A role for foxd3 and sox10 in the differentiation of gonadotropin-releasing hormone (GnRH) cells in the zebrafish Danio rerio

Gonadotropin-releasing hormone (GnRH) is a functionally diverse decapeptide having both reproductive and neuromodulatory roles(Dubois et al., 2002; Somoza et al., 2002). Vertebrates have three forms of GnRH distinguished on the basis of their amino acid sequence, and these forms have been named: GnRH1, GnRH2 and GnRH3(Dubois et al., 2002; Fernald and White, 1999). Several species express variants of GnRH1(Montaner et al., 2002; Somoza et al., 2002), and although GnRH3 is considered to be specific to fishes it has been recently reported in rats (Parhar et al.,2005). These various forms of GnRH have distinct sites of expression in the nervous system. The endocrine form of GnRH, GnRH1, is expressed in hypothalamic cell populations that have been proposed to originate from the olfactory placodes(Schwanzel-Fukuda and Pfaff,1989; Wray et al.,1989a) and, more recently, from the anterior pituitary(Whitlock, 2005; Whitlock et al., 2003). The neuromodulatory form of GnRH found in the midbrain, called GnRH2, is unique in that its amino acid sequence is conserved in taxa from jawed fish to mammals(Millar, 2003). The GnRH2 neurons of the midbrain have been proposed to modulate behaviors relevant to reproduction (Millar, 2003; Temple et al., 2003), and are currently believed to arise from within the neural tube(Amano et al., 2004; Gonzalez-Martinez et al.,2002). The terminal nerve (a cranial nerve) contains cells that express GnRH (GnRH3), and these have a neuromodulatory role in the forebrain and olfactory epithelium (Abe and Oka,2002; Eisthen et al.,2000), and modulate olfactory-mediated behaviors(Wirsig-Wiechmann et al.,2002). The terminal nerve GnRH cells are proposed to originate from the olfactory placodes(Schwanzel-Fukuda and Pfaff,1989; Wray et al.,1989a) and, more recently, from the cranial neural crest(Whitlock, 2004b; Whitlock et al., 2003). Thus the GnRH3 cells of the terminal nerve are functionally more like the GnRH2 cells of the midbrain, in that they appear to modulate behaviors. For this reason, we examined whether GnRH2 cells might originate from neural crest.

The vertebrate head arises through a complex interplay between epithelial and mesenchymal cell populations. Prominent among the latter are cells derived from the cranial neural crest. Depending on the specific structure, neural crest cells may be the sole contributor or may complement contributions from other organs, e.g. mesoderm and placodes(Hall, 1999). Cranial neural crest cells give rise to a myriad of different cell types, including but not limited to: neurons, glia, pigment, and many skeletal elements of the head. Cranial neural crest cells also contribute to the specialized sensory systems,which have become concentrated and elaborated as the structure of the vertebrate head evolved (Le Douarin and Kalcheim, 1999).

Neural crest development is a complex, multi-step process, involving a myriad of inducing signals and responding transcription factors(Heeg-Truesdell and LaBonne,2004). Two genes encoding transcription factors important for neural crest differentiation are foxd3 and sox10. A member of the forkhead family of transcription factors, foxd3 is expressed in neural crest precursor cells. In Xenopus, overexpression of foxd3 results in the induction of neural crest-specific genes and loss of function suppressed the expression of neural crest markers, suggesting a primary role in neural crest cell differentiation(Sasai et al., 2001). Zebrafish foxd3 is expressed in the pre-migratory neural crest,floorplate, somites and tailbud (Odenthal and Nusslein-Volhard, 1998). Subsequently, it has also been shown to be expressed in glia of the peripheral nervous system(Kelsh et al., 2000). The Sry-related transcription factor gene sox10 is expressed in pre-migratory neural crest, and loss-of-function studies support its role in promoting the survival of undifferentiated neural crest cells(Honore et al., 2003; Mollaaghababa and Pavan,2003). In zebrafish, sox10 is expressed in developing pre-migratory neural crest and plays an important role in specifying non-ectomesenchymal (neurons, glia and pigment) neural crest(Dutton et al., 2001b); in mammals, it is important for the maintenance of multipotency in neural crest stem cells (Kim et al.,2003).

Deficits in GnRH (hypogonadic hypogonadism) underlie some human disease. Kallmann Syndrome in humans is characterized by deficits in GnRH that are associated with anosmia (loss of sense of smell). Previous studies have shown that development of the terminal nerve and hypothalamic GnRH cells is disrupted in a human Kallmann embryo(Schwanzel-Fukuda et al.,1989), and that this phenotype results from mutations in the KAL(anosmin 1) gene (Duke et al.,1995; Legouis et al.,1991). Homologues of the kallmann gene have yet to be identified in mouse, but, in the zebrafish, there are two kallmanngenes, kallmann 1.1 and kallmann 1.2 (Kallman syndrome 1a and Kallman syndrome 1b - Zebrafish Information Network). The genes are widely expressed in regions of the developing central and peripheral nervous system, as well as in the reproductive tract and pronephric ducts(Ardouin et al., 2000).

