Spalt4 mediates invagination and otic placode gene expression in cranial ectoderm

In vertebrate embryos, ectodermal placodes are discrete regions of thickened epithelium that form transiently on the head ectoderm. They can be subdivided into two broad categories: sensory placodes that contribute to paired sense organs (nose, ear, lens, lateral line) and neurogenic placodes that contribute to cranial sensory ganglia (trigeminal, epibranchial)(Le Douarin et al., 1986; Webb and Noden, 1993). Initially, all placodes arise from a horse-shoe shaped `preplacodal domain' at the border of the rostral neural plate prior to neurulation(Baker and Bronner-Fraser,2001; Brugmann and Moody,2005; Streit,2004). They subsequently segregate into separate domains designated, from rostral to caudal, as olfactory, lens, trigeminal, otic and epibranchial, each with a distinctive gene expression profile. For example,all placodes express characteristic Pax genes, with the most rostral (nose,lens) expressing Pax6, intermediate level (trigeminal) expressing Pax3 and most caudal (ear, epibranchial) expressing Pax2(Baker and Bronner-Fraser,2001). After thickening of the ectoderm, placodal cells either invaginate or ingress to internalize as the first step in their conversion from epithelia to sensory structures.

Although a number of genes have been implicated in specification of placodal identity, little is known about what imbues placodal ectoderm with the ability to internalize, migrate and contribute to sensory organs and/or ganglia and thus distinguishes it from non-placodal ectoderm. At early stages,ectoderm from other axial levels is competent to form particular placodes when heterotopically grafted in place of the endogenous placode. However, this broad potential to respond to placode-inducing signals becomes limited with time (Groves and Bronner-Fraser,2000). One intriguing possibility is that a factor(s), initially expressed throughout the cranial ectoderm is critical for invagination of ectodermal cells in response to placode inducing signals and that this becomes restricted with time to individual placodes.

Here, we show that a zinc finger transcription factor of the spalt gene family (Sweetman and Munsterberg,2006), chick Spalt4 (also known as Sall4),fulfills these criteria in the cranial ectoderm. Although its early expression is uniform throughout the head ectoderm, its localization later becomes restricted to the otic, lens and olfactory placodes. In Drosophila,mutations in members of the spalt gene family cause defects in both migration and cell fate (de Celis et al.,1999; Elstob et al.,2001; Kuhnlein and Schuh,1996; Rusten et al.,2001). In Xenopus, spalt genes have been shown to be involved in brain development (Onai et al., 2004) and limb regeneration(Neff et al., 2005). Here, we show that expression of Spalt4 in non-placodal ectoderm is sufficient to induce invagination or ingression of cranial ectodermal cells. Knockdown of Spalt4 function in the otic placode results in deficient otic vesicle development whereas overexpression causes ear abnormalities. These results suggest that Spalt4 is important for otic vesicle formation, and may be generally important for the invagination/ingression of placodal ectoderm.

We cloned the complete coding region of Spalt4 into pCIG, which drives expression through a chicken actin promoter and CMV enhancer(Megason and McMahon, 2002). It also contains GFP driven from an IRES sequence downstream of the coding sequence, allowing for coexpression of the desired protein and GFP in the same cell. The sequence coding the amino terminus of Spalt4 including the CCHC zinc finger (amino acids 1-98) was cloned into pCS2+NLS in frame with the nuclear localization signal. The fragment with the NLS and truncated Spalt4was then cloned into pCIG. The Sox10 expression construct was described previously (McKeown et al.,2005). Control electroporations used either empty pCIG or empty pMES (in the Sox10 overexpression experiments) plasmids.

The plasmid was injected into embryos containing three to seven somites(stages 8-9), as well as at later stages (stages 12-13) where indicated. The concentration of DNA was between 2 and 2.5 μg/μl. Eggs were windowed and inked using standard procedures. The needle was placed through the vitelline membrane at the caudal end of the embryo nearly parallel to the embryo, and moved rostrally between the vitelline membrane and the embryo until it lay above the midbrain. The DNA was then injected using a glass needle until the embryo was completely covered with the DNA solution, and then withdrawn. The electrodes were constructed in the lab and were made of two pieces of platinum wire (19 gauge) bent at 45° from the horizontal, about 5 mm from the end and 4 mm apart. The positive platinum electrode was inserted through a hole at the edge of the blastoderm and under the embryo. The negative electrode, 4 mm away from the positive electrode, was placed on top of the vitelline membrane above the embryo at the level of the hindbrain and submerged in albumin but without touching the membrane. Three pulses of 8-10 V for 30 mseconds duration were applied 100 mseconds apart. The electrodes were carefully removed and the egg was sealed and incubated at 38°C for up to 72 hours.

