Mesenchymal Bone Morphogenetic Protein Signaling Is Required for Normal Pancreas Development

In ovo electroporation and in situ hybridization was performed as in previous studies (6), and at least three embryos were analyzed for each condition. RNA probes were transcribed in vitro from cDNA clones provided by BBSRC (Bmp7: ChEST290N8; Bapx1: ChEST177B22; Pitx2: ChEST76C15), except for the Neurog3 cDNA provided by Anne Grapin-Botton and the Bmp4 cDNA provided by Cliff Tabin. Probes for Nkx6.1 and Hes1 were used as in earlier studies (6,17) The Xenopus Noggin cDNA was provided by Richard Harland, and a COOH-terminal cMyc epitope tag inserted by PCR. The mouse dnALK3 was a COOH-terminal truncated receptor (encoding aa 1–237) lacking the kinase domain. A similar construct has previously been used to block BMP signaling in duck and chicken embryos (18). All expression constructs were cloned into the pCIG5 vector (6).

Whole-mount immunoflourescent staining was performed as in (19). Briefly, embryos were cleared in Dent's bleach and incubated overnight with antibodies diluted in blocking buffer. Confocal images were obtained from embryos cleared in benzyl alcohol/benzyl benzoate (BABB) (19) on a Zeiss LSM510 AxioImager. For antibodies, these data can be found in online appendix supplementary Table 1, available at http://diabetes.diabetesjournals.org/cgi/content/full/db09-1010/DC1. Perfusion of embryos with lens culinaris agglutinin was performed with a protocol modified from Jilani et al. (20). Using this protocol, the embryos were dissected in PBS, and 10 μl was injected into the heart with a pulled glass pipette. The embryos were left for 5 min, then fixed and processed for immunohistochemistry. Labeled areas were quantified using ImageJ software by analyzing every second optical section corresponding to an optical depth of 20 μm through the whole pancreas (between 7 and 12 sections from each embryos). Statistical significance was calculated using Student t test. Results were considered statistically significant if P < 0.05.

Lentiviral constructs were generated by gateway recombination cloning, using entry vectors with the same constructs used in the chicken experiments. Organs were cultured and transduced as by Attali et al. (21), except that the epithelium was not separated from the mesenchyme before lentiviral infection. All lentiviral vectors were used at viral concentrations corresponding to 300 ng of p24 capsid protein per explant. The e12.5-pancreas explants were fixed in formalin and embedded in paraffin 7 days after transduction. The explants were sectioned (4 μm) and stained for insulin, amylase, and DNA (DAPI) as described by Haumaitre et al. (22). Quantification was performed every fifth section spanning the entire explant using ImageJ software, and total area, insulin area, and amylase area surface were measured. In the e10.5 gut tube explants 10-μm cryosections of the entire explants were stained for insulin, amylase, and DNA (DAPI). Quantification was performed every second section by counting the nuclei of insulin and amylase-positive cells. We estimated total pancreatic cell numbers by adding insulin and amylase numbers. Statistical significance was calculated using Student t test. Results were considered statistically significant if P < 0.05.

BMP signaling is transduced by Smad1, −5, and −8 which, upon signaling, become phosphorylated and accumulates in the nucleus where they act as transcription factors (23). We used a phosphoSmad1,5,8-specific antibody (24) to characterize endogenous BMP signaling in the embryonic chicken pancreas (Fig. 1) at Hamburger and Hamilton (HH) stages 12, 18, and 21 (25). This corresponds to time points when the pancreas has just been specified in the endoderm, dorsal and ventral buds have formed, and α-cells, β-cells, and amylase-positive cells can be detected (6,17).

At HH st. 12, embryos were whole-mount stained and 200-μm transverse vibratome sections were analyzed (Fig. 1,A–D). Nuclear pSmad1,5,8 could be detected in the dorsal neural tube (Fig. 1,A–D) and in the ventral part of the anterior foregut endoderm (arrows, Fig. 1,A,) as previously reported (24). At the level of the 3rd to 8th somite pairs, where the prospective pancreatic endoderm is located (3,26), pSmad1,5,8 was detected in the lateral plate mesoderm and in the endoderm lateral to the Nkx6.1 expression domain (Fig. 1,B, arrow, and arrowheads in Fig. 1,C and D, supplementary Fig. 1). We could also observe pSmad1,5,8 in the endothelium of the paired dorsal aortas (Fig. 1 C and D).

