Lgr5-EGFP-IRES-creERT2 mice (Stock #008875, Jackson Laboratory) and Rosa26-tdTomato reporter mice (Stock #007914, Jackson Laboratory) of either sex were used in the experiments (Pannier et al., 2009). We performed all animal procedures according to protocols that were approved by the Animal Care and Use Committee of Southeast University and were consistent with the National Institute of Health's Guide for the Care and Use of Laboratory Animals. We made all efforts to minimize the number of animals used and to prevent their suffering.

The tail tips were collected from transgenic mice, and genomic DNA was obtained by adding 180 μl 50 mM NaOH, incubating at 98°C for 60 min, and adding 20 μl 1M Tris-HCl (PH 7.0). The genotyping PCR was carried out by using 2 × Tag Master Mix (Vazyme), and the PCR protocol was as follows: 94°C for 3 min; 37 cycles of 94°C for 30 s, 60°C for 30 s, and 72°C for 45 s; 72°C for 5 min; and holding at 4°C. The genotyping primers were as follows: Lgr5 (F) CTG CTC TCT GCT CCC AGT CT, wild-type (R) ATA CCC CAT CCC TTT TGA GC, mutant (R) GAA CTT CAG GGT CAG CTT GC; tdTomato wild-type (F) AAG GGA GCT GCA GTG GAG T, (R) CCG AAA ATC TGT GGG AAG TC; mutant (F) GGC ATT AAA GCA GCG TAT C, (R) CTG TTC CTG TAC GGC ATG G.

Heterozygous Lgr5-EGFP-creERT2 mice were crossed with homozygous Rosa26-tdTomato mice to trace the fate of Lgr5+ cells in the neomycin-damaged and undamaged cochleae. Postnatal day (P)1 mice were sacrificed, and the cochleae from Lgr5-EGFP-creER/Rosa26-tdTomato double-positive mice were dissected out and cultured in DMEM/F12 medium supplemented with N2 (1:100 dilution, Invitrogen), B27 (1:50 dilution, Invitrogen), heparin sulfate (50 ng/ml, Sigma), and the growth factors bFGF (10 ng/ml, Sigma), EGF (20 ng/ml, Sigma), and IGF-1 (50 ng/ml, Sigma) (full medium). The cochleae were treated with 500 nM 4OH-tamoxifen for 4 days all through the culture. At 12 h after the beginning of the culture, the cochleae were treated with 0.5 mM neomycin (Sigma) or PBS for 12 h. EdU was added to the medium at a final concentration of 10 μM to label dividing cells. The damaged and undamaged cochleae were examined after 4 days of culture.

Approximately 30–40 postnatal day (P)1–2 Lgr5-EGFP-creERT2 mice were sacrificed, and the cochleae were dissected out and cultured in full medium as described above and allowed to recover for a few hours. The cochleae were treated with 0.5 mM neomycin (Sigma) or PBS for 12 h and then allowed to recover in full medium for 24 h. The cochleae were collected and trypsinized by prewarmed 0.125% trypsin/EDTA (Invitrogen) at 37°C for 8 min. The same amount of soybean trypsin inhibitor (10 mg/ml, Worthington Biochem) was then added to terminate the trypsin reaction in the neomycin-damaged and undamaged cochlear samples. Cochleae were separated into single cells by pipetting up and down 80–100 times with blunt tips and then percolating through a 40 μm cell strainer (BD Biosciences). Dissociated cells from damaged and undamaged cochleae were sorted on a BD FACS Aria III using the GFP channel.

Total RNA was extracted from ~20,000 FACS-sorted neomycin-treated Lgr5+ progenitors (NLPs) and 20,000 untreated Lgr5+ progenitors (ULPs) with an RNeasy micro kit (QIAGEN). RevertAid First Strand cDNA Synthesis Kit (Thermo) was used to synthesize cDNA. Real-time PCR was carried out by using the SYBR Green PCR Master Mix (Roche) on a BIO-RAD C1000 Touch thermal cycler (BIO-RAD). Each 25 μL PCR reaction mixture contained 12.5 μL 2 × SYBR Green PCR Master Mix, 0.5 μL forward primer (10 μM), 0.5 μL reverse primer (10 μM), 2 μL template, and 9.5 μL sterilized distilled water. Each group contained three samples, and each PCR was carried out in triplicate. The PCR protocol was as follows: 50°C for 2 min; 95°C for 10 min; 45 cycles of 95°C for 15 s, 60°C for 1 min; and a melting curve was performed starting at 65 up to 95°C with an increase of 0.5°C per 1 s to verify primer specificities. Expression levels of each gene was normalized to the GAPDH in the same samples. The primers were listed in Table 1.

