Glucagon-Like Peptide 1 Increases β-Cell Regeneration by Promoting α- to β-Cell Transdifferentiation

Diabetes is caused by defective control of blood glucose levels resulting from an absolute or relative deficiency of functional pancreatic β-cells. Type 1 diabetes is characterized by absolute deficiency of insulin due to autoimmune-mediated destruction of pancreatic β-cells, whereas type 2 diabetes is characterized by relative deficiency of insulin due to insufficient insulin secretion to compensate for insulin resistance. Thus, strategies to manage diabetes by restoring functional β-cells are under investigation.

Glucagon-like peptide 1 (GLP-1), which is secreted from intestinal l-cells in response to nutrient ingestion, is known to have important physiological roles: it potentiates glucose-stimulated insulin secretion, induces insulin gene transcription and insulin biosynthesis, enhances β-cell proliferation, and inhibits β-cell apoptosis (1,2). In addition, GLP-1 has effects on the regeneration, differentiation, and neogenesis of pancreatic β-cells (3–6). In 70% pancreatectomized mice, the β-cell mass was significantly lower in GLP-1 receptor knockout (KO) mice compared with wild-type mice, suggesting a potential role of GLP-1 in regulation of the β-cell mass (7).

In the pancreatic islets, α-cells primarily produce the hormone glucagon. However, recent observations indicate that α-cells can also produce GLP-1 (8–10). Another important role of α-cells is their ability to transdifferentiate into β-cells under conditions of extreme damage to β-cells (11). As well, increased physiological demand for insulin can result in increased α-cell proliferation (12). These results suggest that α-cells can be a source of newly generated insulin-producing cells and that GLP-1 may act as a stimulus through autocrine signaling (13).

Fibroblast growth factor 21 (FGF21) is a circulating protein that is highly expressed in the liver, and FGF21 protein expression is also detected in both α- and β-cells in the pancreas (14). Constant infusion of FGF21 for 8 weeks in db/db mice results in a higher number of islets per pancreatic section and a higher number of insulin-positive cells per islet without β-cell proliferation compared with control mice (15). As well, glucagon receptor KO mice display α-cell hyperplasia, and there is a correlation with the increase of plasma FGF21 in these mice (14). In addition, a GLP-1 analog increases FGF21 expression and FGF21 activity in insulin-resistant mice (16). Therefore, we hypothesized that GLP-1 might increase FGF21 production, and FGF21 might play a role in the generation of new β-cells from α-cells by GLP-1. In the current study, we sought to determine the effects of GLP-1 on α-cells to generate new β-cells and to examine the mechanisms involved.

RIP-CreER mice (a gift of Dr. D. Melton), Glucagon-rtTA mice (a gift of Dr. P.L. Herrera), Tet-O-Cre mice (The Jackson Laboratory, Bar Harbor, ME), ROSA26 (R26)–yellow fluorescent protein (YFP) mice (The Jackson Laboratory), and FGF21 KO mice (provided by Eli Lilly and Company) were used. We generated two types of mice: RIP-CreER;R26-YFP mice for tracing the β-cell lineage and Glucagon-rtTA;Tet-O-Cre;R26-YFP mice for tracing the α-cell lineage. These mice were maintained at the facility at Gachon University under a 12-h light/12-h dark photoperiod. Animals were fed ad libitum on a standard rodent diet. All animal experiments were carried out under a protocol approved by the Institutional Animal Care and Use Committee at Lee Gil Ya Cancer and Diabetes Institute, Gachon University.

Recombinant adenovirus expressing GLP-1 (rAd-GLP-1) or rAd producing β-galactosidase (rAd-βgal), as a control, was produced as previously described (17). The recombinant adenoviruses were produced and amplified in a human embryonic kidney cell line (HEK-293). After purification of virus by CsCl-gradient ultracentrifugation, viral titer was determined by 50% tissue culture infectious dose.

