Lack of pendrin expression leads to deafness and expansion of the endolymphatic compartment in inner ears of Foxi1 null mutant mice

Hearing loss is a common form of sensory impairment, and several millions of individuals are affected world wide(Wilson, 1985). Around one in a thousand children are born with a significant permanent hearing impairment,and this proportion increases with age until 60% of the population over 70 show an impairment of 25dB or greater(Fortnum et al., 2001;Davis, 1995). In most instances, the inner ear is the primary site of the dysfunction. The inner ear is also one of the most complex organs that has evolved in vertebrates. Originating from the otic placode, a thickening of the surface ectoderm of the head, the inner ear develops gradually through invagination and formation of an otic vesicle by day E9.5 in mice. The sensory hair cells responsible for hearing are located within the organ of Corti in the cochlea. Their function is to convert the energy transmitted by sound waves into neural impulses(Corey and Hudspeth, 1983;Howard and Hudspeth, 1988) and to act as a selective amplifier (Zheng et al., 2000). The vestibular sensory organs, the cristae of the semicircular ducts and maculae of the utricle and saccule, detect angular and linear acceleration, respectively. One crucial aspect of inner ear architecture is the presence of two isolated chambers containing endolymph(high K⁺, low Na⁺) and perilymph (low K⁺,high Na⁺). During sound stimulation, deflection of a specialised array of finger-like projections (stereocilia) from the top of each hair cell opens transduction channels. This allows potassium ions to leave the potassium rich endolymph with its positive endocochlear potential, and enter the hair cell with its negative intracellular resting potential. This flood of potassium depolarises the hair cell and triggers synaptic activity leading to action potentials in the cochlear neurons. The endocochlear potential is maintained by active pumping of potassium by stria vascularis cells fed by rapid recycling of potassium ions within the cochlear duct(Steel, 1999). The endolymphatic duct and sac are structures thought to serve as sites for resorption of endolymph as well as an expansion chamber that can compensate for sudden changes in endolymph volume(Barbara et al., 1988a;Everett et al., 1999). Several ion channel and pump proteins, that assist in maintaining proper ionic composition of the endolymph, are expressed at various specific locations throughout the inner ear epithelium, e.g. the dark cells adjacent to the maculae, stria vascularis of the cochlea and the endolymphatic duct/sac epithelium. Appropriate ionic composition of the endolymph is a prerequisite for normal inner ear function. Mutations in genes encoding such ion channels and pumps have been shown to cause deafness e.g. KCNQ4(Kubisch et al., 1999),KCNQ1 (Neyroud et al.,1997), KCNE1 (Vetter et al., 1996), Slc12a2(Delpire et al., 1999;Dixon et al., 1999),ATP6B1 (Karet et al.,1999) and PDS (Li et al., 1998). Deafness genes may also influence ionic balance of the endolymph by affecting the integrity of the surrounding endolymphatic epithelium or other cell types involved in recycling and/or production of endolymph, e.g. connexin 26, 30 and 31(Steel and Bussoli, 1999) andCOCH [encodes a secreted protein, thought to be involved in some cases of Ménière's disease(Fransen et al., 1999)].Slc12a2 encodes a Na-K-Cl cotransporter expressed in the marginal cells of stria vascularis, the major site of endolymph production, and mutations in this gene result in failure to produce endolymph. This has been demonstrated by gene targeting experiments as well as by identification of mutations in this gene in the mouse mutant shaker-with-syndactylism[sy (Delpire et al.,1999; Dixon et al.,1999)]. In both instances, there are clear signs of decreased endolymph production, such as extensive collapse of Reissner's membrane and vestibular walls together with thinning of the semicircular canals. TheKcne1 mouse mutant shows a similar collapse of endolymphatic compartments (Vetter et al.,1996). Even though several inner ear conditions such as Ménière's disease and Pendred syndrome have been linked to increased volume and/or pressure of the endolymph(Everett et al., 1999), no good model for the pathogenic mechanisms underlying these conditions have been presented.

Here we describe the Foxi1^-/- (Fkh10) mutant in which the endolymphatic compartment is severely dilated. We show that these mice are deaf and lack an endocochlear potential, indicating a primary defect in fluid homeostasis in the inner ear. This is associated with an expansion of the endolymph compartment followed by rupture of the endolymphatic epithelium. Based on morphological, physiological and genetic analysis we propose a molecular mechanism by which the endolymphatic inner ear fluid is allowed to accumulate to abnormal levels.

