lable at ScienceDirect

Hearing Research 361 (2018) 66e79

Contents lists avai

Hearing Research

journal homepage: www.elsevier .com/locate/heares

Research Paper

Transcriptomic analysis of chicken cochleae after gentamicin damageand the involvement of four signaling pathways (Notch, FGF, Wnt andBMP) in hair cell regeneration

Lingling Jiang a, 1, Jincao Xu b, 1, Ran Jin a, 1, Huanju Bai a, Meiguang Zhang b, Siyuan Yang c,Xuebo Zhang c, Xinwen Zhang c, Zhongming Han b, **, Shaoju Zeng a, *

a Beijing Key Laboratory of Gene Resource and Molecular Development, Beijing Normal University, 100875, Chinab Department of Otorhinolaryngology, The General Hospital of the PLA Rocket Force, Beijing, 100088, Chinac College of Life Sciences, Hainan Normal University, Haikou, 571158, China

a r t i c l e i n f o

Article history:Received 19 January 2017Received in revised form22 June 2017Accepted 6 January 2018Available online 17 January 2018

Keywords:TranscriptomicChicken cochleaSignaling pathwayRegenerationAntibiotic toxicity

Abbreviations: MAPK, mitogen-activated protein kfactor; TGF-b, transforming growth factor-b; BMP, bSAGE, serial analysis of gene expression; BrdU, 5-bfragments per kilobase of exons per million fragmentexpressed genes; FDR, false discovery rate; PV, parva* Corresponding author.** Corresponding author.

E-mail addresses: drhanliang@163.com (Z. Han), s1 These authors contributed equally to this study.

https://doi.org/10.1016/j.heares.2018.01.0040378-5955/© 2018 Elsevier B.V. All rights reserved.

a b s t r a c t

Unlike mammalian hair cells, which are essentially unable to regenerate after damage, avian hair cellshave a robust capacity for regeneration. The prerequisite for understanding the above difference isknowing the genetic programming of avian hair cell regeneration. Although the major processes havebeen known, the precise molecular signaling that induces regeneration remains unclear. To address thisissue, we performed a high-throughput transcriptomic analysis of gene expression during hair cellregeneration in the chick cochlea after antibiotic injury in vivo. A total of 16,588 genes were found to beexpressed in the cochlea, of which about 1000 genes were differentially expressed among the fourgroups studied, i.e., 2 days (d) or 3 d post-treatment with gentamicin or physiological saline. Thedifferentially expressed genes were distributed across approximately one hundred signaling pathways,including the Notch, MAPK (FGF), Wnt and TGF-b (BMP) pathways that have been shown to playimportant roles in embryonic development. Some differentially expressed genes (2e3 in each pathway)were further verified by qRT-PCR. After blocking Notch, FGF or BMP signaling, the number of regener-ating hair cells and mitotic supporting cells increased. However, the opposite effect was observed aftersuppressing the Wnt pathway or enhancing BMP signaling. To our knowledge, the present study pro-vided a relatively complete dataset of candidate genes and signaling pathways most likely involved inhair cell regeneration and should be a useful start in deciphering the genetic circuitry for inducing haircell regeneration in the chick cochlea.

© 2018 Elsevier B.V. All rights reserved.

1. Introduction

Millions of people have permanent hearing deficits, which canbe caused by genetic disorders, acoustic trauma, aging process orototoxic drugs (Yorgason et al., 2006; Oishi and Schacht, 2011).

inase; FGF, fibroblast growthone morphogenetic protein;romo-2-deoxyuridine; FPKM,s mapped; DEG, differentiallylbumin

jzeng@bnu.edu.cn (S. Zeng).

Therapy for deafness is challenging because hair cells in maturemammals are essentially unable to regenerate after damage(Selimoglu, 2007; Taylor et al., 2008). However, there is a strongcapacity for hair cell regeneration in the cochlea and utricle oflower vertebrates and a limited regenerative ability in themammalian vestibular organ (Cruz et al., 1987; Baird et al., 1996;Harris et al., 2003; Cafaro et al., 2007; Dror and Avraham, 2010).The molecular signals that guide cell regeneration in non-mammalian ears are of great biological and clinical interest, asthey suggest that once we understand how regeneration is trig-gered by the molecular signals, we could induce similar hair cellregeneration in the mammalian cochlea (Cotanche and Kaiser,2010; Dror and Avraham, 2010; Oishi and Schacht, 2011).

Although the major anatomical events that occur duringregeneration in the avian ear have been known (Roberson et al.,

L. Jiang et al. / Hearing Research 361 (2018) 66e79 67

1996, 2004; Daudet et al., 2009; Cotanche and Kaiser, 2010), theprecise molecular signaling that initiates regenerative proliferationremains unclear. However, it is assumed that many of the genes andsignaling pathways required for the production of embryonic haircells are also needed during the regeneration of hair cells (Cotancheand Kaiser, 2010; Groves et al., 2013). Many studies have shownthat the genes and signaling pathways involved in the productionof embryonic hair cells are highly conserved among vertebrates(Cotanche and Kaiser, 2010; Groves et al., 2013) and are complexwith a multitude of interacting signaling pathways, including theNotch, FGF, Wnt and BMP pathways (Lewis, 1998; Bryant et al.,2002; Daudet and Lewis, 2005; Li et al., 2005, 2015; Millimakiet al., 2007; Jayasena et al., 2008; Daudet et al., 2009; Su et al.,2015). Some studies have investigated the roles of these path-ways in the regulation of cell regeneration (Bermingham-McDonogh et al., 2001; Ma et al., 2008; Daudet et al., 2009;Mizutari et al., 2013; Wu et al., 2016). Nevertheless, how hair cellregeneration is induced after damage and what/how signalingpathways are involved in hair cell regeneration still remainunknown.

To know the signaling circuitry of hair cell regeneration afterdamage, comparing the transcriptional profiles of regenerating haircells is regarded as a significant starting point (Chen and Corey,2002; Tao and Segil, 2015; Schimmang and Maconochie, 2016).By using cDNA microarrays, SAGE or RNA-seq, several studies haveinvestigated the changes in the transcriptional profiles during haircell regeneration in the basilar papilla and utricle of chick orzebrafish (Hawkins et al., 2003, 2007; Schuck et al., 2011; Lianget al., 2012; Jiang et al., 2014; Ku et al., 2014; Schimmang andMaconochie, 2016). These experiments generate large amounts ofdata. However, no consensus has been reached regarding changesin gene expression patterns, probably due to the different methodsapplied. To date, quite a lot of studies have been reported on theroles of some signaling pathways, including Notch, FGF, Wnt andBmp in the embryonic development of hair cells (Li et al., 2005,2015; Millimaki et al., 2007; Petrovic et al., 2015; Munnamalaiand Fekete, 2016). However, the precise molecular mechanismsbywhich the signaling pathwaysmediate hair-cell regeneration areless known. Notably, the comprehensive gene expression of in vivocochlea has not been analyzed during regeneration after damage.Such a study is necessary, considering that proliferating cells (BrdU-labeled) can be detected in the chick cochlea in in vitro culture, evenwithout any treatment with ototoxic drugs (Shang et al., 2010), butare not found in the undamaged in vivo cochlea (Jiang et al., 2016).These data suggest that cell regeneration is actually initiated inin vitro culture even without drug treatment. Thus, to obtain thegenetic expression profile necessary to induce hair cell regenera-tion, the transcriptional differences between the cochlea under-going regeneration after damage and the cochlea lackingregeneration completely need to be compared.

To address the above issue, RNA-seq was employed to analyzethe transcriptomic profiles of in vivo basilar papillae after 2 d or 3 dof treatment with gentamicin or physiological saline (for control).According to previous reports, nearly all hair cells were damaged,and the peak of dividing cells in S phase and the onset of hair cellsregenerated via direct transdifferentiation occur in the proximalpart of cochlea after 2 or 3 d gentamicin treatment (Daudet et al.,2009; Shang et al., 2010). The genes involved in initiating cellproliferation and transdifferentiation during cell regenerationshould be included in this period of time (2e3 d after gentamicintreatment). The results indicated that about 1000 genes weredifferentially expressed in the chicken basilar papillae in thestudied groups, including some members of the Notch, FGF, Wntand TGF-b (BMP) signaling pathways. By pharmacological inhibi-tion or activation of the four pathways, we further studied the roles

of the four signaling pathways in hair cell regeneration in thechicken basilar papilla.

2. Materials and methods

2.1. Animal care and treatment

Neonatal chickens (Gallus gallus domesticus) were obtained fromthe Chinese Academy of Agricultural Sciences. Theywere raised in aheated brooder and provided with adequate food and water at alltimes. Posthatch chicks aged between 6 and 8 d (30e45 g) receiveda single subcutaneous injection of gentamicin per day for 2consecutive days, and the injection dosage was 0.25mg/g bodyweight. Control chicks were treated with physiological saline. Thislevel of gentamicin treatment kills nearly all of the hair cells in theproximal part of the cochlear duct (hair cells in the control animalsare intact) (Jiang et al., 2016). Animal experiments were performedaccording to the Beijing Laboratory Animal Welfare and EthicsReview guidelines, and all experimental procedures were approvedby the Animal Management Committee of the College of Life Sci-ences, Beijing Normal University.

2.2. RNA isolation and preparation

RNA samples were obtained from the sensory epithelia of 70e80chickens from each group. After the chickens were treated withgentamicin or physiological saline, they were allowed to survive for2 d or 3 d (the first injection of gentamicin or physiological salinewas considered time zero for survival). The study included fourtotal groups: 2 and 3 d post-treatment with physiological saline(PS) (PS2d/PS3d), and 2 and 3d post-treatment with gentamicin(PG) (PG2d/PG3d).

