Both adult wild-type and Tg(brn3c:mGFP)^s356t transgenic zebrafish and their spawning embryos were raised in embryo medium (EM) at 28.5°C according to standard protocols (Detrich et al., 1999). Ages of embryos are given as days post-fertilization (dpf). The Animal Care and Use Committee of Fudan University approved for all of the zebrafish experiments.

Neomycin sulfate (Sigma-Aldrich) was added at a final concentration of 400 μM, and the 5 dpf larvae were incubated for 1 h at 28.5°C. The larvae were allowed to recover at 28.5°C followed by three rinses in fresh EM. The H3K9 dimethylation inhibitors BIX01294 (Sigma-Aldrich) and UNC0638 (Sigma-Aldrich) were dissolved in DMSO both at stock concentrations of 10 mM and then diluted to the final concentrations in fresh EM. The neomycin-damaged larvae were incubated with medium containing BIX01294 (2, 5, 10 or 20 μM) or UNC0638 (10 μM) as the experimental groups, and the same concentrations of DMSO were used as controls. BIX01294, UNC0638, or DMSO was added to the larvae immediately after the neomycin was washed away, and the larvae were incubated at 28.5°C for 24 h or 48 h prior to analysis.

BrdU (Sigma) was used to label the proliferating cells. After 1 h neomycin treatment, the larvae were incubated in fresh EM with 10 mM BrdU for 24 h or 48 h at 28.5°C.

FM1-43FX dye (Molecular Probes) was used to observe and image the functional HCs within the neuromasts. FM1-43FX dye can enter the mature HCs through mechanotransduction channels. FM1-43FX dye at a final concentration of 3 μM was added to live 5 dpf larvae for 45 s and then rinsed three times with fresh water. The larvae were anesthetized in 0.02% MS-222 and fixed with 4% PFA overnight at 4°C.

For immunohistochemistry analysis, the fixed larvae were washed several times with PBT-2 (PBS containing 0.5% Triton X-100) for 30 min at room temperature then incubated in blocking solution for 1 h at 37°C. The primary antibodies were anti-Myosin-VI (1:200 dilution; Proteus BioSciences), anti-GFP (green fluorescent protein; 1:1000 dilution; Abcam), anti-dimethyl H3K9 (1:1000 dilution; Abcam), anti-Sox2 (1:200 dilution; Abcam), and anti-cleaved caspase-3 (1:500 dilution; Cell Signaling Technology Inc., Danvers, MA, USA). The larvae were incubated with primary antibodies overnight at 4°C and then washed three times with PBT-2 and incubated with secondary antibodies (1:200 dilution; Invitrogen) for 1 h at 37°C. The larvae were then incubated with 4,6-diamidino-2-phenylindole (DAPI; Invitrogen) for 20 min at room temperature to label the nuclei.

Immunohistochemistry staining was also used to detect the BrdU incorporation. To denature the DNA before antibody staining, the fixed larvae were washed three times with PBT-2 and then treated with 2 N HCl for 30 min at 37°C. The larvae were then incubated in blocking solution (PBT-2 with 10% normal donkey serum) for 1 h at 37°C and then incubated with anti-BrdU antibody (1:200 dilution; Santa Cruz Biotechnology) overnight at 4°C. The larvae were washed three times with PBT-2 and incubated with secondary antibody (1:200 dilution; Invitrogen) for 1 h at 37°C, and specimens were examined with a Leica confocal fluorescence microscope (TCS SP8; Leica).

Total protein was isolated from each of 10 larvae with the AllPrep DNA/RNA/Protein Mini Kit (QIAGEN, Hilden, Germany) according to the manufacturer’s instructions. The protein concentrations were determined using a bicinchoninic acid (BCA) protein kit (Thermo, Fisher Scientific, Rockford, IL, USA). The proteins were separated on sodium dodecyl sulfate (SDS)-polyacrylamide gels and transferred onto polyvinylidene difluoride (PVDF) membranes (Immobilon-P; Millipore, Bedford, MA, USA) after electrophoresis. The membranes were blocked with 5% non-fat dried milk in Tris-buffered saline with Tween 20 (TBST; 50 mM Tris-HCl (pH 7.4), 150 mM NaCl, and 0.1% Tween-20) for 1 h at room temperature and then blotted overnight with anti-dimethyl H3K9 (1:1000 dilution; Abcam) at 4°C with β-actin as control. After rinsing three times (10 min each) with TBST, the membranes were incubated with secondary antibody (1:1000 dilution; Invitrogen) for 1 h at room temperature. The immunoreactive bands were visualized using an ECL kit (Pierce). To quantify the relative levels of the protein, the intensities of the bands were quantified with Image J.