Four- week-old male RIP-CreER;R26-YFP mice were injected with 4-hydroxytamoxifen (1 mg/mouse) (Sigma-Aldrich, St. Louis, MO) daily for 5 days every other week, which was repeated four times. Doxycycline (1.5 mg/mL) (Sigma-Aldrich) was added to the drinking water of 4-week-old male Glucagon-rtTA;Tet-O-Cre;R26-YFP mice for 2 weeks. After doxycycline removal, mice were kept for 14 days without treatment before streptozotocin (STZ) administration. β-Cell destruction was achieved in RIP-CreER;R26-YFP mice, Glucagon-rtTA;Tet-O-Cre;R26-YFP mice, and FGF21 KO mice by i.p. injection of STZ (150 mg/kg) (Sigma-Aldrich), a β-cell–specific toxin. The mice were monitored for the development of hyperglycemia using a glucometer. STZ-induced diabetic mice (blood glucose levels >300 mg/dL for 3 consecutive days) were injected via the tail vein with rAd-GLP-1 or rAd-βgal (3 × 10⁹ plaque-forming units). After viral injection, RIP-CreER;R26-YFP mice were injected with BrdU (100 mg/kg) (Sigma-Aldrich) every day for 4 weeks.

Mice were not fed for 4 h, and a glucose solution (2 g/kg body weight) was injected i.p. Blood glucose levels were measured at 0, 30, 60, 90, and 120 min after glucose injection.

RIP-CreER;R26-YFP mice, Glucagon-rtTA;Tet-O-Cre;R26-YFP mice, and FGF21 KO mice were sacrificed at 4 weeks after rAd-GLP-1 or rAd-βgal treatment. Pancreata were removed, fixed in 10% formalin, and embedded in paraffin. More than 500 serial sections (4-μm thick) were prepared from each pancreas, and every 25th section was used for immunohistochemical analysis. The tissue sections were boiled (100°C for 10 min, 10 mmol/L sodium citrate, pH 6.0) for antigen retrieval and blocked with blocking solution (Dako, Carpinteria, CA). The sections were then incubated with primary antibody solution: guinea pig anti-insulin (1:100) (Dako), rabbit anti-insulin (1:100) (Santa Cruz Biotechnology, Santa Cruz, CA), rabbit anti-glucagon (1:100) (Dako), mouse anti-glucagon (1:100) (Sigma-Aldrich), rabbit anti-GFP, which cross-reacts with YFP (1:50) (Abcam, Cambridge, U.K.) (1:100) (Invitrogen, Vancouver, BC, Canada), or mouse anti-BrdU (1:50) (Dako). FITC-conjugated goat anti–guinea pig IgG (1:200) (Santa Cruz Biotechnology), Texas Red (TR)–conjugated goat anti-rabbit IgG (1:200) (Santa Cruz Biotechnology), or Alexa Fluor 633–conjugated goat anti–guinea pig IgG (1:200) (Thermo Fisher Scientific, Rockford, IL) were used as secondary antibodies. Fluorescence was imaged using a laser scanning confocal fluorescent microscope (LSM 700; Carl Zeiss MicroImaging, Jena, Germany), and colocalization was analyzed by the ZEN 2009 Analysis Program.

Pancreatic islets were isolated from 4–6-week-old male FGF21 KO or C57BL/6 mice as described previously (18). Intact islets were dissociated at 37°C in Accutase (Millipore, Bilerica, MA), given STZ (1 mmol/L) for 15 h, washed, and then cultured with exendin-4 (10 nmol/L) (Sigma-Aldrich) or FGF21 (50 nmol/L) (Sigma-Aldrich) in RPMI 1640 media. The islet cells were fixed in 4% paraformaldehyde, permeabilized in permeabilization buffer (Thermo Fisher Scientific), blocked in blocking solution (Thermo Fisher Scientific), and then incubated with mouse anti-glucagon (1:100) (Sigma-Aldrich), guinea pig anti-insulin (1:100) (Dako), or rabbit anti–pancreatic and duodenal homeobox-1 (PDX-1) (1:25) (Cell Signaling Technology, Beverly MA) antibodies. FITC-conjugated goat anti-rabbit IgG (1:200) (Santa Cruz Biotechnology), TR-conjugated goat anti–guinea pig IgG (1:200) (Santa Cruz Biotechnology), or TR-conjugated goat anti-mouse IgG (1:200) (Santa Cruz Biotechnology) were used as secondary antibodies. Fluorescence was imaged using a laser scanning confocal fluorescent microscope (LSM 700).

