Otx1 and Otx2 activities are required for the normal development of the mouse inner ear

Otx1 and Otx2 are murine orthologues of the Drosophila orthodenticle (otd) gene (Simeone et al., 1992). These genes are transcription factors containing a bicoid-like homeobox (Simeone et al., 1992). In Drosophila, otd is important for head formation and different levels of otd protein are required for the formation of specific subdomains of the adult head (Hirth et al., 1995; Royet and Finkelstein, 1995). otd and its orthologues share sequence homology and have been shown to substitute for each other functionally (Acampora et al., 1998a,b, 1999; Leuzinger et al., 1998). In mouse, both Otx genes are involved in regional specification and proper morphogenesis of brain structures. Otx1 is expressed in the presumptive forebrain and midbrain regions of the neural tube of mice starting at 8 dpc (days post coitum) and the absence of this gene product causes malformations of the telencephalon, mesencephalon and cerebellum (Simeone et al., 1993; Acampora et al., 1996). Otx1^{−/ −} mutant mice also display spontaneous epileptic behavior (Acampora et al., 1996). Otx2 gene expression is activated in the entire ectoderm and visceral endoderm of the mouse before gastrulation, and is restricted to the anterior part of the embryo by 7.5 dpc (Simeone et al., 1993; Ang et al., 1994). Mice homozygous for Otx2 are early embryonic lethal and lack anterior neural tissues (Ang et al., 1996; Matsuo et al., 1995; Acampora et al., 1995). Otx1^{−/ −};Otx2^{+/ −} mice also display a more severe brain phenotype than Otx1^{−/ −} alone suggesting that the dosage of Otx gene products is important for normal development (Acampora et al., 1997).

Both Otx1 and Otx2 are expressed in sense organs such as the inner ear, eye and olfactory epithelium (Simeone et al., 1993). The absence of these gene products affects the formation of these sense organs (Acampora et al., 1996; Matsuo et al., 1995). Otx1^{−/ −} mutant mice showed a rapid turning behavior and their inner ears lacked the lateral semicircular canal (Acampora et al., 1996). This defect of Otx1^{−/ −} mutant mice was not rescued with the otd or hOtx2 cDNA, suggesting a unique role of Otx1 in inner ear development (Acampora et al., 1998, 1999).

In this study, we examined the normal expression of the two Otx genes in the mouse inner ear and the characterization of Otx1^{−/ −}, Otx1^{−/ −};Otx2^{+/ −} as well as hOtx2¹ mutant ears. The mouse inner ear is a complex organ that is derived embryonically from a rudimentary structure known as the otocyst. There are six major sensory organs, three cristae, two maculae and a cochlea, that are responsible for vestibular and auditory functions. Each sensory organ requires a precise pattern and position within the inner ear in order to properly mediate its function. The molecular mechanisms that underlie these various features of sensory organ formation are largely unknown. Here, we demonstrate that the lack of Otx1 and Otx2 gene products affected the development and final positioning of some of the sensory organs within the mouse inner ear, as well as the shaping of its cochlea. The difference in the phenotype of Otx1^{−/ −} and Otx1^{−/ −};Otx2^{+/ −} inner ears and the inability of hOtx2 cDNA to rescue defects of Otx1^{−/ −} inner ears suggest that these two genes play separate roles in inner ear development. These data are also discussed in light of the recently proposed boundary model for sensory organ specification in the inner ear.

Mice were generated and genotyped as described (Acampora et al., 1996, 1997, 1999). Mutant mice have the same mixed genetic background as 129 Sv, BL6 and DBA2. All embryos were individually staged according to the method of Theiler (1989).

The RNA probes for Fng, BMP4 and Brain3.1 (Brn3.1) were prepared as previously described (Morsli et al., 1998). The probe for Otx1 was a 396 nucleotide SacI-PstI fragment (Simeone et al, 1993). Approximately 1 kb of each respective cDNA was used to generate probes for Otx2 (Ang et al., 1994) and TRP-2 (Steel et al., 1992). The probe for the 68 kDa neurofilament protein was a PCR product generated between nucleotides 851 and 1350 (gift of XianJie Yang, UCLA).

