Primary cilia regulate Gli/Hedgehog activation in pancreas

Abstract

Previous studies have suggested that defects in pancreatic epithelium caused by activation of the Hedgehog (Hh) signaling pathway are secondary to changes in the differentiation state of the surrounding mesenchyme. However, recent results describe a role of the pathway in pancreatic epithelium, both during development and in adult tissue during neoplastic transformation. To determine the consequences of epithelial Hh activation during pancreas development, we employed a transgenic mouse model in which an activated version of GLI2, a transcriptional mediator of the pathway, is overexpressed specifically in the pancreatic epithelium. Surprisingly, efficient Hh activation was not observed in these transgenic mice, indicating the presence of physiological mechanisms within pancreas epithelium that prevent full Hh activation. Additional studies revealed that primary cilia regulate the level of Hh activation, and that ablation of these cellular organelles is sufficient to cause significant up-regulation of the Hh pathway in pancreata of mice overexpressing GLI2. As a consequence of overt Hh activation, we observe profound morphological changes in both the exocrine and endocrine pancreas. Increased Hh activity also induced the expansion of an undifferentiated cell population expressing progenitor markers. Thus, our findings suggest that Hh signaling plays a critical role in regulating pancreatic epithelial plasticity.

The Hedgehog (Hh) signaling pathway plays a critical role in pancreas development and function. Loss- and gain-of-function studies in mice have revealed that deregulation of Hh activity affects pancreas morphogenesis and function (reviewed in ref. 1). In these studies, Hh activity was deregulated by either manipulation of Hh ligands, which can signal both to the pancreatic epithelium and mesenchyme, or by the use of conventional mutant mice (1–5). Therefore, these studies do not differentiate between epithelial and mesenchymal Hh signaling. Hh signaling is known to mediate communication between adjacent tissues, and several recent reports have addressed the differential requirement for Hh specifically in the pancreatic epithelium and mesenchyme. Epithelial cell-specific activation of Hh signaling appears not to have any major effect on pancreatic development (6, 7), and Hh elimination in epithelial cells does not result in profound changes during pancreas formation (8, 9). Conversely, pancreatic mesenchyme appears to be the predominate receiving tissue of Hh signaling, and inappropriate stimulation interferes with the epithelial–mesenchymal crosstalk essential for pancreas organogenesis (2, 5). Similar to embryonic development, inhibition of Hh signaling by inactivation of the Hh transducer Smoothened (Smo) in epithelial cells does not affect pancreatic adenocarcinoma formation (8), whereas Smo inactivation in stromal cells results in growth inhibition in pancreatic cancer xenograft models (10). Thus, these observations indicate that mesenchymal Hh signaling plays the major role during pancreas development and pancreatic cancer, and suggest that the developing pancreatic epithelium is insensitive to deregulation of Hh signaling.

However, previous studies from our group have shown that pancreas-specific overexpression of a version of GLI2 lacking the N-terminal repressor domain (GLI2ΔN) in pancreatic epithelial progenitor cells in Pdx1-Cre;CLEG2 mice results in the formation of undifferentiated tumors (7). These findings suggest an additional, cell-autonomous role of activated Hh signaling within the mature pancreas epithelium. To determine whether activation of Hh signaling in the pancreatic epithelium also affects pancreas formation, we have analyzed pancreas organogenesis in Pdx1-Cre;CLEG2 mice.

Surprisingly, we find that ectopic expression of GLI2ΔN fails to efficiently up-regulate Hh pathway within the pancreas epithelium. This observation suggests that mechanisms exist in pancreatic epithelial cells that block inappropriate activation of the pathway. Recent studies have shown that primary cilia, cellular organelles, are critical regulators of the Hh pathway during embryonic development, organ function, and in cancer (11–15). Specifically, cilia ablation increases Hh activation mediated by GLI2ΔN during medulloblastoma and basal cell carcinoma (BCC) formation (11, 15). Our findings indicate that concomitant elimination of cilia in the presence of GLI2ΔN in mice results in overt Hh activation in pancreatic epithelium and, consequently, impaired pancreas formation. These pancreata display a significant loss of both exocrine and endocrine tissue accompanied by the appearance of undifferentiated epithelial cells expressing pancreatic progenitor cell markers. Thus, our study reveals a role for primary cilia in regulating Hh signaling during pancreas formation and demonstrates that excessive Hh activation results in unique phenotypes in the pancreas, underscoring a potential role for Hh signaling in modulating the differentiated state of pancreatic cells.

