Developmental Gene Expression Profiling along the Tonotopic Axis of the Mouse Cochlea

The mammalian cochlear duct is tonotopically organized such that the basal cochlea is tuned to high frequency sounds and the apical cochlea to low frequency sounds. In an effort to understand how this tonotopic organization is established, we searched for genes that are differentially expressed along the tonotopic axis during neonatal development. Cochlear tissues dissected from P0 and P8 mice were divided into three equal pieces, representing the base, middle and apex, and gene expression profiles were determined using the microarray technique. The gene expression profiles were grouped according to changes in expression levels along the tonotopic axis as well as changes during neonatal development. The classified groups were further analyzed by functional annotation clustering analysis to determine whether genes associated with specific biological function or processes are particularly enriched in each group. These analyses identified several candidate genes that may be involved in cochlear development and acquisition of tonotopy. We examined the expression domains for a few candidate genes in the developing mouse cochlea. Tnc (tenacin C) and Nov (nephroblastoma overexpressed gene) are expressed in the basilar membrane, with increased expression toward the apex, which may contribute to graded changes in the structure of the basilar membrane along the tonotopic axis. In addition, Fst (Follistatin), an antagonist of TGF-β/BMP signaling, is expressed in the lesser epithelial ridge and at gradually higher levels towards the apex. The graded expression pattern of Fst is established at the time of cochlear specification and maintained throughout embryonic and postnatal development, suggesting its possible role in the organization of tonotopy. Our data will provide a good resource for investigating the developmental mechanisms of the mammalian cochlea including the acquisition of tonotopy.


Introduction
The ability of the mammalian cochlea to respond to sounds of different frequencies is attributable to numerous specializations along the longitudinal (tonotopic) axis of the cochlea [1,2]. Hair cells in the organ of Corti are the mechanosensory receptors that exhibit different sensitivities to specific frequencies depending on their position along the cochlear duct. Their morphological and mechanoelectrical properties gradually differ along the tonotopic axis, which appears to be critical for the frequency tuning [2]. Spiral ganglion neurons, which relay sound information from the hair cells in the form of electrical impulses to the brain, also display different morphological and electrophysiological characteristics along the length of the cochlea [1,3]. Differences in synaptic structure, function, and number along the tonotopic axis have also been reported [4]. Besides the sensory receptors and neurons, nonsensory structures such as the basilar membrane and tectorial membrane also exhibit graded changes in composition and mechanical features. In fact, the basilar membrane, which vibrates in response to sound, has been considered as a primary determinant that stimulates the hair cells at a specific position with respect to a specific frequency [5]. The tectorial membrane, a viscoelastic structure that covers the stereociliary bundles of hair cells, has also been suggested to play a crucial role in facilitating tonotopic discrimination [6]. Despite numerous structural and cellular specializations in the cochlea that are thought to contribute to tonotopy, the molecular mechanisms that establish these graded specializations during development remain largely unknown.
Several studies have shown that genes and their isoforms are differentially expressed along the base and apex of the cochlea in both mammals and birds. Recently, the microarray technique has been utilized to obtain more comprehensive gene expression profiles along the tonotopic axis of the cochlea [7][8][9]. Sato and colleagues identified genes that are differentially expressed between basal and apical halves of the cochlea in adult mice [7]. In addition, several microarray analyses have compared gene expression levels along the tonotopic axis of the basilar papilla, the avian auditory organ, at hatching or 2-week-old chickens [8,9]. These gene expression profiles will aid in understanding the tonotopic characteristics of the mature cochlea. However, since these analyses were performed after onset of hearing and the cochlea is thought to be mature already, they may not provide insight into how tonotopy is established.
In mice, the cochlea continues to develop after birth. The hair cells in the organ of Corti undergo dramatic architectural remodeling and achieve mechanosensitivity during the first week of postnatal development [10][11][12]. Also, the stereociliary bundles on the apical surface of the outer hair cells attain mature height and shape, leading to the characteristic organization of a stair-case pattern [10,12]. The apical surface of the outer hair cells changes progressively from a rounded-hexagon to a non-convex circumference that correlates with the V-shape of their overlying stereociliary bundles [11]. These morphological changes are closely associated with redistribution of cytoskeletal proteins and spatiotemporal expression patterns of genes involved in mechanotransduction, which leads to maturation of the mechanosensitivity of the hair cells [10,12,13]. Moreover, the supporting cells in the organ of Corti also undergo major morphological and physiological changes during this period [14]. In particular, progenitor-associated genes such as Notch effectors are downregulated in supporting cells during postnatal development [15,16]. The loss of Notch signaling may account for diminishment of stem cell features and lack of regeneration potential in the adult mouse cochlea [17,18]. In contrast, genes associated with supporting cell maturation and function such as the intermediate filament glial fibrillary acidic protein (GFAP) and the glutamate transporter GLAST are upregulated during neonatal development [19,20].
Interestingly, the morphological and functional maturation progresses gradually from the base of the cochlea toward the apex [10][11][12]21]. Such changes precede the onset of auditory function, which suggests that the postnatal development may be crucial to achieving the tonotopic characteristics in the mammalian cochlea. Thus, we obtained comprehensive gene expression profiles along the tonotopic axis at postnatal day (P) 0 and P8, and compared expression profiles along the tonotopic axis (base, mid, apex) as well as between the neonatal stages (P0, P8). Our data should provide a platform for future investigations into the developmental mechanisms of the mammalian cochlea.

