Deafness in mice lacking the T-box transcription factor Tbx18 in otic fibrocytes

The cochlea of the inner ear is a sensory apparatus that converts the mechanical stimulation of sound into electrical activity. A crucial factor in sound transduction, and thus hearing, is the maintenance of the ionic homeostasis of the endolymph, the extracellular fluid of the cochlear duct. The importance of fibrocyte integrity in this process has become apparent by pathological changes caused by inherited disorders or environmental stress. Genetic ablation of certain fibrocyte-expressed genes is known to cause deafness (Minowa et al., 1999; Teubner et al., 2003; Boettger et al., 2003; Delprat et al., 2005), and noise-induced, as well as age-related, hearing deficits are initiated by changes in fibrocyte physiology(Hequembourg and Liberman,2001; Hirose and Liberman,2003).

Otic fibrocytes represent a heterogeneous population of cells with special structural and molecular adaptations according to their location and physiological properties (Spicer and Schulte, 1991). Fibrocytes are found in the spiral limbus at the proximal site of the cochlea, and in the spiral ligament in the cochlear lateral wall, where five subgroups can be distinguished (see Fig. 1A). Type I fibrocytes underlie the stria vascularis, a specialized non-sensory epithelial thickening of the lateral wall, type II fibrocytes are situated under the spiral prominence, type III fibrocytes line, as a thin layer, the otic capsule, type IV fibrocytes are located lateral to the basilar membrane and anchor it to the lateral wall (Henson and Henson,1988), and type V fibrocytes reside above the stria vascularis. Fibrocytes of subtypes I, II and V are highly interconnected, and form a mesenchymal gap junction network. This and an independent epithelial network couple non-sensory supporting cells of the Organ of Corti with basal and intermediate cells of the stria vascularis (for a review, see Kikuchi et al., 2000). Basal cells form a multi-layered epithelial barrier that separates the extracellular spaces of the stria vascularis and the spiral ligament. Neural crest-derived intermediate cells form a discontinuous layer between basal cells and marginal cells that constitutes an epithelial barrier facing the endolymph in the cochlear duct (see Fig. 1B for a scheme of the cellular structure of the stria) (for reviews, see Forge and Wright, 2002; Raphael and Altschuler, 2003). The mesenchymal gap junction network plays a central role in ionic homeostasis. In fact, recycling of K⁺-ions through this network is pivotal for cochlear physiology. Strial marginal cells actively transport K⁺-ions into the endolymph to maintain a very high concentration in this compartment. A voltage gradient between the negative potential inside the sensory hair cells and the positive endocochlear potential (EP) in the endolymph, together with the concentration gradient in the same direction,drives the influx of K⁺-ions through apical mechano-sensitive channels and, thus causes the depolarization of hair cells. After secretion by hair cells and re-uptake by supporting cells, K⁺-ions are thought to travel through the epithelial and mesenchymal gap junction networks back to the stria vascularis (for reviews, see Kikuchi et al., 2000; Wangemann, 2002).

Despite the importance of otic fibrocytes for the physiology and pathology of hearing, little insight has been gained into the genetic circuits regulating fibrocyte development. Mice mutant for the transcription factor gene Pou3f4 (also known as Brn4) show ultrastructural alterations in fibrocyte morphology and exhibit a reduced EP and profound deafness (Minowa et al., 1999; Phippard et al., 1999). In mice mutant for otospiralin (Otos), a gene encoding a small extracellular matrix (ECM) protein of unknown function, fibrocytes type II and IV are degenerated (Delprat et al.,2005). Similar to Pou3f4, the precise role of Otos in fibrocyte differentiation is unknown.

This report defines a critical role in the development of otic fibrocytes for Tbx18, a member of the evolutionary conserved family of T-box transcription factors (Naiche et al.,2005). Mice carrying a null allele of Tbx18 die shortly after birth with severe malformations of the vertebral column and the rib cage(Bussen et al., 2004),prominent hydroureternephrosis (Airik et al., 2006) and defective caval veins(Christoffels et al., 2006),defects that have been traced to crucial functions of the gene in somite patterning, differentiation of the ureteric mesenchyme and myocardialization of caval veins, respectively. Here, we uncover an additional requirement for Tbx18 in the development of the inner ear. We show that adult Tbx18-deficient mice with rescued lethality display profound deafness, and we analyze the electro-physiological, histological and molecular changes that underlie this phenotype. We correlate the expression of Tbx18 in otic mesenchyme with defects in fibrocyte composition and stria vascularis integrity, and characterize the etiology of the defects.