Results

Primary Cilia Prevent Full Hh Activation upon GLI2ΔN Overexpression.

We have recently shown that, in Pdx1-Cre;CLEG2 transgenic mice, GLI2ΔN accumulation is observed in a mosaic fashion within the pancreatic epithelium. The activated GLI2ΔN expressed in CLEG2 mice is fused to a myc-tag in its N terminus, thus allowing for immunodetection by an anti-myc antibody (myc-GLI2ΔN, hereafter) (7). The restricted expression pattern of myc-GLI2ΔN is surprising because the CLEG2 transgene should be transcribed in all pancreatic cells due to the efficient elimination of the preceding lox-GFP-stop-lox cassette that places the myc-GLI2ΔN transgene under direct control of the strong ubiquitous CMV early enhancer/chicken β-actin (CAG) promoter (7). To determine whether expression of the CLEG2 transgene in the pancreas indeed leads to activation of the Hh signaling pathway, we crossed Pdx1-Cre;CLEG2 mice with Ptch1^lacZ/+ mice. Ptch1 is a direct transcriptional target of Hh signaling, and Ptch1^lacZ/+ mice carrying the β-galactosidase (β-gal) gene (LacZ) in the endogenous Ptch1 locus serve as accurate reporters of Hh pathway activity (16). Analysis of β-gal activity in 3-week-old Pdx1-Cre;CLEG2;Ptch1^lacZ/+ mice revealed few cells within the pancreas displaying detectable activity (Fig. 1A), suggesting that control mechanisms present in pancreatic epithelial cells prevent complete activation of Hh signaling in Pdx1-Cre;CLEG2 mice.

Fig. 1.

Primary cilia prevent full Hh activation in pancreas of myc-GLI2ΔN-overexpressing mice. (A) Analysis of β-gal activity in 3-week-old Pdx1-Cre;CLEG2;Ptch1^lacZ/+ mice revealed few cells within the pancreas displaying detectable activity. (B) Ablation of primary cilia through Kif3a inactivation results in a significant increase of both the area and intensity of β-gal staining of positive cells in the pancreata of Pdx1-Cre;CLEG2;Kif3a^lox/lox;Ptch1^lacZ/+ mice. (C) Myc-GLI2ΔN protein accumulates in a limited number of cells in Pdx1-Cre;CLEG2 pancreata (arrowheads). (D) In contrast, the number of pancreatic cells marked by strong expression of the myc-GLI2ΔN fusion protein significantly increases in Pdx1-Cre;CLEG2;Kif3a^lox/lox mice. Note the increased stromal compartment (asterisks) and the presence of epithelial cell nests accumulating high levels of myc-GLI2ΔN (outlined). (E and F) Myc-GLI2ΔN fusion protein localizes to primary cilia in ductal cells of Pdx1-Cre;CLEG2 mice. Note that the areas of strong myc staining in the Pdx1-Cre;CLEG2 sample correspond to one of the few duct areas in which the transgene was active.