Cells in the first five posterior lateral line neuromasts (L1–L5) were counted. All statistical analyses were performed with IBM SPSS (version 21 for Mac) and GraphPad Prism (version 6 for Mac). Data were analyzed using Student’s t-tests (with two comparisons) or one-way ANOVA (with multiple comparisons). All data are presented as mean ± SEM. p < 0.05 was considered statistically significant, and p < 0.001 was considered highly significant.

Regular whole-mount in situ hybridization of zebrafish embryos was performed as previously described (Thisse and Thisse, 2008). All primers for probe synthesis are listed in the Table 1. Embryos were mounted in 100% glycerol.

We first examined the level of H3K9me2 in neomycin-damaged 5 dpf zebrafish after exposure to 20 μM BIX01294 for 48 h. H3K9me2 immunostaining was performed on transgenic line Tg(brn3c:mGFP)^s356t in which differentiated HCs express green fluorescence. Using 20 μM BIX01294 did not increase the death of zebrafish after neomycin damage compared with the controls, and both the 20 μM BIX01294-treated group and the DMSO-treated control group were morphologically normal, suggesting that the larvae did not suffer any drug toxicity during exposure to BIX01294 (Figure 1A). Consistent with previous reports, pharmacological inhibition of G9a/GLP with BIX01294 significantly decreased the H3K9me2 levels in neuromasts compared to controls with or without neomycin damage (Figures 1B–E). Moreover, the significantly decreased H3K9me2 level upon the addition of 20 μM BIX01294 was confirmed by semi-quantitative western blotting analysis (Figure 1F). We also observed fewer HCs when exposed to BIX01294 for 48 h after 1 h neomycin damage in comparison with the neomycin-treated controls (Figures 1D2–E3).

To test whether BIX01294 could modulate the proliferation and regeneration of HCs after HC loss, we quantified the changes in the numbers of newly regenerated HCs in the 5 dpf Tg(brn3c:mGFP) zebrafish lateral line 24 h and 48 h after different doses of BIX01294 treatment. Compared with DMSO controls, both the 24 h and 48 h groups treated with BIX01294 showed significant decreases in the number of GFP-positive HCs (Figures 2A1–A4). There was no significant difference between controls and the 2 μM BIX01294-treated group after 24 h; however, after 48 h co-incubation the difference was significant between DMSO controls and the 2 μM BIX01294-treated group (p < 0.001). We did not find any significant dose-dependent difference between 5 μM and 10 μM BIX01294 treatment groups after 24 h and 48 h co-incubation, while 20 μM BIX01294 treatment induced fewer GFP-positive HCs in both the 24 h and 48 h groups without increasing cell death compared with controls (p < 0.001; Figure 2B).

To further confirm our findings, we also used Myosin-VI immunostaining to specifically label and quantify the HCs of the wild-type larvae. The numbers of HCs in neuromasts L1–L5 were counted in 6–9 fish at both time points. As Figures 2C1–C4, 2C1–C4, D show, treatment with BIX01294 for 24 h after neomycin damage for 1 h resulted in a significant decrease in the number of Myosin-VI⁺ HCs compared to controls (p < 0.001), and the decrease was even greater compared to controls at 48 h (p < 0.001). Both groups showed dose-dependent decreases except for the 5 μM and 10 μM groups after 48 h treatment. All of these results validated the observation that normal HC regeneration was severely impaired in the presence of the G9a/GLP inhibitor BIX01294. Based on the dose-dependent effect analysis of GFP-positive HCs in the Tg(brn3c:mGFP) zebrafish lateral line and the Myosin-VI staining of HCs in the wild-type larvae, we chose the 20 μM BIX01294 treatment as the experimental group in the following experiments.

To test the functionality of the newly regenerated HCs, the wild-type larvae were stained with the vital dye FM1-43FX, which is a marker of functional mechanotransduction channels in HCs. We found significantly fewer FM1-43FX-positive HCs in the BIX01294-treated group than in controls at both 24 h and 48 h after 1 h neomycin treatment (Figures 3A,B), which is consistent with our findings in Tg(brn3c:mGFP) zebrafish and Myosin-VI staining in wild-type zebrafish.