Generation of Tbx18^lacZ and msd::Tbx18transgenic mouse lines was described previously(Bussen et al., 2004; Airik et al., 2006). The generation of a GFP allele of Tbx18 (Tbx18^GFP)that was interchangeably used with the lacZ allele will be described elsewhere. All mouse lines were maintained on an NMRI outbred background. Embryos for Tbx18 expression analysis were derived from matings of NMRI wild-type mice. Tbx18^-/- embryos were obtained from matings of Tbx18 heterozygotes. Mice compound heterozygous for a Tbx18 mutant allele and the msd::Tbx18 transgene were mated to derive viable adult double homozygous mice. Wild-type or heterozygous littermates were used as controls for mutant embryos and mice. Genomic DNA prepared from yolk sac or tail biopsies was used for genotyping by PCR. For timed pregnancies, vaginal plugs were checked in the morning after mating,noon was taken as embryonic day (E) 0.5. Embryos were dissected in phosphate-buffered saline (PBS) and fixed in 4% paraformaldehyde (PFA) in PBS.

Embryos were embedded in paraffin wax and sectioned to 5 μm. Inner ears were dissected from adult temporal bones and fixed in Bouin's fixative for 48 hours, decalcified in 0.5 M EDTA/PBS for 48 hours, dehydrated, paraffin wax embedded and sectioned to 5 μm. Sections were stained with Hematoxylin and Eosin or Picro-Sirius Red (F3BA, Sigma-Aldrich, USA). Histochemistry forβ-galactosidase activity was carried out on cryosections as described(Lobe et al., 1999).

For the preparation of ultra-thin sections, cochleae from anaesthetized 3-week-old Tbx18KO and control animals (one each) were quickly removed and perfused through the round window with 3% glutaraldehyde in cacodylat buffer, and then left in the fixative overnight. Fixed cochleae were decalcified in 10% EDTA (pH 7.3) at 4°C for 3 days, cut in two, post-fixed in 2% osmium tetroxide for 30 minutes, dehydrated in a graded series of ethanol solutions and embedded in Epon. Ultra-thin sections (60 nm) were stained with uranyl acetate and lead citrate and examined with a Zeiss 902 electron microscope.

Adult inner ears were fixed in 4% PFA overnight, decalcified in 0.5 M EDTA/PBS, pH 8, for 48 hours, embedded in paraffin wax and cut to 5 μm. Alternatively, decalcified cochleae were immersed in 30% sucrose/PBS for 12 hours, embedded in tissue-freezing medium (Leica, Germany), and cryosectioned to 5 μm. For the detection of antigens, the following primary antibodies and dilutions were used. Polyclonal rabbit antisera against Aqp1 (1:200,Alomone Labs), Barttin (1:100, gift from Friedhelm Hildebrandt)(Birkenhager et al., 2001),E-cadherin (1:200, gift from Rolf Kemler)(Vestweber and Kemler, 1984),Glut1 (1:200, Dianova), Kcc3 (1:500, gift from Thomas Jentsch)(Boettger et al., 2003), Kcnq1(1:250, gift from Thomas Jentsch) (Dedek and Waldegger, 2001), Kir4.1 (1:200, Alomone Labs), laminin(1:100, Sigma) and otospiralin (1:250, gift from Christian Hamel)(Delprat et al., 2002);polyclonal guinea pig-anti-connexin26 (1:200, gift from Johanna Brandner)(Brandner et al., 2004); and monoclonal mouse antibodies against Atp1a1 (Na-K-ATPase subunit 1a, clone a6F,1:500, developed by Douglas M. Farmbrough and obtained from the Developmental Studies Hybridoma Bank, University of Iowa) and Cldn11 (clone 37E3, 1:500,gift from Alexander Gow) (Gow et al.,2004). Fluorophore-coupled secondary antibodies were purchased from Dianova, Germany (Rhodamine-Red-X-conjugated donkey-anti-mouse,Rhodamine-Red-X-conjugated goat-anti-rabbit, Cy3-conjugated goat-anti-guinea pig), Invitrogen, USA (Alexa-488 conjugated donkey-anti-rabbit) and Santa Cruz Biotechnology, USA (FITC-conjugated goat-anti-mouse, FITC-conjugated goat-anti-rabbit), and used at a dilution of 1:200. Non-fluorescent staining was performed using kits from Vector Laboratories (Mouse-on-Mouse peroxidase kit, Vectastain ABC peroxidase kit (Rabbit IgG), DAB substrate kit). Labeling with the primary antibody was performed at 4°C overnight after antigen retrieval (2 mM EDTA in 0.01 M Tris-HCl, pH9, at 80°C, 1 hour) and blocking in 3% BSA, 3% normal goat serum for 30 minutes in PBST. For monoclonal mouse antibodies an additional IgG blocking step was performed using the Mouse-on-Mouse Kit (Vector Laboratories).