All embryos were harvested and injected as previously described (Morsli et al., 1998). Inner ears injected were as follows: for wild type, five 12 dpc and five 15 dpc; for Otx1^+/ −, four 15.5 dpc; for Otx1^+/ −;Otx2^+/ −, two 15.5 dpc; for Otx1 −/ −, four 12 dpc, two 15 dpc and seven 15.5 dpc and two 16 dpc; for Otx1 −/ −;Otx2^+/ −, four 12 dpc, two 15 dpc and seven 15.5 dpc; for hOtx2¹, seven 16 dpc and two postnatal day 1 (P1).

Whole-mount in situ hybridization was performed as previously described (Riddle et al., 1993) with the following modifications. All embryos were permeabilized with proteinase K (Boehringer Mannheim, Indianapolis, IN) using concentrations from 1 to 15 μg/ml. Hybridization, washings and detection procedures were performed as described by Riddle et al. (1993).

Frozen 12 μm sections for both wild-type and mutant embryos were processed for in situ hybridization as previously described (Morsli et al., 1998). The mutants analyzed were as follows: for Otx1⁻/⁻, two 10.5 dpc, five 11.5 dpc, three 12 dpc, three 13 dpc, three 14 dpc and four 18 dpc; for Otx1⁻/⁻;Otx2^{+/ −}, two 12 dpc, one 14 dpc and five 18 dpc. Each specimen was serially sectioned at the level of the inner ear and adjacent sections were processed for two or more probes.

After analysis of paint-injected inner ears, specimens were transferred and embedded in paraffin. All embryos were sectioned at 10 μm.

The histology of the mouse inner ear becomes distinct at 14 dpc. In order to characterize the domains of Otx1 and Otx2 expression in earlier developmental stages, we have used previously established markers for the presumptive sensory organs. Bone morphogenetic protein 4 (BMP4) was used as a marker for the three presumptive cristae, whereas lunatic fringe (Fng) served as a marker for the cochlea, maculae utriculi and sacculi at early stages, and for all sensory organs including the cristae for embryos 13 dpc and older.

Both Otx1 and Otx2 were expressed in mostly nonsensory regions of the developing inner ear. Otx1 was expressed in a region destined to form the lateral canal, lateral ampulla, utricle and pars inferior. Otx2 was expressed mostly in the region destined to form the pars inferior.

At 10.25 dpc, Otx1 mRNA was detected in the posteroventral quadrant of the lateral part of the otic vesicle (arrow in Fig. 1A). Otx2 transcripts were localized at the most ventral tip of the vesicle, where it was coexpressed with Otx1 (arrow in Fig. 1B). At 11.5 dpc, BMP4 was expressed in the lateral part of the otocyst as an anterior streak which encompassed the presumptive anterior and lateral cristae (Morsli et al., 1998). Fig. 1C′ illustrates the most posterior portion of the anterior streak. Otx1 expression domain, at 11.5 dpc, also remained in the most ventrolateral part of the otocyst (blue stripes in 11.5 dpc drawing of Fig. 1). Its dorsal boundary most likely corresponded to the presumptive lateral canal plate region judging by the position of the anterior streak of BMP4 domain at the same level (compare Fig. 1C and 1C′). It is interesting that the anterior portion of the Otx1 boundary overlapped with the posterior portion of the anterior streak of BMP4 expression (compare Fig. 1C and 1C′). This region most likely gives rise to the lateral crista (Morsli et al., 1998). The Otx1 expression domain extended ventrally to the tip of the cochlear anlage. Otx2 transcripts also remained in the ventrolateral part of the otocyst, thus being coexpressed with Otx1 (red stripes in 11.5 dpc drawing of Fig. 1) starting in the middle of the presumptive utricle (Fig. 1G) and extending ventrally to the tip of the cochlear anlage.

At 12 dpc, the Otx1 expression domain remained in the most ventrolateral part of the developing inner ear (blue stripes in 12 dpc drawing of Fig. 1). Its dorsal boundary was at the level of the lateral canal plate which, at this stage, was fully developed (lp in Fig. 1D). The anterior portion of the boundary overlapped with the presumptive lateral crista, as highlighted by BMP4 expression (lc in Fig. 1D′, compare to Fig. 1D). At the level of the presumptive macula utriculi, as highlighted by the Fng expression pattern, the Otx1 expression domain was lateral but not overlapping with the Fng domain (mu in Fig. 1E′, compare to Fig. 1E). At the cochlear level, Otx1 was also expressed in the lateral wall, not overlapping with the Fng-positive sensory tissue of the cochlea (Fig. 1F and F′). At 13 dpc, the pattern of Otx1 expression remained the same. From 14 dpc and thereafter, no Otx1 transcripts were detected in the inner ear.

