Hedgehog signaling is required for pituitary gland development

The pituitary gland originates through the interaction of two different ectodermal tissues, the neural and oral ectoderm. The neural ectoderm gives rise to the posterior pituitary, whereas part of the oral ectoderm will develop into the anterior and intermediate pituitary gland containing at least six distinct cell phenotypes (Treier and Rosenfeld, 1996; Sheng and Westphal, 1999; Burrows et al., 1999; Dasen and Rosenfeld, 1999).

Recent work has begun to unravel general mechanisms of pituitary organ induction based on defining the obligatory interactions between the neural and oral ectoderm as a prerequisite for pituitary gland formation. A dual induction from the diencephalon is required for formation of Rathke’s pouch, the primordium of the pituitary gland (Ericson et al., 1998; Takuma et al., 1998; Treier et al., 1998). Both a BMP4 signal and FGF activity from the ventral diencephalon are required for the development of a definitive pouch. Two LIM homeobox factors Lhx3/Lhx4 have been shown to be the earliest molecular pituitary markers expressed within the oral ectoderm and required for formation of a definitive pouch (Sheng et al., 1997). Subsequently a ventrodorsal BMP2 gradient, appearing at a Shh boundary (Treier et al., 1998), together with an opposing dorsoventral FGF8 gradient serve to determine ventral/intermediate cell phenotypes (gonadotropes, thyrotropes, somatotropes, lactotropes) and dorsal cell phenotypes (melanotropes and corticotropes), respectively, with attenuation of the BMP2 signal ultimately required for terminal differentiation.

A consequence of these signaling gradients is apparently to establish overlapping expression patterns of several transcription factors in Rathke’s pouch, allowing positional determination of pituitary cell types by E10.5-E12.5. Several factors exhibit spatially restricted domains of expression, including Isl1 (Ericson et al., 1998; Treier et al., 1998), P-Frk (Treier et al., 1998), Brn4 (Sornson et al., 1996; also known as Pou3F4), and Gata2 (Dasen et al., 1999). In the final stages, the pituitary cell types express a series of differentiation markers which appear in distinct temporal and spatial patterns between E14.5-E16.5, including corticotropes, secreting adrenocorticotropin hormone (ACTH); melanotropes, secreting melanocyte-stimulating hormone (MSH); thyrotropes, secreting thyroid-stimulating hormone (TSH); gonadotropes, secreting luteinizing hormone (LH) and follicle-stimulating hormone (FSH); somatotropes, secreting growth hormone (GH); and lactotropes, secreting prolactin (Prl). In addition, an embryonic cell type produces TSH (rostral tip cells). LH, FSH and TSH are heterodimers sharing a common α-glycoprotein subunit (αGSU) and a specific β-subunit (reviewed by Treier and Rosenfeld, 1996). The well-defined nature of these cell types make the pituitary gland an excellent model system in which to investigate the molecular mechanisms that underlie the appearance of specific cell phenotypes during mammalian organogenesis (Treier and Rosenfeld, 1996; Dasen and Rosenfeld, 1999).

The expression pattern of Shh in cells adjacent to Rathke’s pouch during its formation led us to speculate whether SHH could exert essential roles in the early phase of pituitary gland formation. Indeed, Hedgehog proteins have been proved to play important roles in the development of various vertebrate and invertebrate tissues (Basler and Struhl, 1994; Johnson and Scott, 1998; Ruiz i Altaba, 1999). Although SHH has been implicated in proliferative aspects of organ development (Bellusci et al 1997; Duprez et al 1998; Pepicelli et al., 1998; St-Jacques et al., 1998; Rowitch et al., 1999; Chiang et al., 1999; Drossopoulou et al., 2000), less is known about how hedgehog boundaries in epithelia may contribute to differentiation of cell types in mammalian organs in vivo (Dahmann and Basler, 1999). We have therefore used a transgenic approach utilizing the Hedgehog inhibitor Hip (Hedgehog interacting protein) (Chuang and McMahon, 1999) to study a potential role of Hedgehog signaling during pituitary gland formation. Here, we demonstrate a critical role for Hedgehog signaling in early phases of pituitary organogenesis, both in proliferation and differentiation of the gland.