To obtain the inner ears, the chickens were anesthetized byadministration of ethyl carbamate (intraperitoneal, 1.25mg/g bodyweight) and then euthanized by decapitation. The heads wereimmersed in alcohol for a few seconds, and the external or middleears were progressively opened. After the cochlear ducts weredissected from the inner ear, the tegmentum vasculosum overlyingthe sensory epithelium was dissected away. The sensory epitheliain the proximal half of the whole cochlear duct, in which the haircells were nearly completely damaged, were pooled for each timepoint and treatment group. The pooled tissue was then dissolved inTrizol (GeneCopoeia, America, E01010A) and stored at �80 �C forRNA isolation (Ku et al., 2014). Total RNAwas extracted using a TotalRNA Kit (TIANGEN, Beijing, DP419), and the quality was checked.RNA was stored at �80 �C for RNA-seq or qRT-PCR.

2.3. RNA-seq and data analysis

Enriched mRNA was obtained from the RNA sample of eachgroup by using an NEB Next Poly(A) mRNA Magnetic IsolationModule (NEB, E7490). Before RNA-seq, cDNA libraries were con-structed according to the manufacturer's instructions for IlluminamRNA-seq. Briefly, the enrichedmRNAwas fragmented and used togenerate first-strand cDNA with random primers. Second-strandsynthesis, end repair, poly-A tail addition and adaptor ligationwere then conducted. Finally, each RNA-seq library was sequencedby using the Illumina HiSeq™ 2500 system with paired-end(2� 125 bp) sequencing at Beijing Biomarker Biotechnology. Theaverage correlation coefficient between technical replicates was>0.998.

Raw reads were purified by removing the adaptor sequence andlow-quality unknown sequences. The clean reads in the FASTQformat were mapped to the Ensembl Gallus gallus referencegenome (//ftp.ensembl.org/pub/release-75/fasta/gallus_gallus/)

L. Jiang et al. / Hearing Research 361 (2018) 66e7968

using Tophat2 software (Kim et al., 2013). The aligned results in theBAM/SAM format were further analyzed to trim the potentialduplicate molecules and processed with Cufflinks software(Trapnell et al., 2010) to evaluate the gene expression level,expressed in FPKM (fragments per kilobase of exons per millionfragments mapped). Subsequently, EBseq was used to estimate thedifferential gene expression in the comparisons (Leng et al., 2013).The differences in the abundance between the samples wereanalyzed according to the ratios of the FPKM values. Differentiallyexpressed genes (DEGs) were defined as those with a false dis-covery rate (FDR) of <0.01, whichwas used to identify the thresholdfor the p value in multiple tests, along with a fold change of �1.5.

For annotating the DEGs, various complementary approacheswere used. Sequences were compared with that in several proteindatabases, including the NCBI non-redundant protein (Nr), Swiss-Prot and NCBI non-redundant nucleotide sequence (Nt) data-bases, with a cut-off E-value of 10�5 by either BLASTx or BLASTn.The top hits were retrieved for gene annotations. The gene se-quences were mapped to the COG (Clusters of Orthologous Groups)database with Blastall software to predict and sort functions. Genepathway analysis was conducted using the Kyoto Encyclopedia ofGenes and Genomes online with Perl scripts. The STRING (Knownand Predicted Protein-Protein Interactions) database was used toestablish the network of target genes using Cytoscape software. Allsequence data and annotations were submitted to the NCBI.

2.4. Quantitative real-time PCR

First-strand cDNA for qRT-PCR was synthesized using a TIAN-Script RT Kit (TIANGEN, KR104) and was stored at �20 �C. For eachqRT-PCR run, approximately 50 ng of cDNA was used. qRT-PCR wasperformed using TransStart Green qPCR SuperMix (TransGenBiotech, AQ101-03) on an ABI 7500 real-time PCRmachine (AppliedBiosystems). Eleven genes were chosen to validate the results ob-tained through RNA-seq 3 d after treatment with physiologicalsaline or gentamicin. Each qRT-PCR assay was run in triplicate, andthe expression of b-actin was used as a normalization control. Byusing the 2�DDCt method, the normalized expression level of anexamined gene relative to b-actinwas calculated in each group. Theexpression values are presented as the fold change over any of theassigned samples examined (assigned expression value¼ 1.0).Primers for all genes are listed in Table 1.

2.5. Cochlear duct culture and treatment

The cochlear ducts were isolated from chicks 6 to 8 d post-hatching as described above. The cochlear ducts without thetegmentum vasculosum were cultured free-floating in 500 mL ofculture medium per well in 24-well plates, with incubation at 37 �C

Table 1Primers for qRT-PCR.

Gene 50 primer 30 primer

DLL4 AGGGCAGCTATACGTGTTCATG TGCAACTACCGCCGTTCCTAHes5 TATGCCTGGTGCCTCAAAGA GCTTGTGACCTCTGGAAATGGFGF3 TGAACTGGGCTACAACACCT TTTCCCATTGACTGAGACGTFGFR3 GGAGCGAGACCGCCTTTCTG GGGTCAGGCGAGAACGTGCCBDNF CCCACTGCTCTTTCTGCTC TTCTCCGCTGCTGTTACCCFOS TATTCCCGCCCCGCCGTGTT GCCGCTGCCATCTTGTTCCTCMAPK9 GGTTGCATCATGGGAGAATT CCTCACTGTAGGCTGTAGTTTWnt6 GGGGCAACGATGGGAAGAC GCGGTTGGCAGAGCAGAAATFzd10 ACAGGGTTATGCTATGTTGG AGAAAGAATGAAAGAAGTGCCSmad2 CCGAGTGTCTTAGTGATAGCG TTAGGTTGCATCCTGGTGGGSmad7 TGTCCAAGAGCCCTCCCT TGGCTTCTGTTGTCCGAGTTb-actin TTGGCAATGAGAGGTTCAGGT TACGGATGTCCACATCACACT

in 95% air/5% CO2. Half of the volume of the medium was replacedevery day. The culture medium contained Dulbecco's modifiedEagle's medium (DMEM, Gibco, 11995-065), 1% fetal bovine serum(FBS, Gibco, 16000-044), 1% L-glutamine (Gibco, 25030081), peni-cillin G (1500 U/mg, Amresco, E480), and 5-bromo-2-deoxyuridine(BrdU, 1 mM, Sigma-Aldrich). Streptomycin (78 mM, Sigma-Aldrich)was added to the medium for the first 2 d in culture to ablate allauditory hair cells (Shang et al., 2010), and the cochlear ducts werereturned to control media without streptomycin for an additional6 d to allow for recovery. During the regeneration period, inhibitorswere used to block the three signaling pathways, including DAPT(Notch signaling inhibitor, Calbiochem, 565770), SU5402 (FGFsignaling inhibitor, Santa Cruz, SC-204308A) and XAV-939 (Wntsignaling inhibitor, Selleck, S1180). DMSO, as the vehicle for theinhibitors, was used as a control. To target the BMP signalingpathway, recombinant human BMP4 (Sigma-Aldrich, SRP3016) orNoggin (BMP signaling inhibitor, Sigma-Aldrich, SRP4675) wasadded to the culture medium. For each group, at least 6 basilarpapillae were simultaneously cultured. After culture, the cochlearducts were fixed with buffered 4% paraformaldehyde for 30min to1 h and then washed with phosphate-buffered saline (PBS).

2.6. Immunohistochemistry

The cultured whole-mount cochleae were processed for director indirect immunofluorescence using standard methods. Beforethe immunoreactions, the basilar papillae were immersed inblocking solution (5% normal donkey or bovine serum in 0.05%Triton X-100 in PBS, pH 7.4) for 30min at room temperature. Bothprimary antibody and secondary antibody incubations were per-formed for 2 h at room temperature. The following primary andsecondary antibodies were used: rabbit anti-parvalbumin (PV)(1:800, Santa Cruz Biotechnology, sc-7449), rabbit anti- MyosinVI(1:500, Proteus BioSciences, 25e6791), rat anti-BrdU (1:2000, ABDSerotec, OBT0030CX), goat anti-Sox2 (1:300, Santa Cruz Biotech-nology, sc-17320), Alexa 647-conjugated donkey anti-rabbit IgG forMyosinVI or PV labeling (cat. 711-606-152), Alexa 488-conjugatedbovine anti-goat IgG for Sox2 labeling (cat. 805-545-180), andAlexa 594-conjugated donkey anti-rat IgG for BrdU labeling (cat.712-585-150). All secondary antibodies were diluted at 1:400. Thewhole-mount basilar papillae were carefully placed onto micro-scope slides and sealed with a coverslip using Antifade Solution(Applygen).

As Sox2 is only expressed in supporting cells (but not in haircells), and PV is a significantmobile Ca2þ buffer in hair cells (but notin supporting cells) (Heller et al., 2002; Kiernan et al., 2005), theyare often chosen as the markers of supporting (Sox2) or hair (PV)cells. In addition, MyosinVI is also used as a marker of hair cells(Shang et al., 2010). The specificity of the antibodies in the chickcochlea has been previously reported (Jiang et al., 2016). As acontrol, the primary antibody or the secondary antibody wasomitted, but all other immunohistochemical procedures wereperformed as described above.

2.7. Microscopic imaging and quantitative analysis

The immunolabeled whole-mount basilar papillae were exam-ined and imaged with an inverted Zeiss Axio Observer Z1 fluores-cence microscope and an inverted Zeiss LSM700 laser-scanningconfocal microscope. The acquisition and processing programsAxioVision 4.8 and Zen Light Edition were used, and the imageswere converted to TIFF files for publication. ImageJ and AdobePhotoshop were also used to analyze and manage the obtainedimages.