To verify our results and to eliminate the possibility of off-target effects of BIX01294, we treated the 5 dpf neomycin-damaged Tg(brn3c:mGFP) line and the wild-type larvae with another potent and selective G9a/GLP inhibitor UNC0638 for 24 h and 48 h. Similar decreases in HCs were detected by immunostaining of anti-GFP, FM1-43FX, and anti-Myosin-VI in the 10 μM UNC0638 treatment groups compared to corresponding parallel DMSO controls, and this confirmed that inhibition of G9a/GLP with BIX01294 and UNC0638 reduces HC regeneration after neomycin damage in the zebrafish lateral line neuromasts (Figures 4A–F).

In the zebrafish lateral line, 400 μM neomycin exposure for 1 h killed almost all of the mature HCs, and this was followed by the rapid regeneration of HCs (Harris et al., 2003). During regeneration, most of the replacement HCs were regenerated from proliferative cells labeled with BrdU (Ma et al., 2008). To further explore whether BIX01294 treatment affects cell proliferation during HC regeneration, we counted the number of BrdU⁺ cells in L1–L5 neuromasts at 24 h and 48 h following 400 μM neomycin treatment. As shown in Figures 5A1–B1,E, exposure to 20 μM BIX01294 for 24 h led to significant decreases in the number of BrdU⁺ cells per NM compared to controls (the control larvae harbored 15.35 ± 0.260 BrdU⁺ cells/NM and the BIX01294-treated larvae harbored 9.42 ± 0.178 BrdU⁺ cells/NM, p < 0.001). In the 48 h treatment, there were even greater decreases in BrdU⁺ cells per NM (control larvae harbored 21.78 ± 0.383 BrdU⁺ cells/NM and the BIX01294-treated larvae harbored 9.65 ± 0.161 BrdU⁺ cells/NM, p < 0.001; Figures 5C1–D1,E). Thus, BIX01294 significantly inhibited the proliferation of NM cells after neomycin-induced HC loss.

To investigate the mitotic regeneration of HCs, we co-stained the larvae with Myosin-VI and BrdU antibodies. We found fewer Myosin-VI/BrdU double-positive cells at 24 h in the BIX01294-treated neuromasts compared with the controls (Figures 5A2–B2). Furthermore, when comparing the ratio of Myosin-VI/BrdU double-positive cells to total HCs, there was a significant difference between the control and BIX01294-treated groups (the ratio in the controls was 0.47 ± 0.017, and in BIX01294-treated groups the ratio was 0.15 ± 0.026, p < 0.001; Figure 5F). At 48 h, the effect was even greater (the ratio in the controls was 0.55 ± 0.021, and in BIX01294-treated groups the ratio was 0.15 ± 0.018, p < 0.001; Figures 5C2–D2,F). These findings suggest that BIX01294 significantly decreases the mitotic regeneration of HCs after neomycin-induced HC loss in zebrafish neuromasts.

Each NM contains HCs and SCs, and the proliferation of SCs enables the replacement of HCs after damage. Thus, we next evaluated the effect of BIX01294 treatment on SCs after neomycin exposure. Sox2 is highly expressed in the SCs of zebrafish neuromasts, and to quantify the number of SCs we stained the larvae with a Sox2 antibody and counted the Sox2⁺ SCs in L1–L5 neuromasts of 8–12 fish at 24 h and 48 h post-treatment. We found significant reductions in the number of Sox2⁺ SCs at both time points after BIX01294 treatment (control larvae harbored 28.75 ± 0.545 and 34.25 ± 0.434 Sox2⁺ SCs per NM at 24 h and 48 h, respectively, and BIX01294-treated larvae harbored 22.87 ± 0.303 and 23.52 ± 0.248 Sox2⁺ SCs per NM at 24 h and 48 h, respectively, p < 0.001; Figures 5A3–D3,G).

Subsequent co-staining with Sox2 and BrdU antibodies was performed to detect the proliferation rates of NM SCs after neomycin damage at 24 h and 48 h. Consistent with the results described above for the mitotic regeneration of HCs, we found significant decreases in the ratio of Sox2/BrdU double-positive cells to total SCs after BIX01294 treatment at both 24 h and 48 h (the ratio in the controls was 0.36 ± 0.010 and 0.48 ± 0.008 at 24 h and 48 h, respectively, and in BIX01294-treated groups the ratio was 0.22 ± 0.008 and 0.27 ± 0.006 at 24 h and 48 h, respectively, p < 0.001). As the major source of newly regenerated HCs after neomycin damage, SCs were significantly affected by BIX01294 treatment, which could cause a further decrease in HC regeneration (Figures 5A4–D4,H).

