Requirements for endoderm and BMP signaling in sensory neurogenesis in zebrafish

Development of the peripheral nervous system involves the coordinated development of two uniquely vertebrate cell types, placodes and neural crest. During head development, focal thickenings of ectoderm called neurogenic placodes give rise to neurons in the paired sense organs and cranial sensory ganglia associated with cranial nerves(LeDouarin, 1982; Graham and Begbie, 2000; Baker and Bronner-Fraser,2001). These are joined by neurogenic neural crest cells to generate the mature ganglia. Some neurogenic placodes develop close to the neural tube, whereas others form next to segmented outpockets of the endoderm called `pouches' that bud from the lateral walls of the pharynx, which may play important roles in placodal induction and neurogenesis(Goodrich, 1930). There is growing evidence to suggest that, despite their similar appearance, each placode is induced by a different combination of signals from the surrounding tissues (Baker and Bronner-Fraser,2001).

Neurogenic placodes can be split into two groups, dorsolateral (trigeminal and vestibular) and epibranchial (including the facial or geniculate, VII;glossopharyngeal or petrosal, IX; vagal or nodose, X), which innervate taste buds, the heart and other visceral organs. Neural progenitors within the placodes are first specified by the expression of the basic helix-loop-helix(bHLH) transcription factors neurogenin 1 (ngn1; neurog1 – Zebrafish Information Network), neurogenin 2(ngn2; neurog3 – Zebrafish Information Network) and neurod (Sommer et al.,1996; Anderson,1999; Andermann et al.,2002; Begbie et al.,2002). Distinct subsets of placodes express these factors in different species (Schlosser and Northcutt, 2000), and each is required at an early stage in the specification of neural progenitors. Targeted inactivation of Ngn2specifically disrupts epibranchials (Fode et al., 1998), while Ngn1 is required in dorsolateral placodes in mice (Ma et al.,1998), and in all neurogenic placodes in zebrafish(Andermann et al., 2002),further suggesting that separate mechanisms control dorsolateral and epibranchial development. The paired homeodomain transcription factors Phox2a and Phox2b are also required for epibranchial differentiation and survival (Tiveron et al., 1996; Valarche et al.,1993). Thus placodes are distinguished by unique patterns of neurogenesis, and these may reflect their responses to different signals.

Pharyngeal pouches in the endoderm are thought to induce neurogenesis in the epibranchial placodes through expression of bone morphogenetic protein 7(BMP7), a member of the transforming growth factor beta (TGFβ)superfamily (Begbie et al.,1999; Luo et al.,1995). Both endoderm and exogenous BMP7 protein can induce neural progenitors in ectodermal explants in culture. Once epibranchial sensory neurons have established their axonal connections, their survival depends upon glial-derived neurotrophic factor (GDNF), which is also a member of the TGFβ superfamily (Buj-Bello et al.,1995). By contrast, a signal from the prospective midbrain-hindbrain boundary (MHB), possibly a fibroblast growth factor (FGF)is thought to induce the trigeminal (Stark et al., 1997; Baker et al.,1999), and the otic placode depends on FGF signaling from the hindbrain (Phillips et al.,2001). These studies suggest that epibranchial and dorsolateral placodes are induced by different signals depending on their proximity to the endoderm or neural tube.

Pharyngeal endoderm also physically interacts with neural crest mesenchyme that forms cartilage and bone, and promotes skeletal differentiation(Hall, 1980; Le Douarin, 1982). Recent evidence suggests that, in this context, the endoderm plays an instructive role in anteroposterior (AP) patterning, as removing and reinserting the pharyngeal endoderm in a reversed AP orientation can cause mandibular duplications (Couly et al.,2002). Zebrafish casanova (cas; sox32– Zebrafish Information Network)(Kikuchi et al., 2000; Dickmeis et al., 2001)mutants, which lack all endoderm, fail to form pharyngeal cartilages due to a lack of local cartilage-inducing signals(David et al., 2002). The pharyngeal pouches appear to play a crucial role in this interaction, because in van gogh (vgo; tbx1 – Zebrafish Information Network) mutants, as well as mutants that lack integrin alpha 5 function, defects in pouch formation correlate with subsequent cartilage malformations (Piotrowski and Nusslein-Volhard, 2000; Piotrowski et al., 2003; Crump et al., 2004a). The primary defect in all of these cases lies in the endoderm, as reintroduction of endodermal cells into mutants rescues cartilage formation and pharyngeal patterning.