At 12 dpc, Otx2 transcripts remained co- expressed with Otx1 in the ventrolateral part of the developing inner ear (red stripes in 12 dpc drawing of Fig. 1). The dorsal boundary of its expression domain remained in the middle of the utricular anlage (Fig. 1H), extending ventrally to the tip of the coiling cochlea. By 13 dpc, Otx2 transcripts were localized ventral to the macula utriculi, and restricted to the lateral wall of both the saccule and cochlea. This expression pattern remained the same at least until P5, where Otx2 transcripts localized to the lateral wall of the saccule (arrowhead, Fig. 1I) and Reissner’s membrane of the cochlea (arrowhead, Fig. 1J). Otx1 and Otx2 transcripts did not overlap with BMP4 or Fng in the developing cochlea.

By 15 dpc, the mouse inner ear had nearly attained its mature shape (Fig. 2A). The pars superior consisted of all three canals and ampullae, the endolymphatic duct and sac, and the utricle. The pars inferior consisted of the saccule and the coiled cochlea. The saccule connected to the utricle and the cochlea via the utriculosaccular (usd in Fig. 2A) and cochleosaccular (csd in Fig. 2A) ducts, respectively. The dorsal tip of the cochlea is defined as the hook. The cochlea consisted of 1.5 turns at this age and reached its mature 1.75 turns by 16 dpc (data not shown). Membranous labyrinths of all Otx1^{+/ −};Otx2^{+/ −} mutants showed normal morphology at this stage (data not shown).

Both Otx1^{−/ −} and Otx1^{−/ −};Otx2^{+/ −} mutants exhibited common consistent phenotypes (Fig. 2B-F). In the pars superior, the lateral semicircular canal and ampulla were absent. In the pars inferior, the utriculosaccular and cochleosaccular ducts were absent. As a result, the utricle, saccule, and cochlea were not separated from one another. In addition, the hook of the cochlea was poorly defined.

In addition to the consistent phenotypes, both Otx1^{−/ −} and Otx1^{−/ −};Otx2^{+/ −} mutants exhibited variable phenotypes in the pars inferior, which we classified according to severity (Table 1). Both Otx1^{−/ −} and Otx1^{−/ −};Otx2^{+/ −} mutants exhibited mild to severe phenotypes. The mildest of phenotypes displayed a nearly normal saccule and cochlea (Fig. 2B,D). More severe phenotypes showed a saccule, either abnormally small or mis-shaped, and a coiled but abnormal cochlea (Fig. 2C,E). The most severe phenotypes consisted of an abnormally shaped saccule and an aberrant cochlea that failed to coil or coiled in the opposite direction (Fig. 2F). Overall, Otx1^{−/ −} mutants exhibited milder phenotypes than Otx1^{−/ −};Otx2^{+/ −} mutants (Fig. 3). The mixed genetic background of these mutant mice most likely contributed to the variability of the phenotype observed. In support of this, the severity of the phenotypes between the right and left ears within the same animal was comparable (asterisk in Table 1).

Since the inner ear phenotypes of Otx1^{−/ −} and Otx1^{−/ −};Otx2^{+/ −}mutants were scored between 15 and 16 dpc (Table 1), a?cochlea of less than 1.5 turns could have represented a delay in development rather than a true malformation. Therefore, 10 Otx1^{−/ −} mice and 3 Otx1^{−/ −};Otx2^{+/ −} mice were analyzed at P1, focusing on the shape of the cochlea. Results of the dissected membranous labyrinths showed that most of the cochleae in Otx1^{−/ −} mice were fairly normal in shape, whereas all three cochleae in Otx1^{−/ −};Otx2^{+/ −} mice were abnormal. These results suggest that, while some of the Otx1^{−/ −} cochlear phenotypes at

15.5 dpc may represent a developmental delay, most of the more severe cochlear phenotypes in Otx1^{−/ −};Otx2^{+/ −} mice were indeed permanent defects. In addition, the hook of the cochlea of the inner ears analyzed at P1 remained poorly defined (data not shown).