The function of midbrain GnRH2 cells suggested that, like the GnRH cells of the terminal nerve, they may arise from the cranial neural crest. We report that the decrement of foxd3 and sox10 gene function results in the loss of midbrain GnRH2 cells, as well as the loss of terminal nerve GnRH3 cells. In addition, we show, using a sox10-GFP line, that the GFP signal and gnrh2 expression co-localize during a discreet time window in early development, thereby supporting a neural crest origin for these cells. Finally, we demonstrate that the decrement of function of the kallmann 1.1 gene, but not the kallmann 1.2 gene, results in loss of the hypothalamic endocrine GnRH1 cells, but did not affect the neuromodulatory midbrain or terminal nerve GnRH cells. Together, these data suggest that the neuromodulatory midbrain GnRH2 cells, like the terminal nerve GnRH cells, have a neural crest origin.

All zebrafish (Danio rerio) used for these experiments were of the wild-type `Cornell Strain' developed in the Whitlock Laboratory. Animals were maintained on a 14 hour/10 hour light:dark cycle. Embryos were collected on the morning of the day of the experiments from group crosses consisting of three males and three females. The acerebellar (ace) mutant(Reifers et al., 1998) was provided by the Brand Laboratory, and embryos were collected from single-pair crosses of parents heterozygous for ace. All animal procedures were approved by the Cornell University Institutional Animal Use and Care Committee. The sox10-GFP line(Wada et al., 2005) was provided by the Kelsh Laboratory and maintained in our fish facility.

Morpholinos (MOs), modified oligonucleotides that interfere with mRNA translation (Heasman, 2002; Nasevicius and Ekker, 2000),have successfully been used in zebrafish and Xenopus to reduce gene function, including that of genes important for neural crest development(Dutton et al., 2001a; Honore et al., 2003; O'Brien et al., 2004). Here,we used MOs to knockdown the function of the foxd3, sox10, kal1.1 and kal1.2 genes. All MOs and corresponding 5-nucleotide mismatch controls (mm; mismatches indicated in lower case in sequence below) were synthesized by the manufacturer (Gene Tools, OR), using existing sequence information as follows:

foxd3 MO, 5′-CACTGGTGCCTCCAGACAGGGTCAT-3′;

foxd3 mm, 5′-CAgTGcTGCCTgCAGACAGcGTgAT-3′ (GenBank#NM131290);

sox10 MO, 5′-ATGCTGTGCTCCTCCGCCGACATCG-3′;

sox10 mm 5′-ATcCTcTGCTCaTCCGaCGAgATCG-3′ (GenBank#AF402677);

kal1.1 MO, 5′-CCGTCGCGCATCTTGAAGAACAGTA-3′;

kal1.1 mm, 5′-CCcTCcCGCATgTTGAAaAACAcTA-3′ (GenBank#AF163310);

kal1.2 MO, 5′-GCAGAGATTCCTCAAAAGCAGCATC-3′;

kal1.2 mm, 5′-GCAGAcATTgCTgAAAAGgAGgATC-3′ (GenBank#AF163311).

The foxd3 and kal1.1 MOs were tagged with lissamine (red fluorescence), and the sox10 and kal1.2 MOs were tagged with fluorescein (green fluorescence). This allowed for visualization of the MOs in the embryos. The control mismatch MOs were co-injected with rhodamine dextran dye (Molecular Probes), at a final concentration of 0.02%. For each MO and mismatch MO, 1, 5 and 10 nl of MO were tested to determine the most effective concentration. Injections of 10 nl or more of MOs resulted in ∼80%mortality, with embryos exhibiting non-specific necrosis; injections of 5-10 nl resulted in a lower mortality with embryos living through the segmentation stages, but they still displayed high levels of necrosis. For all MOs(foxd3, sox10, kal1.1, kal1.2) 2-4 nl produced specific defects,including those previously reported for sox10(Dutton et al., 2001a), and little general necrosis, and were therefore the volumes injected in the experiments reported here. For all mm MOs, we injected 0.5-1 nl, which produced no observable phenotype; volumes above 2 nl resulted in 80% mortality due to non-specific necrosis. For injections of foxd3+sox10,equal amounts of MOs to foxd3 and sox10 were pre-mixed. Injection of 10 nl or greater of the combined MOs resulted in 80% mortality,as was seen with the individual MOs. Injections were done into one- to two-cell stage embryos (Kimmel et al.,1995). Injection pipettes were pulled using thin walled borosilicate glass tubing with microfilament (OD=1.2 mm, ID=0.94 mm) on a Sutter Puller P-2000 (Sutter Instruments).