The αTC1 clone 9 (αTC1-9) is a pancreatic α-cell line that produces glucagon, but not preproinsulin mRNA (19,20). Cells were obtained from ATCC (CRL-2350) and grown in DMEM with 16.7 mmol/L glucose supplemented with 10% heat-inactivated dialyzed FBS, 15 mmol/L HEPES, 0.1 mmol/L nonessential amino acids, and 0.02% BSA under an atmosphere of 95% humidified air and 5% CO₂ at 37°C.

PCR was carried out in a 7900HT Fast Real-Time PCR system (Applied Biosystems, Carlsbad, CA). The specific PCR primers are given in Supplementary Table 1. The relative copy number was calculated using the threshold crossing point as calculated by the 7900HT Fast Real-Time PCR software combined with the ΔΔ threshold crossing point calculations.

αTC1-9 cells were seeded in 24-well plates at a density of 1 × 10⁵ cells 1,000 μL⁻¹/well and cultured with or without exendin-4 (10 nmol/L) for 7 days (media were changed, and exendin-4 was added daily). The cells were pulsed with [³H]thymidine (1 μCi/well). Eight hours after [³H]thymidine addition, the cells were analyzed for [³H]thymidine incorporation using a scintillation β-counter, the 1450 LSC and Luminescence Counter MicroBeta TriLux (PerkinElmer).

Whole lysates of cells were prepared as described previously (21). Western blots were performed with rabbit anti-cyclin D2 (Cell Signaling Technology), mouse-anti–β-actin (Santa Cruz Biotechnology), goat anti-FGF21 (R&D Systems, Minneapolis, MN), rabbit anti–peroxisome proliferator–activated receptor-α (PPAR-α) (Santa Cruz Biotechnology), rabbit anti–PDX-1 (Millipore), rabbit anti–PDX-1 (Abcam), goat anti-neurogenin 3 (Ngn3) (Santa Cruz Biotechnology), rabbit anti-glucagon (Santa Cruz Biotechnology), and mouse-anti–β-tubulin (Santa Cruz Biotechnology).

Mouse islets were isolated, and STZ (1 mmol/L) was added for 15 h to destroy the β-cells. The STZ-administered islets were given new media and treated with exendin-4 (10 nmol/L added every 24 h) for 72 h. FGF21 secretion was analyzed by ELISA (R&D Systems), with values normalized to protein.

αTC1-9 cells (1 × 10⁵) were seeded in six-well plates in DMEM medium with 16.7 mmol/L glucose. After 24 h, the medium was changed to that with 25 mmol/L glucose, and exendin-4 (0, 5, 10, or 20 nmol/L) was added for 2 h. The medium was removed, and 1 mL of 0.1 mol/L HCL was added for 20 min for lysis. The lysates were centrifuged at 5,000 rpm for 5 min to pellet the cellular debris. The supernatant was assayed for cAMP using ELISA (Enzo Life Sciences, Farmingdale, NY).

FGF21 small interfering RNAs (siRNAs) were purchased from Bioneer (Daejeon, Korea). The target sequences were 5′-CUGAUGGAAUGGAUGAGAU-3′ and 5′-AUCUCAUCCAUUCCAUCAG-3′. Intact islets from C57BL/6 mice were dissociated at 37°C in Accutase (Millipore). The single islet cells were seeded in 24-well plates at a density of 2 × 10⁵ cells 250 μL⁻¹/well and cultured with RPMI 1640 media for 12 h. FGF21 siRNAs were transfected with RNAi Max reagent (Invitrogen) according to the manufacturer’s protocol.

Data are presented as means ± SE. Statistical significance of the difference between two groups or among multiple groups was analyzed by unpaired Student t test or ANOVA followed by Fisher protected least significant difference test, respectively. A P value <0.05 was accepted as significant.

To investigate whether GLP-1–induced new β-cells come from remaining β-cells or non–β-cells, we used RIP-CreER;R26-YFP transgenic mice, bearing the transgenes RIP-CreER (inducible tagger) and R26-YFP as a reporter, to label pre-existing β-cells (11). Administration of 4-hydroxytamoxifen induced the expression of the reporter protein YFP in β-cells, and almost all of the β-cells expressed YFP (Supplementary Fig. 1). STZ-induced diabetic RIP-CreER;R26-YFP male mice were injected via the tail vein with rAd-GLP-1 or rAd-βgal, and then blood glucose levels were monitored for 4 weeks. Blood glucose levels were significantly decreased in rAd-GLP-1–treated mice compared with rAd-βgal–treated mice and became normoglycemic (Supplementary Fig. 2A). The i.p. glucose tolerance tests in normoglycemic RIP-CreER;R26-YFP mice at 2 weeks after rAd-GLP-1 treatment showed that blood glucose levels in rAd-GLP-1–treated mice were properly controlled (Supplementary Fig. 2B).

