pitx3 defines an equivalence domain for lens and anterior pituitary placode

The pituitary gland is a key regulator of hormone secretion in vertebrates. The anterior lobe of the pituitary contains several cell types distinguished by their morphologies, locations and secreted hormones(Bentley, 1998; Dasen and Rosenfeld, 1999a; Dasen and Rosenfeld, 1999b). In mammals and birds, the anterior pituitary forms from Rathke's pouch, an invagination of oral ectoderm beneath the ventral forebrain, whereas in zebrafish, the anterior pituitary derives from a placode that forms between the anterior hypothalamus and ventral ectoderm(Glasgow et al., 1997). Although several studies analyzed anterior pituitary precursor cells(Houart et al., 1998; Kozlowski et al., 1997; Whitlock and Westerfield,2000), the precise origin of precursor cells is not well understood. Most studies suggest a dual origin for anterior pituitary; it derives either from the anterior neural ridge or from the non-neural ectoderm(Couly and Le Douarin, 1988; Dasen and Rosenfeld, 1999b; Eagleson et al., 1995; Eagleson and Harris, 1990; Kawamura and Kikuyama, 1992; Kondoh et al., 2000; Sheng and Westphal, 1999).

Several transcription factors demarcate the presumptive pituitary and lens placodal field. Presumptive lens ectoderm in Xenopus expresses Otx2, Pax6 and Sox3, and in zebrafish, expression of dlx3b demarcates olfactory(Whitlock and Westerfield,2000; Zygar et al.,1998) and otic placode precursor cells(Akimenko et al., 1994; Liu et al., 2003; Solomon and Fritz, 2002). We recently showed that dlx3b, dlx4b, sox9a and fibroblast growth factor(Fgf) signaling interact to specify the zebrafish otic placode(Liu et al., 2003). Because dlx3b and dlx4b are also expressed around the anterior border of the neural plate, these factors might also play a role in formation of the anterior pituitary.

Analysis of zebrafish Hedgehog pathway mutants, you-too(gli2) (Karlstrom et al.,1999), iguana (Kondoh et al., 2000) and smu (smoothened)(Varga et al., 2001), suggests that Hedgehog signaling is required for anterior pituitary placode specification. The observations that zebrafish Hedgehog pathway mutants form ectopic lenses at the expense of pituitary and that early Rathke's pouch expresses lens Delta-crystallin in chick suggest that these apparently unrelated tissues might arise from common precursors and that Hedgehog may be required to specify the pituitary lineage(Karlstrom et al., 1999; Kondoh et al., 2000; Varga et al., 2001). Other studies suggest that pituitary and lens cells `transdifferentiate' in Hedgehog pathway mutants (Kondoh et al.,2000). In zebrafish, overexpression of Shh suggests that Hedgehog signaling is sufficient to pattern the anterior pituitary and induce pituitary specific gene expression (Herzog et al.,2003; Sbrogna et al.,2003). In mouse, Shh is thought to regulate cell proliferation and cell-type specification (Treier et al.,2001) indirectly (Takuma et al., 1998).

Several homeodomain transcription factors may control specification of pituitary cell types in a combinatorial manner and are likely targets of signaling interactions (Dasen and Rosenfeld, 1999a). Lhx3, for example, is expressed initially in all cells of Rathke's pouch, is required for proliferation of ventral cell types (Bach et al., 1995; Zhadanov et al., 1995), and controls gene expression in thyrotrope and gonadotrope cells in synergy with recently identified Pitx genes (Bach et al., 1997; Tremblay et al.,1998). The Pitx family members belong to a Bicoid-related subclass of homeobox genes. Pitx1 binds to and transactivates a cis-acting element required for proopiomelanocortin (Pomc)(Lamonerie et al., 1996) and other pituitary-specific gene expression, including Lhx3 and prolactin (Szeto et al.,1999; Tremblay et al.,1998). In mouse, lens and mesencephalon, but not anterior pituitary, express Pitx3. Mutations in Pitx3 lead to dominant cataracts and malformations of the anterior eye mesenchyme and its derivatives (Nunes et al.,2003; Semina et al.,2000). In Xenopus, stomodeum, lens and pituitary express Xpitx3 (Pommereit et al.,2001).

To understand whether Hedgehog acts directly on pituitary and lens placode precursor cells and whether Hedgehog is sufficient to induce placode formation, we used lineage tracers to define the locations and gene expression patterns of pituitary and lens precursor cells precisely. We then analyzed placode specification in Hedgehog gain- and loss-of-function experiments.

Our results indicate that by the end of gastrulation, pitx3expression demarcates an equivalence domain of cells that gives rise to lens and pituitary. This domain overlaps partially with the dlx3b and dlx4b expression domains at the anterior neural plate border. pitx3 is required at bud stage for pituitary pre-placode formation,whereas dlx3b restricts pituitary placode size during somitogenesis. During mid-somitogenesis, Hedgehog signaling is necessary for pituitary placode specification and is sufficient to induce expression of pituitary genes in pitx3 expressing pre-placode. Hedgehog signaling is also sufficient to block pitx3 expression in presumptive lens precursors and subsequent lens tissue differentiation. We suggest that a non-Hedgehog signal initially induces an unspecified pitx3 expressing pre-placodal field around the anterior border of the neural plate. Later, during mid-somitogenesis Hedgehog signaling specifies pituitary characteristics in the placode that forms from the midline of the pre-placode.

