The Notch Ligand JAG1 Is Required for Sensory Progenitor Development in the Mammalian Inner Ear

In mammals, six separate sensory regions in the inner ear are essential for hearing and balance function. Each sensory region is made up of hair cells, which are the sensory cells, and their associated supporting cells, both arising from a common progenitor. Little is known about the molecular mechanisms that govern the development of these sensory organs. Notch signaling plays a pivotal role in the differentiation of hair cells and supporting cells by mediating lateral inhibition via the ligands Delta-like 1 and Jagged (JAG) 2. However, another Notch ligand, JAG1, is expressed early in the sensory patches prior to cell differentiation, indicating that there may be an earlier role for Notch signaling in sensory development in the ear. Here, using conditional gene targeting, we show that the Jag1 gene is required for the normal development of all six sensory organs within the inner ear. Cristae are completely lacking in Jag1-conditional knockout (cko) mutant inner ears, whereas the cochlea and utricle show partial sensory development. The saccular macula is present but malformed. Using SOX2 and p27kip1 as molecular markers of the prosensory domain, we show that JAG1 is initially expressed in all the prosensory regions of the ear, but becomes down-regulated in the nascent organ of Corti by embryonic day 14.5, when the cells exit the cell cycle and differentiate. We also show that both SOX2 and p27kip1 are down-regulated in Jag1-cko inner ears. Taken together, these data demonstrate that JAG1 is expressed early in the prosensory domains of both the cochlear and vestibular regions, and is required to maintain the normal expression levels of both SOX2 and p27kip1. These data demonstrate that JAG1-mediated Notch signaling is essential during early development for establishing the prosensory regions of the inner ear.


Introduction
The mammalian inner ear is a complex structure consisting of a coiled cochlea, three orthogonally positioned semicircular canals, a central vestibule, and a dorsally projecting endolymphatic duct and sac. With the exception of the endolymphatic duct and sac, the different parts of the ear all contain sensory organs populated by sensory hair cells and their associated supporting cells. There are three different categories of sensory organs: cristae, located at the base of each semicircular canal; maculae, housed within the central vestibule; and the organ of Corti, which lines the cochlear duct. Only one sensory organ, the organ of Corti, is required for hearing; the other five organs are important for balance. Unfortunately, in mammals, if these regions are damaged due to an environmental or genetic insult, they cannot regenerate, leaving a permanent hearing and/or balance impairment.
Although some progress has been made in understanding how the individual cell types within the sensory areas of the ear are formed [1,2], little is known about the molecular mechanisms that establish the prosensory lineage and how the different sensory organ types are formed. Interestingly, the molecular mechanisms that underlie sensory differentiation in the vertebrate inner ear demonstrate strong parallels with Drosophila sense organ development [3][4][5]. For example, during Drosophila external sense organ development, lateral inhibition mediated by Notch signaling is required to restrict the adoption of the sensory organ precursor cell fate, which then gives rise to the entire sensory organ [6][7][8]. Similarly, in the vertebrate ear, lateral inhibition mediated by Notch signaling appears to be important for restricting the number of cells that can adopt the hair cell fate [9][10][11][12][13][14][15]. Lineage analysis has also shown that, at least in the chicken, hair cells and supporting cells arise from a common progenitor [16], consistent with an equipotent epithelium that undergoes lateral signaling to specify cell fates. Unlike in Drosophila, which has a single Notch receptor and two ligands (Delta and Serrate/Jagged), in mammals Notch signaling pathway components include four receptors (Notch 1-4) and five ligands (Delta-like [DLL] 1, 3, and 4, and Jagged [JAG] 1 and 2; for reviews, see [8,[17][18][19]). In the mouse, both DLL1 and JAG2 are expressed in nascent hair cells [10,20] and act synergistically during lateral inhibition [15]. Both DLL1 and JAG2 appear to signal through the NOTCH1 receptor [15]. In Drosophila, prosensory regions that undergo lateral inhibition are first delineated by expression of members of the atonal or acheate-scute family of basic helix-loop-helix transcription factors. Further parallels with Drosophila have arisen with the finding that an atonal homolog, the Math1 gene, is required for hair cell differentiation [21][22][23]. However, the situation is not entirely similar to Drosophila, as MATH1 does not appear to be required to establish the prosensory regions [22,24]. Instead, it has turned out that another type of transcription factor, the HMG-box factor SOX2, is required for sensory organ formation in the inner ear [25]. This finding does not demonstrate direct parallels with Drosophila sense organ formation, since the Drosophila SOX2 homologs SoxNeuro and Dicheate have no known role in peripheral sense organ formation. Instead, both genes play a role in the formation of neural progenitor cells in the central nervous system [26,27]. Consistent with a prosensory role, SOX2 marks sensory progenitors early in development and acts upstream of the Math1 gene during sensory organ formation in the ear [25]. Interestingly, one of the Notch ligands, JAG1, is expressed early in the prosensory regions of the ear [4,20], indicating that Notch signaling also may play a role in early sensory organ formation. Consistent with this finding, mice heterozygous for N-ethyl-N-nitrosourea-induced point mutations in the Jag1 gene show mild sensory organ defects in the ear [28,29]. Unfortunately, embryos homozygous for available mutant alleles of the Jag1 gene do not survive past embryonic day (E) 11.5 due to vascular defects, precluding an analysis of their inner ears. To circumvent this early lethality, we have created a conditional allele of the Jag1 gene using Cre/loxP technology. Using the Foxg1-Cre mouse line to express Cre recombinase in the early otocyst [30,31], we have disrupted JAG1 function in the ear and show that sensory formation in the inner ear is severely attenuated in these mutants. Analysis of the patterns of hair and supporting cell formation in the Jag1-conditional knockout (cko) inner ears suggests that fewer progenitors form in Jag1-cko inner ears. This result is confirmed by analysis of the prosensory markers SOX2 and p27 kip1 , which are down-regulated as early as E12.5, indicating that the Jag1 gene acts early during sensory progenitor formation. These data demonstrate an early role for Notch signaling in establishing the sensory progenitors of the inner ear.