To determine whether β-cells are increased in diabetic rAd-GLP-1–treated mice, pancreatic sections were analyzed by immunostaining. The insulin-positive cell population, including YFP⁺insulin⁺ (new β-cells from surviving β-cells) and YFP⁻insulin⁺ (new β-cells from non–β-cells), was significantly increased in rAd-GLP-1–treated mice compared with rAd-βgal–treated mice (Fig. 1A–C). YFP⁻insulin⁺ cells were 49.38 ± 3.95% of the insulin-positive cells in rAd-GLP-1–treated mice and 23.25 ± 2.79% in rAd-βgal–treated mice. Thus, the proportion of new β-cells originating from non–β-cells (YFP⁻insulin⁺) was significantly higher in rAd-GLP-1–treated mice compared with rAd-βgal–treated mice (Fig. 1D).

To explore the origin of new β-cells originating from non–β-cells, we first examined the alteration of cells in the islets of rAd-GLP-1–treated RIP-CreER;R26-YFP mice after β-cell ablation by STZ injection. To monitor the proliferating cells, the mitotic marker BrdU was injected daily (i.p.) for 4 weeks. BrdU-positive cells in the islets increased in rAd-GLP-1–treated mice compared with rAd-βgal–treated mice (Fig. 2A and B). Both BrdU⁺glucagon⁺ cells (Fig. 2C) and BrdU⁺insulin⁺ (Fig. 2D) cells were significantly increased in the islets of rAd-GLP-1–treated mice compared with rAd-βgal–treated mice, with BrdU⁺glucagon⁺ cells showing the greater increase.

To investigate whether GLP-1 receptor signaling directly increases proliferation in α-cells, we examined the effect of exendin-4 on proliferation of αTC1-9 cells by [³H] thymidine incorporation assay. Exendin-4 significantly increased proliferation of αTC1-9 cells dose dependently (Fig. 2E). Cyclins are important proteins that control the proliferation of cells; thus, we analyzed the expression of cyclins. The expression of cyclin D2 mRNA and protein was significantly increased (Fig. 2F and J); however, the expression of cyclin A2, cyclin E, and cyclin D3 was not changed in exendin-4–treated αTC1-9 cells (Fig. 2G–I). Similarly, BrdU-positive α-cells were increased in exendin-4–treated αTC1-9 cells when the cells were cultured in the presence of BrdU (Fig. 2K).

Recent studies show plasticity between pancreatic α- and β-cells. New β-cells can be produced from α-cells via a bihormonal insulin⁺glucagon⁺ transitional state in animals almost devoid of β-cells (11). Immunostaining of pancreatic sections from rAd-GLP-1– or rAd-βgal–treated RIP-CreER;R26-YFP mice with anti-glucagon and anti-insulin antibodies revealed that insulin⁺glucagon⁺ double-stained bihormonal cells were increased in rAd-GLP-1–treated mice compared with rAd-βgal–treated mice (Fig. 3A and B).

