Control of β-Cell Differentiation by the Pancreatic Mesenchyme

Pregnant Wistar rats were purchased from the Janvier breeding center (Centre d'Elevage René Janvier, Le Genest, France). Pregnant female rats at 13.5 days of gestation were killed by CO₂ asphyxiation, according to the guidelines of the French Animal Care Committee. To study cell proliferation in vivo, pregnant rats at E13.5 and E18.5 were injected with bromodeoxyuridine (BrdU) (100 μg/g body wt) (Sigma) and killed 1 h later.

The embryos were harvested on E13.5. The stomach, the pancreas, and a small portion of the intestine were dissected together; then, either the pancreatic primordium was dissected or the mesenchyme was separated from the pancreatic epithelium as described previously (21).

Dorsal pancreatic rudiments with or without their mesenchyme were laid on Millicell culture plate inserts (Millipore) in 35-mm sterile Petri dishes containing 2 ml RPMI 1640 (Invitrogen) supplemented with penicillin (100 units/ml), streptomycin (100 μg/ml), HEPES (10 mmol/l), l-glutamine (2 mmol/l), nonessential amino acids (1×; Gibco), and 10% heat-inactivated calf serum (Hyclone). Under such culture conditions, the explants grew at the air/medium interface. Cultures were maintained at 37°C in humidified 95% air/5% CO₂. The medium was changed every other day. Recombinant human FGF10 (50 ng/ml; R&D Systems) was used in the presence of heparin (50 μg/ml; Sigma). For cell proliferation, BrdU (10 μmol/l) was added to the culture medium.

Tissues were fixed in 10% formalin and processed for immunohistochemistry, as described previously (21,22). The following antibodies were used: mouse anti-human insulin (1/2,000; Sigma), rabbit anti-insulin (1/2,000; Diasorin), mouse anti-BrdU (1/2; Amersham), and rabbit anti-PDX1 (1/1,000) (22). The fluorescent secondary antibodies were fluorescein isothiocyanate anti-rabbit and Texas-red anti-mouse antibodies (1/200; Jackson Immunoresearch) and Alexa-fluor anti-rabbit antibody (1/400; Biogenex). Photographs were taken using a fluorescence microscope (Leitz DMRB; Leica) and digitized using a C5810 cooled 3CCD camera (Hamamatsu, Middlesex, NJ). No signals were observed when the first antibodies were omitted.

Tissues were fixed at 4°C in 4% paraformaldehyde in PBS and either included in paraffin or cryoprotected in 15% sucrose-PBS at 4°C overnight, embedded in 15% sucrose-7.5% gelatin in PBS, and frozen in isopentane. Paraffin sections (4 μm thick) or cryosections (14 μm thick) were prepared. A Ngn3 probe (726 bp) was used (23) and in situ hybridization was done as previously described (24). No signal was obtained when a sense riboprobe was used.

To quantify the absolute numbers of insulin- and Ngn3-expressing cells, all sections of each pancreatic rudiment were digitized. On every image, the surfaces of insulin or Ngn3 stainings were quantified using IPLab Eval (version 3.2.4; Scanalytics), and the stained areas were summed as previously described (25). Three to four rudiments were analyzed per condition.

To measure proliferation of either β-cells or early progenitors expressing PDX1, we counted the frequency of BrdU⁺ nuclei among 1,000 insulin⁺ cells or 1,000 early PDX1⁺ progenitors per rudiment. Three rudiments were analyzed per condition.

To quantify the maintenance of early PDX1⁺ progenitors, we counted the number of insulin⁻ cells among 1,000 PDX1⁺ cells. The percentage of undifferentiated PDX1-expressing cells was then calculated. Three rudiments were analyzed per condition. Statistical significance was determined using Student's t test.