Neomycin-damaged and undamaged cochleae were fixed in 4% PFA for 1 h at room temperature, washed with PBS, blocked with blocking solution (5% donkey serum, 0.5% Triton X100, 0.02% sodium azide, and 1% bovine serum albumin in pH 7.4 PBS) for 1 h at room temperature and then incubated with primary antibodies diluted in PBT1 (2.5% donkey serum, 0.1% Triton X100, 0.02% sodium azide, and 1% bovine serum albumin in pH 7.4 PBS) at 4°C for overnight. This was followed by washing with 0.1% (v/v) Triton X100 in pH 7.4 PBS three times and incubating with fluorescence-conjugated secondary antibody for 1 h at room temperature. After washing with 0.1% (v/v) Triton X100 in pH 7.4 PBS three times, the cochleae were mounted in antifade fluorescence mounting medium (DAKO). Anti-Myosin7a (Proteus Bioscience, #25-6790, 1:1,000 diluted in PBT1) and anti-Sox2 (Santa Cruz Biotechnology, #17320, 1:400 diluted in PBT1) primary antibodies were used. Donkey anti-rabbit Alexa Fluor 555 and 647 fluorescence-conjugated secondary antibodies (Invitrogen, #A-31572, #A-31573) were used for Myo7a, and donkey anti-goat Alexa Fluor 647 fluorescence-conjugated secondary antibody (Invitrogen, #A-21447) was used for Sox2. All the fluorescent secondary antibodies were diluted 1:400 in PBT2 (0.1% Triton X100 and 1% bovine serum albumin in pH 7.4 PBS). The Click-it EdU imaging kit (Invitrogen) was used after blocking to measure cell proliferation. The fluorescence images were obtained with a Zeiss LSM 710 confocal microscope and were analyzed using ImageJ (NIH) and Photoshop CS5 (Adobe Systems).

Approximately 20,000 NLPs and 20,000 ULPs were used to extract total RNA with an RNeasy micro kit (QIAGEN). The RNA samples from NLPs and ULPs were split into three fractions for separate replicates.

The double-strand cDNA was synthesized from the total RNA obtained from the NLPs and ULPs using a TruSeq® RNA LT Sample Prep Kit v2 (Illumina). Illumina adapters were ligated to the cDNA molecules after end repair. The ligated cDNA was cleaned up with AmpureBeads (Beckman). The library was amplified using 10 cycles of PCR for the enrichment of adapter-ligated fragments. Transcriptome sequencing was carried out with the Illumina-Hiseq2500 system (Illumina).

The TopHat (version 1.3.2) and CuffLinks (version 2.2.1) pipeline was used for the alignment and gene expression counting of the RNA-seq data. The reference genome was mm9. The FPKM (Fragments per kilobase of exon per million fragments mapped) values of all mouse genes were summarized together for all samples (three NLP samples and three ULP sample). Spearman's rank correlation was calculated for all pair-wise combinations of samples based on the FPKM values of all mouse genes. The correlation plot was generated with the corrplot package in R. A total of 46,983 mouse RefSeq transcripts were included in the RNA-seq data, and the means and standard deviations of the normalized data were calculated. A value of p < 0.05 was considered statistically significant. The expression levels of all of the transcriptional units were measured according to their FPKM values, and a cutoff level of 0.1 was chosen as the lowest gene expression level. The Gene Ontology (GO) analysis was done with DAVID GO Annotation. The protein-protein interaction information was extracted from the STRING database. The interaction network graph was drawn by Cytoscope 3.4. Important gene lists of different signaling pathways (Wnt, Notch, TGFβ, Hippo) were determined based on information from the KEGG database.

For each condition, at least three individual experiments were conducted. Data are presented as mean ± standard errors of the means (SEM), and GraphPad Prism6 software was used to analyze the data. Statistical significance was determined using a two-tailed, unpaired Student's t-test. A value of p < 0.05 was considered statistically significant.