Foxi1 mutant mice (Hulander et al., 1998) and wild-type littermates of CD-1 and B16 origin were obtained from heterozygous crosses. Genotypes were determined from tail biopsies using PCR. The forward primer 5′-CCT TCG ACC TCC CAG CGC CTT-3′ and the two reverse primers 5′-GGG CAG TAG CTG CCT CTG CAT-3′ and 5′-GGG CCA GCT CAT TCC TCC ACT-3′ yielded 290 bp wild-type and 320 bp mutated allele products in a 30-cycle reaction (95°C for 30 seconds 65°C for 30 seconds 72°C for 15 seconds).

The morning that vaginal plugs were detected was designated embryonic day 0.5 (E0.5). Tails from embryos were collected for genotyping using PCR as described above. Embryos were fixed overnight in PBS with 4% paraformaldehyde at 4°C. For histological examination, embryos were dehydrated and embedded in paraffin wax and 6 μm sections were cut. Tissue sections were stained with Haematoxylin and Eosin. In situ hybridization on whole-mount mouse embryos (Rosen and Beddington,1993) and in situ hybridization of cryosections(Bostrom et al., 1996) were performed with digoxigenin-labeled antisense cRNA probes. No signals were observed when control sense probes were used. We used plasmids to generate probes for: Pds (Everett et al.,1999), Pax2 (Torres et al., 1996), ErbB3(Britsch et al., 1998),Eya1 (Xu et al.,1999), Hmx3 (Wang et al., 1998), Jag1(Morrison et al., 1999),Otx1 (Acampora et al.,1996) and Coch(Robertson et al., 1997). For histology and in situ hybridization experiments inner ears from at least seven mice of each genotype (Foxi1^+/+,Foxi1^+/- and Foxi1^-/-) of CD-1 background and three of each genotype of B16 background were analyzed.

For BrdU labeling, pregnant female mice were injected intraperitoneally with BrdU (1 ml of 10 mM BrdU per 100 g body weight; Roche). Injected mice were sacrificed 2 hours later and processed for immunohistochemistry,according to protocols supplied by the manufacture. Kits (Roche) were used to identify apoptotic cells by the TUNEL assay according to the manufacturer's instructions. For immunohistochemistry, BrdU labeling experiments and TUNEL assay inner ears from at least four mice of each genotype(Foxi1^+/+, Foxi1^+/- andFoxi1^-/-) of CD-1 background were analyzed.

Paint-filling of inner ears (E11.5: Foxi1^+/+;n=5, Foxi1^+/-; n=3,Foxi1^-/-; n=3, E12.5:Foxi1^+/+; n=3, Foxi1^+/-;n=8, Foxi1^-/-; n=3, E16.5:Foxi1^+/+; n=6, Foxi1^+/-;n=11, Foxi1^-/-; n=5) was performed in a similar way to that described for chicken and mouse embryos(Martin and Swanson, 1993). Briefly, mice (E11.5-E16.5) were decapitated and fixed in Bodian's fixative(75% ethanol, 5% formalin, 5% glacial acetic acid) overnight and dehydrated twice in each of the following solutions: 75%, 95%, 100% ethanol (minimum 2 hours for each wash). Heads were bisected and cleared overnight in methyl salicylate. The endolymphatic compartment of the inner ear was injected via either the common crus and/or cochlea using a pulled glass capillary pipette(20-40 μm diameter) filled with 1% gloss paint in methyl salicylate. The ears were then dissected free of the skull and photographed in methyl salicylate. Temporal bone biopsies were cleared in methyl salicylate for analysis of otoconial crystals.

Three-dimensional reconstructions from inner ears were made using approximately 300 serial sections (6 μm) from an entire inner ear. The inner lumen of the ear was traced and sensory areas, identified by the typical pseudostratified appearance of the epithelium, were outlined in different colors. Areas of interest from each section were traced, aligned and reconstructed into three-dimensional images essentially as has been described earlier (Wu and Oh, 1996). For data processing we used the computer software Spyglass Slicer/T3D (Research System,http://www.researchsystems.com/noesys). Four skulls of each genotype (Foxi1^+/+,Foxi1^+/- and Foxi1^-/-) of CD-1 background were sectioned and all four in each group had identical phenotypes. Two of each genotype were used for three-dimensional reconstruction.

The embryos were fixed in 95% ethanol and stained with Alizarin Red to reveal bone and Alcian Blue to reveal cartilage(McLeod, 1980). They were subsequently cleared by trypsin digestion and KOH treatment, and stored in glycerol.

Heads of E14.5 embryos were fixed immediately after decapitation in a mixture of 2% paraformaldehyde, 2.5% glutaraldehyde, 0.1% sodium azide in 0.05 M sodium cacodylate buffer, pH 7.2. The heads were embedded in 7% agar and serial sections were obtained with an oscillating tissue slicer (Leica VT 1000S; section thickness set at 200 μm, horizontal orientation). The slices were examined in a tissue preparation microscope to identify levels with opened endolymphatic structures. Selected sections were further treated with a repeated sequence of 1% osmium tetroxide and a saturated thiocarbohydrazide solution (OTOTO method), followed by dehydration in ethanol and infiltration in hexamethyldisilazane (HMDS). The specimens were dried by evaporation of the HMDS in a fume hood and were flat mounted on aluminum stubs. They were examined in a Zeiss 982 Gemini field emission scanning electron microscope without prior thin film metal coating.