Wild-type and smu^b641 mutant(Varga et al., 2001) zebrafish(Danio rerio) embryos were obtained and maintained with standard procedures (Westerfield, 2000)at 28.5°C. Developmental stages described by standard terms(Kimmel et al., 1995) or hours post fertilization (h). Gene symbols according to ZFIN(http://zfin.org).

Mouse and human Pitx2 homeodomain sequences were used to design degenerate PCR primers: PitxF1, 5′ CYAGCCAGCAGCTSCASGAGCTGGA 3′; PitxR1,5′ AGGCCCTTGGCDGCCCARTTGTTGTA 3′. These primers amplified an∼275 bp fragment from a 16- to 18-somite zebrafish cDNA library (B. Appel)that we cloned and used as probe to screen the cDNA library. A full-length cDNA was obtained and subcloned into pCS2+ vector (Accession Number AY525643). Parsimony and nearest neighbor joining phylogenetic analyses were performed with Paup (Sinauer Associates), ClustalX(Thompson et al., 1997)(ftp://ftp-igbmc.u-strasbg.fr/pub/ClustalX/)and NJPlot software (M. Gouy, UMR CNRS 5558, Université Lyon).

To predict the locations of cells relative to morphological landmarks and gene expression domains in live, unlabeled embryos, we performed double in situ hybridization on 10-15 size-selected (590±5 μm diameter),staged embryos as described (Varga et al.,1999) and averaged gene expression domains of pitx3 and dlx3b. Single-cell labeling with vital fluorescent dyes and cell fate mapping were conducted (Varga et al.,1999) using a 10×10 grid reticule (19 mm diameter, KR406;Zeiss) in the ocular (10×) of a fixed stage microscope (Zeiss Axioscope)and a differential amplifier (A-M Systems, Neuroprobe Amplifier 1600). At prim-5 stage (24 h), the descendants of the labeled progenitor cells were analyzed using fluorescence and confocal laser scanning microscopy (Zeiss LSM 510). Serial 1.6 μm optical sections (0.5 μm overlap between consecutive sections) were projected onto 2D images using Zeiss Software.

Embryos were fixed and hybridized with one or two mRNA probes(Hauptmann and Gerster, 1994; Varga et al., 1999). We used mRNA probes for pitx3 and dlx3b as landmarks for the prospective placodal field and mRNA probes for pomc, ptc1, ptc2, nkx2.2,lhx3, bmp2b, shh, eyes absent1 (eya1), foxb1.2 and pax6a. Fast Red (Boehringer, Sigma) was the substrate for alkaline phosphatase color reaction to detect one mRNA probe and the lineage tracer and NBT and BCIP (Boehringer Mannheim) were used to detect the other mRNA probe. This allowed us to detect fluorescein labeled mRNA probe and fluorescein-dextran injected cells at the same time. We labeled lens fiber cells with zl-1 (Trevarrow et al.,1990) monoclonal antibody (Zebrafish International Resource Center) as described (Varga et al.,2001).

To label groups of precursor cells, we injected one-cell stage embryos with 1% DMNB-caged fluorescein (Molecular Probes) in 0.2 M KCl. At bud stage, we uncaged cells in regions of the dlx3b and pitx3 expression domains that include lens or pituitary precursors. UV light directed through a pinhole diaphragm, ∼40 μm diameter or larger, was used for photoactivation and we detected uncaged fluorescein at prim-5 stage using fluorescence and confocal microscopy (Zeiss LSM510), or antibody labeling in combination with in situ hybridization as described above.

To study the roles of pitx3, dlx3b and dlx4b in pituitary and lens formation, we injected 5-8 ng of morpholino antisense oligonucleotides into one-cell stage embryos: pitx3 morpholino 5′-AGGTTAAAATCCATCACCTCTACCG-3′; dlx3b morpholino,5′-ATGTCGGTCCACTCATCCTTAATAA-3′; dlx4b morpholino,5′-GCCCGATGATGGTCTGAGTGCTGC-3′(Liu et al., 2003). Per embryo, we injected ∼8-10 ng of 8 μg/μl (in 0.2 M KCl) pitx3 morpholino; 5-7 ng each dlx3b and dlx4bmorpholinos; and 5 ng each pitx3, dlx3b and dlx4bmorpholinos. Per embryo, we injected pitx3:eGFP (600 pg); 5Δpitx3:eGFP (600 pg); shh (10-100 pg); and dominant-negative Protein Kinase A(dnPKA:GFP, 1 ng). mRNAs were synthesized in vitro (mMessage Machine Kits, Ambion). The embryos were analyzed morphologically at bud (10 h), prim-5 and long-pec (48 h) stages, and fixed in 4% paraformaldehyde. We analyzed 16 μm cryosections to identify cells doubly labeled with photoactivated fluorescein and lim3 mRNA probe.