At HH st. 18, pSmad1,5,8 immunoreactivity was detected in the pancreas mesenchyme and at lower levels in the duodenal epithelium (Fig. 1,E and F). The dorsal pancreas epithelium was pSmad1,5,8-negative. One day later, we still observed a mesenchymal pSmad1,5,8 signal (Fig. 1,G–J), but no signal was detected in the dorsal or ventral pancreatic epithelium. The liver epithelium had a strong signal with a sharp border of immunoreactivity in the bile duct (Fig. 1,G arrows). No pSmad1,5,8 reactivity was detected in Nkx6.1-positive cells, which at this stage comprise both undifferentiated epithelial cells and β-cells (17) (Fig. 1 H–J). The absence of labeling in β-cells was confirmed by colabeling with an insulin antibody (supplementary Fig. 1G and H).

To evaluate the status of Smad1,5,8 phosphorylation in the early mouse pancreas, we performed whole-mount staining at e9.5 and e11.5 for pSmad1,5,8 combined with Pdx1 (Fig. 1,K–N). At e9.5, immunoreactivity was detected in the dorsal aortas, intersomitic vessels (Fig. 1,K and L), and the dorsal neural tube (not shown). No pSmad1,5,8 was detected in the Pdx1-positive endoderm. At e11.5, we observed pSmad1,5,8 reactivity in the pancreas mesenchyme, but no reactivity in the pancreas epithelium (Fig. 1 M and N). We conclude that endogenous signaling leading to detectable levels of pSmad1,5,8 in the early chicken and mouse pancreas is entirely restricted to the mesenchyme.

Bmp4, -5, and -7 mRNA have been reported to be expressed in the embryonic mouse pancreas by RT-PCR (15), and Bmp4 and Bmp7 have been localized to the pancreas epithelium (16), but little is known about BMP expression during chicken pancreas development. We performed in situ hybridizations at e3 (HH. st. 18) and e4 (HH. st. 22) to characterize the expression patterns of Bmp2, Bmp4, Bmp5, and Bmp7 (Fig. 2 and data not shown). We found that Bmp4 and Bmp7 are expressed in the chicken pancreas mesenchyme and excluded from the epithelium at the stages examined (Fig. 2 B and D and data not shown). We could not detect Bmp2 or Bmp5 in the pancreas at any stage (data not shown). From these results we concluded that BMP signaling is active in the pancreas mesenchyme, possibly mediated by BMPs expressed within the mesenchyme.

To investigate the requirements for BMP signaling in pancreas development, we applied two complementary approaches. We ectopically expressed the secreted BMP antagonist Noggin or a dominant negative BMP receptor type 1 (dnALK3) in the pancreas epithelium using in ovo electroporation of the endoderm (6,27). We expected that the secreted antagonist Noggin should block BMP signaling in the epithelium as well as in the mesenchyme, whereas dnALK3 should be autonomous to the epithelial cells expressing this construct. Electroporation was performed at HH st. 10–12, and the introduced genes were thus expressed soon after the specification of the pancreas endoderm (17).

First we electroporated with a cMYC tagged Noggin in the pCIG5 vector which has an IRES nuclear enhanced green fluorescent protein (EGFP) sequence allowing direct monitoring of electroporated cells (6). As expected, EGFP was restricted to the epithelium after 32 h. However, cMYC immunoreactivity demonstrated that Noggin was efficiently secreted and spread throughout the surrounding mesenchyme (supplementary Fig. 2). We next assayed for pSmad1,5,8 in the pancreas 24 and 48 h after electroporation (Fig. 3). Electroporation with a control plasmid did not change the pattern of endogenous pSmad1,5,8 (Fig. 3,B and F), but ectopic expression of Noggin resulted in loss of mesenchymal pSmad1,5,8 immunoreactivity around the electroporated area (Fig. 3,D and H). We could still detect pSmad1,5,8 in the mesenchyme at a distance from the electroporated area (Fig. 3,H, arrowheads) and in the dorsal neural tube (not shown). We noted that the Nkx6.1-positive pancreas epithelium appeared less developed than the corresponding controls after 48 h (compare Fig. 3 G–E).