Mice (n=5 controls (2 Foxi1^+/+, 3Foxi1^+/-); 5 Foxi1^-/-, aged 29-30 days; and n=6 Foxi1^+/+; 6Foxi1^+/-; 5 Foxi1^-/- aged 67-68 days)were anaesthetized with urethane, the middle ear was opened, and a recording electrode was placed on the round window of the cochlea. A closed sound system in the external ear canal was used to deliver calibrated tonebursts (15 mseconds duration, 1 mseconds rise/fall time, 100 mseconds interstimulus interval, average 200 repetitions). Thresholds for detection of a cochlear nerve compound action potential (CAP) response were obtained using 2-3 dB steps. The endocochlear potential (EP) was measured in the same animals as well as in additional ones (Foxi1^+/+, n=12;Foxi1^+/-, n=15; Foxi1^-/-,n=14) using a micropipette electrode inserted into the basal turn scala media through the lateral cochlear wall(Steel and Smith, 1992).

In earlier work, we described the otic vesicle-restricted expression pattern of Foxi1 at E9.5 and the cystic dilatation of the inner ears at E18.5 in mice that lack Foxi1(Hulander et al., 1998). As a likely consequence of this dilatation, such mice display hearing and balance problems (Hulander et al.,1998). Here we present a more detailed analysis of the morphological, physiological and genetic consequences for inner ear development that proceeds without the participation of the winged helix geneFoxi1.

From earlier experiments, it was known that otic vesicle formation at E9.5 is unaffected in Foxi1 null mutants(Hulander et al., 1998). To investigate the morphology in wild type and Foxi1-deficient mice during later stages of inner ear development, we compared the histology using sections from E11.5 (not shown), E12.5 (not shown) and E13.5(Fig. 1). At E11.5 formation of the endolymphatic appendage, the cochlear anlagen and the vertical plate of the vestibular system (future anterior and posterior semicircular ducts) is seemingly unaffected in mice that lack Foxi1 (not shown). From E12.5-13.5, we find a normal cochlea as well as normal semicircular ducts in both wild-type and Foxi1^-/- mice (not shown). At E11.5,the combined facial and vestibulo-cochlear ganglion (cranial nerve ganglion VII and VIII) is present and at E12.5 to E13.5 the vestibulo-cochlear ganglion(VIII) has formed in both wild type and null mutants. To assess the general morphology of the membranous labyrinth we used a paint-fill method in which a thin needle was inserted into the most cranial part of the common crus, formed by the anterior and posterior semicircular ducts(Martin and Swanson, 1993). Subsequently white gloss paint was injected to visualize the endolymphatic compartment. The future endolymphatic and semicircular ducts as well as saccule and cochlea showed no signs of abnormality at E11.5 and E12.5(Fig. 2). As illustrated (Figs1,2) no apparent difference in morphogenesis can be seen at early stages of inner ear development.

At E11.5 Foxi1 expression is restricted to the otocyst with most of the staining in structures that will give rise to the endolymphatic duct and sac (Fig. 3). Based on earlier findings (Hulander et al.,1998) and data presented inFig. 3 it is clear thatFoxi1 expression is restricted to the otic vesicle from E9.5 to E11.5. To study the possibility that Foxi1^-/- embryos could be affected in their capacity to express other genes of importance for inner ear development, without causing any detectable morphological aberration(Figs 1,2) at early developmental stages (E11.5-E13.5) we performed whole-mount hybridization on E11.5 wild-type and Foxi1^-/- embryos using cRNA probes encoding five inner ear marker genes. (i) ErbB3, a marker for neural crest-derived cell populations, i.e. presumptive stria vascularis cells(Britsch et al., 1998;Lemke, 1996). (ii)Otx1, necessary for proper induction of the lateral canal as well as utriculosaccular and cochleosaccular ducts(Acampora et al., 1996;Morsli et al., 1999). (iii)Eya1, crucial for inductive signaling events during early otic vesicle maturation (Sahly et al.,1999; Xu et al.,1999). (iv) Pax2, expressed in regions of the otocyst that will give rise to cochlea and saccule(Favor et al., 1996;Hutson et al., 1999;Torres et al., 1996) as well as in prospective endolymphatic epithelium(Lawoko-Kerali et al., 2002). Mice lacking Pax2 are void of cochlea(Torres et al., 1996;Favor et al., 1996). (v)Hmx3 (also known as Nkx5-1) is required for proper formation of the sensory organs of the vestibulum(Hadrys et al., 1998;Wang et al., 1998). These five marker genes showed identical staining pattern in Foxi1^-/-and wild-type litter mates (not shown).