To ensure that the pitx3 morpholinos block translation of pitx3 mRNA, we co-injected pitx3 mRNA (see Fig. S1A-D in the supplementary material), pitx3:eGFP mRNA that encodes the normal 5′ UTR of pitx3 fused to the eGFP open reading frame(Fig. S1E-G), or a 5′UTR that contains five modified bases(5Δpitx3:eGFP) (see Fig. S1H,I in the supplementary material).

Wild-type embryos were injected with dnPKA:GFP mRNA and rhodamine-dextran (2.5%). The dnPKA mRNA and rhodamine-injected embryos were used as donors at dome stage and we transplanted 10 to 20 cells from the animal pole to wild-type host animal poles. Donor and host embryos were maintained as pairs in 24-well dishes in penicillin and streptomycin (1%each) in Danieau's solution. At prim-5 stage, donors were analyzed for loss of lens phenotype that resulted from dnPKA expression; hosts were analyzed only if the donor sibling lacked both lenses.

We cloned a novel zebrafish Pitx gene and studied its phylogenetic relationship, temporal and spatial expression patterns, and its role in pituitary and lens development. Nearest neighbor-joining analysis of the amino acid sequence of the full-length protein shows that the gene encodes zebrafish Pitx3. Like Pitx2, zebrafish Pitx3 is more closely related to XenopusPitx3 than to mammalian Pitx3 orthologs(Fig. 1A). Analysis of the homeodomain amino-acid sequence further supports this assignment: zebrafish, Xenopus and mouse Pitx3 homeodomains are 100% identical, the human homeodomain ortholog is 98% identical. The zebrafish Pitx3 homeodomain is less similar to the Pitx1 and Pitx2 homeodomains of other vertebrates: 98%identical to Xenopus Pitx1, and 97% identical to mouse and human Pitx1 homeodomains.

Cells at the anterior neural plate border and, later, lens and pituitary cells express pitx3. We detect zebrafish pitx3 expression by 95% epiboly in anterior neural plate border cells(Fig. 1B). At bud stage, the pitx3 expression domain resembles two inverted u-shaped regions stacked on top of each other (Fig. 1C). These pitx3-expressing cells are located in a row that is two to four cells wide at the transition between non-neural ectoderm and prospective placode (anterior to the polster) and a row at the presumptive transition between prospective placode and neural plate (posterior to the polster). Farther posteriorly and laterally, presumptive lens placode precursors express pitx3. Ectodermal cells superficial to the polster(between the two pitx3-expressing rows) express pitx3 at low levels (Fig. 1C, asterisk; Fig. 2C,D). At the five-somite stage, pitx3 expression is downregulated in most of the anterior expression domain and only small patches of expression remain(Fig. 1D). At the ten-somite stage, pitx3 is upregulated in a region that presumably corresponds to the border that later forms between forebrain and midbrain(Fig. 1E). In addition, pitx3 is expressed at this stage in the border between non-neural ectoderm and prospective placodes, in presumptive first branchial arch, in ventral head mesenchyme, and between the eye primordia and diencephalon. We also find pitx3 expression in prospective lens ectodermal cells on the surface of the eye primordia (Fig. 1E). At the 18-somite stage (18 h), pitx3 expression becomes more prominent in lens epithelial precursors and in bilateral expression domains at the anterior end of the forebrain that correspond to previously described lim3 expression in presumptive pituitary placode(Fig. 1F, arrowheads)(Glasgow et al., 1997; Herzog et al., 2003; Sbrogna et al., 2003). From the 21-somite stage (19.5 h) through prim-10 stage (27 h), lens, pituitary,head mesenchyme and ventral diencephalic cells express pitx3(Fig. 1G-J) as well as mesenchymal cells of the presumptive first branchial arch(Fig. 1K,L). In trunk somites,presumptive muscle pioneers express pitx3(Fig. 1M). At long-pec (48 h)and protruding-mouth (72 h) stages, ventral diencephalic cells and lens fiber cells downregulate pitx3 expression(Fig. 1N). In addition, we find that presumptive nucleus coeruleus cells in the first rhombomere(Holzschuh et al., 2001) and tegmental cells express pitx3(Fig. 1O). At protruding-mouth(72 h) stage, pituitary cells and the primordia of the first and second mandibular arches express pitx3(Fig. 1P).

To test whether pitx3 expression demarcates pituitary and lens precursors at bud stage, we analyzed the fate of pitx3-expressing cells using embryonic morphology and the pitx3 and dlx3bexpression domains as landmarks for single cell lineage analysis. We correlated the expression domain of pitx3 with zic1, foxb1.2,pax6a (Fig. 2A-C) and otx2 (not shown) that we used previously to fate map the anterior neural plate (Varga et al.,1999). Because Pax6 has been shown to regulate both lens and eye development (Ashery-Padan and Gruss,2001; Kondoh,1999; Zygar et al.,1998), we did not use pax6a and pax6b as neural plate landmarks or in functional assays. However, because dlx3b has been shown to demarcate prospective olfactory(Kondoh, 1999; Whitlock and Westerfield,2000; Zygar et al.,1998) and otic placode cells(Akimenko et al., 1994; Liu et al., 2003), we compared pitx3 expression with dlx3b expression(Fig. 2E,F) and find that the expression domains overlap significantly(Fig. 2F).

