Stabilization of β-catenin impacts pancreas growth

β-Catenin function is essential for the canonical arm of the Wnt signaling pathway (reviewed by Willert and Nusse, 1998). In the absence of Wnt ligand, cytoplasmic pools ofβ-catenin are highly unstable because of multiple phosphorylations in the N terminus of the protein that target the protein for degradation. However,when the Wnt co-receptors Lrp and frizzled are engaged by ligand, the assembly of proteins responsible for this phosphorylation state is inhibited. Consequently, the unphosphorylated form of β-catenin accumulates in the cytoplasm, making it available for entry into the nucleus. Once in the nucleus, β-catenin interacts with Tcf/Lef transcriptional co-activators to promote the expression of target genes(Nelson and Nusse, 2004).

Previous work has established that the third exon of the β-catenin gene encodes the N-terminal phosphorylation sites necessary for degradation of the protein via ubiquitylation (Harada et al., 1999). Therefore, the removal of this exon in transgenic mice using Cre/loxP technology results in a constitutively stabilized, or activated, form of the β-catenin protein(Harada et al., 1999). Theseβ -cat^active mice have proven to be useful in probing the effects of β-catenin signaling on embryonic stem cell differentiation, progenitor cell expansion in the nervous system,epithelial-mesenchymal transition in the epiblast and other phenomena(Kemler et al., 2004; Kielman et al., 2002; Zechner et al., 2003).

Previous studies have demonstrated that Wnt signaling components are dynamically expressed within the developing pancreas, suggesting that canonical Wnt signaling may be involved in pancreas organogenesis(Dessimoz et al., 2005; Heller et al., 2003; Murtaugh et al., 2005; Papadopoulou and Edlund,2005). Two independent laboratories recently reported divergent phenotypes resulting from the conditional deletion of β-catenin within the pancreatic epithelium. In one instance, loss of β-catenin resulted in a reduction in pancreatic endocrine cell numbers, whereas the gross morphology of the organ appeared normal at birth(Dessimoz et al., 2005). However, a separate report demonstrated that loss of β-catenin did not affect pancreatic endocrine cell mass, despite the almost complete loss of the exocrine compartment. Here, we have used theβ -cat^active mouse to help clarify how β-catenin stability affects pancreas development and organ maturation.

In mice, pancreas morphogenesis begins by 9.5 days post coitum (E9.5) when epithelial tissue fated to become the dorsal pancreas buds from the gut endoderm within a mesenchymal cap (Kim and Hebrok, 2001). Emergence of two distinct ventral pancreatic buds occurs slightly later, by E10.25-10.5. Signaling by the mesenchyme is essential for epithelial proliferation and branching. The epithelium eventually gives rise to two distinct tissue compartments: exocrine cells that produce digestive enzymes and endocrine cells that produce hormones essential for regulating blood glucose levels.

Pdx1, a homeobox transcription factor, is one of the earliest genes to be expressed within the developing pancreatic epithelium and is essential for normal organ formation. Moreover, Pdx1-expressing pancreatic progenitor cells have been shown to give rise to all three types of pancreatic tissue: endocrine, exocrine and duct (Gu et al., 2002). A number of independent lines of transgenic mice that express Cre recombinase under the control of Pdx1 promoter fragments have been generated (Gannon et al., 2000; Gu et al.,2002; Herrera,2000). Our characterization of two of these strains indicated that the temporal and spatial activity of Cre-recombinase differed. This allowed us to determine how β-catenin stabilization in these distinct temporal/spatial domains of the pancreatic epithelium affected organogenesis and adult organ function. Interestingly, we observed significantly different pancreatic phenotypes depending on the Cre strain employed. Using the PdxCre mice generated in the laboratory of D. Melton that displayed early and robust Cre recombinase activity (PdxCre^early),we observed a nearly complete loss of pancreatic tissue(Gu et al., 2002). Conversely,slightly delayed and more mosaic Cre recombinase expression in Pdx mice generated in one of our laboratories (PdxCre^late) drives outgrowth of pancreatic tissue, resulting in a grossly enlarged pancreas(Gannon et al., 2000; Herrera, 2000). Thus, in one instance, β-catenin stabilization drives tissue loss, and in the other culminates in an increase in organ size relative to body mass. Therefore,ectopic stabilization of β-catenin blocks/deregulates the normal mechanisms that control embryonic pancreas formation and postnatal organ growth.

