Fgf3 signaling from the ventral diencephalon is required for early specification and subsequent survival of the zebrafish adenohypophysis

The pituitary gland links the nervous system with the endocrine system of vertebrates, responding to signals from the hypothalamus with the production and release of peptide hormones that regulate vital processes in metabolism,reproduction and growth. It consists of two parts that both secrete hormones:the posterior pituitary, also called the neurohypophysis, and the anterior pituitary, called the adenohypophysis. Neurohypophyseal hormones are generated in neuroendocrine cells in the hypothalamus, from where they are brought to the neurohypophysis via axonal transport, whereas adenohypophyseal hormones are generated in pituitary cells themselves. In mammals, these can be subdivided into different cell types, characterized by the hormones they produce: lactotropes producing prolactin (Prl), somatotropes producing growth hormone (Gh), thyrotropes producing thyroid-stimulating hormone (Tsh),gonadotropes producing luteinizing or follicle-stimulating hormone (Lh, Fsh),and corticotropes and melanotropes producing a common precursor peptide,Proopiomelanocortin (Pomc), which is proteolytically cleaved to give rise to adrenocorticotropin (Acth) and melanocyte-stimulating hormone (Msh).

The neurohypophysis is of neuroectodermal origin and derives from the infundibulum, an evagination of the ventral diencephalon. By contrast, the adenohypophysis is a derivative of the non-neuronal, placodal ectoderm. It is initially located at the anterior neural ridge (ANR) and becomes part of the oral roof ectoderm, from where it invaginates toward the presumptive neurohypophysis/infundibulum, forming Rathke's pouch. This co-development of neuro- and adenohypophysis is governed by various signaling processes between infundibulum and adenohypophyseal placode (for reviews, see Treier and Rosenfeld, 1996; Burgess et al., 2002; Scully and Rosenfeld, 2002). For example, the infundibulum generates at least three members of the fibroblast growth factor (Fgf) family, Fgf8, Fgf10 and Fgf18(Norlin et al., 2000; Ohuchi et al., 2000). Gain-of-function studies with fgf8 transgenic mice or pituitary explants treated with Fgf8 beads suggest that Fgf signaling promotes the proliferation and differentiation of dorsal cell types of the adenohypophysis(corticotropes and melanotropes), antagonized by Bmp2, which promotes ventral cell fates (thyrotropes, gonadotropes, somatotropes and lactotropes)(Ericson et al., 1998; Treier et al., 1998). Consistent results were obtained in loss-of-function studies, showing that treatment of embryonic day (E) 10 pituitary transplants with the Fgf receptor inhibitor SU5402 (Mohammadi et al.,1997) selectively blocks the proliferation and generation of dorsal cell types (Norlin et al.,2000). Together, the data suggest a patterning and cell type selective role for Fgf signaling. By contrast, knockout mice deficient in Fgf10 (Ohuchi et al., 2000) or the Fgf receptor 2 isoform IIIB (Fgfr2IIIb)(De Moerlooze et al., 2000)display a poorly formed Rathke's pouch, with numerous apoptotic cells throughout the entire pouch. These defects are already apparent at E10,suggesting an earlier and more general role for Fgf signaling during mouse pituitary development, affecting pouch morphogenesis and the survival and proliferation of all adenohypophyseal progenitor cells.

Here, we describe mutant analyses to uncover the role of Fgf3 during adenohypophysis development in the zebrafish. Recent work has revealed both striking similarities and differences in pituitary development between the different vertebrate classes. Thus, we found that despite differences in the morphogenesis and architecture of their pituitary glands, the role of Sonic Hedgehog during induction, growth and patterning of the adenohypophyseal anlage appears to be largely conserved between zebrafish and mouse(Herzog et al., 2003; Sbrogna et al., 2003). To isolate additional genes required for zebrafish pituitary development, we conducted a screen for ENU-induced mutations affecting adenohypophyseal production of Gh (Herzog et al.,2004). One gene identified in this screen is the pituitary-specific transcription factor Pit1. pit1 mutant zebrafish are semi-viable and show very similar phenotypic traits to those of their human and mouse counterparts (Nica et al.,2004), strengthening the value of the zebrafish system as a clinically relevant model organism. Here, we show that another identified gene, limabsent (lia), encodes zebrafish Fgf3(Kiefer et al., 1996). Using antisense-mediated inactivation, fgf3 has recently been proposed to be necessary for craniofacial development(David et al., 2002; Walshe and Mason, 2003b) and,in partial redundancy with fgf8, telencephalic patterning(Shanmugalingam et al., 2000; Fürthauer et al., 2001; Shinya et al., 2001; Walshe and Mason, 2003a),hindbrain patterning (Liu et al.,2003a; Maves et al.,2002; Walshe et al.,2002) and otic placode induction(Phillips et al., 2001; Leger and Brand, 2002; Maroon et al., 2002; Liu et al., 2003a). Our mutant analyses confirm most, but not all, of the proposed Fgf3-unique roles.

