FGFR3 Is a Negative Regulator of the Expansion of Pancreatic Epithelial Cells

NOD/shi, IFNg.NOD (27), IFNg/FGFR3^−/−, and FGFR3^−/− (kind gift from D. Ornitz, Washington University School of Medicine, Saint-Louis, MO [20]) strains were housed in The Scripps Research Institute animal facility, according to the rules and regulations governed and enforced by the institutional animal care and use committee. Animals on the NOD/shi genetic background were killed at E15.5 and E18 for embryonic detection of FGFRs.

The FGFR3 attenuation studies were performed on both NOD and IFNg.NOD mice (10–12 weeks of age) by injecting a neutralizing rat anti-FGFR3 (200 μg/mouse per injection; RD Systems, Minneapolis, MN) three times a week for 2 weeks. Control mice were injected with 200 μg/mouse per injection of purified rat IgGs (Chemicon International, Temecula, CA). Two independent experiments were carried out, and two animals per group received either FGFR3 antibody or vehicle.

Embryos and adult tissues (10–12 weeks of age) were fixed in Bouin’s fixative and processed into serial paraffin sections using routine procedures. Primary antibodies used were rabbit anti–pancreatic duodenal homeobox-1 (PDX-1) (1/500; Chemicon International, Temecula, CA), rabbit anti-FGFR2 and rabbit anti-FGFR3 (1/500; Sigma, Saint-Louis, MO), rabbit anti-FGF2 (1/200; Santa Cruz Biotechnology, Sant Cruz, CA), anti-FGF9 (1/200; Abcam, Cambridge, MA), guinea pig anti-glucagon (1/500; Linco Research, Saint-Charles, MO), and guinea pig anti-insulin (1/1,000; Dako, Carpinteria, CA). The specificity of the anti-FGFR3 antibody was tested by Western blotting of NOD.IFNg pancreatic protein extracts. A single band of ∼120 kDa was detected (data not shown), which is the predicted size for mature FGFR3 (28). FGFR3 was also detected in pancreatic cell lines (Fig. 4A). Appropriate biotinylated secondary antibodies and the Vectastain ABC-elite kit (1/200; Vector Laboratories, Burlingame, CA) were used to detect the primary antibody presence on the sections. Diaminobenzidine was used as a substrate (Sigma).

For double immunofluorescence staining, anti-rabbit–fluorescein isothiocyanate or anti-rabbit–TEXASRED and anti–guinea pig–TEXASRED or anti-rat–fluorescein isothiocyanate (1/200; all Vector Laboratories) were incubated sequentially. Slides were mounted with antifading solution (Molecular Probe, Carlsbad, CA). Confocal images were taken using a Bio-Rad Radiance 2100 Rainbow laser scanning confocal microscope (BioRad, Hercules, CA).

Apoptotic cells were detected with an in situ cell death detection kit (Roche Molecular Biochemicals, Indianapolis, IN) according to the manufacturer’s protocol, and scoring was based on staining and morphological criteria for apoptotic cells. The positive controls consisted of DNaseI-treated sections, and we omitted the enzyme TdT to generate negative controls.

5-bromo-2-deoxyuridine (BrdU) (Sigma) labeling was initiated by intraperitoneal injection (100 μg · g⁻¹ · body wt⁻¹) in adult mice (10–12 weeks of age), the day before death (15–24 h after injection) of the treated animals. In newborn mice, BrdU was injected 2 h before death. BrdU is an analog of deoxythymidine and is incorporated in the DNA of dividing cells during S-phase and can be detected by a specific antibody (29). Tissues were dissected, fixed by immersion in Bouin’s fixative or 10% neutral buffered formalin (Sigma), and embedded in paraffin. Four-micrometer sections of pancreas were stained with rat anti-BrdU (1/100; Accurate Chemicals, Westbury, NY) and revealed by a biotinylated anti-rat IgG, followed by use of the Vectastain ABC-Elite Kit (1/200; Vector Laboratories). Five fields were scored under a microscope on at least two sections from four animals in each group. Results were expressed as mean of positive BrdU nuclei ± SD.

