Noradrenergic neurons in the zebrafish hindbrain are induced by retinoic acid and require tfap2a for expression of the neurotransmitter phenotype

Noradrenergic neurons of the brain stem play a major role in modulating sensory, motor and arousal tone in vertebrates. Several evolutionary conserved noradrenergic (NA) centers exist in the vertebrate hindbrain(Smeets and Gonzalez, 2000). In the rostral hindbrain, NA neurons located ventrolaterally in the first rhombomere form the locus coeruleus (LC). In the posterior hindbrain, NA neurons are more loosely distributed in the medulla and area postrema. To adopt a given neurotransmitter phenotype, developing neurons begin to express enzymes to synthesize a particular neurotransmitter, as well as other proteins for neurotransmitter storage, degradation or re-uptake. Noradrenergic (NA)neurons synthesize noradrenaline from tyrosine in a three-step process in which tyrosine hydroxylase (TH), L-aromatic amino acid decarboxylase (L-AADC)and dopamine beta-hydroxylase (DBH) act sequentially.

Several factors that control development of NA centers have been identified. Two closely related homeodomain proteins, Phox2a and Phox2b, have been shown to be crucial for the development of the LC(Pattyn et al., 1999; Pattyn et al., 2000b). In contrast to Phox2a, which controls development of NA neurons in the LC, Phox2b is required for differentiation of all NA centers in the brain(Pattyn et al., 2000b). Phox2a and Phox2b are also directly involved in the development of autonomic neurons and hindbrain motoneurons (Valarche et al., 1993; Tiveron et al.,1996; Pattyn et al.,1997; Pattyn et al.,2000a). The proneural protein Mash1 is required for most NA neurons (Hirsch et al., 1998),and has been shown to act upstream of the Phox2 genes(Hirsch et al., 1998; Pattyn et al., 2000b). Recently, another homeodomain-containing protein, Tlx3, has been shown to control the development of nearly all NA neurons in the hindbrain independently of Phox2b (Qian et al.,2001). The transcriptional activator Tfap2a(Williams et al., 1988; Mitchell et al., 1991) is expressed in hindbrain NA neurons. Tfap2a has been shown to bind to the th and dbh promoters, and in cell culture assays may activate expression of the key enzymes for development of the NA neurotransmitter phenotype, TH and DBH(Kim et al., 2001). Inactivation of the Tcfap2a gene in mouse revealed a critical function of Tfap2a for development of the neural tube, craniofacial structures and eyes (Schorle et al.,1996; Zhang et al.,1996; Nottoli et al.,1998). The defects in neural tube closure in Tcfap2amutant mice prevented adequate assessment of the Tcfap2a mutant phenotype during neuronal differentiation. Thus, a potential role of Tfap2a during NA differentiation in vivo has not been previously investigated.

We show that mutations isolated as mont blanc alleles[mob (Neuhauss et al.,1996); allelic with lockjaw(Schilling et al., 1996)]affect zebrafish tfap2a. tfap2a mutant embryos develop defects in neural crest and epibranchial placode derivatives. Analysis of the development of the sympathetic and central nervous systems in tfap2a mutants revealed loss of th expression in the LC, medulla and area postrema,and sympathetic ganglia. Thus, neurons in the hindbrain and peripheral nervous system do not express the noradrenergic transmitter phenotype in the absence of tfap2a function. Furthermore, we show that retinoic acid (RA) is a necessary inducer of NA cell fate in the hindbrain in vivo. necklessmutant embryos, which have a disrupted raldh2 gene and are thus deficient in RA synthesis (Begemann et al.,2001), lack NA neurons in the medulla oblongata. Moreover,exogenously applied retinoic acid can upregulate tfap2a expression and induce ectopic NA neurons in both the hindbrain medulla and the sympathetic nervous system. By contrast, in tfap2a mutant embryos RA fails to induce ectopic th expression in the hindbrain or sympathetic nervous system. Taken together, our results provide evidence that the neurotransmitter phenotype of hindbrain NA neurons is regulated by tfap2a. Our data further reveal distinctive control mechanisms for NA neurons in the LC and medulla: Although LC neurons are specified through a pathway including Fgf8 and phox2a(Guo et al., 1999a), neurons of the medulla are induced by RA, which we have identified as acting upstream of tfap2a.

Zebrafish maintenance and breeding was under standard conditions at 28.5°C (Westerfield,1995). Embryos were staged and fixed at the desired time points(hpf or dpf, hours or days post fertilization)(Kimmel et al., 1995). To avoid formation of melanin pigments, embryos were incubated in 0.2 mM 1-phenyl-2-thiourea (Sigma). The following mutant zebrafish strains were used in this study: neckless (nlsⁱ²⁶)(Begemann and Meyer, 2001) and soulless (sou^m810)(Guo et al., 1999a; Guo et al., 1999b). The craniofacial defects in embryos mutant for the tfap2a alleles mob^m610/780 and lockjaw(low^ts213) have previously been described(Neuhauss et al., 1996; Schilling et al., 1996). A new tfap2a allele, mob^m819, was isolated during a mutagenesis screen performed in our laboratory. G0 males (AB strain) were mutagenized with ethyl-nitrosourea (ENU) as described(Solnica-Krezel et al., 1994). F1 progeny were grown from G0 males crossed to AB strain females. Individual F1 females were used to generate haploid F2 progeny. F2 embryos were fixed at 24 hpf or 60 hpf and screened by whole-mount in situ hybridization for expression of tyrosine hydroxylase (th)(Holzschuh et al., 2001). Egg clutches in which about half of the embryos developed abnormalities in the catecholaminergic systems were identified by screening using a dissecting microscope. Strains were established from the corresponding F1 females and the diploid phenotype investigated in F3 embryos. Among several strains characterized during the screen, one contained mob^m819.

