Notch signaling functions as a binary switch for the determination of glandular and luminal fates of endodermal epithelium during chicken stomach development

The vertebrate digestive tract is a simple tube in the early stages of development, which comprises the endodermal epithelium and surrounding mesenchyme. Distinct digestive organs are specified along the anteroposterior axis and subsequently undergo organ-specific development. In the chicken, the stomach is subdivided into the rostral proventriculus (PV, glandular stomach)and caudal gizzard (muscular stomach). The PV is characterized by the development of compound glands, which secrete digestive enzymes. From stage 29 of Hamburger and Hamilton (Hamburger and Hamilton, 1951) (about day 6 of incubation), the epithelium sequentially begins to invaginate into the surrounding mesenchyme to form simple glands. In this process, homogenous epithelial cells of the PV take either glandular or luminal fate, and luminal cells subsequently express chicken SP (chicken spasmolytic polypeptide)(Tabata and Yasugi, 1998)gene, while glandular cells express Smad8 and later ECPg(embryonic chicken pepsinogen) (Hayashi et al., 1988) genes. Gland cells are also characterized by specific downregulation of several genes (Yasugi and Fukuda, 2000).

It is well established that the epithelial-mesenchymal interaction is important for the development of the PV(Yasugi, 1984; Yasugi and Fukuda, 2000). Recently, several secreted molecules have been identified as signaling factors regulating the epithelial development of the PV. Bone morphogenetic protein 2(BMP2), which is secreted from the surrounding mesenchyme, is an important inducer of gland formation and ECPg expression(Narita et al., 2000). Fgf10, which is expressed in PV mesenchyme(Shin et al., 2005), regulates epithelial cell proliferation and differentiation (M. Shin, S. Noji and S. Yasuji, unpublished). Fukuda et al.(Fukuda et al., 2003) have shown that a downregulation of Sonic hedgehog (Shh)expression in the epithelium is necessary for gland formation. Epidermal growth factor (EGF) signaling can also promote the luminal fate in dispense of gland cell population (Takeda et al.,2002). However, the molecular mechanism(s) that control the binary fate decision and the spatially patterned differentiation of epithelial cells remain to be elucidated.

Notch signaling controls cell fate decision and patterned differentiation in numerous developmental processes (reviewed by Campos-Ortega, 1993; Artavanis-Tsakonas et al.,1999; Kopan,2002). The Notch transmembrane receptors are activated by cell-surface DSL (Delta, Serrate, Lag2) ligands and mediate direct cell-cell communication. Ligand binding results in a proteolytic cleavage to release the intracellular domain of the Notch proteins (NICD), which subsequently translocates into the nucleus and associates with the DNA-binding CSL(CBF1/Su(H)/Lag1, also known as RBPj) proteins to activate target genes such as the Hairy/Enhancer of Split (HES) family of bHLH transcriptional repressors (Beatus and Lendahl,1998; Hu et al.,2003). Recently, the Deltex pathway has also been identified as mediating alternative Notch signaling(Yamamoto et al., 2001; Matsuno et al., 2002; Endo et al., 2003; Hu et al., 2003). Numb is an inhibitory molecule that binds to PEST domain of NICD and disturb nuclear translocation (Wakamatsu et al.,1999; Wakamatsu et al.,2000).

In this study, we have investigated the involvement of Notch signaling in the early gland cell differentiation of the chicken PV. We show that Delta1/Notch1 signaling in the PV epithelium is activated in a scattered fashion prior to epithelial invagination, as the earliest indication of fate segregation in the epithelium. We also demonstrate that an activation of Notch signaling promotes gland-specific gene expression, whereas persistence of Notch signaling prevents progress of epithelial cell differentiation, and that an inhibition of Notch signaling leads to luminal differentiation. Taken together, these results suggest that Notch signaling regulates spatiotemporally patterned differentiation of the glandular epithelium in the developing chicken PV.

Embryos of White Leghorn chicken (Gallus gallus domesticus) were staged according to Hamburger and Hamilton(Hamburger and Hamilton,1951).