Lgr5+ progenitors can generate HCs in the neonatal mouse cochlea both in vivo and in vitro (Madisen et al., 2010; Chai et al., 2012). Here we performed an in vitro lineage-tracing experiment by crossing Lgr5-EGFP-creER mice with the Rosa26-tdTomato reporter strain (Pannier et al., 2009). P1 Lgr5-EGFP-creER/Rosa26-tdTomato double-positive mouse cochleae were dissected out and cultured in full medium with 500 nM 4OH-tamoxifen to lineage trace the Lgr5+ progenitors. The cochleae were damaged by neomycin as described in the Section Materials and Methods (Figure 1A). We found that significantly more tdTomato/Myo7a double-positive HCs were generated from NLPs compared to ULPs in all three turns of the cochlea (Figures 1B–D, p < 0.05, n ≧ 4), suggesting that the Lgr5+ progenitors generated significantly more HCs after neomycin injury in vitro.

To determine the capacity of Lgr5+ progenitors in the damaged and undamaged cochleae to mitotically regenerate HCs, EdU was added to the culture medium from day 0 to day 4 of the culture (Figure 2A). Consistent with previous reports, there were no tdTomato+/EdU+ cells in the undamaged cochleae (Figures 2B,C). In contrast, tdTomato+/EdU+ cells, which represent the mitotically proliferated Lgr5+ progenitors, could be found in the damaged cochleae (Figure 2C), indicating that neomycin treatment induced the proliferation of Lgr5+ progenitors. However, due to the very small number of tdTomato+/EdU+ cells in neomycin-treated cochleae, the increase was not significant compared to the control group (Figure 2B, p = 0.093, n ≧ 5).

P1 Lgr5-EGFP-creER mice were sacrificed, and their cochleae were dissected out, cultured in full medium for 12 h, and then treated with 0.5 mM neomycin for 12 h to damage the HCs. The cochleae were allowed to recover for another 24 h before trypsinization and cell sorting (Figures 3A,B). After cell sorting, 20,000 isolated NLPs and 20,000 ULPs were collected and RNA-seq analysis was performed to determine their gene-expression profiles (supplementary Data Sheet 1). Principal component analysis was performed to assess the reproducibility of the measurements, and the NLP and ULP groups were well-separated by principal component 1 (Figure 3C). After excluding FPKM values below 0.1, 20,362 and 17,123 transcripts were examined separately in the NLPs and ULPs, respectively, and 14,877 transcripts were expressed in both cell populations (Figure 3D).

To determine the expression profiles of the richly expressed genes in NLPs and ULPs, the expression levels and abundance rankings of the most abundantly expressed genes were analyzed. Figure 4A shows the expression levels for the top 200 most abundant transcripts in ULPs (blue bars). The expression levels (red bars) and the abundance rankings (red numbers) of the same transcripts in NLPs are also illustrated for comparison. Similarly, Figure 4B shows the 200 most abundant transcripts in NLPs (red bars) compared to expression levels (blue bars) and abundance rankings (blue numbers) of the same transcripts in ULPs. As shown in both figures, most of the transcripts that were abundantly expressed in NLPs were also abundantly expressed in ULPs. However, Gm10800, Net1, Gm28438, Nr4a1, Krt18, Ler2, and Dpysl2 (NLP rank > 1,000) were only richly expressed in ULPs, and Cdkn1a, Ccng1, and Suco (ULP rank > 1,000) were only richly expressed in NLPs.

In order to characterize the genes that are significantly differentially expressed in NLPs and ULPs, we selected the differentially expressed genes in NLPs and ULPs by comparing their expression levels (fold change > 2.0, p < 0.05). Figure 5A shows an overall picture of the expressed transcripts in NLPs and ULPs. We found 549 genes that were significantly upregulated and 1,817 genes that were significantly downregulated in the NLPs. Figures 5B,C show the top 150 differentially expressed genes in ULPs and NLPs. Among these differentially expressed genes, the functions of some genes have been reported previously. Fgfr3 (Bermingham-McDonogh et al., 2001; White et al., 2012), Egfr (Saleem and Siddiqui, 2015), Frem2 (Nadol et al., 2015), Alms1 (Oshima et al., 2007b; Jagger et al., 2011), and Lif (Marzella et al., 1999; Su et al., 2015) were upregulated in NLPs, while Hes1, Hes5 (Zheng et al., 2000; Zine et al., 2001; Li et al., 2008; Murata et al., 2009; Abdolazimi et al., 2016), Hey1 (Tateya et al., 2011; Korrapati et al., 2013; Benito-Gonzalez and Doetzlhofer, 2014; Petrovic et al., 2015), HeyL (Kamaid et al., 2010), Id1, Id2, and Id3 (Ozeki et al., 2005; Jones et al., 2006; Laine et al., 2010) were downregulated in NLPs. We did not find any functional reports for the other differentially expressed genes in the cochleae, and these should be further studied in the future.