To understand the spatial relationship between Hedgehog-producing cells and placodal precursors, we analyzed where pitx3-positive cells are located relative to median shh-expressing cells. Because polster cells express hgg1 at bud stage just beneath the superficial pitx3-expressing ectoderm (Fig. 2D), we also used hgg1 expression to correlate the locations of median neural plate border cells, ventral ectoderm and shh expression. At bud stage, we find that prechordal plate cells and cells in the midline of the neural plate express shh(Fig. 2G). Thus, the median ectodermal cells that express pitx3 near the polster, are near Hedgehog-secreting cells in the median neural plate(Fig. 2G), suggesting that these cells are in a position to receive significant levels of Hedgehog signal. We also analyzed other genes that demarcate ventral ectoderm and the border of the neural plate such as bmp2b(Fig. 2G) and eya1(Fig. 2H). However, we did not use the eya1 expression domain to establish the pituitary and lens fate map, because eya1 expression at the anterior neural plate border overlaps with the dlx3b expression domain(Fig. 2H) that gives rise to olfactory placode (Whitlock and Westerfield, 2000). Although eya1 is expressed in the olfactory placode at prim-5 (24 h) stage, it is downregulated in lens during late somitogenesis (Sahly et al.,1999).

We have previously shown that smoothened mutant embryos that cannot transduce Hedgehog signals lack pituitary and instead form ectopic lenses (Varga et al., 2001)(see Fig. S2 in the supplementary material). This could occur because median pituitary precursors are mis-specified and form lens at the expense of pituitary or, alternatively, because lens precursors aberrantly invade the midline.

To distinguish between these possibilities, we fate mapped the cells that give rise to pituitary and lens in wild-type and smoothened mutant embryos. We labeled cells at bud stage, using pitx3 and dlx3b as landmarks for lens and pituitary precursors, and analyzed their progeny at prim-5 stage.

In wild-type embryos, we find that pitx3- and dlx3b-expressing cells at the midline of the anterior neural plate border contribute to pituitary (Fig. 3B, and red dots in 3A) or olfactory placodes(Fig. 3A, yellow dots). Precursors closer to non-neural ectoderm mainly form pituitary cells(Fig. 3A, red dots) or ventral ectoderm (Fig. 3A, green dots),and cells closer to the neural plate have a greater tendency to contribute to olfactory placode (Fig. 3A,yellow dots).

In general, we find that single precursor cells at bud stage give rise to progeny with a single cell fate (e.g. lens, pituitary or olfactory cells),indicating a high degree of cell specification(Fig. 3B). However, we find that these separate populations of precursor cells intermingle even though they are individually specified for one fate(Fig. 3A; single-colored dots). In only a few cases, did we observe dual (or split) cell fates of daughter cells (for example olfactory and pituitary) derived from a single precursor(Fig. 3A,C; two-colored dots).

Lens cells derive from lateral positions in the pitx3 and dlx3b expression domains (Fig. 3A, blue dots). They intermingle with precursors of non-neural ectoderm on the ventral side (Fig. 3A, green dots) and with olfactory placode precursors closer to the neural plate (Fig. 3A,yellow dots). Thus, we find that lens precursors(Fig. 3A, blue dots) are usually farther from the neural plate and express pitx3, whereas dlx3b-expressing cells closer to the neural plate mainly contribute to olfactory placode (Fig. 3A,yellow dots). The large degree of intermingling of these two placodal cell fates is consistent with the overlap of the two gene expression domains and indicates that placodal fields have not yet resolved by this developmental stage (Whitlock and Westerfield,2000).

In smoothened mutant embryos, lateral pitx3- and dlx3b-expressing cells also give rise to lens(Fig. 3E,F). In the midline,however, in the region that corresponds to the location of pituitary precursors in wild-type embryos (Fig. 3A,B), precursors (dark blue dots) give rise to daughter cells in ectopic (Fig. 3G), distorted(Fig. 3H) and fused lenses(Fig. 3I). Thus, in the absence of Hedgehog signal transduction, median placode precursors are mis-specified and form lens instead of pituitary.

We tested whether lens precursors invade the midline of the anterior neural plate border in smoothened mutant embryos by photoactivating DMNB-caged fluorescein in lens precursors in wild types and mutants. Because large numbers of cells are labeled, this approach allows us to track cell movements that might have been missed previously with single cell labeling. When we uncage fluorescein in the anterior midline of the pitx3expression domain at bud stage (Fig. 4A,B), we find that pituitary and skin are labeled at prim-5 stage in wild-type embryos (Fig. 4E;arrowhead). In smoothened mutant embryos, however, when we uncage in the anterior midline, the ectopic median lens(Fig. 4F; arrowhead) and skin are later labeled (Fig. 4B,F). When we uncage in cells farther lateral in the pitx3 and dlx3b expression domain (Fig. 4C,D), we find that in both wild-type and smoothenedmutant embryos, skin, lens, olfactory placode, telencephalon and retina are labeled later (Fig. 4G,H). Typically, almost the entire retina associated lens is labeled in wild-type and mutant embryos, whereas in smoothened mutant embryos, the ectopic median lens is never labeled (Fig. 4H; arrowhead). Our results indicate that in smoothenedmutant embryos, lens precursors do not invade the midline aberrantly and that apparently the entire pitx3 expression domain gives rise to lens. Thus, lateral cells contribute to the retinal lens and median cells that form pituitary in wild-type embryos are mis-specified and form ectopic lens in smoothened mutants.

