The role of Six1 in mammalian auditory system development

The mammalian auditory system includes three distinct parts: the outer, the middle and the inner ears. Despite the complexity and multiple functions, the ear forms one anatomical unit that serves both hearing and equilibrium. The earliest morphological evidence for inner ear development is the otic placode,a thickened area of surface ectoderm on each side of the hindbrain. The otic placode invaginates to form the otic cup and vesicle, which subsequently undergoes proliferative growth and eventually differentiates into different regions of the inner ear. In the mouse, the basic architecture of the inner ear is fully established by E14.5 (Morsli et al., 1998).

A large number of otic genes, including transcription factors, secreted factors, receptors, cell adhesion proteins and others have been described;however, functional importance in early morphogenetic processes has only been demonstrated for some genes (Fekete and Wu,2002). The homeobox-containing genes such as the NK-related homeobox gene Nkx5.1 and the paired-box gene Pax2 are expressed in complementary patterns in the otic vesicle, with Nkx5.1dorsolaterally and Pax2 ventromedially(Herbrand et al., 1998). Mutation in the Nkx5.1 gene results in agenesis of the semicircular canals and circling behavior (Hadrys et al., 1998), while mutation in the Pax2 gene leads to agenesis of the cochlea (Torres et al.,1996). The GATA family zinc-finger gene Gata3 shows reciprocal relationships with Pax2 in the regional patterning of the early otocyst and cellular patterning within the sensory epithelia and ears of Gata3-null mouse mutants remain cystic, with a single extension of the endolymphatic duct (Karis et al.,2001; Lawoko-Kerali et al.,2002). The eyes absent gene Eya1, which encodes a transcription coactivator, is also expressed early in the otic epithelium and the inner ear development in Eya1 knockout mice arrests at the otic vesicle stage (Xu et al.,1999a). This is the first described mouse mutant lacking all sensory areas of the inner ear. Secreted factors like the Bmp-family of Tgfβ-like polypeptides, Fgfs and receptor molecules like the Fgfr2 IIIb and Fgfr1 are also expressed in the otic epithelium and serve as signaling molecules in early otic development (Chang et al., 1999; Ohuchi et al.,2000; Pirvola et al., 2000; Noramly and Grainger, 2002; Pirvola et al., 2002). Nonetheless, it is largely unknown how these genes function and respond to the inductive signals from neighboring tissues in the morphogenetic processes of inner ear development.

The murine homeobox Six gene family has been identified on the basis of sequence homology with the Drosophila sine oculis (so) gene. At present, six members (Six1-Six6) of the Six gene family have been isolated and they are suggested to interact with Pax and Eya genes based on their wide co-expression in many tissues during mammalian organogenesis and development (Oliver et al.,1995a; Oliver et al.,1995b; Kawakami et al.,1996; Chen et al.,1997; Pignoni et al.,1997; Xu et al.,1997a; Xu et al.,1997b). However, their functional roles during mammalian inner ear development have not been studied. In this study, we analyzed the expression of Six1 during inner ear development and its role in mouse auditory system development. In the developing inner ear, Six1 is expressed in all sensory epithelia. Inactivation of the Six1 gene led to malformation of the auditory system involving the outer, middle and inner ears. The inner ear development in Six1^-/- embryos arrested at the otic vesicle stage and all components of the inner ear failed to form because of increased cell death and reduced cell proliferation in the otic epithelium. Molecularly, Six1 is not required for the expression of Eya1, Pax2 and Pax8 in the otic epithelium. By contrast, Six1 is required for the normal expression of Fgf3, Fgf10, Bmp4,Gata3 and Nkx5.1 in the otic vesicle, indicating that Six1 is required for the regional specification of the otic vesicle. Finally, we provide evidence for a genetic interaction between Eya1and Six1 during inner ear development. These analyses indicate that similar to Eya1, Six1 is not required for the initiation of otic placode morphogenesis to form otic vesicle, but is required for the normal growth and regional specification of the otic vesicle.

Eya1/Six1 double heterozygous mice were generated by crossing mice carrying mutant alleles of Eya1 and Six1(Six1^lacZ) and genotyping of mice and embryos was performed as previously described (Torres et al., 1995; Xu et al.,1999a; Xu et al.,2002; Laclef et al.,2003).

We used a computer-assisted evoked potential system to obtain ABR thresholds for tone pips at 5, 8, 11, 16, 22, 32 and 45 kHz (tone pip duration 5 mseconds); repetition rate 30/secons and averaged responses to 512 pips of alternating polarity.