Histological sections comparing inner ear morphology in wild-type andFoxi1 mutant embryos reveal an expansion of several inner ear structures at E16.5 (Fig. 4A-D). The entire inner ear seems to be expanded including the posterior semicircular duct, common crus (not shown), ampullae (not shown) and the cochlea. However, the most prominent enlargement is seen in the endolymphatic duct/sac (Fig. 4A,B). The apical turn of the cochlea has been replaced by an apical cyst, the basal turn is also enlarged(Fig. 4C,D). In paint-fill experiments using inner ears of embryos from E16.5 the picture is quite different from that at E11.5 and E12.5. There is a prominent expansion of the entire membranous labyrinth. The endolymphatic duct and sac together with the saccule, utricle, ampullae and cochlea are all severely enlarged whereas the semicircular ducts seem to be less affected(Fig. 5). A three-dimensional image was reconstructed using approximately 300 serial sections representing an entire inner ear at E16.5 (Fig. 6). This experiment allowed us to visualize the membranous labyrinth and at the same time mark the sensory areas of the inner ear. This analysis corroborates the results from histological examination(Fig. 4) and paint-fill experiments (Fig. 5). It also demonstrates that the sensory patches in both the vestibulum and the cochlea are located normally. From these data it is clear that while inner ear formation proceeds without any major aberration up to E13.5, an expansion of the entire compartment is evident at E16.5. Measurements of the diameter of the endolymphatic duct at corresponding anatomical sites reveal an increase in diameter of mutant ducts, expressed as arbitrary units [42.3±4.7 (s.d.,n=3)], compared with wild-type ducts [11.0±1.0 (s.d.,n=3), P<0.01]. The ratio of mutant over wild-type diameter is 3.8, this ratio predicts an increase in volume by a factor of approximately 55 (3.8³=54.9), assuming an equal and proportional increase in all three dimensions.

To investigate whether the expansion noted at E16.5 was transient or progressive in nature, a histological examination of decalcified skulls from postnatal day 12 (P12) mice was performed. From this analysis it is clear that the expansion is progressive and that the entire inner ear is replaced by large irregular cavities (Fig. 7A-D). The saccule and utricle have expanded and formed a large cavity that is continuous with the endolymphatic duct and sac(Fig. 7A,D). The basal turn of the cochlea has expanded and ruptured into a common cavity, and this is also true for the apical turn that forms an apical cyst in the mutant cochlea(Fig. 7B,D). Since the modiolus is missing in the cochlea of Foxi1 mutant mice(Fig. 7D) the phenotype fulfills the criterion for a Mondini malformation of the cochlea as described by Carlo Mondini in 1791 (Mondini,1791). The extreme expansion of the inner ear also seems to compress the cerebellum raising the intriguing possibility that the balance problems seen in Foxi1^-/- mice(Hulander et al., 1998) could be derived to some extent from a direct compression of the cerebellum. The temporal bone, in which the entire inner ear resides, is thinner adjacent to the inner ear (Fig. 7C). A thinning of the endolymphatic epithelium in both the endolymphatic duct(Fig. 4B) and the cochlea(Fig. 4D) is evident already at E16.5. The cystic appearance of the cochlea(Fig. 7D; no sign of perilymphatic compartment) as well as the disrupted compartmentalization of the vestibular portion of the inner ear at P12(Fig. 7C) strongly suggest that the peri- and endolymphatic compartments of the inner ear in Foxi1mutants, at P12, have been replaced by a common irregular cavity.

The otoconia of the utricle and saccule were investigated in methyl salicylate-cleared temporal bone specimens from mutants and controls(Fig. 8). Here we found a complete absence of normal-looking otoconia in the mutant mice whereasFoxi1^+/- mice displayed the normal white otoconial appearance. This might reflect an inability to form proper crystals in the mutant mice. The expansion and altered ability to form otoconia could reflect an altered ionic composition of the endolymph.

