Epibranchial and otic placodes are induced by a common Fgf signal, but their subsequent development is independent

Developmental Biology 303 (2007) 675 – 686 www.elsevier.com/locate/ydbio

Epibranchial and otic placodes are induced by a common Fgf signal, but their subsequent development is independent Shun-Kuo Sun 1 , Chris T. Dee 1 , Vineeta B. Tripathi, Andrea Rengifo, Caroline S. Hirst, Paul J. Scotting ⁎ Children’s Brain Tumour Research Centre, Institute of Genetics, Queen’s Medical Centre, University of Nottingham, Nottingham, NG7 2UH, UK Received for publication 10 May 2006; revised 30 November 2006; accepted 5 December 2006 Available online 9 December 2006

Abstract The epibranchial placodes are cranial, ectodermal thickenings that give rise to sensory neurons of the peripheral nervous system. Despite their importance in the developing animal, the signals responsible for their induction remain unknown. Using the placodal marker, sox3, we have shown that the same Fgf signaling required for otic vesicle development is required for the development of the epibranchial placodes. Loss of both Fgf3 and Fgf8 is sufficient to block placode development. We further show that epibranchial sox3 expression is unaffected in mutants in which no otic placode forms, where dlx3b and dlx4b are knocked down, or deleted along with sox9a. However, the forkhead factor, Foxi1, is required for both otic and epibranchial placode development. Thus, both the otic and epibranchial placodes form in a common region of ectoderm under the influence of Fgfs, but these two structures subsequently develop independently. Although previous studies have investigated the signals that trigger neurogenesis from the epibranchial placodes, this represents the first demonstration of the signaling events that underlie the formation of the placodes themselves, and therefore, the process that determines which ectodermal cells will adopt a neural fate. © 2006 Elsevier Inc. All rights reserved. Keywords: Sox3; Zebrafish; Epibranchial placodes; Fgf signaling; foxi1

Introduction The mechanisms underlying the decision of ectoderm to adopt a neural fate have been a central area of developmental studies for many decades. While research into the origins of the CNS have been abundant, their success has been restricted by the early stage of development when these events occur, and their association with other fundamental processes to which neural induction is linked, such as gastrulation and the specification of the mesendoderm. The neurogenic placodes, on the other hand, appear later after such basic aspects of the embryonic body plan are established, making them a useful alternative system in which to study the mechanisms controlling neural fate. These placodes include the nasal, otic and also the

⁎ Corresponding author. Fax: +44 115 8230350. E-mail address: [email protected] (P.J. Scotting). 1 These authors contributed equally to this paper. 0012-1606/$ - see front matter © 2006 Elsevier Inc. All rights reserved. doi:10.1016/j.ydbio.2006.12.008

epibranchial placodes, which give rise to many of the neurons associated with the sensory ganglia. Placodes are, by definition, patches of thickened ectoderm (Webb and Noden, 1993). They arise outside of the axial ectoderm from which the CNS and neural crest are formed. As such they are the only regions outside of the CNS and the neural crest (derived from the margins of the CNS) that give rise to cells of the nervous system. Understanding the mechanisms by which placodes arise is therefore an important issue in developmental biology. Although identified many years ago, a general lack of studies has meant that understanding of the basic processes underlying the formation and development of the epibranchial placodes has lagged behind the CNS and other placodal structures. Classical embryological studies carried out in the latter part of the 20th century led to a generally accepted picture of the inductive processes underlying the formation of these structures. These studies appeared to demonstrate that formation of all placodes except the epibranchial placodes involves signals

676

S.-K. Sun et al. / Developmental Biology 303 (2007) 675–686

from the CNS, although a role for hindbrain derived neural crest cells in later migration of neurons from the epibranchial placodes has been shown (Baker and Bronner-Fraser, 2001; Begbie and Graham, 2001; Graham and Begbie, 2000; Yntema, 1944). Formation of the epibranchial placodes therefore required an alternative source of inductive signals to be invoked. To date the exact nature of these signals and inducing tissues remains unclear, with the mesoderm or endoderm as the most likely candidate source. Endoderm and endoderm-derived Fgfs and Bmps have been shown to be required for neurogenesis in the epibranchial placodes (Begbie et al., 1999; Holzschuh et al., 2005; Nechiporuk et al., 2005). However, despite showing that these factors could induce the formation of neurons from the ectoderm, they did not determine which signals actually induced the early placodal ectoderm itself. Our recent studies in chick (Abu-Elmagd et al., 2001; Ishii et al., 2001), and the work of others in Xenopus laevis (Penzel et al., 1997), have identified the transcription factor, Sox3, as one of the earliest markers of placode development. In chick, Sox3 is expressed in a broad domain near to the developing ear and only later does this domain become segmented to give rise to the final, four epibranchial placodes (Ishii et al., 2001). Here we show the same pattern of sox3 expression in the epibranchial placodes of the zebrafish, where it represents a unique marker for the early events in placode induction. This has allowed us to re-examine the signals underlying epibranchial placode formation. In particular, based on the location of the early placodal domain adjacent to the developing otic placodes, and data implicating hindbrain derived Fgfs in otic development (Alsina et al., 2004; Ladher et al., 2000; Leger and Brand, 2002; Liu et al., 2003; Maroon et al., 2002; Phillips et al., 2001; Vendrell et al., 2000; Wright and Mansour, 2003), we have shown that the same signals, Fgf3 and Fgf8, are required for development of the epibranchial placodes. Although these data suggest that Fgfs normally act by maintaining sox3 expression from within a larger earlier domain, we show that Fgfs are capable of re-inducing sox3 expression if signaling restarts after an initial block. In addition, we demonstrate that the gene foxi1, which is expressed in both the otic and epibranchial ectoderm and is required for formation of epibranchial placode-derived neurons (Lee et al., 2003), is required for sox3 expression. Our experiments further show that the epibranchial placode domain was unaffected in mutants in which no otic placode forms or when several genes associated with otic development are inhibited. Thus, the initial induction of both otic and epibranchial placodes share common signals, but the subsequent development of these two structures is independent. Methods Maintenance of fish Breeding zebrafish were maintained and embryos were raised (Westerfield, 2000) and staged according to Kimmel et al. (1995). The fgf8/ace mutant has been previously described (Reifers et al., 1998). Homozygous B380 mutant embryos were a kind gift from Monty Westerfield (Liu et al., 2003).