Total RNA was extracted from pools of three to four embryonic pancreases (cultured with or without mesenchyme) using Rneasy microkit (Qiagen). The cDNAs were generated using Superscript reagents (Invitrogen) according to the manufacturer's instructions. Real-time PCR was performed using a 7300 real-time PCR system (Applied Biosystems). Each reaction consist of a mix of Taqman universal PCR master mix (Applied Biosystems) with primers and Taqman-labeled probe specific for each gene (Applied Biosystems) and run after universal thermal cycling protocol (95°C for 10 min followed by 40 cycles of 95°C for 15 s and 65°C for 1 min). Control reactions in the absence of template were included in each assay. Results were normalized to the transcript encoding the housekeeping protein cyclophilin A. Each point represents the mean ± SE of three individual pools. PCR primer sequences are available on request.

We developed a culture model that allowed β-cell differentiation with and without mesenchyme. For this purpose, we dissected pancreatic epithelia at E13.5. At this stage, the pancreatic epithelium is mainly composed of early PDX1⁺ progenitors, with few glucagon⁺ endocrine cells. We cultured pancreatic rudiments at the air/liquid interface on filters floating on culture medium. Without mesenchyme, the epithelium acquired a spherical shape after 1 day in culture, and the size of the tissues did not change significantly during the 7-day culture period (Fig. 1A–E). At the end of the 7 days, the epithelium appeared as a small dark sphere surrounded by a variable number of translucent buds (Fig. 1E). In the presence of mesenchyme, in contrast, the epithelium grew rapidly, spread into the mesenchyme, and developed lobules (Fig. 1F–J).

We next analyzed the in vitro pattern of expression of different pancreatic genes. We focused on Nkx6.2 expressed in vivo at early stages of development, its expression declining in late embryogenesis (26), Ngn3 and Pax4 known to be expressed in pancreatic endocrine progenitors, and NeuroD expressed both in endocrine progenitors and in mature β-cells. We also followed the expression of three genes expressed in mature β-cells: Pcsk1, Pcsk2, and Abcc8. As shown in Fig. 1K, Nkx6.2 was expressed during the first 3 days of culture, its expression decreasing thereafter. The expression of Ngn3 and Pax4 was turned on and next decreased both with and without mesenchyme, as expected for markers of pancreatic progenitors. The expression of NeuroD was rapidly activated and did not decrease with time, as expected for a gene expressed both in progenitors and mature β-cells. Finally, the expression of Pcsk1, Pcsk2, and Abcc8 increased at late time points of in vitro development.

We first compared β-cell development from epithelia cultured with or without mesenchyme. In epithelia cultured with or without mesenchyme, the first insulin-expressing cells were detected after 3 days of culture (Fig. 2C and H), and their number increased thereafter (Fig. 2D, E, I, and J). We next compared β-cell development in both conditions. As shown in Fig. 2K, both with and without mesenchyme, the number of β-cells increased during the first 5 culture days. However, the number of β-cells did not increase further from days 5 to 7 without mesenchyme but continued to increase with mesenchyme. As a result, a significantly larger number of β-cells developed with than without mesenchyme. Individual β-cell size did not vary in the different culture conditions (data not shown). When epithelium was dissected, directly reassociated with mesenchyme, and cultured, its development was similar to the one of whole pancreas (data not shown).

We compared BrdU incorporation in β-cells developed without or with the mesenchyme. After 3 days of culture, β-cells developed without or with mesenchyme did not incorporate BrdU (Fig. 3A and D). After 5 and 7 days of culture, some β-cells developed without mesenchyme incorporated BrdU (Fig. 3B and C); this proliferation was even lower with mesenchyme (Fig. 3E–G for quantification). This low proliferation of β-cells resembles the one found in the rat embryonic pancreas at E18.5 (Fig. 3H). Thus, the increase in the number of β-cells that developed with mesenchyme cannot be explained by an activation of β-cell proliferation by the mesenchyme.

We cultured epithelia for 1 day with or without mesenchyme and added BrdU during the last hour of culture. The percentage of PDX1⁺ cells that incorporated BrdU was 5 ± 2% without mesenchyme (Fig. 4A) and 39 ± 7% with mesenchyme (Fig. 4B, B′, and C for quantification). When epithelium was dissected, directly reassociated with mesenchyme, and cultured for 1 day, the proliferation of early PDX1⁺ progenitors was similar to the one of whole pancreas (data not shown). A similar high level of PDX1⁺-cell proliferation in the presence of mesenchyme can be found in the E13.5 rat embryonic pancreas (Fig. 4D and D′). Thus, the mesenchyme is necessary for maintaining the proliferation of early pancreatic progenitors.