Stage 4 embryos were collected on Whatman filter paper rings and turned ventral side up in Ringer's solution. A small slit was made in the area opaca next to the area pellucida. A bead soaked in 50 μg/ml Fgf2 or BSA (bovine serum albumin) was inserted into the slit(Litsiou et al., 2005) and incubated in modified New culture (Chapman et al., 2001) for 5-7 hours and then collected and fixed overnight in 4% paraformaldehyde.

Embryos were collected in Ringer's solution and fixed in 4%paraformaldehyde overnight. Embryos were washed in PBT and embedded in gelatin for histochemical analysis or dehydrated in methanol for in situ hybridization. In situ hybridization was performed as described previously(Wilkinson, 1992). In situ hybridization on sections was performed using embryos fixed in Carony's fixative, embedded in paraffin and sectioned at 10 μm. In situ hybridization was performed using published procedures(Etchevers et al., 2001).

The anti-GFP (Abcam), anti-Pax2 (Zymed), anti-Pax6 (Covance), HuC/D(Molecular Probes) and TUJ1 (Covance) antibodies were obtained commercially. The Spalt4 antibody was generated using a GST fusion construct that included the region encoding amino acids 654-835 of chicken Spalt4. This region lies between two zinc finger regions and has low sequence homology to other spalt genes. The antibody recognizes nuclei in tissues that express Spalt4RNA but does not recognize cells electroporated with Spalt1-expressing constructs. The pan-Dlx antibody was a kind gift from Jhumuku Kohtz, which was made from a construct from Grace Panganiban(Dong et al., 2000). The polyclonal antibody to Dlx3 (Bailey et al.,2006) gives specific nuclear staining in tissues that normally express Dlx3, such as the otic vesicle and the olfactory epithelium. However, it does not stain other CNS structures that express other Dlx genes(e.g. Dlx1, 2, 5 and 6). Primary antibodies were visualized with Alexa Fluor 488-conjugated donkey anti-goat or Alexa Fluor 594-conjugated donkey anti-rabbit secondary antibodies (Molecular Probes). TUNEL analysis was done using the In Situ Cell Death Detector, TMR kit from Roche according to the manufacturer's instructions.

We examined the pattern of Spalt4 expression by in situ hybridization of chicken embryos from the early primitive streak stage (stage 3) to the time of formation of the otic and nasal placodes (E2 and E3). At stage 3, Spalt4 is expressed in the epiblast throughout the embryo but absent from the hypoblast (Fig. 1A). As the node begins to regress (stage 5), there is increased expression in the presumptive neural plate above the notochord and some scattered cells within the notochord as well as continued expression in the ectoderm (Fig. 1B,D,E). Spalt4 expression in stage 5 embryos(Fig. 1A,N) overlaps the expression of Six1 (Fig. 1O) and Eya2 (Fig. 1P) in the preplacode ectoderm and Sox3(Fig. 1Q) in the prospective neural plate. Later, at the open neural plate stage (stage 8), Spalt4expression is found throughout the head ectoderm and the open neural plate but is lost from the closing neural tube and non-head ectoderm(Fig. 1C,F) The otic placode strongly expresses Spalt4 at stage 10(Fig. 1G) but has no detectable signal in the hindbrain. At stages 13 and 15(Fig. 1H,I) Spalt4expression in the ectoderm is strong in the otic pits and continues as they close to form the otic vesicle, but becomes reduced in the otocyst by stage 18(data not shown). At these stages, expression is observed in the neural crest(arrows in Fig. 1H,I) as previously described (Barembaum and Bronner-Fraser, 2004). Spalt4 can also be detected in the lens (Fig. 1J) and weakly in the olfactory epithelium (Fig. 1K). It is also present in limb mesoderm(Fig. 1L), but not the ectoderm or apical ectodermal ridge (Fig. 1M).