Zebrafish embryos were staged as described in Kimmel et al.(Kimmel et al., 1995). Embryos were fixed in phosphate-buffered 4% paraformaldehyde. Whole-mount in situ hybridization was performed as described in Thisse et al.(Thisse et al., 1993), using single-stranded RNA probes labeled with digoxigenin-UTP (Roche), the only modification being that the pronase (Sigma) permeabilization step was shortened to 3 minutes. The GnRH probes were generated after cloning these cDNAs from zebrafish (Gopinath et al.,2004). Briefly, the gnrh2 gene was initially cloned using heterologous primers to the goldfish gnrh2 gene [GenBank Accession number U30386 (Bogerd et al.,1994); forward primer, 5′-ATGGTGCACATCTGCAGGCT-3′;reverse primer, 5′-GTCATTTTCTCTTTTGGGAATC-3′] and the gnrh3 gene, expressed in the terminal nerve, was cloned using primers to the zebrafish gnrh3 sequence [GenBank Accession number AJ304429(Torgersen et al., 2002);forward primer, 5′-CAGCACTGGTCATATGGTTGGCTTCCCGG-3′; reverse primer, 5′-CACTCTTCCCCGTCTGTCGG-3′]. gnrh2 and gnrh3 cDNA was amplified by RT-PCR, the products were cloned into pGEM-T Easy Vector System I (Promega) and the identity of the resulting clones confirmed by sequencing (Gopinath et al.,2004). Probes to both gnrh2 and gnrh3 included the sequences to the GnRH decapeptide and the GnRH associated protein (GAP) of the pre-pro GnRHs, thereby making the probes highly specific to a given form of GnRH. The pattern of in situ hybridization showed that the gnrh2and gnrh3 probes had non-overlapping expression patterns, thereby confirming this specificity (see Results).

For analysis of gnrh2 and gnrh3 expression in the foxd3, sox10 and foxd3+sox10 MO-injected (morphant)animals, embryos were fixed at 35±1 hours post-fertilization (hpf) and processed for in situ hybridization using probes to gnrh2 or gnrh3, as described previously(Gopinath et al., 2004). For analysis of gnrh2 and gnrh3 expression in the kal1.1 morphant animals, embryos were fixed at 55±1 hpf.

To date the molecular form of GnRH expressed in the migrating hypothalamic cells has not been identified in zebrafish, and there is no antibody that specifically recognizes only this group of GnRH cells (see Results). Thus, for our analysis of hypothalamic GnRH (GnRH1), we used the LRH13 antibody(Park and Wakabayashi, 1986),which recognizes both the terminal nerve and hypothalamic GnRH cell populations. Immunocytochemistry was carried out as previously described(Whitlock et al., 2003),except that the embryos were fixed in 4% paraformaldehyde(Westerfield, 1993) with 7%saturated picric acid for two hours at room temperature.

Embryos positive for sox10-GFP were fixed at 26-28 hpf and protocol 5 of Schulte-Merker(Schulte-Merker, 2002) for immunocytochemistry/in situ hybridization double labeling was followed. The anti-GFP antibody (Molecular Probes, rabbit polyclonal) was used at a dilution of 1:1000 and the in situ probe for gnrh2 was that described above.

The number of cells expressing gnrh2 (midbrain) and gnrh3(terminal nerve) was counted in whole-mount embryos. Cell counts obtained in different groups of fish were compared using the Wilcoxon Rank Sum Test,comparing the medians of non-normal distributions.

In zebrafish, the foxd3 and sox10 genes are expressed in the developing pre-migratory neural crest(Fig. 1A-C), and in the majority of these cells this expression is then downregulated as they migrate anteriorly (Fig. 1D) to contribute to elements of the head (Fig. 1E). These genes are later expressed in neural crest-derived structures (Dutton et al.,2001b; Kelsh et al.,2000), but the lack of continuous expression prevented us from using expression as a de facto lineage tracer. In order to disrupt the development of neural crest cells, we used morpholinos (MOs; modified oligonucleotides) (Nasevicius and Ekker,2000) to knockdown the function of sox10 and foxd3, two genes essential for the differentiation of neural crest(Fig. 1A,B). Embryos injected with MOs to either foxd3 or sox10 alone were distinct in their appearance: foxd3 MO-injected animals (foxd3morphants) had pigment but showed a flexed trunk and tail(Fig. 2B), whereas the sox10 MO-injected embryos (sox10 morphants) lacked body pigment (Fig. 2C). The sox10 morphant phenotype that we obtained was the same as that previously reported (Dutton et al.,2001a; Dutton et al.,2001b). The knockdown of both Foxd3 and Sox10 resulted in a combined phenotype, such that the animals had no body pigment and a flexed trunk and tail (Fig. 2D).

Prior to the migration of neural crest, the cranial neural crest(Fig. 1C, purple) and the region giving rise to the olfactory placodes(Fig. 1C, red) share a common border. We have shown previously, through lineage-tracing techniques, that the neuromodulatory GnRH cells of the terminal nerve(Fig. 1E) come from cranial neural crest (Whitlock et al.,2003) (Fig. 1A-C,asterisks), and that the gnrh3 gene is expressed in these cells starting at 24-26 hpf (Gopinath et al.,2004). Here, we investigated whether the neuromodulatory gnrh2 cells of the midbrain (Fig. 1E), like the gnrh3 cells, also arise from neural crest.