Here, we report that the pharyngeal endoderm is required for the induction of the epibranchial nervous system in zebrafish, and that BMPs play an important role in this process. The sequential formation of pharyngeal pouches correlates precisely with the onset of neurogenesis in the epibranchial placodes, which we show by following endodermal morphogenesis in the living embryo. Mutants that disrupt pouch formation, such as cas and vgo, have corresponding defects in epibranchial, but not dorsolateral, placodes. We further show that the reintroduction of wild-type endoderm can rescue epibranchial development in cas mutants. Several lines of evidence suggest that this interaction depends on BMP signaling,including endoderm-specific inhibition of BMPs, as well as exogenous application of BMP proteins or BMP inhibitors. These are the first studies to show that the endoderm is required to induce neurogenesis in epibranchial placodes (but not dorsolateral placodes) in vivo, and that BMPs other than BMP7 are involved. In addition, they go beyond previous work in demonstrating that endodermal segmentation (pouch formation) controls the spatial patterning of sensory neurogenesis in the embryo.

Embryos were produced by paired matings of zebrafish (Danio rerio), and staged in hours post-fertilization (hpf), as described previously (Kimmel et al.,1995; Westerfield,1995). Embryos homozygous for the casanova(cas^ta56), van gogh(vgo^tm208) and snailhouse(snh^ty68) mutations were derived from matings between heterozygous carriers, and scored by their heart, ear or tail defects,respectively (Kikuchi et al.,2000; Dickmeis et al.,2001; Piotrowski et al.,1996; Dick et al.,2000).

A fragment containing 700 bp of her5 upstream sequence driving egfp (–0.7her5:egfp) was obtained by PCR from a larger construct containing 3650 bp (her5PAC:egfp), as described previously(Tallafuss and Bally-Cuif,2003). This was purified and injected as a linear fragment into fertilized eggs at the one-cell stage at a concentration of 50 ng/μl. Embryos were then raised to adulthood and mated with wild-type adults. F1 embryos expressing egfp were identified and mated with wild types to establish the transgenic lines. Homozygous her5:egfp transgenic fish were then generated in natural matings, and expression verified over at least three generations.

For analysis of pouch formation, pharyngeal endoderm was labeled by the her5:egfp transgene starting at 14-15 hpf. Transgenic embryos were manually dechorionated and anesthetized with ethyl-m-aminobenzoate methane sulfanate (Westerfield, 1995),transferred to 0.5% agarose in embryo medium and then mounted on a coverslip. Approximately 80 μm Z-stacks at 6 μm intervals were captured using a Zeiss LSM510 Meta confocal fluorescence microscope.

Cell transplantations were targeted to the endoderm using either an injection of mRNA encoding the activated Taram-A receptor (tar*), as previously described (Aoki et al.,2002a; Aoki et al.,2002b; David et al.,2002), or cas (sox32) mRNA(Dickmeis et al., 2001). Briefly, wild-type donor embryos were injected at the one-cell stage with a mixture of 2% tetramethylrhodamine-isothiocyanate dextran and 3%lysine-fixable biotin dextran (10,000 M_r, Molecular Probes) together with 1 ng cas or 1 pg tar* RNA (cas/dextran or tar*/dextran, respectively). At late blastula stages, cells from these RNA-injected donors were transplanted to the margins of wild-type or cas mutant hosts. The resulting mosaic embryos were selected using a Leica fluorescence stereomicroscope for those containing large numbers of transplanted cells in the pharyngeal endoderm. Such transplants were typically restricted to one side of the pharynx or to individual pouches, with the contralateral side serving as a control and allowing the identification of rescued cas mutants.

Morpholino oligomers targeted to the translation start sites of bmp2b,bmp7 (Imai and Talbot,2002) and bmp5 (GenBank #NM 201051 –CCACAGAAGTTCCAAATGTTCTCAT) were obtained from Gene Tools, diluted in 1×Danieau's buffer and injected together with the tar*/dextran or cas/dextran mixtures. Volumes injected were calculated for each microinjection needle at a particular injection pressure, using a micrometer imprinted on a glass slide to measure the diameter of a droplet produced at its tip with a single injection pulse. Amounts injected per embryo were then chosen for each morpholino that phenocopied loss-of-function mutations: 300 pg of bmp2b MO phenocopies swirl mutants; 900 pg of bmp7 MO phenocopies snailhouse mutants(Imai and Talbot, 2002). For bmp5, we found that 1 ng of the bmp5-MO caused a slight reduction in head size, but otherwise no clear phenotype on its own, but effectively reduced epibranchials when injected into endoderm. A similar result was obtained using a second, non-overlapping bmp5-MO.

Human recombinant BMP4, BMP5, BMP7 and Noggin proteins (R&D Systems)were used for bead implantation experiments. CM-Affi-Gel Blue beads (diameter:70-100 μm, Bio-Rad) were incubated in each protein solution (BMPs, 1-10μg/ml; NOGGIN, 100-500 μg/ml) at 4°C for 1 hour. Using a tungsten needle, a small slit was made anterior to the otic vesicle on one side of the head at 20 hpf. A protein-coated bead was inserted into the hole, and positioned beneath the ectoderm. Embryos were raised to 48 hpf, and fixed for phox2b analysis by in situ hybridization. As controls, we used beads coated in 1% BSA-PBS.