FLAG epitope-tagged expression vectors of CNIC (constitutively active form of chicken Notch1), CNIC^ΔC89(CNIC that lacks C-terminal end) and chicken Numb have been described previously (Wakamatsu et al.,1999; Wakamatsu et al.,2000). Throughout our study, CNIC and CNIC^ΔC89 were used to activate Notch1. CNIC^ΔC89 was more effective than CNIC,although both of them resulted in the same phenotype. CNIC^ΔC89 was used in 3-day cultivation experiment to obtain stable results, while CNIC was used in 36-hour culture experiments to compare with rescue experiments. pmiwSV-quail Delta1has also been described previously(Maynard et al., 2000; Endo et al., 2002). Xenopus Su(H)^DBM was kindly provided by Dr C. Kintner(Wettstein et al., 1997) and the insert was subcloned into pmiwSV (Endo et al., 2002). pTP-1Venus will be described elsewhere (J. Kohyama, A. Tokunaga, Y. Fujita, H. Miyoshi, T. Nagai, A. Miyawaki, K. Nakao,Y. Matsuzaki and H. Okano, unpublished).

The apparatus used for electroporation of plasmid DNAs into PV epithelium was set up as follows (Sakamoto et al.,2000). Platinum electrodes were fixed on a glass dish and integrated into a resin chamber (7 mm in height, 8 mm in width and 5 mm in length). A vessel made of 1.5% agarose/Tyrode's solution gel was put into an electrode chamber and filled with 14 μl plasmid DNAs in Tyrode's solution. The outside of the gel vessel was filled with Tyrode's solution. Isolated PV were cut open and placed in the vessel with their epithelial sides facing to the cathode to introduce DNAs into the epithelium. For the optimal transfection, 50 msecond pulses of 30 V and 75 msecond durations were generated 10 times for stage 28 PV and 15 times for stage 29 PV using CUY 21(BEX). Tissues were immediately washed with Tyrode's solution and cultured for indicated hours. Appropriate concentration of each plasmid DNAs were examined and determined as follows: 200 nM of pEGFP-C1 for 36 hours cultivation and 300 nM for 72 hours culture; 200 nM CNIC^ΔC89 expression vector for 72 hours cultivation; 50 nM CNIC expression vector for 36 hours cultivation; 200 nM of pTP-1Venus for reporter assay or 100 nM for cotransfection with CNIC; 300 nM expression vector of chicken Numb or Xenopus Su(H)^DBM were used for inhibitory experiments and 400 nM for co-transfection with CNIC; and 200 nM pmiwSV-quail Delta1 for reporter assay.

Explanted PV were laid on a Nuclepore filter (pore size 0.8 μm). This filter was placed on a stainless steel grid, placed into one well of a 24-well culture dish (Falcon, 3047) and cultured at the medium-gas interphase in 5%CO₂ and 95% air at 37°C. The culture medium was 199 with Earle's salt containing an equal volume of embryo extract prepared from stage 38 embryos (Takiguchi et al.,1988).

In situ hybridization with digoxigenin-labeled RNA probe was performed on 8-10 μm cryosections as previously described(Ishii et al., 1997). cRNA probes were generated by in vitro transcription from cDNA fragments of ECPg (Hayashi et al.,1988), chicken SP(Tabata and Yasugi, 1998),chicken Notch1, quail Serrate1, quail Delta1,chicken Numb (Wakamatsu et al.,1999; Wakamatsu et al.,2000), chicken Notch2, chicken Serrate2 [gifts from Dr R. Goistuka (Morimura et al.,2001)], chicken Hairy1, chicken Hairy2 [gifts from Dr O. Pourquié, (Palmeirim et al., 1997; Jouve et al.,2000)], chicken Smad8, Shh [gifts from Dr T. Nohno(Nohno et al., 1995; Ohuchi et al., 1997)], chicken Gata5 (Sakamoto et al.,2000), chicken Gata4, chicken Gata6(Laverriere et al., 1994),chicken Sox2 (Ishii et al.,1998), chicken Sox21(Uchikawa et al., 1999),chicken GK19 (Sato and Yasugi,1997) and chicken Fra2(Matsumoto et al., 1998).

M2 anti-FLAG (mouse IgG1, Sigma) and goat anti-mouse IgG conjugated with Alexa 546 (Molecular Probes) were purchased from commercial suppliers. Immunological staining on sections was performed as described previously(Wakamatsu et al., 1993). Cryosections (8-12 μm) were prepared on VectaBond-coated slides (Vector). Sections treated with antibodies were also exposed to DAPI (Sigma) to visualize nuclei, and subsequently mounted with VectaShield mounting medium(Vector).