Primary cilia regulate the level of Hh signaling during mouse development in different organs and tissues (17, 18), and therefore could also potentially regulate Hh signaling in the pancreas. Importantly, cilia have been recently shown to repress Hh activation mediated by myc-GLI2ΔN during medulloblastoma and BCC formation (11, 15). To address the role of cilia in pancreatic epithelial Hh signaling, we generated compound Pdx1-Cre;CLEG2;Kif3a^lox/lox;Ptch1^lacZ/+ mice characterized by ectopic expression of myc-GLI2ΔN (CLEG2), loss of primary cilia via elimination of the Kif3a gene (Kif3a^lox/lox), one of the key components of the kinesin-2 complex that is required for ciliogenesis (19), and expression of Ptch1-lacZ as a marker for Hh activity (Ptch1^lacZ/+). A significant increase both in the area and intensity of β-gal staining of positive cells was observed in the pancreata of Pdx1-Cre;CLEG2;Kif3a^lox/lox;Ptch1^lacZ/+ compared with Pdx1-Cre;CLEG2;Ptch1^lacZ/+ mice during postnatal (Fig. 1B) and embryonic stages (Fig. S1). Interestingly, Hh activity remained mostly localized to the main ductal branches, one of the ciliated cell types in the pancreas (20), at e15.5 and e17.5, in Pdx1-Cre;CLEG2;Kif3a^lox/lox;Ptch1^lacZ/+ mice. Of note, cilia ablation in Pdx1-Cre;Kif3a^lox/lox;Ptch1^lacZ/+ mice resulted in decreased β-gal activity during embryonic stages compared with Ptch1^lacZ/+ controls (Fig. S1), thus suggesting a role for primary cilia in regulating endogenous Hh activity. Importantly, β-gal assay conditions used at embryonic stages were more sensitive than those used in 3-week-old mice (SI Methods). Thus, Hh signaling in postnatal pancreata is active at low levels in ductal cells and islets (21). Quantitative measurements of expression of Hh target genes in whole pancreata revealed that whereas Gli1 and Ptch1 expression was marginally increased in Pdx1-Cre;CLEG2 tissue, cilia ablation in Pdx1-Cre;CLEG2;Kif3a^lox/lox mice resulted in a robust increase of Hh target genes at early postnatal stages (Fig. S2).

To determine whether the increase in Gli/Hh activity correlates with accumulation of the myc-GLI2ΔN protein, we assessed its level by immunohistochemical analysis of the myc-tag in transgenic mice. As previously reported (7), myc-GLI2ΔN protein only accumulated in a small number of cells in Pdx1-Cre;CLEG2 pancreata during postnatal stages (Fig. 1C). In contrast, Pdx1-Cre;CLEG2;Kif3a^lox/lox mice displayed a significant increase in the number of cells marked by wide expression of myc-GLI2ΔN protein (Fig. 1D). Furthermore, we determined that cilia ablation correlated with myc-GLI2ΔN accumulation, as well as increased Hh activation (as measured by β-gal expression in Pdx1-Cre;CLEG2;Kif3a^lox/lox;Ptch1^lacZ/+ mice; Fig. S3). Importantly, we found that, despite its modification, myc-GLI2ΔN is able to locate to the primary cilia (Fig. 1 E and F).

Our findings indicate that cilia modulate myc-GLI2ΔN accumulation in Pdx1-Cre;CLEG2 mice. In other contexts, cilia have been shown to control Hh signaling activity by specifically regulating the formation of processed Gli3 repressor (Gli3R). To determine whether this regulation occurs in the pancreas, we analyzed Gli3 in protein extracts from Pdx1-Cre;Kif3a^lox/lox and Pdx1-Cre;CLEG2;Kif3a^lox/lox postnatal pancreata. Surprisingly, we found that total Gli3 levels (full length and processed) appeared dramatically increased in both Pdx1-Cre;Kif3a^lox/lox and Pdx1-Cre;CLEG2;Kif3a^lox/lox pancreata compared with control littermates (Fig. S4). Thus, in contrast to recent studies in brain and skin (11, 15), these data indicate that cilia function to curb Gli3 levels in the pancreas.

Cilia Regulate Hh Signaling Downstream of Smoothened.