Creation of a Conditional Allele of the Jag1 Gene
We created an allele for conditional inactivation of JAG1 function by flanking the Delta-Serrate-Lag2 (DSL) domainencoding exon (exon 4) of the Jag1 gene with loxP sites (Figure 1). The DSL domain has been shown to be the region of the DLL and JAG proteins that interacts with Notch family receptors [32,33]; therefore, removing this region of the gene should create a nonfunctional protein. To demonstrate that this allele encodes a nonfunctional protein, we crossed Jag1 flox /þ mice to mice expressing Cre recombinase in the female germline under control of the Zp3 promoter (Zp3-Cre mice); the Zp3-Cre mouse strain has been shown to express Cre recombinase in the growing oocyte prior to the completion of the first meiotic division [34]. Female Zp3-Cre/ þ; Jag1 flox /þ offspring were then crossed to male B6 mice to produce offspring that were heterozygous for the deleted region of the floxed allele (designated Jag1 del2 /þ; see Materials and Methods). Heterozygous Jag1 del2 /þ mice were intercrossed, and Jag1 del2 /Jag1 del2 homozygous offspring were analyzed for defects between E9.5 and E11.5. Jag1 del2 /Jag1 del2 mutant embryos exhibited the same vascular phenotype we described previously in embryos homozygous for a targeted Jag1 null allele, Jag1 del1 [35]. Specifically, the Jag1 del2 /Jag1 del2 mutant embryos exhibited yolk sac vascular remodeling defects and cranial hemorrhaging, and often exhibited an enlarged pericardial sac (Figure 2A-2D). All Jag1 del2 /Jag1 del2 mutant embryos were necrotic by E11.5, and most showed vascular defects by E10.5. RT-PCR of cDNA synthesized from the mutant embryos using primers that span the floxed exon 4 demonstrated that this region was deleted, as expected ( Figure 2E). These data demonstrate that deletion of the Jag1 flox allele yields a nonfunctional Jag1 mutant allele.

Inactivation of Jag1 Function within the Ear
To disrupt JAG1 function within the inner ear, we crossed Jag1 flox /Jag1 flox mice with mice doubly heterozygous for the Foxg1-Cre allele (these mice express Cre recombinase throughout the otocyst, as well as forebrain, eye, and foregut) [30] and the Jag1 del1 allele [35]. Offspring with the genotype Foxg1-Cre/þ; Jag1 del1 /Jag1 flox (hereafter designated Jag1-cko) survived through E18.5 and were analyzed for inner ear defects. We examined the patterns of Cre-mediated excision in Jag1-cko embryos at E10.5 and in cochleae at E16.5-E18.5 by in situ hybridization using a probe that specifically detected the deleted exon 4 ( Figure 3). These results showed that expression in the otocyst was weak or absent by E10.5 ( Figure 3B). In addition, analysis of cochlear expression at later stages showed no expression at E16.5 ( Figure 3F) and E18.5 (unpublished data). These data indicate that, as previously shown for conditional deletion of a Fibroblast growth factor receptor 1 (Fgfr1) floxed allele [31], the Foxg1-Cre line efficiently deletes the Jag1 flox allele early during inner ear development.