During pituitary development, Fgf3 is primarily required to promote early specification steps and the subsequent survival of adenohypophyseal cells. However, we find that morphogenesis and growth of the pituitary anlage do not depend upon Fgf3. In addition, we show that the source of Fgf3 is cells of the ventral diencephalon, presumably encompassing the presumptive neurohypophysis.

Mapping of the lia^t24147 and lia^t24152 mutations was done via PCR analysis of genomic DNA from individual lia mutant embryos, as described(Geisler, 2002). For identification of point mutations, fgf3 cDNA was amplified via RT-PCR, as described (Nica et al.,2004), using the following PCR conditions: 3 minutes at 94°C -32× (30 seconds at 94°C, 30 seconds at 60°C, 1.5 minutes at 72°C) - 10 minutes at 72°C; sense primer 5′GCTTGGAGAGCAGAGTGGAAC; antisense primer 5′ GCCTGGGATCTCCATCTGTGTC. In two alleles, the mutations generate restriction length polymorphisms (RFLP),which were used to genotype single embryos via PCR of genomic DNA. For the fgf3^t26212 allele, PCR conditions were: 2 minutes at 94°C - 34× (30 seconds at 94°C, 30 seconds at 58°C, 1 minute at 72°C) - 5 minutes at 72°C; sense primer, 5′TTCTGCTCTTGTTGTTACTGAGC, antisense primer 5′ TACGCAAGTCCCATGAACTCATC. The obtained 330 bp PCR fragment was digested with Tsp509I, leading to a cleavage of the wild-type fragment to a 120 bp fragment and a 210 bp fragment, while the mutant fragment remained at 330 bp. For the fgf3^t24149 allele, PCR conditions were: 2 minutes at 94°C - 35× (30 seconds at 94°C, 30 seconds at 58°C, 1 minute at 72°C) - 5 minutes at 72°C; sense primer 5′GTCTTCAACCGAGAGTGTG; antisense primer 5′ CTATGCCGGACCCTGTTG. The PCR products were digested with Hpy188I, yielding mutant fragments of 90 bp and 20 bp, while the wild-type fragment remained at 110 bp.

For expression constructs, full-length wild-type and mutant fgf3cDNAs were amplified via RT-PCR with primers containing EcoRI and XbaI restriction sites, and cloned into pCS2+(Rupp et al., 1994). Capped RNA was prepared with the Message Machine kit (Ambion). shh RNA was prepared, and RNA was injected as described(Herzog et al., 2003).

Whole-mount in-situ hybridizations were carried out as previously described(Hammerschmidt et al., 1996; Herzog et al., 2003). For fgf3 in-situ probe synthesis, plasmid pCRII-fgf3 was linearized with KpnI and transcribed with T7 RNA polymerase. In addition, riboprobes of the following cDNAs were used: dlx2, dlx3(Akimenko et al., 1994), eya1 (Sahly et al.,1999), emx1 (Morita et al., 1995), gh, pomc, prl(Herzog et al., 2003), hgg (Thisse et al.,1994), krox20 (Oxtoby and Jowett, 1993), lim3(Glasgow et al., 1997), nkx2.1a (Rohr and Concha,2000), pit1 (Nica et al., 2004), shh(Krauss et al., 1993), spry4 (Fürthauer et al.,2001).

To visualize craniofacial cartilage, 120 hours postfertilization (hpf)embryos were stained with Alcian Blue (Sigma) as described(Schilling et al., 1996).

For detection of apoptotic cells, dechorionated live embryos were incubated for 10 minutes in 5 μg/ml Acridine Orange (Sigma) in embryo medium (E3),followed by fluorescent microscopy (Zeiss axiophot) and photography.

To study the fate of pituitary precursor cells, single cells in the prospective pituitary forming region were injected in tailbud-stage embryos with 3% rhodamine dextran (3 kD) in 0.2 M KCl, using an AM-Systems 1600 Neuroprobe amplifier, and the underlying mesodermal polster as a morphological landmark (Varga et al., 1999)(S. Dutta and Z.M.V., unpublished). Fates of labeled daughter cells were analyzed between 24 and 30 hpf, using conventional fluorescence microscopy, or a Zeiss LSM 510 confocal microscope. Individual embryos were genotyped after photography.

Embryos were incubated at 28.5°C in 12 or 20 μM SU5402 containing E3, prepared from a 3 mM SU5402 (Calbiochem, S72630) stock solution in DMSO. Control embryos were incubated in E3 medium with the corresponding amount of DMSO. After treatment, embryos were washed five times with E3/DMSO and transferred to fresh E3.