Previously, we have shown that gnrh2 is expressed at 24 hpf in cells lying laterally within the mesencephalon(Gopinath et al., 2004). By in situ hybridization using a mRNA probe specific for gnrh2, we found that the reduction of either Foxd3 (Fig. 3B) or Sox10 (Fig. 3C) protein resulted in a significant reduction in the number of gnrh2-expressing cells at 35±1 hpf relative to the normal number of cells observed in mismatch MO-injected controls(Fig. 3A, Table 1). The foxd3morphants had a median of 12.5 cells expressing gnrh2(Fig. 3B, Table 1) compared with 25.0 gnrh2-expressing cells in the mismatch controls(Table 1; P<0.001). The sox10 morphants showed a larger reduction in cell number, showing a median of 7.0 gnrh2-expressing cells(Fig. 3C, Table 1) compared with 23.0 gnrh2-expressing cells in the mismatch controls(Table 1; P<0.001). The numbers of gnrh2- and gnrh3-expressing cells in mismatch controls were the same as we had previously reported for uninjected wild-type animals (Gopinath et al.,2004). Strikingly, injection of both MOs together resulted in an overall loss of gnrh2-expressing cells in the midbrain(Fig. 3D, Table 1, P<0.001). Thus, knockdown of both Foxd3 and Sox10 proteins had an additive effect on the reduction in gnrh2 cells of the midbrain, suggesting that foxd3 and sox10 may act together in the development of these cells.

The loss of function of foxd3 and sox10 resulted in a decrement and/or loss of the midbrain gnrh2 cells, suggesting a non-neural tube origin for these cells. To further exclude the neural tube as a possible origin for these cells, we examined the role of fgf8(fibroblast growth factor 8) in the development of midbrain gnrh2 cells. The fgf8 gene plays a role in the patterning of the midbrain-hindbrain boundary (MHB) region of the neural tube, as well as in the maintenance of gene expression in the developing midbrain(Irving and Mason, 2000; Jaszai et al., 2003; Reifers et al., 1998). We examined gnrh2 expression in the acerebellar (ace)mutant, which carries a null mutation in the fgf8 gene. We found that there was no statistically significant change in the number of cells expressing gnrh2 in ace mutant embryos(Fig. 3F) relative to their sibling controls (Fig. 3E, Table 1, P=0.561). The normal pattern of gnrh2 expression in these embryos mutant for Fgf8 function provided further support that gnrh2 cells do not arise from the neural tube.

In order to better understand the genetic pathways underlying the development of the neural crest-derived gnrh3 cells, we examined foxd3 and sox10 morphant animals for the presence of gnrh3-expressing neurons in the terminal nerve. In foxd3morphants, the median number of cells expressing gnrh3 in the terminal nerves at 35 hpf was reduced from 7.0 (control, Fig. 4A) to 1.5 (morphant, Fig. 4B, Table 1; P<0.001). The number of gnrh3 cells in sox10 morphants was also decreased, showing a median number of 0.5 cells(Fig. 4C, Table 1; P<0.001),compared with that in sox10 mismatch control morphants (median=6.0, Fig. 4A, Table 1). The combined reduction of both Foxd3 and Sox10 proteins also resulted in a decrease in the number of gnrh3-expressing cells(Fig. 4D-F, Table 1; P<0.001)relative to the double-mismatch morphant control(Fig. 4A). The phenotype of embryos injected with both MOs ranged from a complete lack of gnrh3cells (Fig. 4D) to gnrh3 cells present in association with only one developing olfactory organ (Fig. 4E, arrow). In a few preparations there were cells associated with both developing placodes,but they were displaced dorsal to the olfactory organs(Fig. 4F, arrows). This displacement was observed only in the double morphant embryos, and may be due to the disrupted timing of neural crest migration and differentiation. These data suggest that both foxd3 and sox10 play a role in the development of the gnrh3-expressing cells of the terminal nerve.