Because antibiotic damage followed a proximal-to-distal

L. Jiang et al. / Hearing Research 361 (2018) 66e79 69

sequence, nearly all hair cells were damaged in the proximal part ofcochlea at a relatively earlier time (within 24 h after gentamicintreatment) than those in the distal part. We only compared thenumber of regenerated hair cells (PVþ), mitotic cells (BrdUþ) anddifferentiated cells (BrdUþ/Sox2 or BrdUþ/MyosinVIþ) in theproximal region among the studied groups (hair cells damage andregeneration after gentamicin treatment are more homogeneous inthe proximal part than in the distal part). Two or three non-overlapping fields (150 � 150 mm) were captured in the middleregion of the proximal part (the whole-mount cochlea was dividedequally into proximal and distal parts, Fig. 4AeC) under a 40�oilobjective. The data for labeling density were averaged for eachcochlea. One-way ANOVA with Tukey's post hoc test was used tocompare the differences in the numbers of regenerating hair cells inthe proximal region among the studied groups. Student's t-test wasused to compare the differences between groups. The significancewas set at p< .05.

3. Results

3.1. Differential gene expression in the basilar papillae aftergentamicin damage

In the four treatment groups (PS2d, PG2d, PS3d and PG3d), wefound a total of 16,588 genes expressed in the chicken basilarpapillae. Among these genes, 649 were differentially expressed inPS2d vs. PG2d (309 up-regulated, 340 down-regulated), and 805genes were differentially expressed in PS3d vs. PG3d (310 up-regulated, 495 down-regulated) (p< .01, fold change� 1.5). Thesegenes are listed in Dataset S1 in which FDR, log2FC, and COG_class,GO, KOG_class, Pfam and Nr_annotation are shown. Additionally,568 genes were differentially expressed in PG2d vs. PG3d. Incontrast, only 130 genes showed changes in expression betweenPS2d and PS3d (Fig. 1A).

By using the COG (Clusters of Orthologous Groups) database, weperformed gene enrichment analysis for the above differentiallyexpressed genes. The results indicated that the general functioncategory (R in Fig. 1BeE) had a relatively high frequency (>50, thenumber of significantly changed genes in the General functioncategory to the total number of significantly changed genes) in allthe groups except PS2d vs. PS3d (only 10). Compared with thatobserved in PS2d vs. PS3d (inwhich no hair regeneration occurred),obvious changes were found in the other three comparisons (PS2dvs. PG2d, PS3d vs. PG3d, and PG2d vs. PG3d in which substantialhair regeneration occurred) in the following categories of activities:‘Translation, ribosomal structure and biogenesis’ (J), ‘Transcription’(K), ‘Replication, recombination and repair’ (L), ‘Posttranslationalmodification, protein turnover, chaperones’ (O), ‘Inorganic iontransport and metabolism’ (P) and ‘Signal transduction mecha-nisms’ (T) (pointed by arrowheads in Fig. 1BeE). The details of somesignificantly changed genes (fold change> 1.7) mentioned aboveare listed in Dataset S2.

Based on the KEGG (Kyoto Encyclopedia of Genes and Genomes)database, the differentially expressed genes in PS2d vs. PG2d andPS3d vs. PG3d were classified into approximately one hundredsignaling pathways. The top 50 signaling pathways, includingNotch, MAPK, Wnt and TGF-b (BMP), which have beenwell studiedand shown to play important roles in the embryonic developmentof the inner ear (Cotanche and Kaiser, 2010; Groves et al., 2013), areshown in Fig. 2A and B (highlighted). In addition, the number ofdifferentially expressed genes in each signaling pathway and thepercent of these differentially expressed genes from the totalnumber of significantly changed genes in the studied groups arealso indicated (Fig. 2A and B). All of the differentially expressedgenes in the Notch, MAPK, Wnt,TGF-b and Jak-Stat signaling

pathways and cell cycle in PS2d vs. PG2d and PS3d vs. PG3d arelisted in Tables 2 and 3, respectively. These significantly changedgenes (fold change� 1.5) included delta-like 4 and hes5 in the Notchpathway; fgf3, fgfr3, c-fos, and MAPK9 in the MAPK (including FGF)pathway; wnt6 and fzd10 in the Wnt pathway; and smad 2/7 andTHBS1 in the TGF-b/BMP pathway. Based on the differentiallyexpressed genes in PS3d vs. PG3d, the four pathways were con-nected to a network using the STRING (Known and PredictedProtein-Protein Interactions) database (Fig. 3A). The differentiallyexpressed genes in PS2d vs. PG2d and PS3d vs. PG3d in the othertop 50 signaling pathways (total number is 44) after gentamicininjury are shown in Dataset S3 (gene name and down/upregulation).

3.2. Validation of expression changes in RNA-seq analysis with qRT-PCR

To confirm the changes in gene expression obtained from theRNA-seq analysis, a total of 11 differentially expressed genes in theNotch, MAPK (FGF), Wnt, and TGF-b (BMP) pathways were chosenfor further verification by qRT-PCR in the PS3d vs. PG3d group(using 2e3 genes in each pathway). The PS3d vs. PG3d group waschosen because it had the largest number of differentiallyexpressed genes among the studied groups. The chosen genesincluded dll4 (delta-like 4) and hes5 (in the Notch pathway), fgf3,fgfr3, BDNF (brain-derived neurotrophic factor), fos and MAPK9(mitogen-activated protein kinase 9) (in the MAPK/FGF pathway),wnt6 and fzd10 (frizzled class receptor 10) (in the Wnt pathway),and smad2 and smad7 (in the TGF-b/BMP pathway). The qRT-PCRresults confirmed the trends in the gene expression changes inthe basilar papillae after gentamicin damage that were obtainedfrom the RNA-seq analysis (n¼ 3 for each group; Fig. 3B).

3.3. Effects of the Notch, FGF, Wnt and BMP signaling pathways oncell proliferation and differentiation

As shown in the above RNA-seq and qRT-PCR analysis, somegenes were differentially expressed in the Notch, FGF, Wnt andTGF-b signaling pathways. Although these pathways have beenreported to play significant roles in the production of embryonichair cells (for review, see Cotanche and Kaiser, 2010; Groves et al.,2013), previously published studies describing their roles in haircell regeneration are still limited. Thus, through pharmaceuticalsuppression or activation of these signaling pathways, we investi-gated their roles in hair cell regeneration.

Following streptomycin treatment, hair cells were first damagedin the proximal region of the basilar papilla, and almost all of themwere lost in the proximal part after 2 d streptomycin addition.However, many regenerated hair cells (MyosinVIþ or PVþ)appeared in the basilar papilla after 6 d streptomycin treatment(Fig. 4A). To evaluate the effect of pharmacological inhibitors oragonists of the four pathways, we compared the number of re-generated hair cells (PVþ), mitotic cells (BrdUþ) and differentiatedcells (BrdUþ/Sox2 or BrdUþ/MyosinVIþ) in two or three non-overlapping fields (150 � 150 mm for each field), randomly cho-sen in the middle region of the proximal part (approximately theboxed areas in Fig. 4AeC).

3.4. Effect of DAPT on cell proliferation and differentiation

After the addition of DAPT, which inhibits the activity of gammasecretase in the Notch pathway (Daudet et al., 2009), there was adramatic increase in the density of hair cells (PVþ cells) at 6d afterstreptomycin treatment at each of the three doses of DAPT (1, 10and 50 mM), particularly the 10 mM treatment (which resulted in an

Fig. 1. Venn diagram and COG (Clusters of Orthologous Groups of proteins) functional classification of the differentially expressed genes. A: Venn diagram illustrating the overlapsof the differentially expressed genes among the four groups. The differentially expressed genes were determined at a false discovery rate (FDR)< 0.01 and fold change (FC) �1.5.BeE: COG functional classification in the studied groups, i.e., PS2d vs. PS3d (B), PS2d vs. PG2d (C), PS3d vs. PG3d (D) and PG2d vs. PG3d (E). Y-axis: Frequency of the differentiallyexpressed genes in each functional class (ratio of the number of significantly changed genes in each functional class to the total number of significantly changed genes in each of thefour studied groups (A-Z, shown in the insert of at the left bottom). Compared with that observed in the PS2d vs. PS3d comparison, the functional classes with obvious changes inthe other three groups (PS2d vs. PG2d, PS3d vs. PG3d, and PG2d vs. PG3d) are indicated by arrowheads. PS: post-treatment with physiological saline; PG: post-treatment withgentamicin.

L. Jiang et al. / Hearing Research 361 (2018) 66e7970

increase of 147.51%) (n ¼ 3e5; Fig. 5A1-A3’ and H). Although thedensity of Sox2þ/BrdUþ cells showed an increasing trend aftertreatment with DAPT at 10 mM, it was not significant (n ¼ 4e6;Fig. 6A, A’ and G). However, the density of BrdUþ/MyosinVIþ cellswas significantly higher after the administration of DAPT (10 mM)than that observed in the control (n ¼ 4; t ¼ �2.914, p ¼ .033)(Fig. 7A1-A3).

3.5. Effect of SU5402 on cell proliferation and differentiation

After the addition of SU5402 (10 mM), which prevents theautophosphorylation of FGFRs and prevents the expression oftarget genes (Jacques et al., 2012a), the density of hair cells (PVþ) inthe sensory epithelium significantly increased at 6d after strepto-mycin treatment (n ¼ 5; Fig. 5B1-B3’). However, there were no

significant changes induced by the other two doses (5 and 15 mM),thus indicating a dose-dependent effect of SU5402 on hair celldensity after streptomycin damage (n¼ 4e6; Fig. 5I). After SU5402treatment (10 mM), the densities of Sox2þ/BrdUþ and BrdUþ/MyosinVIþ cells were both significantly higher than that in thecontrols (n ¼ 3e5; Fig. 6B, B’ and H for Sox2þ/BrdUþ; Fig. 7B1-B3for BrdUþ/MyosinVI þ).