Adult ears were sectioned after paraffin wax embedding (8 μm) for morphological analysis as described (Xu et al., 1999a). We examined 10 heterozygotes in both 129/Sv and C57BL/6J backgrounds and compared them with sections from five 129/Sv and three C57BL/6J wild-type mice.

The latex paintfilling of the ears at E16.5 and 17.5 was performed as described (Morsli et al.,1998). The paintfilled inner ears were dissected out and photographed.

Embryos for histology and in situ hybridization were dissected out in PBS and fixed with 4% paraformaldehyde (PFA) at 4°C overnight. Embryonic membranes were saved in DNA isolation buffer for genotyping. Histology was performed as described (Xu et al.,1999a). To visualize Six1^lacZ expression,mutant embryos were stained with X-gal and sectioned as described(Xu et al., 2002). To reveal the middle ear ossicles, we performed skeletal staining of cartilage and bone as described (Peters et al.,1998).

For in situ hybridization, we used four wild-type or mutant embryos at each stage for each probe as described (Xu et al., 1997a).

TUNEL assay was performed as described(Xu et al., 1999a). To label the proliferating cells, timed pregnant mice at E8.5 and 9.5 were injected i.p. twice at 2-hour intervals with 5-bromodeoxyuridine (BrdU, Sigma) and embryos were collected as described (Xu et al., 1999b). Paraffin wax embedded sections of 6 μm were prepared and denatured with 4N HCl for 1 hour at 37°C. Mouse anti-BrdU monoclonal antibody and goat anti-mouse IgG coupled with HRP or Cy3 were used for detection. The number of proliferating cells was counted in serial sections from each otic placode or vesicle, and at least five embryos (10 ears) of each genotype were counted.

As the inserted lacZ transgene displayed an expression pattern identical to the Six1 mRNA distribution obtained by in situ hybridization, we analyzed the expression of Six1 gene during inner ear development in Six1^lacZ heterozygotes by staining forβ-galactosidase activity (Fig. 1). During inner ear development, Six1 expression was first detected in the ventral region of the otic pit at around E8.75(Fig. 1A). Its expression domain expands as otic development proceeds and by E9.5, Six1expression became restricted to the middle and ventral otic vesicle within which the vestibular and auditory epithelia form respectively(Fig. 1B,C). By contrast, Six1 expression is excluded from the dorsal region within which the semicircular canals form (arrows, Fig. 1A-C). Six1 is also expressed in the vestibuloacoustic ganglion (gVIII, Fig. 1D). During subsequent stages of inner ear development, Six1 is expressed in all sensory epithelia. In the vestibule, Six1 is expressed in the developing cristae, saccule and utricle(Fig. 1D-F). In the utricle and saccule, Six1 is expressed throughout the neuroepithelium at E12.5(Fig. 1D and data not shown)and by E15.5 its expression appears to be restricted to the hair cells in the middle of the epithelium (Fig. 1F). In the cochlea, Six1 is expressed throughout the future greater and lesser epithelial ridge (GER and LER) of the cochlear duct at E12.5 (Fig. 1G); however,its expression became weaker in a region within which the organ of Corti begins to differentiate at this stage (bracket, Fig. 1G). In mice, the progenitors of hair and supporting cells in the primordial organ of Corti become postmitotic between E12.5 and E14.5(Ruben, 1967). At E14.5-E15.5 when the cochlear duct has made 1.5 turns, hair and supporting cell differentiation initiates in the mid-basal region of the cochlea and hair cell differentiation proceeds until the entire length of the sensory epithelium is patterned into one inner row and three outer rows of hair cells at E17.5-E18.5(Sher, 1971; Lim and Anniko, 1985; Chen et al., 2002). In the apex of E15.5 cochlea, Six1 expression was weakly detected in the supporting cells and started to appear in the inner hair cell (arrowhead, Fig. 1H). By contrast, strong Six1 expression was observed in the GER and LER flanking the developing organ of Corti. In the basal cochlea where the development of the organ of Corti is more advanced than in the apex, Six1 is expressed in the outer and inner hair cells (arrows and arrowhead, Fig. 1I), but was undetectable in the supporting cells beneath the hair cells. Strong Six1expression was maintained in some cells in the GER. In addition, Six1is expressed in the region from which the stria normally develops. Therefore in the developing organ of Corti, Six1 is initially expressed in the progenitors of hair and supporting cells but its expression disappears when the progenitor cells exit cell cycle and later on it is expressed in the terminally differentiated hair cells. Taken together, our data show that Six1 is predominantly expressed in all sensory regions of the inner ear, suggesting a role for Six1 in the morphogenesis of sensory organs during mammalian inner ear development.