In situ hybridization analysis on 10 μm sagittal sections of E18.5 heads was performed following a standard procedure with digoxigenin-labeled antisense riboprobes (Moorman et al.,2001). Details of probes used are available upon request.

Sections were photographed using a Leica DM5000 microscope with a Leica DFC300FX digital camera. Laser scanning microscopy was performed using a Leica TCS SP2 microscope. All images were processed in Adobe Photoshop CS.

Auditory brainstem responses were measured as described previously(Boettger et al., 2002). In cases where no auditory response could be recorded at the highest obtainable level of 137 dB peak equivalent sound pressure level, the hearing threshold was set to this value for statistical analysis. Measurement of the endocochlear potential (EP) followed published procedures(Boettger et al., 2003).

In order to assess the spatial distribution of Tbx18 transcripts during inner ear development, we performed an in situ hybridization analysis of sagittal sections of mouse heads at different embryonic stages. Since we detected Tbx18 expression in otic mesenchyme, we compared on adjacent sections expression of the chondrogenic marker gene Sox9(Lefebvre and de Crombrugghe,1998) and expression of Pou3f4, whose expression marks the entire early periotic mesenchyme(Phippard et al., 1998)(Fig. 2). Inner ear expression of Tbx18 was first detected in ventral aspects of the mesenchyme surrounding the otic vesicle at embryonic day (E) 11.5(Fig. 2A). Tbx18expression partially overlapped Sox9 and Pou3f4 expression,whose domains, however, were clearly more expanded at this stage(Fig. 2B,C). Expression of Tbx18 and Sox9 started to become mutually exclusive at E12.5. Tbx18 expression was confined to the inner zone of the periotic mesenchyme in the whole inner ear, including the future cochlea and the vestibular portion (Fig. 2D,G; data not shown), whereas Sox9 expression was restricted to the outer zone, which undergoes chondrogenic differentiation to form the otic capsule from E13.5 onwards(Fig. 2E,H). Expression of Pou3f4 was found throughout the periotic mesenchyme, overlapping Tbx18 and Sox9 expression until E13.5(Fig. 2F,I). From E15.5 until E18.5, Tbx18 expression was restricted to prospective otic fibrocytes of the spiral limbus and the spiral ligament(Fig. 2J,M). Expression was upregulated in condensing mesenchyme underlying the stria vascularis at E18.5(Fig. 2P). Sox9expression was downregulated in the otic capsule after E15.5, but was found in condensing mesenchymal cells at E18.5 (Fig. 2K,N,Q). Pou3f4 and Tbx18 expression domains overlapped from E15.5 to E18.5 (Fig. 2L,O,R). Tbx18 expression was not detected at postnatal day (P) 21, when inner ear development is completed (data not shown). Hence, Tbx18 expression was restricted to an inner zone of otic mesenchyme fated to differentiate into otic fibrocytes, throughout inner ear development.

Mice homozygous for a null allele of Tbx18 die shortly after birth due to severe malformation of the vertebral column. This phenotype was traced to a requirement for Tbx18 in the anterior-posterior polarization of somites (Bussen et al., 2004). To evaluate the functional significance of Tbx18 expression in otic mesenchyme, we sought to overcome the perinatal lethality of Tbx18-deficient mice by re-introducing somitic expression of Tbx18. We used transgenic msd::Tbx18 mice that express Tbx18 throughout the presomitic and somitic mesoderm, but not in the inner ear (Bussen et al.,2004) (see Fig. S1 in the supplementary material), to generate mice compound mutant for a Tbx18 loss-of-function allele and the msd::Tbx18 transgene. Mice double homozygous for Tbx18 and msd::Tbx18 were born in the expected Mendelian ratio. They were smaller than their littermates and exhibited a general impaired mobility,explainable by skeletal defects. They survived for 3-4 months before they died from hydronephrotic lesions. Although vestibular function seemed normal in these mice, acoustic Preyer reflexes were absent, indicating a severe hearing deficit. We examined auditory function by measuring auditory brainstem responses (ABRs) to clicks, with an upper frequency limit of 5.5 kHz, shortly after the onset of hearing (at P14) in three- and twelve-week-old Tbx18/Tbx18,msd::Tbx18/msd::Tbx18 (Tbx18KO) mice and Tbx18/+,msd::Tbx18/+ (control) littermates. Tbx18KO mice showed an ABR with clicks of about 130 dB peak equivalent sound pressure level(pe SPL) at both time-points; the average increase was >75 dB above the hearing threshold of control animals (Fig. 3A). Thus, Tbx18KO mice had a pronounced hearing loss at three weeks of age.

