Inhibition of Activin Signaling Induces Pancreatic Epithelial Cell Expansion and Diminishes Terminal Differentiation of Pancreatic β-Cells

The IFN-γ transgenic mice used were on the nonobese diabetic (NOD) background. All mice were maintained in a specific pathogen-free facility at The Scripps Research Institute.

Rabbit polyclonal antibody against activin βA monomers and dimers were prepared as described (18). Rabbit anti-cyclic inhibin βB antibody was a kind gift from Dr. Wylie Vale (Salk Institute, La Jolla, CA). This antibody specifically recognizes inhibin βB (6,19). Rabbit anti-human pancreas duodenum homeobox (PDX)-1 antibody was a generous gift from Dr. Chris Wright (Vanderbilt University Medical School, Nashville, TN) (20). Goat anti-activin type I, type IB, type II, and type IIB receptor, anti-Smad2/3, rabbit anti-phosphorylated Smad2/3, and rabbit anti–proliferating cell nuclear antigen (PCNA) antibodies were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). Goat anti-human Cripto-1 and anti-human follistatin antibodies were purchased from R&D Systems (Minneapolis, MN). Monoclonal anti-pan cytokeratin antibody was purchased from Sigma-Aldrich (St. Louis, MO). Rat anti-bromodeoxyuridine (BrdU) antibody was purchased from Accurate Chemical (San Diego, CA).

Ductal preparations were carried out as described (21). For flow cytometry, at least 2.5 × 10⁵ cells/tube were used. Fluorescence-activated cell sorter detection was performed using the fluorescein isothiocyanate (FITC) BrdU flow kit (BD Pharmingen, San Diego, CA). The epithelial cells were stained with CD49f-PE, and 7ADD was used to determine the stage of the cell cycle. Samples were analyzed on a FACScan flow cytometer using Cell Quest software (Becton Dickinson, Heidelberg, Germany).

Immunohistochemistry stainings were carried out as described (18). For double and triple immunofluorescence staining, FITC-conjugated anti–guinea pig, anti-rabbit, or anti-mouse; Cy3-conjugated anti-guinea pig; or Cy5-conjugated anti-rabbit antibodies (Jackson ImmunoResearch Laboratories, West Grove, PA) and Alexa Fluor 488 donkey anti-goat or anti-rat FITC-conjugated (Molecular Probes, Eugene, OR) antibodies were used. Sections were observed using a BioRad MRC 1024 scanning confocal microscope (Richmond, CA) mounted on a Zeiss Axiovert TV-100 with 40× objectives. Negative controls for immunostaining were performed without the primary antibodies. For determination of apoptosis, transferase-mediated dUTP-biotin nick-end labeling (TUNEL) staining was performed using the Roche In Situ Cell Death Detection POD kit (Indianapolis, IN).

Recombinant human (rh)-follistatin was purified from the culture supernatant of human-follistatin cDNA bearing CHO cells as described (10). To analyze the efficacy of exogenous follistatin in the regenerating pancreas, purified rh-follistatin was prepared in a carrier protein-free buffer (0.05 mol/l sodium acetate, pH 5.0) and dissolved in 10% serum from a NOD mouse for a total volume of 150 μl and was injected through the tail vein every other day for 2 weeks. The concentration of rh-follistatin was 0.3 μg/mouse. Control mice were injected with 150 μl of 10% NOD mouse serum in saline.

To examine the regeneration and differentiation of duct epithelial cells, mice were injected intraperitoneally with 100 μg/g body wt BrdU twice a day for 3 days. The mice were killed 12 h after the last injection. For flow cytometric analysis, the mice were fed BrdU (80 μg/ml) in drinking water for 24 h. For colocalization studies of BrdU and activin type IIB receptor or βA subunit, BrdU (100 μg/g) was injected 12 h before killing.

Plasma glucose was measured with a blood glucose meter, Glucometer Elite X (Bayer, Elkhart, IN). Insulin concentration was detected using the Mercodia Ultrasensitive Mouse Insulin ELISA (ALPCO Diagnostics, Windham, NH).

To quantify BrdU-positive cells, the number of BrdU-positive cells in the duct epithelial cells was counted. The labeled cells were evaluated as the percentage of BrdU-labeled cells out of the total number of duct epithelial cells. To quantify the number of BrdU and insulin double-positive cells, the immunolabeled cells were scored on a 400× microscope field. At least 10 fields were counted per section, and three sections were counted per mouse. In total, five mice were examined for each test group. The ratio of BrdU-insulin double-positive cells to the total insulin-positive cells in small islet-like structures and scattered insulin-positive cells was calculated. To quantify the number of islet-like cell clusters (ICCs) (less than eight cells across), ICCs were counted in 20 fields per section and three sections were counted per mouse. Results were expressed as mean ± SE. The significance of differences between two groups was determined using the unpaired Student’s t test.