Most of the 129 Six1 heterozygous mice had certain degree of hearing loss, as determined by auditory-evoked brainstem response (ABR)threshold measurements (n=26; Fig. 2A). Of the 13 Six1^+/- mice analyzed, eight had hearing loss in both ears, two had hearing loss in one ear and three showed mild hearing loss in both ears. Similar observation was obtained in C57BL/6J background.

To determine whether the hearing loss is conductive, sensorineural or both,we sectioned the Six1^+/- ears associated with hearing loss. All Six1^+/- ears with hearing losses revealed middle ear abnormalities: typically, a failure of the ossicles to complete a sound transmission path from the tympanum to the oval window. Although the stapes attached to oval window, the VIIth cranial nerve passed abnormally between the oval window and stapedial artery or over the surface of the cochlea under the stapedial artery (Fig. 2C,D and data not shown). The middle ears also showed morphologically abnormal ossicles, including stapes with a small lumen and the middle ear space was filled up with loose connective tissues (arrow, Fig. 2F and data not shown). The latter could be due to secondary inflammation. The middle ear space appeared to be small (Fig. 2C,D,F,G) and the tympanic membrane was also abnormal (arrowheads, Fig. 2F,G). In four cases from 3 animals, there were no stapedial arteries (data not shown). In the inner ear, the cochlear spiral of all heterozygotes was well formed, although four out of 22 Six1^+/- ears (three out of 11 embryos) showed slightly shortened cochlea by latex paintfilling at E16.5 (see Fig. 8 and Table 1). Nonetheless, our analyses clearly show that there is a failure of sound transmission in the middle ear.

Auditory system abnormalities in Six1 homozygotes involve the outer, middle and inner ears (Fig. 3). The outer ear revealed malformed auricles, preauricular pits and malformed eardrums (Fig. 3A,B and data not shown). In the middle ear, the incus was present but malformed or fused with the malleus (arrowheads, Fig. 3D). The short process of the malleus was typically absent (arrow, Fig. 3D), as was the stapes,while the tympanic cavity is present in Six1 homozygotes (data not shown). In the inner ear, although the otic vesicle forms, it appears to be smaller and abnormal at E10.5 (arrow, Fig. 3F). By E12.5, no inner ear structure or sometimes only severely malformed vestibule-like structure was observed in Six1^-/-embryos (arrow, Fig. 3H). The malformed vestibule-like structure observed in some Six1^-/- embryos at E12.5 also failed to develop further(data not shown). In addition, the vestibuloacoustic (gVIII) and petrosal(gIX) ganglia were absent (asterisk and arrowhead, Fig. 3H). Thus, Six1plays a direct role in the normal development of the mammalian auditory system.

The failure of inner ear development in Six1^-/- embryos was associated with an increased cell death as detected by TUNEL analysis. Although apparent morphological difference of the otic vesicle was not observed between Six1^-/- and wild-type embryos at E9.5(data not shown), numerous apoptotic cells in the lateral wall of Six1^-/- otic vesicle were detected (arrow, Fig. 3J). At E10.5, apoptotic cells were also increased in the medial region of Six1^-/-otic vesicle (data not shown). Thus, Six1 is required for otic epithelial cell survival.