Next, to determine whether cilia regulate Hh activity in pancreatic epithelial cells at different levels of the signaling pathway, we used a transgenic mouse harboring a constitutively active version of Smoothened (SmoM2) that is expressed upon Cre-mediated excision of an upstream lox-stop-lox sequence (22). Activation of Hh signaling using this mouse model has been widely achieved in multiple contexts (23, 24). In agreement with recent reports (7), we found that Hh activity was not significantly increased in Pdx1-Cre;SmoM2;Ptch1^lacZ/+ when compared with control Ptch1^lacZ/+ pancreata (Fig. S5). To determine whether cilia prevented Hh activation at the level of Smo, we eliminated cilia in the context of SmoM2 expression. We found that Ptch1-lacZ activity remained unchanged in Pdx1-Cre;Kif3a^lox/lox;SmoM2;Ptch1^lacZ/+, findings that are in stark contrast to the increased Hh signaling activity observed in Pdx1-Cre;CLEG2;Kif3a^lox/lox;Ptch1^lacZ/+ mice (Fig. S5 and Fig. 1B). To confirm these observations, we conditionally eliminated cilia along with the Hh repressor Patched1, which blocks Hh signaling by inhibiting Smo transduction activity. We found that in Pdx1-Cre;Kif3a^lox/lox;Ptch1^lox/lacZ mice, Hh activity remained similar to that found in the other mice harboring Hh-activating mutations upstream of Gli2 (Fig. S5). Thus, our findings suggest that primary cilia regulate Hh signaling downstream of Ptch1 and Smo, possibly at the level of Gli, in the pancreas epithelium.

Full Hh Activation in the Pancreas Results in Loss of Pancreatic Tissue in Pdx1-Cre;CLEG2;Kif3a^lox/lox Mice.

The finding that cilia ablation in Pdx1-Cre;CLEG2;Kif3a^lox/lox mice results in overt activation of Hh signaling allowed us to evaluate whether cell-autonomous deregulation of Hh signaling in the pancreatic epithelium impacts pancreas formation. Triple-transgenic Pdx1-Cre;CLEG2;Kif3a^lox/lox mice were born at the expected Mendelian ratio and appeared grossly normal. However, by 2–3 weeks of age, Pdx1-Cre;CLEG2;Kif3a^lox/lox mice were smaller, failed to thrive, and showed postnatal lethality. A profound loss of pancreatic tissue, in both the exocrine and endocrine compartments, was observed in Pdx1-Cre;CLEG2;Kif3a^lox/lox animals (Fig. 2D). The severe acinar cell loss was accompanied by expansion of the duct-like epithelium and stromal compartment (Fig. 2D). In addition, the duct-like epithelium present in triple-transgenic Pdx1-Cre;CLEG2;Kif3a^lox/lox mice was composed of tall columnar cells distinct from the normally cuboidal duct morphology (Fig. 2L Inset). These cells are likely derived from the duct-like epithelium, as they were contiguous with tissues staining for ductal markers CK19 and mucin. However, the abnormal cells completely lost expression of these markers and accumulated high levels of myc-GLI2ΔN protein, suggesting an inverse correlation between Hh signaling levels and ductal marker expression (Fig. 2 H and L). Within or adjacent to ducts, we also found solid epithelial cell nests embedded in a matrix of fibrous stroma (Figs. 1D and 2 D and H). The epithelial cell nests were strongly positive for myc-GLI2ΔN and did not express markers for any of the three mature pancreatic cell lineages (amylase, acinar; synaptophysin, endocrine; mucin, duct; Figs. 1D and 2 H and L and Fig. S6), indicating some degree of dedifferentiation. Thus, Pdx1Cre;CLEG2;Kif3a^lox/lox mice develop a dramatic loss of mature pancreatic tissue that is accompanied by a loss of cellular differentiation and expansion of abnormal cells. In contrast, and as previously reported (7, 25), pancreas morphology was only marginally affected in Pdx1-Cre;CLEG2 mice (Fig. 2 B, F, and J), whereas mild duct dilation and reduction in acinar cells were noted in Pdx1-Cre;Kif3a^lox/lox mice by 2–3 weeks of age (Fig. 2 C, G, and K).

Fig. 2.

Full Hh activation results in perturbed pancreas morphology. (A–D) Hematoxylin/eosin-stained sections of 3-week-old mice reveal loss of pancreatic tissue and expansion of duct-like epithelium in Pdx1-Cre;CLEG2;Kif3a^lox/lox mice. Note the increased stromal compartment in Pdx1-Cre;CLEG2;Kif3a^lox/lox mice (asterisks in D). (E–H) Extensive acinar cell loss as shown by the decrease of amylase staining and increased dilation of duct-like structures marked by CK19 staining is observed in Pdx1-Cre;CLEG2;Kif3a^lox/lox mice. (G) Mild ductal dilation is observed in Pdx1-Cre;Kif3a^lox/lox mice. (I–L) Expression of the ductal marker mucin-1 is lost in cells expressing high levels of myc-GLI2 in Pdx1-Cre;CLEG2;Kif3a^lox/lox mice (arrowheads). (J) Myc-GLI2ΔN accumulation is observed in a subset of pancreatic cells in Pdx1-Cre;CLEG2 mice. Epithelial cell nests are outlined in D and H.