To investigate whether GLP-1 receptor signaling directly increases insulin⁺glucagon⁺ bihormonal cells, we isolated islet cells from C57BL/6 mice, treated them with STZ to destroy β-cells, and then treated them with exendin-4. In isolated islets, β-cells were specifically destroyed by STZ treatment, whereas α-cells were preserved. In addition, the pattern of α-cell distribution in islets was similar to that observed in islets of STZ-administered mice (Supplementary Fig. 3). STZ-induced β-cell destruction was confirmed by staining with anti-insulin antibody. Insulin⁺ cells were almost completely destroyed with only 3% of STZ-administered islet cells being insulin positive (Fig. 3C). The STZ-administered islet cells were cultured with exendin-4 for 2 days and stained with anti-glucagon and anti-insulin antibodies (Fig. 3D). Insulin⁺glucagon⁺ bihormonal cells were significantly increased in STZ–exendin-4–administered islet cells compared with STZ-only–administered islet cells (Fig. 3E). To investigate the change of the expression of β- or α-cell–related genes, we measured the expression of insulin, glucagon, PDX-1, and aristaless-related homeobox (Arx) mRNA and protein expression of glucagon and PDX-1 in STZ–exendin-4–administered islet cells. The expression of glucagon and Arx mRNA and glucagon protein was significantly increased at 3 days after exendin-4 treatment in STZ–exendin-4–administered islet cells compared with STZ-administered islet cells. In addition, the expression of insulin and PDX-1 mRNA and PDX-1 protein was significantly increased at 7 days after exendin-4 treatment in STZ–exendin-4–administered islet cells compared with STZ-administered islet cells (Fig. 3F–O).

To investigate the possibility of transdifferentiation of α-cells to β-cells by GLP-1, we used a genetic lineage tracing system. Glucagon-rtTA;Tet-O-Cre;R26-YFP mice have a doxycycline-inducible glucagon-driven reverse tet transactivator to direct Cre recombinase expression from a Tet-O-Cre transgene in glucagon⁺ α-cells. Cre activates YFP transgene expression from the Rosa26 locus. Thus, doxycycline stimulates Cre recombinase expression specifically in glucagon⁺ α-cells. Almost all of the glucagon⁺ α-cells were labeled with YFP in pancreatic sections of Glucagon-rtTA;Tet-O-Cre;R26-YFP mice exposed to doxycycline for 2 weeks (Fig. 4A). After a washout period of 2 weeks after doxycycline treatment, diabetes was induced by STZ injection. The diabetic mice (blood glucose levels >300 mg/dL for 3 consecutive days) were injected via the tail vein with rAd-GLP-1 or rAd-βgal. Blood glucose levels were significantly decreased in rAd-GLP-1–treated mice compared with rAd-βgal–treated mice (Supplementary Fig. 4). At 4 weeks after virus injection, pancreatic sections were analyzed by immunostaining. The insulin-positive cell population, including YFP⁺insulin⁺ (new β-cells from α-cell transdifferentiation) and YFP⁻insulin⁺ (new β-cells from non–α-cells), was significantly increased in rAd-GLP-1–treated mice compared with rAd-βgal–treated mice (Fig. 4B, C, and F). New β-cells from α-cell transdifferentiation, YFP⁺insulin⁺-coexpressing cells, were 32% of insulin-positive cells observed in rAd-GLP-1–injected mice and 22% in rAd-βgal–injected mice (Fig. 4G). In addition, YFP⁺PDX-1⁺ cells were significantly increased in rAd-GLP-1–treated mice compared with rAd-βgal–treated mice (Fig. 4D, E, and H). These results suggest that GLP-1 increases new β-cell generation by promoting α- to β-cell transdifferentiation.

A recent report demonstrated that GLP-1 induces FGF21 production in the liver and adipose tissue (16). As well, pancreatic α-cells express FGF21 and hyperplastic α-cells highly express FGF21 (14). In addition, long-term treatment with FGF21 increases the number of insulin-positive cells per islet (15). To investigate whether FGF21 is truly involved in the increase of β-cells by GLP-1, STZ-induced diabetic wild-type and FGF21 KO mice were injected with rAd-GLP-1, and then blood glucose levels were monitored. Blood glucose levels were decreased in both wild-type and FGF21 KO mice, but the glucose levels in FGF21 KO mice were significantly higher compared with wild-type mice (Fig. 5A). The i.p. glucose tolerance tests at 4 weeks after rAd-GLP-1 treatment showed that blood glucose levels of FGF21 KO mice were significantly higher at 60, 90, and 120 min following glucose injection compared with wild-type mice (Fig. 5B). We then examined pancreatic sections by immunostaining. After rAd-GLP-1 treatment, the insulin⁺ cell population was significantly lower and the glucagon⁺ cell population was significantly higher in FGF21 KO mice compared with wild-type mice (Fig. 5C and D). To investigate the alteration of the expression of β- and α-cell–related genes, we measured the expression of insulin, glucagon, PDX-1, and Arx mRNA in STZ-administered dispersed islets from FGF21 KO or wild-type mice after exendin-4 treatment. The expression of all these mRNAs was significantly reduced in islets from FGF21 KO mice compared with wild-type mice (Fig. 5E–H). These results suggest that increase of β-cells by GLP-1 is, in part, mediated by FGF21.