To investigate whether the expression of activins is induced during pancreatic islet differentiation, immunohistochemical studies were conducted on regenerating and control syngeneic NOD mice utilizing antibodies to activin βA and βB subunits. As shown in Fig. 1A, the immunoreactivity of the βA subunit was detected in centroacinar cells, small duct cells, and within cells of the periphery of islets, and the expression of the βB subunit was scant in the small ducts and islets in the pancreas of NOD mice. In the regenerating pancreas, both βA and βB subunit antibodies demonstrated strong immunoreactivity in duct epithelial cells beginning at 1 month of age, suggesting that during regeneration both subunits of activin are significantly upregulated. To determine the cellular specificity of activin expression, double or triple staining of activins, insulin, and glucagon was performed. As shown in Fig. 1B, the βA subunit colocalized with glucagon, but not with insulin in NOD mice and only rarely with insulin in the regenerating pancreas (Fig. 1C). The βB subunit colocalized with insulin and glucagon in NOD mice. In the regenerating pancreas, βB was extensively expressed in duct epithelial cells and colocalized with insulin and glucagon (Table 1).

Next, we investigated the expression of activin receptors in the regenerating pancreas. In the regenerating pancreas, the expression of both activin type II and type IIB receptors was extensive in duct epithelial cells, whereas they were only weakly expressed in NOD mice. Similarly, type IB receptors were also in a subset of duct epithelial cells (Figs. 2 and Table 2). Furthermore, activin type IIB receptors colocalized with most of the glucagon-producing cells but rarely with insulin-containing cells. However, a subset of the type II receptor–positive cells colocalized with both insulin and glucagon (Table 1). We also investigated the expression of type IIA and type IIB receptors during pancreatic ontogeny using E18 embryos. The expression of type IIA receptors was in the developing pancreas around the duct-like cords, and a subset of these cells coexpressed either cytokeratin or PDX-1, a critical pancreatic islet progenitor transcription factor (19,20) (Fig. 2D). Similar results were also observed with the expression of type IIB receptors (data not shown). These results further support the notion that activin receptor expression reflects islet differentiation during both pancreas development and pancreatic regeneration. We also investigated the coexpression of cytokeratin with activins or activin receptors. Our results showed that some activin receptor-expressing cells coexpressed cytokeratin, indicating they were epithelial cells; this was also observed for the βA and βB subunits of activins (Figs. 3A and Table 2).

If activin inhibition promotes growth, then cells expressing activins may exhibit low levels of expansion. To determine whether activins and/or activin receptors are expressed in mitotically active cells, we investigated the colocalization of activins and their receptors in tissues from BrdU-labeled mice. Of 1,000 activin A–positive cells, the percentage of activin A/BrdU double-positive cells was 3.2%. Furthermore, the type IIB/BrdU double-positive cells comprised 4% of the total type II–positive cells (Fig. 3B). These results demonstrate that neither activin- nor activin receptor–bearing cells are rapidly proliferating nor do they correspond to the main proliferating compartment during pancreatic regeneration, which exhibits ∼26% BrdU incorporation under these labeling conditions.

To investigate the downstream signaling events of activins during pancreatic islet differentiation, we assessed the cellular localization of Smad2/3 and phosphorylated Smad (P-Smad)2/3 in the pancreas of NOD and IFN-γ transgenic mice. Our studies revealed that Smad2/3 was expressed in centroacinar cells and cells in the small ducts with expression patterns paralleling that of activins. Smad2/3 immunoreactivity was located solely in the cytoplasm in NOD and BALB/C mice, but was largely found within the nuclei in the regenerating pancreas (Fig. 3C). The immunoreactivity of P-Smad2/3 was also found inside the nucleus in a proportion of the cells in large ducts, demonstrating that these molecules were in their activated state (Fig. 3D). In addition, some of the P-Smad2/3–positive cells colocalized with insulin (Fig. 3E). As Smad2/3 is a common pathway for both activins and TGF-β, we also examined the expression of TGF-β1. However, there was no significant difference between NOD mice and IFN-γ transgenic mice in TGF-β1 expression (data not shown).