To investigate whether Notch signaling plays a role in early chicken PV development, we first analyzed the spatiotemporal expression patterns of genes previously identified as Notch signaling components during the stages before secretion of zymogens. Expressions of chicken SP and Smad8were also investigated to monitor the epithelial cell differentiation. Serial sections of PV for each stage were obtained from the same samples. Sections for each probe were mounted on the same slide glass to compare intensity of gene expressions. In the stage 28 gut, the PV region is apparent with its bulgy shape but is still in the phase prior to PV-specific differentiation,and no obvious difference in morphology or gene expression profiles was detected among the epithelial cells. Expressions of chicken SP,marker molecule of luminal epithelial cells, and Smad8 were not detected (Fig. 1A,D). Notch1 and Notch2, and their potential target genes, Hairy1 and Hairy2, were expressed in the entire epithelium(Fig. 1J,M,P,S). Delta1, which encodes a Notch ligand, was expressed in the mesenchyme underlying the epithelium, but not in the epithelium(Fig. 1G).

The PV epithelium begins to invaginate into the mesenchyme at stage 29. This stage is a crucial early phase of gland formation: epithelial invagination is still very small, but luminal surface indentation starts to be observed. Chicken SP and Smad8 expressions were first detected at this stage (Fig. 1B,E). Expression of Smad8 is restricted to cells in small glands and cells just before invagination (arrowheads in Fig. 1). Delta1 was also first expressed in a subpopulation of uninvaginated epithelial cells in a scattered fashion (Fig. 1H,arrowheads). Mesenchymal expression of Delta1 was maintained at sites where the overlying epithelium did not invaginate. In the invaginated epithelial cells, the expression of Notch1 decreased and that of Hairy1 disappeared (Fig. 1K,Q). By contrast, expression of Notch2 and Hairy2 was upregulated in the invaginating epithelium(Fig. 1N,T).

At stage 30, obvious simple glands are formed, although newly formed small glands are also observed. Strong chicken SP expression was detected in the luminal epithelium, but chicken SP-negative cells were also found in the luminal epithelium (Fig. 1C,C′, arrows), suggesting that undifferentiated cells remained within the uninvaginated epithelium. Expression of Smad8 was apparently restricted in the invaginated epithelium(Fig. 1F,F′). Expression of ECPg was not detected until a further 1.5 days later (data not shown); thus, these stages are prior to secretion-competent gland maturation. The number of Delta1-positive cells increased in the uninvaginated epithelium (Fig. 1I,I′). The region where Hairy1 was expressed coincided with the Notch1-positive region and was restricted to the uninvaginated epithelium (Fig. 1L,L′,R,R′), whereas Hairy2 was expressed in the same pattern as Notch2 and was restricted to the invaginated glandular epithelium (Fig. 1O,O′,U,U′). Numb was expressed in the entire epithelium throughout these stages (Fig. 1V-X,X′). In chicken, Notch ligand genes, such as Serrate1 and Serrate2, have also been identified, but their mRNAs were not detected in the epithelium during these stages (data not shown). Therefore, of all the possible ligand/receptor combinations only Delta1 and Notch1 expression overlapped from stage 29, when gland formation and luminal epithelial cell differentiation commence.

From the analysis of expression patterns, we concluded that Smad8is a specific marker gene of gland epithelial cells from the early stage of their development, as its expression begins with and persists during the gland formation and is specific to these cells.