Cells in the postnatal mammalian cochlea have exited the cell cycle, and they have very limited capacity for proliferation. In order to promote mitotic HC regeneration, it is important to induce HC progenitors to re-enter the cell cycle and to mitotically regenerate HCs. In the present study, we have demonstrated that neomycin injury could induce the proliferation of Lgr5+ progenitors; however, the detailed mechanism behind this proliferative ability remains unclear. It has been reported that some of the cell cycle genes play important roles in the cochlea. To identify the possible genes regulating the cell cycling of Lgr5+ progenitors, we examined the expression levels of cell cycle genes in NLPs and ULPs. We found that Cdkn1a, Mdm2, Tfdp1, and Wee1 were significantly upregulated in NLPs and that Ccne2, Gadd45g, Nek2, Sfn, and Stmn1 were significantly downregulated in ULPs (Figure 6A). Real-time PCR was also performed to confirm the RNA-seq results, and these two results were consistent (Figure 6D). Only the roles of Cdkn1a (Laine et al., 2007; Laos et al., 2017) and Mdm2 (Mahmoodian Sani et al., 2016) in the inner ear have been described, and there are no reports of the roles of the other cell cycle genes we identified in ULPs and NLPs.

Transcription factors (TFs) are able to bind to enhancer or promoter regions of their downstream target genes and control their expression levels. There are many TFs involved in inner ear development and HC regeneration. In the present study, we have demonstrated that NLPs have significantly greater HC regeneration capacity compared to ULPs (Figure 1, p < 0.05, n ≧ 4). However, the roles of a large number of TFs in the inner ear and in HC regeneration are unknown. To determine the TFs that might be involved in HC regeneration from Lgr5+ progenitors, we compared the expression levels of TFs in the mouse genome between NLPs and ULPs. Figure 6B shows the 88 significantly differentially expressed TFs in NLPs and ULPs (fold change > 2, p < 0.05). Some of the TFs that were downregulated in NLPs, including Hes1, Hes5 (Zheng et al., 2000; Li et al., 2008; Murata et al., 2009; Abdolazimi et al., 2016), Hey1 (Tateya et al., 2011; Korrapati et al., 2013; Benito-Gonzalez and Doetzlhofer, 2014; Petrovic et al., 2015), HeyL (Kamaid et al., 2010), Id1, Id2, and Id3 (Ozeki et al., 2005; Jones et al., 2006; Laine et al., 2010), have been reported to play roles in negatively regulating HC fate and patterning regulation during inner ear development (Figure 6B). Real-time PCR was also performed to confirm the RNA-seq results, and these two results were consistent (Figure 6E). However, a significant number of the differentially expressed TFs have not been characterized in the inner ear before and need to be further studied in the future.

MicroRNAs (miRNAs) are untranslated RNAs that control gene expression by binding to target mRNAs. A few miRNAs have been reported to play important roles in HC protection and HC regeneration (Jen et al., 1997; Li et al., 2010; Wang et al., 2010; Patel and Hu, 2012). We found that 149 miRNAs were uniquely expressed in ULPs, 151 miRNAs were uniquely expressed in NLPs, and 59 miRNAs were expressed in both ULPs and NLPs. Among these miRNAs, eight miRNAs were significantly differentially expressed in NLPs and ULPs (p < 0.05, fold change > 2; Figure 6C). Mir466i was upregulated in NLPs, while Mir7007, mmu-mir-703, Mir107, Mir361, Mir6918, Mir6982, and Mir3099 were downregulated in NLPs. These miRNAs have not been characterized in the inner ear and need to be further studied in the future.