Results and Discussion
Hair Cell Maturation Along the Tonotopic Axis during Neonatal Cochlear Development It has been shown that hair cells in the mouse cochlea continue to differentiate after birth [10][11][12]. They acquire mature morphological and mechanoelectrical characteristics during the first postnatal week and hair cell differentiation progresses from the base of the cochlea towards the apex. We examined the maturation process of the hair cells during early postnatal stages in mice ( Fig. 1). At P0, the basal half of the cochlea was organized into one row of inner and three rows of outer hair cells with distinct stereociliary hair bundles, while the apical region of the cochlea exhibited multiple rows of unorganized immature hair cells (Fig. 1A), At P4, the entire organ of Corti was in an orderly array with stereocilia of all outer hair cells displaying the characteristic V-shaped (Fig. 1A). The angle of the V-shaped hair bundles was wider at the base and smaller towards the apical cochlea (Fig. 1A). This gradual change in bundle angle is clearly shown in the graph plotted as a function of relative distance from the basal end (Fig. 1B&C). At P8, the morphologies of the hair bundles were almost identical with those of P4, but the angles were generally slightly wider and comparable to those of P21 (Fig. 1C). These results showed that hair cells in the developing cochlea gradually mature in a base-to-apex gradient, and the tonotopic variations along the cochlear duct become evident during neonatal development.

Gene Expression Profiling Along the Tonotopic Axis during Neonatal Development
Our results and others suggest that neonatal development is crucial for maturation and acquisition of the tonotopic characteristics in the mouse cochlea ( Fig. 1) [10,11]. Genes that play important roles in these processes may be differentially expressed along the cochlear duct during neonatal stages. In order to obtain comprehensive gene expression profiles along the tonotopic axis of the mouse cochlea, we utilized the microarray technique. Approximately eighty temporal bones were dissected from P0 and P8 mouse pups and cochlear tissues containing the organ of Corti and lateral wall were collected. The cochlear tissues were divided into three equal pieces representing the base, middle and apex, and pooled separately (P0-base, P0-middle, P0-apex, P8base, P8-middle, P8-apex). The same dissection procedure was repeated to obtain a second independent set of pooled cochlear samples from P0 and P8 mice. RNAs were purified from the two biological replicates and were independently subjected to microarray analyses.
The microarray data were analyzed to identify differentially expressed genes along the tonotopic axis at each stage (P0 or P8) by comparing the average expression levels of individual genes. Genes showing at least a 1.5-fold difference (p,0.05) between the base and the apex were considered to be differentially expressed. Interestingly, for most of the genes that were differentially expressed between the base and the apex, expression levels were gradually changed along the cochlear duct, exhibiting either an increasing or decreasing gradient towards the apex. Only a small fraction among the differentially expressed genes (,3%) showed peak expression in the middle cochlear region. Table 1 shows the top fifty genes that showed higher expression in the base or apex at P0 or P8, ranked by the average fold differences (a complete list of the differentially expressed genes can found in Table S1). The lists contain many genes previously associated with inner ear development or function. For example, Tectb, Isl1, Foxg1, Ush2a, Pcdh15, and Atoh1 showed higher expression in the apex at P0, while Gjb6, Cldn11, and Otoa showed higher expression in the base at P0. At P8, genes showing higher expression in the apex included Tecta, Tectb, Tnc, Fgf10, Smpx, Slitrk6, and Otog, while genes showing higher expression in the base included Crym and Otos. In addition, the lists contain many other genes that have not been previously associated with inner ear or hearing function.