Transgenic mice were produced as described (Crenshaw et al., 1989). Construction of transgenes using the 14 kb αGSU promoter or the 9 kb Pitx1 promoter in combination with a vector cassette containing a 5′-globin intron and a 3′-polyadenylation signal from the human growth hormone has been described previously (Treier et al., 1998), by generation of transgenic mice expressing lacZ or CRE under control of this information, we have demonstrated the temporal and spatial specificity of its expression. The Pitx1^HS-Hip transgene contains a mouse Hip cDNA that encodes for a HIP protein lacking the last 22 amino acid residues involved in membrane anchoring. TheαGSU-Shh transgene contains the full-length cDNA for Shh. The αGSU-Fgf8 transgene has been described previously (Treier et al., 1998).

The number of transgenic mice obtained at the reported stage and the number showing an abnormal phenotype (respectively in parentheses) are as follows: Pitx1^HS-Hip E11.5 (12:9), E13.5 (10:8), E17.5 (9:7); αGSU-Shh, E17.5 (7:7); αGSU-Fgf8, E17.5 (9:7).

Tissues were fixed in either 10% formalin for analysis by in situ hybridization or in 10% formaldehyde, 60% ethanol for 2-3 hours for immunohistochemistry. Hybridization with ³⁵S-labeled antisense RNA probes and exposure were all done as previously described (Simmons et al., 1990). In brief, frozen sections (16 mm) were mounted onto Superfrost plus slides. The sections were prehybridized followed by hybridization with ³⁵S-labeled antisense cRNA probes for the respective cDNAs. After hybridization and washing the sections were incubated with RNase A (20 mg/ml) at 37°C for 20 minutes.

RNase A-resistant hybrids were detected by autoradiography using Kodak NTB-2 liquid emulsion. Autoradiographic exposure was for 4-7 days. Immunohistochemistry was done on 5-7 mm thick paraffin sections stained by an indirect immunoperoxidase method. Peroxidase activity was visualized with the DAB/metal enhancer (Pierce; Rockford, Ill.).

Rathke’s pouches were dissected from E9.5 mouse embryos with electrolytically sharpened tungsten needles after collagenase (Sigma, type I) treatment. After 30 minutes recovery in Dulbecco’s minimum essential medium (DMEM) which contained 10% fetal calf serum, Rathke’s pouches were recombined in collagen gel either with control QT6 quail cells or with SHH-producing QT6 cells (a gift from Dr D. Duprez; see Duprez et al., 1998). For recombinant cultures, control QT6 cells and SHH-producing QT6 cells were grown in 35 mm petri dishes until a dense monolayer was formed. Small pieces (approx. 50 μm in diameter) of monolayer were dissected from these cultures with tungsten needles and placed in close proximity to Rathke’s pouches in drops of collagen. Collagen gel from collagen type I (Collaborative Research) was prepared according to the manufacturer’s protocol. Recombinants were cultivated for 48 hours in serum-free DMEM/F12 medium supplemented with penicillin-streptomycin and N2 (Gibco-BRL). They were fixed in 4% formaldehyde in PBS (30 minutes), washed with PBS overnight and stained with rabbit affinity-purified anti-LHX3 antibody (1:2500 dilution in PBS, 10% fetal calf serum, 0.1% Triton X-100, 16 hours at +4°C). Antibodies against LHX3 were a gift from Dr S. Pfaff. After washing in PBS (4 times, 6 hours) cultures were processed with anti-rabbit HRP-conjugated antibodies (dilution 1:500; Chemicon) for 16 hours at +4°C and washed 4 times with PBS (3 hours). Peroxidase was visualized with amino-ethyl carbasole (Calbiochem) according to the manufacturer’s protocol.

PITX1 (Lamonerie et al., 1996; Szeto et al., 1996) is a transcription factor expressed during very early stages in the oral ectoderm and prior to and during Rathke’s pouch formation. A 9 kb HindIII-SrfI (Pitx1^HS) fragment of the upstream region of Pitx1 directs expression of β-galactosidase (β-gal) to early Rathke’s pouch and oral ectoderm. From seven founder transgenics, β-gal expression was detected in four of them. Expression was specific to the oral ectoderm and Rathke’s pouch within the head region in all four of them. The other three founders showed no expression of the reporter transgene, presumably due to integration position effects. Fig. 1B shows whole-mount β-gal staining of 10.5 d.p.c. Pitx1^HS-lacZ founders with transgene expression only in Rathke’s pouch and oral ectoderm and in the branchial arch, without evidence of expression elsewhere within the head region, including an absence of expression in the ventral diencephalon.