For transplanting wild-type cells into offspring of heterozygous mutants,wild-type embryos were injected with a 1% biotin-dextrane, 0.5% fluorescein dextrane solution. Homotopic transplantations of presumptive telencephalic or diencephalic cells were carried out at the shield stage. At 32 hpf, chimeras were fixed and analyzed by in-situ hybridization, followed by anti-biotin stainings with the Vectastain Elite ABC kit (Vector Laboratories) to stain transplanted cells. The tails of chimeras (containing only recipient-derived cells) were genotyped, as described above.

Heparin-coated acrylic beads (Sigma, H-5263) were washed twice in PBS, and incubated with 0.5 μg/μl recombinant human FGF3 (R&D Systems,1206-FG) or 0.5 μg/μl BSA overnight at 4°C. At 19 hpf, offspring from a cross of two t24149/+ carriers were dechorionated, embedded in 1% low melting agarose on agarose-coated petri dishes, and covered with Ringer's solution. Using a fine tungsten needle and forceps, a bead was implanted behind the eye and pushed to the front of the head. After implantation,embryos were left on the plates for recovery for 2 hours at 28.5°C, then carefully taken out of the agarose and incubated for another 5 hours until fixation at 26 hpf.

During a large-scale screen for zebrafish mutants with altered growth hormone expression, we isolated four ENU-induced lia mutations. lia mutant embryos are characterized by the complete absence of lim3 expression in the adenohypophyseal anlage at 32 hpf, and the lack of all hormone-producing adenohypophyseal cell types during further development (Herzog et al.,2004). Mutants usually die between 7 and 9 days after fertilization, most likely due to the additional craniofacial defects (see below). Using bulk segregation analysis, lia was mapped to linkage group 7. Further fine mapping located lia approximately 62 cM from the top of LG7, between the markers Z41496 and Z64508. One of the genes that had previously been mapped to the same region is fgf3(wwwmap.tuebingen.mpg.de). Together with the similar craniofacial phenotypes of lia mutants and fgf3 morphants (see below), this made fgf3 a likely candidate for lia.

PCR amplification and sequencing of fgf3 cDNA from the four different lia alleles revealed three missense mutations and one nonsense mutation, leading to exchanges or deletions of highly conserved amino acid residues of Fgf3 protein (Fig. 1A). In the fgf3^t26212 allele, a T-to-A transition leads to the replacement of an isoleucine residue at amino acid position 76 of the protein by an asparagine (I76N). In fgf3^t24149, a G-to-A transition causes the exchange of glutamate 138 to lysine (E138K). Both mutations generate RFLPs, which were used to confirm the linkage between the lia mutations and the fgf3 gene (no recombination in 436 meioses). In fgf3^t21142, tyrosine 148 is exchanged to a cysteine(T148C), due to an A-to-G transition, while in the fgf3^t24152 allele, a G-to-A transition introduces a premature stop codon after amino acid residue 177. This results in a deletion of the last 79 amino acids of the protein, including 40 amino acids of the central core of 140 amino acids that is highly conserved among the different Fgf family members (Powers et al.,2000).

To determine the effects of the four lia/fgf3 mutations on the activity of Fgf3 protein, we carried out overexpression studies in zebrafish embryos. Injection of wild-type fgf3 RNA leads to a general dorsalization of the embryo, and to a posteriorization of its neuroectoderm(Fürthauer et al., 1997; Koshida et al., 2002). For better quantification of these effects, we defined four phenotypic classes of increasing strength, from mild F1 to strong F4 (see legend to Fig. 1B). When injected at a concentration of 0.1 ng/μl, wild-type fgf3 mRNA was highly effective, whereas all four mutant versions showed hardly any effect(Fig. 1B). Even upon injection at 100-fold higher concentrations (10 ng/μl), fgf3^t24152 and fgf3^t21142 mRNA were ineffective, while fgf3^t26212 and fgf3^t24149 showed some effects that, however, were significantly weaker than those obtained with wild-type mRNA at 0.1 ng/μl(Fig. 1B). Together, these data indicate that the fgf3^t24152 nonsense mutation and the fgf3^t21142 missense mutation are likely to be null mutations (amorphs), while fgf3^t26212 and fgf3^t24149 have retained less than 1% of the normal Fgf3 activity, and can therefore be regarded as very strong hypomorphs or near amorphs. This is consistent with the embryonic defects caused by the fgf3 mutations, which are of indistinguishable strength for all four alleles.

Several recent publications describe fgf3 and combinatorial fgf3 and fgf8 loss-of-function studies, using antisense morpholino oligonucleotide technology (see Introduction). The combinatorial activity of fgf3 and fgf8 was reported to be required for otic vesicle formation, telencephalon and hindbrain patterning, and craniofacial development. Most of these processes were also affected upon single loss of fgf3; however, the effects were more moderate or more restricted, pointing to partial redundancy of Fgf3 and Fgf8.