As the excitability of hair cells depends on the magnitude of the EP, a reduced EP would explain the elevation of the ABR threshold in Tbx18KO mice. Evaluation of the EP in Tbx18KO mice at P21 revealed a complete breakdown of the EP(Fig. 3B,C). Together, these results demonstrate that loss of Tbx18 expression in otic mesenchyme leads to an abnormal otic physiology and a severely compromised auditory function.

Histological analysis of three-week-old Tbx18KO inner ears did not reveal any obvious changes in cochlear shape, but an overall reduction in size was apparent (Fig. 4B), and the spiral ganglion was not completely surrounded by bone tissue of the modiolus(Fig. 4D). The lateral wall,including stria vascularis and otic fibrocytes, of the spiral ligament was severely hypoplastic, the spiral prominence was variably reduced (white arrow in Fig. 4F,J). Suprastrial(type V) fibrocytes were absent, and type IV fibrocytes lateral to the basilar membrane were partially replaced by bone tissue (arrow and arrowhead in Fig. 4F,J). The number of cells underlying the stria vascularis was drastically diminished and the spiral prominence was different from that of controls, indicating additional defects in type I and II fibrocytes (Fig. 4D,F). The stria vascularis of Tbx18KO mice was variably reduced, with a thin stripe of flat marginal cells extending into Reissner's membrane (Fig. 4J). Ultrastructural analysis of the stria vascularis architecture at 3 weeks of age revealed presence of marginal, intermediate and basal cells, but their densely packed membranous infoldings were severely reduced in Tbx18KOanimals (Fig. 4L). Phenotypic changes in the lateral wall gradually increased in severity towards the basal end of the cochlea duct. The Organ of Corti appeared normal at this stage(Fig. 4H). At 12 weeks of age,defects of the lateral wall had increased in severity. The stria vascularis was flattened to a thin layer of simple squamous epithelial cells (see Fig. S2 in the supplementary material).

Hence, the cyto-architecture of the lateral wall, including the fibrocytes of the spiral ligament and the epithelial barrier cells of the stria vascularis, is severely affected by the loss of Tbx18 function in development.

Expression of Tbx18 in otic fibrocytes, disruption of the EP and histological changes in the lateral wall of Tbx18-deficient mice suggested a role for Tbx18 in the differentiation of otic fibrocytes. To explore this possibility in more detail, we analyzed the distribution of proteins defining fibrocyte subtypes by immunohistochemistry.

Otospiralin (Otos), an ECM protein of unknown function, is expressed in all otic fibrocytes (Delprat et al.,2002). In Tbx18KO mice, Otos was present throughout the spiral ligament, albeit with decreased expression, particularly in the region(of type IV fibrocytes) beneath the basilar membrane (arrow in Fig. 5B). The gap junction protein connexin 26 (Cx26; also known as Gjb2 - Mouse Genome Informatics)marks type I fibrocytes that underlie the stria vascularis(Xia et al., 1999). Cx26 staining was absent in Tbx18KO spiral ligament fibrocytes(Fig. 5D). Expression of Atp1a1, the alpha1 polypeptide of the Na⁺/K⁺-transporting ATPase, is confined to marginal cells of the stria vascularis and to type II, IV and V fibrocytes of the spiral ligament (Xia et al.,1999). In Tbx18KO mice, the Atp1a1 expression domain was unchanged but the level of expression in spiral ligament fibrocytes appeared reduced (Fig. 5F). Aquaporin 1(Aqp1) expression, which is normally restricted to bone lining fibrocytes type III (Li and Verkman, 2001)(arrowheads in Fig. 5G), was lost in the mutant (Fig. 5H). Kcc3 (also known as Slc12a6 - Mouse Genome Informatics) represents a potassium-chloride co-transporter whose expression is found in type I, III and V fibrocytes (Boettger et al.,2003). In the mutant, expression of Kcc3 at strongly reduced levels was homogenous in the spiral ligament(Fig. 5J). Together,histological and immunohistochemical analyses revealed that terminal differentiation of fibrocytes into clearly distinct subpopulations was severely disturbed in the spiral ligament of Tbx18KO mice(Fig. 5K).