Shh is uniformly expressed throughout the oral ectoderm, but its expression is restricted from the nascent Rathke’s pouch as soon as it becomes morphologically visible (Fig. 1A1,A2). This creates a Shh boundary within the oral epithelium whereby cells that are destined to form Rathke’s pouch, the primordium of the anterior pituitary, no longer express Shh. In addition, Shh can be detected throughout the ventral diencephalon adjacent to Rathke’s pouch. Shh expression is subsequently lost within the oral epithelium after E12.0 and within the diencephalon after E14.0 (Fig. 1A3). Furthermore, patched 1 (Ptc1; also known as Ptch) (Fig. 1A2), the multipass transmembrane protein receptor for SHH and a direct transcriptional target of the Hedgehog pathway is expressed at high levels in all cells of Rathke’s pouch suggesting that cells of the nascent pituitary gland are receiving a Hedgehog signal.

To test whether SHH may play a role during pituitary organogenesis, we first analyzed Shh^−/− mice (St.Jacques et al., 1998). However, in Shh^−/− mice, we were never able to detect even a rudimentary Rathke’s pouch (data not shown). This is not surprising because Shh^−/− mice exhibit cyclopia and several midline structures of the brain are missing including the T/Ebp/Nkx2.1 expression domain (Pabst et al., 2000). Therefore, the absence of the pituitary gland in Shh^−/− embryos could be a direct or indirect effect of SHH action, as it has been shown that the dysmorphogenesis of the ventral diencephalon in T/Ebp/Nkx2.1 null mice (Kimura et al., 1996) leads to a complete disruption of pituitary gland development. It has been suggested that the observed dysmorphogenesis in T/Ebp/Nkx2.1 null mice results in a displacement of the Fgf8 expression domain within the ventral diencephalon, such that it no longer contacts the oral ectoderm (Kimura et al., 1996). Furthermore, in vitro experiments have shown that FGF8 can maintain Lhx3 expression (Ericson et al., 1998), a LIM homeobox protein essential for pituitary development (Sheng et al., 1996), leading to the suggestion that Lhx3 may be a direct FGF8 target in vivo. Because of the dysmorphogenesis of the ventral diencephalon in Shh^−/− mice, we wanted to investigate a possible direct role for Hedgehog signaling in pituitary organogenesis, which precludes any model disrupting ventral diencephalon morphogenesis.

To permit a transgenic approach, we linked the Pitx1^HSregulatory upstream sequences with the DNA sequence encoding Hip. HIP is a type I membrane protein that binds with high affinity to all mammalian Hedgehog family members, and like PTC1, is a general transcriptional target of the Hedgehog signaling pathway (Chuang and McMahon 1999). Ectopic expression studies indicate that HIP acts as an antagonist of Hedgehog signaling, a property it also shares with PTC1. Ectopic expression of HIP in the developing skeleton generates a limb phenotype closely resembling a loss of Indian hedgehog activity (Chuang and McMahon, 1999). Furthermore, ectopic expression of Hip in the developing zebrafish embryo produces an adaxial muscle phenotype that closely resembles that produced by zebrafish mutants lacking Hedgehog signals (M. Hammerschmidt and A. P. M., unpublished data). Indeed, despite the widespread expression of Hip during a period of fish development known to be regulated by many different classes of signals, the only phenotype detected appears to be a loss of Hedgehog signaling. Thus, Hip appears to be a direct and highly specific antagonist of Hedgehog proteins, exhibiting a functional effect on the Hedgehog proteins analogous to that which Noggin exerts against specific members of the BMP family. We utilized a truncated form of HIP, which lacks the transmembrane domain and has been shown to efficiently become secreted into the extracellular space (Chuang and McMahon 1999). This enabled us to use a transgenic founder analysis strategy to determine the consequences of blocking Hedgehog signaling within Rathke’s pouch and the oral ectoderm for pituitary organogenesis.