To investigate the consequences of proliferation of early PDX1⁺ pancreatic progenitors, we quantified the maintenance of these cells with or without mesenchyme. We cultured pancreatic epithelia with or without mesenchyme and quantified the decrease in the number of early PDX1⁺/insulin⁻ progenitors. At E13.5, the epithelium is composed of early PDX1⁺/insulin⁻ cells (Fig. 5A). In cultures without mesenchyme (Fig. 5B–E and J for quantification), the proportion of early PDX1⁺/insulin⁻ progenitors decreased rapidly, and after 3 and 5 days, 41 ± 4 and 23 ± 5% of PDX1⁺ cells remained insulin negative, respectively. With mesenchyme (Fig. 5F–I and J for quantification), the decrease in the proportion of early PDX1⁺/insulin⁻ progenitors was slower, and after 3 and 5 days, 76 ± 4 and 51 ± 11% of PDX1⁺ cells remained insulin negative, respectively. By knowing the absolute area occupied by β-cells and the percentage of undifferentiated Pdx1⁺ cells, we deduced and compared the absolute number of undifferentiated Pdx1⁺ cells. We found 4.02 and 4.37 times more PDX1⁺ insulin⁻ cells with than without mesenchyme after 3 and 5 days of culture, respectively. These results indicate that, in addition to increasing progenitor cell proliferation, the mesenchyme maintains a pool of early PDX1⁺ progenitors.

We next monitored Ngn3 expression in epithelia grown with or without mesenchyme. At E13.5, very few cells expressed Ngn3 in rat pancreatic epithelium (Fig. 6A). After 1 day of culture without mesenchyme, Ngn3 expression increased dramatically, reaching a peak followed by a plateau until day 3 then by a sharp decrease on days 5 and 7 (Fig. 6B–F). With mesenchyme (Fig. 6G–K), the number of Ngn3⁺ cells increased slightly at day 1 compared with epithelia cultured alone, but interestingly, in 3- and 5-day-old cultures, the number of Ngn3⁺ cells was dramatically higher in epithelia cultured with than without mesenchyme (Fig. 6L). In addition, Ngn3 expression was prolonged with mesenchyme: Whereas Ngn3⁺ cells were extremely scarce after 5 days and undetectable after 7 days without mesenchyme, they were present after 5 and 7 days of culture with mesenchyme. Compare Fig. 6D with I, and compare Fig. 6E with J. Quantifications are shown in Fig. 6L.

Hypotheses to explain the increase in the number of Ngn3⁺ cells in the presence of mesenchyme include increased differentiation of amplified PDX1⁺ progenitors and proliferation of Ngn3⁺ cells. To test the latter hypothesis, we measured BrdU incorporation by Ngn3⁺ cells grown in vitro from epithelia cultured for 3 or 5 days with or without mesenchyme. Extremely rare Ngn3⁺ cells incorporating BrdU could be found in epithelia that developed either with or without mesenchyme (Fig. 7). This poor in vitro proliferation of Ngn3⁺ cells is reminiscent of the low level of Ngn3⁺-cell proliferation in the pancreas at E16.5 and E18.5 (data not shown) and could not explain the major increase in the number of Ngn3⁺ cells in the presence of mesenchyme. Thus, our data strongly suggest that the mesenchyme may amplify Ngn3 expression by increasing the magnitude and duration of development of early PDX1⁺ progenitors into Ngn3⁺ progenitors.