Morpholino injections Morpholino antisense oligonucleotides (Gene Tools, LLC, Corvallis) and controls were as previously described: fgf3 (Phillips et al., 2001), fgf8 (Araki and Brand, 2001), dlx3b and dlx4b (Solomon and Fritz, 2002), and foxi1 (Solomon et al., 2003). Embryos were injected at the 1–4 cell stage at concentrations of 1–8 ng/embryo. Whole-mount in situ hybridisation Whole-mount in situ hybridization on zebrafish embryos was carried out as previously described (Jowett, 2001). For double in situ hybridizations, riboprobes were synthesized either with digoxigenin (DIG) or fluoresceinlabelled nucleotides (Roche). Detection of DIG/fluorescein antibody–alkaline phosphatase conjugate was performed using BM-purple (Roche) for dark blue/ purple stain, Fast Red (Sigma-Aldrich) for red stain, or BCIP (Roche) for green/ pale blue stain. The following probes were derived from previously described cDNA clones: pax2a (Krauss et al., 1991), pax8 (Pfeffer et al., 1998), dlx3b (Ekker et al., 1992) krox20 (Yi-Chuan Cheng, Taipei). Other clones were obtained as ESTs from RZPD: sox3 (GenBank Accession Number: AI959362), foxi1 (GenBank Accession Number: CF997841) neurogenin1 (GenBank Accession Number: CA496091), neuroD (GenBank Accession Number: CD757273). After in situ hybridization, embryos were re-fixed in 4% paraformaldehyde, transferred into 80% glycerol and photographed. Embryos that were to be sectioned were rehydrated, transferred into ethanol and embedded in JB4 methacrylate (Agar Scientific, UK) for microtome (Leica RM2265) sectioning. SU5402 and retinoic acid treatments SU5402 (Calbiochem; in DMSO to give a 2 mg/ml stock solution) was used at a final concentration of 20–50 μM in fish water containing methyl blue. Retinoic acid (Sigma; in DMSO to give a stock concentration of 10 mM) was diluted to a working concentration of 10 μM in fish water containing methyl blue. For both treatments, embryos were treated in their chorions at 28 °C, in the absence of light, for specific time periods as required.

Results Expression of sox3 defines the developing, epibranchial placodes In zebrafish, as has been described in chick (Rex et al., 1997b) and mouse (Wood and Episkopou, 1999), sox3 is initially expressed throughout the epiblast (Okuda et al., 2006). As embryos undergo neural induction, its expression becomes restricted to the prospective CNS (Kudoh et al., 2004; Penzel et al., 1997; Rex et al., 1997a and Fig. 1A). At about the time that the first somites are formed, a small domain of sox3 expression, flanking either side of the prospective hindbrain, also becomes apparent (arrow, Fig. 1B). This domain becomes more pronounced with time (Figs. 1C, D). Initially the domain is symmetrical, but it rapidly takes on an arced shape, with a sox3-negative region most medially, at the same rostrocaudal level as the prospective otic placode (Fig. 1D). As we have shown previously in chick (AbuElmagd et al., 2001; Ishii et al., 2001), this domain of sox3 expression is subsequently divided into a series of patches (Figs. 1E, F, K, L), which are coincident with later neurogenesis as shown by expression of the neurogenic genes, neurogenin1 (Figs. 1G, H, M, N) and neuroD (Figs. 1I, J). Expression rostral to the otic vesicle is lost by 60 hpf, and

S.-K. Sun et al. / Developmental Biology 303 (2007) 675–686

likewise disappears in the region caudal to the ear between 72 and 80 hpf (data not shown). sox3 expression therefore represents a unique marker for the early stages of epibranchial placode development. In zebrafish, the pax genes, pax2a and pax8, have been studied during the development of the adjacent otic placode. These are amongst the earliest definitive markers of the prospective otic placode (Leger and Brand, 2002; Liu et al., 2003; Mackereth et al., 2005; Phillips et al., 2001; Solomon and Fritz, 2002). We therefore compared sox3 expression with that of pax2a and pax8 when they appear at early somite stages (Figs. 2 and 5A). This showed that the patches of sox3 and pax8 adjacent to the hindbrain appear at about the same time (9– 11 hpf) and occupy the same position (Figs. 2D, G). However,

677

by 12 hpf, soon after pax2a appears (Fig. 2B), the domain of ectoderm now expressing both pax2a and pax8 (Figs. 2B, C, E, F) does not express sox3, which now occupies a more lateral position around the prospective otic placodes (Figs. 2H–J). These observations suggest that the otic and epibranchial placodes arise from a common region expressing both Sox3 and Pax8, which becomes subdivided when pax2a expression appears and sox3 expression becomes restricted to the future prospective epibranchial placode region. sox3 expression is maintained by Fgf signals Fgf signaling has been shown to be a positive inducing force in otic development in a variety of species (Alsina et al., 2004; Ladher et al., 2000; Leger and Brand, 2002; Liu et al., 2003; Maroon et al., 2002; Phillips et al., 2001; Vendrell et al., 2000; Wright and Mansour, 2003). These studies propose that in zebrafish, hindbrain-derived Fgf3 and Fgf8 dictate the localized appearance of several otic markers, the transcription factors pax2a, pax8 and sox9a, and the upregulation of dlx3b in the otic region (Leger and Brand, 2002; Liu et al., 2003; Maroon et al., 2002; Phillips et al., 2001). Previous studies in chick have implicated the endoderm, and have been interpreted as excluding the hindbrain, as a source of inducing signals for the epibranchial placodes (Baker and Bronner-Fraser, 2001; Begbie et al., 1999; Yntema, 1944). However, on closer inspection, it is clear that these studies failed to rule out a role for hindbrain-derived signals in induction of the epibranchial placodes. Both studies actually set out to test the role of the neural crest, rather than the CNS as a whole, in placode development. Yntema (1944) only removed the crest and dorsal neural tube and Begbie et al. (1999) appear to have removed the entire hindbrain, but this was done at HH stage 9 when the expression of sox3 is already well established in the

Fig. 1. Expression of sox3 marks the developing epibranchial placodes. (A, B, G, I, K) Embryos viewed laterally with dorsal to right (A, B) or dorsal uppermost (G, I, K). (C–E, M, N) Embryos viewed dorsally with rostral uppermost (C–E) or to the left (M, N). In situ hybridization for sox3 expression shows its appearance in a small bilateral patch (arrow) of ectoderm (see insert in panel B showing t.s. section) either side of the hindbrain between 9 and 10 hpf (sox3 is also strongly expressed throughout the length of the CNS) (A, B). By 11 hpf (2– 3 somites) these patches have formed parallel to the hindbrain adjacent to, but a short distance lateral to rhombomeres 3–5 (C). At 12 hpf (5–6 somites), the patches take on an arced shape with a pronounced sox3-negative domain nearest the CNS (D). At later stages the single domain appears to extend in the rostrocaudal axis and break into a series of smaller patches (E) that are again clearly seen to be ectodermal in sections (more medial expression of sox3 also seen in the endoderm of the pharyngeal pouches) (F). Comparison of single in situ hybridization for ngn1 (G, H), neuroD (I, J) and sox3 (K, L) at 24 hpf shows that these regions the regions of neurogenesis are closely associated with the ectodermal patches of sox3 expression. This is most evident from sections taken at equivalent positions just rostral to the otic vesicle (H, J, L). Double in situ hybridization with the early marker of neurogenesis, ngn1 (red) at 24 hpf (M) and 36 hpf (N), verified that these sox3 domains (blue) are indeed the regions from which neurons later arise (M,N) as shown in the transverse section, inset in (M). (M⁎) and (N⁎) represent enlarged views of boxed region in (M) and (N) respectively. ad, anterodorsal lateral line; av, anteroventral lateral line; f, facial; m/v, middle lateral line/vagal; o/g, octaval/statoacoustic/glossopharyngeal; ov, otic vesicle.