Fig. 1C shows representative examples of the morphological consequences of targeted HIP overexpression within the oral ectoderm, analyzed at various time points during mouse embryonic development. Rathke’s pouch has separated from the oral ectoderm by E11.5 during normal development. In contrast, in Pitx1^HS-Hip transgenic embryos only a rudimentary pouch is visible, resembling the developmental stage equivalent to E9.0. The nine transgenic founders analyzed showed a visible morphological phenotype with varing degrees of organ hypoplasia; a representation of the most severe form is shown (in Fig. 1C). At E17.5, normally the pituitary gland is fully developed containing an anterior and intermediate lobe, as well as the posterior pituitary (Fig. 1C). In contrast, only a cystic rudiment of the pituitary gland can be found in high copy number Pitx1^HS-Hip transgenic embryos.

These results provide strong evidence that Hedgehog signaling plays an essential role during pituitary organogenesis. In this analysis, we have not disturbed the development of the ventral diencephalon, as occurs in Shh^−/− embryos. Not only is the ventral diencephalon morphologically intact, but it’s normal development is further substantiated by the observation that the expression domain of Fgf8, one of the family of signaling factors from the diencephalon required for Rathke’s pouch formation, is not altered in these mice (Fig. 2B), in contrast to alterations observed in T/Ebp/Nkx2.1 null mice. Therefore, the data are most consistent with the conclusion that cells within Rathke’s pouch directly receive a hedgehog signal that is pivotal for pituitary gland formation.

To better understand the failure of pituitary organogenesis in Pitx1^HS-Hip transgenic embryos, we performed in situ hybridization on the nine transgenic founders obtained at E11.5 with probes for molecular markers previous shown to play essential roles in pituitary gland formation. Although at E11.5 development of Rathke’s pouch is severely disturbed in Pitx1^HS-Hip transgenics, Lhx3 expression can be readily detected even in the rudimentary pouch that is seen (Fig. 2A7). These data indicate that the induction of Lhx3 expression through FGF factors is still occurring and further confirms that signals from the ventral diencephalon are not impaired, in contrast to the situation in Shh^−/− or T/Ebp/Nkx2.1^−/− mice (Takuma et al., 1998). However, Prop1 (Sornson et al., 1996), a tissue-specific paired-like homeodomain factor, which is mutated in the Ames mouse and is required for later appearance of the PIT1 lineage, is only minimally expressed, albeit initially detected (Fig. 2A8). Bmp2 is known to be induced at the Shh boundary in a ventral to dorsal gradient opposing the patterning effects of the dorsally expressed FGFs. Performing in situ analysis with a Bmp2-specific probe we could not detect Bmp2 expression in the nascent pouch rudiment in the analyzed Pitx1^HS-Hip transgenic animals. However, Bmp2 expression persists in the surrounding ventral mesenchyme, showing that the block of Bmp2 expression is specific for ectodermal cells of the pouch (Fig. 2A6,A12).

Consistent with this result, we observe that Gata2, which encodes a zinc finger-containing transcription factor previously shown to be a direct transcriptional target of BMP2 in the pituitary gland (Treier et al., 1998; Dasen et al., 1999), fails to be expressed in Rathke’s pouch of all E11.5 Pitx1^HS-Hip transgenic mouse embryos analyzed (Fig. 2A4,A10). In addition, all other ventral transcription factors, exhibiting expression domains in Rathke’s pouch beginning at the ectodermal boundary of Shh exclusion, fail to become expressed, as further exemplified by Brn4 (Sorenson et al., 1996) (Fig. 2A5,A11). In contrast, transcripts of Isl1 (Ericson et al., 1998: Treier et al 1998), which encode a LIM homeobox transcription factor expressed in the oral ectoderm prior to and during Rathke’s pouch formation are not altered (Fig. 2A3,A9). The observation that Isl1 expression across the Shh boundary is unaffected in Pitx1^HS-Hip transgenics demonstrates that the block of Hedgehog signaling may affect only factors that become expressed de novo at the hedgehog boundary after the appearance of Rathke’s pouch.

By E13.0, the arrest of pituitary development in Pitx1^HS-Hip transgenic embryos has become even more apparent (Fig. 3A). In wild-type mice, the developing pituitary gland has undergone a marked increase in organ volume with concomitant expression of Lhx3 (Fig. 3A1). In contrast, in the most severe cases of the eight transgenic Pitx1^HS-Hip embryos obtained at E13.0, only a cystic pouch is present, which remains partly connected to the oral ectoderm (Fig. 3A8). Furthermore, Lhx3 expression has now ceased in the ventral part of the cystic structure and only remains within the dorsal cells, where expression of Prop1, a dorsally expressed factor, is now clearly visible (Fig. 3A11). As was the case at E11.5, no expression of ventral factors including Bmp2, Gata2 or Brn4 is detectable in the rudimentary pouch of any of the transgenic embryos analyzed (Fig. 3A9,A10,A12). This is further underscored by the absence of the first two terminal differentiation markers, POMC and TSHβ, which are normally detectable after E12-E12.5 (Fig. 3A6,A13,A7,A14) during normal pituitary development (Japon et al., 1994). These results demonstrate that Hedgehog signaling is required for both proliferation and appearance of ventral cell type markers in pituitary organogenesis in vivo.