During craniofacial development, Fgf3 from the pharyngeal endoderm was reported to be required for the specification and survival of chondrogenic neural crest cells of the gill arches (pharyngeal arches 3-7)(David et al., 2002), while the crest cells forming mandibular and hyoid (pharyngeal arches 1 and 2) are under the redundant control of Fgf3 and Fgf8(Walshe and Mason, 2003b). Consistent with this notion, Alcian Blue stainings of fgf3 mutants at 120 hpf revealed an almost complete loss of the cartilage of the gill arches,while the cartilage of mandibular and hyoid was present(Fig. 2G,H). These defects are anticipated by a progressive reduction in the number of dlx2-positive chondrogenic neural crest cells of arches 3-7, starting around 26 hpf, while the two anterior dlx2 domains, forming cartilage of arches 1 and 2,appear normal (Fig. 2C,D). In addition, the posterior dlx2 domain fails to subdivide into the streams corresponding to the different gill arches(Fig. 2C,D). Similarly, the pharyngeal endoderm remains unsegmented, as reflected by the fgf3expression pattern (Fig. 2E,F).

We could also confirm the previously reported essential, but partially redundant, role of Fgf3 during inner ear development. Thus, fgf3mutants display otic vesicles of reduced size and fused otoliths(Fig. 2G,H). By contrast,mutants show normal expression patterns of emx1 and dlx2 in the dorsal and ventral telencephalon (Fig. 2A,B), suggesting that in contrast to a previous report (Walshe et al., 2003a), Fgf3 is dispensable for telencephalic patterning. This discrepancy between mutant and morphant data might be due to toxicity and/or cross-reactions of the fgf3 morpholino oligonucleotides, possibly with thus far undescribed fgfs.

By contrast to the processes mentioned above, no role of Fgf3 during adenohypophysis development had been described so far. We carried out double in-situ hybridizations to investigate fgf3 expression relative to the developing adenohypophysis. At the end of gastrulation (10 hpf), fgf3is expressed in the telencephalon, adjacent to the placodal cells of the anterior neural ridge (ANR) that are marked by dlx3 expression(Fig. 3A). The placodal ectoderm itself is devoid of fgf3 transcripts. Similarly, the progressing anterior dorsal mesoderm, called polster and marked by the expression of hgg, lacks fgf3 expression(Fig. 3B). Median ventral regions of the neuroectoderm, marked by the expression of shh(Fig. 3C) (cf. Varga et al., 1999), might show some weak and diffuse fgf3 staining(Fig. 3B,E). However, strong and distinct diencephalic fgf3 expression can first be detected from the 18-somite stage onward (18 hpf), after fgf3 expression in the telencephalon has ceased (Fig. 3F). Within the diencephalon, fgf3 expression is restricted to the ventralmost cell layers, as revealed with double-stainings using the hypothalamic marker nkx2.1a(Fig. 3I,L). This tuberal region of the posterior-ventral hypothalamus is thought to correspond to the presumptive infundibulum of higher vertebrates, which gives rise to the neurohypophysis (Fürthauer et al.,2001; Mathieu et al.,2002). It is in close contact with the adenohypophyseal cells,which at 26 hpf are located in a horseshoe-like pattern around the anterior and lateral borders of the fgf3 expression domain(Fig. 3J,M). During further development, the lateral-posterior cells of the adenohypophyseal anlage converge to the midline (Nica et al.,2004) and become located underneath the fgf3-expressing cells of the presumptive infundibulum (Fig. 3N,O). By contrast, the anteriormost cells of the adenohypophysis,as characterized by the expression of prolactin, are in some distance from the fgf3-positive cells (Fig. 3K).

Comparative in-situ hybridizations with fgf8 further reveal that the fgf3-positive cells of the infundibular region lack fgf8expression (compare Fig. 3Gwith 3H), which instead is weakly expressed in parts of the adenohypophyseal anlage itself and in posteriormost regions of the ventral diencephalon(Fig. 3H). Another site of unique fgf3 expression is the posterior pharyngeal endoderm(Fig. 3G,H), explaining the gill arch defects of fgf3 mutants (see above).

As a first step to characterize the adenohypophyseal defects of fgf3 mutants, we carried out in-situ hybridizations at various stages of development. Placodal ectoderm starts to be specified toward the end of gastrulation, indicated by the expression of eya1, dlx3 or pitx3. All of them are expressed throughout the entire placodal field, while marker genes specific for the different placodes along the anterior-posterior axis of the embryo (adenohypophysis, olfactory, lense, otic etc.) only come up later. At tailbud stage (10 hpf), fgf3 mutants display normal expression of eya1(Fig. 4A,B), dlx3 and pitx3 (data not shown), indicating that Fgf3 is not required for the earliest steps of general placodal specification. Similar unaltered expression patterns were found in fgf3 and fgf8 (ace)(Reifers et al., 1998) double mutants, and in embryos treated with the Fgf receptor inhibitor SU5402 from mid-through late-gastrula stages(Mohammadi et al., 1997) (data not shown), indicating that early placodal development is independent of Fgf signaling in general.