The lateral semicircular canal originated as an outpouch or “lateral plate” from the utricle. Opposing epithelium in the center of the canal plate soon came together, fused and reabsorbed, leaving a mature canal-like structure. Since both Otx1^{−/ −} and Otx1^{−/ −};Otx2^{+/ −} mutants did not possess a lateral semicircular canal and ampulla at 15 dpc, we analyzed inner ears at 12 dpc, a stage when the lateral canal plate was fully developed (Fig. 4A,B). Both Otx1 −/ −?and Otx1^{−/ −};Otx2^{+/ −} mutants exhibited a rudimentary lateral canal plate at 12 dpc, which failed to develop normally (Fig. 4E,F). We then analyzed the generation of both anterior and lateral cristae using BMP4 gene expression as a marker. Fig. 4C illustrates BMP4 expression in a normal inner ear, showing the two presumptive cristae (anterior and lateral), which were still connected but distinguishable from each other (arrows in Fig. 4C). Otx1^{−/ −}inner ears showed a comparable BMP4 expression pattern at 11.5 dpc (arrows in Fig. 4G). However, by 12 dpc, when the anterior and lateral cristae were separate entities in wild-type inner ears (Fig. 4D), BMP4 was expressed as a single domain in both Otx1^{−/ −} (Fig. 4H) and Otx1^{−/ −};Otx2^{+/ −} mutants (data not shown). Tracing this expression pattern in subsequent stages showed that this domain developed into the anterior crista and the lateral crista was not detected in these mutants.

In wild-type mice, by 15 dpc, the utriculosaccular duct is well formed, thus leading to the simultaneous compartmentalization and separation of the utricle and saccule (Figs 2A, 5A). Fig. 5C illustrates Fng gene expression in the supporting cells of the lateral crista, the macula utriculi and the macula sacculi in a wild-type embryo. Both Otx1^{−/ −}and Otx1^{−/ −};Otx2^{+/ −} mutants did not form utriculosaccular ducts (see Fig. 2B-F). As a consequence, the utricle and saccule failed to separate, and thus were housed in a common chamber (Fig. 5B). The macula utriculi was also displaced laterally (compare mu in Fig. 5A,B). In addition, the two sensory organs, macula utriculi and the macula sacculi were partially fused, as highlighted by Fng gene expression (Fig. 5D). The crista shown in Fig. 5D (asterisk) is a part of the anterior crista, which was often found in sections of mutant mice at the level of the utricle.

The histology of most of the Otx1^{−/ −} cochlea were fairly normal and indistinguishable from wild type (n=5). Preliminary results showed that three out of four Otx1^{−/ −} mice tested positive for Preyer reflex indicating that these mice are sound-responsive (data not shown). Because the Otx1^{−/ −};Otx2^{+/ −} mice die at birth, their ability to respond to sound cannot be evaluated. Since the cochlea of Otx1^{−/ −};Otx2^{+/ −} mice were mostly abnormal in shape, we examined the relationship of the spiral ganglion and the cochlea, using neurofilament gene expression as a ganglion marker. Despite the abnormal shape of the cochlea, the spiral ganglion was always present and located underneath the cochlear duct (compare Fig. 6A and B). The sensory regions in the cochlea of Otx1^{−/ −};Otx2^{+/ −} appeared normal, as indicated by the expression of Brn3.1 in the inner and outer hair cells (Fig. 6C), and Fng in the underlying supporting cells (Fig. 6D). Other specialized regions in the cochlea of mutants also appeared normal. The relative location of Hensen’s and Claudius’ cells in relation to the sensory hair cells also appeared normal as indicated by BMP4 gene expression (Fig. 6E). To investigate the early development of the stria vascularis, we examined the expression of tyrosinase related protein 2 (TRP-2). TRP-2 is expressed in all melanocytes and serves as an early marker for the neural crest cells that migrated from the neural tube to form part of the stria vascularis (Steel et al., 1992). Our results show that TRP-2 was expressed in an appropriate location within the cochlea that was destined to form the stria vascularis (Fig. 6F). Taken together, these results suggest that, while the shapes of the cochlea in Otx1^{−/ −};Otx2^{+/ −} mice were aberrant, the relative locations of various sensory and non-sensory components within the cochlea were normal.