To determine whether the reductions in gnrh3 and gnrh2cell number caused by the injection of sox10 and foxd3 MOs were due to non-specific effects, we examined the expression of otx2,fgf8 and six4.1 in MO-injected embryos. The otx2 gene normally has a border of expression at the midbrain hindbrain boundary (MHB)(Li et al., 1994), and fgf8 is expressed in a distinct stripe of cells at the MHB(Scholpp et al., 2003). The MHB is an important signaling center in the developing neural tube, and is the region where the midbrain GnRH2 cells are first found(Amano et al., 2004; White and Fernald, 1998). We found no detectable alterations to the pattern of otx2 or fgf8 expression in sox10+foxd3 morphant embryos,the class of MO-treated embryos most likely to express non-specific defects because of their more extreme phenotype(Fig. 3D, Fig. 4D; Table 1). The posterior border of otx2 expression was unaltered in sox10+foxd3morphant animals (Fig. 5E,arrow) when compared with that of wild-type uninjected fish(Fig. 5A, arrow). Likewise, sox10+foxd3 morphants showed strong otx2 expression in the midbrain (Fig. 5F),similar to that observed in wild-type embryos(Fig. 5B). Consistent with the maintenance of the MHB, expression of fgf8 at 36 hpf was present in the sox10+foxd3 morphant animals(Fig. 5G, arrows) in the same pattern as was observed in wild-type uninjected fish(Fig. 5C, arrows). Thus, both otx2 and fgf8 expression patterns are maintained at the midbrain-hindbrain boundary in the sox10+foxd3morphants.

The six4.1 gene encodes a transcription factor that is expressed in the developing olfactory placodes(Kobayashi et al., 2000)(Fig. 5D,H). As shown in Fig. 5H, the sox10+foxd3 morphant embryos also showed a normal pattern of gene expression (Fig. 5D). Thus, our results suggest that the effects of knocking down sox10 and foxd3 on the number of gnrh3- and gnrh2-expressing cells were not due to a non-specific interference with the development of this region of the head.

The sox10 gene is initially expressed in pre-migratory neural crest (12 hpf; Fig. 1A). As the neural crest starts to migrate sox10 expression is downregulated. At 18-20 hpf, there is a small cluster of cells lying between the neural tube and the backside of the developing optic cup(Fig. 6A, arrows). By 24 hpf,this gene expression is no longer apparent by in situ hybridization. Using the sox10-GFP embryos, we were able to visualize this cluster of cells until 28 hpf because of the perdurance of the GFP protein(Fig. 6B, arrows, anti-GFP antibody). Because gnrh2 is first expressed at 24-26 hpf(Fig. 6C, arrows), as the sox10-GFP expression was waning, we were able to double label for gnrh2 (Fig. 6C,arrows) and the GFP protein (Fig. 6B, arrows). A subset of the gnrh2-expressing cells lying at the edge of the neural tube against the eye(Fig. 6D, purple, arrowhead;see also Fig. 6C) were also positive for sox10-GFP immunoreactivity(Fig. 6D, brown and purple,arrow; see also Fig. 6B),indicating that these cells still expressed sox10-GFP. Not all gnrh2 cells were positive for GFP(Fig. 6D, arrowhead),presumably because some of the cells had already lost their sox10-GFP signal. On average there are 5.33±0.58 gnrh2-positive cells at 24 hpf (Gopinath et al.,2004). In our analysis of ten preparations, 2.9±0.69(±s.e.m.) gnrh2-positive cells also contained GFP. Thus,around 54% of the gnrh2 cells were double labeled at 24-26 hpf.

The expression of sox10 in the first two days of development is very dynamic. At 36 hpf, sox10 expression is initiated in the region of the terminal nerve and at 48 hpf is in the cells surrounding the gnrh3 expression in the terminal nerve(Fig. 7A,B). In general, it appears that the gnrh3 (Fig. 7B, arrowhead, red) and sox10(Fig. 7B, arrow, blue) signals do not overlap, although some preparations showed cells that appeared to co-express gnrh3 and sox10(Fig. 7B, arrow with asterisk). Within the midbrain there was no expression of sox10 at 36 hpf, and at 48 hpf there were sox10-expressing cells in the CNS(Fig. 7C,D, arrows). This sox10 expression in the midbrain(Fig. 7C,D, arrows, blue) did not co-localize with gnrh2 expression(Fig. 7C,D, arrowheads, red),and most likely represents oligodendrocyte precursors(Woodruff et al., 2001).

The earliest sox10 expression(Fig. 8A,B, purple) may play a role in maintaining the pluripotency of the pre-migratory and early migratory neural crest cells. The use of the sox10-GFP line allowed us to follow the lineage of the early sox10-expressing cells by the perdurance of the GFP cells (Fig. 8C, green, yellow), and to visualize the loss of sox10 as the gnrh2 cells were differentiating. By 48 hpf, sox10 is expressed within the developing brain (Fig. 8D, purple) in cells dorsal to the gnrh2 cells(Fig. 8D, red).