Fgf signaling has been implicated in placode formation(Litsiou et al., 2005; Maroon et al., 2002; Martin and Groves, 2006) as well as induction of the otic vesicle(Ladher et al., 2000; Ladher et al., 2005; Liu et al., 2003; Maroon et al., 2002; Vendrell et al., 2000). Furthermore, insertion of Fgf8-coated beads in the area opaca results in the induction of early placode markers such as Dlx5, Sox3 and Eya2 (Litsiou et al.,2005). To analyze the ability of Fgf signaling to induce Spalt4, we implanted an Fgf2-soaked bead(Litsiou et al., 2005) under the area opaca of stage 4 embryos, as this region of ectoderm is competent to respond to placode-inducing signals. Embryos were analyzed for Spalt4expression 5 to 7 hours thereafter. Spalt4 was detected in the ectoderm surrounding the bead in nine out of ten embryos(Fig. 2A,B,C), while none was detected in control embryos (0/8 embryos) in which a BSA-soaked bead was implanted (Fig. 2D-F). In addition to Spalt4, the preplacodal marker Eya2 was induced by Fgf2 (3/5) (data not shown).

To investigate its potential role in placode development, we used electroporation to misexpress a Spalt4 construct that co-expressed GFP in portions of the cranial ectoderm that do not normally contribute to placodes. We always performed parallel electroporations using an empty pCIG vector as controls. As a first step, Spalt4 was targeted to the ectoderm adjacent to the hindbrain of stage 8 chick embryos and up to as late as stage 13. One day after electroporation, we detected GFP-positive cells throughout the ectoderm (data not shown) including cells that will give rise to otic, trigeminal and epibranchial placodes as well as epidermis. As early as 30 hours we were able to detect ectopic vesicles (8/8). By 2 days after electroporation, we noted the formation of multiple small ectopic pits or vesicles outside the endogenous otic forming region(Fig. 3) in most Spalt4-electroporated embryos (123/127; Fig. 3E,G,I) and almost never in those electroporated with the control plasmid (1/49; Fig. 3A,C). The numbers and locations of these vesicles varied from embryo to embryo but 60% of Spalt4-electroporated embryos had five or more ectopic vesicles. The ectopic vesicles were found both rostral and caudal to the endogenous otic vesicle as well as laterally in the ectoderm overlying the branchial arches. Most of these vesicles were small, ranging from five to ten cells in diameter,although a few were as large as 30 cells in diameter. The ectoderm in the ectopic vesicles was generally monolayered, but often appeared thickened compared to the adjacent non-electroporated ectoderm. After 72 hours the ectopic vesicles were sometimes several cell layers thick (data not shown). All of the cells within the ectopic vesicles were GFP-positive. However, other GFP-positive cells remained in the adjacent ectoderm and failed to invaginate and also failed to express otic-specific genes(Fig. 3). Owing to the transient nature of electroporation, we were unable to detect GFP-expressing cells after 96 hours. Ectopic vesicles could be detected in embryos electroporated as late as stage 12 (4/4).

We next tested whether the ectopic vesicles and pits resembled otic vesicles using a number of molecular markers for placodes in general [e.g. the Six-Eya-Dach gene network(Streit, 2004)], early otic markers such as Pax2 (Groves and Bronner-Fraser, 2000), Dlx3(Pera and Kessel, 1999), Nkx5.1 (Herbrand et al.,1998) as well as other transcription factors such as Sox8(Bell et al., 2000), Sox10 (Cheng et al.,2000), Dlx5 (Streit,2002), Tbx1 and Tbx3(Chapman et al., 1996) and signaling molecules such as Bmp4, Notch1 and Lunatic fringe(Adam et al., 1998; Cole et al., 2000) thought to be involved in otic pit/vesicle or later ear development. By using a combination of immunocytochemistry and in situ hybridization, we examined the extent to which these ectopic vesicles mimicked the normal otic vesicle program of gene expression.