Classification of Gene Profiles Based on Spatial and Temporal Expression Patterns
We noticed that there were many genes that exhibited similar or different gradients at P0 and P8 (Table 1 and Table S1). For example, Fst, Otog and Tectb showed higher expression in the apex at both P0 and P8, but Trpa1, P2rx3 and Atoh1 were expressed more in the apex at P0, but no longer differentially expressed along the tonotopic axis at P8. Thus, we further classified the gene profiles according to the spatial (along the tonotopic axis) as well as temporal (between P0 and P8) expression patterns. The entire gene profiles were classified into three groups: 1) genes showing no gradient along the tonotopic axis, 2) genes showing increasing gradient towards the apex and 3) genes showing decreasing gradient towards the apex. Then, the expression patterns were compared between P0 and P8, which yielded nine groups displaying distinct expression patterns ( Figure 2). This analysis revealed that a majority of genes (.95%) were expressed relatively constantly along the tonotopic axis at both P0 and P8. Among these genes (no gradient at both P0 and P8), gene showing significant fold-changes between P0 and P8 (fold-change .2, p,0.05) were classified as Group A: some of them were downregulated (Group A1) or up-regulated (Group A2) during this period. Among the genes showing no gradient at P0, some genes displayed an increasing (Group B) or decreasing (Group C) gradient at P8. Genes showing an increasing gradient toward the apex at P0 were also subcategorized into three groups, comprising genes no longer showing a gradient at P8 (Group D), retaining the increasing gradient (Group E) or showing the opposite (decreasing) gradient at P8 (Group F). Finally, genes showing a decreasing gradient toward the apex at P0 were similarly subcategorized as genes no longer showing a gradient at P8 (Group G), showing the opposite (increasing) gradient (Group H) or retaining the decreasing gradient at P8 (Group I). Examples of genes included in each group were shown in Table 2 (the complete list of the genes in each group can found in Table S2). The possible functional significance of the genes in each group is discussed below.

Validation of the Microarray Data with Quantitative RT-PCR
We validated our microarray data by performing quantitative real time PCR (qRT-PCR) on 22 genes selected from various groups ( Fig. 3 and Table 3). Genes for qRT-PCR were selected among ones showing higher fold changes in each group such as Stmn2, Tecta, Tnmd, Fst, Atp6v0a4, Hoxa2, and A2m, and also among ones previously shown to be associated with cochlear development or function such as Kcnj10, Kcnq1, Ptprq, Tectb, Otog and Ush1c. As an example, qRT-PCR results for Fst (Follistatin: Group E) showed an increasing gradient toward the apex at both P0 and P8 (Fig. 3A). When the qRT-PCR results were plotted as a function of the values from the microarray, close correlation was clearly shown (Fig. 3B, Pearson correlation coefficient, r = 0.982). Similarly, all 22 selected genes generally showed close correlations between the microarray data and qRT-PCR results (Pearson correlation coefficient, r = 0.862) (Fig. 3C).
In Table 3, we calculated fold changes between the apex and base with the results from qRT-PCR, and compared to those from the microarray. For genes from Group A showing no gradient along the tonotopic axis, average expression levels between P0 and P8 were compared. The results from qRT-PCR appeared to be more sensitive than the microarray data in verifying the changes  Figure 2. Schematic representation of distinct gene expression patterns along the tonotopic axis at P0 and P8. Gene profiles obtained from the microarray were classified based on the spatial (along the tonotopic axis) as well as temporal (between P0 and P8) expression changes. Expression levels of individual genes were compared between base and apex at P0, and classified into three categories: relatively constant along the cochlear duct (average fold-changes ,1.5), higher in the apex or base (average fold-changes .1.5). Then, genes in each category were subdivided into three groups base on the expression patterns at P8 (constant, higher in the base or apex), which yielded nine distinct groups. See text for details. doi:10.1371/journal.pone.0040735.g002 between the base and the apex. Nevertheless, the expression patterns from the qRT-PCR results were generally consistent with those from the microarray.