To simplify cross-reference with the mammalian literature, we will use the approved UCL/HGNC/HUGO Human Gene Nomenclature database symbol tfap2afor the zebrafish gene, instead of a gene name derived from the allele designations of the initial isolates mont blanc mob^m610/780,mob^m819 and lockjaw (low^ts213). The use of the gene name tfap2a for the zebrafish gene will also avoid confusion with other genes that have been named AP2 (e.g. adaptor protein 2).

We mapped mob^m819 using bulked segregant analysis(Michelmore et al., 1991) with pooled DNA from homozygous mutants and siblings from a mob^m819 AB × India map-cross. We linked mob^m819 to linkage group 24 between SSLP markers Z23011 and Z13695. Linkage with tfap2a was determined by RFLP analysis of PCR products from DNA of single mob^m819 mutant and wild-type sibling embryos, obtained using primers flanking exon 5. DraII restriction digest of the amplified tfap2a products did not digest the 360 bp PCR product from wild-type embryos, but cleaved that of mob^m819 mutant embryos into two fragments of 123bp and 237bp. Analysis of the mob^m610 and mob^mm780 alleles reported previously(Neuhauss et al., 1996)revealed that they are likely re-isolates of one allele, as both represent the same mutational event: a point mutation in the splice acceptor at the 5′end of exon 7 resulting in the use of a cryptic splice acceptor 14 bp further 3′, which causes a frame shift in the DNA-binding and dimerization domain, and a truncation at amino acid 395 (A.B. and E.K., unpublished).

Sequencing the EST fc31a07 revealed that it contained the full-length ORF of tfap2a isoform 3. Using gene specific primers for this EST, we performed RT-PCR with mRNA from mob^m819 mutant and wild-type embryos to obtain the full-length ORF. Total RNA was isolated using the RNAeasy kit (Qiagen) and RT-PCR was performed with the RevertAid cDNA synthesis kit (MBI Fermentas). PCR products were cloned into pCR4 vector using the TOPO TA kit (Invitrogen). Sequencing was carried out on an automated sequencer (MWG). The full-length sequence of tfap2a isoform 3 has been submitted to GenBank (Accession Number AY166856).

Embryos from wild-type or heterozygous mob^m819 parents were exposed to serial dilutions (3×10^–8 to 3×10^–6 M in egg water) of all-trans retinoic acid (RA,Sigma; stock solution 3×10^–1M in DMSO) for defined time periods in the dark. After treatment, embryos were washed under safety red light three times each for 10 minutes with egg water(Westerfield, 1995) to remove RA, and fixed at desired stages. For th expression analysis zebrafish embryos were exposed to RA for 9 hours between 24 hpf and 33 hpf, or for 24 hours between 24 hpf and 48 hpf. Hox gene expression analysis was performed on zebrafish embryos treated with RA for 9 hours between 24 hpf and 33 hpf.

Cycloheximide (CHX, Calbiochem; stock of 10 mg/ml in 100% ethanol) was applied at a concentration of 100 μg/ml in egg water. Embryos were exposed to CHX from 21 hpf to 32 hpf. For combined CHX and RA treatments, RA was added at 24 hpf to the embryos (final concentration of 3×10^–7M RA) and incubated for 8 hours. At 32 hpf, the embryos were washed as described above and fixed at desired stages. Controls for each experiment included sibling embryos incubated in the same concentration of Ethanol and/or DMSO as the experimental embryos.

Whole-mount in situ hybridization (WISH) was performed according to Hauptmann and Gerster (Hauptmann and Gerster, 1994). Digoxigenin- or fluorescein-labeled antisense RNA probes were prepared using RNA labeling reagents (Boehringer).We prepared RNA probes for the following genes: th(Holzschuh et al., 2001); phox2a (Guo et al.,1999a); hoxa2, hoxb2, hoxb3 and hoxd3(Prince et al., 1998); isl1 (Inoue et al.,1994b); and ret1(Bisgrove et al., 1997). To generate a probe specific for dopamine beta hydroxylase(dbh) expression, we used a PCR-based approach to combine genomic fragments (Sanger Centre, Hinxton, UK) which contain Dbh-coding region, as judged from the sequence two genomic fragments (z35723-a1914e04.q1c and zfishC-a1746c09.p1c) in public databases and from comparison with mammalian Dbh sequences. To generate the probe containing 735 bp of the dbhcoding sequence, we linearized with SpeI and transcribed with T7 polymerase. To make tfap2a probes, the EST fc31a07 was linearized with SalI (MBI Fermentas) and RNA probe was transcribed with Sp6 RNA polymerase (Boehringer). Immunohistochemistry anti-serotonin antibodies(Chemicon), anti HuC/HuD antibodies (16A11, Molecular Probes), and anti-acetylated tubulin antibodies (Sigma) was performed as described(Solnica-Krezel and Driever,1994), except that fixed embryos older than 3 dpf were permeabilized by 30 minutes incubation in proteinase K (Sigma; 10 μg/ml)and successive inhibition of proteinase K with phenylmethyl-sulfonylfluoride(PMSF, Sigma). The cartilage of zebrafish embryos was stained with Alcian Blue as described elsewhere (Neuhauss et al.,1996).