We next performed a reporter assay to confirm whether Notch signaling is indeed activated during the normal development of the PV. pTP-1Venuswas used, a CSL-dependent Venus (a GFP variant) reporter that carries 12 repeats of CSL-binding sites and that efficiently reflects Notch signaling activity (J. Kohyama, A. Tokunaga, Y. Fujita, H. Miyoshi, T. Nagai, A. Miyawaki, K. Nakao, Y. Matsuzaki and H. Okano, unpublished). First,pTP-1Venus was transfected into stage 28 and 29 PV, and the reporter activity was observed after 3 hours of explant culture. As 3 hours cultivation is the shortest time in which to detect reporter-derived proteins, this condition enables us to observe almost `real-time' activity of Notch signaling. Reporter activity was detected in scattered cells in uninvaginated epithelium of stage 29 PV, but was never detected in invaginated epithelium(Fig. 2A). Reporter activity was not detected in stage 28 PV (data not shown), in which no Delta1expression was detected in the epithelium, suggesting that epithelial, not mesenchymal, Delta1 expression was responsible for the Notch signal activation in the epithelium. This result indicated that activation of Su(H)-mediated Notch signaling begins between stage 28 and 29 in cells scattered in the uninvaginated epithelium. When stage 28 and 29 PV were cultured for 24 or 48 hours after transfection, the reporter activity was detected not only in the uninvaginated epithelium in a scattered fashion but also strongly in invaginated gland cells(Fig. 2B,C). In this condition,cells in which Notch signaling had once been activated during cultivation could be Venus positive. Gland formation in stage 28 PV cultivated for 24 hours is comparable with that in transiently cultivated stage 29 PV(Fig. 2A,B). These results suggest the possibility that cells in which Notch signal was transiently activated took gland cell fate and subsequently invaginated. Consistently, Venus-positive cells did not express chicken SP mRNA when stage 29 PV was transfected with pTP-1Venus and cultured for 3 hours(Fig. 3).

As the activation of Notch signaling pathway was observed in a scattered fashion, similar to the expression pattern of Delta1, and as only Delta1 expression was detected in the epithelium among all DSL ligand genes examined, endogenous Delta1 is probably the ligand responsible for the activation of Notch1. To test the ability of Delta1 as a ligand in the PV epithelium, stage 29 PV was transfected with pTP-1Venus and cultured for 3 hours, and the expression of endogenous Delta1 was compared with the reporter activity of pTP-1Venus on the same section(Fig. 4A-C). The Venusreporter was activated in cells neighboring the Delta1-expressing cells (Fig. 4C′),suggesting that endogenous Delta1 actually functions as a Notch ligand to activate Notch signaling. Furthermore, when Delta1expression vector was co-transfected with pTP-1Venus in stage 28 PV and cultured for 36 hours, the reporter activity was not detected(Fig. 4D,E), suggesting that Notch signaling was activated in the neighboring cells but not activated in cells expressing Delta1. Consistent with this, when the reporter construct was first electroporated into stage 28 PV, followed by an additional transfection of Delta1 expression vector 10 hours later, activation of the Venus reporter was found around the Delta1-transfected cells after 14 hours of additional culture(Fig. 4F-H).

The results of reporter assay suggested that Notch signaling is activated when the epithelium is specified as glandular and luminal epithelium. We thus examined the function of Notch signaling in the developing PV epithelium. As Delta1-overexpression does not activate Notch signaling in Delta1-expressing cells (see above), we could not observe any obvious effect on epithelial development (data not shown). Thus, we tried a receptor-mediated strategy to activate Notch pathway. CNIC^ΔC89 encodes the intracellular fragment of chicken Notch1 with a C-terminal truncation, including the PEST domain (which is related to protein degeneration and Numb-binding site), and works as a constitutively active form (Wakamatsu et al., 1999; Wakamatsu et al.,2000; Maynard et al.,2000). CNIC^ΔC89 was transfected into the epithelium of stage 28 PV, and the transfected explants were cultured for 3 days. In control explants, transfected with an EGFP expression vector alone, invagination of epithelium was observed, and differentiation marker genes (chicken SP for luminal epithelium and ECPg for late glandular epithelium) were expressed in a complementary pattern as in the normal development (Fig. 5B,C). By contrast, CNIC^ΔC89 transfection caused severe defects in gland formation. Thus, virtually no epithelial invagination occurred, and the expression of ECPg in the glandular epithelium was not observed (Fig. 5E). In these explants, chicken SP expression was also diminished(Fig. 5F), suggesting that the persistent activation of Notch signaling prevents the progress of differentiation.