Functional Annotation Clustering Analysis
Our microarray analysis yielded an extensive list of genes that are differentially expressed along the tonotopic axis during neonatal development. Analysis of the dynamic expression patterns of individual genes may highlight some key players that are critical to cochlear maturation including the acquisition of tonotopy (Table 2 and Table S2). However, these single-gene analyses may fail to identify important biological processes, which are often regulated collectively by a group of genes involved in common biological function or pathways. Also, biologically significant gene expression changes occurring in a certain group of cells can be missed due to masking by genes expressed in other cell types or inherent noise signals. Thus, in order to evaluate our microarray data at the level of biological function and processes, we performed functional annotation clustering analysis using the Database for Annotation, Visualization, and Integrated Discovery (DAVID) bioinformatics resources [22]. This analysis identified biological functions and processes specifically enriched in the gene list of each group, which were summarized in Table 4.
In Group A1 (no gradient, down-regulated at P8), genes associated with cell cycle process or chromosomal activity were enriched. Such enrichments indicated that cell proliferation occurring at P0 is generally diminished at P8, which is consistent with active maturation processes observed during neonatal cochlear development. Interestingly, genes associated with microtubule cytoskeleton such as kinesin family members were also shown to be down-regulated at P8, which may be related with a reduced cell cycle and with disassembly of the microtubule-based kinocilium in the hair cells during this period.
On the other hand, Group A2 (no gradient, up-regulated at P8) contained many genes associated with extracellular matrix and cell adhesion such as Col4a3, Col4a4 and Emilin2. Upregulation of these genes may explain the maturation of the cytoarchitecture in the cochlea during the postnatal period. Consistent with our results, Emilin2 was shown to be rapidly accumulated during postnatal development in the extracellular matrix surrounding the collagenous fibers in the basilar membrane [23]. Genes associated with ion transport such as Kcnq1, Kcne1 and Kcnj10 were also enriched. It has been shown that Kcnq1 and Kcne1 are expressed in strial marginal cells and that Kcnj10 in intermediate cells plays a critical role in potassium recycling in the endolymph [24]. Thus, enrichment of such gene sets in P8 cochlea is consistent with the establishment of endocochlear potential during the first postnatal week in the mouse cochlea [25].
In Group B (no gradient at P0, increasing base-to-apex gradient at P8), genes associated with neuron differentiation and plasma membrane were enriched. Several genes involved in cochlear development and function were included in this group such as Tecta, Gata2, Slitrk6, Chrna10, Nrcam, Otof and Otog. The higher expression of these genes in the apex at P8 may be related to the base-to-apex wave of cochlear maturation, since the increasing base-to-apex gradient for some genes resulted from downregulation at the base by P8. In support of this notion, two genes in this group, Ptprq and Tecta, which are both important for inner ear development, were no longer expressed in the adult cochlea [26,27]. Interestingly, several genes related to non-sensory structures such as the basilar or tectorial membrane were included in this group. Tecta and Otog, which encode for a-tectorin and otogelin, respectively, that are components of the tectorial membrane [28], were expressed at higher levels in the apex. Tectb, encoding b-tectorin and classified in Group E, also showed higher expression in the apex at P8. The higher expression levels of genes associated with the tectorial membrane in the apex may be related with the decreasing base-to-apex gradient in stiffness of the tectorial membrane, an important factor for tonotopic discrimination [29].
In Group C (no gradient at P0, decreasing base-to-apex gradient at P8), genes related to extracellular matrix were enriched. The biological annotation for proteinaceous extracellular matrix was also shown to be enriched in Group A2 with different sets of genes, suggesting that different types of extracellular matrix components are required at different stages of postnatal cochlear development. Also enriched were genes associated with cation transport. Interestingly, splicing variants of Kcnma1, which encodes for the Maxi-K + (BK) channel have been shown to be differentially expressed along the tonotopic axis and also during developmental stages in mice and chicken [8,30]. Kcnma1 expression patterns from our microarray data, showing no gradient at P0 and higher expression in the base at P8, did not closely match with previous reports. This may be because the microarray technique did not distinguish the splicing variants, but displayed a summed expression level of all the variants expressed at the time of analysis. Since differential expressions of specific splice variants have been suggested to contribute to tonotopy [2], it may be necessary to profile the entire transcriptome along the tonotopic axis using deep-sequencing technologies such as the RNA-seq approach [31].
In Group D (increasing base-to-apex gradient at P0, no gradient at P8), enriched genes were related with multicellular organismal responses to stress including P2rx3 and Trpa1. Consistent with our data, P2rx3, encoding the P2X 3 subunit of ATP-gated ion channels, was shown to be expressed at higher levels in the apical cochlea at P0, and its expression is gradually decreased and no longer detected after the onset of hearing [32]. Similarly, expression of Trpa1, a subunit for transient receptor potential (TRP) channels, was greatly increased in the apex at around P0 and then decreased to show no gradient by P6 [13]. These expression changes precede the onset of hearing, suggesting their roles in the acquisition of mature cochlear function such as the mechanosensitivity of the hair cells.
In Group E (increasing base-to-apex gradient at both P0 and P8), genes associated with neuron differentiation and sensory organ development were enriched, which is consistent with the ongoing maturation processes in the apical cochlea during neonatal stages (Fig. 1) [10,11].
In Group G (decreasing base-to-apex gradient at P0, no gradient at P8), genes related to ion transport and cell adhesion were enriched. For most genes in this group, the expression gradient at P0, higher in the base, was lost at P8 due to increased expression of these genes towards the apex during the postnatal development. Since ion transport and cell adhesion molecules are closely associated with cochlear remodeling and function, this expression pattern may be related with the base-to-apex wave of cochlear maturation. Consistently, these gene sets were also shown to be enriched in Group A2, in which genes are up-regulated at P8, showing no gradient.
Finally, in Groups F (higher expression in the apex at P0 and in the base at P8), H (higher expression in the base at P0 and in the apex at P8) and I (higher expression in the base at both P0 and P8), no functionally related genes were significantly enriched.