To analyze the roles of pitx3, dlx3b and dlx4b in pituitary development, we injected pitx3, dlx3b and dlx4bmorpholinos and tested for expression of pituitary marker genes such as lim3, a pan-pituitary marker(Glasgow et al., 1997), the Hedgehog target gene nkx2.2 and pomc(Herzog et al., 2003), a gene expressed in corticotropes and melanotropes. We find that `knock-down' of pitx3 causes reduction or complete loss of lim3, nkx2.2 and pomc expression in pituitary (Fig. 5A-F), indicating that pitx3 is required for pituitary cell specification. We find that pitx3 is required in a dose-dependent manner; high doses of morpholinos (8-10 ng/embryo) lead to complete loss of pituitary placode (Fig. 5D-F), whereas less morpholino (2-4 ng/embryo) leads to a smaller placode, and variably reduced or lost expression of lim3, nkx2.2 and pomc (Fig. 5D, inset). By contrast, injection of dlx3b and dlx4b morpholinos increases both pituitary placode size and cell number(Fig. 5G-I), indicating that dlx3b and dlx4b function as negative regulators in pituitary placode development. To test epistatic relationships between pitx3and dlx3b-dlx4b, we injected embryos with all three morpholinos. We find that expression of lim3, nkx2.2 and pomc is reduced or entirely lost at prim-5 stage(Fig. 5J-L), comparable with embryos injected with pitx3 morpholinos alone. Together, these observations suggest that pitx3 is epistatic to dlx3b and dlx4b, and that Dlx3b-Dlx4b is required to restrict pituitary cell number and placode size. By contrast, Pitx3 is required for pituitary placode formation and cell type specification.

To learn when pitx3, dlx3b and dlx4b are required, we injected embryos with pitx3 or dlx3b and dlx4bmorpholinos and analyzed pitx3 and dlx3b expression at bud,18-somite and prim-5 stages. Injection of pitx3 morpholino strongly reduces pitx3 expression in prospective lens and pituitary precursors at bud stage (Fig. 6B), and at 18-somite and prim-5 stages (Fig. 6E,H), pitx3 expression is reduced in pituitary pre-placode and placode, but not in head mesenchyme or ventral diencephalon. dlx3b and dlx4b morpholino injection reduces pitx3expression in lateral, prospective lens precursors at bud stage(Fig. 6C, arrowheads), but not in prospective pituitary precursors. In older, dlx3b and dlx4b morpholino-injected embryos, however, we find expanded pitx3 expression in pituitary pre-placode(Fig. 6F), pituitary placode and in ectopic cells close to the olfactory placode(Fig. 6I, arrowheads). By contrast, injection of pitx3 morpholinos did not affect dlx3b expression between bud and prim-5 stages (not shown). Injection of dlx3b and dlx4b morpholinos changes dlx3bexpression, however, consistent with previous results(Solomon and Fritz, 2002). Thus, our results suggest that at the end of gastrulation, pituitary precursors require Pitx3 function for pitx3 expression and subsequent pituitary pre-placode formation, whereas Dlx3b and Dlx4b act on lens precursors at bud stage and on the median pre-placode later, during somitogenesis.

Because partially overlapping domains of cells express pitx3 and dlx3b at bud stage, we also tested whether these genes function in lens formation. Our results indicate that pitx3 and dlx3b(and dlx4b) are required for primary lens fiber cell specification but not for lens placode formation (see Fig. S3 in the supplementary material).

Loss of Hedgehog signal transduction results in mis-specification of pituitary precursor cells and subsequent ectopic lens formation(Fig. 3)(Karlstrom et al., 1999; Varga et al., 2001). Conversely, overexpression of shh suppresses lens formation(Barth and Wilson, 1995) and increases the number of ventral, anterior pituitary cell types(Herzog et al., 2003; Sbrogna et al., 2003). Consistent with these previous observations, we find that increasing amounts(10-100 pg/embryo) of synthetic shh mRNA block lens tissue formation(Fig. 7A-D) and, in parallel,downregulate lens pitx3 expression(Fig. 7E-G,I-K). Although pitx3 expression expands in branchial arches(Fig. 7K) and into the epithalamus (not shown), the size of the pituitary placode as such is unaffected (Fig. 7I-K, insets). Interestingly, at high levels of shh mRNA (100 pg/embryo),prospective ventral ectoderm at bud stage(Fig. 7H) and cells covering the retina at prim-5 stage (Fig. 7L) upregulate pitx3 expression in a pattern reminiscent of normal bud stage pitx3 expression(Fig. 7E, compare with Fig. 7H,L). This observation suggests that pitx3-expressing cells respond to Hedgehog differentially, depending on their position in the embryo and the source of Hedgehog signal. During normal development, median cells might acquire pituitary characteristics because of their proximity to Hedgehog signal,whereas in lateral regions absence of Hedgehog signal might favor lens specification. Thus, at the end of gastrulation, pitx3-expressing precursor cells might be competent to respond to Hedgehog signal directly, or by Hh induced, secondary signaling interactions.