To investigate whether FGF21 directly increases insulin⁺glucagon⁺ bihormonal cells, we treated isolated islet cells from C57BL/6 mice with STZ to destroy β-cells and then treated them with FGF21. FGF21 treatment of β-cell–ablated islets significantly increased the number of insulin⁺glucagon⁺ bihormonal cells (Fig. 6A and B).

To investigate whether exendin-4–induced α-cell transdifferentiation into β-cells is mediated by FGF21, we inhibited the expression of FGF21 by siRNA in β-cell–ablated islets and treated them with exendin-4. Knockdown of FGF21 significantly decreased the number of insulin⁺glucagon⁺ bihormonal cells generated by exendin-4 treatment (Fig. 6C–E). In addition, we confirmed this result using islets from FGF21 KO mice. The generation of insulin⁺glucagon⁺ cells by exendin-4 was significantly reduced in STZ-administered islets from FGF21 KO mice compared with wild-type mice (Fig. 6F and G). These results suggest that generation of insulin⁺glucagon⁺ bihormonal cells by exendin-4 is mediated by FGF21 production.

To investigate whether FGF21 induces the expression of β-cell transcription factors in α-cells, we isolated mouse islets from C57BL/6 mice, ablated the β-cells with STZ, treated the islets with FGF21, and then analyzed mRNA expression of β-cell transcription factors in α-cells. We found that PDX-1 mRNA expression was significantly increased, but the expression of glucagon mRNA was significantly decreased in β-cell–ablated islets treated with FGF21. However, the expression of Arx, MafA, MafB, and insulin mRNA was not changed by FGF21 treatment (Fig. 7A–F). The protein level of PDX-1 and Ngn3, an important transcription factor in endocrine pancreas development and transdifferentiation of α- to β-cells (22), was also significantly increased in FGF21-treated β-cell–ablated islets compared with cells without FGF21 treatment (Fig. 7G–J). To investigate whether the expression of PDX-1 is observed in glucagon⁺ α-cells after FGF21 treatment, we stained β-cell–ablated islet cells with anti-glucagon and anti–PDX-1 antibodies. Immunocytochemical analysis showed that coexpression of glucagon and PDX-1 in α-cells was clearly observed after FGF21 treatment. The intensity of PDX-1 staining in glucagon-producing α-cells was significantly increased in FGF21-treated cells compared with cells without FGF21 treatment as measured by a laser scanning confocal fluorescent microscope (Fig. 7K and L).

It was reported that pancreatic α-cells express FGF21 (14). We found that α-cells are increased by GLP-1 in this study, and FGF21 is involved in GLP-1–induced α-cell transdifferentiation into β-cells. Therefore, we wanted to know whether α-cells produce FGF21 after treatment with GLP-1. In αTC1-9 cells, the expression of FGF21 mRNA was increased at 12 h after exendin-4 treatment (Fig. 8A), and FGF21 protein was increased at 24 h after exendin-4 treatment (Fig. 8C), both dose dependently. FGF21 is regulated by PPAR-α (23). Thus, we investigated the expression of PPAR-α in exendin-4–treated αTC1-9 cells. The expression of PPAR-α mRNA was upregulated at 6 h, and the protein level of PPAR-α was increased at 24 h after exendin-4 treatment (Fig. 8B and C). To investigate whether FGF21 expression is regulated by GLP-1 in α-cells, we used islets in which β-cells had been ablated by administration of STZ. The expression of FGF21 mRNA (Fig. 8D) and FGF21 protein secretion (Fig. 8E) were also significantly increased in STZ–exendin-4–administered islets compared with STZ-administered islets. In addition, we investigated whether the expression of FGF21 is increased in vivo in rAd-GLP-1–injected mice. At 1 week after virus injection of diabetic Glucagon-rtTA;Tet-O-Cre;R26-YFP mice, pancreatic sections were analyzed for FGF21 expression. FGF21⁺ cells were significantly increased in islets of rAd-GLP-1–treated mice compared with rAd-βgal–treated mice (Supplementary Fig. 5).

