Epidermal Growth Factor Increases Undifferentiated Pancreatic Embryonic Cells In Vitro: A Balance Between Proliferation and Differentiation

Pregnant Wistar rats were purchased from Janvier breeding center (CERJ, Le Genet, France). The morning the vaginal plug was discovered was designated as embryonic day (e) 0.5. Animals had free access to food pellets and water. Pregnant rats at 13.5 days of gestation were killed by injection of a lethal dose of pentobarbital (Sanofi, Libourne, France). The embryos were harvested, and the dorsal pancreatic buds were dissected as described (10). The pancreatic epithelium was separated from its surrounding mesenchyme, as described previously (6) with minor modifications. Briefly, the stomach, pancreas, and a small portion of the intestine were dissected together and incubated with 0.5 mg/ml of collagenase A (Boehringer-Mannheim, Mannheim, Germany) at 37°C for 30 min. They were then washed several times with Hanks’ balanced salt solution (HBSS; Gibco) at 4°C. The epithelium was then mechanically depleted from the surrounding mesenchyme using needles on a 0.25% Agar, 25% HBSS, 75% RPMI (Gibco) gel in a Petri dish.

Pancreatic epithelia were embedded into 500 μl of collagen gel (10% RPMI 10× [Sigma-Aldrich], 80% type I rat tail collagen [2.5 mg/ml; Sigma-Aldrich], and 10% sodium bicarbonate in NaOH 0.1 mol/l) into four-well plates (Nunc), as previously described (6,11). Once the gel had polymerized, 500 μl of RPMI 1640 (Gibco) containing penicillin (100 U/ml), streptomycin (100 μg/ml), HEPES (10 mmol/l), l-glutamin (2 mmol/l), and nonessential amino acid (1×; Gibco) were added. The medium was supplemented with 1% heat-inactivated fetal calf serum (FCS) (Hyclone). At this concentration of FCS, the development of the endocrine tissue was identical to the one obtained with 10% FCS. Cultures were maintained at 37°C in a humidified atmosphere of 95% O₂/5% CO₂. The medium was changed every 2 days, and 50 ng/ml human recombinant EGF (Sigma-Aldrich (diluted in phosphate-buffered saline 0.1% bovine serum albumin or its carrier) were added every day. This daily addition and concentration of EGF was chosen because in preliminary experiments it gave the strongest effect on cell differentiation. PD98059 (Calbiochem) was diluted in DMSO and used at 50 μmol/l. Comparable DMSO concentrations were made in culture medium. To label cells in the S phase, bromo-deoxy-uridine (BrdU; 10 μmol/l) was added to the culture medium. At the indicated times, the pancreatic epithelia were photographed and fixed for immunohistochemistry, as described below.

Pancreatic rudiments were photographed and fixed in formalin 3.7% for 1 h, preembedded in an agarose gel (4% of type VII low gelling−temperature agarose [Sigma] in Tris-buffered saline), and embedded in paraffin. Consecutive sections 4 μm thick were collected on gelatinized glass slides, deparaffinized, and microwaved for 12 min. Immunohistochemistry was performed as previously described (6,8,9) using the following antibodies: mouse anti-human insulin (Sigma; 1/2,500), mouse anti-human glucagon (Sigma; 1/2,000), rabbit anti-glucagon (Diasorin; 1/2,000), rabbit anti-human amylase (Sigma; 1/300), mouse anti-human pan-cytokeratin (Sigma; 1/100), mouse anti-BrdU (Amersham; 1/4), and mouse anti-porcine vimentin (Dako; clone V9, 1/100).

The fluorescent secondary antibodies obtained from Jackson Immunoresearch Laboratories were fluorescein anti-rabbit antibodies (1/200), fluorescein anti-mouse antibodies (1/150), Texas Red anti-mouse antibodies (1/200), and Texas Red anti-rabbit antibodies (1/200)

In some experiments, adjacent sections were stained for endocrine plus mesenchymal markers (insulin/glucagon plus vimentin) and acinar plus epithelial markers (amylase and cytokeratin, respectively). False colors were generated using Graphic Converter 3.9.1 (Lemke Software), and the images from adjacent sections were overlaid using Photoshop 5.5 (Adobe).

Confocal microscopy was performed using a CS4D confocal microscope (Leica). Sections 4 μm thick were analyzed every 0.5 μm. Recomposed color images were generated using GraphicConverter 4.0.1 (Lemke Software). Images obtained for the different fluorochromes were then overlaid using Photoshop 5.5 (Adobe).