Mutations of the mont blanc (mob) complementation group were initially isolated based on their prominent craniofacial phenotype(mob^m780/610)(Neuhauss et al., 1996). We isolated the new mob^m819 allele in a haploid genetic screen for mutations affecting the catecholaminergic systems (CA) in zebrafish. All three alleles are characterized by absence of derivatives of the second pharyngeal arch (ceratohyal, basihyal cartilages) and strongly reduced gill arch derivates (basibranchial, ceratobranchials; Fig. 1A,B, Fig. 4I,J). As mob^m610/780, low^ts213 and mob^m819 map to the same region of LG24 (this paper) (A.B. and E.K., unpublished) (Knight et al.,2003), we carried out complementation tests with all three mutations. Both alleles from the previous screens failed to complement with mob^m819 (data not shown).

The main CNS neuronal phenotype of mob^m819 mutant embryos affects the noradrenergic (NA) centers in the hindbrain(Fig. 1C-P). The NA defects of mob^m610/780 and mob^m819 are similar,indicating a similar allelic strength for these alleles (data not shown; A.B. and E.K., unpublished). Analysis of th expression in mob^m819 mutant embryos revealed that in areas of the hindbrain where NA neurons normally develop, the locus coeruleus (LC) and the medulla, as well as in the PNS, fully differentiated th-expressing NA neurons could not be detected at 54 hpf(Fig. 1C-J). By contrast, th expression in the forebrain appears normal. We cloned zebrafish dopamine beta hydroxylase (dbh) to specifically label noradrenergic neurons. At 4 dpf, expression of dbh can be detected in the locus coeruleus, the medulla oblongata/area postrema, sympathetic neurons,as well as a group of arch associated neurons that may later contribute to the carotid body (Fig. 1K,M,N; a detailed analysis of dbh expression will be presented elsewhere). We analyzed expression of dbh in mob^m819 mutant embryos at 4 dpf and found that dbh is not expressed in the locus coeruleus (Fig. 1L,O) and the enteric region (Fig. 1L,P). By contrast, dbh is still expressed in the carotid bodies of mob^m819 mutant embryos(Fig. 1K). In the medulla oblongata/area postrema, a small number of cells express dbh at 4 dpf(Fig. 1K,O). The number of dbh-expressing cells in the medulla oblongata is strongly reduced when compared with wild type. Furthermore, the expression level of dbh in mutant medulla oblongata cells appears lower than in wild type(Fig. 1K,L,M.O; WISH procedure was performed identically on mutants and wild type in the same staining batch– which is confirmed by the fact that expression levels of dbhdetected in the carotid bodies are similar in mutant and wild type). We conclude that mob^m819 does affect th expression and noradrenergic differentiation in the hindbrain; however, in the medulla oblongata, other factors may later during development partially compensate for the absence of mob^m819 activity. To test whether other monoaminergic neurons are affected in mob^m819 mutants, we stained serotonergic neurons in the CNS with anti-serotonin antibodies. Serotonergic neurons appear normal in mob^m819 mutant embryos at 72 hpf (Fig. 1Q-T). The development of reticulospinal neurons of the hindbrain also appeared normal as revealed by backfill labeling with rhodamine-dextran (data not shown). Furthermore, the overall pattern of neural development appeared normal at 1 dpf, as judged from anti-acetylated tubulin immunohistochemistry (data not shown).

To identify the gene affected by mont blanc (mob)mutations, we mapped the mutation using simple sequence length polymorphism markers (SSLP). mob is located between the markers z23011 (2.04 cM from mob; 11 recombinants in 538 meiosis) and z13695 (2.75 cM from mob; 15 recombinants in 546 meiosis) on LG 24(Fig. 2A). Two EST clones map to this region: fc31a07 (W. Talbot, http://zebrafish.stanford.edu)and fb83f04 (I. Dawid, http://dir.nichd.nih.gov/lmg/lmgdevb.htm),both predicted to be homologous to mammalian Tfap2a (S. Johnson, AI584805,AI722745; WUZGR; http//zfish.wustl.edu). Tcfap2a mutant mouse embryos(Schorle et al., 1996; Zhang et al., 1996) develop craniofacial and PNS phenotypes similar to those of mob^m819 zebrafish mutants. Thus, tfap2a appeared to be a likely candidate gene for mont blanc mutations. As in mouse,the zebrafish tfap2a locus encodes three mRNA isoforms, each using different first exons. Zebrafish tfap2a isoforms tfap2a1 and tfap2a2 have been isolated (M. Fuerthauer, C. Thisse and B. Thisse,GenBank Accession Numbers AF457191, AF457192). We cloned and sequenced a full-length cDNA that encodes the zebrafish tfap2a3, a homolog of the mouse Tcfap2a isoform 3. The amino acid sequences of all three Tfap2a isoforms are highly conserved: isoforms 1 and 3 each share 85% amino acid identity between mouse and zebrafish, while isoform 2 shares 82% identity. The three zebrafish isoforms differ from each other only in the first 15 (isoform 1), 21 (isoform 2) or 9 (isoform 3) amino acids, which are encoded by three alternative first exons (Fig. 2). We assembled the genomic structure of tfap2a(Fig. 2B) through in silico chromosome walk on publicly available zebrafish genomic sequences(http://www.sanger.ac.uk/Projects/D_rerio). Sequencing of cDNAs from homozygous mob^m819 embryos identified a point mutation in exon V, close to the splice-acceptor site(Fig. 2C). This A to T transition created a DraII restriction site, which made it possible to confirm linkage to mob^m819 by RFLP analysis(Fig. 2C,D). We did not find a single recombinant in 546 meiosis, indicating that mob and tfap2a map within an interval smaller than 0.18 cM. The mutation introduces a stop codon that truncates the protein at amino acid 264, located at the N-terminal end of the dimerization and DNA-binding domain (DDB; Fig. 2C,E). This domain is necessary for the dimerization of Tfap2 proteins, which in turn is required for DNA binding of Tfap2a protein to the target promoters and the formation of a functional transcriptional activation complex(Williams and Tjian, 1991a; Williams and Tjian, 1991b). Therefore, mob^m819 is most likely a null allele. All three tfap2a isoforms are affected by the mutation.