To evaluate the histological changes in the Tbx18KO stria vascularis more carefully, we analyzed the localization of marker proteins by immunofluorescence. The potassium channel Kcnq1 is confined to the apical surface of strial marginal cells, whereas the chloride channel subunit Barttin is localized baso-laterally (Estevez et al., 2001). Subcellular localization of Kcnq1 and Barttin was unchanged in Tbx18KO mice (Fig. 6B,D), suggesting the presence of marginal cells with normal apicobasal polarity. However, the area of Barttin staining was reduced and exhibited an irregular shape, possibly indicating an improper formation of baso-lateral projections (inset in Fig. 6D). Expression of Kir4.1 (also known as Kcnj10 - Mouse Genome Informatics), an inwardly-rectifying potassium channel of intermediate cells(Ando and Takeuchi, 1999), was reduced in the Tbx18KO stria vascularis(Fig. 6F), indicating that the differentiation of intermediate cells, including the formation of membranous infoldings (inset in Fig. 6F)was severely affected. The glucose transporter Glut1 (also known as Slc2a1 -Mouse Genome Informatics) exhibits strong expression in basal and endothelial cells of the stria vascularis (Ito et al.,1993). In the mutant, Glut1 expression was found only in endothelial cells of the few remaining strial vessels(Fig. 6H). The absence of expression of the basal cell-specific protein claudin 11 (Cldn11)(Gow et al., 2004; Kitajiri et al., 2004) in Tbx18KO mice (Fig. 6J)confirmed the dramatic reduction of the basal cell layer.

Extensive elongation and interdigitation of strial cells relies on the controlled degradation of the basal lamina underlying the marginal cell layer around birth (Sagara et al.,1995; Kikuchi and Hilding,1966). In addition to the basal lamina of strial vessels, laminin staining detected the presence of a basal lamina underlying marginal cells in Tbx18KO mice (Fig. 6L), which is in agreement with the observed reduction of baso-lateral surface projections of marginal cells. Thus, the cellular architecture of the stria vascularis in Tbx18KO is severely disturbed, with a massive decrease of basal cell number, and a vast reduction of membraneous projections of intermediate and marginal cells(Fig. 6M).

The occurrence of strial defects upon loss of Tbx18 in otic mesenchyme suggested a cellular or trophic contribution of the otic mesenchyme to strial development. To distinguish these possibilities, we analyzed Tbx18^-/- inner ears at E18.5, when stria vascularis maturation is initiated (Kiernan et al.,2002; Xia et al.,1999). Histological examination revealed condensation of mesenchymal cells underneath the forming stria in the wild type, whereas in Tbx18^-/- mice cells underneath the stria remained loosely organized at this stage (arrow in Fig. 7B). During our studies, we established that these mesenchymal condensates are characterized by the expression of Sox9, E-cadherin(also known as Cdh1 - Mouse Genome Informatics) and Cx26 in the wild type (Fig. 7C,E,G). Expression of all three genes was lost in the Tbx18-deficient spiral ligament, although the epithelial expression domains of these genes appeared unaffected (Fig. 7D,F,H).

Further markers were used to assess cytodifferentiation of the stria vascularis. Expression of Bsnd, which encodes Barttin, in prospective marginal cells (Birkenhager et al.,2001) was detected in the mutant, suggesting proper temporal regulation of marginal cell differentiation(Fig. 7J). Dct (also known as Trp2) is a marker for prospective intermediate cells(Steel et al., 1992). Dct expression in the mutant was indistinguishable from in the wild type, demonstrating normal homing of future intermediate cells(Fig. 7L). We did not detect changes in the BrdU incorporation of cells of the spiral ligament at E18.5 and E12.5 (n=3 each, data not shown), which suggests that proliferation defects are not a causative agent.

Together, these results suggest that loss of Tbx18 primarily affects the development of the mesenchymal portion of the cochlea, including the cytodifferentiation of otic fibrocytes. Defects of the stria vascularis may originate from the failure of mesenchymal cells to condense and undergo an epithelial transition to form basal cells.