In situ hybridization was carried out as described previously(Thisse et al., 1993). Probes used were: bmp2a, bmp2b and bmp4(Martinez-Barbera et al.,1997), bmp7 (Dick et al., 2000), neurod(Korzh et al., 1998), foxi1 (Nissen et al.,2003; Solomon et al.,2003) and phox2a (Guo et al., 1999). A cDNA encoding bmp5 was isolated from a gridded zebrafish pharyngula-stage (24 hpf) library using human BMP7 as a probe, cloned into the pSPORT vector. To generate a bmp5 probe for in situ hybridization, we linearized with EcoRI and transcribed with SP6. To generate a probe specific for phox2b expression, we amplified by RT-PCR the complete coding sequence of zebrafish phox2b (GenBank#AY166856), using primers designed from phox2b genomic sequence(Sanger Centre, Hinxton, UK). The PCR product was cloned into the pBS-KS vector and verified by sequencing. To generate the probe containing the phox2b complete coding sequence, we linearized with Kpn1 and transcribed with T3 polymerase.

Immunolabeling with the anti-Hu and Zn-8 antibodies was done as described previously (Marusich et al.,1994; Trevarrow et al.,1990). After incubation in the primary antibodies, embryos were incubated with biotin-conjugated secondary antibodies using the Vectastain Kit, following the manufacturers instructions. For double staining, standard in situ hybridization was followed by three washes for 5 minutes each in 0.1 M glycine buffer (pH 2) before proceeding with immunohistochemistry.

For sectioning, embryos labeled by in situ hybridization or immunohistochemistry were embedded in 1 ml of gelatin-albumin (0.025 g gelatin, 1.3 g BSA, 0.9 g saccharose, in 4.5 ml PBS) for 1 hour at room temperature. This mixture was then replaced with 1 ml of gelatine-albumin + 35μl of 50% gluteraldehyde. The embedding mixture was allowed to harden overnight at 4°C. Sections were cut at 10-20 μm with a vibratome(VT1000S, Leica), and mounted in Glyergel (Dako).

Epibranchial placodes in the zebrafish first express the bHLH genes ngn1 and neurod beginning at 24 hpf in three cell clusters(facial, glossopharyngeal and vagal)(Andermann et al., 2002). At 30 hpf, some neuroblasts within the placodes express phox2b and become immunoreactive for anti-Hu (α-Hu) antibody(Fig. 1A,B) (Raible and Kruse,2000). By 48 hpf, all six epibranchial ganglia have formed and enlarged dramatically (Fig. 1C,D). To examine the sequence of neuronal maturation in these ganglia more precisely,we performed double staining with in situ hybridization for neurod or phox2b combined with anti-Hu immunohistochemistry. We found that although neurod and Hu are co-expressed in dorsolateral placodes(trigeminal, auditory, vestibular, lateral line; Fig. 1A), they only partially overlap in the emerging glossopharyngeal/petrosal (gIX) and vagal/nodose(gX_1-4) ganglia. Distal cells within each ganglion express neurod, whereas more proximal cells express Hu, and this pattern persists at 48 and 72 hpf, suggesting that neuroblasts are generated distally and become displaced proximally as they mature(Fig. 1C,C′,E). Consistent with this hypothesis, phox2b mRNA and Hu antigen co-localize in proximal cells in each epibranchial ganglion(Fig. 1D,D′,F). Based on these studies, we can define three distinct cell populations within an epibranchial ganglion in a distal to proximal order: (1) distal cells expressing neurod alone; (2) intermediate cells expressing Hu antigen, phox2a, phox2b and neurod; and (3) proximal cells expressing Hu, phox2a and phox2b, but not neurod. Thus, newly generated neurod+ neural progenitors are born directly adjacent to the dorsal ends of the pharyngeal pouches, and then downregulate neurod and express phox2b as they mature within the ganglion and are displaced proximally (Fig. 1G). Labeling of placodal cells in the chick has demonstrated a similar proximal displacement (Begbie and Graham, 2001).

The initiation of epibranchial neurogenesis around 24 hpf roughly coincides with the time at which the pharyngeal endoderm segments into pouches in zebrafish (Crump et al.,2004b). To correlate these two sets of events more precisely and to better understand the cellular basis of pharyngeal segmentation, we made time-lapse recordings of endodermal development using a transgenic line in which 700 bp of the her5 promoter drive egfp expression in the pharynx. her5 is the earliest known selective marker of pharyngeal cells (Bally-Cuif et al.,2000), and this –0.7her5:egfp transgene (hereafter referred to as her5:egfp) faithfully reproduces expression in the pharyngeal endoderm beginning at 13-14 hpf, with minimal expression in the nervous system (Fig. 2A)(Tallafuss and Bally-Cuif,2003). Expression persists in the forming pouches and medial endoderm (Fig. 2B,D). Confocal imaging of her5:egfp expression between 15-36 hpf revealed that lateral endodermal cells become elongated mediolaterally (16-18 hpf) to form the primordia of the first and second pouches(Fig. 2C). These primordia consist of groups of six to ten cells in which the most lateral cells extend filopodial extensions to contact the surface ectoderm (arrow in Fig. 2E,F). her5:egfpalso marks continuous bands of more medial endodermal cells that connect two adjacent pouches along the AP axis, which surround the unlabeled mesenchyme of each pharyngeal arch. The first two pouches contact the ectoderm almost simultaneously between 20 and 22 hpf (Fig. 2E,F), the third and fourth pouches at 24 and 28 hpf, respectively(Fig. 2G), while the remaining pouches (5-6) form at roughly 3 hour intervals thereafter(Fig. 2I).