We next tested whether Six1^-/- otic epithelial cells proliferate appropriately by assaying BrdU incorporation in the mutant otic placode and vesicle at E8.5 and 9.5, before apparent morphological alteration was seen in Six1^-/- embryos. Four hours after BrdU injection, BrdU-labeled cells were seen throughout the otic placode in wild-type embryos (Fig. 4A). However, in Six1^-/- embryos, the number of BrdU-labeled cells was reduced in the otic placode (arrowhead, Fig. 4B). By E9.5,BrdU-positive cells were largely reduced in the dorsal half of Six1^-/- otic vesicle (above arrowheads, Fig. 4D). Using an image analysis system, we next counted the number of BrdU-positive cells from 10 wild-type and 10 Six1^-/- ears at each stage on serial sections to determine the labeling index(Fig. 4E). At E8.5, the number of BrdU-positive cells in Six1^-/- otic placode was 80% of wild-type embryos (Fig. 4E). By E9.5, the number of BrdU-positive cells in Six1^-/- otic vesicle was reduced to 50% of that in wild-type embryos(Fig. 4E). As the epithelial cells in the lateral wall of Six1^-/- otic vesicle undergo abnormal apoptosis from E9.5 (Fig. 3J), to further clarify whether the reduction of cell proliferation in E9.5 Six1^-/- otic vesicle is due to abnormal cell death, we determined the labeling index from the lateral and medial half of the otic vesicle, respectively. In the lateral half, the number of BrdU-labeled cells in Six1^-/- otic vesicles was 60% of that in wild-type embryos (E9.5L, Fig. 4E). Similarly, in the medial half of the otic vesicle, although no abnormal apoptosis was observed in Six1^-/- embryos at E9.5, the number of BrdU-labeled cells was reduced to 40% of that seen in wild-type embryos (E9.5M, Fig. 4E). Thus, Six1 is required for normal growth of the otic vesicle by regulating cell proliferation during early otic development.

To determine the molecular defects in early otic development of Six1^-/- animals, we first examined whether the expression of the Eya and Pax gene families depends upon Six1. Studies in Drosophila indicate that eya is epistatic to so and both genes reside within the same genetic and molecular pathway downstream of the Pax6 gene ey (Halder et al., 1998). As we previously found that Six1 mRNA expression was undetectable in Eya1^-/- embryos(Xu et al., 1999a), we first analyzed the Six1^lacZ expression by X-gal staining to further confirm this observation. Six1^lacZ expression was also undetectable in Eya1^-/- otic epithelium, further demonstrating that Six1 expression in the otic epithelium is Eya1 dependent (Fig. 5A,B). We next analyzed the expression of Eya1 in Six1^-/- embryos to further clarify their regulatory relationship during early otic development. Eya1 is normally co-expressed with Six1 in the otic epithelium and its expression was unaffected in Six1^-/- otic vesicles at E9.5 and E10.5(Fig. 5C,D and data not shown). This further confirms that Eya1 functions upstream of Six1and that the Eya-Six regulatory pathway elucidated in Drosophila eye imaginal disc is evolutionarily conserved in early mammalian otic development. Because we previously found that the expression of both Pax2 and Pax8 was unaffected in Eya1^-/- otic epithelium(Xu et al., 1999a), we next examined whether the Pax gene expression in the otic epithelium is also Six1 independent. In Six1^-/- embryos, Pax2 and Pax8 are expressed in the otic placode and vesicle at normal levels at E8.5-E10.5 (Fig. 5E-H and data not shown). This indicates that similar to Eya1,Six1 is also not required for the expression of both Pax2 and Pax8 in the otic epithelium.

We next examined the expression of several other well-characterized molecular markers in the otic epithelium at E8.5-10.5. Fgf3, a member of the Fgf superfamily of secreted signals, begins to be expressed in the ventrolateral wall of the otic vesicle at E9.5 and in the delaminating neuroblasts and gVIII (Fig. 6Aand data not shown). Inactivation of Fgf3 results in inner ear defects and its expression in the otic vesicle is Eya1 dependent(Mansour et al., 1993; Xu et al., 1999a). Similarly, Fgf3 expression was undetectable in Six1^-/- otic vesicle from E9.5 (Fig. 6B). By contrast, its expression in the hindbrain was relatively normal in Six1^-/- embryos (Fig. 6A,B). Fgf10, another member of the Fgf superfamily, is expressed in the otic placode and vesicle and by E10.5, its expression became concentrated to a broad region in the ventral half, a small patch in the posterodorsal wall and in the neuroblast cells and gVIII(Fig. 6C) (Pirvola et al.,2000). The expression of both Fgf3 and Fgf10 in the ventrolateral wall of the otic vesicle and in the neuroblasts and gVIII overlaps and both genes are suggested to function through their receptor Fgfr2 IIIb (Pauley et al.,2003). When both are knocked out, the otic vesicle fails to form(Wright and Mansour, 2003). In Six1^-/- embryos, Fgf10 expression was unaffected in the otic placode (data not shown). However by E10.5, residual Fgf10 expression was only observed in the ventromedial wall of Six1^-/- otic vesicle (arrow, Fig. 6D). The expression of both Fgf3 and Fgf10 was undetectable in the gVIII in Six1^-/- embryos, further confirming the absence of this structure in Six1 homozygotes. Bmp4, a member of the Tgfβ superfamily, is expressed in a broad region of the lateral otic vesicle at E9.5 and by E10.5, its expression is restricted to two patches, one in the dorsal and the other in the lateral region of the otic vesicle which mark the sensory anlagen of the cristae(Fig. 6E)(Wu and Oh, 1996). In Six1^-/- embryos, Bmp4 was expressed in the otic vesicle at E9.5 but its expression significantly diminished at E10.5 with the dorsal expression domain disappeared and the lateral domain weakened greatly(arrowhead and arrow, Fig. 6F). Alteration of Fgf3, Fgf10 and Bmp4 expression in Six1^-/- embryos could indicate that Six1regulates the transcription of these genes or that Six1 is required for the specification of the cells that express these genes or both. Nonetheless, these results indicate that Six1 is required for the normal expression of Fgf3, Fgf10 and Bmp4 in the otic vesicle.