We examined the spatiotemporal expression patterns of tfap2a by whole-mount in situ hybridization. Using a full-length probe, we found an expression pattern throughout development that resembles the known expression pattern of tfap2a1(http://zfin.org). tfap2a expression is first detected at gastrula stages in non-neural ectoderm (data not shown), and expression continues in some epidermal cells(Fig. 3D; a potential requirement in Xenopus non-neural ectoderm has already been described(Luo et al., 2002). The neural crest primordia express tfap2a from the one-somite stage on (data not shown). During the segmentation period tfap2a expression begins in the pronephric duct and in migrating neural crest cells(Fig. 3A,B). From late segmentation stages, tfap2a is also expressed in the CNS in two bilateral cell groups in the tegmentum and in bilateral stripes by hindbrain and spinal chord neurons (Fig. 3E,F). During these stages, the migrating cells of the lateral line primordium also express tfap2a(Fig. 3C). By 48 hpf,expression of tfap2a can be detected in the inner nuclear layer of the retina, the epiphyseal region, the pretectum and tectum, the tegmentum,the cerebellum, and two lateral stripes on each side of the hindbrain(Fig. 3G-I). The neuromasts of the lateral line organ continue to express tfap2a(Fig. 3J). From 3 dpf onwards, tfap2a can be detected in cells adjacent to the gut tube, which probably correspond to neurons of the sympathetic ganglia(Fig. 3K-M).

To analyze the cranial ganglia in tfap2a mutant zebrafish, we used expression of dlx2, ret1 (previously c-ret) and phox2a as markers. dlx2 is expressed in neural crest cells contributing to the branchial arches(Akimenko et al., 1994). In mouse embryos, the receptor tyrosine kinase Ret is expressed early together with Phox2a in derivatives of the epibranchial placode(Pachnis et al., 1993; Valarche et al., 1993; Robertson and Mason, 1995; Tiveron et al., 1996). The zebrafish homolog of Ret, ret1, is expressed in the branchial arches(Bisgrove et al., 1997; Marcos-Gutierrez et al.,1997), and we could also detect expression in cranial neural placodes (Fig. 4E). The closely related homeobox containing transcription factors Phox2a and Phox2b are expressed in epibranchial placodes and their derivatives, the VIIth, IXth and Xth cranial sensory ganglia (Tiveron et al., 1996; Pattyn et al.,1997; Fode et al.,1998; Begbie et al.,1999). Loss-of-function experiments in mouse have shown that Phox2a and Phox2b are required for the development of these ganglia, as well as for the development of NA neurons in the hindbrain(Morin et al., 1997; Pattyn et al., 1999).

Expression of dlx2 and ret1 is greatly reduced in tfap2a mutant embryos at 24 hpf(Fig. 4A-D). As in wild-type embryos, at this stage the neural crest and the placodal-derived cells migrate to their final positions. Later in development, when the sensory ganglia of the VIIth (facial), IXth (glossopharyngeal) and Xth (vagal) cranial nerves are forming, expression of ret1 and phox2a is not detectable above the branchial clefts in mutant embryos(Fig. 4E-H). However, the anterior and posterior lateral line ganglia express phox2a. We further analyzed tfap2a mutant embryos by immunohistochemistry with anti-acetylated tubulin antibodies to visualize axon tracts. In tfap2a mutant embryos, the early axon scaffold tracts of the CNS appear normal at 24 hpf (data not shown). By 4 dpf, however, the pattern of peripheral axons in the jaw and pharyngeal regions of mutant embryos is disturbed, suggesting that either the fibers of the VII and IX cranial nerves are absent, or that they fail to reach this region because craniofacial structures are disturbed (Fig. 4I,J).

The homeodomain transcription factor Phox2a is expressed by noradrenergic neurons in the locus coeruleus and is required for their development in mammals and zebrafish (Morin et al.,1997; Pattyn et al.,1997; Guo et al.,1999a). The phox2a gene is disrupted in zebrafish soulless (sou) mutant embryos, in which NA neurons of the LC do not form (Guo et al.,1999a). Tfap2a and tyrosine hydroxylase are co-localized in noradrenergic neurons in the embryonic mouse hindbrain(Kim et al., 2001). To test whether tfap2a and phox2a are co-expressed in zebrafish noradrenergic neurons of the hindbrain, we performed double in situ hybridizations with tfap2a and phox2a(Fig. 5A-C,E-J) or thprobes (Fig. 5D). phox2a and tfap2a are indeed co-expressed in the zebrafish LC (Fig. 5C,G,I). When the first cells in the posterior hindbrain differentiate into NA neurons between 36 and 48 hpf, the expression domains of phox2a and tfap2aoverlap in the ventral half of the hindbrain(Fig. 5E,F). tfap2aexpression also expands to more dorsal areas where NA neurons differentiate. Consistent with the fact that phox2a is not expressed in the part of the medulla/area postrema where NA neurons develop, we find normal NA differentiation in the medulla of soulless/phox2a mutant embryos(Fig. 5M,N).