3.6. Effect of XAV-939 on cell proliferation and differentiation

Treatment with an inhibitor of theWnt signaling pathway, XAV-939, which selectively inhibits Wnt/b-catenin-mediated tran-scription through tankyrase1/2 inhibition and beta-catenin degra-dation by stabilizing axin (Huang et al., 2009), produced adecreasing trend in the density of hair cells (PVþ) at 6d after

Fig. 2. KEGG (Kyoto Encyclopedia of Genes and Genomes) metabolic pathways for the genes showing differential expressions between the gentamicin-treated groups and thecontrols. AeB: The top 50 signaling pathways (in Y-axis) and the frequency of the differentially expressed genes (X-axis, the percent of significantly changed genes in each pathwayfrom the total number of significantly changed genes in the studied group) in the PS2d vs. PG2d (A) and PS3d vs. PG3d (B) groups. Arabic numerals adjacent to the color boxesindicate the number of differentially expressed genes in each pathway. The bars with the same color belong to the same function group (including metabolism, genetic informationprocessing, organism systems and etc). Notch, MAPK (FGF), Wnt and TGF-b signaling pathways are highlighted. PS: post-treatment with physiological saline; PG: post-treatmentwith gentamicin. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

L. Jiang et al. / Hearing Research 361 (2018) 66e79 71

streptomycin treatment at all three doses tested (1 mM, 10 mM and50 mM), with a significant decrease after treatment with 1 mM and10 mMXAV-939 (n¼ 3e5; Fig. 5C1-C3’ and J). The density of Sox2þ/BrdUþ cells also significantly decreased after administration ofXAV-939 (1 mM) (n¼ 4; Fig. 6C, C’ and I). In addition, comparedwiththat of the control group, the density of BrdUþ/MyosinVIþ cells didnot show any significant changes after addition of XAV-939 (1 mM)(n ¼ 5; t ¼ 1.124, p ¼ .324; data not shown).

3.7. Effect of Noggin and BMP4 on cell proliferation anddifferentiation

Treatment with Noggin, an inhibitor of BMP signaling, whichblocks the binding of BMP4 to its receptors (Chang et al., 1999;Gerlach et al., 2000), increased the density of hair cells (PVþ) at6 d after streptomycin treatment at both doses of Noggin tested(0.08 and 0.16 mg/mL) (n ¼ 3e6; Fig. 5D, F1, F2 and K). To furtherstudy the role of the BMP signaling pathway during hair cellregeneration, we added recombinant human BMP4 to the culturemedium and found a decrease in the density of hair cells at bothdoses (10 and 20 ng/mL) (n¼ 3e6; Fig. 5D, E1, E2 and K). However,the above effect of Noggin or BMP4 on the density of hair cells wasnot apparent after the simultaneous addition of Noggin (0.16 mg/mL) and BMP4 (20 ng/mL) to the culture medium (n¼ 5; Fig. 6D, Gand K). The density of cells double-labeled for Sox2 and BrdUsignificantly increased after the addition of Noggin (0.08 mg/mL)(n¼ 5; Fig. 6D, E and J) but significantly decreased after the addi-tion of BMP4 (20 ng/mL) (n¼ 5; Fig. 6D, F and J). The density ofBrdUþ/MyosinVIþ cells had no significant changes after Noggintreatment (0.08 mg/mL) compared to that in the control (n ¼ 3 or 5;t ¼ �1.688, p ¼ .152). In contrast, treatment with BMP4 (20 ng/mL)had no significant effect on the density of BrdUþ/MyosinVIþ cells

(20 ng/mL) (n ¼ 5; t ¼ �0.090, p ¼ .931; data not shown).

4. Discussion

To effectively identify candidate genes underlying hair cellregeneration in the chick, RNA-seq analysis was used in the presentstudy, and about 1000 genes, including some members in theNotch, MAPK (FGF), Wnt and TGF-b (BMP) pathways, were found tobe differentially expressed among the four groups studied. By usingthe STRING database, a probable interplay network of the foursignaling pathways was obtained on the basis of some significantlychanged genes in these pathways (Fig. 3A).

Although large-scale gene expression profiles have beenanalyzed during hair cell regeneration in non-mammals in previousreports (for review, see Schimmang and Maconochie, 2016), therehave only been three in vivo studies: one comparing chick utricleand basilar papilla using human cDNA microarrays (Hawkins et al.,2003) and the others comparing chick basilar papilla after noisedamage using Digital Gene Expression (Liang et al., 2012) orcomparing the zebrafish lateral line after neomycin damage usingRNA-seq (Jiang et al., 2014). Thus, this is the first report on hair cellregeneration of the chick basilar papilla in vivo using RNA-seq. Byutilizing the capabilities of next-generation sequencing, RNA-seqanalysis can indicate a snapshot of RNA presence and quantityfrom a genome in time. In contrast, cDNA microarrays only identifythe allele variants which have been known and designed into themicroarrays. Thus, the number of genes analyzed in cDNA micro-arrays is limited. In addition, RNA-seq has advantage to detect thechanges in gene expression at higher resolution than cDNAmicroarrays (Chu and Corey, 2012).

As mentioned above, cell regeneration actually occurs in in vitroculture of chick basilar papilla even without any drug treatment

Table 2Differentially expressed genes in Notch, MAPK, Wnt, TGF-b and Jak-Stat signaling pathways and cell cycle in 2 d post-treatment with physiological saline vs. gentamicin.

Gene ID Name log2 (Fold Change) up/down regulation FDR

Notch Signaling PathwayENSGALG00000001141 HES5 (transcription factor HES-5) �3.4995 down 0ENSGALG00000027897 DTX1 (E3 ubiquitin-protein ligase deltex 1) �2.0639 down 0.0020ENSGALG00000021593 HEYL (hairy/enhancer-of-split related with YRPW motif-like isoform X1) �1.1900 down 6.88E-07ENSGALG00000008514 DLL4 (delta-like protein 4 isoform X3) �0.7976 down 0.0083ENSGALG00000012075 DTX3L (E3 ubiquitin-protein ligase deltex 3 like) 1.3333 up 4.10E-07MAPK Signaling PathwayENSGALG00000008262 RASGRF1 (ras-specific guanine nucleotide-releasing factor 1 isoform X1) �0.8409 down 0.0021ENSGALG00000019233 PLA2G10 (group 10 secretory phospholipase A2 isoformX2) �0.8244 down 0.0008ENSGALG00000015708 FGFR3 (fibroblast growth factor receptor 3 precursor) �0.8200 down 0.0008ENSGALG00000027830 CACNG2 (voltage-dependent calcium channel gamma-2 subunit isoform X2) 1.6881 up 0.0006ENSGALG00000028037 Fos (proto-oncogene c-Fos) 1.1598 up 2.21E-08ENSGALG00000015598 RASGRF2 (ras-specific guanine nucleotide-releasing factor 2 isoform X2) 1.0996 up 1.21E-07ENSGALG00000028005 GADD45 (growth arrest and DNA damage-inducible protein GADD45 gamma) 1.0405 up 3.71E-06ENSGALG00000004690 MAPK6 (mitogen-activated protein kinase 6) 0.9539 up 7.59E-06ENSGALG00000028143 GADD45B (growth arrest and DNA damage-inducible protein GADD45 beta-like) 0.8052 up 0.0058Wnt Signaling PathwayENSGALG00000012073 SFRP4 (secreted frizzled-related protein 4) 1.5755 up 3.65E-08ENSGALG00000003485 CCND3 (G1/S-specific cyclin-D3) 0.7645 up 0.0071ENSGALG00000010870 JUN (Proto-oncogene c-Jun) 0.8847 up 8.46E-05TGF-b Signaling PathwayENSGALG00000014184 SMAD2 (mothers against decapentaplegic homolog 2-like isoformX1) �2.0269 down 2.80E-14ENSGALG00000006210 ID1 (DNA-binding protein inhibitor ID-1) �0.9110 down 4.26E-05ENSGALG00000009626 THBS1 (thrombospondin-1 precursor) 0.9898 Up 2.08E-06Jak-Stat Signaling PathwayENSGALG00000016906 SPRY2 (protein sprouty homolog 2) �6.4143 down 0ENSGALG00000027786 SOCS3 (suppressor of cytokine signaling 3) 2.6849 up 7.59E-06ENSGALG00000007651 STAT1 (Signal transducer and activator of transcription 1) 1.1005 up 7.03E-08ENSGALG00000003485 CCND3 (G1/S-specific cyclin-D3) 0.7645 up 0.0071Cell cycleENSGALG00000014184 SMAD2 (mothers against decapentaplegic homolog 2-like isoformX1) �2.0269 down 2.80E-14ENSGALG00000017052 CCNA1 (cyclin-A1 isoform X3) 2.6962 up 2.59E-14ENSGALG00000028005 GADD45 (growth arrest and DNA damage-inducible protein GADD45 gamma) 1.0405 up 3.71E-06ENSGALG00000028143 GADD45B (growth arrest and DNA damage-inducible protein GADD45 beta-like) 0.8052 up 0.0058ENSGALG00000003485 CCND3 (G1/S-specific cyclin-D3) 0.7645 up 0.0071

FDR: false discovery rate. For significantly differentially expressed genes, FDR <0.01, and Fold Change �1.5.

L. Jiang et al. / Hearing Research 361 (2018) 66e7972

(Shang et al., 2010). To identify candidate genes involved in earlystage of hair cell regeneration, in vivo study on the chick basilarpapilla was necessary, as some genes do not express in completelyquiescent inner ears, allowing relatively more genes differentiallyexpressed to be distinguished out from quiescent inner ears incontrast with those undergoing regeneration. The identified genesand their located signaling pathways in the present study shouldrepresent a relatively more complete and accurate dataset ofcandidate genes or pathways involved in hair cell regeneration inthe chick cochlea, providing a useful start to determine the geneticcircuitry of hair cell regeneration.