To correlate the time at which a pouch contacts the ectoderm more directly with the specification of neural progenitors in the adjacent placode, pouches were imaged with confocal microscopy in individual her5:egfptransgenic embryos, and then immediately fixed and analyzed for neurod expression by in situ hybridization(Fig. 2H,J). In most cases(17/20) the number of neurod+ ganglia closely matched the number of fully formed pouches. This suggests that neurogenesis in the ectoderm occurs within a few hours of endodermal contact, consistent with a direct interaction.

Pharyngeal endoderm promotes neurogenesis in ectodermal explants(Begbie et al., 1999) but whether or not it is required in vivo remains unclear. Thus, we examined markers of epibranchial neurons in casanova (cas, sox23) and van gogh (vgo, tbx1) mutants, which are defective in endoderm. cas mutants lack all endoderm at 48 hpf, including the pharyngeal endoderm, as shown by immunolabeling with the zn-8 antibody, which recognizes DM-GRASP (Fig. 3A,B). To determine whether epibranchial sensory neurons differentiate in the absence of endoderm we used the α-Hu antibody. cas mutants also lack virtually the entire epibranchial nervous system at 48 hpf, including gVII, gIX and the distal portions of gX_1-4 (Fig. 3C,D). Consistent with a defect in placodal neurogenesis, mutants also lacked any expression of neurod or phox2b in these ganglia(Fig. 3E,F). By contrast,neurons derived from dorsolateral placodes, including the trigeminal (gV) and posterior lateral line (gP) ganglia, as well as the proximal portion of the vagal (gX), appeared to be unaffected in cas mutants, and some additional scattered Hu-immunoreactive cells were detected ventral to gP(Fig. 3D). To determine whether earlier placodal specification is disrupted in cas mutants, we analyzed expression of foxi1, which marks the placodal field during somitogenesis (Fig. 6)(Lee et al., 2003; Nissen et al., 2003; Solomon et al., 2003). No defects were detected in embryos derived from cas^+/–heterozygotes at 20 hpf (Fig. 6A,B; n=50), but, by 48 hpf, foxi1 expression was no longer detected in either the epibranchials or in the pouches in cas^–/– mutants(Fig. 6C,D). These results suggest that endoderm is required specifically for neurogenesis in epibranchial, and not dorsolateral, placodes.

In contrast to cas, vgo mutants form endoderm but their pouches are severely disorganized. Mutants often lack posterior pouches or they appear to fuse along the AP axis, but no defects in cranial ganglia have been reported (Piotrowski and Nusslein-Volhard,2000). Our analysis of phox2b expression, however,revealed consistent defects in epibranchial ganglia in vgo(Fig. 3G,H). Subsets of ganglia were missing or fused along the AP axis in many mutants (92%, n=26),and these defects were not always symmetrical on the left and right sides of the pharynx, similar to the variable pouch deformities in mutants. Surprisingly, in many vgo mutants the ganglia appeared to be enlarged, and extended further ventrally than in wild-type siblings(Fig. 3G,H). To correlate pouch and placodal defects more directly, we analyzed vgo mutants carrying the her5:egfp transgene and raised them individually for analysis of phox2b expression. Consistent with our analyses in wild type at 24 hpf (Fig. 3I), we found a close correlation between enlarged or reduced contacts made by pouches with the ectoderm and epibranchial defects observed in vgo(Fig. 3J). Mutant embryos in which a pouch failed to make this contact showed specific loss of the adjacent ganglion, while in other cases pouches appeared to form larger regions of contact, and these invariably correlated with the presence of more phox2b+ neurons (100%, n=14). This further supports the idea of a contact-dependent signal from the lateral, pouch-forming endoderm that controls the spatial extent of neurogenesis in the overlying ectoderm.