A few signaling pathways have been shown to be involved in inner ear development and HC regeneration. To determine which pathways might be involved in regulating HC regeneration from Lgr5+ progenitors, we compared the expression of genes involved in these pathways between the NLPs and ULPs. The most significantly different expression was in genes involved in the Notch and TGFβ pathways. Among the Notch signaling genes examined here, Hes1, Hes5, Hey1, HeyL, and Notch4 were all significantly downregulated in NLPs compared to ULPs (Figure 7A). Among the TGFβ pathway genes, Tfdp1 and Bmpr2 were upregulated, while Id1, Id2, and Id3 were downregulated in NLPs (Figure 7C). Among the Wnt pathway genes, Wnt7a and Fzd7 were upregulated, while Sfrp1, Ctnnbip1, Mapk10, and Dkk2 were downregulated in NLPs (Figure 7B). Among the Hippo pathway genes, Bmpr2, Wnt7a, and Fzd7 were upregulated, while Id1, Id2, and Id3 were downregulated in NLPs (Figure 7D). Real-time PCR was also performed to confirm the RNA-seq results, and these two results were consistent (Figure 7E). The differential expression of genes in the Notch, TGFβ, Wnt, and Hippo pathways suggests that these pathways might be involved in neomycin-induced HC regeneration. Some studies have shown that the Notch and Wnt pathways regulate the development of inner ear progenitor cells (Chai et al., 2012; Kelly et al., 2012). Thus, although the TGFβ and Hippo pathways are not well-studied they are probably the pathways that regulate HC regeneration.

To view the interactions and connections of genes that are differentially expressed in NLPs and ULPs, we constructed a STRING protein-protein interaction network for the significantly differentially expressed genes (fold change > 2.0, p < 0.05) with the functional categories in the gene ontology (GO) analysis (DAVID; Figure 8B). This comprehensive analysis revealed a complex gene network that might regulate HC regeneration. We also applied GO analysis to genes with altered expression levels in NLPs (fold change > 2.0, p < 0.05; Figure 8A). The genes with altered expression in NLPs were highly enriched in functional categories such as auditory receptor cell fate determination, neuron fate determination, signaling, and extracellular matrix formation and maintenance.

We have demonstrated that NLPs are able to regenerate many more HCs than ULPs, which has been reported previously (Bramhall et al., 2014). To determine the detailed mechanisms behind this difference, we compared the expression levels of all of the transcripts in NLPs with those of ULPs. We identified 549 genes that were significantly upregulated and 1,817 genes that were significantly downregulated in the NLPs compared to the ULPs. The functions of some of the differentially expressed genes have been reported previously. Egfr governs the regenerative proliferation of auditory p75+ SCs in birds and mammals after HC damage (Saleem and Siddiqui, 2015). Mutation of Fgfr3 causes hearing loss and inner ear defects and might be involved in regulating the proliferation of SCs (Bermingham-McDonogh et al., 2001; White et al., 2012). Mutations in Frem2 have been linked to Fraser's syndrome, which is a rare autosomal recessive disorder with a spectrum of malformations, including malformations of the ear (Nadol et al., 2015). Mutations in Alms1 cause Alstrom's syndrome, which is an autosomal recessive syndromic genetic disorder with sensorineural hearing loss (Bermingham-McDonogh et al., 2001; White et al., 2012). Lif controls neural differentiation and maintenance of stem cell-derived murine spiral ganglion neuron precursors (Marzella et al., 1999; Su et al., 2015). Hes1, Hes5, Hey1, and HeyL are downstream effectors of the Notch pathway and have been reported to negatively regulate HC differentiation and regeneration (Zheng et al., 2000; Zine et al., 2001; Li et al., 2008; Murata et al., 2009; Abdolazimi et al., 2016). Id1, Id2, and Id3 are downstream targets of the TGFβ and Hippo pathways and regulate HC formation during inner ear development (Ozeki et al., 2005; Jones et al., 2006; Laine et al., 2010; Zhan et al., 2017). These results support our hypothesis that NLPs have a much greater potential to generate HCs in the neonatal cochlea than ULPs. However, it should be noted that not all of the differentially expressed genes that we identified have been characterized, so there might still be mechanisms at work that we are not yet aware of.