During mouse inner ear development, the semicircular canals form from the dorsal region of the otic vesicle, the vestibular epithelia form from the middle region and the auditory epithelia form from the ventral region of the otic vesicle, respectively (Li et al.,1978). Because only severely affected semicircular canal-like structure was observed in some Six1 homozygotes at E12.5, we further analyzed additional markers that are well characterized and localized along the dorsoventral axis of the otic vesicle. Nkx5.1 is expressed early in the otic placode and its expression is restricted to the dorsolateral otic vesicle, which will give rise to the vestibular apparatus of the inner ear(Fig. 6G)(Hadrys et al., 1998; Wang et al., 1998). Inactivation of Nkx5.1 leads to agenesis of the semicircular canals and circling behavior (Hadrys et al.,1998). In Six1^-/- embryos, Nkx5.1expression was normal in the otic placode at E8.5 (data not shown). However in the otic vesicle, Nkx5.1 expression was excluded from the dorsolateral region at E10.5 (arrow, Fig. 6H) and its expression shifted or expanded ventrally, including the ventral-most wall (Fig. 6G,H). Gata3 is normally expressed throughout the otic placode and by E10.5 its expression is restricted to two regions in the otic vesicle, strongly in the dorsolateral region and weakly in the ventromedial region (Fig. 6I)(Lawoko-Kerali et al., 2002). In Six1^-/- embryos, no significant difference of Gata3 expression was detected by E9.5 (data not shown). However similar to Nkx5.1, Gata3 expression in the dorsolateral region was shifted ventrally in Six1^-/- otic vesicle at E10.5(Fig. 6I,J). By contrast, its expression in the medial region was unaffected in Six1^-/-otic vesicle (Fig. 6I,J). The failure to express Nkx5.1 and Gata3 in the dorsolateral region of the otic vesicle lacking Six1 is unlikely to be due only to abnormal cell death, as apoptotic cells were detected throughout the lateral region at E9.5 (Fig. 3J). Although it is unclear whether the ventrolateral region of Six1^-/- otic vesicle expressing Nkx5.1 and Gata3 will give rise to the vestibule-like structure observed in some Six1^-/- embryos at E12.5, these data indicate that Six1 regulates the establishment of regional specification of the otic vesicle.

Because the Pax2 mutant inner ear phenotype is less severe than that seen in Eya1^-/- or Six1^-/- mice(Torres et al., 1996), it is unclear whether Pax genes function in the same genetic pathway with Eya1 and Six1. To further clarify their genetic relationships during early inner ear development, we next examined the expression of Eya1 and Six1 in Pax2^-/-embryos. Surprisingly, the expression of both Eya1 and Six1was unaffected in the otic vesicle and its derivative gVIII of Pax2^-/- embryos (Fig. 7A-D and data not shown), indicating that Six1 and Eya1 expression in the otic epithelium does not require Pax2function. Taken together, our results suggest that Pax2 may function independently or in parallel with Eya1 and Six1 during early mammalian otic development.