Because highest shh concentrations caused ectopic pitx3expression at prim-5 stage, we examined whether Hedgehog also induces pituitary genes in pitx3-expressing precursors. In embryos injected only with high amounts of shh mRNA (100 pg/embryo), we find that both lim3 and pomc are expressed ectopically between retina and epidermis (Fig. 7N,Q,R), as previously reported in similar studies(Herzog et al., 2003; Sbrogna et al., 2003). Co-injection of shh mRNA (100 pg per embryo) and pitx3 MO (8 ng per embryo) completely eliminates lim3 expression from median and ectopic locations (Fig. 7O)suggesting that pitx3 is required for pituitary cell specification when Hedgehog is overexpressed. In response to high levels of Hedgehog,ectopic pomc-expressing cells are scattered on top of the forebrain and retina (Fig. 7Q,R) and in some embryos, form small, flattened clusters of cells that resemble miniature lenses (Fig. 7R, arrowheads). This observation supports the idea that lens precursors form in their normal positions but express pomc in response to shh.

We further tested the possibility that Hedgehog induces pituitary gene expression in lens precursors by co-injecting caged-fluorescein and shh mRNA (100 pg per embryo) into one-cell stage embryos. We uncaged fluorescein in prospective lens and skin precursors(Fig. 7S) and analyzed these embryos by in situ hybridization with probes for lim3 and by antibody labeling to detect caged-fluorescein. We find that shh mRNA-injected embryos lack lens and have ectopic lim3-expressing cells near the eyes doubly labeled with fluorescein (16/16 embryos; Fig. 7T, arrowheads). Because lens is entirely absent in these embryos, doubly labeled cells indicate that(at least some) lens/nasal precursors express pituitary genes in response to shh mRNA injection. Thus, loss of lens tissue and ectopic pitx3,lim3 and pomc expression suggest that Hedgehog signaling suppresses lens development and induces gene expression characteristic of the pituitary cell lineage in derivatives of pitx3 expressing lens precursors.

To test whether pitx3-expressing cells are competent to respond to Hedgehog and, as a result, express genes characteristic of pituitary, we analyzed whether shh induces ptc genes in pituitary or lens precursors. At bud stage, ptc1 is expressed in median neural plate cells (Fig. 8A,C), but not in pitx3-expressing cells at the anterior neural plate border that give rise to lens or pituitary. In embryos injected with shh mRNA (100 pg/embryo), pituitary and lens precursors that express pitx3, but not dlx3b, do not upregulate ptc1(Fig. 8B,D). By contrast, cells throughout the neural plate and dlx3b-expressing neural plate border cells, upregulate ptc1 in response to shh injection(Fig. 8D). Thus, even though Hedgehog induces pituitary gene expression, is required for pituitary specification, and is sufficient to block lens formation, placodal precursor cells that express only pitx3, do not express ptc1 and ptc2 (data not shown) in response to shh mRNA injection.

Hedgehog patterns the pituitary placode(Herzog et al., 2003; Sbrogna et al., 2003) during mid-somitogenesis when pituitary and lens placodes form. However, the observations that shh overexpression fails to upregulate ptcexpression in pituitary precursors at bud stage or to affect median placode size suggest that Hedgehog acts indirectly on pituitary and lens precursors at bud stage. To test this idea, we injected one-cell stage donor embryos with biotin-rhodamine dextran and dnPKA that activates the Hedgehog signaling cascade cell autonomously. We then transplanted groups of cells from these donor embryos into uninjected blastula stage hosts. We find that ptc1 is upregulated before the 18-somite stage in transplanted cells on top of the eye vesicle (Fig. 8E). However, at prim-5 stage, we find that donor derived cells that express dnPKA contribute to lens(Fig. 8F) and do not express pituitary genes (not shown), even though all corresponding donor embryos completely lack lens. We also find that dnPKA-expressing cells contribute to lens, regardless of whether large groups(Fig. 8F) or only a few cells(not shown) have been transplanted. Thus, even though overexpression of Hedgehog blocks lens formation and induces ectopic pituitary gene expression(Fig. 7), cell-autonomous activation of the Hedgehog signaling cascade does not induce pituitary marker genes or prevent cells from contributing to lens. Because injected mRNA is degraded during the first day of development, this result supports the hypothesis that Hedgehog signaling acts indirectly on lens precursors during early stages of development. Presumably, block of lens formation and induction of ectopic pituitary cells in shh mRNA-injected embryos is an event downstream of Hedgehog signaling due to changes in other cellular interactions and signaling systems regulated by Hedgehog.