AR42J cells (30) and BTC-Tet cells (kind gift from S. Efrat, Tel Aviv University, Israel [31]) were maintained in Dulbecco’s modified Eagle’s medium containing 10% fetal calf serum (Invitrogen, Carlsbad, CA). For in vitro activation experiments, 10⁵ cells/well were incubated in 12-well plates overnight at 37°C in a 5% CO₂-enriched atmosphere. The following day, they were serum starved overnight and FGF2 or FGF9 (50 ng/ml; Sigma and Abcam, Cambridge, MA, respectively) in combination with 10 μg/ml heparin (Sigma) and blocking FGFR3 antibody (20 μg/ml) were applied on the cells for 10 min. After treatments, the cells were washed with PBS and lysed and processed as already described (32). Membranes were sequentially incubated with rabbit anti-phosphorylated 42/44 ERK (1/1,000; Cell Signaling, Danvers, MA) and anti-42/44 ERK (1/1,000; Santa Cruz) or rabbit anti-phosphorylated AKT (1/1,000; Cell Signaling) and anti-AKT (1/1,000; Santa Cruz). Primary antibodies were detected with specific anti–rabbit-IgG horseradish peroxidase (1/2,000; Vector Laboratories). Proteins were visualized by an enhanced chemiluminescence detection system (Amersham Pharmacia Biotech, Piscataway, NJ).

Pancreata from adult (10–12 weeks of age) FGFR3^−/− (n = 4) and wild-type (n = 3) mice were extracted, fixed in 10% normal buffered formalin, and sectioned (4 μm sections) after paraffin embedding. Five slides at 90-μm intervals were stained with hematoxylin and eosin. Pictures were taken under a Zeiss Axiophot microscope (final magnification ×200), connected with a video camera. Islet contours were traced manually with the polygon tool of an image analysis system (ImageJ 1.31v), and islet areas were measured on three fields for each section. Total islet areas were determined for each mouse and are expressed in arbitrary units ± SD. For the determination of islet distribution within the pancreas, islet areas were sorted by size, and the percentages of small (1–30,000 arbitrary surface units [ASU]), medium (30,001–49,999 ASU), and large (>50,000 ASU) were calculated over the total islets scored (n = 157 for FGFR3^−/− and n = 108 for FGFR3^+/+). Change in islet cell size was evaluated on confocal images by determining the number of nuclei (stained with DAPI) per insulin + glucagon–positive islet area (33). The ratio of islet area to number of nuclei was calculated and is indicated in results as means ±SD for wild-type (n = 3) and FGFR3^−/− (n = 4) mice.

Insulin and glucagon double immunofluorescence detections were performed to determine the percentages of pancreatic insulin- and glucagon-positive areas in five wild-type and five FGFR3^−/− adult mice (8–12 weeks of age). Image J 1.31v software was utilized to determine the hormone-positive and total islet areas.

Venous blood glucose levels were measured with Glucometer Elite XL (Bayer, Pittsburgh, PA). Plasma insulin concentrations were determined with an Ultrasensitive Mouse Insulin ELISA (Alpco Diagnostics, Salem, NH). In both cases, numeric data are presented as means ± SE.

For glucose tolerance tests, overnight-fasted 10- to 14-week-old mice were injected with d-glucose (1 mg/g i.p.). Blood glucose levels were measured from tails at regular intervals during 150 min after glucose injection. For insulin tolerance test, mice were fasted for 3–4 h previous to the insulin injection (0.75 units/kg i.p. Humulin; Lilly, Indianapolis, IN). Blood glucose levels were measured from tails.

Statistical significance was determined using two-tailed, unpaired Student’s t test. Results were considered significant at P ≤ 0.05.

We defined the expression pattern of FGFR3 during pancreatic development in parallel with FGFR2 as a control for the specificity of staining patterns. Immunohistochemical analyses were performed on E15 and E18 embryos, when vigorous growth and differentiation occur in the pancreas.