In mice, Fgf10 is expressed in the mesenchyme adjacent to the early pancreatic epithelial buds and required for normal development of the pancreas (17). In rat, FGF10 is expressed in the pancreatic mesenchyme between E12.5 and E18 (27; data not shown). We thus tested whether the effects of the mesenchyme on early progenitor cell proliferation and differentiation could be mimicked by FGF10. Epithelia were cultured for 1 day with or without FGF10. As shown in Fig. 8A–C, FGF10 treatment significantly increased the proliferation of PDX1⁺ progenitors. To test the effect of FGF10 on progenitor cell differentiation, epithelia were cultured for 1, 3, or 5 days alone, with mesenchyme, or with FGF10. Without mesenchyme and FGF10, Ngn3 expression increased after 1 day of culture, reached a plateau at day 3, and decreased thereafter. On the other hand, in the presence of either mesenchyme or FGF10, the pattern of expression of Ngn3 was amplified and prolonged (Fig. 8D–M). For example, whereas after 5 days of culture, extremely rare Ngn3⁺ cells were detected in epithelia cultured alone (Fig. 8G), Ngn3⁺ cells were abundant in epithelia cultured with FGF10 (Fig. 8M). Such effects of FGF10 on Ngn3 expression were further confirmed by real-time PCR (data not shown).

Finally, we quantified the absolute number of β-cells developed in epithelia cultured for 5 days in the absence or presence of FGF10. As shown in Fig. 8N–P, FGF10 treatment significantly increased the absolute number of β-cells that develop. Such effect of FGF10 was also confirmed by real-time PCR (data not shown).

Our results show that the mesenchyme positively regulates the final number of β-cells developed from embryonic pancreas. This effect was not due to a direct action of the mesenchyme on the proliferation of either β-cells or Ngn3⁺ endocrine progenitors. Instead, the mesenchyme acted by increasing the proliferation of early PDX1⁺ pancreatic progenitors and amplifying and prolonging the formation of Ngn3⁺ endocrine progenitors.

We used an in vitro model that allows β-cell development from embryonic pancreatic epithelium cultured with and without mesenchyme (28). In earlier work, we showed that β-cells developed from embryonic pancreatic epithelium cultured in collagen gel without mesenchyme (20), whereas β-cells developed poorly when cultured with mesenchyme (21). This inhibitory effect of mesenchyme was due to delayed Ngn3 induction and required a functional Notch pathway (24). A working hypothesis was that hypoxia occurred in collagen gel, activating the Notch pathway and inhibiting endocrine differentiation, a mechanism recently established for other cell types (29).

Here, we cultured rat embryonic pancreatic epithelia at the air/liquid interface. Under these conditions, β-cells developed in epithelia cultured with or without mesenchyme. β-Cells never developed from mesenchyme cultured alone (data not shown). We then found that β-cell development in this in vitro model mimicked in vivo development. For example, Ngn3 expression was turned on rapidly then turned off a few days later, in keeping with the reported in vivo pattern (11,30). Also similar to in vivo events, in vitro, Ngn3 induction was followed by the activation of other genes, such as Pax4 and NeuroD, and finally by the induction of markers of terminally differentiated β-cells, such as Pcsk1 and Pcsk2, two endoproteolytic enzymes necessary for proinsulin processing, and Abcc8, a sulfonylurea receptor (31–33). It is also interesting to note that in vitro cell proliferation faithfully replicated in vivo events (17,26,34). Specifically, with mesenchyme, early PDX1⁺ progenitors proliferated at a fast pace. When the cells entered the endocrine pathway and expressed Ngn3, proliferation decreased sharply. Finally, the first β-cells that developed after 3 days of culture did not proliferate. This simple in vitro model in which proliferation and differentiation mimic in vivo events is extremely useful for elucidating the effects of the mesenchyme at each step of β-cell development.

The exact role for the mesenchyme in coordinating progenitor cell proliferation and differentiation is incompletely understood (15,16). Here, we demonstrate that the mesenchyme increases the proliferation of early PDX1⁺ pancreatic progenitors, amplifying and prolonging the formation of Ngn3⁺ endocrine progenitors and ultimately increasing β-cell development, and FGF10 mimics the effect of the mesenchyme. However, FGF10 does not provide complete growth of the explants when compared with mesenchyme. We are currently testing whether in addition to FGF10, the mesenchyme generates additional signals that are important for pancreatic epithelial development.