The major pathway for GLP-1 receptor signaling is via a cAMP-dependent pathway (24). Thus, we determined whether the cAMP pathway is involved in FGF21 production by exendin-4 in αTC1-9 cells. The production of cAMP was increased at 2 h after exendin-4 treatment in αTC1-9 cells (Fig. 8F). The expression of FGF21 mRNA was significantly increased by exendin-4 treatment, but it was inhibited by KH7, which is a selective inhibitor of soluble adenylyl cyclase (Fig. 8G). These results suggest that GLP-1 increases FGF21 expression through cAMP in α-cells. Next, we determined the expression of FGF receptors after exendin-4 treatment. The mRNA expression of FGFR1, FGFR4, and coreceptor β-klotho was significantly increased in exendin-4–treated αTC1-9 cells (Fig. 8H–L). These results suggest that GLP-1 increases both FGF21 and FGF receptor expression in α-cells.

Diabetes is caused by absolute or relative deficiency of functional pancreatic β-cells, resulting in defective control of blood glucose. Thus, strategies to increase the β-cell mass—for example, the proliferation of remaining β-cells, differentiation of progenitors of β-cells, and transdifferentiation of non–β-cells into β-cells—have been investigated for management of diabetes.

GLP-1 is known to have potential effects in regulation of the β-cell mass through regeneration, differentiation, and neogenesis of pancreatic β-cells (3–5). However, GLP-1 is rapidly inactivated by the enzyme dipeptidyl peptidase 4 (25). We previously constructed a recombinant adenovirus containing cytomegalovirus promoter and albumin leader sequence, followed by GLP-1 cDNA (rAd-GLP-1) to facilitate GLP-1 secretion in the circulation (26). In mice treated with rAd-GLP-1, serum levels of GLP-1 are dramatically increased and remain significantly higher for at least 4 weeks compared with rAd-βgal–treated mice (17), indicating that a substantial amount of circulating GLP-1 could effectively act throughout the whole body including pancreas. In this study, hyperglycemia was rapidly reversed in <2 days, which might be due to the large amount of GLP-1 produced after rAd-GLP-1 administration. Reduced blood glucose levels were maintained for 28 days when the experiment was terminated. These effects can be due to the increase of newly generated β-cells.

In the current study, we analyzed the contribution of pre-existing β-cells to become new β-cells using the tamoxifen-dependent Cre/loxP system: transgenic mice bearing the transgenes RIP-CreERT (inducible tagger) and R26-YFP as a reporter. After β-cell destruction by STZ, we found that both existing β-cells and non–β-cells contributed to the generation of new β-cells in response to GLP-1 produced by rAd-GLP-1. Interestingly, we found that a larger proportion of insulin⁺ cells originated from non–β-cells than from pre-existing β-cells in rAd-GLP-1–treated mice compared with rAd-βgal–treated control mice.

Besides the production of glucagon, α-cells in the pancreatic islets can be a source of new β-cells (11). α-Cell hyperplasia occurs in response to hyperglycemia resulting from injury of β-cells as well as in pancreatic islets of diabetic animals and human patients with diabetes (27–29), suggesting that the new α-cells might serve as a source for β-cell regeneration. In our study, we found increased proliferation of α-cells in the pancreas of rAd-GLP-1–treated mice compared with rAd-βgal–treated mice after ablation of β-cells by STZ. Similarly, we found that treatment of αTC1-9 cells, which produce glucagon but not preproinsulin mRNA (20,30), with exendin-4, a GLP-1 receptor agonist, increased the proliferation of α-cells.