Confirmation of Expression Patterns for Selected Genes by in Situ Hybridization
We next performed in situ hybridization for selected genes to confirm their differential expression patterns along the tonotopic axis. We selected a few genes previously shown to be expressed in the cochlea such as Chrna10, Ptprq, Kcnj10, Tectb, and Tnc, as well as a few genes that exhibited higher fold changes along the tonotopic axis among the newly identified genes from our analysis such as A2m, Nov, and Fst.
We first performed in situ hybridization on Chrna10, Ptprq, Kcnj10, and Tetcb, whose expression patterns have previously been reported, to confirm our microarray results [26,27,33,34]. Expression domains of these genes within the neonatal cochlea were consistent with the previous reports, and their graded expression patterns along the tonotopic axis generally matched with the microarray data (see Fig. S1 and Fig. S2 for details). Expression pattern of A2m was also consistent with the microarray data (Fig. S2).
We next examined the expression patterns of the few newly identified genes from our microarray analysis. We found that Nov (nephroblastoma overexpressed gene), identified in Group B, was expressed in the basilar membrane with no obvious gradient at  P0 and in an increasing base-to-apex gradient at P8, consistent with the microarray results (Fig. 4C,D, arrows). The Nov (also called as CCN3) gene encodes for a matricellular protein that associates with the extracellular matrix and mediates a variety of cellular functions including proliferation, differentiation, survival, migration and cytoskeleton reorganization [35]. Since the structural organization of the extracellular matrix regulates the mechanical properties of the basilar membrane, gradual changes in Nov expression may contribute to the frequency tuning of the basilar membrane [5]. The expression pattern of Nov in the basilar membrane was compared with that of Tnc, which encodes the glycoprotein tenascin C and has been shown to be expressed in the developing mouse cochlea including the basilar membrane [36].
Tnc was strongly expressed in the hair cells and the basilar membrane along the cochlear duct at P0 (Fig. 4A,B). While Tnc expression in the hair cells was downregulated at P8 (Fig. 4A,B, arrowheads), its expression in the basilar membrane continued in an increasing base-to-apex gradient at P8, consistent with the microarray results ( Table 1). The specific expression in the basilar membrane with gradual changes along the cochlear duct of Nov and Tnc suggest that these extracellular matrix proteins may be involved in the developmental regulation of the basilar membrane, specifying differential properties along the tonotopic axis. We also examined the expression patterns of Fst (Follistatin), a TGF-b/BMP antagonist, which was identified in Group E and displayed high fold-changes along the tonotopic axis at both P0 and P8 (Table 1 and 2). Fst expression was barely detected in the basal turn of the cochlear duct, began to be observed in the lateral side of the lesser epithelial ridge (LER) in the middle turn and then expanded to the entire LER in the apical turn at both P0 and P8 (Fig. 5A,B, brackets). Since Bmp4 and Bmp7 are also expressed in the LER region (Fig. 5D) [37], the gradual expansion of the Fst expression domain implies gradual decreases of BMP activities toward the apex. We then asked when this graded Fst expression pattern is established. At E15.5 when cochlear morphogenesis is almost complete, the graded expression patterns of Fst were comparable to those observed at P0 and P8 (Fig. 5C). Thus, we examined the inner ears at E10.5, when the primordial cochlear structure just begins to form in the ventral aspect of the otocyst. When examined from dorsal to ventral sections of the otocyst, Fst expression was not detected in the vestibular structures (Fig. 5E1-E3), and began to be faintly observed at the level where Bmp4 and Lfng were expressed in the lateral and medial side of the developing cochlea, respectively (Fig. 5F1-F3) [38]. Fst expression became stronger and broader toward the apical cochlear sections, while Bmp4 and Lfng expression patterns were not changed ( Fig. 5G1-H3, brackets). However, at E9.5 when the cochlear anlage has yet to be developed, Fst expression was not detected in the otocyst (data not shown). These results indicate that the longitudinal gradient of Fst expression is established at the time of cochlear specification and maintained throughout cochlear development including embryonic and neonatal stages. Considering the role of BMP signaling in hair cell differentiation and inner ear development [37,39,40], the graded expression patterns of Fst suggest a possible role for Follistatin in the specification of gradually different properties along the cochlear duct such as the tonotopy.