We previously demonstrated that FGF10 is produced by embryonic pancreatic mesenchymal cells and is required for the proliferation of early pancreatic progenitors (17). We suggested that failure of the progenitor pool to expand in Fgf10 mutants might lead to a reduced number of progenitors available for differentiation, ultimately resulting in an inadequate number of cells to produce a normal pancreas. However, the detrimental effects of FGF10 deletion hindered the interpretation of the role for FGF10 in later stages of development (17). However, in transgenic mice ectopically expressing FGF10 in the pancreatic epithelium, epithelial cell proliferation increased but pancreatic differentiation of all cell types was markedly decreased (18,19). Such a phenotype was also observed when E10 mouse pancreatic epithelium was cultured in matrigel in the presence of FGF10 (35). This inhibition of differentiation was unexpected, and the contradiction between the phenotype of FGF10-deficient mice and transgenic mice with ectopic FGF10 expression or E10 mouse pancreatic epithelium grown with FGF10 was unexplained. For example, the phenotype of FGF10-deficient mice suggested that FGF10 did not inhibit endocrine cell differentiation. In the absence of FGF10, there was no excessive expression of early markers for endocrine cells, such as Isl1 and Ngn3. Moreover, glucagon-expressing cells did not form prematurely or in excess (17). In contrast, endocrine cell differentiation was strongly inhibited in transgenic mice overexpressing FGF10 or in E10 mouse pancreatic epithelium grown with FGF10 (18,19,35). One hypothesis for explaining these apparent contradictions involves the timing of FGF10 expression in transgenic mice. In these mice, FGF10 expression is controlled by the Pdx1 promoter, which is expressed very early in the pancreas, before the mesenchyme condenses around the gut tube. At this early stage, FGF10 treatment of dorsal pancreatic endoderm explants was sufficient to increase the expression of Ptf1a, a transcription factor needed to complete pancreatic specification (36,37). Thus, one possibility is that early FGF10 misexpression in the pancreas modifies the pancreatic cell-differentiation program, a mechanism recently suggested to explain the pancreatic phenotype of mice that overexpress Wnt1 under the control of the Pdx1 promoter (38). These mice have pancreatic agenesis. Comparisons of their phenotype with that of mice deficient in β-catenin suggest that Wnt1 overexpression at early development stages may reflect alternative roles for Wnt signaling, such as specifying the fate of the intestine (39). Difference in the stage of development of the explants could also explain the differences between our study and the one by Miralles et al. (35). We used E13.5 rat pancreas, whereas Miralles et al. used E10 mouse pancreas. Culture conditions could also explain the differences. Although, here, we grew explants on filters at the air/liquid interface, Miralles et al. performed cultures in matrigel that contains basement membrane proteins that are not fully characterized.

Taken together, our data indicate that 1) the pancreatic mesenchyme increases the final β-cell population by activating the proliferation of early PDX1⁺ pancreatic progenitors; and 2) the effects of the mesenchyme on proliferation and differentiation can be mimicked by FGF10. We propose that early PDX1⁺ pancreatic progenitors represent an expansion pool characterized by a significant proliferative potential that should be exploited for expansion with FGF10 and differentiation into β-cells. Interestingly, Baetge and colleagues (40) recently directed differentiation of human embryonic stem cells into insulin-producing cells by mimicking embryonic development. However, as noted by the authors, the protocol must be further improved to produce therapeutic β-cells. The information we provide here should be useful to define the exact timing during which FGF10 should be used to amplify pancreatic progenitors and thus increase the final number of β-cells that will develop.

M.A. received support from the French Ministry for Research and Technology and from the Association Pour la Recherche Médical. This work was supported by Juvenile Diabetes Research Foundation (JDRF Center for Cell Therapy in Europe), INSERM/Fondation pour la Recherche Médicale/JDRF (Grant 4DA03H), the National Institutes of Health Beta Cell Biology Consortium (DK 072495-02), the 6th European Union Framework Program (β-Cell Therapy Integrated Project), INSERM-JDRF Grant (AIP Cellules souches A03139MS), the French National Program of Research on Diabetes, and the Association Française des Diabetiques.