The first currently known adenohypophysis-specific zebrafish marker genes[pit1 (Nica et al.,2004); lim3 (Glasgow et al., 1997)] start to be expressed around the 19-somite stage(18.5 hpf), slightly after the initiation of fgf3 expression in the infundibular region (see above). In fgf3 mutants, expression of pit1 and lim3 fails to be initiated(Fig. 4C-F for pit1 at 19 hpf, lim3 at 24 hpf). Similarly, no transcripts of adenohypophyseal hormones can be detected at any stage(Fig. 4K,L for pomc at 72 hpf, and data not shown).

The infundibular region itself appears to be less affected by the fgf3 mutations. Infundibular cells of mutant embryos lack expression of the Fgf target gene and autocrine feedback antagonist sprouty4(spry4) (Fig. 4I,J)(Fürthauer et al., 2001). However, fgf3 mutants display normal infundibular expression of fgf3 at 36 hpf (Fig. 2E,F) and later (48 hpf, 72 hpf, data not shown), indicating that infundibular cells are maintained, and that in contrast to Fgf8(Reifers et al., 1998; Heisenberg et al., 1999), Fgf3 is not required for the maintenance of its own expression.

Like the infundibulum, which is part of the tuberal, posterior-ventral hypothalamus, other regions of the ventral diencephalon appear to develop independently of Fgf3, too. This is indicated by the unaltered expression of shh (Fig. 4G,H), a marker for the anterior-dorsal hypothalamus(Mathieu et al., 2002), and by the normal expression of pomc in endorphin-synthesizing hypothalamic neurons of mutant embryos (Fig. 4K,L).

As described above, adenohypophyseal cells are exposed to Fgf3 signals from different sources (telencephalon, diencephalon) and at different stages of development (end of gastrulation, midsegmentation)(Fig. 3). To determine the source of Fgf3 required for adenohypophyseal development, we generated mosaic embryos, transplanting presumptive telencephalic or diencephalic cells from early gastrula wild-type donor embryos into the same regions of fgf3mutant recipients (Fig. 5B). We found that wild-type cells in the ventral region of the diencephalon rescued lim3 expression in adjacent adenohypophyseal cells of fgf3mutants (Fig. 5E,F; n=9). By contrast, even largest clones of wild-type cells in the telencephalon of fgf3 mutants failed to rescue lim3expression (Fig. 5G,H; n=8). This indicates that it is Fgf3 from the ventral diencephalons that is required in a non-cell autonomous fashion to instruct adenohypophyseal specification in the underlying placodal ectoderm.

To narrow down the time window during which Fgf3 from the forebrain is required for adenohypophysis development, two kinds of experiments were carried out. To block Fgf signaling in a temporally controlled fashion, we treated wild-type embryos at various stages with the Fgf receptor inhibitor SU5402 (Mohammadi et al.,1997). Embryos treated during midsegmentation stages (18-22 hpf)show a complete loss of lim3, pomc and prl expression at 34 hpf, while head morphology appears normal(Fig. 6E-H, and data not shown)(284/286 embryos without adenohypophyseal marker gene expression; 6 independent experiments). In comparison, SU5402 treatment during gastrulation and early segmentation stages (8-12 hpf) led to milder adenohypophyseal defects (66/138 embryos without, 72/138 with, reduced adenohypophyseal marker gene expression; data not shown), although such embryos display strong brain defects (data not shown) (cf. Walshe et al., 2003a). In sum, the data show that Fgf signaling during midsegmentation stages is absolutely required for adenohypophyseal development, while the importance of earlier Fgf signaling is less clear (see Discussion).

A requirement of Fgf signaling during midsegmentation stages is also revealed by bead implantation studies. When implanted into the diencephalon of mutant embryos between the 18- and 20-somite stages (19 hpf), Fgf3-loaded beads led to a significant rescue of adenohypophyseal lim3 expression at 26 hpf (Fig. 6B,C) (9/9; 2 independent experiments). By contrast, BSA-loaded control beads had no effect(Fig. 6D) (0/5).

In sum, these experiments indicate that adenohypophysis development depends on Fgf3 during stages when it is normally expressed in the presumptive infundibular region.