Noon of the day when vaginal plugs are detected is treated as E0.5 day post coitum. Mice carrying the floxed exon 3 allele of β-catenin(β-cat^active)(Harada et al., 1999) were crossed with strains expressing Cre-recombinase under the control of the Pdx1 promoter (PdxCre^early, PdxCre^late,PdxCre^ER) (Gannon et al.,2000; Gu et al.,2002; Herrera,2000), all of which were maintained in a mixed background. In all experiments presented in this work, the floxed exon 3 allele of β-catenin was maintained as a heterozygote to insure that the level of β-catenin signaling pathway activation was consistent. Thus, theβ -cat^active nomenclature used refers to mice containing one wild-type and one floxed exon 3 allele of β-catenin. To verify the expression pattern of Cre-recombinase, a lacZ reporter line was used (R26R) (Soriano,1999).

Tamoxifen (10 mg/ml) (Sigma, T5648) was dissolved in corn oil (Sigma,C8267) following 30 minutes of incubation at 37°C and vigorous vortexing. Intraperitoneal injections (100 μl, 1 mg/mouse) were made into the pregnant female at the indicated developmental timepoint using a 21-gauge needle.

Embryonic tissues were fixed and paraffin wax imbedded as previously described (Kawahira et al.,2003). Hematoxylin/Eosin staining, immunohistochemical and immunofluorescence analyses were performed as previously described(Kim et al., 1997). The primary antibodies used in this study are listed in Table 1. For immunohistochemistry, a biotinylated anti-goat (Vector; BA-9500) was used at a dilution of 1:200. Staining for diaminobenzidine (DAB) was performed with the ABC Elite immuoperoxidase system (Vector). The Alexa series of secondary antibodies from Molecular Probes was used for the immunofluorescent analysis performed in this study. However, in order to amplify the signal from theβ-galactosidase antibody, we found it necessary to use the TSA Plus Fluorescence system (Perkin Elmer, fluorescein NEL741). Slides were mounted with Vectashield mounting media containing the nuclear stain, DAPI (Vector). Bright-field images were acquired using a Zeiss Axio Imager D1 scope;fluorescent images were captured using a Leica DMIRE2 SP2 confocal microscope.

Fixed PdxCre^early or PdxCre^lateR26R were incubated overnight at room temperature in phosphate-buffered saline (PBS) supplemented with 5-bromo-4-chloro-3-indolyl-D-galactopyranoside(X-gal: 1 mg/ml), 0.02% Triton X-100, 5 mM potassium ferricyanide, and 5 mM potassium ferrocyanide. For whole-mount in situ hybridization, the gastrointestinal cavity of intact embryos was opened to allow greater penetration of fix and probe, and then incubated in 4% paraformaldehyde overnight at 4°C. Whole-mount in situ hybridization with digoxigenin-labeled Fgf10 riboprobe was performed as previously described (Chuang et al.,2003). Stained tissues were photographed with a Leica MZ FL3 dissecting microscope equipped with a Leica IM500 system.

Pancreatic paraffin wax-embedded sections (6 μm) were cut from both the dorsal and ventral pancreas. Following immunofluorescent staining for the proliferation marker phospho-histone H3, positive cells were then scored from 20 non-overlapping fields at 20× magnification from the dorsal and ventral sections of four control and four PdxCre^lateβ -cat^active mice. The average number of cells per field was then normalized against the control. In order to determine endocrine and exocrine cell density, Hematoxylin/Eosin-stained tissues were used to count cell nuclei from 20 non-overlapping fields at 40× magnification isolated from four control and four PdxCre^lateβ -cat^active mice. The average number of nuclei present in the field was then normalized against control.

For P0 pancreata, the whole pancreas was sectioned and aliquoted as described previously (Kawahira et al.,2003), in order to obtain representative results. Islet area was assayed at P0 as described previously(Hebrok et al., 2000; Perez et al., 2005). Error bars represent s.e.m., and confidence intervals were determined using a Student's t-test analysis.

Six control and six PdxCre^lateβ -cat^active mice were fasted for 14 hours before intraperitoneal injection of a 20% glucose (w/v) solution at a dose of 2 g per kg body mass. Venous blood glucose readings were then taken at the indicated intervals using a Bayer Ascensia Elite XL to analyze samples collected from tail nicks. Error bars represent standard error of the mean.