Wild-type cells can form endoderm when transplanted into a casmutant, because cas acts downstream of Nodal signaling(Aoki et al., 2002a; Aoki et al., 2002b). Using this approach, we tested whether the reintroduction of endoderm into casmutants was sufficient to restore epibranchial formation. Cells transplanted into the blastula margin in zebrafish contribute to the mesodermal layer, but rarely to the endoderm. However, we can drive cells into the endoderm and rescue the formation of pouches by injecting the donors with an activated form of the Nodal receptor TaramA (tar*) (David et al., 2002). Injections of cas mRNA itself also drives donor cells to the endoderm, demonstrating that this is not simply due to the presence of tar* (Fig. 4B). cas or tar* RNA was co-injected into wild-type donor embryos together with rhodamine and biotin-conjugated dextrans (10,000 M_r;Molecular Probes) as lineage tracers at the 1-cell stage, and cells were transplanted at early blastula stages into host embryos derived from two cas^+/– heterozygotes(Fig. 4A). We then analyzed phox2b and neurod expression at 40 hpf and compared it with the locations of grafted cells. Transplants of this type form clones of endoderm in various positions, including the pouches(Fig. 4 C-E). Injection of cas or tar* alone caused no defects in epibranchial ganglia(Fig. 4B; Fig. 7A).

Grafting of cas- or tar*-injected wild-type cells into cas mutant hosts efficiently restored phox2b (and neurod) expression in epibranchial placodes wherever endodermal cells formed pouches (Fig. 4C-E;100%; n=29; Table 1). No rescue was observed when transplanted cells were located medially in the endoderm or outside of the pharynx. Confocal imaging of several grafts,revealed a close correlation between the organized movements of endodermal cells into pouches and the locations of phox2b+ cells, further suggesting that pouch formation in these mosaics is crucial for rescue. These results demonstrate that the restoration of endodermal pharyngeal pouches in cas mutant embryos rescues epibranchial development.

BMP7 has been shown to induce Phox2b expression in chick explants of pharyngeal ectoderm (Begbie et al.,1999) but it is not known: (1) whether BMP signaling is required for this process in vivo; or (2) whether BMPs expressed in endoderm control the spatial localization of placodal neurogenesis. In order to address these issues, we first examined the expression patterns of several zebrafish relatives of BMP7 in the pharyngeal pouches, including bmp2a, bmp2b,bmp4 and bmp7. All but bmp7 show expression in the pouches during the stages of epibranchial placode formation. At 22 hpf, bmp2a, bmp2b and bmp4 are expressed in the lateral pouches,near the junction between the endoderm and placodal ectoderm(Fig. 5A-C). By 30-40 hpf, bmp2a and bmp2b are expressed throughout the pouches,whereas bmp4 remains more distally restricted near the epibranchials(Fig. 5E-G).

In addition, we identified a zebrafish bmp5 gene that has not been described previously and that is also expressed in the pouches(Fig. 5D,H). Amino acid identities of mature bmp5 are 84% to mouse and human BMP5,67% to zebrafish bmp7, 74% to human BMP7, 54% to zebrafish bmp4, and 52% to bmp2a and bmp2b. Furthermore, bmp5 is not expressed during gastrula or early segmentation stages,unlike its close zebrafish relatives bmp2b and bmp7, which are required for early dorsoventral patterning. Expression is first detected in sensory patches within the otic placode, and in the pharyngeal region at 22 and 30 hpf (Fig. 5D,H). Pharyngeal expression of bmp5 includes both arch mesenchyme and pharyngeal pouches, with expression becoming restricted to the dorsal- and ventral-most pouches by 30-40 hpf (Fig. 5H,L).

Consistent with their endodermal defects, expression of all four BMP family members is reduced in cas mutants by 40 hpf(Fig. 5I-P). Expression is still present but the pattern is variably disorganized in vgo mutants(data not shown). Thus, several BMPs are expressed in the pouches at the appropriate place and time to influence epibranchial neurogenesis, and loss of expression correlates with epibranchial defects in mutants.

As Bmp7 plays a role in epibranchial formation in the chick, we examined epibranchials in the bmp7 zebrafish mutant snailhouse(snh) (Dick et al.,2000). Analysis of foxi1 at 20 hpf revealed variable reductions or loss of the early placodal field and pouches(Fig. 6E,F). By 48 hpf, snh mutants displayed a severe reduction in epibranchial neurons as determined by the expression of neurod(Fig. 6G,H). However snh mutants have severe defects in DV patterning and gastrulation that may secondarily cause these defects in neurogenesis.

Therefore, to test more local requirements for BMPs in epibranchial development, we blocked BMP signaling focally in the endoderm by co-injecting cas or tar* mRNA and mRNA encoding the Xenopus form of the BMP inhibitor noggin (nog)(Smith and Harland, 1992),into wild-type donors and transplanting these cells into either wild-type or cas mutant hosts. As a control for activity, we showed that injection of 500 pg of nog mRNA at the one-cell stage was capable of dorsalizing zebrafish embryos, and we used similar or slightly lower amounts for co-injection with cas or tar* (hereafter referred to as cas/nog or tar*/nog). Such nog-injected transplanted cells contributed to pharyngeal pouches in equal numbers to cells injected with cas or tar* alone, and were generally located on only one side of the pharynx, leaving the contralateral side as an internal control(see Fig. 4). We found that cells co-expressing cas/nog or tar*/nog caused reductions in the number of phox2b-expressing cells in adjacent epibranchial ganglia on the same side in wild-type hosts at 48 hpf (37%, n=26; Fig. 7B-D; Table 1). Furthermore, similar transplants into cas mutants were completely unable to rescue epibranchial development, even when pharyngeal pouches were clearly restored on the transplanted side (93%, n=15; Fig. 7H,I; Table 1). These results indicate that BMP signaling is required locally to specify the locations of epibranchial placodes, and further suggest that the crucial source of the signal is the endoderm.