Because activins and their receptors promote the differentiation processes, we investigated whether the natural activin inhibitors were present during pancreatic differentiation. In NOD mice, follistatin expression was found inside the islets, consistent with previous reports (6,22). In the regenerating pancreas, follistatin immunoreactivity was in the islet-like clusters and in a subpopulation of the ductal cells in large ducts (Fig. 4A). Cripto immunoreactivity was found very modestly in some of the blood vessels and duct cells in the NOD pancreas. However, in the regenerating pancreas, intense Cripto immunoreactivity was found in numerous cells within smaller ducts, similar to the expression pattern of the βA subunit of activin (Figs. 4B–C). Only very occasional staining of Cripto was found in enlarged ducts. These results cumulatively suggest that activin signaling may be repressed during the aggressive growth that occurs in the transition from small to large ducts. To determine the relationship between Cripto and endocrine differentiation, we performed colocalization studies of Cripto and insulin. Our results showed that during pancreatic regeneration, insulin-positive cells are adjacent to Cripto-producing cells. Only in very rare cells were insulin and Cripto colocalized. However, the Cripto immunoreactivity was diminished (Fig. 4D). Of particular importance is that the expression of follistatin and Cripto does not increase during the destructive inflammation in NOD mice, suggesting that their expression does not directly result from inflammation. To assess whether this relationship was also relevant to pancreatic ontogeny, the expression of Cripto in the fetal pancreas was investigated. As shown in Fig. 4E, the expression of Cripto was in the duct-like cords in both E15.5 and E18.0 when the growth of the pancreas is vigorous, supporting the idea that activin inhibition occurs in epithelial progenitor cells that undergo rapid expansion.

To determine whether activins are involved in epithelial cell expansion, recombinant follistatin was administrated intravenously every other day for 2 weeks into 6-week-old transgenic mice and syngeneic NOD mice. BrdU was injected intraperitoneally and visualized by immunohistochemistry. We have found that a 2-week treatment interval is sufficient to observe changes in the expansion and cellular composition of the regeneration response (21). In addition, the bioactivity of the recombinant follistatin used in this experiment has been confirmed by previous studies (23,24). The results of this analysis showed strongly increased BrdU-positive cells in the ducts of transgenic follistatin-injected mice compared with those of the saline-injected mice (Fig. 5A). In saline-treated mice, an average of 26.9% cells in the duct epithelium were BrdU positive (n = 5) compared with 47.2% BrdU-positive cells in the ducts of follistatin-treated mice (n = 8, P = 0.014). Importantly, syngeneic NOD mice that received the same amount of follistatin as the transgenic mice did not exhibit significantly increased BrdU incorporation compared with control saline-injected mice (n = 3, data not shown), indicating that activins regulate growth processes in a subset of cells that are absent from the normal pancreas. This suggests that follistatin promotes epithelial cell expansion in pancreatic islet progenitor cells. We also investigated the effects of follistatin on the growth of isolated duct cells treated with exogenous activin A. Our results showed that the growth inhibition by activin A is reversed by administration of recombinant follistatin (data not shown), suggesting that the recombinant follistatin used in our experiments acts by inhibiting activins.

To confirm our results, three-color flow cytometry studies were conducted. We investigated the percentage of BrdU-positive epithelial cells that were in the S-phase of the cell cycle. In the follistatin-treated IFN-γ transgenic mice, the S-phase BrdU-positive cells comprised 16.7% of the gated epithelial cells, whereas in the saline-treated mice, S-phase BrdU-positive cells accounted for 9.1% of the gated epithelial cells (n = 3, P = 0.031) (Fig. 5B). Furthermore, epithelial cells in the G1 phase were decreased (25.7% in follistatin treated vs. 38.4% in saline treated, P = 0.040), and cells in the G2 phase were increased (34.5% in follistatin treated vs. 22.2% in saline treated, P = 0.037) in follistatin-treated mice. These results suggest that follistatin increases the number of epithelial cells undergoing mitosis.

Interestingly, activins have been reported to induce apoptosis in a variety of tissues and cells (23–25). We asked whether the increased BrdU incorporation after follistatin treatment could reflect changes in cell survival. Therefore cells that were TUNEL positive and fit the morphologic criteria for apoptosis (condensed or fragmented nuclei) were quantitated (Fig. 5C). Our results showed that the percentage of apoptotic cells in the duct epithelial cells in the IFN-γ saline-treated mice was 1.69 ± 0.7%. However, in the follistatin-treated IFN-γ mice, the percentage of apoptotic cells in the duct epithelial cells was reduced to 0.68 ± 0.2% (P = 0.0015). This result indicates enhanced epithelial cell survival in the follistatin-treated mice, suggesting that activins regulate the turnover of epithelial cells during islet development. Inhibition of activin binding supports the survival of islet progenitor cells.