To assess pancreas morphogenesis, we performed whole-mount immunofluorescent staining of electroporated embryos at 24, 48, and 72 h. Three-dimensional projections of confocal z-stacks were generated showing green fluorescent protein (GFP), Nkx6.1, and glucagon/insulin (Fig. 4). At 24 h after electroporation with control plasmid or Noggin, the dorsal and ventral pancreata were both formed normally, and endocrine cells were found exclusively in the dorsal pancreas (Fig. 4,A–D). At 48 h, the pancreata electroporated with empty vector developed normally with primary branching, two distinct ventral pancreas buds (5), and endocrine cells confined to the dorsal pancreas (Fig. 4,E and F). In contrast, development of the Noggin electroporated pancreata was severely disturbed (Fig. 4,G, H, M). There was no primary branching of the epithelium. Instead the dorsal pancreas appeared as a tubular structure extending anteriorly along the stomach (Fig. 4,G and H). Quantifying the amount of epithelial and endocrine areas showed that the total endocrine area was unchanged, but the epithelium was reduced by ∼50% (supplementary Table 2). Precocious endocrine cells could be observed in the ventral pancreas which was hypoplastic (Fig. 4,G and H, arrow; supplementary Table 2). In some embryos, the pancreas was almost entirely converted into a large endocrine cluster (Fig. 4 M).

At 72 h after electroporation with control vector, endocrine cells were still confined to the dorsal pancreas and branching of the epithelium was prominent (Fig. 4,I and J). Conversely, the pancreata of Noggin electroporated embryos were hypomorphic, and large endocrine clusters detached from the pancreas could be observed in the stomach mesenchyme (Fig. 4,K and L, arrowheads). There was no epithelial branching and endocrine cells were found in the ventral pancreas (Fig. 4,K and L, arrow). Cell-autonomous inhibition of BMP signaling in the endoderm via electroporation with dnALK3 had no effect on pancreas development (Fig. 4 N), suggesting that the effect elicited by Noggin is through inhibition of BMP signaling in the mesenchyme. In neural tube electroporation experiments, dnALK3, as well as Noggin, was able to block Smad1,5,8 phosphorylation in the dorsal neural tube. confirming that the dnALK3 construct worked as expected (supplementary Fig. 2C–H).

The pancreas of Noggin electroporated embryos was further analyzed on sections 48 h after electroporation (Fig. 5). The anterior extension of the pancreas contained Nkx6.1, glucagon, and insulin-positive cells (Fig. 5,A and B). Control pancreata contained scattered exocrine cells marked by amylase expression (Fig. 5,C), but no exocrine cells were found in Noggin electroporated embryos (Fig. 5 D).

The results presented so far indicate that BMP signaling is required in the mesenchyme and that the effect on the epithelium is indirect. This was difficult to demonstrate directly since electroporation in the chicken embryo does not permit cell-autonomous inhibition of BMP signaling in the mesenchyme. Therefore, we took advantage of established protocols for in vitro development of mouse pancreas explants and lentiviral gene transduction (28). We explanted e12.5 pancreata transduced with lentivirial vectors encoding either GFP, Noggin, or dnALK3. When whole pancreas explants were exposed to lentiviral vectors, exclusive infection of the mesenchyme was achieved (Fig. 6 H and I), allowing inhibition of mesenchymal BMP signaling in a cell-autonomous manner with dnALK3.

