4f3 cKO) from the mean HC number of the control group (N = 3, from the latest timepoint assessed for each age of Pou4f3 deletion) and dividing that number by the mean HC number of the control group. The result was then converted to a percentage. For each age of Pou4f3 deletion, a two-way ANOVA was used to assess HC loss where control samples were treated as a pseudo-pre-injection timepoint. This allowed for analysis of the main questions of interest (effect of Pou4f3 deletion, effect of time post-injection, and effect of cochlear turn) without having to test three-way interaction terms which were not the focus of these studies. Significant main effects were followed by two Tukey's post-hoc tests. One post-hoc test compared each timepoint with control as well as to other timepoints, but within each cochlear turn and the second post-hoc test compared cochlear turns within each timepoint. Note that IHCs and OHCs were compared separately for the neonatal deletion of Pou4f3 in Figure 1. The number of SCs were compared across genotype and across cochlear turn using a two-way ANOVA followed by Sidak-corrected post-hoc comparisons. The percentage of SGN loss was calculated in a similar manner as percent HC loss using the mean SGN number from control samples as the denominator. The percentage of SGN loss was compared across genotypes using a two-way ANOVA followed by Sidak's post-hoc test. qPCR data are presented in the graphs as fold change from control samples and each gene was compared to its own control using a Student's t-test. Statistical analyses were conducted using GraphPad Prism 7.0 software (La Jolla, CA).

To delete Pou4f3 from HCs at birth, we used Atoh1-CreER^TM mice which show CreER activity in the majority of both IHCs and OHCs after tamoxifen induction at P0 and P1 (Chow et al., 2006; Weber et al., 2008; Cox et al., 2014). We validated this expression pattern using Rosa26^tdTomatoreporter mice, and found that 98%−100% of OHCs expressed tdTomato (100% ± 0% in apex, 98.5% ± 1.5% in middle, and 99.3 ± 0.7% in basal turn, N = 4, Supplementary Figure 1) and 92–100% of IHCs expressed tdTomato (100% ± 0% in apex, and 93.0% ± 6.2% in middle, and 93.7% ± 5.4% in basal turn, N = 4, Supplementary Figure 1).

HCs in the mouse cochleae are immature at birth and undergo various morphological and electrophysiological changes during the first postnatal weeks (Anniko, 1983; Eatock and Hurley, 2003). To determine the role of Pou4f3 in immature HCs, we generated Atoh1-CreER^TM:Pou4f3^loxP/loxP mice (referred to hereafter as Atoh1-Pou4f3cKO) and injected them with tamoxifen at P0 and P1, followed by assessment of auditory function using auditory brainstem response (ABR) measurements at 4 weeks of age (Figure 1A). Controls were CreER-negative littermates which contained the Pou4f3^loxP/loxP allele and also received tamoxifen. Compared to controls, Atoh1-Pou4f3cKO mice had significant ABR threshold elevations (~35–40 dB SPL) at all four frequencies tested (Figure 1B). Analysis of the cochleae showed rapid and progressive loss of both IHCs and OHCs in Atoh1-Pou4f3cKO mice (Figures 1D–G) compared to controls (Figures 1C–C″,F–G) which had the normal pattern of four rows of HCs with POU4F3 present in all HC nuclei (Figures 1C–C″). In Atoh1-Pou4f3cKO mice, OHC loss was observed as early as 5 days post-tamoxifen with more than 50% loss in all cochlear turns (64.8 ± 4.2% in apex; 55.9% ± 1.2% in middle, and 58.4% ± 15.9% in base, N = 3), which progressed to >95% OHC loss by 4 weeks post-tamoxifen (95.2% ± 1.0% in apex, 96.7% ± 2.1% in middle, and 97.5% ± 1.3% in base, N = 3; Figure 1F, Supplementary Table 2A). IHC loss was also observed at 5 days post-tamoxifen in Atoh1-Pou4f3cKO mice, but IHCs died at a faster rate than OHCs with ~94%−98% loss by 1 week post-tamoxifen (98.2 ± 1.8% in apex; 96.6 ± 2.3% in middle; 93.9% ± 3.5% in base, N = 3; Figure 1G, Supplementary Table 2B).