The predicted murine OTX1 and OTX2 proteins show extensive homology between the N-terminal and the end of the homeodomain (Simeone et al., 1993). Beyond the homeodomain, OTX1 protein contains blocks of amino acid insertions that are absent in OTX2. The similarity between the murine and human OTX2 proteins is even more extensive, with a single amino acid substitution of a serine to a threonine, out of 289 amino acids. To test if Otx2 can functionally rescue the absence of Otx1, a full- coding region of hOtx2 cDNA was introduced into a disrupted Otx1 locus (Acampora et al., 1999). The inner ear of hOtx2¹/hOtx2¹ mice did not show rescue of the lateral canal and ampulla defect (Table 1; Fig. 7; Acampora et al., 1999). In regions where Otx1 and Otx2 are co-expressed, a spectrum of partial rescues was observed (Table 1, Fig. 3). Fig. 7A illustrates a hOtx2¹/hOtx2¹ inner ear with the most extensive rescue. It consisted of a fairly normal utriculosaccular duct (long arrow in Fig. 7A), cochleosaccular duct (short arrow in Fig. 7A), and the hook of the cochlea (compare to Fig. 2A). In contrast, the inner ear in Fig. 7B showed no rescue in any of these structures. In addition, the cochlea remained shortened, suggesting the inability of hOtx2 to rescue the cochlear phenotype in this case. However, the shape of the saccule was normal in all hOtx2¹/hOtx2¹ inner ears examined (Table 1).

When Otx genes were activated in the developing otocyst, the domain of Otx1 expression was broader than that of Otx2. Only Otx1 was expressed in the presumptive lateral crista and canal area, and these structures were consistently affected in the Otx1 knock-out mice. Otx1 most likely mediates the normal development of these structures in a cell autonomous fashion. Previously, using BMP4 as a presumptive sensory marker for cristae, we proposed that the anterior and lateral cristae originated from a common streak of BMP4-positive area in the anterior portion of the otocyst, which later split to form two separate sensory organs (Morsli et al., 1998). The presumptive lateral crista most likely existed in the Otx1 mutants initially, based on the early pattern of BMP4 expression. However, the lateral crista either remained fused to the anterior crista, or it split from the common streak but failed to continue to develop into a normal dome-shaped crista.

Starting from the mid-level of the presumptive utricle to the ventral tip of the inner ear, the expression of Otx1 overlapped with that of Otx2 at 11.5 dpc. In Otx1 mutants, the sensory organs of the utricle and saccule were partially fused, and the fusion was also more severe in Otx2 heterozygous background (data not shown). Both Otx genes were acting cell non- autonomously to specify the final position of the two maculae within the inner ear since the presumptive maculae were not within the Otx expression domain. The partial fusion of the two maculae in the Otx mutant mice also supports our previous hypothesis that these two sensory organs share a common origin (Morsli et al., 1998).

The cochlear phenotype in the Otx1^{−/ −} mice was variable and was much more consistent in Otx1^{−/ −};Otx2^{+/ −} mice. Otx1 expression was no longer detectable in the inner ear by 14 dpc, which was 2 days earlier than the completion of the cochlear coiling. However, Otx2 expression remained at least until P5, and was expressed in the lateral wall of the saccule and Reissner’s membrane of the cochlea. Based on the gene expression pattern and the severity of the phenotype, Otx2 may play a more significant role than Otx1 in the development of the cochlear region as well as the saccule. Unfortunately, the loss of Otx2 functions alone cannot be evaluated easily since Otx2 mutant mice die around 10 dpc, well before inner ear morphogenesis.

Otx genes are expressed in a similar ventral region of the developing inner ear among zebrafish (Li et al., 1994), frogs (Kablar et al., 1996), chicken (Wu et al., 1998) and mouse (Simeone et al., 1993). However, zebrafish and frog inner ears do not contain a cochlea, and the cochlea in chicken, the basilar papilla, is not a coiled structure. Therefore, although the lack of proper coiling of the cochlea was one of the manifestations in Otx mutant mice, the role of Otx genes in inner ear development may be more general, involving specification of the ventral compartment of the inner ear.

In Otx1^{−/ −} mutants, the smaller size of the telencephalon compared to wild type was attributed to a reduction in cell proliferation (Acampora et al., 1998, 1999). It remains to be tested whether cell proliferation is affected in the inner ears of Otx1^{−/ −} mutants. However, a reduction in growth can easily account for all the phenotypes observed. The presence of primordial structures such as the lateral canal plate and sensory patches which did not develop properly in mutants are consistent with this hypothesis.