Conclusions
In this study, we applied the microarray technique to identify genes that are differentially expressed along the longitudinal axis of the mouse cochlea during neonatal development. Our analyses identified several candidate genes, which may be involved in cochlear development and acquisition of tonotopy, but have not been appreciated previously. Although the biological relevance of these candidates in cochlear development remains to be determined, the gene expression profiles obtained in this study will be a useful resource for studying various aspects of mammalian cochlear development.

Tissue Dissection and RNA Extraction
All animals were handled in accordance with the guidelines for the Care and Use of Laboratory Animals of Yonsei University College of Medicine. The protocol for obtaining cochlear samples was approved by the Committee on Animal Research at Yonsei University College of Medicine (#10-104). Eighty temporal bones were dissected from C57BL/6 mice at P0 and P8. After removing the otic capsule, the cochlear tissues including the organ of Corti, spiral limbus, and lateral wall was dissected in sterile, chilled Hanks' balanced salt solution under a stereomicroscope (Olympus SZX10). The cochlear tissues were divided into three equal pieces representing the base, middle and apex, and pooled separately (P0base, P0-middle, P0-apex, P8-base, P8-middle, P8-apex). The pooled cochlear tissues were homogenized with a micro-homogenizer (Microtec, Chiba, Japan), and total RNA was extracted using an RNeasy Mini Kit (Qiagen, Hilden, Germany). The quality and quantity of the extracted total RNA was measured using the Total RNA Nano Assay (Agilent Technologies, Palo Alto, CA, USA). The same procedure was repeated to obtain additional biological replicates for each cochlear sample (P0-base, P0-middle, P0-apex, P8-base, P8-middle, P8-apex).