Malformation of the Inner Ear in Jag1-cko Mutants
To examine the morphology of the Jag1-cko inner ears, paintfilling of the inner ears of mutants and controls was

Synopsis
Deafness and adult-onset hearing loss are significant health problems. In most cases, deafness or vestibular dysfunction results when the sensory cells in the inner ear, known as hair cells, degenerate due to environmental or genetic causes. In the mammalian inner ear, the hair cells and their associated supporting cells can be found in six different patches that have particular functions related to hearing or balance. Unfortunately, unlike in birds or fish, mammalian hair cells show little ability to regenerate, resulting in a permanent hearing or balance disorder when damaged. Here, the authors show that a protein called JAG1, a ligand in the Notch signaling pathway, is required for the normal development of all six sensory regions in the mammalian inner ear. In ears that lacked JAG1, some of the sensory patches were missing completely, whereas others were small and lacked particular cell types. The authors showed that JAG1 is required by the sensory precursors, progenitor cells that give rise to both the hair cells and the supporting cells. By understanding how the sensory areas develop normally, it is hoped that molecular tools can be developed that will aid sensory regeneration in the mammalian inner ear.
performed at E15.5 ( Figure 4). Results of this analysis showed a severe disruption in the structure of the Jag1-cko inner ears compared to their littermate controls ( Figure 4C and 4D). Specifically, the semicircular canals were largely absent, with the exception of a portion of the anterior and lateral semicircular canals. In addition, the utricle appeared small, the saccule was misshapen, and the cochlea was undercoiled. In contrast, the parts of the inner ear that are not associated with sensory formation, including the endolymphatic duct and sac and the common crus, appeared relatively unaffected.

Sensory Defects in Jag1-cko Mutant Inner Ears
Since the Jag1 gene is expressed in the sensory areas of the ear, and because the structural malformations observed in Jag1-cko mutant inner ears appeared to primarily affect regions of the ear that contained sensory organs, we examined the sensory regions of the ear for defects. We examined the organ of Corti, the sensory organ of the cochlea, at E18.5 by scanning electron microscopy (SEM) ( Figure 5). By this stage, all hair cells within the organ of Corti have exited the cell cycle, and most are well-differentiated although not fully mature [36]. Severe hair cell patterning defects were apparent by SEM within the Jag1-cko mutant cochleae. This phenotype was most striking in the basal turns of the cochlea, where no hair cell formation was observed ( Figure 5D). In the midbasal regions of the organ of Corti, hair cells formed in patches, within which there was no clear formation of rows or distinction between inner and outer hair cells ( Figure 5F). More apically, hair cells appeared more continuous along the organ of Corti ( Figure 5H). However, although hair cells were present in the apical region, their numbers were clearly reduced; rather than the normal, perfectly ordered four rows of hair cells, there were only two rows of loosely arranged hair cells of indistinct type.