Pancreata from E15 embryos demonstrated modest epithelial staining for FGFR3 (Fig. 1A). However, at E18 the receptor was more strongly expressed in the emerging islets of the developing pancreas, whereas it was absent in the acinar tissue and ducts (Fig. 1C). Confocal studies showed that the FGFR3-expressing cells were immunopositive for insulin (Figs. 1E–G), and none of the glucagon-positive cells contained FGFR3 antigen (data not shown). Thus, in embryos, FGFR3 expression is first epithelial, then restricted to the β-cell compartment. FGFR2 was found in budding islets (Fig. 1B) at E15 and at E18 (Fig. 1D). Rare duct-associated cells contained FGFR2 immunoreactivity at E15 (Fig. 1B, arrow), and these were not detected later in development. At E18, most of the insulin-producing cells were immunopositive for FGFR2 (Figs. 1H–J), in accordance with published data (9), but the glucagon-positive cells (data not shown) and the surrounding acinar tissue (Fig. 1D) were immunonegative. Remarkably, some cells positive for insulin did not coexpress FGFR2 (Figs. 1H–J, white arrows), suggesting that FGFR2 expression is heterogeneous within the E18 islets.

FGFR3 immunoreactivity was not detectable by immunohistochemistry in the adult pancreas (Fig. 2A). As previously published, strong FGFR2 expression was detected in the islets of the NOD/shi pancreas (Fig. 2G). These data indicate that unlike FGFR2, FGFR3 is present only in the immature embryonic β-cells, and its expression disappears in fully differentiated islets.

For the next series of immunohistochemistry studies, we utilized transgenic IFNg.NOD mice that demonstrate ongoing islet formation in adults, displaying transcription factors and cellular intermediates comparable with those expressed during embryonic pancreas development (34–36).

During pancreatic regeneration, FGFR3 was found in the ductal network (Figs. 2B and C). Interestingly, confocal images showed that clusters of insulin-immunonegative ductal cells displayed an intense FGFR3 signal, particularly at the periphery of the cells, suggesting that the receptor was preferentially localized to the plasma membranes in this ductal cell subpopulation (Figs. 2B–F, arrow heads). In neo-islet structures (Figs. 2D–F, arrows), some insulin-positive cells contained FGFR3. Moreover, no glucagon-positive cells were positive for FGFR3 (data not shown). FGFR2 was found in islet-like structures (Figs. 2H and I) and some ductal cells. Confocal images showed that the FGFR2 receptor was colocalized with insulin but not with glucagon (Figs. 2J–L, arrow heads, and data not shown). However, some cells expressing insulin remained immunonegative for FGFR2 (white arrows), suggesting that as observed in the E18 embryo, β-cells are heterogeneous for the presence of FGFR2 during endocrine cell formation. Thus, the distribution of both receptors in embryonic and adult pancreas is recapitulated during adult regeneration.

We next examined whether the FGFR3-positive cells were actively proliferating. After BrdU injection in mice, coimmunodetection of FGFR3 and BrdU was performed in pancreatic ducts. Qualitative estimation of the number of FGFR3⁺/BrdU⁺ cells revealed that only a small fraction (<10%) of FGFR3-expressing cells was actively proliferating (Figs. 2M–O, white arrows), but the majority of FGFR3-expressing cells were BrdU immunonegative. As only a minor subpopulation of FGFR3-positive cells were dividing in the expanding ducts of the regenerating pancreas, we conclude that FGFR3 expression is more prominent in the growth-arrested cells.

FGF signaling is mediated via intracellular pathways that include the ras/raf/mitogen-activated protein kinase cascade as an initial response to ligand binding (37). We have examined the cellular events occurring following FGF-dependent activation of two distinct pancreatic cell lines, and using a neutralizing antibody, we assessed whether FGFR3 was active in pancreatic cell lines. To validate the approach in these experiments, we first tested AR42J cells, which are a pancreatic carcinoma cell line displaying progenitor cell properties (30), and the BTC-Tet pancreatic β-cell line (31) for expression of FGFR3. Western blots on cellular extracts from these cell lines revealed that both lines expressed the FGFR3 (Fig. 3A). Noticeably, the two cell lines express different isotypes of the receptor as evidenced by the difference in FGFR3 mobility (Fig. 3A).