Microarray Data Analysis
Gene expression profiles were generated using the Affymetrix GeneChipß Mouse Gene 1.0 ST oligonucleotide arrays (Affymetrix, Santa Clara, California). RNA samples isolated from two biological replicates were sent to the local Affymetrix GeneChipß service provider (DNA LINK, Seoul, Korea). Three hundred nano-grams of high quality total RNA was converted into doublestranded cDNA using a random hexamer conjugated with a T7 promoter sequence. Amplified RNA (cRNA) was generated from the double-stranded cDNA using an in vitro transcription (IVT) reaction and then re-converted into cDNA with a mix of dNTPs containing dUTP. This cDNA was then fragmented by uracil-DNA glycosylase (UDG) and APE1 restriction endonucleases and labeled with a biotinylated dideoxynucleotide via a terminal transferase reaction. End-labeled cDNA was hybridized for 16 hours at 45u and 60 rpm according to the Gene Chip Whole Transcript (WT) Sense Target Labeling Assay Manual (Affymetrix). After hybridization, the chips were stained and washed in a Genechip Fluidics Station 450 (Affymetrix) and scanned using a Genechip Array scanner 3000 7G (Affymetrix). Affymetrix Command Console Software version 1.1 was used to process the raw images into CEL files. Gene expression estimates were normalized using the robust multiarray averaging (RMA) method. Microarray data were deposited in the GEO database (#GSE34187) and followed MIAME requirements.
Genes showing an average fold-change greater than 1.5 (p,0.05, unpaired t-test) between the base and the apex and exhibiting a log-transformed gene expression level greater than 6 Figure 5. Expression patterns of Fst in the cochlea during embryonic and neonatal development. Fst transcripts were present in the lesser epithelial ridge (LER) region of the cochlea in an increasing base-to-apex gradient during neonatal (A,B) and embryonic development (C-H). (A,B) At P0 and P8, Fst expression was barely detectable in the basal cochlear turns (A1,B1, asterisks), but was evident in the lateral side of the LER in the middle turns (A2,B2, brackets) and its expression domain included the entire LER region in the apical turns (A3,B3, brackets). Insets in A1-B3 show expression domains of Myo15 or Atoh1 in the adjacent sections to indicate the location of hair cells. Arrowheads indicate the lateral border of differentiating hair cells. (C,D) Fst expression patterns at E15.5 were comparable to those observed at P0 and P8. Fst expression domains largely overlapped that of Bmp4 in the LER area. (E1-H3) The increasing gradient of Fst expression patterns in the cochlea was already apparent in the cochlear primordium at E10.5 (F3,G3,H3, asterisk, brackets). Fst expression was not detectable in the presumptive vestibular organs including the lateral crista (E1-E3). Expression domains of Bmp4 (E1,F1,G1,H1) and Lfng (E2,F2,G2,H2) indicated the lateral and medial regions of the developing cochlea, respectively. The schematic diagram indicates the level of sections for the pictures shown in E1-H3. Scale bar in A (200 mm) applies to B-D; scale bar in A1 (50 mm) applies to A1-A3, B1-B3, C1-C3, and D1-D3; Scale bar in E1 (100 mm) applies to E1-H3. doi:10.1371/journal.pone.0040735.g005 were considered to be significantly differentially expressed. The differentially expressed genes were further classified into nine groups based on the fold-changes along the tonotopic axis (base vs. apex) and during neonatal development (P0 vs. P8) (Fig. 2). Gene lists in each group were subjected to functional annotation clustering analysis using the Database for Annotation, Visualization, and Integrated Discovery (DAVID) (http://david.abcc. ncifcrf.gov/home.jsp). The analysis was performed by following the procedure previously described [22]. Biological annotations with an enrichment score of .1.5, P-value (,0.01) and false discovery rate (FDR, #5%) were considered biologically interesting (Table 4).

Quantitative Real-time PCR
Quantitative real-time PCR (qRT-PCR) was performed to confirm the microarray data. RNAs purified from three biological replicates for each of the cochlear samples (two biological replicates used for the microarray analysis and one additional (third) biological replicate collected on a small scale) were used for qRT-PCR. cDNAs were synthesized from each RNA (2 mg) sample using poly (dT) 20 and the Superscript III RT kit according to manufacturer's instructions (Invitrogen). No-RT controls were not included. cDNAs were diluted three times with water and the same amount of cDNAs were used for all PCR reactions. qRT-PCR for 22 selected genes was performed in triplicate wells using SYBR Green PCR Master Mix and the ABI 7500 machine (Applied Biosystems, Carlsbad, California). The samples were denatured at 94uC for 10 minutes and then subjected to 40 cycles of denaturing at 94uC for 15 seconds and annealing and extension at 60uC for 1 minute. We normalized gene expression level (DCt) with b-actin (Actb). Relative expression levels were determined by comparative methods (DDCt). Relationships between the microarray data and qRT-PCR were assessed with Pearson correlation coefficient (r) ( Table 3). Primers used for qRT-PCR were designed by using the Primer3 program [41]. At least one primer (either forward or reverse) was designed to include an exon-exon junction sequences to avoid amplification of genomic DNA. All the primer sequences are listed in Table S3.