Spalt4-induced ectopic vesicles at the level of the hindbrain expressed Sox10 (Fig. 3F), Notch1 (Fig. 3H) and EphA4 (Fig. 3J), though these markers were missing in electroporated ectoderm of control embryos (Sox10 in Fig. 2D and data not shown for Notch1 or EphA4). In situ hybridization performed on tissue sections showed that Spalt4was expressed throughout the ectopic vesicles(Fig. 4A). Similarly, Sox10 (Fig. 4B), Notch1 (Fig. 4C), Lunatic fringe (Fig. 4D), EphA4 (Fig. 4E), Tbx1 (Fig. 4F) and Dlx5 (Fig. 4G) were expressed throughout the ectopic vesicles, though Tbx1 was only detected in ectopic vesicles caudal to the otic vesicle. Interestingly, in the ectopic vesicles, with the exception of Dach2, we were unable to detect genes in the Six-Eya-Dachnetwork (Table 1), including Six1 (Fig. 4H).

In order to provide a marker independent of the IRES-GFP to verify which cells express the Spalt4 transgene, we produced a polyclonal antibody to a unique region of chicken Spalt4 protein (covering amino acid 654-835)allowing us to discriminate between expressing and non-expressing cells at high resolution. We found a good correlation between the expression patterns of Spalt4 and GFP proteins (compare Fig. 4I,J). GFP in the ectopic vesicles also correlated with Pax2(Fig. 4K,L) and Dlx3(Fig. 4M,N) protein expression. By contrast, we were unable to detect the early neuronal marker, NeuroD, 48 hours following electroporation or neuron-specific β-tubulin (TUJ1) 72 post-electroporation (Fig. 4O). Cytokeratin 19 was downregulated in the ectopic vesicles, but not in other parts of the GFP-positive ectoderm, possibly because of indirect effects (data not shown).

The above results show that expression of Spalt4 in naïve ectoderm causes ectopic vesicle formation. We next investigated whether excess Spalt4 would alter development if overexpressed in the endogenous placode region. To this end, we targeted electroporations to the otic placode itself. This resulted in profound morphological malformations in the developing ear. The otic epithelium appeared enlarged and failed to close properly. Frequently, electroporation of Spalt4 into the otic placode resulted in a flat or opened otic pit (29/42) rather than closed otic vesicles as in control embryos (0/31 with abnormal vesicles). The morphological alterations were variable, perhaps due to differences in amount or distribution of construct, but our data cumulatively suggest that excess Spalt4 causes severe patterning defects in the developing ear. Molecular markers that selectively localize in different domains within the developing ear confirmed that cells overexpressing Spalt4 assumed gene expression characteristic of the ear, though the spatial distribution was sometimes altered. After electroporation of Spalt4, Sox10, which is usually spatially restricted to the dorsal-lateral half of the otocyst(Fig. 5A,C,D), was expressed throughout the thickened ectoderm (Fig. 5B,E,F). Sox10 expression was not limited to the cells expressing GFP (compare Fig. 5E with F), suggesting the effects may be non cell-autonomous, perhaps as a result of mispatterning of the otic ectoderm and failure to downregulate endogenous Sox10 in the ventral-medial half as happens in control embryos. Similarly, Pax2, normally expressed medially in control electroporated otocysts, was observed in lateral regions after Spalt4overexpression (Fig. 5G,H). However the expression of Six1, which is not induced by Spalt4 overexpression in the ectopic vesicles(Fig. 4H), remains expressed in a spatially restricted manner in the thickened otic ectoderm(Fig. 5K,L). In rare cases(<5%), multiple smaller vesicles were observed (data not shown). In contrast to electroporation at early stages, those performed at later times(stage 12 and 13) show normal otic vesicle formation (0/5 abnormal).

At the level of the midbrain, we observed some thickened ectoderm and a few ectopic vesicles after electroporation. In addition, Spalt4electroporated ectoderm failed to contribute to the trigeminal ganglia at the same levels as did stage-matched control electroporated trigeminal placode ectoderm. This resulted in the reduction of the placodal contribution to the trigeminal ganglia as evidenced by a reduction in NeuroD expression in whole mounts by in situ hybridization(Fig. 6A-D). This was particularly apparent in the ophthalmic branch, as well as in a split in the maxillomandibular branch (Fig. 6B,D,F). In addition, the Spalt-expressing cells at midbrain level failed to express either Pax2 or Pax6 proteins (data not shown). Further rostrally, GFP-positive cells maintained Pax6 expression but failed to express Prox1 or δ-crystallin, which are characteristic of the lens (data not shown).

