All-Trans Retinoic Acid Induces Differentiation of Ducts and Endocrine Cells by Mesenchymal/Epithelial Interactions in Embryonic Pancreas

All animal experiments were performed in accordance with the guideline for animal experiments of Kyoto University. Male and female ICR mice were mated overnight. The presence of a vaginal plug the next morning was indicative of pregnancy, and noon of that day was designated as day 0.5. Dorsal embryonic pancreata were harvested on e11.5, e12.5, e14.5, e16.5, and e18.5 using microdissection techniques as described previously (19). Two types of tissue were procured from e11.5 rudiments for in vitro culture, epithelia only (PE) and epithelia with mesenchyme (PEM).

E11.5 rudiments (PE or PEM) were grown under various culture conditions in 0.4-μm Millipore filter inserts containing 150 μl collagen I gel (Vitrogen; Cohesion, Palo Alto, CA) or basement-rich gel (Matrigel; Collaborative Research, Bedford, MA) and placed in a standard 24-well plate under sterile conditions. Each well was filled with 500 μl Dulbecco’s modified Eagle’s medium F-12 (Gibco BRL, Grand Island, NY) containing 10% fetal bovine serum and 1% antibiotic/antimycotic solution (10,000 units/ml penicillin G, 10,000 μg/ml streptomycin sulfate, and 20 μg/ml amphotericin B [Gibco BRL]). Three different final concentrations (1, 10, and 100 μmol/l) of atRA (Sigma, St. Louis, MO) were supplemented later in the media and replaced every 48 h. Control tissues were cultured in collagen I or Matrigel with media alone or media with DMSO (the diluent for atRA). In addition, in a different series, the same three concentrations of atRA were added only on the 6th day of culture and harvested after 24 h. All of the tissues were cultured in 5% CO₂ at 37°C for 4–7 days.

Harvested tissues were fixed in 4% paraformaldehyde for 4 h then cryoprotected overnight in 30% sucrose, embedded in Tissue Tek OCT compound, frozen in liquid nitrogen, and cut into 6-μm sections using Microm HM 505E cryostat. Most of the sections were pretreated with peroxidase blocking reagent (Dako, Carpinteria, CA) to minimize nonspecific binding to endogenous peroxidase enzymes. Primary antibodies to the following antigens were used at the indicated dilution: insulin (guinea pig anti-swine; Dako) 1:700, glucagon (rabbit anti-human; Dako) 1:600, cytokeratin MNF 116 (mouse anti-human; Dako) ready to use, carbonic anhydrase II (anti-human; The Binding Site, Birmingham, U.K.) 1:500, α-amylase (rabbit anti-human; Sigma) 1:200; PDX-1 (gift from Prof. C.V.E. Wright, Vanderbilt University Medical School, Nashville, TN) 1:1,200, RALDH2 (gift from Dr. Peter McCaffery, University of Massachusetts Medical Center, Worcester, MA) 1:2000, Bax (P-19; Santa Cruz Biotech, Santa Cruz, CA) 1:100, Bcl-2 (N-19; Santa Cruz Biotech) 1:100, Bcl-x_L (H-62; Santa Cruz Biotech) 1:100, and anti-single stranded DNA (rabbit anti-human; Dako, Kyoto, Japan) 1:100. Primary antibodies were incubated at room temperature for 30 min for insulin/glucagon/cytokeratin/Bax/Bcl-2/Bcl-x_L/ssDNA, 1 h for CAII/PDX-1, and overnight at 4°C for α-amylase and RALDH2. For secondary antibodies, ready-to-use Envision peroxidase system (Dako) for insulin/glucagons/PDX-1/Bax/Bcl-2/Bcl-x_L/ssDNA, biotin-labeled donkey anti-sheep (Chemicon International, Temecula, CA) for CA II, ready-to-use Labeled Streptavidin Biotin (LSAB2) kit (Dako) for cytokeratin, and FITC/Cy3 conjugated goat anti-rabbit (Jackson ImmunoResearch Lab, West Grove, PA) for α-amylase/RALDH2/PDX-1 were used. Immunoperoxidase was detected by DAB kit (Dako) or AEC (Sigma), alkaline phosphatase by New Fuchsin (Dako) substrate chromogen system, and fluorescently labeled samples were imaged by using a fluorescent microscope. For PDX-1:insulin double staining, sections were first stained with PDX-1 using Envision peroxidase system visualized by DAB (brown precipitate), then briefly treated with 0.2 mol/l HCl at room temperature, and insulin staining was performed using LSAB2 alkaline phosphatase system and visualized by New Fuchsin substrate chromogen (red precipitate).