Expression of Progenitor Cell Markers in Hh-Active Pancreatic Epithelial Cells.

The abnormal epithelial nests budding from the ducts in Pdx1-Cre;CLEG2;Kif3a^lox/lox mice did not express markers normally present in mature pancreatic cell types, and thus we decided to characterize them in more detail. Expression of the cell-adhesion protein E-cadherin confirmed the epithelial nature of the high-myc-GLI2ΔN-expressing cells (Fig. 3B). Though markers typical of any of the three mature pancreatic cell types were missing, we did detect expression of FoxA2 and Sox9 (Fig. 3 B and D), proteins normally restricted to islets or ducts during postnatal stages (Fig. 3 A and C). However, these transcription factors are also expressed during pancreatic organogenesis in progenitors before the secondary transition (26–28), suggesting that the high-myc-GLI2ΔN-expressing cells might have acquired a progenitor-like state.

Fig. 3.

Activation of Hh signaling results in expression of progenitor markers in the ductal-like epithelium of Pdx1-Cre;CLEG2;Kif3a^lox/lox mice. (A and B) Expression of the embryonic pancreatic marker Sox9 in undifferentiated cells in 2- to 3-week-old Pdx1-Cre;CLEG2;Kif3a^lox/lox mice. E-cadherin expression confirms the epithelial nature of those cells. (C and D) FoxA2 is expressed in undifferentiated cells in Pdx1-Cre;CLEG2;Kif3a^lox/lox mice. (E and F) Myc-GLI2ΔN-expressing cells in Pdx1-Cre;CLEG2;Kif3a^lox/lox mice are highly proliferative as determined by staining with phospho-Histone H3. (G and H) Activation of Notch signaling in undifferentiated cells of Pdx1-Cre;CLEG2;Kif3a^lox/lox mice as determined by Hes1 expression. (I and J) Pdx-1 expression is excluded from undifferentiated cells expressing high levels of myc-GLI2ΔN in Pdx1-Cre;CLEG2;Kif3a^lox/lox pancreata. (K and L) Undifferentiated cells are first observed at P0 in Pdx1-Cre;CLEG2;Kif3a^lox/lox pancreata. (M and N) Abnormal mucin-1 expression in undifferentiated cells of P0 Pdx1-Cre;CLEG2;Kif3a^lox/lox pancreata. Mucin-1 expression is decreased and not restricted to the apical membrane in myc-GLI2ΔN-expressing cells of Pdx1-Cre;CLEG2;Kif3a^lox/lox mice (arrows). Mucin-1 is properly localized to the apical membrane in Pdx1-Cre; CLEG2 mice. Note the low number of myc-GLI2ΔN-expressing cells observed in Pdx1-Cre; CLEG2 mice. (O and P) Coexpression of exocrine (amylase) and ductal markers (Mucin-1) in myc-GLI2ΔN-expressing cells of P0 in Pdx1-Cre;CLEG2;Kif3a^lox/lox mice. The area marked by arrowheads is shown at higher magnification in inset in P.

Further evidence for this hypothesis comes from the observation that the abnormal cells are highly proliferative (Fig. 3F) and display active Notch signaling, as evidenced by immunofluorescence staining against Hes1 (Fig. 3H), a Notch-target gene that is expressed in embryonic progenitor cells (29). Hes1 was found only in centroacinar cells and a few ductal cells in adult pancreata of control mice (30) (Fig. 3G). Interestingly, expression of Pdx-1, an embryonic pancreatic transcription factor and β-cell marker (31, 32), was excluded from undifferentiated cells expressing high levels of myc-GLI2ΔN in Pdx1-Cre;CLEG2;Kif3a^lox/lox pancreata (Fig. 3J). Thus, these findings suggest that cell-autonomous activation of Hh signaling in pancreatic epithelium induces dedifferentiation and expansion of cells expressing a subset of progenitor markers.