We next asked whether FGF2, FGF9, and FGF18, all ligands for FGFR3, could activate these pancreatic cell lines through FGFR3. Our experiments demonstrated that after 10 min of FGF exposure, ERK phosphorylation was enhanced (Fig. 3B, P < 0.05) for both cell lines, with a stronger effect in the presence of FGF9, especially in the BTC-tet cell line. Furthermore, we observed that the ERK phosphorylation was attenuated in the presence of the blocking antibody, especially in BTC-tet cells following FGF9 treatment (Fig. 3B), suggesting that FGF9 signals through FGFR3 in this β-cell line. FGF2-induced ERK phosphorylation was attenuated to a lesser extent in the presence of the neutralizing antibody in both cell lines. FGF18 enhanced ERK phosphorylation in the β-cell line, but this effect was not modulated by the addition of the anti-FGFR3. We further analyzed the phosphorylation status of PKB/AKT, since AKT inhibition is related to growth arrest of chondrocytes (38). FGFs were not able to modulate the levels of phosphorylated AKT in these tested cell lines. Thus, in mitotically active β-cells, FGFR3 is expressed and activated by FGF9.

Because FGF2 and FGF9 were capable to activate FGFR3 in vitro, we wondered whether they were expressed in pancreatic tissues. Thus, we performed double immunodetection of these FGFs and insulin. Confocal images revealed that both ligands were absent from the exocrine tissue (Fig. 4). FGF2 expression was restricted to the β-cell compartment in embryos (Figs. 4A–C, E18) as well as in adult tissues (Figs. 4D–I). Interestingly, the FGF9 pattern of expression was broader, since it was expressed in the majority of insulin-positive cells at E18 (Figs. 4J–L) and was also present in other endocrine cells but was not found in pancreatic ducts. In adults, FGF9 expression was restricted to islet non–β-cells (Figs. 4M–N). Interestingly, during adult pancreatic regeneration, the FGF9 pattern resembled that observed at E18 (Figs. 4P–R), suggesting that enhanced pancreatic growth is associated with FGF9 elicitation. Thus, where FGF2 expression was consistently found in β-cells in the pancreas, FGF9 expression appeared to be adjusted according to the proliferative state of the pancreatic epithelium, as FGFR3.

FGFR3 is expressed during pancreas formation and regeneration and is active in pancreatic cell lines. Thus, to examine the role of FGFR3 on the morphogenesis of the pancreas in vivo, we performed histological analysis using the FGFR3^−/− mouse strain (20). We first tested the impact of FGFR3 on newborn pancreata by labeling newborn mice with BrdU, which is incorporated during S-phase, and collecting their pancreata. FGFR3^−/− pancreata displayed increased BrdU-positive nuclei within their pancreatic epithelium (Figs. 5A and B, arrows). Interestingly, FGFR3-deficient newborn mice exhibited a 1.6-fold increase in BrdU-positive epithelial cells within their pancreatic ducts (Table 1, P = 0.034). In mutant and wild-type mice, BrdU⁺ cells within the islets were not numerous enough to perform a statistically meaningful quantitative evaluation of the impact of the absence of FGFR3 on proliferation of newborn pancreatic β-cells. However, we did not observe a noteworthy difference between the two groups. In addition, no difference in apoptosis, as assessed by transferase-mediated dUTP nick-end labeling (TUNEL) staining, was evident in total pancreatic epithelium or in β-cells (data not shown). Consequently, the absence of FGFR3 positively regulates the expansion of pancreatic epithelial cells in newborn mice.