678

S.-K. Sun et al. / Developmental Biology 303 (2007) 675–686

Fig. 2. Relationship between the expression of sox3 and the otic markers pax2a and pax8. All panels viewed dorsally at the level of the hindbrain with rostral to the top. In situ hybridization for pax2a (A–C) or pax8 (D–F) expression. (G–J) Double in situ hybridization showing pax8 expression (blue) and sox3 expression (red). pax8 is expressed in the prospective otic placodes (arrowheads) first appearing as discrete domains between 9 and 11 hpf (D) when pax2a is barely detectable in the prospective otic region (arrowhead in A). Double in situ hybridization shows that the initial domain of pax8 expression (arrowhead) is largely coincident with the first expression of sox3 (arrow) in this region. By 12 hpf, however, when pax2a is strongly expressed (B, C), the two domains are almost mutually exclusive with pax8 (and pax2a, see also Fig. 5A) occupying a more medial domain and sox3 an arc of expression surrounding the pax8 expressing ectoderm (G–I). (I) Enlarged view of boxed domain in panel G.

ectoderm adjacent to the hindbrain (Abu-Elmagd et al., 2001; Rex et al., 1997b) and signals may already have been produced that could have established the initial placodal domains. This stage is morphologically equivalent to 12–13 h of development in zebrafish (based on somite number and krox20 expression in the hindbrain; Nieto et al., 1995), when the placodal expression of sox3 is also well established in zebrafish (see Fig. 1D). Recent experiments in zebrafish, in which endoderm development was disrupted also suggest that an additional source of inducing signals must exist (Holzschuh et al., 2005; Nechiporuk et al., 2005). Since the domain of sox3 expression in the prospective epibranchial placodes occupies a region immediately abutting the otic domain, we carried out experiments to determine whether this sox3 expression was also dependent on the same Fgf signaling that drives otic placode formation. Phillips et al. (2001) have shown that treatment of embryos with retinoic acid (RA) resulted in an expansion of the otic marker pax8 and the formation of ectopic otic vesicles, presumed to be due to posteriorization of the embryonic head and consequent expansion of the hindbrain (Phillips et al., 2001). This effect of RA was blocked by SU5402, an inhibitor of Fgf signaling. We used this same approach to determine whether expansion of the hindbrain caused a similar expansion of the sox3-expressing placodal domain. Embryos were treated

with RA from 6–7 hpf and subjected to in situ hybridization at 11 hpf when the single large bilateral domains of sox3 expression were most apparent in control embryos (Fig. 3A). Such RA treatment not only expanded the hindbrain, as shown by displacement of the rostral band of krox20 expression to the very rostral tip of the CNS (Fig. 3C), and the domain of otic markers such as pax8 as previously published (Phillips et al., 2001), but also resulted in a similar expansion of sox3 expression (Fig. 3B) as compared to DMSO treated controls (Fig. 3A). As with pax8, this effect was blocked by co-treatment with the Fgf signaling inhibitor, SU5402 (Fig. 3E). In support of the role of Fgf signaling in this expansion of sox3 expression, we found that Fgf expression (Fgf3 and 8) was seen throughout most of the rostral part of the CNS in the RA treated embryos (Fig. 3D) as was also reported by Phillips et al. (2001). Thus, the data support a model in which RA expands the hindbrain and as a consequence expands the region expressing Fgfs and so expands the expression of Fgf-dependent genes, such as sox3. However, although we have also shown that the expansion of sox3 expression in these experiments is indeed Fgf-dependent (the effect was lost when embryos were cotreated with SU5402) this does not rule out a role for other factors derived from the hindbrain or other tissues. In order to gain more insight into the role of Fgf signaling in determining the domain of placodal sox3 expression, we set out

S.-K. Sun et al. / Developmental Biology 303 (2007) 675–686

679

Fig. 3. sox3 expression is affected by disruption of Fgf signaling. In situ hybridization for sox3 expression at 11 hpf in DMSO treated control embryos (A) or following treatments to alter Fgf signaling. Treatment with retinoic acid (RA) from 6–7 hpf causes expansion of the hindbrain, as shown by rostral displacement of the rhombomere 3 band of krox20 expression (C) and consequently expression of fgf3/8 throughout most of the rostral CNS (D). This treatment results in a striking expansion of sox3 expression, in the most severe cases resulting in a complete arc of expression around the anterior of the embryo (B). Expression of sox3 outside of the CNS, even in the presence of RA, is completely blocked when Fgf signaling is inhibited using the small molecule inhibitor, SU5402 (E–H). SU5402 causes a loss of placodal sox3 expression when treated from 4–11 hpf (F), but also after shorter treatments at 4–6 hpf (G) and 6–8 hpf (H).

to determine when Fgf signaling was required. Addition of SU5402 to the water surrounding the developing zebrafish embryos is sufficient to block Fgf signaling, allowing us to block Fgf signaling over a range of stages of development. In each experiment the effect upon the single bilateral domains of sox3 expression at 11 hpf were assayed. Treatment of embryos with SU5402 from 4–11 hpf completely abolished the placodal patches of sox3 expression (Fig. 3F). Thus, the signals promoting sox3 expression appear very similar to those dictating the appearance of otic markers (Alsina et al., 2004; Ladher et al., 2000; Leger and Brand, 2002; Liu et al., 2003; Maroon et al., 2002; Phillips et al., 2001; Vendrell et al., 2000; Wright and Mansour, 2003). This same effect could be achieved by blocking Fgf signaling for shorter periods prior to 11 hpf (4–6 and 6– 8 hpf) suggesting that there was not a single critical period when Fgf signaling was required (Figs. 3G, H). However, it is not clear from these data whether Fgf signaling is required only to initiate sox3 expression in the prospective epibranchial placodes or is a continuous requirement for that expression. Treatment of embryos between 4 and 6 hpf resulted in a delay in the appearance of sox3 transcripts until 12 hpf (Fig. 4B) (1 h delayed as compared to controls, Fig. 4A). Similarly, treatment from 6–8 hpf or 7–9 hpf delayed expression until 13–14 hpf (Figs. 4C, D). Thus, both treatments inhibited sox3 expression, but expression recovered 4–6 h after the Fgf inhibitor was removed. In order to determine if this requirement was continuous and to gain clearer insight into the dynamics of the Fgf-blocking regime, we treated embryos at a later time point after sox3 expression was initiated. When inhibitor was present from 11–13 hpf, sox3 expression appeared normally at 11 hpf, was lost from 14 hpf but reappeared at 16 hpf (Fig. 4E). Together, these results indicate that there was a 2–3 h delay

between the addition of SU5402 and loss of sox3 expression. This is presumed to reflect the time taken for the full inhibition of Fgf signaling and decay of sox3 transcripts already present. Likewise, there was always a period of 3–6 h after removal of SU5402 before sox3 expression reappeared, presumed to reflect time to fully wash the blocking agent out, reactivate Fgf signaling and consequently reactivate sox3 expression. Overall, these data show that there is a continuous requirement for Fgf signaling to maintain sox3 expression, but when Fgf signaling reappears after a period of blocking, sox3 expression is induced de novo. sox3 expression is dependent on Fgf3 and Fgf8 Blocking expression of both fgf3 and fgf8 is sufficient to cause loss of several otic placode markers (Leger and Brand, 2002; Liu et al., 2003; Maroon et al., 2002; Phillips et al., 2001). We have used the fgf8 mutant, ace (Reifers et al., 1998), and morpholinos against both fgf8 and fgf3 to determine whether these same Fgfs are necessary for sox3 expression in the prospective epibranchial placodes. Loss of Fgf8 (either in the ace homozygous mutants (Figs. 5B, D) or following injection with a single morpholino, Figs. 5H, K) or Fgf3 (following injection with a single morpholino, Figs. 5E, G, J) caused little reduction in expression of either sox3 (Figs. 5E, J, K), or the otic marker, pax2a (Figs. 5E, G, H). However, absence of both Fgfs (again using either the ace mutant (Fig. 5F) or fgf8 morpholino (Figs. 5I, L)) led to a complete loss of placodal sox3 expression (Figs. 5F, L) and absence of pax2a expression in the otic region (Figs. 5F, I). These effects were restricted to the placodal domains with CNS expression of sox3 and pax2a remaining intact. Thus, like the otic markers, either Fgf3 or Fgf8