To further study the primary response of epithelial cells to Notch signaling, we examined the effect of CNIC in shorter time of culture. Thirty-six hours cultivation of stage 28 PV was chosen because epithelium of explanted PV does not express ECPg at this point but segregation of epithelial cell fates can be visualized by detecting chicken SP or Smad8 mRNAs. To investigate the identity of Notch-activated cells,expression of various genes normally expressed in the PV epithelium, with different intensity in gland and luminal epithelium, were compared with that in normal PV before and after the beginning of gland formation. Throughout Figs 6 and 7, serial sections of explants were made from the same samples. These sections for each probe were mounted on the same slide glass together with serial sections of normal PV to compare expression intensity. We have always confirmed the efficient activation of Notch signaling by the expression of the co-transfected pTP-1Venusreporter in neighboring or next neighboring sections of each panel presented(Fig. 6B, Fig. 7B,K: small insets and not shown; see Discussion). In Fig. 6, genes whose expression is similar to gland cell are listed. CNIC transfection inhibited expression of chicken SP, Shh,Fra2 and GK19 genes, all of which are specifically downregulated in the glandular epithelium of normally developing PV(Fig. 6F-Y). As Shh,Fra2 and GK19 were expressed in the undifferentiated epithelium prior to the gland invagination (Fig. 6M,R,W), and as CNIC transfection promoted the expression of Smad8, which is specific to the invaginated glandular epithelium in the normal PV (Fig. 6A-E), CNIC-transfected cells were similar in character to the early gland cells and were clearly distinct from the undetermined epithelial cells or the luminal epithelial cells. Expression of other genes examined (Gata4,Gata5, Gata6, Sox2, Sox21, Hairy1 and Hairy2), however, was not obviously identical to gland cells, but it was found that this expression was comparable with those of undetermined epithelial progenitor cells(Fig. 7). Therefore, CNIC-transfected cells are probably intermediate between gland cells and undetermined cells. These results suggest that activation of Notch signaling promotes the initial phase of glandular specification but prevents progression of further differentiation and invagination. The distribution of reporter-positive cells, Delta1-expressing cells and chicken SP-negative cells supports this idea.

We next tried to rescue the phenotype obtained from the forced expression of CNIC by co-transfection of expression vectors such as Numb or dominant-negative form of Suppressor of hairless,Su(H)^DBM, into stage 28 PV, and the explants were cultured for 36 hours. Both constructs interfered with the activation of reporter activity of pTP-1Venus by CNIC(Fig. 8B,D; compare with 8F),and gland formation and the expression of chicken SP were nearly recovered (Fig. 8A,C; compare with 8E). These results confirmed that the phenotype obtained by overexpression of CNIC was dependent on Su(H) activity. It is also indicated that overexpression of Numb or Su(H)^DBM is useful to inhibit Notch signaling, while endogenous Numb allows endogenous Notch activation, as shown in Fig. 2.

Next, to investigate whether Notch signaling is necessary for glandular cell differentiation, Numb and Su(H)^DBM were overexpressed in the stage 28 PV and the transfected explants were cultured for 36 hours. Epithelial invagination was morphologically confirmed(Fig. 9, arrowheads), and luminal epithelium differentiation was monitored by detecting expression of chicken SP in neighboring sections(Fig. 9, rightmost column). Transfected cells were identified by immunological detection of FLAG-tag on the transgene-derived Numb protein (Fig. 9G′,I′), or by expression of co-transfected EGFP (Fig. 9C′,E′). In the explants with high efficiency of transfection, these transgenes inhibited gland formation and induced luminal differentiation (Fig. 9E-F,I-J). The majority of explants with lower efficiency of transfection, however, often formed some invaginated glands, while transfected cells contributed only to the luminal epithelium and did not participate in the invaginated glands (Fig. 9C-D,G-H). As initial transfection efficiency of these inhibitor constructs and EGFP expression vectors were almost comparable when explants were cultivated only for 12 hours (data not shown), cells that escaped from transfection seemed to differentiate only into gland cells. By contrast, overexpression of a dominant-negative form of Dtx1(Yamamoto et al., 2001) did not affect overall gland formation, and dominant-negative Dtx1-transfected cells could contribute to both glandular and luminal epithelium (data not shown). These results suggest that Su(H)-mediated Notch signaling is necessary for the glandular epithelium differentiation.