For each pancreatic epithelium that was analyzed, all sections (50−100 per rudiment) were numbered using a Hamamatsu C5810 cooled tri-CCD camera. All images were taken at the same magnification. On every image, the surfaces of the different stainings were quantified with IPLab (version 3.2.4, Scananalytics). The surfaces were then added up per rudiment. When a section was missing (<2% of the total section number), we took the mean value of the staining in the adjacent sections into account. To quantify the total size of the rudiments, the surface of all the sections of the rudiment prepared for immunohistochemistry were measured, and areas of all sections of each rudiment were added. Statistical analysis was performed using StatView 5.0 (SAS Institute). Because we could not assume that the data followed a binomial law, we used nonparametric Mann-Whitney U tests to compare the different groups.

We first tested the effects of PD98059, an inhibitor of the mitogen-activated protein (MAP) kinase pathway, on cell proliferation and differentiation. Epithelial rudiments were cultured for 3 days with or without PD98059. As shown in Figs. 1A and B, in the presence of PD98059, the size of the rudiments and the number of BrdU-positive cells were decreased when compared with controls grown for the same period in the absence of PD98059. Moreover, the number of BrdU-positive cells per surface unit was also decreased (P < 0.05) in the presence of PD98059. In fact, while the size of the tissue increased during the 3-day culture period in control conditions (Fig. 1B, top left panel; compare columns 1 and 2), no increase in terms of size of the tissue was observed during a 3-day culture period in the presence of PD98059 when compared with uncultured rudiments (Fig. 1B, top left panel; compare columns 1 and 3).

At the same time, the absolute mass of endocrine cells that differentiated in the presence of PD98059 was increased when compared with the ones that developed in the absence of PD98059 (Figs. 1A and B, bottom panel). In fact, while the endocrine cell mass was increased by 1.8-fold during a 3-day culture period in the absence of PD98059, it was increased 3.3-fold during a 3-day culture period in the presence of PD98059 when compared with uncultured rudiments (Fig. 1B, bottom panel).

When e13.5 pancreatic epithelial buds were grown in culture in a collagen gel in the absence of EGF over 7 days, the size of the tissue increased 1.5-fold. On the other hand, in the presence of EGF, the size of the tissue increased by 2.8-fold during a 7-day culture period. Thus, with EGF, after 7 days in culture, a 1.9-fold increase in size (P = 0.0040) was seen in EGF-treated epithelial buds compared with controls buds grown in the same conditions (Fig. 2). Immunohistochemistry analysis was next performed to define the effect of EGF treatment on the absolute mass of endocrine and acinar tissue that developed. EGF treatment did not modify the absolute amylase-expressing cell mass (Fig. 3A and data not shown). On the other hand, in the presence of EGF, the absolute mass of insulin- and glucagon-expressing cells that developed was strongly decreased when compared with controls (2.96-fold decrease for insulin, P < 0.005; 5.23-fold decrease for glucagon, P < 0.005) (Fig. 3).

After 7 days of culture in the presence of EGF, a large number of cells stained negative for endocrine and acinar markers when compared with rudiments grown in the absence of EGF (Figs. 4A and B). As shown in Fig. 4B, such negative cells expressed high levels of cytokeratins and were more numerous in EGF-treated buds than in buds grown in the absence of EGF (compare Figs. 4A and B). In fact, 4.65 times more cells (P < 0.01) stained negative for endocrine and acinar markers in cultures performed in the presence of EGF than in the absence of EGF. On the other hand, both in EGF-treated and control cultures, very few cells expressed vimentin, a marker of mesenchymal cells, indicating that EGF has no effect on the few remaining pancreatic mesenchymal cells (Figs. 4A and B).

To define whether, in the presence of EGF, cells expressing high levels of cytokeratins proliferate, a 6-h pulse of BrdU was performed on day 7 of the culture. As shown in Figs. 4D and E, cells that stained positive for BrdU and cytokeratin could be detected in rudiments grown with EGF, indicating that cells expressing high levels of cytokeratins proliferated upon EGF treatment. On the other hand, in the absence of EGF, the few cells expressing high levels of cytokeratins did not proliferate (Fig. 4C). The few cells that did proliferate in the absence of EGF were amylase-expressing cells (data not shown).

We next asked whether cells expressing high levels of cytokeratins that proliferated in the presence of EGF had the ability to differentiate into insulin-expressing cells. For that purpose, two sets of experiments were performed.