Our previous research confirmed the protective effect of BIX01294 on apoptosis from aminoglycoside-induced HC loss (Yu et al., 2013), suggesting that the decreased HC regeneration with BIX01294 might be mainly due to the inhibition of cell proliferation but not to increased cell death. To confirm our hypothesis, we used an antibody against cleaved caspase-3 to detect cell death during the HC regeneration process. We failed to find cells that labeled with cleaved caspase-3 both in the BIX01294 treatment groups and the controls in the zebrafish lateral line neuromasts (Supplementary Figure 1). Moreover, we failed to observe any significant difference in the apoptosis index (the number of apoptosis bodies/total number of DAPI cells) between the BIX01294-treated group and the control group. These results suggest that BIX01294 did not increase cell death in the lateral line neuromasts of zebrafish compared to the controls and that the inhibition of cell proliferation, but not apoptosis, was the main cause of decreased HCs in the course of regeneration.

Wnt and Fgf signaling play important roles in the development and regeneration of HCs in the zebrafish lateral line (Wright and Mansour, 2003; Munnamalai and Fekete, 2013). In this study, we investigated whether BIX01294 treatment reduced HC regeneration by disrupting the regulation of Wnt and/or Fgf signaling in the NM. At 24 h after neomycin treatment, the mRNA levels of wnt4a, wnt10a, fgf10a, and fgf3 were significantly up-regulated compared with the non-neomycin treatment group, indicating that the Wnt and Fgf pathways were activated in the control larvae. In the two groups without neomycin-induced damage, the mRNA levels of wnt4a, wnt10a, fgf10a, and fgf3 were of not significantly different between the control and BIX01294-treated larvae, suggesting that BIX01294 had no effect on the mature HCs. We found a striking down-regulation of Wnt ligand (wnt4a and wnt10a) and Fgf ligand (fgf3 and fgf10a) expression after exposure to 20 μM BIX01294 for 24 h after neomycin treatment compared with the neomycin-treated controls (Figure 6, Supplementary Figure 2). We next investigated the expression of representative downstream targets of Wnt and Fgf signaling in the NM. We also found a significant decrease in the expression of the Wnt signaling downstream target gene axin2 in BIX01294-treated neuromasts after neomycin damage, and the expression of the Fgf signaling-dependent gene pea3 was also strongly reduced in BIX01294-treated neuromasts compared with the controls after neomycin treatment. We examined the expression of the Fgf receptor fgfr1 in the regenerated neuromasts and found a similar down-regulation of fgfr1 expression in the BIX01294-treated group after neomycin damage. Furthermore, in situ hybridization data showed that the mRNA expression of atoh1a, a HC marker, was also significantly reduced in the BIX01294-treated groups after neomycin exposure (Figure 6). Taken together, these results suggest that BIX01294 treatment significantly down-regulated the Wnt and Fgf signaling pathways, which could lead to the inhibition of cell proliferation and HC regeneration after neomycin-induced HC loss in the zebrafish lateral line.

To confirm our hypothesis that the decreased HCs in the course of regeneration were mainly due to the down-regulation of the Wnt and Fgf signaling pathways, we increased Wnt and FGF signaling with the addition of 6-bromoindirubin-3-oxime (BIO; 1 μM) and bFGF (2 ng/ml) to the co-incubations with DMSO or BIX01294 for 48 h following 1 h neomycin treatment.

We found a significant increase in the number of HCs by staining of Myosin-VI after the addition of bFGF and BIO compared to BIX01294 treatment zebrafish (the mean number of HCs in the neuromasts of the neomycin + BIX01294 + bFGF + BIO group was 9.28 ± 0.124 compared to 6.05 ± 0.119 in the BIX01294-only treatment group, p < 0.001). The addition of bFGF and BIO also increased HCs in the DMSO control (the control larvae harbored 9.40 ± 0.149 Myosin-VI⁺ HCs per NM, and the neomycin + bFGF + BIO treatment group harbored 11.25 ± 0.220 HCs per NM, p < 0.001). There was no significant difference in the number of HCs per NM between the neomycin + BIX01294 + bFGF + BIO group and the controls (p > 0.05; Figures 7A,B). These data together suggest that BIX01294 treatment decreased HCs mainly due to down-regulating the Wnt and Fgf signaling pathways, which can be rescued by the addition of the effectors of the Wnt and Fgf signaling pathways.