The PdxCre^late transgenic mouse was generated using a 4.5 kb promoter fragment, while the PdxCre^early transgenic mouse used a 5.5 kb fragment of the promoter. Because the size of the promoter fragment used, the site of integration, and other factors that can affect the expression of the transgene, we directly compared the activity of the Cre-recombinase by crossing both PdxCre strains to the reporter line R26R, which carries a β-galactosidase gene whose transcription is activated after Cre-mediated recombination(Soriano, 1999). In this manner, we were able to map the timing and expression of pancreatic Cre-recombinase. By E10.5, robust lacZ staining was detected in the pancreatic epithelium of PdxCre^early R26R animals at the gross and histological levels (Fig. 1C,F). Moreover, expression was detectable in the majority of Pdx1+ cells observed (Fig. 1F). Interestingly, no lacZ staining was observed in the PdxCre^late R26R animals at this time point(Fig. 1B,E). By E11.5, a few lacZ-positive cells were found in the pancreatic epithelium of PdxCre^late R26R, a delay of ∼24 hours (data not shown)from the onset of expression in the PdxCre^early. Furthermore, comparison of lacZ/Pdx1 co-stained pancreatic tissue revealed that even at E12.5, the Cre expression in the PdxCre^late strain was more mosaic than in the PdxCre^early (Fig. 1H,I).

Analysis of lacZ stained pancreas sections in adult PdxCre^late R26R mice suggested that Creexpression within the terminally differentiated exocrine and endocrine cells remained mosaic in the PdxCre^late strain(Fig. 1K). Interestingly,β-galactosidase can rarely be detected within the adult pancreatic ducts in PdxCre^late R26R mice(Fig. 1N). By comparison, the majority of endocrine, exocrine and ductal cells in PdxCre^earlyR26R mice exhibit β-galactosidase activity(Fig. 1L,O). The higher number of lacZ+ cells in the adult PdxCre^early R26R mice strongly indicates that the onset of Cre expression was not only earlier in timing, but also targeted a greater portion of the pancreatic epithelium. Control tissue was stained for each of the time points assayed to demonstrate the specificity of the lacZ staining reaction/immunofluorescence (Fig. 1A,D,G,J,M). Therefore, the PdxCre^late and PdxCre^early mice were used in this study as tools to probe the effect of increased β-catenin signaling on distinct temporal/spatial populations of cells in the early embryonic pancreas.

In order to assess the consequences of increased β-catenin signaling on pancreas organogenesis, we crossed β-cat^activeanimals with the PdxCre^late or PdxCre^early mice. Because previous studies have shown that the loss of the third exon of one allele of β-catenin is sufficient to drive strong increases in Wnt pathway activity, all experiments were carried out using mice that were heterozygous for the floxed β-catenin allele(Harada et al., 1999; Kemler et al., 2004; Zechner et al., 2003). Analysis of gross morphology and pancreas architecture at E18.5 in the PdxCre^late β-cat^active mice did not reveal any overt changes compared with control littermates(Fig. 2A,B,D,E). By contrast,the PdxCre^early β-cat^activeanimals displayed near total pancreas agenesis, and the pancreatic remnant contained multiple large cysts (Fig. 2C,F). Histological examination revealed a significant reduction in the epithelial derived exocrine and endocrine tissues. Consequently, PdxCre^early β-cat^active survive on average only 7 days after birth, whereas PdxCre^lateβ -cat^active are viable and reproductively active.

Nuclear β-catenin localization was abundant and easily detected by confocal microscopy in both PdxCre^lateβ -cat^active and PdxCre^earlyβ -cat^active pancreata at E18.5(Fig. 2H,I). In control samples, β-catenin was detected only at the plasma membrane(Fig. 2G). Therefore, as has been shown in other tissues (Jamora et al., 2003; Miller and Moon,1997; Tolwinski and Wieschaus,2004), stabilization of β-catenin leads to increased nuclearβ-catenin signaling and, presumably, hyperactivation of the canonical Wnt signaling pathway in the pancreas.

In addition to its role in the canonical Wnt signaling pathway, theβ-catenin protein also participates in cell adhesion at adherens junctions. β-Catenin links the cytoplasmic domain of transmembrane cadherins to the actin cytoskeleton via its association with the adaptor protein α-catenin (reviewed by Bienz,2005). Therefore, β-catenin stabilization may also impact cell adhesion. However, despite clear evidence of nuclear localization ofβ-catenin in PdxCre^lateβ -cat^active and PdxCre^earlyβ -cat^active mice, E-cadherin remained properly localized to the plasma membrane (Fig. 2K,L inset), suggesting that adhesion has not been disrupted. Similarly, other studies using the β-cat^active mouse strain have not found adhesion defects in cells expressing the stabilized form of β-catenin (Gounari et al.,2002; Harada et al.,1999). Using E-cadherin as a marker of epithelial cells,immunohistochemical analysis revealed the extensive loss of epithelial tissue mass in the PdxCre^earlyβ -cat^active organ remnant. As expected, the developing cysts are E-cadherin positive, indicating that the cysts were epithelial in origin (Fig. 2L).