For determining the number of cells undergoing proliferation in cultured rudiments, a final concentration of 100 μmol/l bromo-deoxyuridine (BrdU; Sigma) was added 12 h before fixation in 10% formaldehyde. The proliferating cells were demonstrated by anti-BrdU (mouse monoclonal; Dako) staining. For BrdU:insulin/glucagon double staining, slides were treated with 1 mol/l HCl at 37°C for 30 min, primary antibody mixture was incubated for 1 h at room temperature, then the mixture of fluorescent secondary antibodies (goat anti-rabbit FITC/goat anti-mouseCy3) was applied and visualized under a fluorescent WIB filter Olympus microscope. Cy3 was seen as yellowish red, and FITC was seen as light green.

The protocol used by Miralles et al. (7) was pursued to quantify PDX-1 and endocrine cells. In brief, serial 6-μm sections were cut and collected in glass slides. One of two consecutive sections was analyzed (to avoid counting the same cell twice) by immunohistochemistry for the given antigen. Only immunoreactive cells with visible nucleus were counted. The sum of β- and α-cells was regarded as total endocrine cells. For ducts/acini/BrdU-positive cells, three positive samples per rudiment were randomly selected and analyzed by point counting using Densitograph software version for Macintosh. The results were obtained by quantifying at least four pancreatic rudiments in each category. These were compared and analyzed by ANOVA, and P < 0.05 was considered statistically significant.

Tissues were fixed in modified Karnovsky’s fixative (2% paraformaldehyde, 1.5% glutaraldehyde in 0.1 mol/l cacodylate buffer [pH 7.4]) for 1–4 h, postfixed in 1% osmium tetroxide for 1 h, dehydrated in graded ethanol, and embedded in Epoxy. Thick sections were cut 0.5–1 μm and stained in toluidine blue. Thin sections were overlaid in copper grid, stained by uranyl acetate and lead citrate, and viewed with a Hitachi electron microscope.

RNA was extracted from pooled e11.5 epithelia and mesenchyme separately, e12.5 PEM, e14.5 PEM, e16.5 PEM, and e18.5 PEM using the SV Total RNA isolation kit (Promega, Madison, WI), and cDNA was prepared from 1 μg RNA by using RT and random hexamer (Advantage RT-for-PCR kit; Clontech, Palo Alto, CA) according to the manufacturer’s instructions. PCR was carried out with a mixture consisting of 100 ng cDNA, 20 pmol each of forward and reverse primers, 2 mmol/l dNTP mix, 10× PCR buffer (100 mmol/l Tris-HCl [pH 8.8], 500 μmol/l KCl, 0.8% NP40) with 15 mmol/l MgCl₂, and recombinant Taq DNA polymerase (MBI fermentas, Vilnius, Lithuania). β-Actin was used as internal control, cDNA prepared without RT was used as negative control, and adult mouse liver for RARs (RARα, RARβ, and RARγ) and chicken ovalbumin upstream promotor transcription factor (COUP-TF II), adult mouse testis for RALDH2, and e13.5 liver for cellular retinoic acid binding protein II (CRABP II) were used as positive control samples. PCR reactions were performed as follows: 1 cycle of 94°C for 5 min; then 40 cycles of 94°C for 30 s, annealing temperature (AT) for 30 s, 72°C for 30 s; and finally 1 cycle of 72°C for 7 min. Products of amplification were separated on 2% agarose gel and photographed after ethidium bromide staining. Each positive result was confirmed by repeating at least three times. We selected primers using published criteria (20), and AT for each primer set was calculated using conventional formulas. Primer sequences used are listed as forward then reverse 5′ to 3′. β-actin primers 5′GGCATCGTGATGGACTCCG3′ and 5′GCTGGAAGGTGGACAGCG3′ amplify a product of 612 bp; RARα primers 5′CAGTTCCGAAGAGATAGTACC3′ and 5′TACACCATGTTCTTCTGGATTGC3′ amplify a product of 167 bp; RARβ primers 5′TCGAGACACAGAGTACCAGC3′ and 5′GAAAAAGCCCTTGCACCCCT3′ amplify product of 155 bp; RARγ primers 5′GCCTCCTCGGGTCTACAAG3′ and 5′ATGATACAGTTTTTGTCGCGG3′ amplify a product of 155 bp RXRβ; primers 5′TCAACTCCACAGTGTCGCTC3′ and 5′TAAACCCCATAGTGCTTGCC3′ amplify a product of 174 bp; RALDH2 primers 5′TTGCAGATGCTGACTTGGAC3′ and 5′TCTGAGGACCCTGCTCAGTT3′ amplify a product of 200 bp; CRABP II primers 5′TGATGAGGAAGATCGCTGTG3′ and 5′TTCCACTCTCCCATTTCACC3′ amplify a product of 191 bp; COUP-TF II primers 5′GTGGAGAAGCTCAAGGCACTG3′ and 5′CGTGCGGAGGGAAGGGAGA3′ amplify a product of 221 bp; and PDX-1 primers 5′CCTTGATATCGCTGCCACCATGAACAG3′and 5′CTGCGCTCGAGTGTAGGCAGTACG3′ amplify a product of 484 bp.