Although some studies have investigated the roles of the Notch,FGF, Wnt and BMP pathways in regulating embryonic production ofhair cells or cell regeneration after damage (Bermingham-McDonogh et al., 2001; Ma et al., 2008; Daudet et al., 2009;Mizutari et al., 2013; Wu et al., 2016), the present study providessome unreported data or further strengthens previous reports.

4.1. Notch signaling pathway

Notch signaling plays a crucial role during inner ear develop-ment and regeneration (Zine et al., 2000; Daudet et al., 2009).Among Notch target genes, onlyHey1 andHes5 are expressed in theinner ear of the chicken embryo, and both Hey1 and Hes5 repressDll and Hes5 (self inhibition) expression (Petrovic et al., 2015).These reports are consistent with our RNA-seq analysis to indicatethat both Hes5 and Dll4 were down-regulated in chick hair cellsafter gentamicin damage, but the othermembers of Notch signalingpathway were not found to change significantly. In addition, it has

been shown that Hes5, Delta1 and Notch1 are down-regulated inchick hair cells at early stages of regeneration after gentamicindamage (Daudet et al., 2009), and Notch pathway is inhibited 1 hafter neomycin treatment, but later activated in regenerating sup-port cells of the zebrafish lateral line (Jiang et al., 2014). Thesefindings support a hypothesis that down-regulation of Notchsignaling is needed during the early stages of hair cell developmentor regeneration to allow support cells to transdifferentiate into haircells, but up-regulation of Notch signaling is required to facilitatethe fate specification of hair and support cells in later stages (Jianget al., 2014). It is needed to point out that all hair cells in thezebrafish lateral line can be killed acutely within 30min, and pro-liferation of inner support cells begins within the first 2 h and mostregeneration completes within 24 h, whereas cell injury and cellregeneration occur in a relatively longer period in the inner ears ofchicks or mice (Jiang et al., 2014). This might be the reason why theexpression of up-regulated genes in the Notch pathway does notshow significant changes in the inner ear of chicks or mice as thosein the zebrafish lateral line in later stages of regeneration.

In addition, our results further indicated that blocking the Notchpathway by DAPT caused the increases in the density of hair cells(PVþ cells) and BrdUþ/MyosinVIþ cells in the proximal part of thecochlea at 6 d after streptomycin treatment, but not in the densityof Sox2þ/BrdUþ cells. Our study recapitulates previous studies todemonstrate that inhibition of the Notch pathway leads to anexcess of hair cells in the zebrafish, chick and mouse (Zine et al.,2000; Daudet et al., 2009; Mizutari et al., 2013; Jiang et al., 2014),strengthening an important role of Notch pathway in hair cellregeneration.

Table 3Differentially expressed genes in Notch, MAPK, Wnt, TGF-b and Jak-Stat signaling pathways and cell cycle in 3 d post-treatment with physiological saline vs. gentamicin.

Gene ID Name log2 (Fold Change) up/down regulation FDR

Notch Signaling PathwayENSGALG00000001141 HES5 (transcription factor HES-5) �1.2161 down 0.0019ENSGALG00000008514 DLL4 (delta-like protein 4 isoform X3) �1.2032 down 7.39E-08ENSGALG00000004284 LFNG (beta-1,3-N-acetylglucosaminyltransferase lunatic fringe precursor) 0.8411 up 0.0001MAPK Signaling PathwayENSGALG00000012163 BDNF (brain-derived neurotrophic factor precursor) �2.2461 down 1.71E-05NSGALG00000009740 RASGRP1 (RAS guanyl-releasing protein 1 isoform X1) �1.5654 down 1.50E-13ENSGALG00000005332 CACNA1D (voltage-dependent L-type calcium channel subunit alpha-1D) �1.4609 down 4.44E-16ENSGALG00000026853 FGF3 (fibroblast growth factor 3 precursor) �1.3198 down 2.88E-06ENSGALG00000028037 FOS (proto-oncogene c-Fos) �1.2706 down 1.94E-12ENSGALG00000015708 FGFR3 (fibroblast growth factor receptor 3 precursor) �1.0414 down 5.78E-09ENSGALG00000008410 RALGAPA2 (ral GTPase-activating protein subunit alpha-2 isoform X5) �1.0110 down 1.92E-06ENSGALG00000020249 RAPGEF2 (rap guanine nucleotide exchange factor 2 isoform X3) �0.8955 down 2.88E-05ENSGALG00000015725 FGFRL1 (fibroblast growth factor receptor-like 1 precursor) �0.8466 down 2.75E-05ENSGALG00000008591 CACNB2 (voltage-dependent L-type calcium channel subunit beta-2 isoform X3) �0.7995 down 0.0083ENSGALG00000016406 RPS6KA3 (ribosomal protein S6 kinase alpha-3 isoform X10) �0.7727 down 0.0019ENSGALG00000028143 GADD45B (growth arrest and DNA damage-inducible protein GADD45 beta-like) 1.2619 up 2.45E-09ENSGALG00000001926 HSPB1 (heat shock protein beta-1-like) 1.1095 up 2.59E-07Wnt Signaling PathwayENSGALG00000012654 NFATC1 (nuclear factor of activated T-cells, cytoplasmic 1 isoform X4) �1.1369 down 3.55E-06ENSGALG00000002652 FZD10 (frizzled-10 precursor) �0.9567 down 6.14E-06ENSGALG00000003485 CCND3 (G1/S-specific cyclin-D3) 1.7788 up 0ENSGALG00000011358 WNT6 (protein Wnt-6-like) 0.8006 up 0.0041ENSGALG00000013761 MAPK9 (mitogen-activated protein kinase 9) �0.7076 down 0.0037ENSGALG00000012456 RAC2 (ras-related C3 botulinum toxin substrate 2) 1.1858 up 0.0098TGF-b Signaling PathwayENSGALG00000014184 SMAD2 (mothers against decapentaplegic homolog 2-like isoformX1) �5.8859 down 0ENSGALG00000027843 SMAD7 (Mothers against decapentaplegic homolog 7) �4.1254 down 0.00117589ENSGALG00000009626 THBS1 (hrombospondin-1 precursor) 1.3945 up 0Jak-Stat Signaling PathwayENSGALG00000016906 SPRY2 (protein sprouty homolog 2) �1.4803 down 2.53E-08ENSGALG00000002335 DOCK3 (dedicator of cytokinesis protein 3) �1.3853 down 2.22E-16ENSGALG00000000311 DOCK5 (dedicator of cytokinesis protein 5 isoform X2) �0.8202 down 0.0002ENSGALG00000012836 PRC1 (protein regulator of cytokinesis 1) 2.4007 up 8.94E-06ENSGALG00000003485 CCND3 (G1/S-specific cyclin-D3) 1.7788 up 0Cell cycleENSGALG00000014184 SMAD2 (mothers against decapentaplegic homolog 2-like isoformX1) �5.8859 down 0ENSGALG00000014378 ANAPC4 (anaphase-promoting complex subunit 4 isoform X1) �0.7606 down 0.0074ENSGALG00000006110 PLK1 (serine/threonine-protein kinase PLK1) 2.9149 up 7.23E-05ENSGALG00000006051 CDC7 (cell division cycle 7-related protein kinase isoform X5) 2.4150 up 0.0015ENSGALG00000025810 CCNB3 (G2/mitotic-specific cyclin-B3) 2.3658 up 1.59E-06ENSGALG00000008233 BUB1 (mitotic checkpoint serine/threonine-protein kinase BUB1) 1.9627 up 0.0004ENSGALG00000003085 CDK1 (cyclin-dependent kinase 1) 1.8490 up 2.84E-13ENSGALG00000004838 BUB1B (mitotic checkpoint serine/threonine-protein kinase BUB1 beta) 1.7933 up 4.96E-07ENSGALG00000003485 CCND3 (G1/S-specific cyclin-D3) 1.7788 up 0ENSGALG00000017052 CCNA1 (cyclin-A1 isoform X3) 1.4905 up 0.0016ENSGALG00000004161 CCNB2 (G2/mitotic-specific cyclin-B2) 1.4396 up 3.17E-06ENSGALG00000028143 GADD45B (growth arrest and DNA damage-inducible protein GADD45 beta-like) 1.2619 up 2.45E-09ENSGALG00000003045 E2F1 (transcription factor E2F1) 1.2042 up 0.0001ENSGALG00000028005 GADD45 (growth arrest and DNA damage-inducible protein GADD45 gamma) 1.1385 up 1.59E-09ENSGALG00000016676 MCM3 (DNA replication licensing factor MCM3) 1.0977 up 0.0027ENSGALG00000012348 MCM6 (DNA replication licensing factor MCM6) 1.0428 up 2.35E-06ENSGALG00000012546 MCM5 (DNA replication licensing factor MCM5) 0.9740 up 0.0019

FDR: false discovery rate. For significantly differentially expressed genes, FDR <0.01, and Fold Change �1.5.

L. Jiang et al. / Hearing Research 361 (2018) 66e79 73

4.2. FGF signaling pathway

The Fgf family includes more than 20 ligands and fourmembrane-bound receptors, of which Fgfr2 is not expressed withinthe cochlear prosensory of mice, and fgfr4 is not reported in liter-ature (Hebert, 2011; Jacques et al., 2012a). SU5402 has equivalentinhibitory effects on all the Fgf receptors (Mohammadi et al., 1997;Hebert, 2011). During embryonic development, Fgfr3 iscontinuously expressed in progenitor cells, but it is subsequentlylimited to expression only in supporting cells (Bermingham-McDonogh et al., 2001). Fgf3 and Fgfr3 are down-regulated inregenerating support cells of the zebrafish lateral line (Jiang et al.,2014) or the chick inner ear (Bermingham-McDonogh et al.,2001). The Fgfr3 mutant mouse has deficiencies in support celldifferentiation, and the inner or outer pillar cell development

(Hayashi et al., 2007). In contrast, Fgf9 or Fgf8mutantmouse has thenormal cytoarchitecture of the sensory and nonsensory epithelia,although the mesenchyme of scala vestibule is abnormally devel-oped in Fgf9 mutant mouse (Pirvola et al., 2004), and the pillar celldevelopment is disrupted in Fgf8 mutant mouse (Jacques et al.,2007). Our RNA-seq results indicated that Fgfr3, Fgf3 and Fgfr1(fibroblast growth factor receptor-like 1 precursor) were down-regulated, but the other Fgf ligands and receptors were notchanged significantly (Table 3). These data indicate that, of Fgffamily members, Fgf3 and its receptor might play crucial roles inhair cell development and regeneration.