Previously, we have shown that the endocrine GnRH cells populating the hypothalamus arise from the anterior pituitary placode, and that these cells are lost in mutants lacking the gli1 and gli2 signaling molecules, that disrupt anterior pituitary development(Whitlock et al., 2003). In order to determine whether the development of the endocrine GnRH cells of the hypothalamus was affected by knocking down foxd3 and sox10function, we assayed for the presence of these GnRH cells in sox10+foxd3 morphant embryos. In zebrafish, the anti-GnRH antibody LRH13 labels both the terminal nerve and the hypothalamic GnRH cells(Fig. 9E, arrowheads), whereas in situ probes to gnrh3 only label the terminal nerve cells(Fig. 9A). This suggests that hypothalamic cells express a different form of GnRH, as occurs in fish as well as some mammals (Lethimonier et al.,2004; Montaner et al.,2002; Somoza et al.,2002). To date, the gene encoding the form of GnRH expressed in this migratory population of GnRH cells has not been identified in zebrafish. Therefore, we used an antibody recognizing GnRH(Park and Wakabayashi, 1986)to visualize the hypothalamic GnRH cells in sox10+foxd3morphant embryos. Previously, we have shown that GnRH-IR is not expressed in hypothalamic GnRH cells until 52-54 hpf. As shown in Fig. 9F, hypothalamic GnRH cells were present in sox10+foxd3 morphant embryos when scored at 56 hpf. In spite of the altered morphology of the head, the hypothalamic GnRH cells were found in a group(Fig. 9F, arrows) near the post-optic commissure (Fig. 9F,asterisk). Thus, the combined knockdown of the Foxd3 and Sox10 proteins did not result in the loss of the GnRH cells of the hypothalamus, suggesting that these cells do not depend on foxd3 and sox10 gene function for their development.

We used MOs to knockdown kal1.1 and kal1.2 gene function,and assayed the effects on the GnRH populations of the hypothalamus (GnRH1),midbrain (GnRH2) and terminal nerve (GnRH3), at 55±1 hpf. We found that injection of kal1.2 MO had no effect on the number of neuromodulatory GnRH2 or GnRH3 cells, but caused a slight decrease in the number of endocrine GnRH cells (H.K. and K.E.W., unpublished). Surprisingly, we found that the kal1.1 MO resulted in a complete loss of the endocrine GnRH cells of the hypothalamus (Fig. 9H, n=35), an effect not seen in the kal1.1 mismatch control(Fig. 9G). By contrast, the number of GnRH3 cells of the terminal nerve was unaffected in the kal1.1 morphants, as judged both by immunocytochemistry(Fig. 9H, arrowheads) and by gnrh3 gene expression at 55±1 hpf(Fig. 9B, arrowheads). Although the two clusters of terminal nerve GnRH cells could sometimes be found in abnormal positions (Fig. 9B,arrowhead with asterisk), this defect was most likely the result of disrupted development of the olfactory system (H.K. and K.E.W., unpublished). Likewise,the gnrh2-expressing cells of the midbrain(Fig. 9C,D, arrows) were unaffected in the kal1.1 morphants, when scored at 55±1 hpf. Thus, only the endocrine GnRH cells of the hypothalamus were affected by the loss of function of the kal1.1 gene. These data support a model where the development of the neuromodulatory midbrain and terminal nerve GnRH cells is controlled by signaling pathways separate from those regulating the development of the endocrine GnRH cells.

Vertebrate brains contain up to three distinct populations of gonadotropin-releasing hormone (GnRH) cells. Both the GnRH3 cells of the terminal nerve, as well as the GnRH2 cells of the midbrain have a neuromodulatory function in the adult animal(Pawson et al., 2003), whereas the hypothalamic GnRH cells play an endocrine role in the hypothalamic/pituitary axis. The neuromodulatory GnRH3 cells of the terminal nerve arise from cranial neural crest (Von Bartheld and Baker, 2004; Whitlock, 2004b; Whitlock et al., 2003),whereas the endocrine GnRH1 cells of the hypothalamus arise from the anterior pituitary placode (Whitlock and Westerfield, 2000; Whitlock et al., 2003). The data presented here suggest a cranial neural crest origin for the GnRH2 cells of the midbrain, much like the origin of the terminal nerve GnRH3 cells. Cranial neural crest cells follow stereotyped migratory routes (Le Douarin and Kalcheim,1999): one route passing rostrally over the dorsal prosencephalon(Fig. 10D, purple arrow, a),and another route passing ventrally caudal to the developing eye(Fig. 10D, purple arrow, b). These are the most likely routes followed by the precursors of the gnrh3 (Fig. 10A) and gnrh2 (Fig. 10B)cells, respectively. The hypothalamic GnRH1 cells(Fig. 10C) were unaffected by the knockdown of sox10 and/or foxd3 function, further supporting an origin independent of neural crest.