Abnormal Hair and Supporting Cell Patterns in Jag1-cko Inner Ears
To determine which sensory cell types were differentiating in the Jag1-cko mutant cochleae, specific markers were used to identify hair cell and supporting cell subtypes throughout the ear ( Figure 6). In the cochlea, we used an antibody against MYO7A to label all hair cells and an antibody against S100A1 to label inner hair cells, Deiter's supporting cells, and inner phalangeal supporting cells [24]. When both markers were used in combination, inner hair cells, outer hair cells, and some supporting cell types could be distinguished ( Figure  6A-6F). This analysis showed that in the apex of the cochlea, inner hair cells were present and usually formed as doublets ( Figure 6B). Their associated supporting cells, the inner phalangeal cells, were also present. Outer hair cells and their associated supporting cells, the Deiter's cells, were not present in this region. In the middle portions of the cochlea, both inner and occasionally outer hair cells were present, although their patterning was clearly abnormal ( Figure 6D). In addition, the tunnel of Corti was not apparent, and there were often doublets of inner hair cells and increased numbers of outer hair cell rows without accompanying Deiter's supporting cells. As shown by SEM, both hair cells and supporting cells were absent in the very basal regions of the cochlea ( Figure 6F).
Using the same markers we also examined the vestibular sensory organs in Jag1-cko mutant inner ears ( Figure 6G-6J). Consistent with the lack of semicircular canal and ampulla ) in the Jag1 del2 /Jag1 del2 yolk sacs (B) similar to other Jag1 loss-of-function mutants. (C and D) E11 embryos demonstrating a small, necrotic Jag1 del2 /Jag1 del2 embryo (D). White arrow in (D) indicates an enlarged pericardial sac (enlarged, inset), which is frequently observed in mutants exhibiting cardiovascular defects. RT-PCR results using primers that span exon 4 using RNA extracted from E10.5 control (þ) and ZP3-Cre deleted embryos (À) (E). The upper band (541 bp) indicates that the wild-type allele is present. The lower bands (286 bp) indicates the Jag1 del2 mutant allele that does not contain exon 4. DOI: 10.1371/journal.pgen.0020004.g002 Figure 3. Conditional Jag1 Inactivation Using the Foxg1-Cre Line (A-D) Low-and high-power views of E10 embryos processed for wholemount in situ hybridization using a Jag1 exon 4-specific probe. White arrows (A) point to the Jag1 signal in the otocyst (left arrow) and the eye (right arrow), two structures where Cre recombinase is expressed. In Jag1-cko mutants at E10, this signal is either absent or extremely weak. Black arrows (A and B) point to expression in the spinal cord and nephric duct, regions where Cre recombinase expression has not been reported in Foxg1-Cre mice. However, expression is consistently weaker in these areas in Jag1-cko embryos, indicating that there may be low levels of widespread expression of Cre recombinase in Jag1-cko embryos. In (D), the otocyst and the eye are outlined by a dotted line. Very little expression is observed in these regions, consistent with Foxg1-Cre expression. (E and F) In situ hybridization of E16.5 cochleae demonstrating Jag1 expression in wild-type (E) and Jag1-cko cochleae (F), where expression is entirely absent. Scale bars ¼ 500 lm. DOI: 10.1371/journal.pgen.0020004.g003 formation observed by paintfilling, there was no evidence of crista formation. The Jag1-cko utricular macula was extremely small with very few differentiating hair cells ( Figure 6H). Surprisingly, the saccule and its macula were only mildly affected in the Jag1-cko inner ears ( Figure 6J). Hair cell differentiation appeared relatively unaffected, although the entire saccular structure was shaped differently than in the controls, a feature that was also observed in the paintfilled specimens (see Figure 4C and 4D). These data show that all sensory organs within the inner ear are affected to varying degrees in Jag1-cko inner ears. However, some sensory organs, such as the cristae, appear to be more sensitive to the loss of JAG1 function.
To examine whether aberrant hair cell patterning in Jag1cko cochleae was due to defects in hair cell formation or in subsequent differentiation, we examined hair cell patterning at an earlier stage (E16.5). Using a lectin that binds to hair cell stereocilia, we examined whether the patterns of hair cell formation at E16.5 looked similar to the patterns at E18.5 ( Figure 7). At E16.5 in wild-type cochleae, a gradient of hair cell differentiation was evident ( Figure 7A, 7C, 7E, and 7G); in the basal regions both inner and outer hair cells could be recognized (unpublished data), while in the middle regions only inner hair cells were clearly detected by most markers (Figure 7A and 7C). In the more apical regions, little to no hair cell differentiation had taken place by this stage ( Figure  7E and 7G). In Jag1-cko cochleae, the patterns looked similar to those at E18.5, with patches of hair cells in the midbasal regions ( Figure 7B and 7D) and a complete absence of hair cells in the very basal regions ( Figure 7B). These data suggest that the Jag1-cko mutants have defects in hair cell formation rather than differentiation. In addition, the apical regions in the Jag1-cko cochleae did not appear more differentiated than the controls ( Figure 7E-7H), arguing against precocious differentiation as an explanation for the reduced numbers of hair cells observed in the mutant cochleae.