We analyzed the mRNA and protein expressions implicated in retinoid metabolism. The expression of RA synthesizing enzyme RALDH2 mRNA was detected exclusively in mesenchyme at e11.5 (Fig. 1 I). This also held true at the posttranscription level, but the expression level was very low compared with that in the mesenchyme of stomach at e11.5 and 13.5 (Fig. 1 IIC and F). The RALDH2 mRNA was also expressed in PEM rudiments throughout the rest of the developmental period (Fig. 1 I). The expression of mRNAs coding for RARs were analyzed next in e11.5 rudiments. RARα was expressed in epithelia, with very low expression in mesenchyme; RARγ was expressed only in the mesenchyme; and RARβ expression was not detected at all. The RA binding protein CRABP II was detected only in mesenchyme, and strong expression of RARs suppressor receptor COUP-TF II was seen in mesenchyme with very low expression in epithelia (Fig. 1 I). PDX-1 was used as a pancreatic epithelia marker in both RT-PCR and immunohistochemistry, demonstrating a good separation of epithelium and mesenchyme in these procedures (Fig. 1 I and IIB and E).

First, e11.5 PE rudiments were cultured in a three-dimensional gel system of collagen I or matrigel as controls. PE did not survive in collagen I gel but developed into cystic ductal structures when cultured in matrigel (Fig. 2A, C, and E). When exposed to exogenous atRA, the rudiments survived but failed to show overt differentiation in collagen I gel, but those cultured in matrigel showed very early duct differentiation with increase in endocrine cells (Fig. 2B, D, and F). The increase in endocrine cells was quantitative and statistically significant (P < 0.05; Fig. 5A).

PEM control rudiments underwent various morphological changes redolent of normal development in vivo when cultured in these two gel systems. They showed active cell proliferation as shown by BrdU staining (Fig. 4G). At day 4 of culture, endocrine cells were present as solitary cells, PDX-1 expression was seen both in acinar and endocrine cells, and acinar cells began to appear. By day 7, endocrine cells began to coalesce (Fig. 3D and G), acinar differentiation became more pronounced (Fig. 4D), and PDX-1 expression was seen almost only in insulin-producing cells (Fig. 3J). Ductal differentiation was seen more in matrigel specimen, probably owing to the presence of insoluble basement membrane components (e.g., laminin, type IV collagen, and heparan sulfate proteoglycans). These observations are analogous to the findings described in the previous experiment (19).

PEM rudiments treated with exogenous atRA showed increase in the expression of PDX-1 and the number of endocrine cells in a dose-dependent manner (Fig. 3E, F, H, I, K, and L). PDX-1 expression was detected in the ducts and endocrine cells at day 4 but almost only in insulin-producing cells after 7 days of culture (Fig. 3K and L). Most of the differentiated endocrine cells were in islet orientation (i.e., spheroidal in shape with insulin-producing cells occupying the center), few were seen as isolated solitary in or emerging from the ducts and coexpressed with PDX-1 (Figs. 3L and 6E and F). The maximum numbers of endocrine cells were seen at the 10 μmol/l concentration of atRA, and the proportion was more than double the control rudiments. The differences were significant at the 95% confidence level with P < 0.0001. However, no significant differences were found between the proportions of endocrine cells in rudiments cultured in the matrigel or collagen I gel system treated with exogenous atRA (Fig. 5B). Similarly, the pervasiveness of cystic structures increased commensurately with increasing doses of atRA (Figs. 3B and C and 5D). These cystic structures were confirmed to be ducts by cytokeratin and CA II staining (Fig. 6A and B). On the contrary, acinar differentiation became sparse in rudiments treated with increasing doses and completely arrested at a higher (100 μmol/l) dose of atRA as validated by α-amylase staining (Figs. 4C and F and 5D). At 100 μmol/l, BrdU-positive proliferating cells were seen mainly in ducts, protodifferentiated endocrine cells, and a few fully differentiated endocrine cells (Fig. 4J–L).