To examine the loss-of-function phenotype of chick Spalt4, we designed a truncated construct encoding the first zinc finger of Spalt4 at the amino terminus. A similar truncated construct was previously used in Xenopus embryos to knock-down function of the spalt protein XsalF and shown to function as a dominant negative(Onai et al., 2004). Constructs were introduced into the presumptive otic epithelium by electroporation in a similar manner to that used for the full-length construct.

The majority of embryos examined 2 days after electroporation with the dominant-negative construct (19/22) had significant reduction in the size of one or both otic vesicles after efficient levels of electroporation (as judged by high levels of GFP expression; Fig. 7C,D). By comparison, few embryos (4/25) electroporated with GFP-vector alone had smaller ears and none of these were as small as the experimental ears (Fig. 7A,B). GFP staining revealed uniform distribution of the electroporated construct throughout the ectoderm of both control and experimental embryos, including the miniature otic vesicles (data not shown). The reduced size of the vesicle in dominant-negative embryos was particularly dramatic in the rostrocaudal dimension, sometimes giving it a tightly squeezed look. There was a reduction of 30% (P<0.0001) in the length of the normal otocyst, with dominant-negative electroporated embryos averaging 212±41 μm(±s.e.m.; n=12) in length along the rostrocaudal axis compared with 301±30 μm (n=8) in controls. In some embryos, the vesicle lost the endolymphatic duct. In a few cases, we achieved unilateral electroporation. In these embryos, the electroporated side had a markedly smaller vesicle compared with a normal vesicle on the contralateral side (data not shown). In situ hybridization revealed that Sox10(Fig. 7C,D) retained its normal pattern (compare with Fig. 7A,B). Notch and Lunatic fringe also retained their normal pattern, though Lunatic fringe expression was lost in the smallest vesicles (data not shown).

We next examined whether the changes in vesicle size were due to increased cell death and/or decreased proliferation. To this end, we performed TUNEL staining (Fig. 7E,F,G,H). At 8 hours after electroporation, we observed approximately twice as many TUNEL-positive cells in the ectoderm of dominant-negative electroporated embryos as in control electroporated embryos (9.4±5.4; ±s.e.m., n=8 TUNEL-positive cells per section compared with 5.25±2.4, n=8 in controls, P<0.05). By contrast, no significant alterations in phosphohistone H3 levels (P<0.4) were noted between dominant-negative [n=5; 1.94±0.4 (±s.e.m.) positive cells in the placode per section] and control (n=6; 1.71±0.18 positive cells) embryos. These data suggest that the decrease in otic vesicle size caused by electroporation of truncated Spalt4 may be due to an early increase in cell death but not to changes in cell proliferation.

In addition to the otic level, the truncated construct was electroporated at midbrain-rostral hindbrain level ectoderm, which does not normally express Spalt4. In these embryos, GFP-labeled cells migrated into the trigeminal ganglia in a manner similar to normal embryos (data not shown). This contrasts with embryos electroporated with full-length Spalt4,where GFP-positive cells remained in the ectoderm and contributed few or no cells to the trigeminal ganglion.

Spalt4-induced vesicles have ectopic Sox10 expression. In a recent study in Xenopus, Sox10 overexpression induced some vesicle-like structures in the vicinity of the ear(Taylor and Labonne, 2005). To test if a similar function was present in birds and if Sox10 would phenocopy Spalt4, we electroporated a construct encoding Sox10 (McKeown et al.,2005) into the cranial ectoderm. Similar to Spalt4overexpression, this generated multiple ectopic vesicles that expressed Notch1 (Fig. 8C,D) and EphA4 (Fig. 8E,F)suggesting that Sox10 is epistatic to Spalt4. However, we did detect some weakly Spalt4-positive ectopic vesicles(Fig. 7G,H), suggesting a more complicated gene regulation. One difference between Spalt4 and Sox10 is that the latter generated ectopic vesicles adjacent to the trigeminal ganglia whereas Spalt4 did not (data not shown).