As Eya1 and Six1 genes function in the same genetic pathway in early otic development and both gene products physically interact in vitro and in cultured cells (Buller et al., 2001), we further tested whether these two genes interact in a molecular pathway during mammalian inner ear morphogenesis by examining the inner ear gross structure of Six1^+/-/Eya1^+/- compound heterozygotes using latex paintfilling (Table 1 and Fig. 8). At E16.5, the membranous labyrinth developed to its mature shape and the cochlea reached 1.75 turns (Morsli et al.,1998). The inner ear phenotypes in each single or double heterozygous mice were variable (Table 1). Each single heterozygote alone on 129 background revealed malformed endolymphatic duct and sac and ∼18% of Six1^+/- or Eya1^+/- ears exhibited slightly shortened cochlea (arrowhead and arrow, Fig. 8B; data not shown). Some Six1^+/- ears also revealed a small or mis-shaped saccule(asterisk, Fig. 8B). By contrast, 19 out of 40 (47.5%) Eya1 and Six1 compound heterozygous ears revealed more severely affected cochlea(Table 1). The severe phenotype showed a coiled but abnormal cochlea, which only completed less than 1 turn at E16.5 with abnormally shaped distal tips (insets, Fig. 8C,D). In addition, seven out of 40 (17.5%) Eya1/Six1 double heterozygous ears showed a missing posterior ampulla and a truncation or complete absence of the posterior semicircular canal (asterisk, Fig. 8D). This defect was not seen in each single heterozygote on 129 background (Table 1). No enhancement of the defects of the endolymphatic duct and sac and the saccule was observed in the compound heterozygous inner ears(Table 1). Similar observation was obtained in C57BL/6J background (data not shown). In summary, these data suggest that Eya1 and Six1 genetically interact during inner ear morphogenesis and this interaction is crucial for the normal morphogenesis of the cochlea and the posterior ampulla.

We next analyzed early otic development in Eya1/Six1 double homozygous embryos (Fig. 8E-H). At E9.5, the otic vesicle in Eya1^+/-/Six1^+/- double heterozygotes was well formed at rhombomere 5 (r5) level and showed a restricted expression of Six1^lacZ (Fig. 8E). In the double homozygotes, the otic vesicles appeared to be formed in the correct place at r5 level but were severely hypoplastic(Fig. 8F-H). Because Six1^lacZ expression in the otic vesicle is Eya1dependent (Fig. 5A,B), the otic vesicle of Eya1/Six1 double homozygous embryos lacked the Six1^lacZ expression (arrows, Fig. 8F-H). Although the neural tube of the hindbrain was significantly reduced in size in Eya1^-/-/Six1^-/- embryos, it appeared to be patterned correctly and Six1^lacZ expression was ectopically turned on in r2, r4 and r6(Fig. 8F-H). These data further indicate that both Eya1 and Six1 are not required for the initiation of otic placode morphogenesis to form the otic vesicle, but are required for the normal growth of the otic vesicle.

Inner ear development begins with the induction of the otic placode. Once the otic placode initiates to form the otic vesicle, extensive morphogenetic processes and cellular events, such as differentiation, proliferation and apoptosis, take place to ensure the formation of the highly organized structures of the adult inner ear. Six1 expression is turned on in the invaginating otic placode at around E8.75 and in the absence of Six1, the otic vesicle formed without an apparent morphological alteration at E9.5. Thus, Six1 is unlikely to be involved directly in either the induction of the otic placode or the initiation of the otic placode morphogenesis to form the otic vesicle. This conclusion was further strengthened by the fact that the otic vesicle also formed in the correct position in Eya1/Six1 double homozygous embryos. Taken together,these findings demonstrate that both Eya1 and Six1 are not required for the initiation of mammalian inner ear organogenesis.

Our studies clearly demonstrate that Six1 controls the inner ear morphogenesis by regulating the programmed cell death and proliferative growth of the otic epithelium, directly or indirectly. Secreted diffusable factors are proposed to play roles in the growth regulation of the inner ear. Of particular interest, retinoic acid (RA) and the growth factor Bmp4 have been shown to influence regional patterning and specification of the inner ear,particularly for the vestibular structures(Chang et al., 1999; Dupe et al., 1999; Gerlach et al., 2000; Niederreither et al., 2000; Pasqualetti et al., 2001; Merlo et al., 2002). Interestingly, the inner ears of Six1^-/- embryos bear some similarities to the phenotype displayed by mice exposed to excess all-trans RA(at-RA) or isotretinoin (13cis-RA), or mice lacking Raldh2,a gene that catalyzes RA formation (Burk and Willhite, 1992; Niederreither et al., 2000). In the otic vesicle of both Six1^-/- and Raldh2^-/- embryos, Nkx5.1 expression is expanded ventrally (Niederreither et al.,2000). RA is also able to rescue the Hoxa1^-/-inner ear phenotype, including vestibular malformation and a lack of cochlear duct outgrowth (Pasqualetti et al.,2001). Recently, Bmp4 has been suggested to function together with RA through the same pathway or intersecting pathways, as RA represses Bmp4 transcription in otocyst cells(Thompson et al., 2003). However, it is unclear how exactly the RA and Bmp4 signaling controls the patterning of the inner ear. In the present study, we found that the maintenance of Bmp4 expression in the otic vesicle requires Six1 function. Therefore, it is possible that Six1 regulates normal growth of the otic vesicle through the RA-Bmp4-signaling pathway. This could explain why the morphogenetic defects in Six1^-/-mice are not restricted to the ventral inner ear but extend to the dorsal inner ear where Six1 is not expressed. Thus, the disruption of Six1 exerts some indirect, nonautomous effects on developing inner ear structures. Such long-distance influence may also be regulated by the expression of Six1 in the periotic mesenchyme. It will be interesting to test whether RA can rescue the Six1^-/- inner ear phenotype.