By E18.5, endocrine islets in mice start to form a stereotypical structure consisting of a core of insulin producing β-cells surrounded by glucagon producing α-cells (Fig. 2M). Although the PdxCre^lateβ-cat^active mutants displayed islet architecture(Fig. 2N) and islet area (data not shown) equivalent to control, PdxCre^earlyβ -cat^active mutants had few insulin⁺ or glucagon⁺ cells (Fig. 1O). In addition, these endocrine cells were found scattered throughout the remaining organ, rather than organized into discrete islet structures (Fig. 2O).

Thus, early and widespread upregulation of the β-catenin signaling pathway in the PdxCre^earlyβ -cat^active mouse strain prevents normal formation of the exocrine and endocrine compartments of the pancreas. However, the delayed and more mosaic upregulation of the β-catenin signaling pathway in the PdxCre^late β-cat^active mutants appears to be well tolerated and does not result in any obvious developmental defect.

Other differences between the PdxCre^early and PdxCre^late mouse strains, beyond the delay in onset of Cre expression, might account for the difference in phenotypes observed. For example, differences in the specific subset of cells targeted by each Cre strain after E12.5 when both lines are active might be responsible for driving the pancreas hypoplasia observed. Therefore, we directly assessed the temporal dependence of the phenotype seen in the PdxCre^earlyβ-cat^active mice by crossing theβ -cat^active mouse to an inducible PdxCrestrain (PdxCre^ER) (Gu et al., 2002) containing the same 5.5 kb promoter fragment as the PdxCre^early mouse. In this system, the Crerecombinase is expressed as a fusion protein with the estrogen receptor, and remains inactive in the cytoplasm in the absence of tamoxifen ligand. However,once bound to tamoxifen, the Cre recombinase enters the nucleus where it can catalyze recombination (Gu et al.,2002).

Injection of tamoxifen at E11.5 induced accumulation of β-catenin in the nucleus of a large number of pancreatic epithelial cells in PdxCre^ER β-cat^active mice(Fig. 3J), resulting in a severe reduction of pancreas mass in all mutants observed (n=9, Fig. 3B). Nuclear localization of β-catenin was not observed in control pancreata(Fig. 3I). Moreover, cystic structures similar to those seen in PdxCre^earlyβ -cat^active are apparent in histological sections of the pancreas (Fig. 3F,indicated by arrows).

Conversely, injection of tamoxifen at E12.5 induced more variable phenotypes. The majority of PdxCre^ERβ -cat^active mutants exhibited an intermediate phenotype with some reduction of the ventral and dorsal pancreas obvious in examination of the gross morphology of the organ(Fig. 3C). The pancreas histology of these intermediate PdxCre^ERβ -cat^active mutants(Fig. 3G) appeared equivalent to control (Fig. 3E). However,a smaller number of PdxCre^ERβ -cat^active mutants from litters injected with tamoxifen at E12.5 displayed either pancreas hypoplasia similar to those in the E11.5 injection group (Fig. 3B) or appeared unaffected (exact proportion of mutants in each category summarized in Fig. 3M). Despite the variability in phenotype, significant numbers of cells with nuclear β-catenin localization were present in the pancreatic epithelium in all mutants analyzed (Fig. 3K). Injection of tamoxifen at E13.5 did not disrupt pancreas formation in any of the eleven PdxCre^ERβ -cat^active mutants observed(Fig. 3D,H), despite clear nuclear localization of β-catenin in a large number of pancreatic cells(Fig. 3L).

Taken together, these results indicate that β-catenin stabilization can adversely affect pancreas development only during the earlier stages of organ formation. After E12.5, the increased levels of canonical Wnt signaling that the stabilized β-catenin presumably induces are well tolerated. Therefore, it appears likely that the pancreas phenotype seen in the PdxCre^early β-cat^active is dependent on the earlier onset of Cre recombinase expression seen when compared with the PdxCre^late mouse strain.

In order to analyze the defects in PdxCre^earlyβ -cat^active mice in more detail, we performed a series of histological and molecular assays aimed at characterizing the early progression of the phenotype and identifying the underlying molecular mechanisms involved. By E12.5, stabilization of β-catenin within the pancreas of PdxCre^earlyβ -cat^active mice caused the pancreatic epithelium in some mutants to exhibit abnormal dilation not seen in control tissues,representing the earliest histological change that we could detect (data not shown). By E15.5, the rudimentary clusters of exocrine cells seen in control embryos are almost completely absent in the PdxCre^earlyβ -cat^active (data not shown). In addition, abnormally enlarged pancreatic ducts are frequently observed at this time point in PdxCre^early β-cat^activepancreatic tissues and it is likely that these dilated ducts later form the cystic structures seen at E18.5 (data not shown).