In addition to Fgfs, the diencephalon is known to be the source of other crucial signals driving adenohypophysis development. One of the identified diencephalic signals governing pituitary induction, patterning and growth in mouse and zebrafish is Sonic Hedgehog (Shh) from the anterior-dorsal hypothalamus (Treier et al.,2001; Herzog et al.,2003; Sbrogna et al.,2003) (see Fig. 4G). In light of results obtained for other developmental processes such as limb/pectoral fin development, proposing a role of Fgf signaling to activate shh expression(Fischer et al., 2003), we wanted to investigate whether neurohypophyseal Fgf3 might regulate adenohypophyseal development via Shh. For this purpose, fgf3 mutant embryos were injected with shh mRNA. However, shh-injected fgf3 mutants still lacked endogenous pomc and prlexpression in the adenohypophyseal domain(Fig. 7B), indicating that Shh cannot compensate for the loss of Fgf3 and suggesting that Fgf3 acts in parallel to, rather than upstream of, Shh.

By contrast to the adenohypophysis itself, injected shh mRNA has a striking effect on adenohypophyseal gene expression in ectopic positions. In wild-type embryos, forced shh expression leads to ectopic pomc and prl expression in more lateral/posterior regions of the placodal ectoderm (Fig. 7A,C). These cells most probably represent trans-fated cells of the lense and otic placodes (Herzog et al., 2003) (note segmented pattern of prl expression in Fig. 7C). A similar ectopic induction of pomc and prl expression was also obtained in fgf3 mutants (Fig. 7B), indicating that in such ectopic locations, adenohypophyseal cells can specify independently of Fgf3 activity. They appear to require other Fgfs that act instead of, or in addition to, Fgf3. This is indicated by the effect of the Fgf receptor inhibitor SU5402, which, by contrast to fgf3 mutations, blocks both the endogenous and the Shh-induced ectopic prl expression to similar extents(Fig. 7D).

In Hedgehog signaling pathway mutants, the adenohypophysis is trans-fated to lens material (Kondoh et al.,2000; Varga et al.,2001). Several experiments were carried out to investigate the fate of non-specifying adenohypophyseal cells in fgf3 mutant embryos. At 24 hpf, the ANR region of mutant embryos appeared morphologically normal,despite the lack of lim3 expression(Fig. 4E,F). Using Nomarski optics, the adenohypophysis of wild-type embryos became visible as a distinct organ at approximately 25 hpf (Fig. 8A). An organ of similar size and shape, and at the same position,was also present in fgf3 mutants at 25 hpf(Fig. 8B). However, by 28 hpf and 29 hpf, the size of the adenohypophysis in mutants had become progressively smaller (Fig. 8D,H), while in wild-type siblings, it was similar to the size at 25 hpf (Fig. 8C,G).

To investigate the reasons for this regression of the organ in fgf3 mutants, we performed Acridine Orange stainings to detect apoptotic cells. At the 20-somite stage (19 hpf; 4 genotyped mutants) and at 24 hpf (5 genotyped mutants), apoptosis rates in the ANR region of mutant and wild-type sibling embryos were low and undistinguishable (data not shown). However, at 28 hpf, fgf3 mutants showed significantly more apoptotic cells in and around the shrunken adenohypophysis(Fig. 8D; 5 genotyped mutants)than wild-type siblings (Fig. 8C). The Acridine Orange signals outside the adenohypophysis could stem from either apoptotic cells of adjacent tissues or adenohypophyseal cells with debris extruded from the organ. To distinguish between these possibilities, we performed cell-tracing experiments, following the fate of single cells from the ANR region after labeling at tailbud stage(Fig. 8E,F). Live cells that contribute to olfactory epithelium, facial ectoderm and head mesenchyme were labeled with similar frequencies in mutant and wild-type sibling embryos(Fig. 8E,F). However, while most of the labeled cells in wild-type embryos gave rise to the adenohypophysis (Fig. 8E,G),not a single live adenohypophyseal cell was obtained in mutants at 28-30 hpf. Instead, cell debris was found in and around the disintegrating organ(Fig. 8F,H). Time course analyses showed that this debris - even that outside the pituitary gland -stems from adenohypophyseal cells that appear alive at 25 hpf(Fig. 8I) but become apoptotic and Acridine Orange-positive within the next hours(Fig. 8J). Together, these data indicate that in the absence of Fgf3 signaling, adenohypophyseal cells undergo apoptosis, while other cell types coming from the ANR region develop normally. Adenohypophyseal cell death starts a few hours after the failed transcriptional activation of lim3 and pit1, pointing to a possible correlation between cell specification and cell survival processes.