The explants were cultured for 3 or 7 days (Fig. 6,A–F). Stereomicrographs of the explants cultured for 7 days showed that dnALK3 and Noggin-transduced explants had a similar morphology and were of similar size as controls (Fig. 6,A–F). Costaining for E-cadherin and GFP on GFP-transduced explants confirmed the mesenchymal-specific expression after 3 and 7 days in culture (Fig. 6,H and I). We quantified the number of exocrine cells and β-cells after 7 days in culture (Fig. 6,J–L, supplementary Table 3) and found a 2.8-fold increase in the relative area of β-cells in dnALK3-transduced explants compared with controls. This expansion was even more pronounced in Noggin-transduced explants which showed a 3.6-fold increase in relative β-cell area. In contrast, the relative exocrine area was diminished in dnALK3 and Noggin-transduced explants compared with controls (not statistically significant). There was no obvious branching defect in these studies, which could be a result of different timing of BMP inhibition compared with the studies performed in chicken. We therefore repeated the experiments with e10.5 explants prior to branching morphogenesis. E10.5 gut tubes were dissected; transduced with GFP, Noggin, or dnALK3; and cultured for 4 and 8 days. After 4 days, the total explant volume was reduced by 5.5- and 1.8-fold when transduced with Noggin and dnALK3, respectively (compared with control GFP transduction; supplementary Fig. 3E). Noggin transduction resulted in a subtle change in the overall branching pattern, visualized by whole-mount Mucin-1 staining, as main ducts were not well defined and appeared diffuse compared with controls (supplementary Fig. 3A–D). After 8 days in culture, mesenchymal Noggin expression increased the total number of insulin-positive cells by 3.8-fold, whereas the total number of amylase-positive cells were reduced by 1.6-fold (Fig. 6 M–O, supplementary Table 3). Similarly, mesenchymal expression of dnALK3 increased the number of insulin-positive cells by 2.5-fold, demonstrating that cell autonomous BMP signal reception in the mesenchyme is required for a normal rate of endocrine cell differentiation to occur. We did observe a small but statistically significant effect on total exocrine cell numbers in Noggin-transduced explants and in relative numbers of exocrine cells in both Noggin and dnALK3 transduced (supplementary Table 3). The effect of mesenchymal expression of dnALK3 on branching morphogenesis was ambiguous; some explants exhibited a phenotype similar to the Noggin-transduced explants, whereas some had a normal appearance (not shown). We speculate that this may be attributed to variable transduction efficiency.

Since our results suggested that Noggin inhibited BMP signaling in the mesenchyme, we next investigated mesenchymal gene expression patterns. Pitx2 is normally expressed in the left side of the mesenchyme of vertebrate embryos and is involved in asymmetrical organ development (29,30). Bapx1 was recently demonstrated to be involved in the left-sided lateral growth of the e10.5 mouse dorsal pancreas because of a requirement for Bapx1 in the formation of a specialized mesothelial structure, the splancnic mesodermal plate (SMP) (31). It is normally expressed in the left side of the mesenchyme, and it is required for the separation of the splenic and pancreatic mesenchyme and maintenance of the pancreas character in the distal part of the dorsal pancreas (31,32). To evaluate the expression of Pitx2 and Bapx1 in relation to the chicken pancreas, we first performed mRNA in situ hybridizations on serial sections for Nkx6.1 and Shh, together with Pitx2 and Bapx1 (supplementary Fig. 4). We found that, as in the mouse, Bapx1 was initially expressed bilaterally in the pancreas region (supplementary Fig. 4P), but from e3 strong expression was restricted to the Pitx2-positive mesenchyme on the left side of the pancreas (supplementary Fig. 4I, J, L, and M). Bapx1 appeared to be symmetrically expressed at the level of the stomach (supplementary Fig. 4H) as in the mouse (31).