Numerous studies have shown the importance of mesenchymal-epithelial interactions for proper pancreas formation. One of the mesenchymal molecules known to regulate expansion of pancreatic epithelial cells is fibroblast growth factor 10 (Fgf10) (Bhushan et al.,2001). Fgf10 expression is normally detectable within the mesenchyme surrounding the pancreatic bud beginning at E9.5 and peaking at E10.5. By E12.5, Fgf10 expression can no longer be detected. Loss of Fgf10 in mice has been shown to disrupt pancreas formation(Bhushan et al., 2001), a phenotype similar to the one we observed in PdxCre^earlyβ -cat^active mice. Therefore, we asked whether stabilization of β-catenin resulted in changes in mesenchymal Fgf10 expression. Whole-mount in situ hybridization at E10.5 revealed that Fgf10 expression is decreased in the pancreatic mesenchyme of PdxCre^early β-cat^active animals(Fig. 4C) when compared with control (Fig. 4A). Fgf10 expression appeared normal in the PdxCre^late β-cat^active mutants(Fig. 4B). This observation supports our R26R expression analysis, which indicated that Cre recombinase is not active in PdxCre^lateβ -cat^active mice at this time point. Thus,β-catenin stabilization in pancreas epithelium before E11.5 disrupts an important component of the signaling exchange that occurs between the mesenchyme and epithelium, which at least partially explains the dramatic defects in organogenesis observed.

Wnt signaling has also been implicated in maintaining the expression of hedgehog (Hh) ligands in both Drosophila wing discs and mouse limb buds (Parr and McMahon, 1995; Pinson et al., 2000; Tabata and Kornberg, 1994). In addition, ectopic Wnt signaling in the epidermis has been shown to upregulate the expression of Hh signaling pathway components(Silva-Vargas et al., 2005). Hedgehog signaling, in turn, stabilizes Wnt expression in Drosophilawing (Tabata and Kornberg,1994), indicating that both pathways can crossregulate the activity of one another. A previous study has shown that overexpression of the Hh ligand, Shh, under the Pdx1 promotor results in pancreas agenesis (Apelqvist et al.,1997), a phenotype similar to that observed in the PdxCre^early β-cat^active mice. To determine whether stabilization of β-catenin activates Hh signaling in pancreatic tissue, we analyzed E12.5 pancreas tissue from wild-type, PdxCre^late β-cat^active and PdxCre^early β-cat^active mice. By immunohistochemistry, protein levels of the Hh receptor Ptch1, a direct transcriptional target of Hh signaling(Goodrich et al., 1996), are significantly upregulated within the pancreatic epithelium of the PdxCre^early β-cat^active mice(Fig. 4F) when compared with control or PdxCre^late β-cat^active(Fig. 4D,E). Protein levels of Hh ligand, are also increased in the PdxCre^earlyβ -cat^active (Fig. 4I) when compared with control or PdxCre^lateβ -cat^active mice(Fig. 4G,H). Therefore,stabilization of β-catenin in the PdxCre^earlyβ -cat^active mice results in upregulation of the Hh signaling pathway within the early pancreatic epithelium, which may also partially account for the dramatic changes observed during organ formation/morphogenesis.

A previous study has indicated that Fgf10^-/- mice exhibit a marked loss of the Pdx1⁺ progenitor cells by E12.5-E13.5(Bhushan et al., 2001). Additionally, loss of Pdx1 in mice has been shown to cause pancreas agenesis(Offield et al., 1996; Jonsson et al., 1994). Therefore, we examined Pdx1 expression within the E12.5 pancreatic epithelium. Interestingly, PdxCre^earlyβ -cat^active mice exhibited a 76% reduction in Pdx1⁺ progenitor cells (Fig. 4L). By contrast, PdxCre^lateβ -cat^active mice had numbers of Pdx1⁺progenitor cells (Fig. 4K) that were equivalent to control (Fig. 4J). Moreover, only cells that did not display elevated levels ofβ-catenin retained Pdx1 expression (data not shown). Presumably, these cells escaped recombination. Cells with a clear increase in cytoplasmic/nuclear β-catenin were rarely detected in the PdxCre^late β-cat^active mutant pancreatic epithelium at E12.5 (data not shown).

Thus, the pancreatic hypoplasia and cyst formation in the PdxCre^early β-cat^active appears to be mediated by loss of Fgf10 signaling, concomitant with an increase in Hh signaling. Together, these perturbations in early pancreas specification may contribute to the loss of Pdx1⁺ pancreatic progenitors.