In addition, previous reports have shown that the FGF pathwayaffects the induction of the otic placode, the formation of otocysts,the generation of hair cells and the formation of the mosaicstructure between hair cells and supporting cells (Hayashi et al.,

Fig. 3. Gene network and qRT-PCR for 11 differentially expressed genes in the Notch, MAPK (FGF), Wnt and TGF-b pathways. A: The interaction network of the differentiallyexpressed genes in the Notch, MAPK (FGF), Wnt and TGF-b pathways. This network is obtained by using the STRING (Known and Predicted Protein-Protein Interactions) database.Thicker lines indicate relatively stronger correlations. B: Comparison of the relative expression levels of the 11 differentially expressed genes between RNA sequencing (RNA-seq, inblue) and quantitative real-time PCR (qRT-PCR, in shallow red) analyses. PS3d: post-treatment with physiological saline for 3 d; PG3d: post-treatment with gentamicin for 3 d. Thenormalized expression level of an examined gene relative to b-actinwas calculated by using the 2�DDCt method, and the expression values are presented as the log2 fold change (FC)over any of the assigned samples examined (assigned expression value¼ 1.0). (For interpretation of the references to color in this figure legend, the reader is referred to the Webversion of this article.)

Fig. 4. Labeling for regenerating hair cells, supporting cells and proliferating cells in the chick basilar papilla at 6 days after streptomycin treatment and the regions analyzed forquantitative examination of labeling densities. AeC: Labeling for regenerating hair cells (MyosinVIþ, A), supporting cells (Sox2þ, B), and proliferating cells (BrdUþ, C). DeF: Theboxed areas in A-C are shown at higher magnification, respectively. G: Merged image of D-F. Most of regenerating hair cells are double-labeled for Sox2, but not for BrdU, suggestingthat they are from transdifferentiation from supporting cells. The areas analyzed for quantitative examination of labeling densities are from two or three non-overlapping fields(150 � 150 mm for each field), randomly chosen in the middle region of the proximal part (approximately boxed areas in Fig. 4AeC). The whole basilar papilla is divided into theproximal (to the left) and distal (to the right) parts by the middle line. Scale bar in C ¼ 200 mm for A-C; and in G ¼ 100 mm for D-G.

L. Jiang et al. / Hearing Research 361 (2018) 66e7974

2007; Jacques et al., 2007; Schimmang, 2007). In SU5402 treatedembryonic basilar papillae of chick, there is a dose dependent in-crease in hair cell density, a slight addition in BrdU-positive haircells and Sox2 cells (Jacques et al., 2012a). Here, our results indi-cated that the density of hair cells increased by 32.95%, after inhi-bition of FGF signaling, extending the study on the development ofembryonic basilar papillae to the hair cell regeneration in post-

hatch chicks.

4.3. BMP signaling pathway

BMP pathway has been reported to be involved in the embry-onic development of inner ears (Ohta et al., 1996; Fekete and Wu,2002; Ohta et al., 2010, 2016), and the regeneration of hair cells

Fig. 5. Effects of exogenous regulation of the Notch, FGF, Wnt and BMP pathways on the regenerating hair cells 6d after streptomycin injury. A1-G: Regenerating hair cells labeledfor parvalbumin (PV). A’1-A’3: After addition of DAPT, an inhibitor of the Notch pathway; B’1-B’3: After addition of SU5402, an inhibitor of the FGF pathway; C’1-C’3: After additionof XAV-939, an inhibitor of the Wnt pathway; D-E2: After addition of Noggin, an inhibitor of the BMP pathway; F1 and F2: After addition of BMP4; G: After addition of Noggin andBMP4. HeK: Densities of the regenerating hair cells (PVþ) after blocking or enhancing the Notch (H), FGF (I), Wnt (J) and BMP (K) pathways. Asterisks (*) indicate p < .05, and errorbars indicate the standard error of the mean. Scale bar in G ¼ 20 mm for A1-G.

L. Jiang et al. / Hearing Research 361 (2018) 66e79 75

after damage (Wu and Oh,1996; Cole et al., 2000; Jiang et al., 2014).Of BMP pathway, BMP4 is relatively well studied: BMP4 has aparticular temporal and spatial expression pattern during thedevelopment of sensory epithelium in the inner ear (Ohta et al.,1996), and during the regeneration of hair cells (Jiang et al.,2014). For example, BMP4 decreases at 1 h, but increases at 3 or5 h in regenerating support cells of the zebrafish lateral line afterneomycin damage (Jiang et al., 2014). BMP4 can regulate the

proliferation of progenitor cells in the inner ear by affecting theexpression of EGFR, Id1 and Pax2 (Li et al., 2005; Chen et al., 2013),and antagonizing BMP4 by Noggin in vitro can cause the decrease ofthe numbers of hair cells in the vesicle of embryonic chicken ear(Pujades et al., 2006), leading to developmental deformities of thesemicircular canals and otic capsule (Chang et al., 1999; Gerlachet al., 2000). Our RNA-Seq analyses indicated that inhibitorySmad2 and Smad7 were down-regulated (but the other

Fig. 6. Effects of exogenous regulation of the Notch, FGF, Wnt and BMP pathways on the mitotic division of supporting cells after streptomycin injury. AeF: Cells double-labeled forSox2 (green) and BrdU (red) after addition of DAPT (an inhibitor of the Notch pathway) (A, A’), SU5402 (an inhibitor of the FGF pathway) (B, B’), XAV-939 (an inhibitor of the Wntpathway) (C, C’), Noggin (an inhibitor of the BMP pathway) (D, E) and BMP4 (F). The cells with single labeling (BrdU in red, and Sox2 in green), indicated by arrowheads in A-F, arefurther shown in the insert at the right bottom of each panel. GeJ: Densities of cells double-labeled for Sox2 and BrdU after blocking or enhancing the Notch (G), FGF (H), Wnt (I)and BMP (J) pathways. Asterisks (*) indicate p < .05, and error bars indicate the standard error of the mean. Scale bar ¼ 20 mm for A-F. (For interpretation of the references to color inthis figure legend, the reader is referred to the Web version of this article.)

L. Jiang et al. / Hearing Research 361 (2018) 66e7976

components in BMP pathway were not found to be changedsignificantly, Table 3). Although BMP4 expressionwas not shown tochange significantly, the density of hair cells or of cells double-labeled for Sox2 and BrdU increased after treatment with Noggin,and decreased after treatment with BMP4, suggesting that BMPpathway is also involved in hair cell regeneration of inner ear(Pujades et al., 2006).

4.4. Wnt signaling pathway

Wnt/b-catenin signaling is necessary for the developing andregenerating inner ears (Jacques et al., 2012b, 2014; Munnamalaiand Fekete, 2013). Enhancing Wnt signaling during embryonicdevelopment has been shown to increase the number of prolifer-ating supporting cells, subsequently increasing the number of haircells (Alvarado et al., 2011; Jacques et al., 2012b, 2014; Head et al.,

2013). In contrast, suppressing Wnt signaling decreases the num-ber of proliferating cells in both chicken basilar papillae andzebrafish lateral lines (Jacques et al., 2014). Although some keymembers of Wnt/b-catenin signaling are not found significantly toexpress during hair cell regeneration in zebrafish lateral lines afterneomycin damage, Gsk-3b and Wnt10a are induced to express inhigh levels (Jiang et al., 2014). Our RNA-seq analysis indicated that,among the members of Wnt pathway, Wnt6, Sfrp4 (secretedfrizzled-related protein 4) was up-regulated, and Fzd10 was down-regulated at 2 or 3 d after gentamicin treatment. In addition, aftertreatment with XAV-939, the densities of hair cells (PVþ cells),Sox2þ/BrdUþ cells, and BrdUþ/MyosinVIþ cells decreased, sug-gesting that the Wnt pathway plays a role in hair cell regeneration,although it remains unknown how the Wnt pathway is involved.

Fig. 7. Effects of exogenous regulation of the Notch and FGF pathways on the regeneration of hair cells through mitotic division after streptomycin injury. A1-A2: Cells double-labeled for MyosinVI (green) and BrdU (red) after the addition of DAPT (an inhibitor of the Notch pathway); B1-B2: Cells double-labeled for MyosinVI and BrdU after the addi-tion of SU5402 (an inhibitor of the FGF pathway). The cells with single labeling (BrdU in red, and MyosinVIþ in green), indicated by arrowheads in A1- B2, are further shown in theinsert at the right bottom of each panel. Asterisks (*) indicate p < .05, and error bars indicate the standard error of the mean. Scale bar ¼ 20 mm for A1-B2. (For interpretation of thereferences to color in this figure legend, the reader is referred to the Web version of this article.)