To detect whether activins affect insulin secretion, we examined the plasma insulin concentration in follistatin-treated mice. The insulin concentration was 8.75 ± 0.8 ng/ml in the saline-treated mice (n = 5) and 6.65 ± 1.0 ng/ml in the follistatin-treated mice (n = 7, P = 0.026). In NOD mice, the insulin concentration was 9.5 ± 1.0 ng/ml and 8.9 ± 0.5 in the follistatin- and saline-treated mice, respectively (P > 0.05), (Fig. 6A). Our results show that in the follistatin-treated mice, insulin levels were significantly reduced. This result suggests that follistatin negatively regulates the differentiation of β-cells or affects their function. Therefore, we counted ICCs in the follistatin-treated and saline-treated mice. The number of ICCs is significantly decreased in the follistatin-treated mice. The average ICC number per section in follistatin-treated mice was 16.6 ± 3.1, and the average ICC number per section in saline-treated mice was 26.3 ± 9.4 (P = 0.0105) (Fig. 6B).

To determine whether this reflected a paucity of newly forming islet cells, we counted BrdU/insulin double-positive cells to determine whether follistatin treatment influences their proportion within the BrdU-positive population. We found that in the follistatin-treated group, although the BrdU-positive cells were increased, the BrdU/insulin double-positive cells were significantly decreased. In the saline-treated group, the BrdU/insulin double-positive cells were 4.7 ± 1.7% of the total insulin-positive cells from areas of overt islet neogenesis, whereas in the follistatin–treated mice, the BrdU/insulin double-positive cells were 1.8 ± 1.2% (P = 0.027) of the total insulin positive cells from areas of overt islet neogenesis (Fig. 6C). The decreased number of BrdU/insulin-positive cells might result from the dilution of the BrdU label in dividing cells. To exclude this possibility, we also counted the PCNA/insulin double-positive cells. The results were similar to those obtained in the BrdU analysis. In the pancreas of saline-treated mice, the PCNA/insulin double-positive cells were 4.5 ± 1.3% of the total insulin-positive cells, but in the follistatin-treated regenerating pancreas, the PCNA/insulin double-positive cells were 2.3 ± 1.1% (P = 0.021) of the total insulin positive cells. Together, our results indicate that by blocking the actions of activins, β-cell differentiation is decreased, demonstrating that activin signaling is required for the terminal differentiation of pancreatic β-cells.

In the present study, we report that follistatin treatment promoted the expansion of duct epithelial cells but inhibited their terminal differentiation into endocrine cells in the regenerating pancreas. We suggest that the balance between the activation and inhibition of activin pathways plays a critical role during pancreatic islet cell differentiation.

IFN-γ transgenic mice demonstrate dramatic proliferation of epithelial cells and continuous neogenesis of islet and/or ICCs from the progenitor cells in the duct in a process similar to embryonic islet morphogenesis (15–17). Activin signaling is involved in the induction of PDX-1 (26). Exogenous activin treatment increases the proportion of insulin cells in the developing chick pancreas (27). Activin A also induces differentiation of human fetal pancreatic endocrine cells (28). Interestingly, activins are upregulated in duct cells following partial pancreatectomy and streptozotocin injection, suggesting that activins might be involved in the initiation of β-cell neogenesis following distinct stimuli in adulthood (18,29). Activins repress the expansion of pancreatic cancer cells, paralleling our observation (30) in the regenerating pancreas. Genetic deletion of the activin type II receptor has been reported (31) in some pancreatic cancers. Therefore, in adults activins specifically inhibit the expansion of immature pancreatic cells found within the regenerating pancreas and in pancreatic cancer. Such parallels support the idea that the extensive growth we observed in the regenerating pancreas approaches the early events of these devastating cancers. However, pancreatic growth during regeneration is distinct from pancreatic cancer by the propensity of these cells to fully mature in vivo. This terminal differentiation appears to be positively regulated by activins. Along these lines, activins induce pancreatic tumor cell differentiation into insulin-producing cells through the Smad2 and P38 mitogenic-activated protein kinase pathway (32–34). In this respect, these pancreatic tumor cells act like progenitor cells and may have similar characteristics to the progenitor cells in IFN-γ transgenic mice. One mechanism explaining the inhibition of activins on pancreatic cancer and regeneration is through activin-induced turnover of these primitive cells. Our studies demonstrated that follistatin treatment enhanced epithelial cell survival. Therefore, inhibition of activin signaling may favor the survival of progenitor populations that would ordinarily be lost to apoptosis during tissue remodeling.