In this study, we report the involvement of Notch signaling in the regulation of primary phase of binary fate decision of glandular and luminal epithelium, and subsequent differentiation of gland cells in the chicken PV. The first morphological sign of gland development is an invagination of the epithelium to form simple glands. Prior to such morphological changes,however, specification of luminal and glandular epithelial cells from homogeneous epithelial cells has to occur. In our study, Delta1-expressing cells and chicken SP-negative cells are scattered in uninvaginated epithelium, like the `salt-and-pepper' pattern of Delta-expressing neuroblasts that later specify the fate of neighboring cells segregating from the neuroectoderm of the Drosophila embryo (Kunisch et al., 1994; Campos-Ortega,1995; Artavanis-Tsakonas et al., 1999). Hence, Notch signal-mediated lateral specification may occur also during the fate determination in the epithelium of developing PV. Consistently, with this idea, cells in which Su(H)-mediated Notch signaling is activated are found adjacent to the Delta1-expressing cells. As Notch activation leads to the expression of Smad8, an early marker for the glandular epithelium, and Numb or dominant-negative Su(H) transfected cells differentiate only into luminal epithelium,it is conceivable that Notch signaling functions as a binary switch for the glandular and luminal epithelial fate determination. Because the way in which Notch receptors mediate different signaling cascade has not been clarified, we cannot distinguish functions of Notch1 and Notch2 with only an inhibition experiment. However, Notch2 is strongly expressed in the invaginated epithelium, whereas reporter-positive cells are localized in the uninvaginated epithelium, suggesting a possibility that Notch2 might contribute later steps of gland development. We have also tried to investigate Notch2 function by overexpression of NICD derived from chicken Notch2, which resulted in severe apoptosis (data not shown). It seems that Notch1 mainly contributes to the initial phase of gland formation in PV.

We used pTP-1Venus reporter construct to visualize Notch signaling that was introduced into epithelial cells together with CNIC by electroporation method. This made it possible to examine expression of marker genes in cells transfected with CNIC to activate Notch signaling. The reliability of the method depends on the efficiency of transfection. We detected(Fig. 6) expression of marker genes in sections next to ones in which reporter activity was visualized, and found that almost all epithelial cells showed reporter activity. It is therefore logical to conclude that expression patterns of marker genes faithfully reflect the effect of CNIC.

In our assay, transfection of NICD (CNIC and CNIC^ΔC89) inhibits the invagination of the glandular epithelium, as well as the expression of late markers of the gland epithelium such as ECPg, while the activation of Notch signaling allows the expression of Smad8. As BMP2 has previously been shown as an important inducer of gland and ECPg expression(Narita et al., 2000), and as Smad8 is a mediator directly regulated by BMP receptor, Smad8expression pattern and function closely reflects the early state of glandular epithelium before secretion of zymogen. Consistently, a transient reporter assay revealed that Notch-activated cells were interspersed only in uninvaginated epithelium, but that such cells later contribute to gland epithelium. We therefore conclude that Notch signaling prevents maturation of glandular epithelium, while it promotes initial glandular specification. A similar phenomenon is reported in pancreatic development. In the mouse pancreas, Notch activity is required for the commitment of precursor cells to the exocrine lineage (Apelqvist et al.,1999; Jensen et al.,2000), but also represses the maturation of these cells to express digestive enzymes by preventing the function of Ptf1-P48 complex(Esni et al., 2004). Similarly, in developing dorsal root ganglia of chicken embryo, continuous and strong Notch activation inhibits both neuronal and glial terminal differentiation to maintain progenitor population(Wakamatsu et al., 2000),while transient activation of the signaling promotes glial differentiation(Morrison et al., 2000). Many studies have reported a short time activation and downregulation of Notch signaling, especially in somitogenesis (reviewed by Pourquie, 2003; Aulehla and Herrmann, 2004). Thus, it is reasonable to speculate that Notch signaling initially functions as a fate determination switch, and subsequently acts to control the differentiation of immature gland cell precursors.

It has been known that CSL protein mediates Notch signaling by associating with NICD and activates target gene transcription (reviewed by Iso et al., 2003). Recently,however, Notch signaling has also been shown to use an alternative pathway mediated by Deltex/Dtx proteins (Yamamoto et al., 2001; Matsuno et al.,2002; Endo et al.,2003; Hu et al.,2003). In our study, the inhibition of Notch signaling by Su(H)^DBM results in a luminal cell fate, while dominant-negative Dtx has no effect, suggesting the `canonical'Su(H)-mediated pathway is predominantly used in this process. Consistently, Numb overexpression, which interferes with the nuclear translocation of the Notch intracellular domain(Wakamatsu et al., 1999), also inhibited the glandular differentiation cell-autonomously.