To test whether the absence of FGFR3 affected the pool of terminally differentiated islet cells in the adult pancreas, we performed morphometric analysis. Interestingly, the islet areas from adult FGFR3^−/− mice appeared larger than those from the wild-type islets (Figs. 5C and D, respectively). Quantitative analysis showed that the percentages of islet areas were significantly increased in the absence of FGFR3 (twofold increase, Fig. 5E, P = 0.049). This was likely due to a higher percentage of large islets in the FGFR3^−/− mice rather than a higher overall number of islets per mouse (Fig. 5F, P < 0.001 by χ² test). To test whether the increase in islet size was due to an enhanced capacity of mature islet cells to divide, BrdU incorporation was examined in the adult islets. We found that the mitotic capacity of mature islet cells was low in both strains and not different between wild-type and FGFR3 knockout mice (data not shown). Therefore, the increased islet area could not readily be accounted for by increased mitosis of mature islet cells. In addition, the cellular density was not different in wild-type (n = 3) and FGFR3^−/− (n = 4) mice (122.7 ± 14.8 and 130 ± 7.1 μm², respectively, P = 0.38). Taken together, these results suggest that the greater epithelial expansion early in life observed in the absence of FGFR3 signaling pathway may provide a larger pool of islet precursors, leading to islet hypertrophy.

We hypothesized that, as the absence of FGFR3 affects the size of mature islets, the distribution of endocrine cells within the islets might be disturbed. The percentages of insulin- and glucagon-positive areas in adult islets were determined. We found that the organization of the islets as well as the ratio of α- and β-cells was not affected by the absence of FGFR3 (Table 1). Moreover, blood glucose levels were measured in newborn FGFR3^−/− (n = 3) and wild-type (n = 10) mice, and no significant difference was observed (129 ± 61 and 89 ± 64 mg/ml, respectively, P = 0.35 by t test). In the same way, FGFR3^−/− (n = 10) and wild-type (n = 20) adult mice (9–14 weeks of age) had similar nonfasting glycemia (140 ± 15 and 158 ± 29 mg/ml, respectively, P = 0.07 by t test). Consistently, fasting blood glucose levels were similar in both groups (74.8 ± 13.3, n = 4, and 88 ± 14.8 mg/ml, n = 5, respectively, P = 0.20 by t test). To test whether the absence of FGFR3 modifies the capacity of the animals to clear glucose, glucose tolerance tests were performed. No statistically significant changes between the two groups were observed (online appendix Fig. S1A [available at http://diabetes.diabetesjournals.org]), and, as expected, glucose-induced insulin secretion was similar in both groups (online appendix Fig. S1B). Finally, insulin tolerance tests showed that despite their increased islet mass, FGFR3^−/− mice did not develop insulin intolerance (online appendix Fig. S1C). Thus, the lack of FGFR3 in adults does not alter glucose homeostasis in normal conditions.

During pancreatic regeneration, the processes of epithelial cell expansion and islet differentiation occur simultaneously. We next determined the effect of modulation of FGFR3 during pancreatic regeneration. Therefore, neutralization studies were conducted to determine whether FGFR3 signaling could modulate the expansion of epithelial cells of the adult pancreas in vivo. Rat IgGs or FGFR3 neutralizing antibodies were injected intravenously into IFNg.NOD adult mice, and BrdU was injected the day before death. The pancreata were examined for proportions of BrdU-incorporating cells. Interestingly, an approximate twofold increase in nuclear BrdU labeling was observed in the pancreatic ducts of the regenerating pancreas when mice were treated with FGFR3 neutralizing antibody (P < 0.001, Table 2). To confirm this result, IFNg.NOD mice were bred with FGFR3 null mice (20). The offspring were then intercrossed and the mice labeled with BrdU, and the pancreata from IFNg/FGFR3^−/− and IFNg/FGFR3^+/+ mice were examined for their BrdU nuclear labeling. When FGFR3 was absent, the proportion of BrdU-positive ductal cells increased 1.5-fold compared with IFNg/FGFR3^+/+ mice (P < 0.001, Table 1), confirming that FGFR3 blockade promotes the growth of pancreatic epithelial cells in vivo.

We asked whether the increase of proliferating ductal cells was due to FGFR3 promotion of cell survival in the pancreas. TUNEL analysis was performed to quantitate apoptotic events in ductal cells after FGFR3 neutralizing antibody injections. By evaluating histological sections for stained cells that fit the morphological criteria of apoptosis, no significant differences could be observed in response to the treatment (data not shown), suggesting that, as in newborns, the FGFR3 signaling pathway does not overtly modulate apoptotic cascades in ductal pancreatic epithelium.