Disrupted Prosensory Development in Jag1-cko Inner Ears
To determine how the JAG1 ligand functions during sensory development, we used several markers of the  prosensory domain, including p27 kip1 and SOX2, and examined their expression patterns in both wild-type and Jag1-cko mutant cochleae (Figure 8). At E14.5, the majority of hair cells and supporting cells in the organ of Corti have completed their final division, and hair cells are beginning to differentiate in the basal portions of the cochlea [36]. p27 kip1 , a cell-cycle inhibitor, is required for the cochlear sensory progenitors to exit the cell cycle on time, and is an established marker of the prosensory domain in the cochlea [22,37]. p27 kip1 begins to be expressed in a discrete domain within the cochlea as the hair cells and supporting cells exit the cell cycle around E13.5 to E14.5 ( Figure 8A, 8B, 8D, and 8E). Recently it has been shown that the SRY-related transcription factor SOX2 is required for establishment of the prosensory regions in the inner ear [25]. Using fluorescence immunocytochemistry double labeling, we examined the relationship between these markers and JAG1 protein expression in both wild-type and Jag1-cko cochleae. As previously reported [22], JAG1 was not expressed within the prosensory domain as assessed by p27 kip1 expression at E14.5, but instead was expressed immediately adjacent (possibly with some slight overlap) in the inner (neural) portion of cochlea (Kö lliker's organ; Figure 8A and 8D). In contrast, SOX2 did show a largely overlapping domain with p27 kip1 (Figure 8B and 8E), as originally described [25]. However, the SOX2 expression domain was slightly larger than the p27 kip1 domain, extending into Kö lliker's organ and overlapping with the JAG1 domain. Despite the fact that JAG1 was not expressed within the prosensory domain at E14.5, both p27 kip1 and SOX2 expression was absent in the basal regions of the cochlea (Figure 8C), indicating that prosensory formation is already disrupted in these ears. In the apex, weak expression of both markers was observed ( Figure 8F), consistent with the fact that some sensory differentiation occurs in this region of the Jag1-cko cochlea.
In order to determine if JAG1 is ever expressed in the prosensory region of the cochlea, we examined an earlier age (E12.5) and compared the JAG1 domain to the SOX2 domain (since p27 Kip1 is not expressed in the inner ear prior to E13.5 to E14.5). Adjacent sections from both wild-type and Jag1-cko cochleae were immunostained to detect either JAG1 or SOX2 protein ( Figure 9). This analysis showed that in the basal regions of the wild-type cochlea, where sensory progenitors were still dividing, JAG1 expression did overlap with the SOX2 domain ( Figure 9A and 9B), indicating that JAG1 is initially expressed within the prosensory domain. However, in the apical regions, where the sensory precursors have ceased dividing, expression of JAG1 and SOX2 did not overlap ( Figure 9D and 9E). In the Jag1-cko cochlea, SOX2 was absent from the basal regions and significantly downregulated in the apical regions ( Figure 9C and 9F). These data demonstrate that JAG1 is expressed within the prosensory domain of the cochlea at early stages, and that, in the absence of JAG1 function, sensory formation is disrupted prior to cell cycle exit and differentiation of sensory hair cells and nonsensory supporting cells.
We also compared JAG1 and SOX2 expression in the vestibular regions of the inner ear in both wild-type and Jag1cko mutant embryos. JAG1 and SOX2 exhibited largely overlapping expression domains that corresponded to the locations of the five sensory organs in the vestibular portion of the ear (Figure 10). The two expression domains only differed significantly in the anterior and posterior cristae, where JAG1 expression had a negative patch in the middle of its expression domain, whereas SOX2 expression did not show this same patch ( Figure 10A, 10B, 10G, and 10H). The JAG1-negative region may correspond to the eminentia cruciatum, a nonsensory region present in the middle of both the anterior and posterior cristae, although it is not clear why SOX2 would be expressed there. In the Jag1-cko vestibular sensory patches, SOX2 expression was consistent with the patterns of sensory differentiation observed at E18.5. For example, the Jag1-cko saccule displayed fairly normal SOX2 expression ( Figure 10C), consistent with the almost normal development of the saccular macula. In contrast, SOX2 expression in the utricle was very weak and the expression domain was much smaller than in controls ( Figure   10F), consistent with the severe disruption of differentiation of the utricular macula in Jag1-cko inner ears. There was no SOX2 expression in the Jag1-cko cristae, and in fact the entire ampullae appeared to be missing or severely disrupted even at this early stage ( Figure 10C and 10I; dotted line regions), consistent with the lack of cristae and ampullae observed at later stages.

Discussion
We have demonstrated that Notch signaling, mediated by the JAG1 ligand, is required early in development for the formation of the sensory regions of the ear. By comparing expression of JAG1 to two markers of the prosensory domain, SOX2 and p27 kip1 , we have shown that JAG1 marks all prosensory regions of the ear from early time points (E12.5), but becomes down-regulated in the organ of Corti by E14.5, when the sensory progenitors exit the cell cycle and begin differentiating into hair cells and supporting cells. Both SOX2 and p27 kip1 are down-regulated in the affected prosensory regions of the Jag1-cko inner ear, demonstrating that JAG1 is necessary for the development of early sensory progenitors in the inner ear.