We suggest a new model for specification and patterning of the anterior cranial placodes that form the pituitary, nose and lens. Initially, precursor cells for these placodes occupy overlapping fields demarcated by overlapping pitx3- and dlx3b-expression domains at the neural plate border (Fig. 2). Our results indicate that Pitx3-expressing, non-neural ectoderm cells at bud stage later give rise mainly to lens and anterior pituitary placodes(Fig. 3). Pitx3 is required for its own expression and also for pituitary placode formation (Figs 5, 6). By contrast, Dlx3b and Dlx4b are mainly expressed in olfactory precursor cells in this region(Whitlock and Westerfield,2000) and function to limit the size of the pituitary placode(Figs 5, 6). Both Pitx3 and Dlx3b-Dlx4b are required for pituitary and lens specification and differentiation(Fig. 5; see Fig. S3 in the supplementary material). Our results support the idea that pitx3demarcates an equivalence domain, because in the absence of Hedgehog signaling, median Pitx3-expressing cells are mis-specified and contribute to ectopic or medially fused lenses (Kondoh et al., 2000; Sbrogna et al.,2003; Varga et al.,2001). In addition, we show that Hedgehog signaling induces genes in lens precursors characteristic of pituitary placode, further supporting the view that pitx3-expressing precursor cells have equivalent potential to form lens or pituitary (Figs 7, 8). Our results also indicate that cell fate depends on the level of Hedgehog signal, even though lens and pituitary precursors are not competent to respond to Hedgehog at the end of gastrulation (Figs 7, 8). Pituitary precursors do not express Ptc genes at bud stage, and lens precursors form lens in spite of cell-autonomous activation of the Hedgehog signaling pathway. Thus,overexpression of Hedgehog might lead to activation of other mechanisms that affect patterning and specification of the placodal field. We suggest that initially an unspecified placodal field is formed, in a Hedgehog independent manner, mediated by Pitx3, Dlx3b and Dlx4b interaction. The placodal field is subsequently subdivided into an outer, pitx3-expressing pre-placode with lens epithelial cell character and an inner, dlx3b-dlx4b-expressing olfactory pre-placode(Fig. 9). During mid-somitogenesis, Hedgehog is sufficient and required to induce pituitary genes in the outer pre-placode, and thus directs it to form median pituitary and lateral lens placodes.

Zebrafish pitx3 may have subsumed the functions of murine Pitx1 and Pitx3. In mice, non-neural ectoderm and its derivatives, including olfactory placodes, Rathke's pouch and subsequently all endocrine adenohypophyseal cells, express Pitx1(Lanctôt et al., 1997; Gage et al., 1999). Pitx1 expression is required for development of Rathke's pouch and activation of pituitary pomc and prl expression(Szeto et al., 1999; Tremblay et al., 1998). However, midbrain dopaminergic neurons, eye mesenchyme and lens, but not pituitary, express Pitx3. Mouse aphakia mutants and mice lacking functional Pitx3 promoter elements fail to develop differentiated lenses (Nunes et al.,2003; Semina et al.,2000) but have no apparent pituitary defects. Consistent with these studies, human fetal mesencephalon, but not anterior pituitary,expresses PITX3(Pellegrini-Bouiller et al.,1999). By contrast, the Xenopus ortholog, Xpitx3, is expressed in presumptive stomodeum, lens and pituitary(Pommereit et al., 2001); and our results show that, in zebrafish, pitx3 is also expressed in prospective lens and pituitary and that embryos injected with pitx3morpholinos fail to differentiate primary lens fiber cells or pituitary placode (Figs 5, 6).

Our results provide a genetic mechanism to explain the placode forming competence of non-neural ectoderm. In Xenopus, apparently all non-neural head ectoderm, including the midline, has a transient lens-forming bias (Henry and Grainger,1987; Zygar et al.,1998). We show that median placodal cells are mis-specified and give rise to lens in smoothened mutant embryos(Fig. 3). This suggests that the pitx3 expression domain demarcates an equivalence domain at the end of gastrulation, because the placodal field gives rise to a single (fused)placode with lens character. Formation of ectopic and fused lenses in smoothened mutants further indicates that Hedgehog signaling from neural plate puts an end to the lens-forming capacity of median non-neural ectoderm and specifies pituitary fate instead. Consistent with this idea,overexpression of Shh suppresses lens formation(Barth and Wilson, 1995) and induces pituitary genes in lens derivatives (Figs 7, 8).

We also show that loss of Hedgehog signal transduction leads to increased olfactory placode size (see Fig. S1 in the supplementary material). In lateral regions of the placodal field, we find that lens and olfactory precursor cell populations intermingle only where the pitx3 and dlx3bexpression domains overlap. By contrast, median pitx3 and dlx3b domains overlap significantly and olfactory and pituitary precursors intermingle to a high degree in this region, presumably because the placodal fields of olfactory and pituitary placode have not segregated from each other by bud stage (Whitlock,2004). In smu mutant embryos, olfactory and ectopic lens precursor cells intermingle near the midline. Although we do not find increased numbers of olfactory precursor cells in smu mutants(Fig. 3B), which might provide an explanation for the larger olfactory placodes, we can not exclude the possibility that median precursor cells have an equivalent potential to contribute to lens, pituitary and olfactory placode before segregation of median placodal fields. Nevertheless, our lineage analysis of median, pitx3-positive precursor cells strongly supports an equivalence domain for at least lens and pituitary fates.