Pitx1^HS-Hip transgenic mice display a pituitary phenotype similar to that of mice with a targeted deletion of Lhx3 (Sheng and Westphal 1996), and exhibit a delayed ventral exclusion of Lhx3 expression. This prompted us to speculate that Lhx3 expression may not only be regulated by FGF signaling, but may in addition be controlled by Hedgehog signaling. To test whether SHH is able to regulate Lhx3 expression, we cocultured Rathke’s pouch explants with SHH-expressing QT6 cells (Duprez et al., 1998). Only when the pouch explants were cocultured with SHH-expressing QT6 cells could we see a strong induction/maintenance of LHX3 immunoreactivity, whereas explants cocultured with control QT6 cells only showed barely detectable LHX3 levels (Fig. 3B). This result suggests that Hedgehog signaling, directly or indirectly, modulate, LHX3 expression levels. The discrete expression of Lhx3 in only that part of the oral ectoderm that is destined to become Rathke’s pouch and subsequent expression in the developing gland therefore appears to result from a combinatorial regulation by two signaling molecules, FGF8 and SHH.

To complement the in vivo Hedgehog loss-of-function studies with gain-of-function studies, we again utilized an in vivo transgenic approach to express the full-length coding region of Shh linked to the regulatory sequences of αGSU, which targets expression into Rathke’s pouch (Fig. 4A and Kendall et al., 1994; Treier et al., 1998). Comparison of pituitary glands from wild-type and αGSU-Shh transgenic mouse embryos at E17.5 revealed a striking threefold increase in organ volume and cell number within the pituitary gland of αGSU-Shh transgenic embryos (Fig. 4B). Lhx3 expression seems slightly increased in the ventral part of the gland, further supporting the model that this domain of Lhx3 expression is under SHH control. Most strikingly, Bmp2 transcripts, which are normally no longer detectable in the wild-type gland at this stage, are clearly present in lateral parts of the gland where αGSU regulatory sequences drive the highest expression of the transgene (Fig. 4B2,B9). These data suggest that pituitary cells can respond to SHH activity with upregulated or sustained Bmp2 transcription. The increase in organ size is mainly due to an increase of the αGSU-expressing cell population, consisting of thyrotropes, producing TSHβ, and gonadotropes, producing LHβ (Fig. 4B3,B10,B6,B13,B7,B14). The field of both terminal differentiation markers is broadly expanded, with premature onset of leutinising hormone β gene expression. Whereas corticotrope cell number and distribution appears unchanged (Fig. 4B5,B12), the diminished field of growth hormone-expressing cells is pushed more dorsally (Fig. 4B15,B16). This result demonstrates that SHH has, at least, a positive effect on gonadotrope and thyrotrope cell type proliferation and determination.

Recent experiments have shown that Lhx3 expression can be maintained in Rathke’s pouch explants by addition of FGF8 (Ericson et al., 1998). Furthermore, mice lacking the IIIb form of fibroblast growth factor receptor 2 have severe defects in pituitary organogenesis, demonstrating that a FGF activity is required for outgrowth of the pituitary gland (De Moerlooze et al., 2000). We and others (Crossley and Martin, 1995) have shown earlier that FGF8 is expressed in a correct spatial and temporal pattern in the infundibulum to be a good candidate for this activity. Nevertheless, FGF8 does not bind with high affinity to the IIIb form of fibroblast growth factor receptor 2, although it has been suggested that accessory molecules might differentially modify the relative activity. In contrast, FGF10 does bind to FGFR2b, and has been reported to be expressed at later stages (E14) in the rat infundibulum (Yamasaki et al., 1996). Therefore, we compared directly the spatial and temporal expression profiles of three FGF family members in the infundibulum: Fgf8, Fgf10 and Fgf18. All three genes are already expressed by approx. E9.0 in the ventral diencephalon, opposite the region of the oral ectoderm from which Rathke’s pouch will morphologically appear and in which expression of Lhx3 is first observed (Fig. 5A). Therefore, Fgf10 most likely constitutes the most critical endogenous mitogenic activity for pituitary progenitor cells signaling through FGFR2b.