Phalloidin Staining and Measurement for the Angles of Stereociliary Bundles
Stereociliary bundles were visualized by staining the organ of Corti with Alexa FluorH 488 Phalloidin (Invitrogen) at P0, P4, P8 and P21. Pictures were taken from 10%, 30%, 50%, 70%, and 90% positions from the basal end of the cochlea duct using a confocal microscopy (Zeiss LSM700). The angle formed by two lines drawn along the medial side of the hair bundles of outer hair cells was measured using the Image J program (National Institutes of Health, USA) (Fig. 1B). Angles for at least thirty outer hair cells were measured for each cochlear region and plotted as a function of relative distance from the basal end of the cochlea.

In Situ Hybridization
In situ hybridization was performed as previously described [38]. Probes for Atoh1 [42], Myo15 [43], Bmp4 and Lfng [38] were prepared as previously described. RNA probes for Fst were generated from a 313 base pair (bp) mouse Fst cDNA containing the +36 to +348 coding region (MN_008046); for Tnc, from a 404 bp mouse Tnc cDNA containing the +5835 to +6060 coding region and 178 bp 39 untranslated region (NM_011607); for Nov, from a 966 bp mouse Nov cDNA containing the 38 bp 5' untranslated region and +1 to +928 coding region (NM_010930); for Chrna10, from a 645 bp mouse Chrna10 cDNA containing the +882 to +1344 coding region and 182 bp 39 untranslated region (NM_001081424); for Ptprq, from a 455 bp mouse Ptprq cDNA containing the 2 bp 5' untranslated region and +1 to +453 coding region (NM_001081432); for Kcnj10, from a 714 bp mouse Kcnj10 cDNA containing the +749 to +1140 coding region and 322 bp 39 untranslated region (NM_001039484); for Tectb, from a 406 bp mouse Tectb cDNA containing the +801 to +990 coding region and 216 bp 39 untranslated region (NM_009348); for A2m, from a 1018 bp mouse A2m cDNA containing the 180 bp 5' untranslated region and +1 to +1000 coding region (NM_175628). Figure S1 Expression patterns of Tectb, Chrna10, and Ptprq in the cochlea during neonatal development. Expression patterns of Tectb (A,B), Chrna1 (C,D) and Ptprq (E, F) were examined by in situ hybridization at P0 (A,C,E) and P8 (B,D,F). (A,B) Tectb was classified in Group E showing increasing gradients towards the apex at both P0 and P8. Consistent with the microarray data, Tectb showed higher expression towards the apex at both P0 (A1-A3) and P8 (B1-B3). In the middle turn of P0 cochlea, Tectb expression was found in the lateral margin of greater epithelial ridge, pillar cells, and the third row of Deiter's cells, consistent with a previous report [26]. Expression domains of Tectb were either diminished in the base or expanded in the apex (A1-A3). At P8, tectb expression in the pillar cells was disappeared and its increasing gradient towards the apex maintained (B1-B3). (C, D) Chrna10 was classified in Group B showing no gradient at P0 and an increasing gradient towards the apex at P8. Consistently, Chrna10 is expressed in the differentiating hair cells relatively constantly along the tonotopic axis at P0 (C1-C3) and in an increasing gradient towards the apex at P8 (D1-D3). (E, F) Ptprq was also classified in Group B showing no gradient at P0 and an increasing gradient towards the apex at P8. However, Ptprq transcripts are observed in the differentiating hair cells with slightly higher expression in the base at P0 (E1-E3, arrowheads), which represents the qRT-PCR results more rather than the microarray data (Table 3). At P8, Ptprq transcripts are observed in an increasing gradient towards the apex at P8, consistent with the microarray and qRT-PCR results (F1-F3). Both Ptprq and Chrna10 were down-regulated in the base at P8, which may reflect the baseto-apex progression of hair cell maturation. (TIF) A2m was classified in Group I showing decreasing gradients towards the apex at both P0 and P8. Consistent with the microarray data, A2m transcripts are observed in the Reissner's membrane in a decreasing gradient toward the apex at P0 (A1-A3). Interestingly, A2m expression is no longer detected in the Reissner's membrane at P8, but is observed in the spiral ligament area retaining the decreasing gradient toward the apex (B1-B3). Kcnj10 was classified in Group A1 showing no gradient at both P0 and P8 with up-regulation at P8. Consistent with the microarray data, Kcnj10 expression was barely detected in the P0 cochlea, but clearly observed in the stria vascularis and the organ of Corti at P8.

(TIF)
Table S1 Complete list of the differentially expressed genes between apex and base of the mouse cochlear at P0 or P8. (XLSX)