Mammalian cochlear SCs do not enter the cell cycle or proliferate after birth under normal circumstances. We demonstrated that Lgr5+ progenitors that re-enter into cell cycle and proliferate could be found in the neomycin-damaged cochlea, but no such cells could be found in the control group. To identify the possible genes regulating the cell cycling of Lgr5+ progenitors, we compared cell cycle gene expression in NLPs and ULPs. Tfdp1 (Vairapandi et al., 2002; Yasui et al., 2003; Lu et al., 2016), which was upregulated in NLPs, is a positive regulator of the cell cycle, while Gadd45g and Sfn, which were downregulated in NLPs, are negative regulators of the cell cycle (Liu et al., 2010; Aktary et al., 2013; Vogel and Herzinger, 2013; Phan et al., 2015). However, Cdkn1a (Duan et al., 2005; Laine et al., 2007; Mollapour et al., 2010; Laos et al., 2017), Wee1 (Lin et al., 2006; Tominaga et al., 2006; De Schutter et al., 2007; Frum et al., 2009), and Mdm2 (Helps et al., 2000; Giono and Manfredi, 2007; Shangary et al., 2008), which were upregulated in NLPs, have been reported to play roles in regulating cell proliferation, and Nek2 (Schultz et al., 1994; Fry et al., 1995; Nabilsi et al., 2013; He et al., 2016), Stmn1 (Johnsen et al., 2000; Wang et al., 2011; Li X. et al., 2015; Guo et al., 2016; Zhou et al., 2016), and Ccne2 (Chen et al., 2015; Clausse et al., 2016; Gorjala et al., 2016), which were downregulated in NLPs, have been reported to negatively regulate cell proliferation. Interestingly, these genes (Cdkn1a, Mdm2, Wee1, Nek2, Stmn1, and Ccne2) are all involved in p53-dependent cell cycle arrest (Fry et al., 1995; Giono and Manfredi, 2007; Kiernan, 2013; Clausse et al., 2016; Zhou et al., 2016; Laos et al., 2017), and the changes in expression of these genes might be because neomycin injury also slightly activates the p53 pathway in Lgr5+ progenitor cells. The expression changes of Tfdp1, Gadd45g, and Sfn promote cell cycle progression, while the expression changes of Cdkn1a, Mdm2, Wee1, Nek2, Stmn1, and Ccne2 repress cell cycle progression, which might be the reason for the lack of significant proliferation in the neomycin treated cochleae.

TFs, which bind to the promoter region of their downstream target genes and regulate gene expression, are important factors involved in development, cell proliferation, differentiation, and other cellular functions. Hes1, Hes5, Hey1, and HeyL are downstream effectors of Notch signaling, which is a well-known signaling pathway regulating HC fate and patterning (Malgrange et al., 2002; Li et al., 2003; Saito et al., 2009; Hartman et al., 2010; Kamaid et al., 2010; Pan et al., 2010; Jeon et al., 2011), and inhibition of Notch induces significant HC regeneration in newborn mice (Li et al., 2003; Kamaid et al., 2010). Id1, Id2, and Id3 (inhibitors of differentiation and DNA binding) regulate HC formation during development by negatively regulating Atoh1 (Ozeki et al., 2005; Jones et al., 2006; Laine et al., 2010; Zhan et al., 2017). These data support our hypothesis that these TFs participate in the increased HC regeneration of NLPs. Furthermore, we have identified many TFs that have not been characterized in the inner ear before. Croxs (Calderon et al., 2012), Lcor (Yu et al., 2014), Nfil3 (Seillet et al., 2014a,b; Malishkevich et al., 2015), Adnp (Nakajima et al., 2008; Oz et al., 2012), and Tfdp1 (Vairapandi et al., 2002; Yasui et al., 2003; Lu et al., 2016) were upregulated in NLPs, and these genes have all been shown previously to have a stimulatory effect on the cell cycle or on the growth of some tumor cells and some normally proliferative tissues and/or on neurodevelopment and lymphoid cell development. Some of the TFs that were downregulated in NLPs, including Esx1 (Asanoma et al., 2015), Bhlhe41 (Cui et al., 2016), and Dmrt1 (Krentz et al., 2009; Zou et al., 2016), have been reported to play critical roles in negatively regulating cancer cell and stem cell growth in other tissues. The involvement of these genes in the differential HC regeneration capacity of NLPs and ULPs should be investigated in the future.

miRNAs bind to target mRNAs and signal their degradation, and they play a key role in the control of gene expression and the regulation of cellular differentiation, proliferation, and apoptosis. Several miRNAs have been reported to play important roles in inner ear development (Jen et al., 1997; Li et al., 2010; Wang et al., 2010; Patel and Hu, 2012). We found eight significantly differentially expressed microRNAs in NLPs and ULPs (p < 0.05, fold change > 2). Mir466i was upregulated in NLPs, while Mir7007, mmu-mir-703, Mir107, Mir361, Mir6918, Mir6982, and Mir3099 were downregulated in NLPs. Among these miRNAs, Mir107 (Chen et al., 2013; Song et al., 2015; Xia et al., 2016; Yang et al., 2016) and Mir361 (Wu et al., 2013; Jacques et al., 2014; Chen et al., 2016; Sun et al., 2016) have been reported to suppress tumor growth and stem cell growth. However, none of the eight miRNAs have been reported previously in the inner ear and need to be further studied in the future.