While the morphology and histological architecture of pancreata dissected from PdxCre^late β-cat^activeappeared normal at the end of development(Fig. 2B,E), pancreas mass was not. Despite the fact that mutant pups were equivalent in body mass to control littermates, pancreas mass at day 0 was increased by 53%(Fig. 5B). Pancreas mass continued to increase independent of animal mass after birth, resulting in a grossly enlarged organ (Fig. 5A) that was 2.7-fold greater than control pancreata at 98 days postnatally (Fig. 5B). In transgenic mice older than 1 year, the pancreas had grown 4.6-fold larger than those found in control littermates (Fig. 5B). Strong nuclear β-catenin staining could be detected within the adult exocrine pancreas, suggesting that increased Wnt activity in the PdxCre^late β-cat^active mice was responsible for the increase in organ size after birth. In support of this hypothesis, we found a 2.5 fold increase in the number of proliferating adult exocrine cells in the PdxCre^lateβ -cat^active (Fig. 5I), as revealed by staining for the mitotic marker phosphohistone H3 (Fig. 5E,F).

Interestingly, PdxCre^ERβ -cat^active mice injected with tamoxifen 4 weeks after birth, also showed a twofold increase in pancreas mass by 1 year of age(n=8, data not shown). Injection of PdxCre^ER R26Rmice with tamoxifen results in clear Cre recombinase activity (as indicated by lacZ staining) in a significant proportion of the exocrine pancreas at this stage (data not shown). Thus, as has been shown for other tissues(Bierie et al., 2003; Huelsken and Birchmeier,2001), β-catenin signaling, as a part of the canonical Wnt signaling pathway, can act as a proliferative signal in mature exocrine pancreas. In addition, the density of cells within the exocrine pancreas, as determined by counting the number of nuclei present in a defined histological field, is also increased 1.8-fold in PdxCre^lateβ -cat^active when compared with controls(Fig. 5G,H,J). By contrast, the density of endocrine cells within the mutant is normal. Although these data suggest that exocrine cell size is reduced in PdxCre^lateβ -cat^active mutants, the mechanism driving this phenomenon is unclear. Stabilization of β-catenin within the adult exocrine tissue may directly deregulate cell size. However, it is also possible that the rapidly proliferating exocrine cells do not need to reach their normal size before dividing again. Electron microscopy will be used in future studies to further characterize how this mutation affects pancreatic cellular ultrastructure.

Despite the dramatic increase in exocrine cell proliferation and the resultant increase in organ mass, pancreatic tumors have not been observed in any of the PdxCre^late β-cat^activemutants, up to 1 year of age. The mice are viable and fertile, and appear healthy.

As shown in Fig. 1K, islets within the PdxCre^late strain were mosaic for Cre activity. Therefore, only a subset of β-cells were targeted for stabilization ofβ-catenin in the PdxCre^lateβ -cat^active mouse. Immunofluorescence staining indicated that PdxCre^lateβ-cat^active adult mutants exhibit stereotypical murine islet architecture, consisting of a central core of insulin producingβ-cells and a periphery of glucagon-producing α-cells(Fig. 6A,B). Despite the increase in overall organ size at birth, the total cross sectional area of endocrine islets remained equivalent to control (data not shown), suggesting that the ratio of endocrine content to animal mass was not perturbed. In addition, analysis of relative hormone content in pancreas homogenates at three months of age indicated that the total amount of insulin and glucagon present in the PdxCre^lateβ -cat^active is equivalent to control (data not shown). Thus, in contrast to the vast expansion of the exocrine compartment of the pancreas, the size of the endocrine compartment remains largely unaffected.

As expected from the R26R expression data, the majority ofβ-cells in adult PdxCre^lateβ -cat^active mutants exhibited plasma membrane localization of β-catenin that appeared equivalent to control(Fig. 6C,D, higher magnifications are shown in Fig. 6F,G) and only a subset of β-cells was marked by upregulatedβ-catenin levels. However, unlike the exocrine compartment(Fig. 5D), confocal analysis revealed that Cre-mediated excision in β-cells did not result in a significant increase in nuclear β-catenin, but instead in an increase in cytoplasmic β-catenin levels (Fig. 6E, higher magnification can be found in Fig. 6H).

Within the β-cells marked by high levels of cytoplasmic β-catenin protein, expression of insulin (Fig. 6H), adult β-cell transcription factors Pdx1(Fig. 6J) and Nkx6.1 (data not shown), and the glucose transporter Glut2 (data not shown) appear equivalent to control (Fig. 6F,I). These results demonstrate that the β-cells are fully differentiated, despite the dramatic increase in cytoplasmic β-catenin. Additionally, fasting glucose tolerance is normal in PdxCre^lateβ -cat^active animals(Fig. 6J), suggesting that either β-cell function is normal in these mutants or that the number ofβ-cells expressing the stabilized form of β-catenin is insufficient to affect this physiological assay. Future experiments using cultured islets are necessary to determine what biochemical mechanisms might protect pancreatic endocrine cells from accumulating nuclear β-catenin.