First, e13.5 epithelial buds were grown in culture in the presence of EGF. After 7 days, EGF was removed and the rudiments were kept in culture for an additional week. A 8.93-fold increase (P = 0.0027) in the number of insulin-expressing cells was observed in epithelial buds treated for 7 days with EGF followed by a 7-day period in the absence of EGF when compared with epithelial buds kept in culture for 7 days with EGF (Fig. 5A [compare panels a and b] and B [compare columns 1 and 2]). On the other hand, no major increase was seen in the absolute number of insulin-expressing cells that developed during 7 or 14 days in the absence of EGF (Fig. 5A [compare panels c and d] and B [compare columns 3 and 4]).

Next, the fate of the cells that proliferated during the first week of culture with EGF was followed. For that purpose, the epithelial buds were grown for 7 days with EGF. At the end of the seventh day, a BrdU pulse was performed. Then 6 h later, the epithelial buds were either fixed for immunohistochemistry or washed, transferred to a new gel, and grown for an additional 7-day period in the absence of EGF. As shown in Fig. 6B, cells that stained positive for both insulin and BrdU could be found after 7 additional days in culture, whereas such double-positive cells were not present at the end of the BrdU pulse at day 7 (Fig. 6A). Only a fraction (∼50%) of BrdU-positive cells were insulin positive after a pulse at day 7 and 7 additional days in culture (Fig. 6B), suggesting that not all the potential stem cells that were proliferating with EGF did differentiate into insulin-expressing cells upon EGF removal.

Taken together, these data strongly suggest that insulin-expressing cells did differentiate in the absence of EGF from precursor cells that were proliferating in the presence of EGF.

The insulin cell mass that developed during 14 days in the absence of EGF was compared with the one that developed when epithelial buds were cultured during the first week in the presence of EGF followed by 1 week in the absence of EGF. As shown in Fig. 5A (compare panels b and d) and B (compare columns 2 and 4), more insulin-expressing cells (1.78-fold increase; P < 0.05) developed when the epithelial buds were cultured during the first week with EGF. This indicated that EGF increased the mass of precursor cells that then differentiated into insulin-expressing cells.

We demonstrated that in vitro, EGF has the ability to expand the pool of embryonic pancreatic epithelial precursor cells. In the meantime, their differentiation into endocrine tissue is repressed. Once expanded and after EGF is removed, these precursor cells differentiate and form endocrine cells by a default pathway.

It is well established that during embryonic life, endocrine and acinar cells derive from precursor cells located in the pancreatic epithelium (12). Such precursor cells, which are not yet characterized, will proliferate and differentiate into mature cells. It has been shown in vivo that when isolated embryonic pancreatic epithelium is grafted under the kidney capsule, it develops into endocrine tissue, but acinar tissue does not develop. These experiments indicated that the development of the endocrine tissue can occur in the absence of signals from the mesenchyme that surrounds the epithelium (13). It has also been shown in vitro that when e12−e13 pancreatic epithelium is cultured in the absence of mesenchyme, pancreatic endocrine cells develop properly (6,7). In fact, in vitro, more insulin-expressing cells develop in the absence of mesenchyme, when epithelial cell proliferation is low, than in the presence of mesenchyme, when proliferation is high, suggesting that signals from the mesenchyme activate the proliferation of immature epithelial cells and repress their differentiation into endocrine tissue (6). Taken together, these results strongly suggest that activation of the proliferation of immature embryonic epithelial cells represses their differentiation into endocrine cells. Cell proliferation would thus act as a repressor of endocrine cell differentiation. The data presented in this study further support this hypothesis. Indeed, when e13.5 pancreatic epithelium is cultured in the presence of EGF, epithelial cell mass increases, whereas endocrine cell differentiation is repressed. On the other hand, in the presence of PD98059, an inhibitor of the MAP kinase pathway, cell proliferation is decreased, but endocrine cell development is activated. However, the possibility cannot be fully excluded that the MAP kinase inhibitor or EGF acts directly on both cell proliferation and endocrine differentiation, without any link between proliferation and differentiation.