L. Jiang et al. / Hearing Research 361 (2018) 66e79 77

4.5. Regulation of cell signaling pathways in combination withmitogenic signals

Complete cochlear regeneration requires stimulating supportcells to re-enter the cell cycle and self-renew, and to differentiateinto hair cells, and the failure to proliferate is one of the reasonsmammalian support cells do not regenerate after hair cell loss(Groves, 2010; Burns and Corwin, 2013). It is necessary to deter-mine the changes in the expression of cell cycle regulators, andcorrelate these changes to the signaling pathways during hair cellregeneration. These pathways might be candidate ones involved inthe onset of regeneration. As shown in Tables 2 and 3, some cellcycle genes are up-regulated, including Cyclin (Cyc)-A1, B, D, CDK1,Cdc7 and Plk1, agreeing with the increasing activity of cell prolif-eration (which can be reflected from BrdU labeling, Figs. 6 and 7).The increasing activity of these cell cycle regulators may be causedby the downregulation of cdk inhibitors, owing to the inhibition ofNotch and Fgf signaling, as suggested by Liu et al. (2012). However,blocking the Notch pathway does not affect the density of Sox2þ/BrdUþ cells, suggesting that Notch signaling pathway is notinvolved in triggering or inhibiting proliferation (Mizutari et al.,2013, present data).

In addition, although Wnt activation and Notch signaling inhi-bition can lead to an excess of regenerating chick hair cells, the twosignaling pathways alone cannot regenerate significant amounts ofnew hair cells (Daudet et al., 2009; Head et al., 2013; Jacques et al.,2014). However, after the co-regulation of Wnt and Notch signalingor after the activation of b-catenin, deletion of Notch1, and

overexpression of Atoh1, more regenerated hair cells are producedthan those after activation of a single pathway after damage to themouse inner ear (Li et al., 2015; Ni et al., 2016; Wu et al., 2016).Thus, it is probable that regeneration of hair cells, like in embryonicdevelopment, should also be orchestrated by the interplay of manygenes and signaling pathways (Kelley, 2006; Munnamalai andFekete, 2013), as a recent report to indicate that Wnt signalingtakes effect on the organizational pattern of mouse cochlea atseveral stages of embryonic development through complex cross-talk with the Bmp and Notch pathways (Munnamalai and Fekete,2016). Here, a probable interplay network of the four signalingpathways was provided (Fig. 3A).

Acknowledgments

This work was supported by the National Natural ScienceFoundation of China through grants to ZM Han (No: 81371352), SJZeng (No: 31372200, 31672283), XW Zhang (No: 31560275 and31360517), and XB Zhang (No: 31360243). We thank Drs. Jin-Liuand Chao-Xi in the Experimental Technology Center for Life Sci-ences and the College of Life Sciences, Beijing Normal University, fortechnological assistance.

Appendix A. Supplementary data

Supplementary data related to this article can be found athttps://doi.org/10.1016/j.heares.2018.01.004.

L. Jiang et al. / Hearing Research 361 (2018) 66e7978

Conflict of interest

The authors declare no competing financial interests.

References

Alvarado, D.M., Hawkins, R.D., Bashiardes, S., Veile, R.A., Ku, Y.C., Powder, K.E.,Spriggs, M.K., Speck, J.D., Warchol, M.E., Lovett, M., 2011. An RNA interference-based screen of transcription factor genes identifies pathways necessary forsensory regeneration in the avian inner ear. J. Neurosci. 31, 4535e4543.

Baird, R.A., Steyger, P.S., Schuff, N.R., 1996. Mitotic and nonmitotic hair cell regen-eration in the bullfrog vestibular otolith organs. Ann. N. Y. Acad. Sci. 781, 59e70.

Bermingham-McDonogh, O., Stone, J.S., Reh, T.A., Rubel, E.W., 2001. FGFR3expression during development and regeneration of the chick inner ear sensoryepithelia. Dev. Biol. 238, 247e259.

Bryant, J., Goodyear, R.J., Richardson, G.P., 2002. Sensory organ development in theinner ear: molecular and cellular mechanisms. Br. Med. Bull. 63, 39e57.

Burns, J.C., Corwin, J.T., 2013. A historical to present-day account of efforts to answerthe question: “What puts the brakes on mammalian hair cell regeneration?”.Hear. Res. 297, 52e67.

Cafaro, J., Lee, G.S., Stone, J.S., 2007. Atoh1 expression defines activated progenitorsand differentiating hair cells during avian hair cell regeneration. Dev. Dynam.236, 156e170.

Chang, W., Nunes, F.D., De Jesus-Escobar, J.M., Harland, R., Wu, D.K., 1999. Ectopicnoggin blocks sensory and nonsensory organ morphogenesis in the chickeninner ear. Dev. Biol. 216, 369e381.

Chen, Z.Y., Corey, D.P., 2002. An inner ear gene expression database. J. Assoc. Res.Otolaryngol. 3, 140e148.

Chen, J., Sun, W., Zheng, Y., Xiong, H., Cai, Y., 2013. Bone morphogenetic protein 4,inhibitor of differentiation 1, and epidermal growth factor receptor regulate thesurvival of cochlear sensory epithelial cells. J. Neurosci. Res. 91, 515e526.

Chu, Y., Corey, D.R., 2012. RNA sequencing: platform selection, experimental design,and data interpretation. Nucleic. Acid 22, 271e274.

Cole, L.K., Le Roux, I., Nunes, F., Laufer, E., Lewis, J., Wu, D.K., 2000. Sensory organgeneration in the chicken inner ear: contributions of bone morphogeneticprotein 4, serrate1, and lunatic fringe. J. Comp. Neurol. 424, 509e520.

Cotanche, D.A., Kaiser, C.L., 2010. Hair cell fate decisions in cochlear developmentand regeneration. Hear. Res. 266, 18e25.

Cruz, R.M., Lambert, P.R., Rubel, E.W., 1987. Light microscopic evidence of hair cellregeneration after gentamicin toxicity in chick cochlea. Arch. Otolaryngol. HeadNeck Surg. 113, 1058e1062.

Daudet, N., Lewis, J., 2005. Two contrasting roles for Notch activity in chick innerear development: specification of prosensory patches and lateral inhibition ofhair-cell differentiation. Development 132, 541e551.

Daudet, N., Gibson, R., Shang, J., Bernard, A., Lewis, J., Stone, J., 2009. Notch regu-lation of progenitor cell behavior in quiescent and regenerating auditoryepithelium of mature birds. Dev. Biol. 326, 86e100.

Dror, A.A., Avraham, K.B., 2010. Hearing impairment: a panoply of genes andfunctions. Neuron 68, 293e308.

Fekete, D.M., Wu, D.K., 2002. Revisiting cell fate specification in the inner ear. Curr.Opin. Neurobiol. 12, 35e42.

Gerlach, L.M., Hutson, M.R., Germiller, J.A., Nguyen-Luu, D., Victor, J.C., Barald, K.F.,2000. Addition of the BMP4 antagonist, noggin, disrupts avian inner eardevelopment. Development 127, 45e54.

Groves, A.K., 2010. The challenge of hair cell regeneration. Exp. Biol. Med. 235,434e446.

Groves, A.K., Zhang, K.D., Fekete, D.M., 2013. The genetics of hair cell developmentand regeneration. Annu. Rev. Neurosci. 36, 361e381.

Harris, J.A., Cheng, A.G., Cunningham, L.L., Macdonald, G., Raible, D.W., Rubel, E.W.,2003. Neomycin-induced hair cell death and rapid regeneration in the lateralline of zebrafish (Danio rerio). J. Assoc. Res. Otolaryngol. 4, 219e234.

Hawkins, R.D., Bashiardes, S., Helms, C.A., Hu, L., Saccone, N.L., Warchol, M.E.,Lovett, M., 2003. Gene expression differences in quiescent versus regeneratinghair cells of avian sensory epithelia: implications for human hearing and bal-ance disorders. Hum. Mol. Genet. 12, 1261e1272.

Hawkins, R.D., Bashiardes, S., Powder, K.E., Sajan, S.A., Bhonagiri, V., Alvarado, D.M.,Speck, J., Warchol, M.E., Lovett, M., 2007. Large scale gene expression profiles ofregenerating inner ear sensory epithelia. Plos One 2, e525.

Hayashi, T., Cunningham, D., Bermingham-McDonogh, O., 2007. Loss of Fgfr3 leadsto excess hair cell development in the mouse organ of Corti. Dev. Dynam. 236,525e533.

Head, J.R., Gacioch, L., Pennisi, M., Meyers, J.R., 2013. Activation of canonical Wnt/b-catenin signaling stimulates proliferation in neuromasts in the zebrafish pos-terior lateral line. Dev. Dynam. 242, 832e846.

Hebert, J.M., 2011. FGFs: neurodevelopment's Jack-of-all-trades e how do they do it?Front. Neurosci. 5, 133.

Heller, S., Bell, A.M., Denis, C.S., Choe, Y., Hudspeth, A.J., 2002. Parvalbumin 3 is anabundant Ca2þ buffer in hair cells. J. Assoc. Res. Otolaryngol. 3, 488e498.

Huang, S.M., Mishina, Y.M., Liu, S., Cheung, A., Stegmeier, F., et al., 2009. Tankyraseinhibition stabilizes axin and antagonizes Wnt signalling. Nature 461, 614e620.

Jacques, B.E., Montcouquiol, M.E., Layman, E.M., Lewandoski, M., Kelley, M.W., 2007.Fgf8 induces pillar cell fate and regulates cellular patterning in the mammaliancochlea. Development 134, 3021e3029.

Jacques, B.E., Dabdoub, A., Kelley, M.W., 2012a. Fgf signaling regulates developmentand transdifferentiation of hair cells and supporting cells in the basilar papilla.Hear. Res. 289, 27e39.

Jacques, B.E., Puligilla, C., Weichert, R.M., Ferrer-Vaquer, A., Hadjantonakis, A.K.,Kelley, M.W., Dabdoub, A., 2012b. A dual function for canonical Wnt/b-cateninsignaling in the developing mammalian cochlea. Development 139,4395e4404.

Jacques, B.E., Montgomery IV, W.H., Uribe, P.M., Yatteau, A., Asuncion, J.D.,Resendiz, G., Matsui, J.I., Dabdoub, A., 2014. The role of Wnt/b-catenin signalingin proliferation and regeneration of the developing basilar papilla and lateralline. Dev. Neurobiol. 74, 438e456.