680

S.-K. Sun et al. / Developmental Biology 303 (2007) 675–686

Fig. 4. Fgf signaling is continuously required for maintenance of placodal sox3 expression. Embryos were treated with SU5402 to inhibit Fgf signaling for various periods of embryonic development, fixed at a range of time points and sox3 expression analysed by in situ hybridization. (A) sox3 expression in DMSO treated controls. (B) Inhibition of Fgf signaling from 4–6 hpf. Placodal expression of sox3 is delayed, only appearing at 12 hpf after which time expression is similar to wild type. Inhibition of Fgf signaling from 6–8 hpf (C) or 7–9 hpf (D) further delayed placodal expression of sox3, appearing at 13 hpf after which time expression is similar to wild type. (E) Inhibition of Fgf signaling after placodal sox3 expression was initiated, from 11–13 hpf. Placodal expression of sox3 is lost between 14 and 15 hpf, after which time it reappears in a pattern similar to wild-type.

is required for the placodal expression of sox3. Since the hindbrain is the only area where these two Fgfs are coexpressed in zebrafish at this stage (Phillips et al., 2004), this strongly implicates the hindbrain as an epibranchial placode-inducing tissue, although low level expression of these Fgfs elsewhere cannot be excluded. The epibranchial placodal domains of sox3 expression develop independently of the otic placodes The data above show that the placodal domain of sox3 expression not only develops in the same region as otic markers, but under the same influence of Fgf signaling. We therefore set out to determine whether there was a close relationship with the expression of other otic markers. In particular, the expression of pax2a and pax8 in the otic domain has been shown to be Fgfdependent (Leger and Brand, 2002; Liu et al., 2003; Maroon et al., 2002; Phillips et al., 2001), and also requires the earlier expression of dlx3b (for pax2a expression) and foxi1 (for both pax8 and pax2a exression) (Hans et al., 2004; Liu et al., 2003). Comparison between expression of sox3 and dlx3b by double in situ hybridization, demonstrated that the sox3 expression does indeed appear within the larger preplacodal

domain of dlx3b expression (Fig. 6A). We therefore tested whether expression of dlx3b and/or dlx4b, which is expressed in a similar domain (Solomon and Fritz, 2002), was required for the expression of sox3 adjacent to the hindbrain. dlx3b and dlx4b morpholinos (Solomon and Fritz, 2002) were injected separately or in combination, and sox3 expression was analysed at the 3 somite stage. As a positive control, we also analysed pax2a expression, which has been shown to be dependent upon these two Dlx factors (Liu et al., 2003). Knock-down of either dlx gene reduced pax2a (not shown). However, even when these genes were both knocked down, we saw no effect upon sox3 expression even when pax2a expression was completely lost (Fig. 6C) as compared to controls (Fig. 6B). These data show that neither dlx3b, dlx4b or pax2a expression (which was lost in the absence of Dlx function) are needed to establish the normal pattern of sox3 expression. Give that this treatment disrupted some but not all features of otic development, we set out to determine whether other signals/factors present in the later developing otic placode itself might play a role in establishing the correct expression domain of sox3. This was achieved using the B380 mutant, which, due to a large deletion including the dlx3b, dlx4b and sox9a genes, expresses no markers of the definitive otic placode by the 3 somite stage

S.-K. Sun et al. / Developmental Biology 303 (2007) 675–686

681

Fig. 5. Loss of both Fgf3 and Fgf8 is needed to inhibit all placodal sox3 expression. (A–F) Wild type or ace/fgf8 mutant embryos were analysed by double in situ hybridization at 12 hpf for sox3 (red) and pax2a (blue) expression with or without injection of a morpholino targeting expression of Fgf3 (or its mismatch control). Right hand of each panel shows enlarged view of left side placode. Injection of either wild type or ace mutants with the fgf3 mismatch control had no effect on either sox3 or pax2a expression (C, D). Injection of wild type embryos with the fgf3 morpholino resulted in a reduced domain of expression of both pax2a and sox3 expression (E) as compared to untreated wild type embryos (A). The ace mutant embryos did not appear to differ from wild type embryos in their sox3 or pax2a expression (B). However, when ace mutant embryos were injected with the fgf3 morpholino, complete loss of sox3 and pax2a expression was seen (F). (G–L) In situ for either pax2a (G–I) or sox3 (J–L) expression in wild type embryos injected with morpholinos that targeted either fgf3 (G,J), fgf8 (H,K) or both fgf3 and fgf8 together (I, L). As seen with injection of ace mutants with the fgf3 morpholino (F), only loss of both fgfs resulted in complete loss of pax2a or sox3 expression in the prospective placodes (I, L).

and consequently lacks otic development (Liu et al., 2003; Solomon and Fritz, 2002). Analysis of these mutant embryos, identified by lack of expression of pax2a, revealed that the domain of sox3 expression was still present and showed little difference to wild-type embryos, although there may have been some expansion of the sox3 domain medially, where otic markers were now absent (Fig. 6D). Thus, it seems that the precise localization of the sox3 expression domain is independent of all aspects of definitive otic development analysed here. Our data indicate that despite

being initiated by the same signals as early otic markers, sox3 expression in the epibranchial placodes is independent of the adjacent developing otic placode and vesicle. Foxi1 is required for normal sox3 expression in the epibranchial placodes foxi1 expression is one of the earliest markers of the ectodermal region in which both the otic and epibranchial placodes arise. It is first expressed in a broad domain of

682

S.-K. Sun et al. / Developmental Biology 303 (2007) 675–686

Fig. 6. Relationship between dlx3b and sox3 expression. (A) Double in situ hybridization illustrating the appearance of placodal sox3 expression (red) at the caudal end of the pre-placodal domain of dlx3b expression (blue) at the 3-somite stage of development (11 hpf). (B–D) Double in situ hybridization for sox3 expression (red) and pax2a expression (blue) at 12 hpf. In untreated embryos expression of pax2a occupies a domain adjacent to, and medial to sox3 expression (B). Injection of embryos with morpholinos directed against both dlx3b and dlx4b results in loss of pax2a expression in the otic region, but has no effect on sox3 expression (C). Embryos homozygous for the B380 mutation show an identical loss of pax2a expression and normal sox3 expression (D).