HC maturation occurs over the first two postnatal weeks leading to the onset of hearing in mice at ~P12–14 (Ehret, 1976). Since Atoh1 expression is down-regulated within the first postnatal week, we instead used Prestin^CreERT2mice which specifically target OHCs of the cochlea (Fang et al., 2012; Cox et al., 2014). We validated the expression pattern of Prestin^CreERT2 using Rosa26^tdTomato reporter mice and found that >90% of OHCs expressed tdTomato in all three turns of the cochlea after tamoxifen injections at P12 and P13, 4 weeks of age, or 8 weeks of age (Supplementary Figure 2).

To determine whether mature OHCs still require Pou4f3 for survival, we generated Prestin^CreERT2::Pou4f3^loxP/loxP mice (Prestin-Pou4f3cKO), injected them with tamoxifen at P12 and P13 (juvenile age), and performed similar assessments as outlined above (Figure 2A). Controls were CreER-negative littermates which contained the Pou4f3^loxP/loxP allele and also received tamoxifen. Prestin-Pou4f3cKO mice also showed significant ABR threshold elevations (~33–43 dB SPL) at all four frequencies tested compared to CreER-negative control littermates when tested at 4 weeks after deletion of Pou4f3 (Figure 2B). Compared to controls, we observed progressive loss of OHCs in all three turns of Prestin-Pou4f3cKO cochleae between 1 and 6 weeks post-tamoxifen (Figures 2C–F). However, IHCs remained intact with all expressing nuclear POU4f3 (Figures 2D,E″). This was expected since Prestin^CreERT2 targets OHCs exclusively. Comparison of OHC loss across cochlear turns in Prestin-Pou4f3cKO mice at 1 week post-tamoxifen showed greater loss in the basal turn (84.9 ± 2.9%) than the middle (47.6 ± 16.8%) and apical turns (24.2 ± 10.1%; Figure 2F, Supplementary Table 3). Yet, OHC loss progressed in all cochlear turns during the following weeks resulting in almost complete absence of OHCs in the middle and basal turns (97.5 ± 1.7% in middle, and 97.2 ± 1.5% in base), but only ~60% loss in the apical turn (61.5 ± 5.3%) at 6 weeks post-tamoxifen (Figure 2F, Supplementary Table 3).

Next, we investigated the role of Pou4f3 in OHCs when mice have mature hearing (at 4 weeks of age), or after they were sexually mature at (8 weeks of age) (reviewed in Ohlemiller et al., 2016). For both experiments, we again used Prestin-Pou4f3cKO mice to delete Pou4f3 from OHCs and performed ABR measurements and histological analyses of the cochlea at multiple timepoints (Figures 3A, 4A). Controls were CreER-negative littermates which contained the Pou4f3^loxP/loxP allele and also received tamoxifen. After Pou4f3 deletion at 4 weeks of age, we observed significant elevation of ABR thresholds in Prestin-Pou4f3cKO mice at all frequencies tested at 2 weeks post-tamoxifen compared to their control littermates (Figure 3B). One week after Pou4f3 deletion, OHC numbers in the apical and middle turns of Prestin-Pou4f3cKO cochleae were not different from controls, but there was significant and robust OHC loss (85.6 ± 2.9%) in the basal turn (Figures 3C–F). Two weeks after Pou4f3 deletion, OHC loss was significantly higher in all cochlear turns compared to controls with the highest loss observed in the basal turn (85.8 ± 7.7%) followed by the middle turn (55.2 ± 13.4%), but there was only a mild loss (15.1 ± 3.4%) in the apex. OHC loss in apical and middle turns continued to progress over time with the middle turn reaching >80% loss (85.8 ± 7.7%) at 4 weeks post-tamoxifen and the apical turn reaching similar loss level (82.1 ± 5.4%) at 6 weeks post-tamoxifen (Figure 3F, Supplementary Table 4).

Finally, we deleted Pou4f3 from OHCs at 8 weeks of age using Prestin-Pou4f3cKO mice to assess the role of Pou4f3 in sexually mature mice and to test whether Pou4f3 is required for OHC survival throughout the lifespan (Figure 4A). Similar to Pou4f3 deletion at all other ages, we observed significant elevation in ABR thresholds in Prestin-Pou4f3cKO mice compared to control mice at 2 weeks post-tamoxifen at all frequencies tested (Figure 4B). However, OHC loss was delayed compared to Pou4f3 deletion at other ages (Figures 4D–F). There was no significant difference in the number of OHCs in Prestin-Pou4f3cKO cochleae compared to controls at 1 week post-tamoxifen. However, at 2 weeks post-tamoxifen a majority of the OHCs were missing in the basal turn (84.5 ± 9.8%; Figure 4F). By 4 weeks post-tamoxifen, OHC loss progressed to the middle (81.0 ± 17.4%) and apical (58.1 ± 15.8%) turns as well (Figure 4F, Supplementary Table 5).