Another possible mechanism explaining the expanded inner ear phenotype would be an altered rate of apoptosis and/or cell proliferation. In a TUNEL labeling experiment, performed on sections from E16.5, only one minor difference in labeled cell populations were detected(Fig. 9). A small mesenchymal cell population adjacent to the lateral border of the endolymphatic duct showed increased TUNEL labeling (Fig. 9). A BrdU labeling experiment was also performed at E16.5 and no difference in the number of labeled cells could be detected (not shown). These results suggest that only minor differences exist in the rate of apoptosis and cell proliferation at E16.5, a time point at which the inner ear has initiated its expansion. Yet another explanation for the inner ear phenotype observed inFoxi1 mutants could be a defect in epithelial-mesenchymal interactions during condensation of the periotic mesenchyme. To address this,bone and cartilage staining was performed on E18.5 embryos. Normal integration of the inner ear into the temporal bone anlagen was observed in wild-type andFoxi1^+/- embryos whereas null mutants displayed an abnormal integration associated with ectopic cartilage and/or bone formation(Fig. 10). In conclusion,these experiments demonstrate an extremely expanded inner ear with lack of a proper crystallization of the otoconia, which points toward an altered ionic composition of the endolymph as the underlying cause of this phenotype. According to this hypothesis, the increased amount of apoptotic cells adjacent to the endolymphatic duct (Fig. 9) as well as the abnormal integration of the inner ear in the temporal bone (Fig. 10)probably should be regarded as phenomena secondary to the expansion.

Previously, we have shown that Foxi1 mutant mice do not respond in a Preyer reflex test, which is a crude single frequency hearing test,indicating profound hearing impairment(Hulander et al., 1998). In order to test the hearing of these mice with better accuracy we performed round window recordings. This enabled us to record compound action potentials(CAP) as the result of sound stimulation of various intensities and frequencies. As shown in Fig. 11A, no thresholds could be established for any of the frequencies tested in Foxi1^-/- mice, a finding compatible with complete hearing loss. The thresholds measured forFoxi1^+/+ and Foxi1^+/- mice were not significantly different from each other but did show relatively high thresholds compared with other strains of mouse, especially at high frequencies (Fig. 11A)(Shone et al., 1991). Endocochlear potentials (EP) were also measured and as can be seen inFig. 11B they were normal(approximately +115 mV) in wild-type and heterozygotes whereas null mutants have an EP of about 0 mV.

With these results at hand, we sought a molecular explanation for the morphological and physiological findings observed inFoxi1^-/- mice. Pendred syndrome (PDS) is a human congenital sensorineural deafness syndrome caused by mutations in thePDS gene, characterized by, among other things, a widened vestibular aqueduct, Mondini malformation of the cochlea as well as an enlarged endolymphatic duct and sac (Everett et al., 1997; Johnsen et al.,1986; Phelps et al.,1998). Since the phenotype described for Pendred syndrome and that observed in Foxi1^-/- mice show clear similarities, we used a mouse Pds cRNA probe for hybridization experiments on inner ear tissue sections from wild type and null mutants(Fig. 12A-C). The wild-type expression patterns of the Pds and Foxi1 genes are very similar, being scattered within the endolymphatic duct/sac epithelium(Fig. 12A). At this developmental stage (E16.5), Foxi1 expression is restricted to the endolymphatic duct/sac epithelium while the Pds gene is expressed in the external sulcus of the cochlea as well as adjacent to the maculae of the saccule and utricle (Fig. 12C). In Foxi1^-/- mutants there is a complete lack of Pds gene expression in the endolympahtic duct/sac epithelium(Fig. 12B). It is interesting to note that only in locations where expression of Foxi1 andPds overlap can we demonstrate a lack of Pds gene expression, i.e. the endolymphatic duct/sac epithelium(Fig. 12B). In spite of altered morphology of the cochlea and utricle in Foxi1^-/-mice, Pds gene expression can be demonstrated at these locations(Fig. 12C). The COCHgene is thought to encode a secreted protein with similarities to extracellular matrix proteins (Robertson et al., 1997). It is expressed at high levels in the cochlear and vestibular organs, in the stroma beneath the epithelium as well as in spindle shaped cells along the pathway of auditory nerve fibers(Robertson et al., 1998).Coch is expressed in a scattered pattern in the endolymphatic duct/sac epithelium of wild-type embryos whereas this expression is absent inFoxi1^-/- mice (Fig. 12D). To our knowledge this is the first time that Cochexpression has been identified in endolymphatic duct/sac epithelium. In other locations such as the cochlea and saccule Coch expression is unaffected in Foxi1 null mutants(Fig. 12E,F). The same is true for expression of the Notch ligand Jag1 [also known asSerrate1 (Morrison et al.,1999)], in that Jag1 is expressed in a punctuate pattern in the endolymphatic duct/sac epithelium and the expression at this location is absent in Foxi1^-/- animals(Fig. 12G). Jag1expression, at other locations of the inner ear, such as the prospective sensory patches at E13-17 and later in supporting cells, is unaffected in theFoxi1 null mutant (Fig. 12G,H). Furthermore, scanning electron microscopy of the surface of the endolymphatic sac/duct epithelium shows that the scattered cells with long microvilli projecting into the lumen of the normal sac/duct are not detected in the Foxi1^-/- mutants(Fig. 13). This is a cell configuration that fits well with the scattered pattern of in situ hybridization signals for Foxi1, Pds, Coch and Jag1(Fig. 12). To confirm viability and biological functionality of the endolymphatic duct/sac epithelium we showed that antisera directed against E-cadherin and ZO-1(zonula occludens), markers for biological active epithelium(Stevenson et al., 1986;Torres et al., 1996), stain wild-type and Foxi1^-/- endolymphatic duct/sac epithelium equally well (not shown).