It is interesting that the Drosophila otd shares sequence homology with Otx1 and Otx2, and these genes have been shown to substitute functionally for each other (Acampora et al., 1998a,b, 1999; Leuzinger et al., 1998). While most of the defects in brain and other organs of Otx1 mutant mice are rescued by otd and hOtx2, these genes fail to rescue the lateral canal and ampulla phenotypes (Acampora et al., 1998, 1999). Also, only partial rescue was observed in regions of the inner ear where Otx1 and Otx2 are normally co-expressed. These results suggest that Otx1 serves a unique role in the patterning of the inner ear. At this point, we cannot rule out the possibility that the lack of rescue by hOtx2 in Otx1^{−/ −} inner ears is due to insufficient levels of HOTX2 protein in these mice (approx. 35% less than the normal OTX1 levels, Acampora et al., 1999). However, it is noteworthy that when both copies of the Otx2 gene were replaced by human Otx1 cDNA, the specification of anterior neural plate and proper gastrulation were recovered, but the forebrain and midbrain identities failed to be maintained (Acampora et al., 1998b). Taken together, these data suggest that Otx1 and Otx2 have overlapping as well as separate roles during embryogenesis.

The lateral canal and ampulla first appeared in gnathostomes and were absent in agnatha (for review, see Wersall and Bagger-Sjoback, 1974). It would also be interesting to see if the appearance of the lateral canal and ampulla structures in gnathostomes is also correlated with the initiation of Otx1 expression in inner ears.

Recently, a boundary model was proposed for the specification of sensory organs in the inner ear (Fekete, 1996; Kiernan et al., 1997). This model proposes that broad domains of gene expression (selector genes) divide the otocyst into sectors. Depending on the combination of selector genes at a particular position, the location and possibly the identity of a particular sensory organ are specified, with various sensory organs forming at boundaries of interacting genes. In the case of Otx1, its expression was restricted in the posterolateral portion of the otocyst and several sensory organs were located at its boundary. For example, the two maculae and the cochlea were located outside the Otx1 domain but abutting its boundary. However, the lateral crista was located within, but at the outer limits of the Otx1 expression domain. Therefore, the expression pattern of Otx1 and its relationship to the locations of these presumptive sensory organs qualifies Otx1 as a candidate for a selector gene. In support of this role for Otx1, the absence of the Otx1 gene product affected the development of the lateral crista and the normal positioning of the two maculae. However, the identity of these sensory organs including the presumptive lateral crista, did not appear to be affected based on the early expression patterns of BMP4 and Fng in Otx1^{−/ −} mutants (Fig. 4 and data not shown). Furthermore, in wild-type inner ears, these two sensory markers were differentially expressed in the otocyst prior to the onset of Otx1 gene expression (compare Fig. 2 in Morsli et al., 1998 to Fig. 1 presented here). Assuming the differential expression of the two sensory organ markers reflect a committed state, the fate of the sensory organs may be specified independently of Otx1 and, possibly, by other selector genes. Taken together, the phenotype of Otx1^{−/ −} mutants supports the boundary model as it relates to positioning of the sensory organs, and further suggests that the identity and final positioning of a sensory organ are conferred by separate developmental events.

A similar scenario can be applied to the role of Otx2 in inner ear development, except the Otx2 domain of expression was more restricted than that of Otx1. Both Otx genes are bicoid- like transcription factors. So far, there is no evidence that either of these genes directly activates a secreted ligand or other downstream targets within the inner ear that formalize their role as selector genes in this model. However, several genes such as engrailed and wnt are thought to be regulated by Otx in Drosophila head formation and vertebrate brain development (Royet and Finkelstein, 1995, 1996; Ang and Rossant, 1993; Ang et al., 1994; Rhinn et al., 1998; Acampora et al., 1997, 1998b). Therefore, Otx genes could mediate their role as a selector gene through the regulation of engrailed and wnt in the inner ear, and the expression of these genes during inner ear development warrant further examination.

The authors thank Siew-Lan Ang, Ian Jackson, Allen Ryan and XianJie Yang for probes, Drs. Susan Sullivan, Thomas Friedman for critical reading of the manuscript and Ms Mirene Boerner for editing. This work was supported by the Intramural Program of National Institute on Deafness and Other Communication Disorders, EC BIOTECH Programme, Italian Telethon Programme and Italian Association for Cancer Research.