Next, we sought to investigate the mechanism of cochlear HC death caused by Pou4f3 deletion. Using cochleae from Atoh1-Pou4f3cKO mice injected with tamoxifen at P0 and P1 and Prestin-Pou4f3cKO mice injected with tamoxifen at 4 weeks of age, we performed the terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assay to detect cells undergoing apoptosis. In Atoh1-Pou4f3cKO samples where CreER targeted both OHCs and IHCs, we observed TUNEL-positive HCs of both types at 5 days post-tamoxifen (Figures 5B′,B″). Similarly, 1 week after Pou4f3 deletion, we observed many TUNEL-positive OHCs in Prestin-Pou4f3cKO cochleae (Figures 5D′,D″). No TUNEL-positive HCs were observed in any of the control samples (Figures 5A–A″,C–C″).

To further investigate potential mediators of cell death in response to the deletion of Pou4f3, we performed real-time qPCR of pro-apoptotic genes and two known POU4F3 target genes using the same mouse models and collection timepoints as used for the TUNEL assay (Atoh1-Pou4f3cKO mice were injected with tamoxifen at P0 and P1 and samples collected 5 days post-tamoxifen; Prestin-Pou4f3cKO mice were injected with tamoxifen at 4 weeks of age and samples collected at 1-week post-tamoxifen). After deletion of Pou4f3 we observed the downregulation of two downstream targets of Pou4f3, Gfi1 (p = 0.0003 for Atoh1-Pou4f3cKO, and p = 0.0009 for Prestin-Pou4f3cKO at 4 weeks, N = 9–10) and Lhx3 (p = 0.0427 for Atoh1-Pou4f3cKO, N = 10; Figures 5E,F) (Hertzano et al., 2004, 2007). Five days after neonatal Pou4f3 deletion using Atoh1-Pou4f3cKO mice, we also observed a significant upregulation of the pro-apoptotic gene, Bak1 (p = 0.0398; Figure 5E, N = 9–10). One week after Pou4f3 deletion at 4 weeks of age using Prestin-Pou4f3cKO mice, we observed upregulation of the pro-apoptotic gene Bax (p = 0.0059, N = 8). Together these data suggest that HCs undergo apoptotic cell death after deletion of Pou4f3.

Previous studies have shown that HC loss can produce secondary effects on the organ of Corti which includes loss of supporting cells and/or delayed SGN loss (Leake et al., 1997; Izumikawa et al., 2008; Kujawa and Liberman, 2009; Barclay et al., 2011; Taylor et al., 2012; Yu et al., 2014). In addition, previous studies showed a loss of SGNs when Pou4f3 was deleted from the germline (Xiang et al., 2003; Pauley et al., 2008). Here, we investigated whether HC loss caused by Pou4f3 deletion impacts these two cell types. Using Atoh1-Pou4f3cKO mice injected with tamoxifen at P0 and P1 and Prestin-Pou4f3cKO mice injected with tamoxifen at 8 weeks of age, we collected samples at 16 weeks post-tamoxifen and performed immunostaining to detect SOX2, a supporting cell-specific marker in the mature cochlea (Figures 6A–D) (Hume et al., 2007). Quantification of SOX2-positive supporting cells revealed no significant differences from control samples for both ages of Pou4f3 deletion (Figures 6E,F).

To investigate the impact of Pou4f3 deletion from HCs on the SGNs, we performed mid-modiolar sectioning of the temporal bones from the same two mouse models at 16 weeks (or 4 months) post-tamoxifen and stained them with the neuron specific anti-Tuj1 antibody (Lee et al., 1990; Barclay et al., 2011; Sun et al., 2018). While the CreER-negative control samples showed normal densities of SGN cell bodies (Figures 7A,C,E), cochleae from mice where Pou4f3 was deleted from both IHCs and OHCs at P0/P1, had 85.3 ± 1.1% loss of SGN cell bodies (Figures 7B,D,K). However, when the deletion was OHC-specific and occurred at 8 weeks of age, there was a 38.9 ± 10.7% loss of SGNs (Figures 7F,K). Additionally, we observed reduction in the neuronal fibers in Pou4f3cKO cochleae (Figures 7H,J) compared to control cochleae (Figures 7G,I) at both ages of Pou4f3 deletion.