Tbx18 expression shows an early restriction to the inner ring of otic mesenchyme that will give rise to otic fibrocytes. We wondered whether the observed defects in otic fibrocyte differentiation might be caused, or at least be affected, by a mis-patterning of the otic mesenchyme, indicating an early role for Tbx18 in compartmentalization of this tissue. Improper radial patterning of the otic mesenchyme should reflect in misallocation of cells to the inner and outer compartment, i.e. to otic fibrocytes and the otic capsule,and/or in improper boundary formation. Histological analysis of E18.5 inner ears revealed a local thickening of the otic capsule adjacent to the spiral ligament of Tbx18^-/- mice (arrow in Fig. 8E), and an altered appearance of otic fibrocyte precursors in the spiral ligament of the basal coil (Fig. 8E). Fibrocytes appeared highly condensed with a preferential parallel orientation to the border of the otic capsule. Collagen staining with Picro-Sirius Red showed a distinct boundary between the fiber systems of the bony otic capsule and the spiral ligament in the wild type at this stage (arrowhead in Fig. 8B). By contrast, in Tbx18^-/- mice, the collagen fiber network appeared to be continuous between the two regions (Fig. 8F). Following the indications of severe histological changes at the interface between otic capsule and otic fibrocytes, we analyzed expression of periostin (Postn), a gene expressed in certain types of fibrous connective tissue, such as the bone-lining periosteum(Horiuchi et al., 1999). Postn expression was restricted to cells lining the outer border of the otic capsule and to a small domain in the proximobasal part of the spiral ligament in the wild type at E18.5 (Fig. 8C). In Tbx18^-/- inner ears, Postnwas expressed in a distal to proximal gradient throughout the spiral ligament(Fig. 8G). By contrast,expression of Coch, which we found to be restricted to a proximal region of the otic mesenchyme underlying the epithelium of the lateral wall in the wild type at this stage (Fig. 8D), was absent in the mutant(Fig. 8H). Expression of Pou3f4 and Otos, which was found throughout the otic mesenchyme in the wild type, was unchanged in Tbx18^-/-mice at E18.5 (data not shown). Together, these findings suggest that in the Tbx18^-/- inner ear, the outer compartment of the otic capsule has expanded at the expense of the inner compartment. Mesenchymal cells of the inner compartment have acquired some basal characteristics of otic fibrocytes, but seem to differentiate into periosteum-like connective tissue rather than into distinct fibrocyte subtypes.

We next wished to determine more directly how and when compartmentalization of the otic mesenchyme is affected in the Tbx18^-/- inner ear. Ideally such an analysis would rely on a genetic fate labeling system that traces the descendants of the inner and outer mesenchymal compartment during development. Presently, such tools that would, for example, be based on Tbx18 and Sox9 alleles with inducible cre activity are not available. To circumvent this restriction, we employed β-galactosidase activity from the Tbx18^lacZ allele for short-time lineage tracing, to follow the fate of Tbx18-expressing cells in early inner ear development (Fig. 8I-N). In the heterozygous control (Tbx18^lacZ/+),β-galactosidase activity was restricted to the inner ring of the otic mesenchyme from E12.5 to E14.5, suggesting that compartmentalization of prospective otic fibrocytes occurs as early as E12.5 during development(Fig. 8I-K). By contrast, in Tbx18 mutant mice with normalized β-galactosidase activity from the Tbx18^lacZ/Tbx18^GFP genotype, theβ-galactosidase-positive area expanded into the outer ring of condensing mesenchymal cells at E13.5 (Fig. 8M) and reached the outer lining of the cartilagenous otic capsule at E14.5 (Fig. 8N). Hence,cells that normally express Tbx18 are no longer restricted to the inner compartment of the otic mesenchyme.

The establishment and maintenance of the EP as an essential prerequisite for normal hearing depends on the structural and functional integrity of the fibrocyte network of the lateral wall and the stria vascularis. This study shows that the T-box transcription factor gene Tbx18 is necessary for the proper maintenance of compartment boundaries in the periotic mesenchyme and the differentiation of fibrocyte subtypes from a common precursor population, and provides evidence that basal cell formation is a crucial step of stria vascularis formation that relies on proper fibrocyte integrity.