Temporal analysis revealed that these undifferentiated cells were first observed at P0. At this time point, a few abnormal ductal structures could be found in Pdx1-Cre;CLEG2;Kif3a^lox/lox pancreata (Fig. 3L). These structures displayed abundant myc-GLI2ΔN and expressed low level of ductal mucin-1, which was not appropriately localized to the apical membrane, further suggesting that Hh activation interferes with the differentiation state of duct cells (Fig. 3N). Furthermore, we also found unusual cells that accumulated myc-GLI2ΔN and coexpressed amylase and CK19 located within dilating ductal structures (Fig. 3 O and P). The coexpression of both exocrine and ductal markers is suggestive of a cell caught in an intermediate stage of transdifferentiation that could contribute to the expansion of ductal structures observed in older Pdx1-Cre;CLEG2;Kif3a^lox/lox mice. Dysmorphic ducts displaying high Hh activity were detected as early as e15.5 in Pdx1-Cre;CLEG2;Kif3a^lox/lox;Ptch1^lacZ/+ mice (Fig. S1), thus suggesting impaired ductal formation during embryogenesis in the context of up-regulated Hh signaling.

Endocrine Defects in Pdx1-Cre;CLEG2;Kif3a^lox/lox Mice.

In addition to the exocrine and ductal defects, we observed profound defects in the endocrine compartment in Pdx1-Cre;CLEG2;Kif3a^lox/lox mice. A 75% reduction in endocrine area in Pdx1-Cre;CLEG2;Kif3a^lox/lox mice was found by postnatal day 5 (P5), in contrast to the 50% reduction observed in Pdx1-Cre;CLEG2 mice (Fig. 4K). Moreover, the architecture of α- and β-cells in the remaining islets appeared disorganized in Pdx1-Cre;CLEG2;Kif3a^lox/lox mice (Fig. 4D).

Fig. 4.

Increased Hh activity results in disruption of islet formation in Pdx1-Cre;CLEG2;Kif3a^lox/lox mice. (A–D) Islets are reduced in number and their architecture is affected in Pdx1-Cre;CLEG2;Kif3a^lox/lox mice at P5. Islet morphology is shown at higher magnification in insets. (E–H) Islet morphology appears unaffected in Pdx1-Cre;CLEG2;Kif3a^lox/lox mice at P0. (I–K) Quantification of islet area in E15.5 embryos and P0 and P5 mice. Note the dramatic reduction of islet area in Pdx1-Cre;CLEG2;Kif3a^lox/lox mice at P5. (L and M) Quantification of ngn3-positive cells and epithelial area in E15.5 embryos. Both epithelial area and neurogenin3 endocrine precursors are reduced in Pdx1-Cre;CLEG2 and Pdx1-Cre;CLEG2;Kif3a^lox/lox mice. (N and O) Quantification of endocrine cell proliferation based on phospho-Histone H3 expression at P0 and P5. (P and Q) Quantification of endocrine cell death based on cleaved caspase-3 expression at P0 and P5. Note the sustained 2- to 3-fold increased cell death rate in Pdx1-Cre;CLEG2;Kif3a^lox/lox mice. All values are relative to control littermates, whose averages were considered to be one. Sample numbers represent n ≥ 3. (See Table S1 for sample numbers.) Error bars represent SD. *P < 0.05 and **P < 0.005.

To evaluate the timing of endocrine cell loss, we analyzed islet area starting at e15.5, when endocrine cell neogenesis peaks, and at P0, when islet cells begin reorganizing to achieve final islet architecture. A substantial 50% reduction in the endocrine area in Pdx1-Cre;CLEG2;Kif3a^lox/lox mice was observed at e15.5 and P0, comparable to that seen in Pdx1-Cre;CLEG2 mice (Fig. 4 E–J). Therefore, these data indicate that endocrine defects in Pdx1-Cre;CLEG2;Kif3a^lox/lox mice deviate from Pdx1-Cre;CLEG2 mice after birth.