Our observations provide evidence for a novel mechanism that regulates the size of the pituitary placode. We find that Pitx3 is required for pituitary placode specification and formation (Fig. 5D-F) at the end of gastrulation (Figs 5, 6). Previous studies have implicated Dlx3b and Dlx4b in olfactory and otic placode formation(Whitlock and Westerfield,1998; Whitlock and Westerfield, 2000; Solomon and Fritz, 2002; Liu et al.,2003; Hans et al.,2004). We find that at bud stage, Pitx3, Dlx3b and Dlx4b are expressed in partially overlapping domains, and that reducing Dlx3b and Dlx4b increases the size of the pituitary placode and prevents lens fiber cell specification (Fig. 5; see Fig. S3 in the supplementary material). Our results indicate that Dlx3b-Dlx4b and Pitx3 may not interact directly, because loss of Dlx3b and Dlx4b functions leads to downregulation of pitx3 expression in lens precursors at bud stage, but does not affect pituitary pitx3 expression. Similarly, the size of the pituitary pre-placode increases during mid-somitogenesis when Dlx3b and Dlx4b function is reduced (Fig. 6). Because Dlx3b and Dlx4b are required for olfactory placode formation, the enlargement of the pituitary pre-placode in their absence further supports the idea that intermingled olfactory and pituitary precursors at bud stage might have equivalent potential to form nose or pituitary. This possibility is also supported by split pituitary and olfactory fates that arise from single precursor cells that express both pitx3 and dlx3b (Fig. 3).

We suggest that the initial overlap between the pitx3, dlx3b and dlx4b expression domains in the placodal field may eventually result in an outer pre-placodal domain, demarcated by pitx3 expression(Fig. 9) and fated to give rise to pituitary and lens, and in an inner dlx3b and dlx4bpositive domain, fated to give rise to olfactory placode(Whitlock and Westerfield,2000). Initially, the entire outer pre-placodal domain may have potential to form lens (Kamachi et al.,1998; Ueda and Okada,1986). Observations in chick that the early anterior pituitary anlage expresses the lens protein, δ-crystalline(Kamachi et al., 1998; Ueda and Okada, 1986) are consistent with this interpretation. Pax6 and other transcription factors are thought to mediate the competence of head ectoderm to form lens in response to inductive signals (Kondoh,1999). Our results show that pitx3 is required for specification of lens (see Fig. S3 in the supplementary material) and pituitary placodes (Fig. 5). Moreover, pitx3 is required for lim3 gene expression during normal development. Therefore, we suggest that Pitx3, in addition to Pax6 and,presumably, other factors, mediates the response of placode precursor cells and placodal cells to signaling interactions.

Previous studies showed that loss of Hedgehog signaling results in a conversion of pituitary to lens (Karlstrom et al., 1999; Kondoh et al.,2000; Varga et al.,2001) and have led to the generally accepted hypothesis that a major function of anterior midline Hedgehog expression is to specify the pituitary placode and prevent it from forming a median lens. The observation that ectopic, cell non-autonomous activation of Hedgehog signaling induces pituitary gene expression in at least some lens precursors is consistent with this interpretation. However, we find that overexpression of Shh does not increase pituitary placode size and that in smoothened mutants, a median placode (albeit of lens character) still forms. This indicates that Hedgehog signaling is neither sufficient nor required for median placode formation. Nevertheless, changes in Hedgehog signaling affect pituitary cell proliferation and specification in mouse and zebrafish(Herzog et al., 2003; Sbrogna et al., 2003; Treier et al., 2001). We suggest (Fig. 9) that formation of the lens and pituitary pre-placode is independent of Hedgehog signaling;however, once the placode forms, Hedgehog from ventral forebrain specifies pituitary character in the median placode(Herzog et al., 2003; Sbrogna et al., 2003; Treier et al., 2001). We further suggest that non-Hedgehog signaling interactions from neighboring tissues specify or block lens character in the lateral placodes, because cell-autonomous activation of the Hedgehog pathway does not prevent lens formation (Fig. 8). Several lines of evidence suggest that BMP is required and sufficient to promote lens fiber cell differentiation in lens epithelial cells(Belecky-Adams et al., 2002; Faber et al., 2002). Thus,localized BMP, together with Hedgehog signaling, may mediate the subsequent differentiation of unspecified pre-placodal cells into lens fiber cells or hormone-producing pituitary cells.

We thank Soojin Ryu, Andrzej Nasiadka and Dirk Meyer for discussions;Wolfgang Driever, Roland Nitschke and the Life Imaging Facility of the SFB592/Z2 for excellent support; Stephanie Heyl (Rulfs), Sophie Seibel,Jeanette Müller, Elvira Geterle, Roswitha Koppa, Angela Naumann and Sabine Götter for technical help; Eric Glasgow, Christine Petit, Wiebke Herzog, Matthias Hammerschmidt, Steve Devoto, Anne Ungar, Kate Lewis, Dong Liu and the Zebrafish International Resource Center for plasmids, antibodies and morpholinos. M.W. supported by NIH DC04186 and HD22486. S.D. and Z.M.V. are supported by DFG Sonderforschungsbereich SFB592/A5. G.A. supported by APART Fellowship from OEAW.