In control embryos, we found that Bapx1 was expressed left sided in the pancreas mesenchyme together with Pitx2 after 48 h (Fig. 7,C and E). In contrast, Noggin-electroporated embryos displayed a duplication of Bapx1 expression, whereas Pitx2 remained left sided, suggesting that Bapx1 and Pitx2 expression is uncoupled at least under these conditions (Fig. 7 D and F). These results are in agreement with a role for BMP signaling in development of the pancreas mesenchyme. Since Bapx1 is involved in the formation of the SPM in the mouse embryo, we hypothesized that this structure could be affected in the BMP-inhibited embryos. We therefore stained for Isl1 and laminin or Nkx6.1/Nkx2.2 with DAPI nuclear counter stain to visualize this structure (supplementary Fig. 5). In control embryos, the SPM was asymmetric with a condensation of the mesothelium on the left side at HH st. 19, 32 h after electroporation (supplementary Fig. 5A and B). Noggin electroporated embryos displayed a duplication of the SPM (supplementary Fig. 5C and D), suggesting that downregulation of right-sided Bapx1 expression is required for establishment of the asymmetric SPM.

The observation of precocious endocrine differentiation in Noggin-electroporated embryos prompted us to analyze the expression of Hes1. We found that the unbranched tubular dorsal pancreas was Hes1-negative 48 h after electroporation, suggesting that the excessive endocrine differentiation is coincident with downregulated Hes1 expression (Fig. 7 E and K). We therefore looked after 32 h of development and earlier (supplementary Fig. 6 and data not shown). Most of the Nkx6.1-positive dorsal pancreas epithelium had differentiated after 32 h and was Hes1-negative, but the remaining Nkx6.1-positive cells expressed normal levels of Hes1. At earlier time points when the pancreas was more normal, we could not detect a general downregulation of Hes1 in the epithelium (not shown). We can therefore not distinguish between loss of Hes1 followed by differentiation or vice versa, likely because our ISH protocol does not allow quantitative measurements of mRNA expression levels.

The bilateral expression of Bapx1 in the mesenchyme of Noggin electroporated embryos was notable. In chicken and mice, the right omphalomesenteric vein is normally found at this position, and the developing pancreas grows around this vessel (4,5). Gene expression changes in this structure might indicate that the normal intimately joined development of the pancreas and the vein was disturbed. To visualize the vasculature and the pancreas, we injected embryos with fluorescently-labeled lens culinaris agglutin, a lectin with high affinity to chicken endothelial cells (20), and subsequently we stained for Nkx6.1 by whole-mount immunohistochemistry (Fig. 8).

At 3 days of development (HH st. 19), the paired omphalomesenteric veins run on each side of the gut tube and join anteriorly at the level of the liver in the ductus venosus (33) (supplementary Fig. 7C). The bilateral, ventral Nkx6.1 domains are in close contact to these veins and the dorsal pancreas has grown asymmetrically over the right omphalomesenteric vein (supplementary Fig. 7C). At this stage. the dorsal aorta is well separated from the pancreas by the pancreatic mesenchyme. During the next day, a series of anastomoses of the veins occur and most of the left omphalomesenteric vein regresses, leaving behind the right omphalomesenteric vein which curls around the gut at the level of the pancreas (5,33) (supplementary Fig. 7D and F). The dorsal pancreas and one of the ventral pancreata lie in close contact to the right omphalomesenteric vein (supplementary Fig. 7F) (5).

We analyzed the vascular development in control and Noggin electroporated embryos 48 h after electroporation. The electroporated control embryos had normal development of the omphalomesenteric vein (Fig. 8,A–C). Transverse vibratome sectioning revealed the asymmetric pancreas mesenchyme (Fig. 8,C), and, as expected, the dorsal aorta was well separated from the dorsal pancreas (Fig. 8,A). In Noggin electroporated embryos, the vasculature developed abnormally (Fig. 8 E–G). The omphalomesenteric vein was found on the ventral side of the pancreas and gut (n = 6) and, in some embryos, the dorsal aortas failed to fuse and move dorsally (2 of 6 embryos). To assess the development of the microvasculature, we made Flk1 ISH on sectioned embryos. This revealed a normal microvasculature around the dorsal pancreas 32 h after electroporation when the omphalomesenteric vein is already displaced to the ventral side in BMP-inhibited embryos (supplementary Fig. 8), suggesting that there is no general inhibition of angiogenesis under these conditions.

Development of the pancreas depends on interactions between the epithelium and the mesenchyme as recognized since the studies by Golosow and Grobstein (12), but few endogenous signaling factors have been identified.