Placodes initially arise from a common preplacodal domain in the early embryo that subsequently becomes subdivided into individual placodes of either sensory (olfactory, lens, otic) or neurogenic (forming cranial sensory ganglia) character (Streit,2004). Members of the Six-Eya-Dach pathway are expressed in a crescent-shaped domain of anterior ectoderm that is postulated to be of general placodal character (Brugmann and Moody, 2005; Streit,2004). Later, individual placodes become distinguishable in the cranial ectoderm by thickening of the epithelium and expression of characteristic genes. For example, Pax6 is expressed in lens and olfactory placodes (Bhattacharyya et al.,2004), Pax3 in trigeminal placodes(Baker et al., 1999) and Pax2 in otic and epibranchial placodes(Baker and Bronner-Fraser,2001; Groves and Bronner-Fraser, 2000). No single gene imbues placodes with a specific identity; instead, it is probable that co-expression of several genes is required for placode formation and differentiation.

In this study, we show that Spalt4 expression overlaps that of Six-Eya-Dach genes in the preplacodal ectoderm, and is detected as early as stage 3 in the chick. Spalt4 subsequently resolves to the presumptive otic, lens and olfactory placode regions by stage 10, concomitant with the time during which non-placodal ectoderm loses competence to form otic placode (Baker et al., 1999; Groves and Bronner-Fraser,2000).

One of the best-studied placodes is the otic placode, which forms the inner ear (Solomon et al., 2004; Streit, 2001). It initiates as a patch of thickened ectoderm on either side of the hindbrain that invaginates to form otic vesicles (otocysts). Subsequently, the otocyst becomes regionalized, giving rise to the complex inner ear including the cochlea, the different parts of the vestibular system and the endolymphatic duct. A number of different cell types originate from this epithelium, including mechanosensory hair cells, support cells and various other specialized cell types. Other cells delaminate from the placode and migrate next to the neural tube to form the acoustic ganglion. Signals from neighboring tissue, such as Fgf (Ladher et al., 2000; Ladher et al., 2005; Liu et al., 2003; Maroon et al., 2002; Vendrell et al., 2000) and Bmp4 (Chang et al., 1999; Gerlach et al., 2000; Merlo et al., 2002), appear to induce the otic placode and/or activate specific patterns of gene expression. We have shown that Spalt4 is also induced in stage 4 ectoderm by Fgf2. Recent results have also implicated Wnt signaling in otic vesicle induction (Ladher et al.,2000; Ohyama et al.,2006).

Electroporation of Spalt4 results in formation of small ectodermal pits near the otic vesicle or even laterally in the branchial arch ectoderm. Adjacent to the hindbrain, Spalt4 induced ectopic vesicles that morphologically resemble otic vesicles and express otic markers. It is interesting to note that not all Spalt4-electroporated ectodermal cells invaginate. In almost all embryos, some remain in the ectoderm adjacent to the pits. Perhaps this reflects some intrinsic limits to ectopic vesicle size, density of electroporated cells, or requirement for additional signals in some cell populations. These results suggest that Spalt4 alone is not sufficient to induce invagination in all ectoderm.

Ectopic vesicles express a number of genes characteristic of the otic vesicle and important for normal ear development; these include Notch(Adam et al., 1998), Lunatic fringe (Cole et al.,2000), Bmp4 (Cole et al., 2000), Dlx3(Pera et al., 1999), Dlx5 (Streit, 2002), Sox8 (Bell et al.,2000), Sox10 (Cheng et al., 2000), Tbx1(Chapman et al., 1996), Tbx3 (Chapman et al.,1996) and Nkx5.1(Herbrand et al., 1998).