To test whether absence of the LC in mob/tfap2a mutants might be due to changes in phox2a expression, we examined the expression of phox2a in mob/tfap2a mutant embryos. phox2a is expressed in the LC of mob/tfap2a mutant embryos at all developmental stages analyzed (Fig. 5K,L;also data not shown). To find out if tfap2a could be a downstream target of phox2a, we assayed tfap2a expression in soulless/phox2a mutant embryos. We could not detect any alterations in the hindbrain expression pattern of tfap2a (data not shown). Our data indicate that tfap2a and phox2a act in parallel to control NA identity in the LC.

Dorsal root ganglia (DRG), enteric and sympathetic ganglia are neuronal derivatives of trunk neural crest. The absence of th and dbhexpression from the intestinal tract of mob/tfap2a mutants could either reflect a differentiation defect resulting in absence of terminal differentiation markers, or could be caused by the mutation affecting neural crest precursor cells, and thus resulting in migration defects or reduced survival. To test whether precursors of trunk neural crest-derived peripheral neurons initially form, we analyzed Hu-antigen and isl1 expression. The anti-Hu antibody recognizes a subset of ELAV-related proteins expressed very early in neuron differentiation(Marusich et al., 1994a). At 72 hpf the presence of Hu-immunoreactive cells reveals that precursors of sympathetic ganglia as well as enteric neurons of the anterior intestine are present at similar cell numbers in wild-type and mob/tfap2a mutant embryos (Fig. 6A,B). In contrast, we did not detect enteric neurons colonizing the posterior part of the gut tube in mob/tfap2a mutant embryos(Fig. 6C,D). DRG and Rohon-Beard cells (RB; dorsal primary sensory neurons of lateral neural plate origin in close lineage relation with neural crest)(Artinger et al., 1999) develop in mob/tfap2a mutant embryos, thought at 48 hpf they appear slightly reduced in cell number (Fig. 6E,F; for further discussion of DRGs in mob see A.B.-G. and E.W.K., unpublished). The LIM homedomain containing transcription factor Isl1 is expressed during early differentiation of dorsal root and sympathetic ganglia, as well as other neurons (Ericson et al., 1992; Inoue et al.,1994a). In 72 hpf old mob/tfap2a mutant embryos cells express isl1 at the location where sympathetic neurons form in wild type (Fig. 6G,H). Furthermore, phox2a is expressed in mob/tfap2a mutant embryos in cells at the location where sympathetic neurons form in wild type(Fig. 6I,J). Taken together our data suggest that the lack of th and dbh expression for the noradrenergic transmitter phenotype is not caused by a failure of neural crest formation or migration, but reflects a block in terminal differentiation of sympathetic neurons.

Tfap2a is expressed in many embryonic structures that are influenced by the level of all-trans-retinoic acid during development(Luscher et al., 1989; Philipp et al., 1994; Xie et al., 1998). The expression of Tcfap2a has been shown to be modulated by retinoic acid treatment in cell cultures (Williams et al., 1988; Luscher et al.,1989), suggesting a link between RA and the regulation of Tcfap2a. Therefore we investigated whether RA signaling plays a role in NA neuron differentiation in the hindbrain. To assay the effects of RA, we incubated groups of zebrafish embryos with different, increasing concentrations of RA. As RA primarily affects anterioposterior patterning and establishment of the Hox code in the neural tube during gastrulation and until mid-somitogenesis (Holder and Hill,1991; Hill et al.,1995; Yan et al.,1998), embryo at these stages were not assayed. Exposure of zebrafish embryos to RA for 9 hours between 24 hpf and 33 hpf, or for 24 hours between 24 hpf and 48 hpf induced ectopic th-expressing cells in the hindbrain and the sympathetic nervous system, which we detected at 33 hpf, 48 hpf and 72 hpf (Fig. 7I,J,M,N;see Fig. 9E,G,I,K,M,O, data not shown). Induction of ectopic NA neurons was never found rostral to rhombomere 2, and we did not detect changes in the LC(Fig. 7E,F), or the DA neurons of the forebrain (Fig. 7A,B). At 33 hpf we counted 10.7±0.9 (mean±s.d.; n=3) th-expressing sympathetic cells in wild-type embryos, while we detected 23.7±3.4 (n=3) in embryos treated with RA between 24 hpf and 33 hpf. At 72 hpf we counted 44.1±4.5 (n=8) th-expressing sympathetic cells in wild-type embryos, and 116±16.2 (n=8) at 72 hpf in embryos treated with RA between 24 hpf and 33 hpf. To investigate effects of reduced RA levels during NA differentiation, we analyzed zebrafish neckless (nls)mutants, which are defective in retinaldehyde dehydrogenase type 2 (RALDH2), a key enzyme for RA synthesis. DA neurons in the forebrain and NA neurons in the LC of nls mutants are not affected by reduced RA signaling in nls mutants (Fig. 7C,G; also data not shown). By contrast, nls mutant embryos lack th-expressing neurons in the medulla/area postrema(Fig. 7K). Furthermore, the number of NA neurons in the sympathetic nervous system is strongly reduced in nls mutant embryos (Fig. 7O). At 72 hpf we detected only 14.6±6.0 (n=8)sympathetic th-expressing cells in nls mutants. To test whether tfap2a may mediate RA dependent induction of thexpression, we analyzed th expression in RA-treated tfap2amutant embryos. In tfap2a mutant embryos, RA does not lead to ectopic expression of th, and is not able to rescue the NA phenotype(Fig. 7H,L,P). Thus, tfap2a may be required downstream of RA to induce NA differentiation.