In pre-migratory neural crest, the expression patterns of foxd3and sox10 overlap extensively, but not completely(Dutton et al., 2001b; Odenthal and Nusslein-Volhard,1998). foxd3 is downregulated upon migration(Kelsh et al., 2000) (K.E.W.,unpublished), and sox10 is maintained in the medial migratory route giving rise to glia of the cranial ganglia(Dutton et al., 2001b). We were able to use sox10-GFP embryos to follow the development of the sox10 cells past 18-20 hpf, when the sox10 gene was no longer expressed in the clusters of cells between the neural tube and eye. The co-localization of GFP with gnrh2 at 26 hpf suggests that the gnrh2 cells arise from this specific group of cranial neural crest-derived cells. The expression of sox10 has been extensively discussed as being important in the differentiation of peripheral glia and central nervous system oligodendrocytes(Woodruff et al., 2001). We have shown that sox10 is expressed in cells clustered around the gnrh3 cells of the terminal nerve at 36 hpf, and these cells may indeed be peripheral glia. More recently, sox10 has been shown to play a role in neuronal differentiation in rat neural crest stem cells(NCSCs). Constitutive expression of SOX10 allows NCSCs to maintain their proliferative activity, thus allowing for the induction of the proneural gene MASH1 (Kim et al., 2003). Here, we show that an MO-induced decrement in Sox10 expression results in a loss of gnrh2 cells, suggesting that, like in rat NCSCs, sox10 expression may be important for the differentiation of these neurons. Therefore, as previously reported(Kim et al., 2003), sox10 expression may allow for the maintenance of the proliferative plasticity necessary for GnRH2 neural differentiation and, unlike in glial subtypes, the expression is not maintained after the neurons differentiate.

At 48 hpf, sox10 expression is maintained in the terminal nerve and is also expressed in a cluster of cells in the midbrain(Fig. 7C,D; Fig. 8D). The midbrain sox10 expression does not co-localize with the gnrh2 cells and may represent differentiating oligodendrocytes previously described in the developing spinal cord of the zebrafish(Park and Appel, 2003; Park et al., 2002). In zebrafish, the co-expression of sox10 and olig2 within the spinal cord is not evident until 48 hpf(Park et al., 2002), a time well after we scored our morphants (36 hpf). Thus, in our experiments we would not have been able to observe any effects on oligodendrocyte differentiation.

The effects of reducing Foxd3 and/or Sox10 on the development of the midbrain and terminal nerve GnRH cells suggest a neural crest origin for both cell types. However, the relative contributions of these gene products to the development of the two populations differed: we observed, on average, a complete loss of gnrh2 cells in the double morphant, whereas the gnrh3 cell population was less affected, suggesting that other genes may be involved. This is consistent with previous work showing that the signals patterning cranial neural crest differ, such that the crest cells contributing to the frontal mass express different genes than those contributing to the branchial arches (for a review, see Santagati and Filippo, 2003). The roles of specific transcription factors, such as sox10 and foxd3, in neural crest development are complex. For example, in Xenopus, sox10 is required for early development of the neural crest,and its suppression results in the loss of foxd3 and slugexpression (Honore et al.,2003). In zebrafish, loss of sox10 results in the elimination of melanocytes, as well as in the disruption of enteric ganglia(Dutton et al., 2001a; Dutton et al., 2001b). The foxd3 gene appears to act as a transcriptional repressor in Xenopus, where overexpression can lead to the loss of neural crest(Pohl and Knochel, 2001), and,in chick, it appears to repress melanogenesis(Kos et al., 2001). Both foxd3 and sox10 are necessary for the genesis of glia and neurons in the peripheral nervous system of animals ranging from fish to mammals (Britsch et al., 2001; Dutton et al., 2001b; Honore et al., 2003; Southard-Smith et al., 1998). The knockdown of foxd3 and sox10 most likely disrupts both the initial genesis of neural crest, as well as the development of neuronal and glial lineages. Future analysis of the role of other transcription factors important for neural crest development, such as AP2(Hilger-Eversheim et al.,2000; Mitchell et al.,1991), will allow us to further define the signaling pathways involved in segregating the GnRH2 and GnRH3 neuronal populations in zebrafish. Whether the neural crest origins for the GnRH2 and GnRH3 cells, and the underlying genetic pathways reported here for zebrafish, apply to all vertebrates will require further investigations.

The hypothesis that GnRH2 cells originated from the neural tube was based on the location of these cells in developing and adult animals. In the flounder and the zebrafish, GnRH2 cells are found in a region of the midbrain(midbrain tegmentum) in juvenile and adult animals(Amano et al., 2004; Gopinath et al., 2004). In addition, ablation studies support the idea that the GnRH2 cells have an origin independent from that of the GnRH1 and GnRH3 cells(Northcutt and Muske, 1994). It was subsequently proposed that this population arises from the germinal zone around the third ventricle in the mesencephalic region of the neural tube(Amano et al., 2004; Parhar et al., 1998; White and Fernald, 1998). The posterior border of the midbrain and regions immediately adjacent to it are specified by fgf8, which is expressed in the midbrain-hindbrain boundary (MHB). Establishment of midbrain and hindbrain progenitors is independent of fgf8, but the maintenance of midbrain gene expression appears to be dependent upon fgf8(Jaszai et al., 2003; Reifers et al., 1998). We used the zebrafish ace mutant, which lacks fgf8 function, to disrupt patterning of the neural tube. Our data show that the loss of fgf8 does not result in a measurable change in the number of gnrh2 cells, suggesting that gnrh2 cells either do not arise from the neural tube, or arise from a region that is not dependent upon fgf8 for patterning information.