The controls showed mostly acinar components, identified by pyramidal epithelial cells containing a large number of secretory granules compatible with zymogen granules in the apical cytoplasm, and crowded with closely spaced parallel cisternae of granular endoplasmic reticulum in the basal region. However, in rudiments treated with atRA, large numbers of duct cells were prevalent and were identified by their typical simple epithelium, uniformity in cellular and nuclear size, and more condensed nuclear chromatin. The cytoplasm was scanty with poorly developed cytoplasmic organelles. The apical surface projected short microvilli, and lateral surfaces of neighboring duct cells exhibited more interdigitations (Fig. 6C). The endocrine cells were identified by size and shape of endocrine granules and were different from acini or duct cells. A few of the duct cells also exhibited some endocrine granules in their cytoplasm. The ultrastructure findings were compared with previously published articles (21,22) and were compatible with the immunohistochemical findings shown above.

Analysis of cultured pancreas with atRA added on day 6 of culture showed profound dysmorphogenesis of acinar components (Fig. 7E). These cells demonstrated overexpression of proapoptotic protein Bax in their cytoplasm (Fig. 7G) with increase in ssDNA-positive nuclear fragmentations (Fig. 7H). However, the endocrine cells were almost free from the apoptotic effects of atRA. There was almost no expression of antiapoptotic protein Bcl-2 or Bcl-x_L in atRA-treated rudiments (data not shown), and no obvious signs of Bax or ssDNA expression were noted in the controls (Fig. 7C and D). Many PDX-1 overexpressing acinar cells as well as endocrine cells with preislet morphology were seen in atRA-treated samples (Fig. 7F). The quantitative study demonstrated the significant increase in the number of PDX-1–positive cells (Fig. 7).

RA has been well acknowledged as a signaling molecule for mesenchymal/epithelial interactions of various organs (11–14), and here we have demonstrated its role in the development of pancreas, especially in differentiation of ducts and endocrine cells. Our study focuses on e11.5 pancreatic rudiments when the local mesenchyme is well developed and yet can be separated mechanically from epithelium. This gives a good ground to explore the type of signaling molecules involved in the interaction between these two surfaces.

Mesenchymal interaction could be due to secretion of inducing or transforming factors, production of extracellular matrix, and information exchanged through cell-to-cell contact during development. The ability of exogenous atRA to induce differentiation in both PE and PEM and its exclusive endogenous expression in mesenchyme suggests that it is one of the candidates of mesenchymal factors for endocrine differentiation. This hypothesis supports the permissive effect of mesenchyme on the development of the endocrine pancreas (23).

RA is synthesized from retinaldehyde mostly by NAD-dependent dehydrogenase (24). Among the three dehydrogenases, RALDH2 exhibits the greatest substrate specificity, and its distribution provides the most accurate guide to the localization of atRA in the embryonic tissues (25,26). Our highly sensitive RT-PCR and immunohistochemistry results show that RALDH2 is expressed solely in the mesenchyme at e11.5 (Fig. 1 I and IIC and F) and in whole pancreas throughout the rest of the developmental period. This result is similar to the findings that RALDH2 is widely expressed in mesenchyme in developing gut including the foregut (27,28).

The expression of CRABP II in mesenchyme also illustrates the dominant role of mesenchyme in retinoid metabolism; in addition, its expression contributes to the synthesis of RA because it has been shown to be associated with cells that synthesize RA (29,30). CRABPs may have some role in development of islets because they are localized in adult islets (31), and transgenic mice overexpressing CRABP I die prematurely as a result of tumor originating from islet of Langerhans (32).

Retinoid signaling is transduced by two families of nuclear receptors, RAR (α, β, and γ) and RXR (α, β, and γ) isoforms. All-trans form is the natural ligand for the RAR, and 9-cis form is the natural ligand for RXR, although the latter binds to both receptor families (26). A previous article by Kadison et al. (33) showed that at e12.5, RARα expression is seen only in mesenchyme and RARβ/RARγ in epithelia by Western blot analysis, but our highly sensitive RT-PCR results at e11.5 show few differences. Here we have tried to be meticulous by using absolute epithelial marker PDX-1 and appropriate positive controls to derive our conclusions. Therefore, these discrepancies could be due either to the method and time of analysis or to the quality of samples. The expression of RARs in both epithelia and mesenchyme raises the possibility that endogenous atRA acts in both an autocrine and a paracrine manner through these receptors. However, the strong expression of RAR repressor COUP-TF II in mesenchyme suggests that endogenous RA primarily prefers paracrine action because most of the autocrine actions may be repressed by these orphan receptors in mesenchyme (34).