To address which specific BMPs are required in epibranchial induction, we used a similar approach, but in this case donors were injected with antisense morpholino oligonucleotides (MOs) to disrupt the expression of zebrafish bmp2a, bmp2b, bmp4, bmp5 and bmp7, either individually or in combination. All of these except for zebrafish bmp7 are expressed in the pouches during epibranchial induction (see Fig. 5). We co-injected morpholinos together with cas or tar* RNA to create endodermal donors lacking the functions of individual BMPs or combinations thereof (hereafter referred to as cas/bmp-MO or tar*/bmp-MO). To control for morpholino effectiveness, we injected amounts into the donors that produce a complete loss-of-function phenotype for the bmp2b-MO (300 pg per embryo phenocopies the swirl mutant) (Imai and Talbot, 2001)and bmp7-MO (800 pg per embryo phenocopies the snailhousemutant) (Imai and Talbot, 2001). Only the combination of both bmp2b-and bmp5-MOs into donors consistently caused reductions in epibranchials adjacent to transplanted pouches in wild-type hosts (62%, n=26; Fig. 6E,F; Table 1) and completely eliminated the ability of these cells to rescue epibranchials in casmutants (n=19, Fig. 7K,L; Table 1). cas or tar* mRNA + bmp2b-MO or bmp5-MO alone slightly reduced the ability of transplanted endodermal cells to rescue epibranchial development in cas mutant hosts (64% rescued; n=13; Fig. 7J), and had no effect when transplanted into wild types(Table 1). Finally, cas or tar* mRNA + bmp2a-, bmp4- or bmp7-MOs had little to no effect on epibranchial development (data not shown). These results suggest that bmp2b and bmp5 are both required within the endoderm for epibranchial induction. Based on their overlapping patterns of expression, these two BMP genes may function redundantly with one another,as well as with bmp2a and bmp4.

To target the stage at which epibranchial induction occurs more precisely,we implanted beads coated with recombinant human BMP or NOG proteins, and assayed the pattern of phox2b expression in neural progenitors(Fig. 8). Beads approximately 70-100 μm in diameter, soaked in either 1 μg or 10 μg/ml BMP4(similar results were obtained with 100 μg/ml BMP5 and BMP7), were placed anterior to the otic vesicle on one side of the head at 20 hpf, embryos were then raised to 48 hpf for phox2b analysis by in situ hybridization. Control embryos that received beads soaked in BSA-PBS showed no changes in the pattern of phox2b expression (Fig. 8B). By contrast, embryos that received beads soaked in 10μg/ml BMP4 showed ectopic phox2b+ neural progenitors in the vicinity of the bead (53%; n=29; Fig. 8C; Table 2), whereas beads soaked in 1 μg/ml BMP4 produced a much weaker effect (8%; n=36; Fig. 8D; Table 2). This effect was very local, occurring only when beads were placed in the branchial region, and ectopic epibranchial induction did not spread into ectoderm outside of the arches. Conversely, beads soaked in 500 μg/ml recombinant NOG and implanted into this region consistently showed a reduction in epibranchials on the implanted side (92%, n=12; Fig. 8E,F; Table 2). A slightly weaker effect was observed with beads soaked in 100 μg/ml NOG(70%; n=10; Fig. 8D; Table 2). These results suggest that at 20 hpf, as pharyngeal pouches are beginning to contact the surface,pharyngeal ectoderm is competent to respond to BMPs and form neurons. They also show that BMP signaling is required within the pharyngeal region at this stage for epibranchial induction.

In this study, we have shown that signals from the endoderm in zebrafish are essential for induction of the epibranchial subdivision of the peripheral nervous system, and that BMPs are one component of this signaling. Epibranchial ganglia develop as ectodermal placodes adjacent to each of the pharyngeal pouches, serially reiterated outgrowths of the endodermal wall of the pharynx that are the first morphological signs of pharyngeal segmentation(Goodrich, 1930). Our experiments suggest that contact between a pouch and the adjacent ectoderm is essential to specify the locations of these placodes and to promote neurogenesis, and that the induction in zebrafish is mediated in part by bmp2b and bmp5 (Fig. 9). Little is known about the signals that induce the formation of other primary sensory neurons in vertebrates, but these may involve similar interactions.