Several signaling pathways have been shown to be involved in inner ear development and HC regeneration (Malgrange et al., 2002; Yamamoto et al., 2006; Bermingham-McDonogh and Reh, 2011; Chai et al., 2012; Kelly et al., 2012). Among these signaling pathways, Wnt and Notch are the two most well-studied pathways in HC regeneration (Bermingham-McDonogh and Reh, 2011; Chai et al., 2012; Kelly et al., 2012). Overexpression of Wnt increases SC proliferation and Lgr5+ cell clustering and leads to increased numbers of EdU+/Lgr5-EGFP+ cells (Zhao et al., 2006; Madisen et al., 2010; Chai et al., 2012; Bohnenpoll et al., 2014). Inhibition of Notch significantly increases HC differentiation from SCs/Lgr5+ progenitors (Malgrange et al., 2002; Saito et al., 2009; Hartman et al., 2010; Pan et al., 2010; Jeon et al., 2011). Notch inhibition also increases HC regeneration through induction of the Wnt pathway (Li et al., 2003). Other pathways, such as Shh (Liu et al., 2002; Loh et al., 2014), Hippo (Murillo-Cuesta et al., 2015), and TGFβ (Kawamoto et al., 2003; Butts et al., 2005; Yang et al., 2009; McLean et al., 2017), also play important roles in inner ear development. In a recent report, a TGFβ receptor inhibitor increased Lgr5+ cell expansion in vitro (Du et al., 2013). To determine which pathways might be involved in regulating HC regeneration from Lgr5+ progenitors, we examined the differences in expression of pathway-related genes that might play a role in inner ear development between the NLPs and ULPs.

Hes1, Hes5, Hey1, HeyL, and Notch4 are genes of the Notch signaling pathway, which is a well-known signaling pathway regulating HC fate and patterning (Zheng et al., 2000; Zine et al., 2001; Zine and de Ribaupierre, 2002; Li et al., 2008; Murata et al., 2009; Tateya et al., 2011; Korrapati et al., 2013; Ku et al., 2014; Petrovic et al., 2015; Abdolazimi et al., 2016), and were significantly downregulated in NLPs. Inhibition of Notch can lead to HC regeneration mainly by inducing SCs to transdifferentiate into HCs (Malgrange et al., 2002; Saito et al., 2009; Hartman et al., 2010; Pan et al., 2010; Jeon et al., 2011). Although, there is no direct evidence for regulating HC fate and patterning, HeyL is thought to be a target and potential Notch effector of Notch signaling (Kamaid et al., 2010; Bui et al., 2017). Notch4 is involved in the migration and invasion of several kinds of cancers (Melchor et al., 2009; Qian et al., 2016).

Five genes of the TGFβ pathway were differentially expressed. Tfdp1 and Bmpr2 were upregulated, while Id1, Id2, and Id3 were downregulated in NLPs. Id1, Id2, and Id3 regulate HC formation during development by negatively regulating Atoh1 (Ozeki et al., 2005; Jones et al., 2006; Laine et al., 2010; Zhan et al., 2017). Tfdp1 encodes a TF that binds to the promoter regions of a number of genes whose products are involved in cell cycle regulation or in tumor proliferation (Vairapandi et al., 2002; Yasui et al., 2003; Liu S. et al., 2016; Lu et al., 2016). Bmpr2 encodes a member of the bone morphogenetic protein receptor family of transmembrane serine/threonine kinases that play important roles in stem cell differentiation (Zeng et al., 2012; Larabee et al., 2015; Ramos-Solano et al., 2015). The roles of these genes in HC regeneration remain unclear and need to be studied in the future.