Our ex vivo observations demonstrated the endogenous presence of retinoids. How does this RA affect the pancreatic development and differentiation? To discern the answer, we cultured undifferentiated e11.5 pancreatic rudiments at different concentration of exogenous atRA. The results obtained upon atRA administration in the three-dimensional organ culture system strongly suggest that it plays an active role in the development of pancreas. Besides the regressive effect on acinar differentiation, both quantitative and significant augmentation of endocrine cells at the posttranscriptional level was appreciated. The increase in differentiation of ductal structures (confirmed by cytokeratin/CAII staining and electron microscopy) is also important because they are necessary for branching morphogenesis and for origin of endocrine cells. In our model, endocrine cells are seen to originate from newly differentiated ducts, which was validated by double staining of insulin and BrdU/PDX-1 (Fig. 6D–F). This finding resembles and supports the in vivo and in vitro model of endocrine pancreas development (35,36). Kadison et al. (33) showed that exogenous atRA and 9-cis RA have completely different ontogenic effects during pancreas differentiation. However, few comparative studies between RA isomers have shown 9-cis to be almost 10 times more potent than the all-trans form (37,38). Our results also suggest that atRA has a similar type of ontogenic effects as 9-cis RA at a higher dose.

With sharp changes in the fate of pancreas cultured under the influence of exogenous atRA, we tried to seek the mechanisms underlying the inhibition of acinar cells and induction of ducts or endocrine cells. Our results indicate that upregulation of PDX-1 by atRA plays the central role on these phenotypic selections. The possible underlying mechanism for inhibition of acini is via apoptosis. The PEM rudiments pulse treated with atRA 24 h before harvesting showed dysmorphogenesis of acinar cells and overexpression of the proapoptosis protein Bax in their cytoplasm with increase in anti-ssDNA positive DNA fragmented cells. The demonstration of apoptosis in these cells could be due to the direct effect of atRA itself or the upregulation of PDX-1, because persistent upregulation of PDX-1 destroys the acini population by fatty infiltration and apoptosis (39).

Acinoductal transdifferentiation is the most likely cause of ductal induction in atRA-treated rudiments (40). This hypothesis is supported by some amylase-positive ducts in atRA-treated rudiments and duct cells exhibiting increase in PDX-1 expression.

The mechanism for induction of endocrine cells may be due either to stimulating the expression of insulin mRNA levels probably by binding to the RARE located upstream of the insulin gene enhancer (41) or to the induction in endocrine transcription factors by atRA. RA has been implicated in foregut A-P patterning through expression of homeobox genes (42,43). The homeobox gene PDX-1 plays a critical role in development of pancreas (44,45). Recent studies show that atRA induces the XIHbox-8 gene in Xenopus (46,47), the amphibian homologue of PDX-1. In our model, the effect of atRA on induction of endocrine cells seems to be partially mediated by the upregulation of PDX-1 gene. There is a growing list of experiments in different organs, cell lines, and stem cells, where the role of RA is as an inducer of pancreatic endocrine transcription factors. The role of Pax-6 in RA signaling of the developing eye (48), Xlim-1 (Xenopus equivalent to human Lim-1) mRNA induction by RA (49), and upregulation of Neuro D on exposure of RA (50) are a few of these examples. However, more studies are needed to clarify the detailed relationship between RA and these endocrine transcription factors during pancreas development.

In conclusion, the overall findings demonstrate that atRA is an important signaling molecule in the development of the embryonic pancreas and plays a great role in differentiation of endocrine and ductal cells, predominantly through paracrine actions and upregulation of PDX-1 gene. Understanding the full picture of the developmental process of the endocrine pancreas is still incomplete, but we believe that identification of signaling molecules, information on their proliferative or inductive signals, and elucidating their roles with transcription factors will certainly unravel the greater details in the future.

This study was supported in part by a grant from the Ministry of Education of Japan.

We extend our deep appreciation to Dr. Yoichi Tani (Dako, Japan) for expert opinion in immunohistochemical techniques and Michiharu Kurino (Laboratory of Pathology, Kyoto University Hospital) for technical assistance with electron microscopy. We also thank Prof. C.V. Wright and Dr. Peter McCaffery for generously providing the antibodies.