In the present study, in control cultures performed in the absence of EGF, most of the cells differentiated into endocrine or exocrine cells during a 7-day culture period. Little additional cell differentiation occurred when the epithelium was grown during an additional week in the absence of EGF to activate endocrine cell differentiation (Fig. 5B, columns 3 and 4). Thus it seems that in these conditions, the pool of precursor cells was strongly depleted. On the other hand, in the presence of EGF, a large number of cells proliferated and remained undifferentiated after 7 days in culture, and the number of endocrine cells that developed had decreased. These undifferentiated cells develop massively into insulin-expressing cells by a default pathway after EGF removal. It is important that significantly more insulin-expressing cells were present after 14 days of culture when EGF was added during the first week and removed during the second week, than after 14 days of culture in the absence of EGF. Thus, in vitro, the development of a normal number of endocrine cells in the pancreas requires a proper proliferation of precursor cells and their differentiation at the right time. Perturbation of one of these events decreases the development of the endocrine tissue. This is the case in vivo in mice with a perturbed Notch signaling pathway in the pancreas, such as mice overexpressing neurogenin-3 under the control of the PDX-1 promoter (14) or mice deficient in Hes1 (15). In these mutant mice, an early and massive differentiation of the pool of precursor cells occurs. Such a pool is depleted and the final mass of endocrine cells that will develop is decreased. This could also be the case for mice deficient in EGFR that have a small pancreas with an abnormal development of the endocrine tissue (5).

In vitro, when pancreatic epithelium is grown in control conditions, the majority of the cells differentiate into endocrine or acinar tissue after a 7-day culture period. On the other hand, in the presence of EGF, a large number of cells remain negative for endocrine and acinar markers. In the present study, we provided different arguments showing that cells that stain negative for endocrine and acinar markers are precursor cells. First, such cells that stain negative for endocrine and acinar markers express a high level of cytokeratin; however, during development, endocrine and acinar cells derive from the endodermal epithelial cells that stain positive for cytokeratins (16,17). Second, the cells expressing high levels of cytokeratin can proliferate upon EGF treatment and next differentiate into insulin-expressing cells. Moreover, when epithelial buds were grown with EGF, pulsed with BrdU at the end of the 7^th day, and grown for an additional week in the absence of EGF, cells that stained positive for both insulin and BrdU were found at the end of the culture. Such double-positive cells were not present at the end of the BrdU pulse at day 7. Third, the expansion of such a pool of putative precursor cells gives rise to a high amount of insulin-expressing cells when differentiation occurs upon EGF removal.

We demonstrated here that the cells that express high levels of cytokeratin and stain negative for endocrine and acinar markers do proliferate upon EGF treatment. Such an effect of EGF on cell proliferation resembles the one found when pancreatic duct cells from adult guinea pigs are treated with EGF (18). Our data also demonstrate that at the same time, differentiation into endocrine cells is repressed. However, the expanded cells keep the capacity to differentiate after EGF removal. The effects of EGF presented here resemble the ones found when EGF is added to cultures of embryonic or adult precursor cells derived from the nervous system. In this tissue, EGF has been found to act as a mitogen for precursor cells, inducing their proliferation while repressing their differentiation. When EGF is removed, cell differentiation into mature cells such as neurons or glia occurs (19,20,21,22,23). Ligands of receptor tyrosine kinases such as EGF should thus be useful in generating a large pool of precursor cells. Such a pool is useful for studying in detail these expanded pancreatic precursor cells by characterizing specific markers expressed on their surface. This type of marker will be necessary to purify multipotent pancreatic embryonic cells, as is the case for other tissues such as hematopoietic stem cells (24) or neural crest stem cells (25).

This study demonstrated that proliferation and differentiation can be balanced and that EGF can control this equilibrium. Whether this is operational in vivo needs to be demonstrated. However, EGF is certainly efficient in vitro, and could be used to increase the amount of differentiated β-cells in vitro. Recently, because of its therapeutic potential, the expansion of β-cells in vitro has been the focus of a large number of studies. In the vast majority of such studies, specific growth factors have been tested for their ability to increase the proliferation of pre-existing β-cells. The growth effect of factors such as hepatocyte growth factor, growth hormone and prolactin, EGF, and platelet-derived growth factor has been tested on mature β-cells from rodents and humans (26,27,28). In these studies, no major increase in β-cell mass has been found using these factors. We demonstrated in the present study that an alternative strategy is to increase the proliferation of precursor cells using ligands of receptor tyrosine kinases such as EGF. Once EGF is removed, such amplified precursor cells differentiate en masse into endocrine cells.

C.C.-M. was a recipient of a fellowship from the Fondation de France. This work was supported by grants from the Juvenile Diabetes Foundation International.

We thank Dr. M. Kedinger for her guidance with the establishment of the mesenchyme depletion procedure and for her useful advice. I. Le Nin is acknowledged for her technical assistance and Arnaud Mailleux (Institute Curie, Paris) is acknowledged for help with confocal microscopy.