Spalt4 is sufficient to recapitulate some but not all of the molecular events necessary for normal ear development. For example, we were unable to detect the expression of many genes in the Six-Dach-Eyapathway. One possibility is that the Six-Eya-Dach pathway is upstream of Spalt4. Another possibility is that regulation of this pathway may involve a gene network independent and perhaps parallel to that induced by Spalt4. Six1 and Eya1 mutants have been shown to have poorly developed auditory systems (Li et al.,2003; Ozaki et al.,2004; Xu et al.,1999; Zheng et al.,2003). Though the ears progress to the otic vesicle stage, they fail to form middle and inner ear structures or neurons, as if stalled at the vesicle stage rather than differentiating further(Zheng et al., 2003). Similarly, we were unable to detect neurons in the ectopic vesicles 72 hours after Spalt4 electroporation. An intriguing possible reason why Spalt4-induced vesicles do not progress beyond the otic vesicle stage and fail to generate neurons is because of their failure to upregulate Six1 and Eya1. Alternatively, the Six-Eya-Dachgenes may require region-specific signals that are absent at the sites where ectopic vesicles form. The latter possibility seems likely in the case of Dach2, since it was expressed only in ectopic vesicles next to the hindbrain. Tbx1 and Tbx3 are also influenced by other factors since they were expressed only in ectopic vesicles caudal to the endogenous otic vesicle.

Recent studies have highlighted interesting similarities between the vertebrate inner ear and Johnston's organ in Drosophila(Boekhoff-Falk, 2005). Many of the genes necessary for specification or function of the auditory cells in Drosophila are also required in the vertebrate inner ear. In Drosophila spalt and spalt-related are required for the formation of Johnston's organ (Dong et al., 2003). However, the roles of spalt genes in vertebrate and fly auditory development may not be completely analogous. In Drosophila,spalt has been shown to be downstream of distalless(Dong et al., 2002). By contrast, we find that chick Dlx genes are upregulated by misexpression of chick Spalt4. Furthermore, Dlx3 or Dlx5 overexpression fails to induce Spalt4 expression (data not shown). The Iroquois homologues Irx1 and Irx2 are upregulated by Spalt4 (Table 1), though they are repressed by spalt in Drosophila wing development (de Celis and Barrio, 2000).

Ectopic Fgf2, Fgf3 or Fgf8 induce ectopic vesicles that express otic markers such as Notch1, Pax2 and Nkx5.1(Adamska et al., 2000; Vendrell et al., 2000). In addition to Fgf, other signals may be involved in regulation of Spalt4 and/or other placodal determinants. Vitamin A-deficient chick embryos lack posterior hindbrain, but develop ectopic Pax2-positive vesicles (Kil et al., 2005). Overexpression in Xenopus of the secreted phospholipase Rossy induces ectopic olfactory vesicles, and, by microarray analysis, has been shown to upregulate a member of the spalt family, Xspalt1(Munoz-Sanjuan and Brivanlou,2005). Overexpression of other transcription factors has been show to result in ectopic vesicles in other species. Pax6 overexpression generates ectopic lens vesicles (Altmann et al., 1997) and Sox10 can generate ectopic otic vesicles(Taylor and Labonne, 2005). Ectopic Six3 expression in mice leads to ectopic vesicle formation near the otocyst (Lagutin et al.,2001), and injection of Sox3(Koster et al., 2000) in medaka gives rise to vesicles of either otic or lens character. However, we failed to detect upregulation of either Six3 or Sox3 after misexpression of Spalt4; this could reflect species differences or a lack of epistatic interactions between these transcription factors. It is currently unclear if any of these factors directly regulates Spalt4in the chick. Interestingly, constitutively active Notch has been shown to cause the formation of ectopic structures expressing ear-specific genes,consistent with the possibility that Spalt4 is upstream of Notch in this cascade (Daudet and Lewis, 2005). We have found that ectopic Sox10 expression can generate ectopic vesicles in the chick, similar to results previously described in Xenopus (Taylor and Labonne, 2005). These ectopic vesicles also express Spalt4. The exact relationship between these two genes has yet to be determined, but may involve a feedback loop. One difference is that Spalt4 was unable to generate ectopic vesicles expressing otic-specific genes at the level of the trigeminal ganglia, whereas Sox10 was able to do so. This indicates that Spalt4 and Sox10 respond differently to the signals in the ectoderm at midbrain level and implies that Sox10 may act downstream of Spalt4.