Several lines of evidence support this hypothesis. First, each pharyngeal pouch contacts the ectoderm within an hour or two of the onset of neurogenic gene expression in the adjacent placode. Second, endodermal defects in cas or vgo mutants correlate with placodal defects. Third,endodermal grafts locally rescue neurogenic gene expression in epibranchial placodes in cas, demonstrating that the loss of placodes in this mutant is due to loss of endoderm and not defects in neural crest cells or other cell types in the vicinity of the epibranchials. This promotion of epibranchial development by underlying endodermal pouches serves to coordinate the pattern of cranial ganglia with the pharyngeal arches that they innervate.

Our results support those of Begbie et al.(Begbie et al., 1999), who performed explant studies in chick and showed that Phox2b+ neural progenitors form in ectodermal explants when co-cultured with endoderm in collagen gels. From these studies, however, it remained unclear whether such interactions occur in vivo, or are necessary for neurogenesis. We have used genetic ablation and reintroduction of endoderm in zebrafish to demonstrate an in vivo requirement. Whether or not this induction is direct, or requires an intermediate signal via neighboring tissues remains unclear. However, the close spatial proximity between pharyngeal pouches and epibranchial placodes within an arch, as well as the coincidence between early epibranchial defects in cas and vgo mutants, supports the idea of a direct signal(Fig. 9). Because these endodermal mutants only show defects in epibranchial and not dorsolateral placodes, our results also support the model proposing that these two groups of sensory neurons are induced by different signals(Graham and Begbie, 2000). This would also help to explain the presence of additional placodes associated with the ventral regions of the pouches in some species, such as the hypobranchial ganglia in Xenopus(Schlosser, 2003).

To visualize this interaction between the pouches and placodes, we followed the expression of her5:egfp in the pharyngeal endoderm in living embryos (Tallafuss and Bally-Cuif,2003). Confocal time-lapse analysis of her5:egfpexpression revealed dramatic and rapid changes in endodermal cell shape during pouch morphogenesis. Lateral endodermal cells of the pharynx align mediolaterally to form pouches and extend filopodia toward the surface,eventually contacting the overlying ectoderm. This contact occurs at almost exactly the same stage at which we could detect the first signs of epibranchial neurogenesis, with contact occurring a few hours prior to the onset of neurod expression in most cases. These results further reinforce the idea that pouches regulate both the spatial and temporal formation of neural progenitors in the placodes. Consistent with this hypothesis, we show that ectopic epibranchials in vgo mutants are associated with abnormal extensions of pouches(Piotrowski and Nusslein-Volhard,2000). In addition, ectopic neurod expression was induced by endodermal transplants into cas mutants that formed ectopic contacts with the ectoderm. These results indicate that a broader region of pharyngeal ectoderm is competent to form placodal neurons, but only does so in response to endoderm.

Many cranial sensory ganglia have a dual embryonic origin from both neural crest and placodes, in contrast to those of the trunk, which are purely crest derived. Crest cells form the proximal epibranchial ganglia in chick, whereas placodally derived cells lie further distally, and the proportions of these differ in each ganglion (Ayer LeLievre and LeDouarin, 1982; D'Amico-Martel and Noden,1983; Webb and Noden,1993). Placodally derived neurons differentiate early and establish the first peripheral and central axonal projections of the sensory nerves. Epibranchial neurons also are displaced inwards along neural crest migratory pathways as they mature, and require the crest to establish their appropriate innervation in the hindbrain(Begbie and Graham, 2001). Our analyses of gene expression within the epibranchial ganglia in zebrafish confirm that there is a similar proximodistal sequence of neurogenesis within the placodally derived neurons in zebrafish, in which newborn neurod+neurons form adjacent to the pharyngeal pouch and older phox2b and Hu+ neurons lie further proximally (Fig. 8).

Both cas and vgo mutants disrupt distal neurogenesis in the vagal ganglion complex (gX_1-4), but retain a more proximal population of neural progenitors that develops independently of endodermal influences. These may be the neural crest-derived equivalent in fish of the proximal nodose ganglion (gX) in chick. Consistent with this hypothesis,proximal gX is less affected than other epibranchial ganglia in Ngn2^–/–(Fode et al., 1998) and Phox2a^–/–(Morin et al., 1997) mutant mice, as well as in foxi1^–/– mutant zebrafish(Lee et al., 2003). Future cell tracing studies are necessary to determine which portions of these ganglia are derived from neural crest in zebrafish, but our previous lineage studies suggest that at least some of the proximal neurons are crest derived(Schilling and Kimmel, 1994). Our results suggest that these proximal epibranchial neurons are BMP-independent or, alternatively, may be induced by BMPs from other sources,such as the otic vesicle.