ectoderm outside of the prospective neural plate prior to tailbud stage (Riley and Phillips, 2003) which seems to lie just lateral to the other preplacodal markers such a the six and dlx genes (Matsuo-Takasaki et al., 2005). From tailbud stage, foxi1 expression becomes restricted to the region around the prospective otic placode. Like dlx3b, foxi1 expression has been shown to be a prerequisite for otic development (Solomon et al., 2003). Foxi1 appears to play an early role in otic development, giving competence for expression of pax8 in response to Fgf signaling from the hindbrain. In addition, foxi1 is expressed in the epibranchial placodes and has been shown to be both necessary and sufficient for epibranchial neurogenesis (Lee et al., 2003). We therefore examined the relationship between foxi1 and sox3 in more detail. The domain of foxi1 expression at tailbud precedes the appearance of sox3 transcripts in the prospective epibranchial region (Fig. 7H), and sox3 expression first appears within the caudal part of the foxi1 domain by about 11 hpf (Fig. 7I). However, the appearance of sox3 expression coincides with a significant downregulation of foxi1 expression (Figs. 7I, J). Hence, by the 3–5 somite stage (12 hpf), foxi1 expression weakens considerably, with a stronger patch only seen at its most anterior limit (Figs. 7D, J). The location of these domains is clearly shown by comparison to krox20 expression, which supports the observation that the rostrocaudal position of the overall foxi1 domain is static, but strong expression becomes restricted only to the most anterior part of that domain (Figs. 7A–D). The full domain of foxi1 expression, as described by Lee et al., (2003) was revealed, as seen in Figs. 7B–D, in the absence of double in situ hybridization for sox3 expression which tended to mask the weaker expression

domain of foxi1. The close association of sox3 and foxi1 expression in the placodal ectoderm is maintained at 24 hpf (Figs. 7E, F, K–N). A major role for foxi1 in later epibranchial placode development has been shown by analysis of the no soul mutant in which a point mutation is believed to interfere with DNAbinding of the Foxi1 protein. These mutants exhibit a specific loss of epibranchial and lateral line neurogenesis (Lee et al., 2003). We therefore tested whether foxi1 was required for sox3 expression. Injection of a foxi1 morpholino (Lee et al., 2003; Mackereth et al., 2005) resulted in loss of pax2a in the otic domain, as has been shown previously in foxi1 mutants (Solomon et al., 2003) (Figs. 7O, P). These injected embryos also failed to express sox3 in the placodal region, while sox3 expression in the CNS was unaffected (Figs. 7Q, R). Thus, foxi1 appears to give competence for the ectoderm to express both pax8 and sox3 in response to Fgf signaling. Discussion The role of Fgf signaling in formation of the epibranchial placodes Our data support a model for the epibranchial placodes in which their initial cranial location is established via in response to Fgf signaling. The hindbrain represents a likely source for these Fgfs. As the head grows and the ectoderm consequently expands, the placodal ectoderm is displaced laterally and ventrally such that it no longer lies in close proximity to the CNS, but the placodal state is then maintained by Fgf signaling from the branchial endoderm (Nechiporuk et al., 2005). A consequence of this is that the segmental structural features of the fgf expressing endoderm imposes a segmental pattern on the placodal ectoderm, which in turn might in part explain the segmental arrangement of the cranial ganglia. The role of the pharyngeal endoderm in neurogenesis during placode development is well established (Begbie et al., 1999; Holzschuh et al., 2005; Nechiporuk et al., 2005). However, as pointed out by Streit (2004), experiments that showed Bmp signaling from the endoderm could induce neurogenesis used ectoderm in or near to regions fated to become placode, and ‘it was therefore possible that these tissues had already received other signals…which may have established a placodal bias’ (Streit, 2004). Our data suggest that these early signals are Fgfs and that sox3 expression reflects this placodal bias. It is stated in the most recent reviews of placode development that ‘the epibranchial placodes have a single specific inducing tissue, the pharyngeal endoderm.’ (Graham and Begbie, 2000) and that ‘Neither neural tube nor neural crest are required for formation of the epibranchial placodes.’ (Baker and Bronner-Fraser, 2001). These statements arise from two studies in chick embryos. In the study of Yntema (1944) only the neural crest and dorsal neural tube were removed, thus not testing a requirement for the hindbrain itself (Yntema, 1944). More recently, Begbie et al. (1999) did actually ablate the neural tube in the region of the hindbrain, primarily in order to remove the neural crest, and saw no effect on placodal neurogenesis (Begbie et al., 1999). However, these experiments were carried

S.-K. Sun et al. / Developmental Biology 303 (2007) 675–686

683

Fig. 7. Relationship between foxi1 and sox3 expression. All panels viewed dorsally at the level of the hindbrain with rostral to the top, except (A, B, E, G, H, K, M) which are viewed laterally. (A, B, G, H) dorsal to the right, (E, K, M) rostral to left. (A–F) Double in situ hybridization for foxi1 (blue) expression illustrating its changing expression in relation to krox20 (red) which marks rhombomeres 3 and 5 in the hindbrain as a positional point of reference. foxi1 expression develops from a single large domain either side of the hindbrain (A), gradually elongating and finally separating into almost distinct patches but always extending from a position slightly rostral to a position slightly caudal to rhombomeres 2–5 (B–F). However, from early somite stages (B) onwards, the level of expression decreases in the more caudal and lateral regions with strong expression maintained only in the rostral-most region (this is particularly evident in panels (I, J)). (G–L) Double in situ hybridization for sox3 (red) and foxi1 (green). (M, N) Single in situ hybridization for sox3 (blue). foxi1 expression outside of the CNS precedes placodal sox3 expression which only appears at tailbud/early somite stages (G–I). Prior to this stage, foxi1 exists as a shrinking domain either side of the hindbrain that does not directly abut the CNS expression of sox3 (G, H). As sox3 expression appears within the larger foxi1 domain of expression, there is a concomitant down regulation of foxi1 in the region of overlap (G, H). This is less apparent in (C, D) since in situ hybridization for foxi1 has been overdeveloped to reveal its full domain of expression. At later stages (K–N) the domains of sox3 and foxi1 remain overlapping with the same rostrocaudal pattern, although the sox3 expression lies more medially and the foxi1 extends more laterally. Note that at 24 hpf expression of sox3 in the underlying endoderm is seen as a particularly dark stain (arrows) below the more diffuse patches of ectodermal expression (outlined in panel N) when viewed dosally (L, N). (O–R) Effects of foxi1 morpholino. In situ hybridization at 12 hpf shows placodal expression of pax2a (O) and sox3 (Q) in wildtype embryos (arrowheads), is lost in embryos injected with the foxi1 morpholino, (P) and (R) respectively.

out at Hamilton Hamburger stage 9 (6–9 somites) when the patches of Sox3-expressing, thickened ectoderm are already established and might now be supported by alternative sources of Fgf such as the epibranchial endoderm. This is consistent with the expression of Fgf3 and Fgf8 in the pharyngeal pouches of chick (Hidalgo-Sanchez et al., 2000; Mahmood et al., 1995) and the recent study of Nechiporuk et al. (2005) showing that Fgf3 from endoderm was sufficient to induce epibranchial neurogenesis in zebrafish and only fgf3 (not fgf8, fgf4 or fgf24) was required to maintain this neurogenesis. Thus, in the study of Begbie et al. (1999) the chick hindbrain was ablated, not necessarily completely, after the stage equivalent to the 3 somite stage of zebrafish when we have shown that Fgf signaling from the hindbrain is required to support sox3 expression in the prospective epibranchial placodes. It therefore appears that these earlier experiments might have failed to show a requirement for the hindbrain because manipulations were carried out after the requirement had passed. However, Litsiou et al. (2005) have shown that Fgf signaling is also a necessary part of the mesoderm derived signal which induces the preplacode in chick embryos (Litsiou et al., 2005). Our data do not exclude the