RA signaling is involved in hindbrain patterning and rhombomere specification (reviewed by Gavalas and Krumlauf, 2000; Gavalas,2002). To rule out the possibility that our conditions of RA treatment interfered with the specification of rhombomere identities, we analyzed the rhombomere-specific expression domains of several Hox genes and epha4 (Prince et al.,1998). Expression domains of hoxa2 in rhombomere 2-5(r2-r5), hoxb2 in r3-r5, hoxb3 in r4-r7 and hoxd3in r6-r7 are similar in wild-type controls and embryos treated with RA between 24-33 hpf (Fig. 8A-H). In addition, no alterations are seen in epha4 expression in r1, r3, and r5 in RA-treated embryos at 33 hpf (Fig. 8I-L). These results demonstrate that RA treatment between 24 and 33 hpf does not change hindbrain patterning or posteriorize rhombomere identities. Thus, the ectopic th expression we observed is not due to an expansion of posterior rhombomere identities into more anterior territories.

Taken together, these results reveal an important role for RA in the development of NA neurons in the medulla/area postrema and in the sympathetic nervous system. In both NA neuronal populations, Tfap2a may be involved in mediating the effect of RA.

RA signaling is mediated by two classes of retinoid receptors, RARs and RXRs (Chambon, 1995). To determine whether RA may directly regulate the expression of ththrough its receptors, or whether it may act through the expression of intermediary proteins, we tested whether RA can induce ectopic thexpression in the absence of new protein synthesis. To inhibit protein synthesis, we exposed zebrafish embryos to the protein synthesis inhibitor cycloheximide shortly before and during RA treatment. Embryos treated with cycloheximide alone had normal th expression in the CNS(Fig. 9B,F,J), but only a few th-expressing sympathetic neurons formed(Fig. 9N). Embryos treated with cycloheximide and RA did not express th ectopically in the hindbrain or peripheral nervous system (Fig. 9 compare C,G,K,O with D,H,L,P). Thus, protein synthesis is required to mediate the effect of RA on NA differentiation, indicating the involvement of intermediate targets of RA. To investigate whether tfap2a may mediate RA induced ectopic expression of th in the hindbrain, we analyzed tfap2a expression in RA-treated embryos. After RA exposure, tfap2a expression was upregulated in a restricted region of the hindbrain, in which th is also ectopically expressed in response to RA treatment (Fig. 9Q,R). We also examined the expression of phox2a and tfap2a in nls mutant embryos to elucidate if a reduction of RA signaling affects the expression pattern of these genes. At 34 hpf, tfap2a and phox2a expression was reduced in the hindbrain posterior to rhombomere 5 (Fig. 9S,T). These findings correlate with the lack of th-expressing cells in the posterior hindbrain and the normal development of the LC in nlsmutants.

The failure of RA to induce ectopic th expression in the hindbrain when protein synthesis was blocked indicates that the response to RA treatment requires intermediary factors. The upregulation of tfap2a expression after RA treatment and the downregulation of tfap2a expression in RA-deficient nls mutant embryos suggest that Tfap2a may mediate RA induced th expression in the posterior hindbrain.

The molecular mechanisms that lead to specification of noradrenergic neurons in the hindbrain, and to expression of the proper neurotransmitter phenotype in the LC and medulla, are poorly understood. Zebrafish embryos mutant for alleles of mont blanc are defective in NA neuron differentiation in the brainstem and in the sympathetic nervous system, and are also affected in visceral sensory ganglia and craniofacial development. We report that the pleiotropic mob phenotype is caused by a nonsense mutation in the tfap2a gene. Our analysis revealed a crucial role of tfap2a for expression of the neurotransmitter phenotype of NA neurons in the hindbrain. Expression of phox2a, a factor crucial for the specification of the locus coeruleus (LC), was not affected in mob/tfap2a mutant embryos, and tfap2a was not affected in phox2a/sou mutants. Thus, phox2a and tfap2a may represent parallel inputs into NA differentiation in the LC. Furthermore, our experiments provide the first in vivo evidence that tfap2a-dependent NA neuronal specification in the hindbrain is controlled by the retinoic acid signaling pathway.

Neural crest cells arise from the dorsal neural tube, migrate throughout the body and differentiate into many different derivatives, including peripheral nervous system and craniofacial skeleton (reviewed by Le Douarin and Kalcheim,1999). Loss-of-function experiments in mouse revealed that tfap2a contributes to a wide variety of developmental events,including skeletal formation in the thorax and limbs, aspects of organogenesis, formation of the peripheral nervous system, and neural tube closure (Schorle et al., 1996; Zhang et al., 1996). The analysis of tfap2a chimeric mice has shown that Tfap2a is indeed autonomously required for each of the pleiotropic aspects of the mutant phenotype investigated to date (Nottoli et al., 1998). In zebrafish, mutations in mob/tfap2a also cause extensive defects in the development of neural crest derivatives,including the craniofacial skeleton and pigment cells(Neuhauss et al., 1996; Schilling et al., 1996).