Here, we used chicken in ovo endoderm electroporation timed to express the pancreatic transgenes after specification of the pancreas. We reasoned that this would allow us to study BMP signaling during pancreas organogenesis. Detection of pSmad1,5,8 shows that early pancreatic mesenchyme receives a BMP signal, possibly through ALK3, which is expressed in the chicken pancreas mesenchyme (34). Since we find BMP expression in the mesenchyme, it is possible that an autocrine signaling mechanism is occurring. Epithelial expression of the secreted BMP antagonist Noggin resulted in the loss of pSmad1,5,8 immunoreactivity in the adjacent mesenchyme and affected epithelial and mesenchymal development. The epithelium displayed reduced branching, anterior extension of the dorsal pancreas into the stomach mesenchyme, and precocious endocrine differentiation.

Inhibiting BMP signaling compromised angiogenic remodeling that positions the right omphalomesenteric vein adjacent to the pancreas. Instead, we observed a persistence of mesenchymal Bapx1 expression, a bilateral SMP, and failure of the dorsal aortas to fuse in the most severely affected embryos. We cannot distinguish a direct effect on the endothelial cells from an indirect effect acting through the mesenchyme. However, the microvasculature seems normal in BMP-inhibited embryos, suggesting that some aspects of angiogenesis occur normally and the defect appears limited to major vessels. Notably, the mouse explants corroborate the findings in chick, arguing against involvement of the major vessels. Additionally, mesenchymal Isl1 expression, dependent on signals from blood vessels (35), is maintained.

The results presented herein appear to conflict with a recent report by Goulley et al. (16). In that study, BMP signaling was perturbed by the expression of Noggin and BMP4 in the pancreas epithelium using the Pdx1 promoter, with no profound effect on pancreatic development. We speculate that this could be due to different levels of expression since establishment of high-expressing transgenic mouse lines used in their studies would be prevented by a severe vascular or pancreatic phenotype caused by embryonic or perinatal lethality, and thus a selection for low expressers. A similar study expressing Bmp6 under the control of the Pdx1 promoter resulted in complete pancreas agenesis at a late gestational age, and these mice displayed a hypomorphic spleen, suggesting that the mesenchyme was also affected (15).

The present study suggests that there is no epithelial BMP signaling in the pancreas at the investigated stages, and that the effect on epithelial development is secondary to inhibition of BMP signaling in the mesenchyme. This is supported by the observation that cell autonomous BMP-signaling inhibition in the mouse pancreas mesenchyme leads to an increase in endocrine differentiation concomitant with a decrease in exocrine development. It is possible that the effect of BMP inhibition on the mesenchyme can be subdivided into an inhibition of an autocrine mesenchymal signal and an associated effect on vascular remodelling. BMP signaling in the mesenchyme may be required for the mesenchyme to express a secondary factor acting on the epithelium. Fgf10 is one candidate because mesenchymal Fgf10 is required for epithelial branching and progenitor cell proliferation (9,–11). However, there is little Fgf10 expression left at e12.5, the time of explanting in some of the organ culture experiments, suggesting that other factors could be involved.

This work was made possible by grants from the Juvenile Diabetes Research Foundation, the European Union's 6th Framework Program, and the Beta Cell Biology Consortium (grant U19 DK072495).

No potential conflicts of interest relevant to this article were reported.

J.A.-R. designed and carried out all cloning and chicken embryo manipulations and analyses. P.R., C.P.-G., and R.S. generated lentiviral constructs and performed the lentiviral transductions of pancreatic explants and carried out data analysis. J.A.-R. and P.S. conceived the study, participated in its design and coordination, performed data analysis, and drafted the manuscript.

The authors thank Anne Grapin-Botton, Swiss Institute for Experimental Cancer Research, Ecole/Polytechnique Fédérale de Lausanne, and Jacob Hecksher-Sørensen, Hagedorn Research Institute, for helpful discussions, and are grateful to Cliff Tabin, Harvard Medical School, and Richard Harland, University of California, Berkeley, for probes and Noggin cDNA, respectively, and for expert technical help from Malene Jørgensen.