possibility of an alternative source of Fgf signals in the early phase of placodal induction. Does this model fit with studies in other species? In this respect, studies of otic development are pertinent. Investigation of the role of Fgfs in otic development has identified some differences between chick, mouse and zebrafish. While only Fgf3 and Fgf8 emanating from the hindbrain appear to be required for otic development in zebrafish, other Fgfs, including Fgf19 or Fgf10 from the mesenchyme, appear to play a role in chick and mouse respectively (Ladher et al., 2000; Wright and Mansour, 2003). Kil et al. (2005) also showed that in chick, the positional identity of the hindbrain adjacent to the forming otic placode is not critical in determining where the placode forms (Kil et al., 2005). Thus, it seems quite feasible that the situation with respect to the epibranchial placodes may similarly differ somewhat between species. There is evidence for a role of the mesenchyme in zebrafish, in that loss of mesendoderm in the one eyed pinhead (oep) mutant zebrafish results in delayed otic development, although this may be an indirect effect (Mendonsa and Riley, 1999). However, expression patterns suggest that the later phase of epibranchial placode patterning

684

S.-K. Sun et al. / Developmental Biology 303 (2007) 675–686

might be well conserved between species. Bmp signaling from the pharyngeal endoderm appears to be a significant signal inducing neurogenesis in both chick (Begbie et al. 1999) and zebrafish (Holzschuh et al., 2005). Similarly fgf3 is expressed in the pharyngeal endoderm of zebrafish, chick and mouse immediately adjacent to the region where the placodes develop. Thus, although the role of Fgf3 in placodal neurogenesis has been tested in zebrafish (Nechiporuk et al., 2005) its role in maintaining the epibranchial placodes themselves has not been studied in detail. However, Fgf3 remains a likely candidate for this role across species. Patterning of the ectoderm Although we have shown that the prospective epibranchial placodal domain of sox3 expression, like the otic placode, is dependent on Fgf signaling, the actual regions of ectoderm that adopt these fates are restricted by factors making ectoderm competent to respond. In the case of the otic placodes, dlx3b, dlx4b and foxi1 have been implicated (Hans et al., 2004; Nissen et al., 2003; Solomon et al., 2003). However, our data show that the domain of sox3 expression is independent of most definitive otic development and otic genes. Since the genomic deletion in the B380 mutant results in complete absence of all later aspects of ear development (Liu et al., 2003; Solomon and Fritz, 2002), our experiments rule out almost all known genes to be expressed in this region. These experiments disrupting the formation of the otic placode provide strong evidence that induction of the placodal sox3 expression is a direct effect of Fgf signaling rather than an indirect effect in which a secondary placode-inducing signal emanates from the Fgf-induced otic tissue. Fig. 8. Proposed models for ectodermal patterning. (A) Early patterning of the ectoderm. Solid arrows indicate clear positive regulation. Broken arrows represent positive regulation which is not necessary for basic level expression, but does increase expression in the placodal domains. Broken T lines indicate potential negative regulation based on expression patterns. Fgf signaling induces the expression of pax2a, pax8 and sox3. This induction is dependent upon the prior expression within the ectoderm of dlx3b or dlx4b for pax2a expression and foxi1 in the case of pax8 and sox3 expression. Pax8 and Pax2a are also known to positively regulate their own expression and are necessary to maintain expression of the later otic markers sox9a and sox9b (Hans et al., 2004). pax8 and sox3 expression appear first in a common domain. pax2a expression appears later in a Pax8 dependent manner, with pax8 expression becoming restricted to that same domain which then loses expression of sox3. These interactions explain the final pattern of gene expression and location of the otic and epibranchial placodes, where the otic domain acquires a position immediately adjacent to the sox3-positive epibranchial placodes. (B) Later maintenance of placodes and the initiation of neurogenesis. As growth displaces the future placodal ectoderm ventrolaterally away from the CNS, Fgfs (red) derived from the pharyngeal endoderm maintain the placodal state of the surface ectoderm, including sox3 expression (orange). Sox3 expression in the CNS and the pharyngeal endoderm itself is also shown. As Bmp (blue) expression begins in the same pharyngeal endoderm this promotes neurogenesis with activation of neurogenin 1 expression (green) in the surface ectoderm, delamination and inward migration of those cells and concomitant loss of sox3 expression. Thus, maintenance of sox3-expressing ectoderm versus neurogenesis is decided through the opposing effects of Fgf versus Bmp signaling respectively. Nearest to the endoderm, Fgf signaling overrides Bmp signaling whereas further from the endoderm the effects of Bmp signaling override the effects of Fgfs. Thus, neurogenesis only occurs at the edge of the patch of sox3-expressing ectoderm as has been described previously (Abu-Elmagd et al., 2001).

We have, however, found that expression of sox3 is dependent on the presence of foxi1. With respect to the ear, foxi1 is also required for expression of pax8 and later markers such as pax2a (Solomon et al., 2003). The effect on later markers may be due to their dependence on Pax8 function. Thus, although foxi1 expression precedes development of both the otic and epibranchial placodes, initially exhibiting a domain of expression much broader than the placodal region, this factor does seem to be a prerequisite for the development of both types of placode. Indeed, later development of the derivatives of epibranchial placodes is severely disrupted by loss of foxi1 (Lee et al., 2003) and the epibranchial placodes, as defined by their thickened ectodermal morphology have been shown to arise from foxi1-expressing ectoderm (Nechiporuk et al., 2005). An additional observation is that foxi1 expression decreases (but is not lost) in the region where sox3 expression appears. Since the appearance of sox3 expression within the foxi1 expressing ectoderm is dependent upon Fgf signaling, this could explain the earlier observation that foxi1 expression actually

S.-K. Sun et al. / Developmental Biology 303 (2007) 675–686

685

increases when Fgf3 is blocked (Nechiporuk et al., 2005). A conflicting observation in the study of Nechiporuk et al. (2005) is that treatment with SU5402 resulted in loss of foxi1 expression, but did not cause loss of thickened ectoderm in the region of the epibranchial placodes. However, the effects of SU5402 are rarely as pronounced as blocking with a combination of fgf morpholinos, suggesting that the doses of SU5402 are insufficient to completely inhibit the Fgf response. Also, Nechiporuk et al. only treated the embryos for a period of 4 h from 22 hpf to 26 hpf, and this might have been insufficient time to reveal the full consequences of loss of Fgf signalling. Therefore, the retention of thickened ectoderm in these experiments could reflect some residual Fgf signaling. Our data allow us to propose a model of the signaling events and transcriptional interactions that pattern the ectoderm in the region of the otic/epibranchial placodes (Fig. 8A). In this model, both sox3 and pax8 expression are induced by Fgf signaling in a manner dependent on the presence of Foxi1. The region of ectoderm that expresses sox3 and pax8 is therefore determined by the location of both the Fgf-producing hindbrain and the ectoderm expressing foxi1. This common region of sox3/pax8 expression (as seen in Fig. 2G) represents the prospective field for both placodes, and overlaps the previously described ‘preplacodal’ domain of expression of dlx3b/dlx4b. The existence of a common primordium for the otic and epibranchial placodes is consistent with data from other species where fate maps indicate that these two placode types initially occupy a common domain (Streit, 2004; Washausen et al., 2005). The next step in patterning this region is the induction by Fgf signaling of pax2a in the most medial part of this sox3/pax8 expressing ectoderm, in a manner dependent upon both pax8 and dlx3b/dlx4b. Our data suggest that the appearance of pax2a expression is soon followed by loss of sox3 expression in that domain and restriction of pax8 expression to the same medial domain as pax2a (Hans et al., 2004). Hence this now defines the two adjacent placodal regions; medially the prospective otic region expressing pax2a and pax8 and laterally the prospective epibranchial region expressing sox3. Subsequent otic specific events are then dependent upon Pax2a and Pax8 while the role of Sox3 in subsequent epibranchial development has yet to be tested.