We next examined whether anti-FGFR3 treatment affected cellular differentiation in the pancreas. PDX-1 is an early and critical marker of pancreatic cells and is maintained in mature insulin-producing cells (39). We evaluated the expression of PDX-1 in duct-associated islets (Figs. 5G and H). Interestingly, a higher proportion of cells were immunopositive for PDX-1 in islet-like clusters in anti-FGFR3 antibody–treated mice, compared with mice treated with rat IgG (Table 2, P < 0.001). Moreover, cells showing immunoreactivity for PDX-1 but not for insulin within the ducts and the periductal region were more numerous when FGFR3 was attenuated (Figs. 5I and J, arrows, and Table 2), suggesting that the absence of FGFR3 signaling provides an enlarged pool of potential islet progenitors. In parallel, we determined that the ratio of differentiated insulin- and glucagons-positive cells in ducts to total ductal cells (Table 2) was not significantly different between anti-FGFR3–treated and control groups (P = 0.61 for insulin and P = 0.08 for glucagon), suggesting that the rate of terminal differentiation is not affected by FGFR3 blockade.

The results presented here demonstrate the role of FGFR3 signaling in the regulation of the expansion of pancreatic cells. We found that FGFR3 marks a population unique to the embryonic and the adult regenerating pancreas. In the adult regenerating pancreas, only a small fraction of the receptor-positive cells were actually dividing, as determined by BrdU incorporation. Moreover, FGFR3 was still present in a subpopulation of insulin-positive cells in islet clusters, as in E18 embryos. We hypothesize that these cells are likely immature β-cells, since FGFR3 expression is shut down in adult islets. Importantly, FGFR3 attenuation by either genetic deletion or immune blockade led to a significant increase in epithelial cell expansion in pancreatic ducts. Therefore, FGFR3, when present, negatively controls the proliferation of this population. The fact that pancreatic ductal cells serve as progenitors during islet formation is still controversial but remains a valid hypothesis (40,41). Thus, according to the FGFR3 expression pattern in the pancreatic epithelium, it is possible that FGFR3 activity is needed to control cell division before the cells commit to the endocrine fate (see below). This interpretation is consistent with a previous study (42) showing that a decline in the expression of FGFR3 was associated with progress toward the acquisition of a differentiated state in the colon-derived CaCo2 cells.

To determine how FGFR3 activity impacts islet formation, we studied terminally differentiated cells in FGFR3 null mice and in anti-FGFR3–treated mice. In both cases, the proportion of insulin- or glucagon-expressing cells remained unchanged compared with wild-type or control mice. However, we demonstrate that in adult FGFR3^−/− mutants, enlargement of mature pancreatic islets is observed. However, we only observed enhancement of mitosis in the ducts of neonatal FGFR3-deficient pancreata and not in adult islets. Moreover, apoptosis was not affected in the mutant newborns. Therefore, we speculate that the increased mass of mature islets is a consequence of the expansion cells in the pancreatic epithelium in newborn mutant mice. Similarly, in the adult regenerating pancreas, the FGFR3 receptor was expressed prominently in nonmitotically active ductal cells. PDX-1–positive/insulin-negative cells, which reflect β-cell progenitors (rev. in 43), were more numerous following FGFR3 antibody treatment. Taken together, these results demonstrate that FGFR3 signaling inhibits expansion and promotes the accumulation of immature cells.

The absence of FGFR3 does not affect glucose homeostasis or glucose clearance following glucose challenge. In addition, despite enlarged islets, FGFR3^−/− adults showed no sign of insulin intolerance. Thus, FGFR3 does not seem to be involved in islet function in mature β-cells. This is in agreement with the fact that FGFR3 is not detected in mature adult islets. However, it would be interesting to test the possibility of an effect of FGFR3 absence in spontaneous or induced onset of diabetes.