As the otic placode invaginates, a population of otic epithelial cells near the center of the otic cup and ventral otic vesicle emigrates into the underlying mesoderm. These cells are neuroblasts for the vestibuloacoustic ganglion (gVIII) and they are the first cell lineage specified within the otic epithelium before leaving the otic epithelium. The basic helix-loop-helix(bHLH) transcription factors neurogenin 1 (Ngn1) and Neurod1 have been shown to be essential for the formation of gVIII(Ma et al., 1998; Liu et al., 2000; Kim et al., 2001). Ngn1 was proposed to play a role for the determination of neuroblast precursor fate (Ma et al.,1998), while Neurod1 is required for the delamination of neuroblasts and for their survival during differentiation process(Liu et al., 2000; Kim et al., 2001). Recently,the transcription coactivator Eya1 has been shown to be required for the formation of the gVIII, as this structure failed to form in Eya1^-/- mice (Xu et al., 1999a). The zinc-finger protein Gata3 is expressed in the neuroblasts and also plays a role in the formation of gVIII(Karis et al., 2001). Growth factors Fgf3 and Fgf10 and their receptor Fgfr2 IIIb also play a role in the formation of the gVIII, as mutations in each of these genes led to severe hypomorphic development of the gVIII(Mansour et al., 1993; Pirvola et al., 2000). However, the molecular mechanisms controlling the specification of neuroblast cell fate are currently not well understood. In the present work, we found that Six1 is expressed in the ventral otic epithelium within which the neuroblast precursors are specified and in the gVIII. In the absence of Six1, the gVIII failed to form, similar to that observed in Eya1^-/- embryos. Thus, Six1 is likely to play a direct role in the determination of neuroblast cell fate and this cell lineage may not be specified in the absence of Six1. This hypothesis was further supported by loss of specific marker expression, including Fgf3 and Fgf10 in the neuroblasts and gVIII of Six1^-/- embryos (Fig. 6 and data not shown).

As both Eya1 and Six1 are required for the activation of Fgf3 expression in the otic epithelium and both proteins physically interact in vitro and in cultured cells(Buller et al., 2001), it is possible that Fgf3 is a common downstream target for both Eya1 and Six1. Based on these observations, we propose that both Eya1 and Six1 control the initial selection of neuroblast precursors by regulating the expression of Ngn1, Neurod1and Fgf3, directly or indirectly. This hypothesis will be strengthened if Ngn1 and Neurod1 expression in the otic epithelium and neuroblasts also requires both Eya1 and Six1function. Expression studies of Ngn1 and Neurod1 in both Eya1 and Six1 mutants are under way in our laboratory.

The commitment of the otocyst to form prospective vestibular and auditory sensory areas is controlled by patterning genes. However, it is unclear exactly how these genes are involved in this complicated process. Our studies show that Six1 is expressed in the middle and ventral otic vesicle and all middle and ventral derivatives failed to form in the absence of Six1. To analyze the molecular defects in Six1^-/-otic vesicle, we examined the expression of ventromedial otic markers,including Pax2, Pax8, the zinc-finger gene Sall1 and dachshund 1 (Dach1), in Six1^-/- embryos and failed to detect obvious changes in their expression between wild-type and Six1^-/- embryos. In addition, Gata3 expression in the ventromedial region was also unaffected in Six1^-/-embryos. Further studies are required to establish the molecular mechanisms by which Six1 acts to regulate the patterning of this region in the inner ear.