We also show a requirement for BMP signaling in the control of epibranchial neurogenesis, and specifically implicate bmp2b and bmp5 in this process in zebrafish. This is based on several lines of evidence: (1) bmp2b and bmp5 are expressed in the pouches as they form;(2) targeted knockdown of bmp2b and bmp5 expression with morpholinos specifically in the endoderm locally disrupts placode induction;and (3) the BMP inhibitor NOGGIN locally disrupts epibranchial ganglia while misexpression of BMP proteins using beads induces ectopic neurons. Begbie et al. (Begbie et al., 1999)showed that Bmp7 in the chick is expressed in the pharyngeal endoderm and that recombinant BMP7 protein was sufficient to induce neurogenesis in ectodermal explants. They also found that Follistatin blocks the induction of neurogenesis by endoderm in these cultures, further indicating a role for TGFβ signaling in this process. Our results are the first to demonstrate a specific requirement for BMP signaling in vivo, and suggest that multiple members of this family, including bmp2b and bmp5, are essential.

Bmp4 (a close relative of bmp2b) and Bmp7 have been implicated in many aspects of neurogenesis and peripheral nervous system(PNS) development. For example, they mediate interactions between the neural plate and epidermal ectoderm that induce neural crest cells, as well as secondary sensory neurons within the spinal cord(Liem et al., 1995). Later,the same two BMPs are secreted by the dorsal aorta and promote a subset of adjacent neural crest cells to express Phox2a and to form sympathetic neurons (Reissmann et al.,1996). Yet, Bmp4^–/– mutant mice die prior to PNS formation and no PNS defects have been described in Bmp7^–/– mutants. Our results may help to explain this apparent discrepancy by showing that at least two BMPs, bmp2b and bmp5, are partially redundant for this process.

Other signaling molecules have been implicated in neurogenic placode induction, and it is important to understand how their functions relate to those of BMPs. Foremost among these are the FGFs. For example, signals from the MHB (possibly FGF8) are thought to induce the trigeminal (gV), a dorsolateral placode that we have shown does not require endodermal BMP signals in zebrafish. FGF3 and FGF8 produced in the hindbrain are involved in induction of the otic placode (Phillips et al., 2001), which also develops independently of endoderm. Both fgf3 and fgf8 are expressed in both the pharyngeal endoderm and ectoderm, and potentially act together with BMPs in epibranchial induction(David et al., 2002; Crump et al., 2004b). However,in the chick Fgf8 is only expressed in a ventral, posterior domain within each pouch, which abuts the domain of Bmp7 and is not in contact with the placode (Graham and Begbie, 2000), suggesting that Fgf8 and Bmp7interact to define territories of expression within the endoderm. Recent evidence in zebrafish has implicated fgf3 in epibranchial development(D. Raible, personal communication). Thus, it will be interesting to determine whether Fgfs act synergistically with BMPs in this interaction, or if different growth factors give qualitatively different responses in placodal cells. BMP2 and FGF act synergistically to induce neuronal differentiation of PC12 cells (Hayashi et al.,2001). The augmentation of FGF-induced differentiation by BMP2 occurs through the upregulation of FGFR1 several hours after BMP2 expression(Hayashi et al., 2003). Similarly, BMPs might also commit epibranchial progenitor cells to neuronal differentiation induced by FGFs. The combined actions of different growth factors may underlie differences not only between types of placode (e.g. FGF in dorsolaterals, BMP in epibranchials), but also in the specification of the distinct types of neurons that form in each ganglion within a class(Vogel and Davies, 1993).

Segmentation of the foregut into pouches is a fundamental feature of the head and there is growing evidence that this plays a crucial patterning role. The segmental characteristics of the pharyngeal pouches develop independently of the presence of neural crest cells(Veitch et al., 1999). Pharyngeal slits are also found in non-vertebrate chordates, such as amphioxus, suggesting that their appearance predated that of placodes or neural crest during evolution. The formation of pouches during embryogenesis coincides with and affects the segmental development of many cell types within the pharyngeal arches, including both neurogenic placodes and neural crest(Begbie et al., 1999; LeDouarin, 1982; David et al., 2002; Couly et al., 2002). Endoderm may provide some guidance cues for neural crest migration; however, most evidence suggests that it plays a more important role later, in local interactions between pouches and immediately adjacent neural or skeletal progenitors, and our results are consistent with this. The size, shape and orientation of neural crest-derived cartilages are prefigured in the shapes of certain pouches, both in fish (Crump et al., 2004a; Crump et al.,2004b) and in chick (Couly et al., 2002). Likewise, we have shown that the location and size of epibranchial sensory ganglia are prefigured by contacts between pouches and the surface ectoderm. Defects in these endoderm-dependent processes appear to underlie human craniofacial malformations such as DiGeorge syndrome (often caused by mutations in TBX1), and our results would suggest that some cranial sensory nerve deficits in humans might also reflect defects in endoderm. Our studies establish a genetic context in zebrafish in which to now examine how these signals control sensory neurogenesis.

We thank I. Blitz for the Xenopus noggin expression construct. We also thank A. Graham, I. Blitz and members of the Schilling Laboratory for critical reading of the manuscript. This work was supported by the NIH(NS-41353, DE-13828), March of Dimes (1-FY01-198) and Pew Scholars Foundation(2615SC) to T.F.S., and a VolkswagenStiftung Junior Research Group Grant to L.B.-C.