To our knowledge, Tbx18 is the earliest molecular marker that shows radially restricted expression in the periotic mesenchyme, demonstrating that molecular subdivision into an inner and outer compartment occurs shortly after mesenchymal aggregation around the otic vesicle. Although we did not formally address the fate of the Tbx18-positive otic mesenchyme by genetic lineage tracing experiments, it is likely that Tbx18expression from at least E12.5 onwards marks prospective otic fibrocytes. Evidence derives from the mutually exclusive expression domains of Tbx18 and the chondrogenic marker gene Sox9, which in turn becomes quickly restricted to the outer ring of mesenchyme destined to form the otic capsule. In addition, short-time lineage tracing experiments with a lacZ reporter in the Tbx18 locus revealed an early restriction of β-galactosidase activity to the inner compartment of mesenchymal cells.

The spatial restriction of Tbx18 expression in the developing inner ear is compatible with a role of this transcriptional regulator either in patterning of the periotic mesenchyme in capsule versus fibrocyte compartments, or in specification/differentiation of otic fibrocytes. Dramatic changes in subtype composition of lateral wall fibrocytes in the Tbx18-deficient inner ear support a function of Tbx18 in fibrocyte differentiation. However, for several reasons, we suggest that defects in fibrocyte differentiation are subordinate to a primary requirement of Tbx18 in patterning or compartmentalization of the otic mesenchyme. First, loss of type I, III and V, and reduction of type IV fibrocytes shows that populations residing in close proximity to the bony capsule are predominantly affected by the loss of Tbx18, although the expression of Tbx18 appears homogenous. More importantly, failure to restrict cells that normally express Tbx18 to the inner compartment in the mutant after E12.5, local loss of a distinct capsule-fibrocyte boundary at E18.5 coupled with histological alterations of fibrocytes, ectopic expression of Postn, and an absence of Cx26 and Coch expression at E18.5 indicate a disturbance of radial patterning of the periotic mesenchyme and a failure to maintain the boundary between mesenchymal compartments destined to give rise to otic capsule and otic fibrocytes. The acquisition of a fate of periosteum-like connective tissue may hamper further differentiation of fibrocytes into the different subtypes.

The failure to restrict cells that normally express Tbx18 to the inner compartment of the otic mesenchyme may have one of several molecular explanations. First, Tbx18^-/- cells no longer recognize a repulsive signal emanating from the outer mesenchymal compartment to restrict their migration or intermingling with these cells. Second, Tbx18-mutant cells have lost selective adhesiveness. Alternatively,the loss of an inhibitory signal from the Tbx18-mutant cells leads to ectopic induction of Tbx18 in the outer ring of mesenchymal cells. Notably, Tbx18 has also been implicated in the maintenance of compartment boundaries in somites and in the ureteric mesenchyme. In Tbx18^-/- somites, posterior somite characteristics expand anteriorly (Bussen et al.,2004), whereas in the metanephric field ureteric mesenchymal cells fail to coalesce onto the ureteric epithelium(Airik et al., 2006). It remains to be explored whether these phenotypes can be rationalized by the disruption of a common molecular program.

To date, a requirement for the proper differentiation of otic fibrocytes has only been demonstrated for Pou3f4(Minowa et al., 1999; Phippard et al., 1999). Similar to Pou3f4^-/- mice, differentiation of fibrocytes is severely compromised in Tbx18-deficient mice. However, reduction but not loss of Pou3f4 expression and more severe phenotypic changes in Tbx18^-/- animals suggests that there is not a simple epistatic relationship between the two transcription factor genes. More likely, the two genes act in parallel genetic circuits regulating patterning and differentiation of otic mesenchyme.

Despite the pivotal role of the stria vascularis in auditory function,little is known about the developmental processes and the underlying molecular pathways involved in its formation. Our study provides molecular evidence that the basal cell layer of the stria vascularis forms by aggregation from, and subsequent mesenchymal-epithelial transition (MET) of, adjacent otic fibrocytes.

Evidence is provided by the co-regulation of Sox9, Cx26 and E-cadherin in condensing mesenchymal cells beneath the stria vascularis. Sox9, a gene encoding an HMG-type transcription factor, has been implicated in the regulation of mesenchymal cell condensation in various contexts (Bi et al., 2001; Akiyama et al., 2004), thereby supporting a similar role in this setting. A direct regulation of Sox9 by Tbx18 is possible. However, it is more likely that the loss of Sox9 expression is secondary to a prior differentiation defect of the otic mesenchyme. Intriguingly, Cx26 expression in this domain precedes the onset of K⁺-cycling through the gap junction network of the spiral ligament (Sadanaga et al., 1995; Yamasaki et al., 2000). Hence,the expression of Cx26 might indicate a role in compartmentalization of cells by gap junction-mediated cell-cell communication (for a review, see Levin, 2007) (see also Ackert et al., 2001). Finally,E-cadherin expression has been implicated in a variety of tissue remodeling processes. The forced expression of E-cadherin in fibroblasts is sufficient to drive them into epithelial cells, whereas the loss of E-cadherin in epithelial cells has been linked to mesenchymal transitions (for a review, see Thiery and Sleeman, 2006) (see also Vanderburg and Hay,1996). The absence of E-cadherin expression in condensing mesenchymal cells beneath the stria vascularis in the Tbx18^-/- inner ear strongly supports a functional role of E-cadherin in MET as a prerequisite for basal cell layer formation.