The developing embryo is a study in cell migration, and the neural tube is notable for the migration of the neural crest from its dorsal surface. In addition, there have been reports of ventrally emigrating neural tube cells(VENT) that leave the neural tube and contribute neurons to the peripheral nervous system, as well as neurons and glia to the enteric nervous system(Dickinson et al., 2004). Furthermore, cells associated with the brain vasculature (vascular smooth muscle cells) migrate into the central nervous system and may arise from neural crest-derived head mesenchyme re-entering the developing brain(Korn et al., 2002). Thus,there may be precedents for neural crest-derived non-neuronal cell types re-entering the neural tube during development.

Our results indicate that the gnrh2-expressing cells of the midbrain arise from cells whose site of origin and early molecular characteristics are similar to those of neural crest cells. This is not the first report of centrally located neurons arising from neural crest and migrating into the neural tube. For example, the neurons of the mesencephalic nucleus of the trigeminal (Mes5) have been reported to arise from neural crest(Narayanan and Narayanan,1978) and to then migrate to their adult location in the mesencephalon. However, we first see gnrh2-expressing cells at 24 hpf lying laterally within the neural tube, and, by 56 hpf, they are in two clusters lying on either side of the midline(Gopinath et al., 2004). Thus,they do not appear to be `unique primitive population along the dorsal midline', as has been described for Mes5 neurons(Sanchez et al., 2002). Furthermore, loss of fgf8 in the isthmus did not affect the number of gnrh2 cells, as it does for the Mes5 neurons(Hunter et al., 2001). Our data support the idea that gnrh2 cells arise from neural crest-derived cells that migrate into the neural tube during development.

In zebrafish, the terminal nerve is an easily identifiable telencephalic population of GnRH cells in the forebrain that expresses gnrh3 early in development (Gopinath et al.,2004; Whitlock et al.,2003). In medaka, the differences in timing between the onset of expression of gnrh3 in the terminal nerve and gnrh1 in the hypothalamus led to the proposal that these populations do not share a common origin in the olfactory placode (Dubois et al., 2002; Parhar et al.,1998). Consistent with this hypothesis, we have found that, in zebrafish, GnRH3 cells of the terminal nerve have their origin in the cranial neural crest (Whitlock, 2004a; Whitlock, 2004b; Whitlock et al., 2003). Whether the terminal nerve arises from neural crest or the olfactory placode has been the subject of debate for the last century(Von Bartheld and Baker, 2004; Whitlock, 2004b). The data presented here support and extend our previous findings that GnRH3 cells of the terminal nerve arise from cranial neural crest, by defining foxd3and sox10 function as being necessary for the differentiation of the gnrh3 cells of the terminal nerve.

Originally, it was proposed that the endocrine and neuromodulatory GnRH cells of the forebrain of mammals arose from the olfactory placode and migrated to their adult location using routes along the developing olfactory and vomeronasal nerves (Schwanzel-Fukuda and Pfaff, 1989; Wray et al.,1989a; Wray et al.,1989b). Subsequently it was suggested, in both fish and mammals,that there are multiple origins for these populations of GnRH cells(Amano et al., 2004; Dubois et al., 2002; Parhar et al., 1998; Quanbeck et al., 1997). Our results further define the different developmental mechanisms underlying the differentiation of the neuromodulatory GnRH cells and the endocrine GnRH cells. Strikingly, we have shown that the knockdown of the kal1.1gene resulted in a loss of the endocrine GnRH1 cells, suggesting a crucial role for one isoform of the kallmann gene in the differentiation of the endocrine GnRH1 cells.

The idea of a neural crest origin for GnRH2 cells is in agreement with previous reports of GnRH2 in the neural crest-derived sympathetic ganglia in amphibians (Troskie et al.,1997). gnrh2 expression has also been reported in non-neuronal, mesodermally derived structures such as bone marrow, kidney and reproductive tissues (for a review, see Millar, 2003), clearly indicating multiple origins for gnrh2-expressing cells beyond what we have described here. Recently, it was demonstrated through an analysis of the sequences of reported structural variants of GnRH that the midbrain gnrh2 form is more closely related to the form found in the terminal nerve of fishes (gnrh3) than it is to the hypothalamic forms(gnrh1) (Lethimonier et al.,2004). Thus, the neuromodulatory midbrain GnRH2- and terminal nerve GnRH3-containing cells have molecularly similar forms of GnRH (when compared with the endocrine hypothalamic form, GnRH1), and our data support a common embryonic origin for these populations in the cranial neural crest.

We thank members of the Whitlock laboratory for helpful discussions; the Brand and Kobayashi laboratories for providing the otx2 and fgf8, and six4.1 cDNAs, respectively; Dr Robert Kelsh for providing the sox10-GFP line; and Drs D. Noden and J. Ewer for critical reading of the manuscript. This work was supported by NIH/DC0421801,NIH/HD050820, and a New York State Hatch Grant (K.E.W.).