The different members of the family of Fgf ligands can be subgrouped according to sequence similarities and receptor specificities(Ornitz et al., 1996; Powers et al., 2000). Fgf3 mainly binds to the IIIb isoform of Fgf receptor 2 (Fgfr2IIIB), as does Fgf10. In mouse, mutants lacking Fgf10 (Ohuchi et al., 2000) or Fgfr2IIIb (Min et al., 1998; Sekine et al.,1999; De Moerlooze et al.,2000) display very similar defects during the formation of numerous organs, including the pituitary, suggesting that Fgf10 is the major ligand of Fgfr2IIIb. In comparison, Fgf3 deficient mice have rather subtle and more restricted defects during the late stages of tail and inner ear development (Mansour et al.,1993). Comparing the phenotypes of mutants in mouse and zebrafish,it appears that the functions of Fgf3 and Fgf10 have partly interchanged. During pituitary development, zebrafish Fgf3 seems to play the same role as mouse Fgf10. This is in contrast to pectoral fin/limb bud development, during which Fgf10 appears to be at play both in mouse and zebrafish(Min et al., 1998; Sekine et al., 1999; Ng et al., 2002).

The mouse infundibulum expresses at least three different Fgfs, Fgf10 of the Fgfr2IIIb-binding subgroup, and Fgf8 and Fgf18 of the IIIc group (Maruoka et al., 1998). The requirement of Fgf8 signaling for adenohypophyseal development has not been addressed as yet, due to early embryonic lethality of Fgf8 mutant embryos(Meyers et al., 1998; Sun et al., 1999). However, Fgf10 mutants are viable until birth, and their phenotype suggests an essential role of Fgf10 affecting all adenohypophyseal cell types before definitive pouch formation (E10) (Ohuchi et al., 2000). Furthermore, results obtained upon treatment of E10 pouch explants with the Fgf receptor inhibitor SU5402 suggest an additional later role of Fgf signaling to pattern the definitive pouch along its dorsoventral axis (Norlin et al.,2000). In zebrafish, our expression pattern analyses even suggest a third, much earlier, phase of Fgf signaling, which might occur during late gastrulation stages, when adenohypophyseal progenitor cells are located at the ANR and in close proximity to fgf3- and fgf8-expressing cells of the presumptive telencephalon(Fig. 3). Our rescue experiments via transplantation of wild-type cells or implantation of Fgf3 beads, as well as our experiments with temporally controlled blockage of Fgf signaling by SU5402, indicate that the later Fgf3 from the ventral diencephalon is absolutely necessary for adenohypophysis formation. Whether the earlier Fgf3 from the telencephalon is required, too, is less clear. The rescue of lim3 expression by diencephalic wild-type cells or late Fgf3 bead implants is only partial, never yielding lim3 expression in the entire adenohypophyseal domain. Similarly, early SU5402 treatment of embryos during gastrulation and early segmentation stages also leads to a strong reduction of adenohypophyseal marker gene expression, which, however,could be a secondary consequence of the effect of SU5402 on hypothalamus development. In sum, we can neither prove, nor rule out, an involvement of early telencephalic Fgf3 signaling on adenohypophysis development.

In addition to this general role affecting all pituitary cell types, Fgf3 signaling might also have a cell type-discriminating function to pattern the zebrafish adenohypophysis, similar to the proposed morphogenetic role of Fgf8 in mouse. As in mouse, the different cell types of the zebrafish adenohypophysis display a differential distribution along the dorsoventral axis of the anlage (Liu et al., 2003b; Nica et al., 2004), with dorsal cells being closer to the infundibulum than ventral cells. In addition,the zebrafish adenohypophysis is patterned along its antero-posterior axis(Herzog et al., 2003), with posterior cell types being closer to the infundibular Fgf3 domain than anterior cells. In conclusion, different adenohypophyseal cell types might indeed be exposed to different doses of Fgf3 signals. It is currently unclear whether this has any effects on the patterning of the adenohypophysis. This question cannot be addressed via regular analyses of fgf3 mutants,due to the earlier requirement of Fgf3 signaling for all adenohypophyseal cell types. Approaches allowing temporally controlled blockage of Fgf signaling have to be entertained, such as late applications of the inhibitor SU5402.

During mouse adenohypophysis development, Fgf signaling is supposed to have diverse effects on target cells, regulating organ morphogenesis and patterning, cell proliferation, cell survival and cell specification (see Introduction). Here, we investigated how these different effects might be correlated, studying the time course with which the different phenotypic aspects arise during adenohypophyseal development of fgf3 mutant zebrafish. First defects are concerned with early adenohypophysis-specific specification steps. While the general specification of the placodal ectoderm at the interphase of neural and non-neural ectoderm (10-12 hpf) (for a review,see Baker and Bronner-Fraser,2001) occurs normally in mutant embryos, later (18.5 hpf) they fail to initiate expression of the adenohypophyseal-specific marker genes lim3 and pit1 in the ANR, the anteriormost region of the placodal domain. This, however, neither affects the proliferation of adenohypophyseal progenitor cells nor further pituitary morphogenesis. At 25 hpf, fgf3 mutants display a pituitary gland of relatively normal size and shape. Also, we failed to detect any differences in BrdU incorporation studies (W.H. and M.H., unpublished data). However, we found dramatic apoptosis of adenohypophyseal cells after organ formation, starting at 25 hpf,and leading to a complete loss of the organ within 5 hours.