The almost complete loss of basal cells in Tbx18^-/-mice offered the opportunity to study the role of these cells for marginal and intermediate cell differentiation and maturation. Intermediate cell projections were reduced in the Tbx18^-/- stria vascularis at three weeks of age, although precursor cells were established normally at E18.5. This favors the idea that expansion, maintenance and/or differentiation of intermediate cells depends on the presence of an intact basal cell layer. Marginal cells showed a mature phenotype with the formation of basolateral processes and expression of the Na⁺/K⁺-ATPase-α1-isoform(Erichsen et al., 1996). The collapse of Reissner's membrane that has been reported for certain mutants with severely impaired endolymphatic K⁺-secretion was not apparent in Tbx18^-/- mice, suggesting that marginal cells became functional (Delpire et al.,1999; Vetter et al.,1996). We noted, however, that the extension of basolateral processes of marginal cells and interdigitation with intermediate cells was severely compromised, showing that parts of the differentiation program of marginal cells depend on basal and intermediate cells. We propose that basal cell-mediated degradation of the basal lamina is a prerequisite for these processes to occur.

The measurement of ABRs revealed profound deafness in Tbx18-deficient mice. At three weeks of age, i.e. shortly after the onset of hearing, the EP was completely abolished. Several of our findings suggest that the defect of otic fibrocyte differentiation in Tbx18KOmice structurally and functionally interferes with the establishment of a normal EP, both at the level of its strial generation and with the recycling of K⁺-ions by the mesenchymal gap junction network.

First, it has recently been demonstrated that expression of the potassium inwardly-rectifying channel Kir4.1 in strial intermediate cells is required for EP generation in the mouse (Marcus et al., 2002). Second, functional integrity of the basal cell layer of the stria vascularis is necessary to establish a distinct intrastrial compartment that, in turn, is required for generation of the EP. Previous studies using Cldn11 mutant mice showed that loss of the tight junction barrier in basal cells causes a strong decrease of the EP(Gow et al., 2004; Kitajiri et al., 2004). Third,fibrocytes of the spiral ligament have been implicated in the uptake and transport of perilymphatic K⁺-ions (for a review, see Kikuchi et al., 2000). Fourth,mice mutant for connexin 30, which exhibits overlapping cochlear expression and forms heteromeric gap junction channels with connexin 26, do not establish a normal EP, illustrating the importance of the spiral ligament gap junction network (Forge et al., 2003; Teubner et al., 2003). Hence,the lack of Kir4.1-positive intermediate cells and the almost complete absence of Cldn11-expressing basal cells in the stria vascularis, the failure in terminal fibrocyte differentiation and the absence of Cx26 expression in the spiral ligament collectively explain the complete abrogation of the EP in Tbx18-deficient mice.

In conclusion, we have shown that lack of the T-box transcription factor Tbx18 in otic mesenchyme leads to changes in the compartmentalization and differentiation of otic fibrocytes, and to subsequent defects in stria vascularis formation. Cooperatively, these defects result in a failure to generate a normal EP, dramatically demonstrating the importance of the mesenchymal cells of the lateral wall for auditory function.

We thank Johanna Brandner, Lorraine Everett, Andrew Forge, Alexander Gow,Christian Hamel, Hiroshi Hibino, Thomas Jentsch, Rolf Kemler, Tetsuo Noda,William J. Pavan and Nahid Robertson for reagents, and Rannar Airik, Henner Farin and Achim Gossler for discussion and critical reading of the manuscript. The monoclonal mouse antibodies against Atp1a1 (Na-K-ATPase subunit 1a, clone a6F, developed by Douglas M. Farmbrough were obtained from the Developmental Studies Hybridoma Bank maintained by the University of Iowa. This work was supported by a grant from the German Research Council (DFG KI728/2) to A.K.