neurogenesis in the chick epibranchial region is initiated under the influence of Bmp7 from the pharyngeal endoderm. Together with our study, these data support a model in which Fgf signaling, initially from the hindbrain and later from the pharyngeal endoderm, determines which regions of ectoderm will adopt a sox3-expressing, neurogenic competence. Neurogenesis is then initiated by Bmp signaling from the same endoderm. foxi1 is initially expressed independently of Fgf signaling (Lee et al., 2003), but its expression is necessary for the Fgf-induced expression of sox3. Hence, this might be why loss of foxi1 also results in loss of neurogenesis in response to Fgf. As mentioned above, we have shown previously in the epibranchial placodes that sox3 expression must be lost in order for neural cells to realize their neurogenic potential (Abu-Elmagd et al., 2001), a feature shared with the CNS (Bylund et al., 2003). Since the pharyngeal endoderm produces Fgf3 (which promotes sox3 expression) and Bmps (which stimulate neurogenesis with concomitant loss of sox3 expression) the precise location of cells undergoing neurogenesis may result from competition between these two signals. In support of this hypothesis, our earlier studies in chick showed that delaminating neurons arise from the dorsal edge of the sox3-expressing placodes at a distance from the pharyngeal endoderm. Thus, we propose a model (Fig. 8B) in which the ectoderm nearest to the endoderm is maintained in a sox3-expressing, placodal state by high concentration of Fgf3 (which overrides the effects of the endoderm derived Bmp signal), while more dorsally, the concentration of Bmp signaling is still sufficient to initiate neurogenesis and override the sox3-inducing effects of a lower concentration of Fgf3. A precedent for competition between Bmp and Fgf signaling in neural development has recently been described. In the earliest stages of induction of the CNS it has been shown that Fgf signaling antagonizes Bmp signaling (Pera et al., 2003; Sater et al., 2003). This antagonism is at least in part due to Fgf activated MAPK causing inactivation of Smad. Thus, in both the CNS and placodes, the outcome of these signaling events depends upon the relative concentrations of the two classes of signaling factors.

A model for the sequential steps in development of the epibranchial placodes

We are very grateful to Monty Westerfield and his colleagues for providing B380 mutant embryos. Thanks to Yi-Chuan Cheng for gene probes and Roger Patient for gene probes and support of the zebrafish facility. Thanks also to Elaine Thorpe and Dave Reffin for their technical help.

We have shown that an initial large domain of sox3expressing ectoderm forms as a result of Fgf3 and Fgf8 signaling likely to emanate from the hindbrain. As the embryo grows, this initial domain is subdivided into the final series of epibranchial placodes in the vicinity of the pharyngeal pouches. Two recent studies have demonstrated in zebrafish, as has been shown in chick embryos, that the subsequent epibranchial neurogenesis is dependent upon signals from the endoderm. Nechiporuk et al. (2005) recently showed that Fgf3 signaling from the endoderm is required for epibranchial neurogenesis. Holzschuh et al. (2005) on the other hand, showed that either Bmp2b or Bmp5 are required for the initiation of neurogenesis, consistent with the data of Begbie et al. (1999) who showed that

Acknowledgments

References Abu-Elmagd, M., Ishii, Y., Cheung, M., Rex, M., Le Rouedec, D., Scotting, P., 2001. cSox3 expression and neurogenesis in the epibranchial placodes. Dev. Biol. 237, 258–269. Alsina, B., Abello, G., Ulloa, E., Henrique, D., Pujades, C., Giraldez, F., 2004. FGF signaling is required for determination of otic neuroblasts in the chick embryo. Dev. Biol. 267, 119–134. Araki, I., Brand, M., 2001. Morpholino-induced knockdown of fgf8 efficiently phenocopies the Acerebellar (ace) phenotype. Genesis 30, 157–159. Baker, C., Bronner-Fraser, M., 2001. Vertebrate cranial placodes I. Embryonic induction. Dev. Biol. 232, 1–61.

686

S.-K. Sun et al. / Developmental Biology 303 (2007) 675–686

Begbie, J., Graham, A., 2001. Integration between the epibranchial placodes and the hindbrain. Science 294, 595–598. Begbie, J., Brunet, J.-F., Rubenstein, J., Graham, A., 1999. Induction of the epibranchial placodes. Development 126, 895–902. Bylund, M., Andersson, E., Novitch, B.G., Muhr, J., 2003. Vertebrate neurogenesis is counteracted by Sox1–3 activity. Nat. Neurosci. 6, 1162–1168. Ekker, M., Akimenko, M.A., Bremiller, R., Westerfield, M., 1992. Regional expression of three homeobox transcripts in the inner ear of zebrafish embryos. Neuron 9, 27–35. Graham, A., Begbie, J., 2000. Neurogenic placodes: a common front. Trends Neurosci. 23, 313–316. Hans, S., Liu, D., Westerfield, M., 2004. Pax8 and Pax2a function synergistically in otic specification, downstream of the Foxi1 and Dlx3b transcription factors. Development 131, 5091–5102. Hidalgo-Sanchez, M., Alvarado-Mallart, R.-M., Alvarez, I., 2000. Pax2, Otx2, Gbx2 and FGF8 expression in early otic vesicle development. Mech. Dev. 95, 225–229. Holzschuh, J., Wada, N., Wada, C., Schaffer, A., Javidan, Y., Tallafuss, A., BallyCuif, L., Schilling, T.F., 2005. Requirements for endoderm and BMP signaling in sensory neurogenesis in zebrafish. Development 132, 3731–3742. Ishii, Y., Abu-Elmagd, M., Scotting, P.J., 2001. Sox3 expression defines a common primordium for the epibranchial placodes in chick. Developmental 344–353. Jowett, T., 2001. Double in situ hybridization techniques in zebrafish. Methods 23, 345–358. Kil, S.H., Streit, A., Brown, S.T., Agrawal, N., Collazo, A., Zile, M.H., Groves, A.K., 2005. Distinct roles for hindbrain and paraxial mesoderm in the induction and patterning of the inner ear revealed by a study of vitamin-Adeficient quail. Dev. Biol. 285, 252–271. Kimmel, C.B., Ballard, W.W., Kimmel, S.R., Ullmann, B., Schilling, T.F., 1995. Stages of embryonic development of the zebrafish. Dev. Dyn. 203, 253–310. Krauss, S., Johansen, T., Korzh, V., Fjose, A., 1991. Expression of the zebrafish paired box gene pax[zf-b] during early neurogenesis. Development 113, 1193–1206. Kudoh, T., Concha, M.L., Houart, C., Dawid, I.B., Wilson, S.W., 2004. Combinatorial Fgf and Bmp signalling patterns the gastrula ectoderm into prospective neural and epidermal domains. Development 131, 3581–3592. Ladher, R.K., Anakwe, K.U., Gurney, A.L., Schoenwolf, G.C., Francis-West, P.H., 2000. Identification of synergistic signals initiating inner ear development. Science 290, 1965–1967. Lee, S.A., Shen, E.L., Fiser, A., Sali, A., Guo, S., 2003. The zebrafish forkhead transcription factor Foxi1 specifies epibranchial placode-derived sensory neurons. Development 130, 2669–2679. Leger, S., Brand, M., 2002. Fgf8 and Fgf3 are required for zebrafish ear placode induction, maintenance and inner ear patterning. Mechanisms of Development 119, 91–108. Litsiou, A., Hanson, S., Streit, A., 2005. A balance of FGF, BMP and WNT signalling positions the future placode territory in the head. Development 132, 4051–4062. Liu, D., Chu, H., Maves, L., Yan, Y.L., Morcos, P.A., Postlethwait, J.H., Westerfield, M., 2003. Fgf3 and Fgf8 dependent and independent transcription factors are required for otic placode specification. Development 130, 2213–2224. Mackereth, M.D., Kwak, S.J., Fritz, A., Riley, B.B., 2005. Zebrafish pax8 is required for otic placode induction and plays a redundant role with Pax2 genes in the maintenance of the otic placode. Development 132, 371–382. Mahmood, R., Kiefer, P., Guthrie, S., Dickson, C., Mason, I., 1995. Multiple roles for FGF-3 during cranial neural development in the chicken. Development 121, 1399–1410. Maroon, H., Walshe, J., Mahmood, R., Kiefer, P., Dickson, C., Mason, I., 2002. Fgf3 and Fgf8 are required together for formation of the otic placode and vesicle. Development 129, 2099–2108. Matsuo-Takasaki, M., Matsumura, M., Sasai, Y., 2005. An essential role of Xenopus Foxi1a for ventral specification of the cephalic ectoderm during gastrulation. Development 132, 3885–3894. Mendonsa, E.S., Riley, B.B., 1999. Genetic analysis of tissue interactions required for otic placode induction in the zebrafish. Dev. Biol. 206, 100–112. Nechiporuk, A., Linbo, T., Raible, D.W., 2005. Endoderm-derived Fgf3 is