Early (E9.5-13.5) morphogenesis and patterning of the inner ear appear to proceed normally in mice lacking Foxi1 (Figs1,2). Expression patterns of marker genes, known to be involved in patterning of the otic vesicle [e.g.Eya1 (Xu et al.,1999)], vestibular development [e.g. Otx1, Hmx3(Acampora et al., 1996;Hadrys et al., 1998;Wang et al., 1998)], cochlear maturation [e.g. Pax2 (Favor et al., 1996; Torres et al.,1996)] and a marker for neural crest cell contribution to the developing inner ear [ErbB3(Britsch et al., 1998;Lemke, 1996)], show no aberration in Foxi1^-/- mutants. Although it is possible that some aspects of early inner ear morphogenesis, not identified by the methods used here, are affected — the otic vesicle-restricted expression of Foxi1 at E9.5 (Hulander et al., 1998) seems to be redundant for early development of the inner ear. At later stages (E16.5) of inner ear maturation Foxi1expression becomes restricted to the endolymphatic duct/sac epithelium(Fig. 12A), thus delimiting this cell population as a unique pool of Foxi1-positive cells in contrast to the remaining otic epithelium. This epithelium is also characterized by a scattered expression of Pds(Fig. 12A), Coch(Fig. 12D) and Jag1(Jagged1; Fig. 12G). The sites of wild-type Coch expression, in the cell layers beneath the otic epithelium in the cochlea and vestibulum reinforces the unique properties of the endolymphatic duct/sac epithelium, where Coch is expressed instead in the epithelial cells (Fig. 12D-F). While the fate and function of the different cell populations within this epithelium remain unclear, the punctate expression pattern of Foxi1, Jag1 (Morrison et al., 1999) Coch and Pds might delimit a subset of endolymphatic duct/sac epithelium cells that could serve a common function in the mature epithelium such as resorption of endolymph. Detailed morphological analysis of the endolymphatic duct/sac epithelium, using electron microscopy, suggests the presence of at least three distinct cell types. One of these cell types project microvilli into the endolymphatic duct/sac lumen (Barbara et al.,1988b). Even though our expression data do not allow any conclusions to be made with regard to which cell type(s) express Foxi1,Jag1, Coch and Pds (Fig. 12), our experiments do provide evidence for the presence of several distinct cell types within this epithelium. The lack of cells with long microvilli projecting into the lumen of the endolymphatic sac in theFoxi1 mutants suggests that this cell type may also expressFoxi1, Pds, Coch and Jag1 (Figs12,13). Thus, it is possible that early Foxi1 expression, at E9.5-13.5, is important for correct patterning of endolymphatic duct/sac epithelium. We would like to speculate that this process could involve Jag1/Notch interactions and thatFoxi1 expression is needed for proper execution of such signals.