Interestingly, in addition to the loss of ventral cell fates, the absence of Six1 impacts the positioning of dorsolateral markers, including Nkx5.1 and Gata3. Although Nkx5.1 appears to be necessary for the correct expression of Bmp4 in the otocyst and for regional control of apoptosis and Bmp4 was suggested to have a role in the specification of sensory organ formation(Oh et al., 1996; Morsli et al., 1998; Cole et al., 2000; Merlo et al., 2002), it is unclear how these genes function together to control the morphogenesis of the sensory system. We found that Six1 is expressed in all sensory regions of the inner ear and Six1^-/- mice lacked all sensory organ formation. The expression domain of Bmp4, which marks the sensory anlage of the posterior crista was lost and its expression level in the other sensory anlagen was also largely reduced in Six1^-/- embryos. Coincidentally, Eya1 and Six1 interaction critically affects the morphogenesis of the posterior ampulla and Bmp4 expression was also lost in Eya1^-/- embryos at E10.5 (data not shown). Therefore, it is likely that both Eya1 and Six1 regulate the expression of Bmp4 the dosage of which is crucial for the morphogenesis of the sensory organs, particularly for the posterior ampulla. Nonetheless, our results indicate that Six1 is probably an early regulator for the specification of all sensory organs of the inner ear.

In the ear, both Eya1 and Six1 are co-expressed during mammalian auditory system development and their mutant mice had similar defects in all three parts of the ear (Xu et al., 1999a). Our studies have clearly demonstrated that the Drosophila Eya-Six regulatory cassette is evolutionarily conserved during mammalian inner ear development.

In Drosophila eye imaginal disc, the fly Pax6 gene ey has been shown to function upstream of both eya and so (Halder et al.,1998). In mammalian inner ear, Pax2 expression overlaps with Eya1 and Six1 in the medial otic vesicle and the inner ear phenotype in Pax2^-/- mice is less severe than that seen in Eya1^-/- or Six1^-/- mice(Torres et al., 1996). Pax8, a paralog of Pax2, is also expressed in the otic placode (Pfeffer et al.,1998). Although Pax8 mutants do not exhibit an otic phenotype (Mansouri et al.,1998), Pax2 and Pax8 may function redundantly during early otic morphogenesis with Pax2 alone executing later functions. This could explain why the Pax2 mutant phenotype appears to occur slightly later. If Pax2 and Pax8 function redundantly in early otic development and a crucial threshold of Pax2/Pax8 protein expression in otic epithelium regulates Eya1 and Six1 expression, Eya1and Six1 expression should be reduced or lost in Pax2/Pax8double homozygotes. We are currently testing this hypothesis by generating Pax2/Pax8 compound mutants in C3H/He background, as Pax2/Pax8 compound heterozygous females in either 129 or C57BL/6J strain had a blind-ending vagina, similar to the recent observation by Bouchard et al. (Bouchard et al.,2002). Alternatively, Pax2 and Pax8 could function in parallel or independently of Eya1 and Six1 in early otic morphogenesis, as Eya1 and Six1 expression was unaffected in Pax2^-/- otic vesicle. Evidence obtained from the analysis of the cochlea phenotype in Eya1/Pax2 compound heterozygous mice suggests that Eya1 and Pax2 may interact during cochlear development, because the cochlea phenotype is enhanced in Eya1/Pax2 compound heterozygotes than in each single heterozygote(data not shown). It should also be noted that our recent studies indicate that the genetic and regulatory relationship between Pax, Eya and Six genes varies between different organs during mammalian development(Xu et al., 2003). Probably,the Pax, Eya and Six genes function in the same or parallel pathway but with different combinations of regulatory relations in different organs. Detailed examination of inner ears in Pax2/Pax8, Pax2/Eya1, Pax2/Six1 or Eya1/Six1/Pax2 compound knockouts will enhance our understanding on the possible molecular and genetic interactions between these transcription factors during early mammalian inner ear morphogenesis.

We thank P. Gruss for the Pax2 and Pax8 mutant mice, B. Fritzsch and G. Carlson for helpful comments, and G. Sajithlal and W. Provance for technical assistance. Photomicroscopy and image analysis was made possible by equipment purchased with a grant from the M.J. Murdock Charitable Trust. This work was supported by NIH RO1 DC05824, NCRR P20RR15583, Oberkotter foundation and Evan Drammis Research Foundation (all to P.-X.X.).