Primary sensory neurons of the trunk arise exclusively from the neural crest. By contrast, only some of the cranial sensory neurons are derived from neural crest. The majority of cranial sensory neurons have their origin in the dorsolateral trigeminal and vestibular placodes, and the distal epibranchial placodes (Ayer-Le Lievre and Le Douarin,1982; D'Amico-Martel and Noden, 1983; Noden,1993). The loss of the epibranchial derived sensory ganglia in tfap2a mutant embryos could be explained by defects in neural crest derived cranial structures. It was shown that neurogenic placodes form under the influence of the surrounding cranial tissues, including neural crest and pharyngeal endoderm (Vogel and Davies,1993; Begbie et al.,1999). The reduction of phox2a and ret1expression in tfap2a mutants indicates that both components of the sensory ganglia, namely neural crest and neurogenic placodal cells, are directly affected by lack of tfap2a. We have not detected an increase in apoptosis in any of the cranial ganglia in mob/tfap2a mutants(data not shown). This is in contrast to Tcfap2a knockout mice, which develop increased apoptosis in the trigeminal ganglia(Zhang et al., 1996).

tfap2a is expressed in neural crest cells and required for their normal development (Luo et al.,2003), and thus it is likely that neural crest defects in tfap2a mutants mediate the sensory ganglia phenotype. By contrast,the phenotype in the sympathetic nervous system may reflect a direct involvement of tfap2a in specification of the NA neurotransmitter phenotype. Hu and isl1 expression occurs early in the maturation process and is observed in S-phase neural precursor cells of the sympathetic ganglia (Vogel et al., 1993; Marusich et al., 1994b; Avivi and Goldstein, 1999). The expression of Hu and isl1 in cells in the sympathetic ganglia of tfap2a mutants indicates that neural crest cells migrate to their proper location, proliferate and start to differentiate, but fail to adopt their neurotransmitter phenotype. Thus, it is likely that expression of tfap2a in sympathetic precursors is required for correct further development and differentiation towards the noradrenergic phenotype.

Two main groups of noradrenergic (NA) neurons, a caudal rhombencephalic group (A1-A3, in medulla and area postrema) and a rostral rhombencephalic group (A4-A7, in the locus coeruleus), can be recognized in the hindbrain of higher vertebrates (Smeets and Reiner,1994; Smeets and Gonzalez,2000) as well as in zebrafish(Ma, 1994a; Ma, 1994b; Ma, 1997; Holzschuh et al., 2001). The control of noradrenergic development in the mouse hindbrain has mainly been studied in the locus coeruleus (reviewed by Goridis and Rohrer, 2002),where important roles have been demonstrated for Mash1(Hirsch et al., 1998), Phox2a(Morin et al., 1997), Phox2b(Pattyn et al., 1999) and Tlx3(Qian et al., 2001). Although Mash1, Phox2a and Phox2b appear to act in a linear cascade(Hirsch et al., 1998; Pattyn et al., 2000a), Tlx3 is thought to act independently of Phox2b during LC development(Qian et al., 2001). Tcfap2a is expressed in NA and AD neurons in the mouse and has been implicated to contribute to activation of the TH and DBH promoters in cell culture systems (Kim et al.,1998; Kim et al.,2001); however, a function of Tcfap2a for NA differentiation in vivo was unknown. We demonstrate a novel requirement of Tfap2a for NA neuron differentiation in the LC.

The development of the NA groups in the posterior hindbrain has not been studied extensively. It was shown that Tlx3 and Phox2b are required for the development of medullary NA neurons(Pattyn et al., 2000a; Qian et al., 2001). We show that expression of th is missing in the entire hindbrain of tfap2a mutant embryos at 2 dpf when noradrenergic differentiation occurs in wild type. Thus, our studies reveal tfap2a as an additional transcription factor required for early NA neuronal differentiation throughout the hindbrain. However, low level expression of dbh can be detected in a small group of cells in the medulla/area postrema from 4 dpf onwards,which indicates that some NA neuronal precursors can bypass a requirement for tfap2a during late development. We can only speculate that other transcription factors with later onset of expression, possibly additional members of the Tfap2 family, may substitute for tfap2a.

A loss of NA identity of the LC has also been reported for zebrafish soulless/phox2a mutant embryos, in which LC progenitor cells have been shown to exist (Guo et al.,1999a). We find that the medullary and area postrema NA neurons develop normally in soulless/phox2a mutants, in contrast to tfap2a mutant embryos. The genetic requirement for both tfap2a and phox2a to control th expression in the LC could be explained either by both of them providing independent inputs for NA specification, or by tfap2a acting downstream of phox2ain specification of the NA neurotransmitter phenotype. The latter idea would be consistent with our findings that the loss of functional Tfap2a does not cause a loss of phox2a expression in the LC, while thexpression does not occurs in tfap2a mutant embryos. Thus, tfap2a could be a downstream target of soulless/phox2a. However, several observations argue against this hypothesis being true for the whole hindbrain. First, tfap2a expression is not detected in many phox2a-expressing cells in the hindbrain and vice versa. Second, the expression pattern of tfap2a is not altered in soulless/phox2a mutant embryos. Third, phox2a and tfap2a are not co-expressed in cells that give rise to NA neurons of the medulla/area postrema. Therefore, tfap2a and phox2a may provide parallel inputs into NA differentiation of LC neurons, while upstream factors other than phox2a, e.g. phox2b, may control tfap2a expression in the medulla/area postrema.