Jayasena, C.S., Ohyama, T., Segil, N., Groves, A.K., 2008. Notch signaling augmentsthe canonical Wnt pathway to specify the size of the otic placode. Development135, 2251e2261.

Jiang, L., Romero-Carvajal, A., Haug, J.S., Seidel, C.W., Piotrowski, T., 2014. Gene-expression analysis of hair cell regeneration in the zebrafish lateral line. Proc.Natl. Acad. Sci. U. S. A. 111, E1383eE1392.

Jiang, L., Jin, R., Xu, J., Ji, Y., Zhang, M., Zhang, X., Zhang, X., Han, Z., Zeng, S., 2016.Hair cell regeneration or the expression of related factors that regulate the fatespecification of supporting cells in the cochlear ducts of embryonic and post-hatch chickens. Hear. Res. 332, 17e28.

Kelley, M.W., 2006. Regulation of cell fate in the sensory epithelia of the inner ear.Nat. Rev. Neurosci. 7, 837e849.

Kiernan, A.E., Pelling, A.L., Leung, K.K., Tang, A.S., Bell, D.M., Tease, C., Lovell-Badge, R., Steel, K.P., Cheah, K.S., 2005. Sox2 is required for sensory organdevelopment in the mammalian inner ear. Nature 434, 1031e1035.

Kim, D., Pertea, G., Trapnell, C., Pimentel, H., Kelley, R., Salzberg, S.L., 2013. TopHat2:accurate alignment of transcriptomes in the presence of insertions, deletionsand gene fusions. Genome Biol. 14, R36.

Ku, Y.C., Renaud, N.A., Veile, R.A., Helms, C., Voelker, C.C., Warchol, M.E., Lovett, M.,2014. The transcriptome of utricle hair cell regeneration in the avian inner ear.J. Neurosci. 34, 3523e3535.

Leng, N., Dawson, J.A., Thomson, J.A., Ruotti, V., Rissman, A.I., Smits, B.M., Haag, J.D.,Gould, M.N., Stewart, R.M., Kendziorski, C., 2013. EBSeq: an empirical Bayeshierarchical model for inference in RNA-seq experiments. Bioinformatics 29,1035e1043.

Lewis, J., 1998. Notch signalling and the control of cell fate choices in vertebrates.Semin. Cell Dev. Biol. 9, 583e589.

Li, H., Corrales, C.E., Wang, Z., Zhao, Y., Wang, Y., Liu, H., Heller, S., 2005. BMP4signaling is involved in the generation of inner ear sensory epithelia. BMC Dev.Biol. 5, 1e11.

Li, W., Wu, J., Yang, J., Sun, S., Chai, R., Chen, Z.Y., Li, H., 2015. Notch inhibition in-duces mitotically generated hair cells in mammalian cochleae via activating theWnt pathway. Proc. Natl. Acad. Sci. U. S. A. 112, 166e171.

Liang, J., Wang, D., Renaud, G., Wolfsberg, T.G., Wilson, A.F., Burgess, S.M., 2012. Thestat3/socs3a pathway is a key regulator of hair cell regeneration in zebrafish.J. Neurosci. 32, 10662e10673.

Liu, Z., Walters, B.J., Owen, T., Brimble, M.A., Steigelman, K.A., Zhang, L., MelladoLagarde, M.M., Valentine, M.B., Yu, Y., Cox, B.C., Zuo, J., 2012. Regulation ofp27Kip1 by Sox2 maintains quiescence of inner pillar cells in the murineauditory sensory epithelium. J. Neurosci. 32, 10530e10540.

Ma, E.Y., Rubel, E.W., Raible, D.W., 2008. Notch signaling regulates the extent of haircell regeneration in the zebrafish lateral line. J. Neurosci. 28, 2261e2273.

Millimaki, B.B., Sweet, E.M., Dhason, M.S., Riley, B.B., 2007. Zebrafish atoh1 genes:classic proneural activity in the inner ear and regulation by Fgf and Notch.Development 134, 295e305.

Mizutari, K., Fujioka, M., Hosoya, M., Bramhall, N., Okano, H.J., Okano, H., Edge, A.S.,2013. Notch inhibition induces cochlear hair cell regeneration and recovery ofhearing after acoustic trauma. Neuron 77, 58e69.

Mohammadi, M., McMahon, G., Sun, L., Tang, C., Hirth, P., Yeh, B.K., Hubbard, S.R.,Schlessinger, J., 1997. Structures of the tyrosine kinase domain of fibroblastgrowth factor receptor in complex with inhibitors. Science 276, 955e960.

Munnamalai, V., Fekete, D.M., 2013. Wnt signaling during cochlear development.Semin. Cell Dev. Biol. 24, 480e489.

Munnamalai, V., Fekete, D.M., 2016. Notch-Wnt-Bmp crosstalk regulates radialpatterning in the mouse cochlea in a spatiotemporal manner. Development 143,4003e4015.

Ni, W., Zeng, S., Li, W., Chen, Y., Zhang, S., Tang, M., Sun, S., Chai, R., Li, H., 2016. Wntactivation followed by Notch inhibition promotes mitotic hair cell regenerationin the postnatal mouse cochlea. Oncotarget 11479.

Ohta, S.H., Johnson, R., Wu, D.K., 1996. Differential expression of bone morphoge-netic proteins in the developing vestibular and auditory sensory organs.J. Neurosci. 16, 6463e6475.

Ohta, S., Mansour, S.L., Schoenwolf, G.C., 2010. BMP/SMAD signaling regulates thecell behaviors that drive the initial dorsal-specific regional morphogenesis ofthe otocyst. Dev. Biol. 347, 369e381.

Ohta, S., Wang, B., Mansour, S.L., Schoenwolf, G.C., 2016. BMP regulates regionalgene expression in the dorsal otocyst through canonical and non-canonicalintracellular pathways. Development 143, 2228e2237.

Oishi, N., Schacht, J., 2011. Emerging treatments for noise-induced hearing loss.Expert Opin. Emerg. Drugs 16, 235e245.

Petrovic, J., G�alvez, H., Neves, J., Abell�o, G., Giraldez, F., 2015. Differential regulationof Hes/Hey genes during inner ear development. Dev. Neurobiol. 75, 703e720.

Pirvola, U., Zhang, X., Mantela, J., Ornitz, D.M., Ylikoski, J., 2004. Fgf9 signalingregulates inner ear morphogenesis through epithelial-mesenchymal

L. Jiang et al. / Hearing Research 361 (2018) 66e79 79

interactions. Dev. Biol. 273, 350e360.Pujades, C., Kamaid, A., Alsina, B., Giraldez, F., 2006. BMP-signaling regulates the

generation of hair-cells. Dev. Biol. 292, 55e67.Roberson, D.W., Kreig, C.S., Rubel, E.W., 1996. Light microscopic evidence that direct

transdifferentiation gives rise to new hair cells in regenerating avian auditoryepithelium. Audit. Neurosci. 2, 195e205.

Roberson, D.W., Alosi, J.A., Cotanche, D.A., 2004. Direct transdifferentiation givesrise to the earliest new hair cells in regenerating avian auditory epithelium.J. Neurosci. Res. 78, 461e471.

Schimmang, T., 2007. Expression and functions of FGF ligands during early oticdevelopment. Int. J. Dev. Biol. 51, 473e481.

Schimmang, T., Maconochie, M., 2016. Gene expression profiling of the inner ear.J. Anat. 228, 255e269.

Schuck, J.B., Sun, H., Penberthy, W.T., Cooper, N.G., Li, X., Smith, M.E., 2011. Tran-scriptomic analysis of the zebrafish inner ear points to growth hormonemediated regeneration following acoustic trauma. BMC Neurosci. 12, 409e427.

Selimoglu, E., 2007. Aminoglycoside-induced ototoxicity. Curr. Pharm. Des 13,119e126.

Shang, J., Cafaro, J., Nehmer, R., Stone, J., 2010. Supporting cell division is notrequired for regeneration of auditory hair cells after ototoxic injury in vitro.J. Assoc. Res. Otolaryngol. 11, 203e222.

Su, Y.X., Hou, C.C., Yang, W.X., 2015. Control of hair cell development by molecular

pathways involving Atoh1, Hes1 and Hes5. Gene 558, 6e24.Tao, L., Segil, N., 2015. Early transcriptional response to aminoglycoside antibiotic

suggests alternate pathways leading to apoptosis in sensory hair cells in themouse inner ear. Front. Cell. Neurosci. 9, 190.

Taylor, R.R., Nevill, G., Forge, A., 2008. Rapid hair cell loss: a mouse model forcochlear lesions. J. Assoc. Res. Otolaryngol. 9, 44e64.

Trapnell, C., Williams, B.A., Pertea, G., Mortazavi, A., Kwan, G., van Baren, M.J.,Salzberg, S.L., Wold, B.J., Pachter, L., 2010. Transcript assembly and quantifica-tion by RNA-seq reveals unannotated transcripts and isoform switching duringcell differentiation. Nat. Biotechnol. 28, 511e515.

Wu, D.K., Oh, S.H., 1996. Sensory organ generation in the chick inner ear. J. Neurosci.16, 6454e6462.

Wu, J., Li, W., Lin, C., Chen, Y., Cheng, C., Sun, S., Tang, M., Chai, R., Li, H., 2016. Co-regulation of the Notch and Wnt signaling pathways promotes supporting cellproliferation and hair cell regeneration in mouse utricles. Sci. Rep. 6, 29418.

Yorgason, J.G., Fayad, J.N., Kalinec, F., 2006. Understanding drug ototoxicity: mo-lecular insights for prevention and clinical management. Expert Opin. Drug Saf5, 383e399.

Zine, A., Van De Water, T.R., de Ribaupierre, F., 2000. Notch signaling regulates thepattern of auditory hair cell differentiation in mammals. Development 127,3373e3383.