To further investigate the reduction in endocrine cells detected at e15.5 in Pdx1-Cre;CLEG2 and Pdx1-Cre;CLEG2;Kif3a^lox/lox mice, we performed immunofluorescence staining against neurogenin3 (ngn3), which marks all endocrine precursors, and found that their number is similarly reduced in both Pdx1-Cre;CLEG2 and Pdx1-Cre;CLEG2;Kif3a^lox/lox mice (Fig. 4L). In addition, quantification of epithelial area at e15.5 revealed a significant 50% reduction in Pdx1-Cre;CLEG2 and Pdx1-Cre;CLEG2;Kif3a^lox/lox compared with control mice (Fig. 4M). Thus, though the ratio of endocrine precursors to epithelial area is normal, the number of endocrine precursors is reduced as a consequence of impaired expansion of the pancreatic epithelium. Interestingly, CLEG2 transgene expression resulted in similar embryonic defects in the presence and absence of cilia, suggesting that the level of Hh activation achieved in Pdx1-Cre;CLEG2 pancreata is sufficient to impair expansion of the embryonic pancreas epithelium.

In contrast to embryonic expansion of endocrine cells via neogenesis, postnatal expansion of islet cells is predominantly achieved via proliferation. To address the question of whether the postnatal loss of endocrine tissue may be attributed to changes in islet cell proliferation and/or apoptosis rates, we quantified phospho-Histone 3 and cleaved caspase-3-positive cells, respectively, in postnatal pancreata. We found transient changes in proliferative capacity in transgenic islets, with a 2-fold increase in Pdx1-Cre;CLEG2 and Pdx1-Cre;Kif3a^lox/lox pancreata at P0 and a marginal increase in Pdx1-Cre;CLEG2;Kif3a^lox/lox islets at P5 (Fig. 4 N and O). In contrast, a significant 2- to 3-fold increase in islet cell death was observed in Pdx1-Cre;CLEG2;Kif3a^lox/lox islets between P0 and P5 (Fig. 4 P and Q), indicating that postnatal endocrine cells cannot sustain deregulated Hh signaling.

Discussion

To determine the role of epithelial Hh signaling on pancreas formation, we have overexpressed an activated version of GLI2, to activate Hh signaling in a cell-autonomous manner specifically in the pancreatic epithelium. Despite the forced expression of this Hh-activating allele, we did not observe efficient Hh activation. To determine whether primary cilia could regulate Hh signaling in the pancreatic epithelium, we eliminated these organelles in mice carrying distinct Hh gain-of-function mutations. Ablation of primary cilia resulted in strong Hh activation in mice harboring a GLI2-activating mutation but not in SmoM2 and Ptch1^lox/LacZ mutant mice, indicating that primary cilia regulate Hh activity downstream of Smo in the pancreas.

The intriguing fact that Hh activating mutations upstream of Gli2 (i.e., SmoM2) do not efficiently activate Hh signaling suggests that pancreatic epithelial cells are equipped with regulatory mechanisms that prevent Hh activation. Our study shows that primary cilia do not fulfill this role, because their ablation in the context of SmoM2 or loss of Ptch1 do not lead to pancreatic abnormalities observed in Pdx1-Cre;CLEG2;Kif3a^lox/lox mice. Our findings suggest that overexpression of GLI2ΔN in CLEG2 mice bypasses these additional modulators, but that GLI2ΔN activity is still subject to cilia-mediated regulation. Thus, elimination of cilia in Pdx1-Cre;CLEG2;Kif3a^lox/lox mice is necessary for full activation of the Hh pathway. The presence of additional regulators that curb signaling activity in the presence of Smo activation might also partly explain the resilient response of pancreatic epithelial cells to Hh ligands in vitro as previously reported (8, 10). These observations are in contrast to recent reports in medulloblastoma and BCC, where expression of SmoM2 is sufficient to induce tumors in a cilia-dependent manner (11, 15). The pancreas further differs from those tissues in that expression of CLEG2 in Pdx1-Cre;CLEG2 mice only lead to a modest increase in Hh signaling activity, which was dramatically boosted by cilia ablation. In cerebellum and skin, CLEG2 expression with or without cilia ablation resulted in similar levels of Hh activation, albeit cilia ablation accelerated tumor formation. Furthermore, whereas cilia regulate generation of Gli3R in cerebellum and skin (11, 15), they modulate total Gli3 protein levels in the pancreas. Together, these differences suggest that regulatory mechanisms other than cilia modulate Hh activity in a tissue-specific manner.