Six genes of the Wnt pathway were differentially expressed. Wnt7a and Fzd7 were upregulated, while Sfrp1, Ctnnbip1, Mapk10, and Dkk2 were downregulated in NLPs. Wnt7a, a gene coding for one of the Wnt genes (Chiu et al., 2010; Qu et al., 2013; King et al., 2015; Qiu et al., 2016), and Fzd7 (Sienknecht and Fekete, 2008; Yang et al., 2011; Song et al., 2014; Deng et al., 2015; Wang K. et al., 2015), one of the Wnt protein receptors, were both upregulated in NLPs and have been reported previously to be expressed in the inner ear (Wang K. et al., 2015). Wnt7a and Fzd7 are both reported to induce cell proliferation and differentiation in other tissues and cell types (Sienknecht and Fekete, 2008; Chiu et al., 2010; Yang et al., 2011; Song et al., 2014; Deng et al., 2015; King et al., 2015; Qiu et al., 2016), but their roles in the inner ear remain unclear and need to be further studied in the future. Sfrp1, which codes for a secreted Wnt antagonist that directly interacts with Wnt ligand (Satoh et al., 2008; Lee et al., 2010; Tong et al., 2015), is downregulated in NLPs. Ctnnbip1, which is downregulated in NLPs, encodes a protein that negatively regulates Wnt signaling by preventing the interaction between β-catenin and TCF/LEF family members (Guo et al., 2015; Qi et al., 2015; Li and Luo, 2017). Mapk10, a target of miR-27a-3p, is envolved in nasopharyngeal carcinoma cell proliferation and migration (Phillips et al., 2011). Dkk2, which is downregulated in NLPs, encodes a protein that antagonizes canonical Wnt signaling by inhibiting LRP5/6 interactions with Wnt (Mukhopadhyay et al., 2006; Fleury et al., 2010).

Six genes of the Hippo pathway were differentially expressed. Bmpr2, Fzd7, and Wnt7a were upregulated in NLPs, while Id1, Id2 and Id3 were downregulated in NLPs. Id1, Id2, and Id3, as mentioned above, have been reported to regulate HC formation during inner ear development (Ozeki et al., 2005; Jones et al., 2006; Laine et al., 2010; Zhan et al., 2017). Bmpr2, as mentioned above, plays important roles in stem cell differentiation (Zeng et al., 2012; Larabee et al., 2015; Ramos-Solano et al., 2015). The roles of these genes and the Hippo pathway in HC regeneration remain unclear and need to be studied in the future.

We used the GO analysis to determine the functional categories of the differentially expressed genes in NLPs and ULPs, and the STRING database was used to construct a protein-protein interaction network for the differentially expressed genes. Importantly, NLPs and ULPs have very different expressions of genes involved in inner ear development, neuron differentiation, signaling pathways, and extracellular matrix. Among the genes involved in inner ear development, Fgfr3 (Bermingham-McDonogh et al., 2001; White et al., 2012), Egfr (Saleem and Siddiqui, 2015), Frem2 (Nadol et al., 2015), Alms1 (Oshima et al., 2007b; Jagger et al., 2011), and Lif (Marzella et al., 1999; Su et al., 2015), which are all positively involved in inner ear development and HC differentiation, were upregulated in NLPs. Hes1, Hes5 (Zheng et al., 2000; Zine et al., 2001; Li et al., 2008; Murata et al., 2009; Abdolazimi et al., 2016), Hey1 (Tateya et al., 2011; Korrapati et al., 2013; Benito-Gonzalez and Doetzlhofer, 2014; Petrovic et al., 2015), HeyL (Kamaid et al., 2010), Id1, Id2, and Id3 (Ozeki et al., 2005; Jones et al., 2006; Laine et al., 2010), which all negatively regulate inner ear development and HC differentiation, were downregulated in NLPs. It would be interesting to investigate the involvement of these genes in regulating HC regeneration of Lgr5+ progenitor cells in the future.

In summary, we found that NLPs have a greater capacity to regenerate HCs and a slightly greater capacity to proliferate compared to ULPs. We investigated the differences in the transcriptomes between the NLPs and ULPs, and we identified several differentially expressed genes that might regulate the ability of Lgr5+ progenitor cells to proliferate and to regenerate functional HCs. Lastly, to further analyze the interactions and connections of the differentially expressed genes in HC regeneration, we constructed a STRING protein-protein interaction network. The transcriptomes of the NLPs and ULPs reported here provide numerous target genes that should be characterized for HC regeneration in the future.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.