Surprisingly, analysis of β-catenin expression patterns in 1-year-old mice revealed the presence of cells within a small subset of islets that exhibited nuclear localization of β-catenin(Fig. 7B, higher magnification is in Fig. 7D), a phenomenon not seen in younger mice. Despite their location in the center of islets,where β-cells typically are found, these cells do not express insulin(Fig. 7B, higher magnification is in Fig. 7D). These cells also lack expression of β-cell markers, including Pdx1(Fig. 7F, higher magnification is in Fig. 7H), Nkx6.1 and Glut2, or markers of other endocrine cell populations (glucagon, somatostatin,PP; data not shown). Moreover, these cells do not express the pancreatic duct marker mucin 1 or the exocrine marker, amylase (data not shown). Fasting glucose tolerance is also normal in PdxCre^lateβ -cat^active mutants at 1 year of age (data not shown), indicating that the presence of these abnormal cells in a subset of islets has not impacted this physiological assay. Given the lack of expression of pancreatic endocrine, exocrine or ductal markers, these cells potentially represent a subpopulation of cells that have dedifferentiated as a result of the accumulation of nuclear β-catenin. Further genetic experimentation is necessary to identify the lineage of this cell population.

Our studies underscore the fact that important differences in Cre-activity can be found in mouse Cre-lines that use similar promoter elements of the same gene. Differences in the site of integration, the length of the promoter fragment used, and gene silencing can all exert significant influence over the transgenic expression of Cre recombinase. Here, we show how differences in the temporal/spatial stabilization of the β-catenin protein within the developing pancreas result in dramatically different phenotypes: organ ablation versus organ enlargement. The PdxCre^early mouse strain efficiently targets all three pancreatic lineages - endocrine, exocrine and duct - and is expressed early in pancreatic development. By contrast, the PdxCre^late strain targets a subset of the endocrine and exocrine cell lineages, but does not appear to be expressed in a significant number of duct cells.

Our observation that stabilization of β-catenin in PdxCre^early βcat^active mice results in upregulation of Hh signaling components concomitant with a loss of Fgf10 suggests that the activities of these three major embryonic signaling pathways are intimately linked during this stage of pancreas development. Although we cannot exclude the possibility that Wnt signaling in the epithelium directly affects the expression of mensenchymal Fgf10, it is interesting to note that upregulation of Hh signaling resulting from loss of the Hh inhibitor Hhip is known to cause a delay in Fgf10expression in the pancreatic mesenchyme(Kawahira et al., 2003). Therefore, our results are consistent with the idea that at least some of the defects observed in epithelium and mesenchyme are mediated via increased Hh activity. A reduction of pancreas size has also been observed in transgenic mice expressing Wnt1 under the control of the Pdx1 promoter(Heller et al., 2003). The phenotype observed in the PdxCre^earlyβ cat^active mice results from upregulation ofβ-catenin signaling in a cell-autonomous fashion within the epithelium. However, transgenic expression of the soluble Wnt1 ligand presumably exerts a direct effect on both the pancreatic mesenchyme and epithelium. Therefore, it would be interesting to compare whether Hh and FGF signaling are also affected in the Pdx-Wnt1 mice. Moreover, future studies are necessary to determine the precise signaling hierarchy between the Wnt, Hh and FGF signaling pathways during normal pancreas specification and organogenesis.

Despite the 4.6-fold increase in mass, the pancreas of PdxCre^late β-cat^active mutant remains functional. PdxCre^lateβ -cat^active animals have a body mass equivalent to control littermates, suggesting that the duodenum still receives the pancreatic enzymes necessary for digestion, and their glucose tolerance profile is normal, indicating that the endocrine function of the pancreas is intact. Moreover, the shape of the organ is also conserved; it is simply bigger. Despite the increased proliferation, the PdxCre^late β-cat^active mice do not appear to develop tumors.