In mice that lack Foxi1 expression, an expansion of the endolymphatic chamber becomes evident at E16.5 with the endolymphatic duct/sac as the most expanded structure (Figs4,5,6). The absence of Pds gene expression in the epithelium of this structure (Fig. 12B)establishes Foxi1 as an upstream regulator of Pds gene expression at this location. There are two possible reasons for the lack ofPds expression in the endolymphatic duct/sac: Foxi1 may in a direct way regulate Pds expression, or without Foxi1, cells that normally express Pds may be missing. Pds gene expression in the cochlea and vestibulum is unaffected by the lack ofFoxi1, suggesting that expression at these locations is regulated by other factors. Both Coch and Jag1 expression are also dependent on Foxi1 in the endolymphatic duct/sac while expression at other sites in the developing inner ear is not(Fig. 12D-H). In the inner ear of wild-type mice, Pds gene expression is first visualized in the prospective endolymphatic duct/sac epithelium at approximately E13. Two days later at E15 Pds gene expression is also found in the non-sensory regions adjacent to the maculae of the saccule and utricle as well as in the cells beneath the spiral prominence on the lateral wall of the external sulcus in the cochlea (Everett et al.,1999). It is interesting to note that the expansion of the endolymphatic compartment, in Foxi1^-/- embryos, starts between E13.5 and E16.5 (Figs4,5,6) — the same time span during which Pds gene expression first becomes detectable in wild-type mice (Everett et al.,1999). Taken together these findings support the idea that the inner ear phenotype in Foxi1^-/- mice is, at least to some extent, due to lack of pendrin expression in the endolymphatic duct/sac epithelium. This hypothesis is further supported by the fact that patients with Pendred syndrome display enlargement of the endolymphatic duct and sac as a constant feature (Phelps et al.,1998). As a consequence of this, it seems that Pds gene expression at inner ear locations other than the endolymphatic duct/sac epithelium is not sufficient to substitute for the lack of Pdsexpression in the endolymphatic duct/sac epithelium and hinder the abnormal expansion of the endolymphatic compartment. Alternatively, the expression at these locations, initiated at E15, is not capable of reversing the presumably established state of dysfunction in the endolymphatic duct/sac epithelium,involving not only Pds gene expression but also Coch andJag1. The exact order of events underlying the phenotype described here is not resolved. It is however possible that Notch-mediated patterning of the endolymphatic epithelium, supported by early (E9.5) Foxi1expression, is disturbed because of a lack of Jag1 and that this leads to lack of a specialized cell type that expresses pendrin and serves as a major site for endolymph resorption(Fig. 13). Alternatively, it is also conceivable that lack of pendrin expression leads to a state of epithelial dysfunction not compatible with Jag1 and Cochexpression, at a later stage of inner ear development. Such dysfunction could not be too pronounced since expression of both E-cadherin and ZO-1, markers of biologically active epithelium, are unaffected by the absence ofFoxi1 expression (not shown). Nevertheless a gene regulatory hierarchy has been established in which Foxi1 is upstream of Pds,Coch and Jag1. Further support for our suggestion that abnormalPds expression may underlie the pathology we observe inFoxi1 mutants comes from the Pds mutant mouse, which shows a similar early expansion of the endolymphatic compartments including endolymphatic duct and sac and resultant malformation(Everett et al., 2001).

Only minor differences in TUNEL (Fig. 9) and BrdU (not shown) labeling can be demonstrated. The abnormalities revealed by bone and cartilage staining(Fig. 10) are interesting since epithelial-mesenchymal interactions are necessary for condensation of the periotic-mesenchyme during formation of the otic capsule(Frenz and Van De Water, 1991;Van de Water and Ruben, 1971). Results from bone and cartilage staining demonstrate an altered gross morphology of the otic capsule in Foxi1 null embryos at E18.5. This may be due to increased endolymphatic pressure, leading to expansion of the endolymphatic compartment, resulting from defective pendrinmediated chloride ion resorption. It should be noted that a defect in epithelial-mesenchymal interaction could not be ruled out as the primary defect of the abnormal otic capsule. However, it seems unlikely since Foxi1 expression at E16.5 is restricted to the endolymphatic duct/sac epithelium, which is located outside the otic capsule. One consequence of altered function of the endolymphatic duct/sac epithelium is most likely a change in ionic composition of the endolymph as demonstrated by lack of proper crystallization of the otoconia (Fig. 8). The complete hearing loss together with an endocochlear potential (EP) of approximately 0 mV in Foxi1 null mutants (Fig. 11A,B) suggest a severe `ion channelopathy' possibly in combination with a rupture of the endolymphatic epithelium allowing peri- and endolymph, to mix. However, we cannot decisively determine which of these two factors is the major cause of deafness. We would like to speculate that the findings of (i) a progressive thinning of the endolymphatic epithelium (Figs4,7), (ii) lack of a perilymphatic compartment in the cochlea(Fig. 7D), (iii) expansion of endolymphatic compartments in the vestibulum(Fig. 7C) and (iv) an EP of 0 mV (Fig. 11B) all are compatible with a more or less total rupture of the endolymphatic epithelium. Nevertheless, it is most likely that the initial event is an expansion of the endolymphatic compartment due to a defective ionic composition, possibly including an increased osmolarity of the endolymph, which in turn severely affects the ability of stria vascularis to maintain a proper EP. A rupture of the endolymphatic epithelium at P12, as suggested by the results inFig. 7, would further hamper this function.