Distinctive Patterns of Hair Cell Formation in Jag1-cko Inner Ears Suggest Progenitor Cell Numbers Are Reduced
One intriguing result from our studies was that the six sensory regions were not equally affected by the loss of Jag1 function. For example, in the Jag1-cko vestibular system, the cristae were lacking altogether, and only a small number of hair cells differentiated in the utricular maculae. In contrast, the saccular maculae exhibited little disturbance in hair cell formation, although the overall shape of the organ was abnormal. In the Jag1-cko cochlea, hair cell differentiation patterns varied based on their apical or basal location. For example, in the apical regions of the cochlea only inner hair cells formed, and these were often arranged in multiple rows rather than the normal single row. In the middle and midbasal turns of the cochlea, patches of hair cells with nonsensory intervening regions were frequently observed. Within these patches, outer hair cells were sometimes present, although the patterning was abnormal and S100A1labeled Dieter's cells were not present. In the very basal regions of the cochlea, neither hair cells nor supporting cells were present.
The patches of hair cells found in the basal regions of the cochlea and the differential effect of the mutation on the basal and apical portions of the cochlea were particularly interesting, as similar defects have been found in at least two other mouse mutants of genes known to play a role in the generation of the sensory precursors of the ear. For example, both a hypomorphic allele and a conditionally deleted allele of the Fgfr1 gene exhibited patches of hair cells in portions of the cochlea [31]. Similar to the Jag1-cko phenotype, these patches in the Fgfr1 conditional mutants contained mostly inner hair cells that were often arranged in multiple rows, with very few outer hair cells. Unlike the Jag1-cko phenotype, Fgfr1 function was required only in the cochlea. Another mouse mutant, a hypomorphic allele of the Sox2 gene (yellow submarine; Sox2 ysb ), also displayed patches of hair cells in the basal portions of the cochlea and a milder phenotype in the apical regions of the cochlea [25]. More similar to the Jag1-cko phenotype, SOX2 was required for both the cochlear and vestibular sensory regions [25]. The finding that primarily inner hair cells differentiate in these mutant cochleae may be due to the fact that inner hair cells are the first to differentiate [10,37], suggesting that if there are reduced numbers of progenitor cells, they would likely differentiate as inner rather than outer hair cells. Similarly, the milder phenotype in the apical regions of these mutants may be due to the fact that cells exit the cell cycle earliest in the apex [36]. This may mean that, if there are reduced numbers of progenitor cells, they would reside in the apical rather than the basal portions of the cochlea.
The multiple rows of inner hair cells observed in Jag1-cko and other mutants could be explained by a number of different scenarios. One possibility is that the multiple rows are not a result of actual increases in inner hair cell numbers, but rather are caused by defects in their eventual arrangement due to the shorter Jag1-cko cochlea. Recent studies of mouse mutants with defects in planar cell polarity and convergent extension (a term referring to the intercalation of cells, leading to growth of tissue in one dimension in the absence of proliferation) indicate that multiple rows of hair cells can be obtained in this way and are frequently observed in the apical regions of the cochlea [38][39][40][41]. An alternative possibility is that the multiple rows are a result of a second, later function of the JAG1 ligand, distinct from its prosensory role described here. A third possibility is that the downregulation of p27 kip1 , a protein that inhibits continued proliferation of the precursor cells in the cochlea, leads to continued cell division of the remaining sensory progenitors, ultimately resulting in excess numbers of inner hair cells in the regions where they form. Taken together, these data suggest that sensory progenitors are reduced in the Jag1-cko inner ears and that Notch signaling, fibroblast growth factor signaling, and the transcription factor SOX2 all act in either common or parallel pathways involved in the production of sensory progenitors in the inner ear.

A Prosensory Role for Notch in the Ear
An examination of early prosensory markers, including p27 Kip1 and SOX2, demonstrated that prosensory establishment is disrupted in Jag1-cko inner ears, consistent with the suggestion that progenitors are reduced in these mutants. Our data show that JAG1 plays an early prosensory role in ear development, quite unlike the role played later during development by the other Notch ligands, DLL1 and JAG2, which are involved in lateral signaling and differentiation [10,15]. These data are consistent with an early role for Notch signaling in progenitor cell maintenance in the inner ear. In a number of other systems, including the nervous system and more recently in the intestinal epithelium, it has been demonstrated that Notch signaling is involved in maintaining cells in an undifferentiated state [42][43][44][45][46]. In the mammalian nervous system it has been shown that loss of Notch signaling leads to premature differentiation and a reduction in the progenitor pool [42]. Consistent with these findings, in vitro studies have demonstrated that the frequency of neurosphere production was reduced in Notch signaling mutants [47,48], indicating a loss of stem cell potential. Moreover, studies have also shown that Notch signaling promotes radial glial identity, a cell type that has been shown to act as a progenitor cell in the central nervous system [49][50][51][52]. Our results suggest that, similar to the nervous system, Notch signaling via JAG1 is important for sensory precursor formation or maintenance in the inner ear. However, unlike the nervous system, we see no evidence for precocious differentiation, suggesting instead that JAG1 may affect the specification, survival, or proliferative capacity of the sensory precursors.
Recent evidence from the chick indicates that JAG1 may be important for the initial sensory specification events. By expressing a constitutively active form of Notch (Notch1-ICD), Daudet et al. [14] demonstrated that ectopic sensory patches could be induced, indicating that early Notch signaling may be important for the induction of sensory areas, and not just for their maintenance. However, it should be noted that ectopic sensory areas formed only in certain areas of the ear, indicating that some sensory competence is required for this effect. A similar result was obtained by overexpressing an activated form of b-catenin, an essential component of the canonical Wnt signaling pathway, in the chicken inner ear [53]. As in the Notch1-ICD studies, ectopic sensory regions were obtained, but again, only in certain regions of the ear. However, unlike the Notch gain-offunction studies, overexpression of b-catenin also led to a change in sensory region character (i.e., cochlear to vestibular), indicating that Wnt signaling governs not only whether a sensory region will form but also the type of sensory region that will form. In Drosophila, interactions between Notch and Wingless, a member of the Wnt family of signaling molecules, are well established [54,55], and evidence of an interaction has begun accumulating in vertebrates as well [45,56,57]. Bone morphogenetic protein (BMP) signaling may also be important for sensory formation, particularly for the sensory cristae, as BMP4 has been shown to mark the mouse cristae from very early in development [58]. Experiments in the chicken have shown that blocking BMP signaling sometimes leads to disturbances in sensory development [59]. Taken together, these data indicate that, based on expression patterns, previous studies, and the evidence presented here, JAG1 is the ligand responsible for the prosensory function of the Notch pathway in the ear. Furthermore, the Notch pathway likely interacts with other signaling pathways such as the Wnt, FGF, and BMP pathways to create sensory organs of the proper size, organization and character.