The increased Hh activity achieved in Pdx1-Cre;CLEG2;Kif3a^lox/lox mice revealed an unknown role of epithelial Hh signaling during pancreas development and differentiation. One of the most intriguing malformations observed in Pdx1-Cre;CLEG2;Kif3a^lox/lox mice is the formation of cells with abnormal morphology displaying high myc-GLI2ΔN levels. These cells expressed pancreatic progenitor markers, such as Hes1, FoxA2, and Sox9, suggesting that activation of Hh impairs the ability of pancreatic cells to maintain a differentiated state. Interestingly, it has been shown that inactivation of epithelial Hh signaling blocks pancreas regeneration upon injury due to the inability of metaplastic cells to redifferentiate back to exocrine cells, further supporting a role for Hh signaling in modulating pancreatic cell plasticity (9).

Our previous studies have shown that despite the low level of myc-GLI2ΔN accumulation in Pdx1-Cre;CLEG2 mice, prolonged expression of myc-GLI2ΔN results in formation of undifferentiated, pancreatic tumors. These undifferentiated tumors display loss of epithelial markers such as E-cadherin, suggestive of an epithelial-to-mesenchymal transition. Although the undifferentiated cells observed in Pdx1-Cre;CLEG2;Kif3a^lox/lox mice expressed E-cadherin, we cannot rule out the possibility that these cells represent an early transitional state toward tumor cells found in older Pdx1-Cre;CLEG2 mice. Though the undifferentiated tumors in these mice are distinct from pancreatic adenocarcinoma, it is intriguing that primary cilia are also absent both from adenocarcinoma precursor lesions and tumor cells (33), another difference to medulloblastoma and BCC, which are ciliated tumors (11, 15). Unfortunately, the compromised health of Pdx1-Cre;CLEG2;Kif3a^lox/lox mice prevents analysis in older animals. Focal/temporal inactivation of cilia and GLI2ΔN overexpression will be required to clarify whether undifferentiated cells in Pdx1-Cre;CLEG2;Kif3a^lox/lox mice represent early stages of pancreas neoplasia.

Increased Hh signaling also results in dramatic defects in endocrine cell formation, consisting of a significant reduction in islet area, largely during postnatal stages. In contrast to what we observed in the exocrine compartment, transdifferentiation does not appear to play a role in the postnatal loss of endocrine area. Rather, our findings suggest that increased apoptosis is most likely responsible for this process, indicating that deregulation of Hh signaling compromises endocrine cell health. Previous reports from insulinoma cells have shown a positive role for this pathway in regulating β-cell function (34, 35). Unfortunately, the early postnatal lethality of the Pdx1-Cre;CLEG2;Kif3a^lox/lox mice has prevented us from analyzing the effects of increased Hh signaling on mature endocrine cell function, and these questions await manipulation of the pathway in adult islet cells.

In summary, we have shown that primary cilia regulate Hh activity downstream of Smo and possibly at the level of the Gli transcription factors in the pancreatic epithelium. Our findings also show that increased Hh signaling in pancreatic epithelial cells leads to loss of differentiated state and activation of embryonic foregut and pancreas progenitor markers. These findings suggest that modulation of epithelial Hh signaling controls differentiated phenotypes and therefore might have important implications for reprogramming of mature exocrine cells, a process involved in pancreatic tumor formation and a potential source of insulin-producing cells (36).

Methods

Mice used in this study were maintained in the barrier facility according to protocols approved by the Committee on Animal Research at the University of California, San Francisco. Pdx1-Cre, CLEG2, Ptch1^lacZ/+, Ptch1^lox/lox, SmoM2, and Kif3a^lox/lox mice have been described previously (7, 16, 22, 32, 37, 38).

Reagents and procedures are described in detail in SI Methods.