FGFR3 is generally considered an inhibitor of cell proliferation. Indeed, FGFR3 is involved in chondrocyte maturation (44), pillar (26), and retinal ganglion cell growth arrest and differentiation (45). Moreover, the observation that FGFR3 ligation inhibits pancreatic cell expansion appears unique among the FGFRs. Indeed, most of the FGF members (ligand and receptors) promote growth and branching morphogenesis of the pancreas. A critical factor in this process is FGF10, the regulation of which is crucial for proper pancreas formation (46). Several FGF ligands could potentially act through FGFR3. We show in this study that FGFR3 signal transduction is weakly responsive to FGF2. FGF2 can bind to other FGF receptors, e.g., FGFR2, which is expressed in β-cells. Moreover, FGF2 is not the sole ligand of FGFR3 (13). FGF18 has been shown to be a potential ligand of FGFR3 (47), but injection of FGF18 in our transgenic mouse model of pancreatic regeneration did not affect epithelial cell proliferation in the pancreas (S.A.-D., unpublished observations). Furthermore, FGF18 did not activate FGFR3 in the pancreatic cell lines we studied. Interestingly, FGF9 has been previously shown to be a specific ligand of FGFR3 (48–50). We found that pancreatic cells, especially transformed β-cells, are highly reactive to FGF9 treatment, and we show that this effect is inhibited when FGFR3 is blocked. Moreover, FGF9 and FGFR3 expression are concomitant in the developing pancreas, and their patterns of expression are recast in adults when pancreatic regeneration occurs. This supports the notion that ligation of FGF9/FGFR3 is possibly involved in the regulation of the expansion of islet cell precursors.

We demonstrated that FGFs activate ERKs and that FGF9-induced ERK phosphorylation is modulated by FGFR3 in pancreatic cell lines. Interestingly, ERK activation was downregulated during the in vitro formation of ductular structures from pancreatic islets, paralleling the development of pancreatic cancer (51). Conversely, ERK activation was found to be positively associated with an increase in the nuclear translocation of NeuroD1 (52), a crucial transcription factor involved in the acquisition of terminal endocrine phenotype (53). Taken together, these results suggest that ERK activation, possibly through FGF9/FGFR3 induction, is associated with the differentiation of endocrine pancreatic cells.

The increase in cell numbers observed when FGFR3 is inhibited could result from the inhibition of cell death. Indeed, FGFR3 has diverse effects on apoptosis. For example, during bone formation, FGFR3 over-activation can lead to increased apoptosis (22,54). Conversely, in multiple myeloma cells, FGFR3 promotes cell survival (55). We performed TUNEL assays on anti-FGFR3–treated or FGFR3^−/− pancreata and rat IgG-treated or wild-type pancreata and did not observe differences in the apoptotic rate of the pancreatic cells. We also performed TUNEL assays on newborn pancreata from wild-type and FGFR3^−/− mice in combination with insulin detection and found that newborn islets seldom displayed apoptotic cells in both strains (data not shown). At the molecular level, PKB/AKT signaling, which impacts cell survival, was not activated by FGFs, consistent with the observation that apoptosis was not modulated by FGFR3 in pancreatic cells. Thus, the enhancement of epithelial cell number does not seem to result from a prosurvival effect in the absence of FGFR3 signal but more likely from a direct modulation of the dividing capacities of the pancreatic cells.

In conclusion, our work demonstrates that FGFR3 expression is connected to pancreas organogenesis. Moreover, signaling through the receptor limits the expansion of pancreatic epithelial cells. FGF9 is a good candidate for the induction of FGFR3 in the pancreas. In the absence of the receptor, the expanded pancreatic cell proliferation results in increased islet mass. Our work supports the importance of FGF family members in regulating key sequences of events involved in the formation of pancreatic islet cells by establishing a critical balance between the induction and inhibition of growth.

This work was supported by a grant from the National Institutes of Health (NIH) (DK060746-02). S.A.D. was supported by the International Fellowships for Beginning Investigators (ACS/05/015) fellowship of the International Union Against Cancer, and Y.Q.Z. received an NIH training grant (T32HL0795).

We are grateful to the Sarvetnick lab’s members for their valuable comments on the manuscript.