Sensory Formation Still Occurs in Jag1-cko Inner Ears
One somewhat puzzling question is that, if JAG1 is important for sensory progenitor development, why does any sensory formation occur in Jag1-cko inner ears? One possibility is that another Notch ligand is compensating for the loss of JAG1 function. This explanation seems unlikely since none of the other Notch ligands shows a similar expression pattern to JAG1 in the ear. For example, both the Dll1 and Jag2 genes are expressed in nascent hair cells after they exit the cell cycle and begin differentiating. However, in addition to hair cell expression, there is also early expression of the Dll1 gene in the anteroventral portion of the otocyst at about E10.5 [4,20], that likely overlaps with at least part of the JAG1 domain (see Figure 2) [60]. This expression domain has previously been thought to be related to the formation of the neuroblasts that delaminate from the otic epithelium and later differentiate into the neurons that will innervate the hair cells [4]. It has been shown in zebrafish that correct neuroblast formation requires Notch-mediated lateral signaling [9]; however, in mammals it has not been shown definitively that this is the role that the Dll1 gene plays at early stages. This leaves open the possibility that this early domain of DLL1 expression may be at least partially involved in prosensory specification, similar to the JAG1 expression domain.

Nonsensory Defects in Jag1-cko Inner Ears
In addition to the defects in sensory formation in the Jag1cko inner ears, the mutant inner ears also exhibited nonsensory defects. Specifically, the semicircular canals were largely absent, with the exception of portions of the anterior and lateral canals. In addition, all three ampullae were absent, the utricle was small, and the cochlea was undercoiled. Based on recent studies, it is likely that these defects are secondary to the sensory defects. For example, it has been shown that FGFs expressed in the sensory cristae promote semicircular canal formation through up-regulation of BMP2 [61]. Thus, loss of the cristae would be expected to have a severe affect on canal formation. Emerging evidence from mouse mutants has demonstrated that genes involved in sensory formation result in severely malformed inner ears. For example, mutations in the Sox2 gene lead to malformations very similar to those described here in Jag1-cko mutants. The inner ears of embryos homozygous for two different mutant alleles of Sox2, Sox2 lcc/lcc and Sox2 ysb/ysb , showed disrupted canal formation; smaller utricular and saccular compartments; and thinner, undercoiled cochleae [25]. In addition, FGF10 mouse knockouts also showed disrupted cristae development associated with loss of canal structures [62]. However, unlike the canal structures, cochlear formation does not appear to be strictly dependent on development of the organ of Corti, as a cochlea, albeit short and thin, will form in the absence of any sensory formation [25]. However, normal cochlear length appears to be dependent on sensory formation, at least partially through convergent extension. Recently, a number of genes have been found in the cochlea that lead to defects in planar cell polarity as well as a shortened cochlea, presumably because of defects in convergent extension [38,41]. Therefore it is likely that the shortened cochlea observed in Jag1-cko mutants is at least partially a result of failure of convergent extension caused by a reduction in the number of sensory precursors.
The data presented here demonstrate that the Jag1 gene is required for sensory precursor development in the inner ear. Further studies are required to establish the exact role that JAG1-mediated Notch signaling plays in early sensory progenitors, and also its relationship to the roles played by FGF signaling and SOX2 expression. Understanding how the sensory precursors form is an important prerequisite for regeneration studies that may provide molecular tools to treat hearing loss and vestibular disorders [63].