In conclusion, it appears that Fgf3 signaling primarily induces the activation of genes involved in early steps of adenohypophyseal specification,subsequently affecting cell survival. Pituitary morphogenesis and growth is affected even later, most probably as a consequence of the death of adenohypophyseal cells. However, it still remains largely unclear how Fgf3-induced cell specification and cell survival processes are correlated. pit1 and lim3 might be direct target genes of Fgf3 signaling, as their expression appears to be initiated shortly after the onset of fgf3 expression in the presumptive neurohypophysis (18 hpf versus 18.5 hpf). This would be consistent with the proposed role of Fgf8 in activating Lhx3 expression in mouse(Takuma et al., 1998; Ericson et al., 1998). In mouse, Pit-1 and Lhx3 have been shown to be directly involved in the transcriptional activation of hormone-encoding genes(Anderson and Rosenfeld, 2001; Lamolet et al., 2001; Liu et al., 2001), explaining why zebrafish fgf3 mutants fail to initiate prl and pomc expression at 24 hpf. It remains unclear whether the subsequent death of adenohypophyseal cells between 25 hpf and 30 hpf is a default consequence of their failed specification/differentiation, or whether Fgf3 activates the transcription of additional genes, particularly regulating cell survival.

In midsegmentation zebrafish embryos, such survival factors appear also to be present in other tissues, where they are generated in response to other or additional Fgf signals. This is indicated by our observation that Shh can induce the formation of viable lactotropes and corticotropes in ectopic placodal locations in an Fgf3-independent, but SU5402-sensitive, manner. According to their position, the intermediate domains of ectopic prl-positive cells in shh-injected embryos most probably represent trans-fated lense precursor cells, which might survive due to the redundant function of Fgf3 and Fgf8 from the optic stalk. In shh-injected embryos, this optic stalk is laterally enlarged(Macdonald et al., 1995), and therefore close to the trans-fated placodal lense precursor cells.

The shh overexpression studies further show that Fgf3 does not simply act via an activation of shh, as Shh is not capable of rescuing pituitary cell types in the adenohypophyseal region itself. Rather,Fgf3 and Shh appear to act in parallel and to have rather distinct effects. Thus, ectopic Shh can induce the expression of adenohypophyseal genes in ectopic locations, whereas Fgf3 cannot(Fig. 7; W.H. and M.H.,unpublished observations). Furthermore, Shh has mitogenic activity, indicated by the reduced growth of the adenohypophysis in shh mutants without any sign of increased cell death, whereas the adenohypophysis of fgf3mutants shrinks due to loss of cells by apoptosis.

We are very grateful to the other members of the Freiburg screening group(M. Schorpp, D. Diekhoff, T. Franz, M. Leicht, E. Nold, T. Nolting, C. Riegger, W. Wiest), and of the Tuebingen screen consortium (MPI for Developmental Biology: F. van Bebber, E. Busch-Nentwich, R. Dahm, H.-G. Frohnhöfer, H. Geiger, D. Gilmour, S. Holley, J. Hooge, D. Jülich,H. Knaut, F. Maderspacher, H.-M. Maischein, C. Neumann, T. Nicolson, C. Nüsslein-Volhard, U. Schönberger, C. Seiler, C. Söllner, M. Sonawane, A. Wehner, C. Weiler; Exelixis Germany GmbH: P. Erker, H. Habeck, U. Hagner, C. Hennen, E. Kaps, A. Kirchner, T. Koblitzek, U. Langheinrich, C. Loeschke, C. Metzger, R. Nordin, J. Odenthal, M. Pezzuti, K. Schlombs, J. deSatana-Stamm, T. Trowe, G. Vacun, B. Walderich, A. Walker, C. Weiler) for carrying out the screen during which the lia alleles were isolated. We are also very grateful to Filippo Del Bene and Jochen Wittbrodt for teaching us the bead implantation technique. M.H. thanks Thomas Boehm, Z.M.V. thanks Wolfgang Driever, and H.H.R. thanks Christiane Nüsslein-Volhard for support and discussions. We are grateful for the help and technical support from Roland Nitschke and the Life Imaging Facility of the SFB 592 (Z2)during confocal microscopy. Many thanks also to Michael Brand (ace, fgf8,krox20), Chi-Yao Chang (gh), Igor Dawid (lim3), Mark Ekker (dlx2, dlx3), Christine Petit (eya1), Corinne Houart(emx1), Phil Ingham (shh), Klaus Rohr (nkx2.1), and Bernard and Christine Thisse (fgf3, hgg, spry4) for mutants and plasmids. Work in the laboratories of M.H. and H.H.R. was supported by the Max-Planck Society, work in Z.M.V.'s laboratory by the Sonderforschungsbereich SFB 592 (A5).