The molecular mechanisms governing transformation of the near spherical otocyst into a highly complex sensory organ — the inner ear —remain to a large extent obscure. One obvious task during this process is to guide morphogenetic events that produce the intricate three-dimensional structure of the mature inner ear. In many instances during embryogenesis,formation of complex structures is achieved through preprogrammed alterations in rate of cell proliferation and/or apoptosis, e.g. limb formation. However,a recent study addressing this in the developing inner ear of chicken failed to demonstrate a clear increase in cell proliferation during outgrowth of endolymphatic duct and canal plates (Lang et al., 2000). Instead differences in the number of cells in various regions of the otic vesicle was noted, suggesting that cells were redistributed. For instance, the dorsal aspect of the otic vesicle (giving rise to the endolymphatic duct and sac) was found to undergo thinning of the epithelial surface with only a slightly increased cell proliferation(Lang et al., 2000). It is intriguing that several aspects of the inner ear phenotype seen inFoxi1^-/- mice bear resemblance to that seen inHoxa1^-/- and Fgf3^-/- animals(Chisaka et al., 1992;Lufkin et al., 1991;Mansour et al., 1993). During early embryonic development the endolymphatic duct/sac fails to form inHoxa1^-/- and Fgf3^-/- mice and the inner ear develops into large irregular cavities or has a distended and swollen appearance, respectively. This finding underscores the importance of an intact endolymphatic duct/sac, for normal inner ear development. Although there are many differences between Foxi1, Hoxa1 and Fgf3mutants, they might all reflect the consequences of inappropriate endolymph turnover resulting from a lack of (Hoxa1^-/-,Fgf3^-/-) or inadequate function(Foxi1^-/-) of the endolymphatic duct/sac epithelium. The phenotypic differences could be explained by the temporal appearance of these defects, early at the otic vesicle stage in Hoxa1 null mutants and late in the case of Foxi1^-/- mice. These models also emphasize the need for tightly regulated endolymph secretion/resorption during normal inner ear development. It is possible that normal inner ear development is dependent upon `pressure driven' events. This idea is supported by the finding that the initial events in outgrowth of the endolymphatic duct are not correlated with increased cell proliferation but rather thinning of the epithelial surface (Lang et al.,2000).

The endolymphatic duct and sac are crucial for maintaining proper osmotic pressure and ionic composition of the endolymph. These structures can function as an expansion chamber and expand beyond the inner ear compartment into the interdural space. In this way variations in the production of endolymph can be tolerated without any significant increase of inner ear pressure(Barbara et al., 1988a;Everett et al., 1999). This is emphasized by the fact that one experimental approach to induce inner ear hydrops is by obliteration of the endolymphatic duct/sac(Kimura and Schuknecht, 1965). The endolymphatic duct/sac epithelium is also important for absorption of endolymph (Everett et al.,1999) — several ion transporters are expressed here such as the chloride transporter pendrin (Everett et al., 1999). The relevance of this for the pathogenesis of human inner ear diseases is exemplified by conditions such as Ménière's disease and Pendred syndrome which are both, to some extent, associated with inner ear hydrops (i.e. accumulation of endolymph in the inner ear resulting in hampered inner ear function) and an expanded endolymphatic duct/sac (Everett et al.,1999; Phelps et al.,1998; Schuknecht and Gulya,1983). We would also like to point out that mutations in theCOCH gene have been linked to Ménière's disease(Fransen et al., 1999). The finding of a specific cell type, with projecting microvilli, in the endolymphatic duct/sac epithelium that most likely expresses Foxi1, Pds,Coch and Jag1 raises the intriguing possibility that these FORE(forkhead related) cells(Fig. 13) are important for maintaining proper volume and/or composition of the endolymph. The FORE cell,here genetically defined, has many similarities with a morphologically identified cell type called `mitochondria-rich' cell.(Qvortrup and Bretlau, 2002). We would like to speculate that the FORE cell could prove to play a role in diseases characterized by altered endolymph turnover/composition, e.g. Pendred syndrome and Ménière's disease.

FOXI1 (also known as FKHL10), the human homologue ofFoxi1, has been mapped to the chromosomal localization of 5q32-34(Larsson et al., 1995). To this date no human disease has been linked to mutations in this gene. As much as 10% of all hereditary deafness could be due to Pendred syndrome(Fraser, 1965) and as many as 23% of the patients with a clinical diagnosis of Pendred syndrome lack mutations in the PDS gene(Reardon et al., 2000). This implies that mutations in FOXI1 could prove to be a cause of human deafness.

We thank Karen B. Avraham, Guy Van Camp, Claes Möller and William Reardon for valuable discussions, C. Birchmeier, E. Green, P. Gruss, C. Hodgetts, T. Lufkin, A. Matthews and C. Morton for sharing probes with us, H. Semb for making antisera available and Tim Folkard for help with setting up the three-dimensional reconstruction system. This work was made possible with support from the UK MRC and EC contract CT97-2715, The Swedish Medical Research Foundation, The Arne and IngaBritt Lundberg Foundation, The Juvenile Diabetes Foundation, The Wallenberg Foundation, The Swedish Foundation for Strategic Research (Nucleic Acid Program), a Junior Individual Grant from The Swedish Foundation for Strategic Research to S.E., an EMBO short term fellowship to M.H.