Materials and Methods
Construction of the Jag1 floxneo allele. To construct the Jag1 floxneo allele, bacterial artificial chromosome clones containing the Jag1 genomic locus were isolated from a RPCI-22 (129S6/SvEvTac) mouse bacterial artificial chromosome library (filters obtained from Research Genetics) by hybridization to a 1.8-kb mouse Jag1 cDNA probe. To make the shorter 59 homology region of the targeting vector, a 2.2kb KpnI fragment upstream of exon 4 was isolated, blunt-ended, and subcloned into the SmaI site of a modified pBS vector that contained a loxP-FRT-PGKneo-FRT cassette. A 1.5-kb KpnI fragment that contained exon 4 was also subcloned into the loxP-FRT-PGKneo-FRT cassette. To construct the longer 39 homology region, a 3.5-kb KpnI-SmaI fragment containing exon 5 was blunt-ended and subcloned into the EcoRV site of another modified pBS vector that contained a single loxP site. A 3.5-kb SmaI-SalI fragment from this construct was then cloned into the SmaI-XhoI site of a pKO 905 vector containing a diphtheria toxin gene for negative selection. A 5.7-kb SalI-NotI fragment from the loxP-FRT-PGKneo-FRT construct was then cloned into the SalI-NotI sites of the pKO 905 vector containing the 39 homology region to generate the final Jag1 floxneo targeting vector (see Figure 1).
Generation of Jag1 flox mice. The Jag1 floxneo targeting construct was linearized with Not1 and electroporated into CJ7 embryonic stem (ES) cells, as described previously [64]. DNA from 288 ES cell clones was screened by PCR using an internal/external primer set, and positive clones were then confirmed by Southern blot by probing EcoRI-digested DNA with an external 1.7-kb Stu1-EcoRI fragment located 39 to the targeting construct (see Figure 1). This probe also detected partial recombination events in which the distal loxP site was lost; in these cases a slightly larger fragment (11 kb rather than 9.3 kb) was obtained (see Figure 1A). The presence of the distal loxP site was further confirmed by PCR using primers that flanked the loxP site (DSLF and J1LoxR1; see below for sequences). Correctly targeted clones were injected into C57BL/6J (B6) blastocysts, and chimeric mice were obtained. Chimeric male mice were mated to B6 females and the agouti progeny were assayed for the presence of the Jag1 floxneo allele by PCR using Jag1 floxneo specific primers. Jag1 floxneo /þ mice were intercrossed to create homozygous Jag1 floxneo /Jag1 floxneo offspring. Homozygous Jag1 floxneo /Jag1 floxneo mice appeared normal and healthy, suggesting that the neomycin resistance cassette (PGKneo) did not adversely affect Jag1 expression. To control for any possible effects from the presence of the PGKneo cassette, the FRT-flanked PGKneo cassette was deleted by mating Jag1 floxneo mice to FLPe-recombinase expressing mice (Gt[ROSA]26Sor tm1(FLP1)Dym ; Jackson Laboratory, Bar Harbor, Maine, United States) to produce Jag1 flox mice. Both Jag1 floxneo and Jag1 flox mice were used in these experiments, and no differences in the resulting phenotype were observed. To differentiate the deleted form of this allele from our previously reported Jag1 null mutant allele (Jag1 del1 ) [35], we designate the Jag1 allele generated by Cre recombinase-mediated deletion of the Jag1 flox or Jag1 floxneo alleles the Jag1 del2 allele.
Paintfilling and scanning electron microscopy. The paintfilling of the Jag1-cko inner ears was performed at E15.5. The technique was performed as previously described [28]. Inner ears were prepared for SEM as described previously using a version of the osmium tetroxidethiocarbohydrazide method [65]. Specimens were examined with a Hitachi 3000N scanning electron microscope (Hitachi, Tokyo, Japan).
In situ hybridization and RT-PCR. Since the Jag1 del2 mutant allele creates an in-frame deletion of the DSL domain, an in situ probe was designed for detection of the floxed region of the Jag1 flox allele by in situ hybridization. The probe was created by amplifying a 433bp product encompassing exon 4 (primers: J1-420F, 59-CGACCG TAATCGCATCGTAC-39 and J1-853R, 59-ATGCACTTGTCGCAGTA CAG-39) and subcloning the product into the PCR II vector (Invitrogen). For whole-mount embryos in situ, embryos were fixed overnight in 4% paraformaldehyde. In situ hybridization was performed essentially as described [66]. For cochlear in situ, inner ears were dissected from the head and fixed overnight in 4% paraformaldehyde. After washing in PBS, the bony shell and stria were removed from the cochleae and the samples were dehydrated in methanol. In situ hybridization was performed as described [67], with the exception of the posthybridization washes, which were done according to [68]. For RT-PCR, total RNA was extracted from the E10.5 control and Jag1 del2 /Jag1 del2 embryos, using an RNAeasy kit (Qiagen, Valencia, California, United States) and following the manufacturer's instructions. First-strand cDNA synthesis was performed using the AMV reverse transcriptase (Promega, Madison, Wisconsin, United States) with specific primer J1-961R (59-AGTCC CACAGTAATTCAGATC-39). Products were amplified from cDNA using primers that flanked exon 4 (J1-420F, 59-CGACCGTAATCG CATCGTAC-39 and J1-961R).