This prompted us to investigate whether FGFs, including FGF8, indeed do regulate Lhx3 in vivo expression as suggested through in vitro experiments. Using a transgenic approach to express Fgf8 under the αGSU promoter reveals that Lhx3 expression is ectopically activated in the oral ectoderm, driving multiple invaginations within the oral ectoderm. Fig. 5B,C shows in situ hybridization with a Lhx3-specific probe. Lhx3 expression in the wild-type gland is restricted from initial invagination to Rathke’s pouch, with no expression visible within the oral ectoderm (Fig. 5B). Overexpression of FGF8 within Rathke’s pouch however leads to a dysmorphogenic gland (Treier et al., 1998). When we analyze the invaginated structures using a Lhx3-specific hybridization probe, we find that the multiple invaginations that have formed small pouch cavities or long open cysts are still expressing high levels of Lhx3 transcripts at a time when Lhx3 levels are becoming undetectable in the wild-type gland. More surprisingly, when we analyze more lateral sections within the same embryo, where the oral ectoderm exhibits no morphological alteration, we also find strong Lhx3 expression within the oral ectoderm (Fig. 5C). These results indicate that FGF8 is indeed sufficient to activate Lhx3 expression in the mouse oral ectoderm in vivo.

Mutations in the human sonic hedgehog gene (Odent et al., 1999), the mouse Gli1 and Gli2 genes (Park et al., 2000) as well as the you-too mutation in zebrafish affecting a Gli2 related transcription factor (Karlstrom et al., 1999), together with our described expression profile of Shh during the appearance of Rathke’s pouch (Treier et al., 1998), suggested that Hedgehog signaling may play an instrumental role in pituitary organogenesis. We analyzed the significance of Hedgehog signaling during pituitary gland development by using a Pitx1^HS enhancer to overexpress Hip, a hedgehog inhibitor within the oral ectoderm and Rathke’s pouch. This approach allowed us to avoid disturbing the development of the ventral diencephalon as shown for Shh^−/− embryos (Pabst et al., 2000) and Gli1/Gli2 double mutant embryos (Park et al., 2000). This is crucial as the ventral diencephalon expresses Fgf8, Fgf10, Fgf18, as well as Bmp4 in addition to Shh, which have been shown to provide crucial instructive signals for pituitary organogenesis. Pitx1^HS-Hip transgenic embryos display pituitary hypoplasia and show a loss of induction of ventral transcription factors expressed at the Shh boundary.

The observed proliferation defect may be due to the combined block of Hedgehog signaling from both the ventral diencephalon and the oral ectoderm. Further, we provide evidence that Lhx3 is under dual control of FGF and SHH signaling. Lhx3^−/− knock out mice display a phenotype similar to Pitx1^HS-Hip transgenics suggesting that the maintenance of high Lhx3 expression levels is crucial for proliferation of pituitary progenitor cells, and ultimately requires a synergistic signaling input from both FGF and SHH, respectively.

Thus, in the pituitary gland, as in other organs, proliferation and terminal differentiation are closely linked. We have, however, been able to show that factors with expression domains initiated at the hedgehog boundary and that determine ventral cell types in Rathke’s pouch are selectively affected by the Pitx1^HS-Hip transgene. We postulate that this is most likely due to a failure of Bmp2 induction within Rathke’s pouch, indicating that the Shh boundary within the oral epithelium is instrumental in the induction of ventral factors expressed in Rathke’s pouch. This scenario is reminiscent of the compartment boundaries organizing anteroposterior patterning in the Drosophila wing (Zecca et al., 1995; Lawrence and Struhl, 1996). In that case, posterior cells organize growth and patterning in both compartments by secreting Hedgehog protein that then induces a stripe of neighboring anterior cells across the compartment boundary to secrete Decapentaplegic. In the case of pituitary development, we would suggest that SHH induces Bmp2 expression in ventral Rathke’s pouch cells, with BMP2 serving as the critical signal for at least Gata2 gene expression. We favor this interpretation of our results, because a FGF factor expressed within the anterior pituitary that is required for pituitary development has not yet been identified, although in other organs and in the vertebrate limb (Sun et al., 2000; Drossopoulou et al., 2000) both SHH/BMP and SHH/FGF signaling relays exist.