necessary and sufficient for inducing neurogenesis in the epibranchial placodes in zebrafish. Development 132, 3717–3730. Nieto, M.A., Sechrist, J., Wilkinson, D.G., Bronner-Fraser, M., 1995. Relationship between spatially restricted Krox-20 gene expression in branchial neural crest and segmentation in the chick embryo hindbrain. EMBO J. 14, 1697–1710. Nissen, R.M., Yan, J., Amsterdam, A., Hopkins, N., Burgess, S.M., 2003. Zebrafish foxi one modulates cellular responses to Fgf signaling required for the integrity of ear and jaw patterning. Development 130, 2543–2554. Okuda, Y., Yoda, H., Uchikawa, M., Furutani-Seiki, M., Takeda, H., Kondoh, H., Kamachi, Y., 2006. Comparative genomic and expression analysis of group B1 sox genes in zebrafish indicates their diversification during vertebrate evolution. Dev. Dyn. 235, 811–825. Penzel, R., Oschwald, R., Chen, Y., Tacke, L., Grunz, H., 1997. Characterization and early embryonic expression of a neural specific transcription factor xSOX3 in Xenopus laevis. Int. J. Dev. Biol. 41, 667–677. Pera, E., Ikeda, A., Eivers, E., De Robertis, E., 2003. Integration of IGF, FGF, and anti-BMP signals via Smad1 phosphorylation in neural induction. Genes Dev. 17, 3023–3028. Pfeffer, P.L., Gerster, T., Lun, K., Brand, M., Busslinger, M., 1998. Characterization of three novel members of the zebrafish Pax2/5/8 family: dependency of Pax5 and Pax8 expression on the Pax2.1 (noi) function. Development 125, 3063–3074. Phillips, B.T., Bolding, K., Riley, B.B., 2001. Zebrafish fgf3 and fgf8 encode redundant functions required for otic placode induction. Dev. Biol. 235, 351–365. Phillips, B.T., Storch, E.M., Lekven, A.C., Riley, B.B., 2004. A direct role for Fgf but not Wnt in otic placode induction. Development 131, 923–931. Reifers, F., Bohli, H., Walsh, E.C., Crossley, P.H., Stainier, D.Y., Brand, M., 1998. Fgf8 is mutated in zebrafish acerebellar (ace) mutants and is required for maintenance of midbrain–hindbrain boundary development and somitogenesis. Development 125, 2381–2395. Rex, M., Church, R., Tointon, K., Scotting, P., 1997a. Combination of nonisotopic in situ hybridization with detection of enzyme activity, bromodeoxyuridine incorporation and immunohistochemical markers. Histochem. Cell Biol. 107, 519–523. Rex, M., Orme, A., Uwanogho, D., Tointon, K., Wigmore, P., Sharpe, P., Scotting, P., 1997b. Dynamic expression of the chicken Sox2 and Sox3 genes in ectoderm induced to form neural tissue. Dev. Dyn. 209, 323–332. Riley, B.B., Phillips, B.T., 2003. Ringing in the new ear: resolution of cell interactions in otic development. Dev. Biol. 261, 289–312. Sater, A.K., El-Hodiri, H.M., Goswami, M., Alexander, T.B., Al-Sheikh, O., Etkin, L.D., Uzman, J.A., 2003. Evidence for antagonism of BMP-4 signals by MAP kinase during Xenopus axis determination and neural specification. Differentiation 71, 434–444. Solomon, K.S., Fritz, A., 2002. Concerted action of two dlx paralogs in sensory placode formation. Development 129, 3127–3136. Solomon, K.S., Kudoh, T., Dawid, I.B., Fritz, A., 2003. Zebrafish foxi1 mediates otic placode formation and jaw development. Development 130, 929–940. Streit, A., 2004. Early development of the cranial sensory nervous system: from a common field to individual placodes. Dev. Biol. 276, 1–15. Vendrell, V., Carnicero, E., Giraldez, F., Alonso, M., Schimmang, T., 2000. Induction of inner ear by FGF3. Development 127, 2011–2019. Washausen, S., Obermayer, B., Brunnett, G., Kuhn, H.J., Knabe, W., 2005. Apoptosis and proliferation in developing, mature, and regressing epibranchial placodes. Dev. Biol. 278, 86–102. Webb, J., Noden, D., 1993. Ectodermal placodes: contributions to the development of the vertebrate head. Am. Zool. 33, 434–447. Westerfield, M., 2000. The Zebrafish Book. A Guide for the Laboratory Use of Zebrafish (Danio rerio). University of Oregon Press, Eugene. Wood, H.B., Episkopou, V., 1999. Comparative expression of the mouse Sox1, Sox2 and Sox3 genes from pre-gastrulation to early somite stages. Mech. Dev. 86, 197–201. Wright, T.J., Mansour, S.L., 2003. Fgf3 and Fgf10 are required for mouse otic placode induction. Development 130, 3379–3390. Yntema, C., 1944. Experiments on the origin if the sensory ganglia of the facial nerve in the chick. J. Comp. Neurol. 81, 147–167.