Distemporaneous responses to activins during midgestation would have deleterious effects on pancreatic progenitor expansion. Similarly, activin responses during normal adulthood may limit the ability of the tissue to respond to injury. Interestingly, we did not find any significant changes in either BrdU incorporation or β-cell neogenesis in NOD mice that received follistatin. These data suggest that endogenous activin-responsive cell populations are missing from the normal adult pancreas due to remodeling or are inhibited by Cripto expression. Activins have been reported (25) to affect the phosphorylation state of the RB protein, a key regulator of the cell cycle at the late G1 phase to inhibit cell proliferation. In addition, growth inhibition has been shown (35) to promote differentiation of insulin-producing tissue from embryonic stem cells. Thus, the proliferation, apoptosis, and differentiation of pancreatic cells are maintained in a very tight balance within the regenerating and normal pancreas.

Recent studies on Cripto have generated important insights on the balance between cell proliferation and its inhibition governed by activins. Cripto has been shown (36) to transform mammary epithelial cells. The increased expression of Cripto is also found in many solid cancers, including pancreatic cancer (37). Cripto is also associated with the formation of duct-like structures in the human pancreas (38). In the regenerating pancreas, Cripto is expressed in cells within small ducts, in the same area as activins. Indeed, the presence of activin inhibitors, particularly Cripto, in ductal progenitors suggests that activin inhibition is required for the expansion of the progenitor cell pool during islet regeneration. We hypothesize that in some cells of small ducts, inhibition of activin signaling by Cripto promotes cellular expansion. However, in the subset where nuclear Smad2/3 staining is found, the activin signal may balance the increased growth induced by the Cripto signal. In the large ducts, however, Cripto expression is restricted, therefore facilitating the differentiation of β-cells through exposure to activins. Of particular interest is the expression of Cripto during pancreatic ontogeny, suggesting that Cripto may promote the expansion of duct epithelial progenitor cells, inhibiting their early differentiation into hormone-secreting cells.

Follistatin binds to activins and blocks their functions (10,11). Interestingly, follistatin is one of the mesenchymal factors required for the development of exocrine tissue while exerting a repressive role on the differentiation of the endocrine cells (22). Our data are consistent with these results because we show that activin positively regulates the differentiation of β-cells in the regenerating pancreas. After administration of exogenous follistatin, we found that in a large population of duct cells the expression of Smad2/3 was localized in the cytoplasm, suggesting that the increased proliferation in the follistatin-treated pancreas is enacted by blocking the activin/Smads signaling pathway. Taken together, our data provide evidence that activin inhibition is required for optimal progenitor cell expansion in the pancreas.

It is possible that follistatin exerts its actions by a mechanism other than the inhibition of activins. It should be noted that in addition to activins, follistatin also binds to bone morphogenetic proteins, albeit with significantly lower affinity compared with activins (K_d 23 vs. 0.5–0.7 nmol/l) (26,39). We therefore cannot exclude the possibility that follistatin exerts its functions by binding to endogenous bone morphogenetic proteins in the regenerating pancreas. In addition, a previous report (40) has shown that follistatin enhances the effect of epidermal growth factor on DNA synthesis in a mouse hepatic cell line, but not in cells expressing dominant-negative activin receptors.

Interestingly, activins have been reported (41–43) to stimulate and modulate insulin secretion in both rat and human islets, suggesting that the activin signal is not only crucial to the developing pancreas but might also be involved in the maintenance of the normal function of pancreatic β-cells. Our results demonstrate that activins and their counterregulation clearly play a pivotal role during pancreatic progenitor expansion and their subsequent differentiation. These findings could provide a potential therapeutic axis to promote β-cell neogenesis and have important implications for the treatment of diabetes.

This work was supported by National Institutes of Health Grants from the National Institute of Diabetes and Digestive and Kidney Diseases, DK55230 and DK60746.

We thank Dr. Wylie Vale from the Salk Institute for his generous gift of rabbit anti-cyclic inhibin βB antibody and Dr. Christopher V.E. Wright from Vanderbilt University Medical School for his generous gift of rabbit anti-human PDX-1 antibody. We also thank all of the members of the Sarvetnick Lab for critically reviewing this manuscript.

This is The Scripps Research Institute manuscript number 16099-IMM.