Overexpression of Spalt4 within the otic vesicle itself causes alterations in morphology and patterns of gene expression in the developing ear. Defects include formation of multiple vesicles resembling otic vesicles,and the failure to form a closed otic vesicle. In both cases, Pax2,Lunatic fringe and Notch1 are expressed in a non-regionalized fashion. In the most extreme cases where the otic vesicle fails to close, Sox10 is expressed throughout the otic ectoderm rather than being confined to the lateral half as in control electroporated embryos. Normally,the expression of Spalt4 in the closed vesicle begins to be downregulated at stage 16. However, in electroporated embryos, Spalt4expression is maintained. This may lead to altered expression of other genes,the loss of regionalizing signals and the observed abnormalities in the otic vesicle. Interestingly, activation of canonical Wnt signaling as well as Fgf overexpression in the ear also leads to formation of open, oversized ears(Ladher et al., 2000; Ohyama et al., 2006; Vendrell et al., 2000).

Expression of Spalt4 at the midbrain level interfered with the normal ingression of placode cells into the trigeminal ganglia. Few GFP cells contributed to the ganglia and the number of placode-derived cells was also reduced, resulting in a malformed ganglia. A similar effect was seen in the failure of Spalt4-expressing neural crest cells to contribute to the trigeminal ganglia (Barembaum and Bronner-Fraser, 2004). Furthermore, we were unable to detect Pax3 in the GFP-expressing thickened ectoderm (data not shown). At epibranchial placode levels, Spalt4-electroporated cells also formed ectopic vesicles and failed to contribute to the ganglia derived from the epibranchial placodes after 48 hours. Since ectopic vesicles were found in ectoderm that would normally give rise to neurons, this probably reflects a change in cell fate from neurogenic to sensory. The cell fate switch is reminiscent of the activity of spalt in Drosophila where it affects fate determination in a number of different lineages(de Celis et al., 1996; Elstob et al., 2001; Rusten et al., 2001).

Our results show that Spalt4 is not only sufficient for vesicle formation but also necessary for proper otic development. Introduction of a truncated, dominant-negative Spalt4 results in abnormal otic vesicles that are drastically reduced in size. In general, otic gene expression remains the same and vesicles retain a regionalized pattern. The reduction in vesicle size appears to be caused by increased cell death, as assayed by TUNEL. That otic vesicles do form in the presence of the dominant-negative Spalt4, albeit reduced in size, may indicate that dominant-negative Spalt4 may not fully abrogate endogenous Spalt4 activity. Spalt4 is normally expressed well before the time that we introduce the dominant-negative construct. Thus, it is likely that we do not achieve full knockdown of transcription factor activity. Also we cannot rule out the possibility that other genes may be acting on the ectoderm to partially compensate for the loss of Spalt4. It is also worth noting that whereas Sox10 can be induced in ectopic vesicles, it is not reduced in otic vesicles electroporated with the dominant-negative construct. A possible explanation is that Sox10 may not be directly induced by Spalt4.

Consistent with our observations in chick, humans with Okihiro syndrome, in which SALL4 is mutated, have hearing defects as well as abnormalities of the heart, kidney and limbs (Kohlhase et al., 2005). By contrast, no hearing defects have been reported in heterozygous mutant mice (Sakaki-Yumoto et al., 2006), though some hearing defects have been detected in mice with a truncated Spalt4(Warren et al., 2007).

The finding that chick Spalt4 is expressed earlier than otic placode markers such as Pax2, Dlx3 and Dlx5 in the otic placode raises the intriguing possibility that Spalt4 may have a role in establishing the placode domain. Spalt4 alone, however, is not sufficient for normal otic vesicle formation since it fails to form ectopic vesicles of normal size that express all ear markers. It is more likely that Spalt4 is an important component in the multiple steps leading to the formation of the ear. Our functional analysis suggests that it may initiate the process of invagination in the early ectoderm in response to region-specific signals along the rostrocaudal axis and to upregulate appropriate gene expression, such as ear-specific genes, in the vesicles in the hindbrain region.

We thank Drs Tatjana Sauka-Spengler, Vivian Lee and Sujata Bhattacharyya for helpful comments on this manuscript. This work is supported by USPHS grant DE16459 to M.B.F.