The suggestion that SHH plays a role both in proliferation and in cell-type determination of the pituitary gland is further supported by αGSU-Shh overexpression studies that result in an induction of Bmp2 and an expansion of ventral cell types producing LHβ and TSHβ. This phenotype mimics the pituitary phenotype we have previously reported for mice harboring a mutation in the coding sequence of the transcription factor Pax6 (Kioussi et al., 1999). Together, this situation is reminiscent of results obtained in the ventral mouse neural tube (Ericson et al., 1997; Briscoe et al., 1999). In the ventral tube of Pax6 mutant mice, progenitor cells missing PAX6 generate neurons characteristic of cells normally exposed to greater SHH activity. This raises the possibility that SHH at the boundary may even have polarizing activity and specific pituitary cell types may likewise be determined from a progenitor cell population by responding to graded SHH/BMP signaling leading to a ventral to dorsal cell fate transformation in αGSU-Shh transgenics. Two additional observations support this view. First, we do not see growth hormone-positive cells intermingled with thyrotropes and gonadotropes in the most rostral part of the gland as in the wild-type gland, where Shh expression is highest. Second, the overall number of growth-hormone-positive cells is relatively reduced, suggesting that more pituitary progenitor cells may have adopted a ventral fate.

While there seems to be no striking upregulation of Gata2 in αGSU-Shh transgenic pituitaries, which may be due to the fact that Bmp2 levels are much less elevated compared to those in transgenic mice overexpressing Bmp2 (Dasen et al., 1999), we still observe a similar phenotypic consequence – an expansion of ventral cell types (gonadotropes and thyrotropes).

Thus, exploring the role of SHH in pituitary organogenesis has further elucidated the early complementary and reciprocal signaling events involved in pituitary development, suggesting a model (Fig. 5D) where the interplay of several classes of signaling molecules including SHH, BMP4 and FGF8/FGF10, is responsible for the induction and patterning of a mammalian organ. It appears that the exclusion of Shh from the primordium of organs that arise by budding morphogenesis may be a common, but clearly not universal, strategy (Hebrok et al., 1998; Pepicelli et al., 1998) through which organ size later becomes determined; in addition, creating a Shh boundary can direct initial organ patterning (Hogan, 1999; Kim and Melton, 1998). It is tempting to speculate that one critical role of BMP4, expressed in the ventral diencephalon, is to suppress Shh expression in the adjacent region of the oral ectoderm destined to become Rathke’s pouch, analogous to the role of BMP4 on Shh gene regulation recently described in tooth development (Zhang et al., 2000). The cells in the invaginating oral ectoderm, which no longer express Shh, are able to respond appropriately to FGF8, FGF10, and SHH signals by expressing the LIM homeobox gene Lhx3, which we have shown is under dual control of FGF8 and SHH signaling. Therefore, Lhx3 may provide the positional cue that governs the fate of oral ectodermal cells to become Rathke’s pouch and, in combination with other transcription factors, may direct proliferation and the expression of terminal markers within the pituitary gland.

In summary, we have shown that SHH exerts effects on both proliferation and cell type determination during pituitary gland development analogous to the role of IHH in bone development. Whether SHH, in addition, displays polarizing activity in the patterning of pituitary progenitor cells in Rathke’s pouch remains to be determined.

We thank Peggy Myer for her expertise and generous assistance in preparation of illustrations. For reagents and plasmids, we thank B. Hogan, N. Itoh, G. Martin, S. Yamamoto. We thank Dr Sam Pfaff (The Salk Institute for Biological Studies) for antibodies against LHX3, Dr Delphine Duprez (Institut d’Embryologie Cellulaire et Moleculaire, Nogent Sur Marne) for SHH-producing QT6 cells and control QT6 cells. We also thank the National Hormone and Pituitary Program, NIDDK (Rockville, MD). M. T. was supported by a Boehringer Ingelheim Fonds postdoctoral fellowship. A. S. G. by a NRSA from NIH. M. G. R. is an investigator with the Howard Hughes Medical Institute. This work was supported by grants from the NIH to A. P. M. (NS33642) and to M. G. R. (NIDDK).