The capacity of RA to induce neuronal differentiation in cell culture has been well documented for several cell lines (reviewed by Guan et al., 2001). Furthermore, the expression of Tfap2a is enhanced during RA induced neuronal differentiation in a human teratocarcinoma cell line(Williams et al., 1988; Luscher et al., 1989). In cell culture, RA may induce the expression of the noradrenaline transporter, a protein specifically expressed in NA neurons(Matsuoka et al., 1997). RA may also increase the number of TH-positive cells in quail neural crest culture (Rockwood and Maxwell,1996). These findings suggest that the RA signaling pathway may play a role in NA differentiation. To investigate a potential role of RA during zebrafish NA development in vivo, we analyzed embryos that are genetically deficient in one of the RA synthesis pathway enzymes,retinaldehyde dehydrogenase 2 (Raldh2), as well as embryos that have been exposed to high levels of RA. Incubation of embryos in the presence of high RA concentrations induced ectopic th expression in the anterior hindbrain caudal to rhombomere 1, as well as in the sympathetic nervous system. The capacity of RA to induce ectopic th-expressing cells was clearly restricted to rhombomere 2 to 7, and never occurred in rhombomere 1 or in the proximity of the LC. The nuclear transducers of RA signaling are the RAR- and RXR-type retinoid receptors, which bind to retinoic acid response elements (RARE) in target gene promoters as RXR homodimers, or as RXR heterodimers either with RAR, or with other nuclear receptors(Gudas, 1994; Chambon, 1995). Several zebrafish retinoic acid receptor genes have been cloned(Joore et al., 1994; White et al., 1994; Jones et al., 1995). At 24 hpf, rara is expressed in the hindbrain with a sharp posterior border at rhombomere 6 to 7. rarg is also expressed in the ventral part of the hindbrain and in two lateral patches of head mesenchyme(Joore et al., 1994). Following RA treatment, the expression of rarg is induced in the entire brain, whereas rara is ubiquitously induced in the embryo(Joore et al., 1994). As rara and rarg are both expressed in the posterior hindbrain and induced ectopically by RA, both are potential mediators for RA induced th expression in the hindbrain. The fact that both receptors are not expressed in rhombomere 1 may explain why RA is unable to induce thexpression anterior to rhombomere 2. When exposing zebrafish embryos to the protein synthesis inhibitor cycloheximide, we prevented the induction of ectopic th expression by RA. Therefore, translation of an intermediary target of RA signaling is required to induce thexpression, rather than th being directly activated by retinoid receptors.

We showed that, after treatment of embryos with excess RA, tfap2aexpression is upregulated in the posterior hindbrain. Furthermore, in tfap2a mutant embryos, exogenously applied RA is not able to ectopically induce th expression or to rescue the deficiency in NA-expressing cells. These findings implicate that RA signaling may induce th expression via activation of tfap2a expression. Indeed, a requirement for RA was confirmed when we analyzed zebrafish neckless/raldh2 mutant embryos, which are deprived of RA. Raldh2 is involved in the biosynthesis of RA from vitamin A (reviewed by Duester, 2000). Loss of function of Raldh2 in mouse mimicked the most severe phenotypes associated with vitamin A deficiency (VAD), implicating Raldh2 as the main source of RA in the vertebrate embryo (Niederreither et al., 1999; Niederreither et al., 2000). Zebrafish neckless/raldh2 mutant embryos resemble many aspects of VAD (Begemann et al., 2001). In nls mutants, we find that the NA neurons are missing in the posterior hindbrain, and that the expression of tfap2a is downregulated.

It is well established that RA signaling is involved in AP patterning of the hindbrain by regulating the expression of Hox genes(Holder and Hill, 1991; Hill et al., 1995) (reviewed by Gavalas and Krumlauf, 2000; Gavalas, 2002). In our experiments, however, embryos were exposed to RA only from 24 hpf onwards,when rhombomere identity has already been determined in the zebrafish. Indeed,we can demonstrate that the AP patterning of the hindbrain is not changed in our experiments: Anterioposterior expression borders of hoxa2, hoxb2,hoxb3, hoxd3 and epha4 were unaffected in our RA-treated embryos. Similarly, treatment of chick embryos with an RA antagonist at late or post-somitogenesis stages has not resulted in any changes of rhombomere identity (Dupe and Lumsden,2001).

In summary, our data provide evidence that two separate mechanisms control noradrenergic development in the zebrafish hindbrain. In the locus coeruleus,signals from the midbrain-hindbrain-boundary organizer initiate a cascade of transcription factors, including Phox2a and Phox2b(Guo et al., 1999a), which ultimately requires Tfap2a for NA precursor cells to express the NA neurotransmitter phenotype. In the posterior hindbrain, the medulla and area postrema, retinoic acid is an important signal, both necessary and sufficient to induce NA differentiation, and precursor neurons share a requirement for Tfap2a in order to be able to express the NA neurotransmitter phenotype. Thus,although the inductive signals may be different, hindbrain NA neurons of the LC and the posterior groups share a requirement for Phox2b, Tlx3 and Tfap2a to establish their noradrenergic identity. Furthermore, our data reveal that noradrenergic precursors of the peripheral nervous system rely on tfap2a to adopt their neurotransmitter phenotype.

We thank F. Biemar and A. Bodenhausen for their help during the genetic screen, and R. Geisler and G. J. Rauch (Tübingen) for help and advice during the linkage group assignment. We thank S. Guo for phox2a plasmids and probes. We thank J. von Lintig, S. Ryu, Z. Varga, D. Meyer and K. Lunde for helpful critique of the manuscript and discussion. S. Götter and R. Schlenvogt provided expert care of the fish. This study was supported by Deutsche Forschungsgemeinschafts Grant SFB 505 TP B7 and SFB 592 TP A3, a Forschungsschwerpunktprogramm Baden-Württemberg and DeveloGen AG(W.D.).