In our study, primary effector molecule(s), which function under Notch signaling, cannot be determined, as Hairy1 and Hairy2 were not significantly upregulated by Notch activation. Thus, there is a possibility that Notch1 targets other gene(s), such as unidentified Hairy genes and/or Herp family genes (reviewed by Iso et al., 2003).

It has long been shown that an induction by the mesenchyme is necessary for gland formation of PV epithelium (Haffen et al., 1987; Mizuno and Yasugi, 1990; Yasugi and Fukuda, 2000). Narita et al.(Narita et al., 2000) have previously reported that Bmp2, which is specifically expressed in the PV mesenchyme, can induce gland formation. Recently, both synergistic and antagonistic relationship between BMP and Notch has been reported(Dahlqvist et al., 2003; de Jong et al., 2004; Itoh et al., 2004). In mouse neuroepithelial development, BMP2 induces expression of Hes5 and Hesr1, primary target genes of Notch signaling; Smad1, one of the mediator of BMP signaling, and NICD form a protein complex, which in turn regulates transcription of the target genes(Takizawa et al., 2003). In BMP signaling, ligand-induced heterotetrameric receptor complex directly phosphorylates Smad proteins such as Smad1, Smad5 and Smad8. Thus, Smad8 would mediate the inductive effect stimulated by BMP2 in gland formation of the PV. In our study, activation of Notch signal derived from CNIC induced expression of Smad8. Hence, induction of the downstream component might be one of the interfaces between the BMP and Notch signals that contribute to gland formation.

Shh was shown to induce luminal fate and its downregulation is necessary for glandular differentiation, probably through BMP7, which is expressed in the mesenchyme adjacent to the luminal epithelium(Fukuda et al., 2003). In our study, the activation of Notch signaling downregulates the epithelial expression of Shh. Therefore, it is possible that Notch signaling regulates the glandular differentiation partly through the repression of Shh signaling.

We also demonstrated that EGF signal induces the luminal fate(Takeda et al., 2002). Antagonistic interaction between Notch and EGF pathways have been described in other systems (de Celis et al.,1997; zur Lage and Jarman,1999; Culi et al.,2001). For example, Drosophila ebi, a target gene of EGF receptor, and a mammalian ortholog of TBL1 associate with Su(H) and SMRTER/SMRT, a nuclear co-repressor protein, and regulate transcription of target genes (Culi et al.,2001). Thus, the recruitment of Su(H) to the repression complex may compete with the NICD/Su(H) complex, which activates target genes. Such antagonistic relations may prevent excess gland formation.

In the normal development, gland formation is a sequential process and the number of glands increases from stage 29 until about stage 33, during the period before ECPg expression begins. In our model(Fig. 10A,B), the epithelial cells are homogenous and undifferentiated(Fig. 10, light blue) until stage 28. From stage 29–, slightly before stage 29, Delta1/Notch1 signaling is activated and commits epithelial cells to keep gland progenitor cells (Fig. 10, green) in a scattered fashion, whereas other epithelial cells differentiate into luminal cells (Fig. 10, dark blue). This function of Notch1 might be completed by luminal effectors such as EGF and/or Shh, which prevent formation of excess glands and act in cooperation with inducers of gland cell differentiation such as BMP2. Notch1 activation is sequentially ceased, and progenitor gland cells begin to invaginate and differentiate into early gland cells(Fig. 10, yellow). Notch1 signaling also keeps progenitor gland cells in the undifferentiated state;these cells can then be released to differentiation as the PV becomes large enough to form additional glands. Thus, Delta1/Notch1 signaling definitively regulates early phase of gland formation before the secretion-competent gland maturation in normal development.

We thank Dr H. Saiga for comments on the manuscript. We also thank Drs C. Kintner, O. Pourquié, R. Goitsuka and T. Nohno for plasmids. We are grateful to Dr J. Lewis for valuable discussions. This work was supported in part by the Sasagawa Scientific Research Grant from the Japan Science Society to Y.M., and by a Grant-in-aid on Priority Area from the Ministry of Education, Science, Sport and Culture of Japan (13044002) to S.Y.