Interestingly, although the pancreas becomes grossly enlarged, the exocrine cells that comprise it appear smaller. Modulation of Wnt signaling has been implicated in increasing the size of skeletal muscle cells after mechanical overload (Armstrong and Esser,2005). In addition, Wnt activation via overexpression of a canonical Wnt ligand has been shown to result in chondrocyte hypertrophy(Day et al., 2005). Although the opposite was observed in the PdxCre^lateβ -cat^active mice, it is possible that deregulation of Wnt signaling in pancreatic exocrine cells may directly impact cell size. However, exocrine cells contain a large number of granules whose digestive enzyme content is emptied into the gut in response to feeding. Because the PdxCre^late β-cat^active mutant pancreas contains significantly more exocrine cells than are present in the wild type, a feedback mechanism might exist that causes a reduction in the number of granules in each cell, thereby reducing cell size. Ultrastructural analysis is required to precisely determine how β-catenin stabilization affects pancreatic exocrine cell morphology.

The absence of a proliferative response or enlargement of the β-cell compartment, within PdxCre^lateβ -cat^active mutants may reflect differences in the way in which exocrine and endocrine cells respond to activation ofβ-catenin. Alternatively, β-cells may have a mechanism that allows for the active exclusion of stabilized β-catenin from the nucleus, which then prevents downstream activation of the canonical Wnt signaling pathway. A recent study involving β-catenin stabilization in preimplantation embryos also observed similar cytoplasmic localization of β-catenin(Kemler et al., 2004). The authors propose several mechanisms that might prevent localization of the stabilized form of β-catenin in the blastomere nuclei. Among these are the presence of an alternative ubiquitylation pathway that can degrade even the stabilized form of β-catenin or the activity of endogenous inhibitors that might prevent β-catenin from exerting its nuclear activity. Endocrine cells may lack, or have low levels of, nuclear transport proteins required for normal translocation and retention of β-catenin in the nucleus. Experiments using cultured islets and known inhibitors of nuclear transport might prove instructive in uncovering the mechanism underlying the resistance of the pancreatic endocrine cell to nuclear accumulation ofβ-catenin.

One of the truly puzzling findings of our study is that a low number of cells with clear nuclear localization of β-catenin can be found only in the islets of one year old PdxCre^lateβ -cat^active animals. Thus, it is possible that pancreatic endocrine cells possess some kind of innate mechanism to prevent nuclear localization of β-catenin that was gradually circumvented over time. Alternatively, it might be a matter of the cellular dose of the stabilized β-catenin; it might take time for enough of the protein to accumulate to affect cell function.

Moreover, the location of these cells within the center of the islet raises the intriguing possibility that these are β-cells that have undergone de-differentiation, after accumulating sufficient levels of nuclearβ-catenin. Genetic cell lineage tracing experiments would need to be performed to confirm this hypothesis. The absence of Pdx1 expression in these cells is consistent with the fact that the pancreatic epithelial cells with nuclear β-catenin accumulation in the PdxCre^earlyβ -cat^active also display a loss of Pdx1 expression at E12.5. However, the absence of normal pancreatic cell lineage markers makes it impossible to conclude what these cells were/are without further experimentation.

Given the fact that fewer than 30% of β-cells are marked by cytoplasmic β-catenin staining, it is also possible that potentialβ-cell defects in β-catenin-activated cells are masked by the remaining wild-type cells. Therefore, although the results presented here are intriguing, future studies using a transgenic strain that expresses Cre in a higher percentage of β-cells are necessary to clarify these findings regarding the effect of β-catenin stabilization on β-cell differentiation, expansion and function.

Our findings further illustrate the dual nature of β-catenin signaling. In one context, β-catenin activation prevents proper differentiation and expansion of early pancreatic progenitor cells; in another context, β-catenin activation acts as a proliferative cue that induces gross enlargement of the exocrine pancreas. Conversely, deletion ofβ-catenin has been shown to cause loss of exocrine cell mass(Murtaugh et al., 2005) and,in a different context, reduction of endocrine islet mass(Dessimoz et al., 2005). Taken together, these results highlight the complex roles β-catenin signaling plays in pancreas organ growth and the determination of the exocrine:endocrine cell ratio.

We are indebted to Drs Doug Melton for providing the PdxCre^earlymouse strain and Mike German for providing antibody reagents. We also thank Drs Mike German, Gail Martin, Ingrid Rulifson and Seung Kim for helpful discussion; Drs David Cano, Shigeki Sekine and Grace Wei for critical reading of the manuscript; and Heather Heiser for her assistance in maintaining our mouse stocks. Work in M.H.'s laboratory was supported by grants from the NIH(U19 DK61245, DK60533-01A1, CA112537-01), the Juvenile Diabetes Research Foundation and the Larry L. Hillblom Foundation. P.W.H. and J.L. are part of the UCSF Biomedical Science graduate student program. P.L.H. was supported by grants from the Juvenile Diabetes Research Foundation, the NIH/NIDDK (U19 DK42502, VUCM CA#9860) and the Swiss National Science Foundation(3100A0-103867).