An increased α-cell mass might be expected to result in an increase of glucagon production, which promotes hepatic glucose output and deterioration of metabolic control in the diabetic condition. Although α-cells were increased after rAd-GLP-1 treatment in STZ-induced diabetic mice, serum glucagon levels were not significantly different from rAd-βgal–treated mice (Supplementary Fig. 6). As well, normal blood glucose levels were maintained in rAd-GLP-1–treated STZ-induced diabetic mice for 4 weeks, suggesting that the proliferation of α-cells caused by GLP-1 did not adversely affect metabolic control of glucose. Similarly, continuous administration of GLP-1 for 6 weeks in patients with type 2 diabetes did not change plasma glucagon concentrations and decreased hemoglobin A_1c (31). In our study, we found that the pancreatic α-cells are a target for GLP-1 action. GLP-1 promotes α-cell proliferation in vitro and in vivo. However, the pancreatic α-cell proportion was similar between STZ–rAd-βgal–administered mice and STZ–rAd-GLP-1–administered mice, even though the proliferation of α-cells was increased in STZ–rAd-GLP-1–treated mice. Furthermore, the pancreatic β-cell proportion was increased in rAd-GLP-1–treated mice (Supplementary Fig. 7). These results suggest that α-cells may transdifferentiate into β-cells in rAd-GLP-1–treated mice. In addition, rAd-GLP-1 treatment in vivo or exendin-4 treatment in vitro increased insulin⁺glucagon⁺ bihormonal cells after β-cell ablation. In conditions of extreme loss of β-cells, bihormonal cells are frequently observed, and they are not from original β-cells but are generated from pre-existing α-cells that start to produce insulin and transdifferentiate into β-cells (11). In our study, bihormonal cells were actually quite rare in vivo (Fig. 3A and B) compared with the in vitro experiment (Fig. 3D and E). This difference may be due to observational time point. In vivo, we measured bihormonal cells after 4 weeks of rAd-βgal or rAd-GLP-1 injection. At that time, many bihormonal cells might already be transdifferentiated to new β-cells. However, in vitro, we measured bihormonal cells after 2 days of exendin-4 treatment, perhaps before complete transdifferentiation occurred. Although α- to β-cell transdifferntiation is known to occur in the case of extreme β-cell loss, other mechanisms, such as generation of new β-cells from progenitor cells or the trans-differentiation of pancreatic polypeptide or δ-cells to new β-cells, could also explain the larger proportion of insulin⁺ cells originated from non–β-cells than from pre-existing β-cells in rAd-GLP-1–treated mice compared with rAd-βgal–treated control mice. Lineage-tracing studies using Glucagon rtTA;Tet-O-Cre;R26-YFP mice showed that new β-cells originating from α-cells were significantly increased in rAd-GLP-1–injected mice compared with rAd-βgal–injected mice. These results indicate that pre-existing α-cells transdifferentiated into insulin-producing β-cells by GLP-1 treatment. Thus, α-cell expansion and transdifferentiation by GLP-1 may contribute to β-cell compensation. However, the mechanism for the transdifferentiation of α- to β-cells remains enigmatic. In addition, rAd-GLP-1 treatment induced proliferation in non–α/non–β-cells compared with rAd-βgal treatment (Fig. 2B–D). rAd-GLP-1 treatment might increase proliferation of pancreatic polypeptide-, ε-, or δ-cells, which also might transdifferentiate to new β-cells. However, the percentage of non–α/non–β-cells in the islets did not increase by exendin-4 (Supplementary Fig. 8). The identity of non–α/non–β-cells as a source of transdifferentiated cells needs further study.

GLP-1 is produced in pancreatic α-cells by activation of PC1/3 under demands for β-cell regeneration such as pregnancy, ob/ob, or db/db conditions and in prediabetic NOD mice (32). In addition, more GLP-1 is released from freshly isolated islets of hyperglycemic animals than from normoglycemic animals; healthy islets secrete more GLP-1 following culture with high glucose (33). In our study, we found that exendin-4 treatment increased PC1/3 expression and GLP-1 production in αTC1-9 cells and β-cell–ablated islets (Supplementary Fig. 9). Locally produced GLP-1 in the pancreatic islets contributes to maintaining β-cell function (9), and GLP-1 restores leucine-induced α-cell dysfunction (34). Thus, local production of GLP-1 from pancreatic α-cells on demand may be beneficial to protect and regenerate β-cells.

A GLP-1 analog was shown to increase FGF21 protein in liver and plasma (16). FGF21 protein expression is detected in both α- and β-cells in the pancreas (14), and there is a correlation between the increase of plasma GLP-1 and FGF21 in glucagon receptor KO mice (14). Constant infusion of FGF21 for 8 weeks in db/db mice normalized blood glucose levels and increased plasma insulin levels. In addition, FGF21-treated mice showed a higher number of insulin-positive cells per islet compared with control mice without β-cell proliferation (15). Therefore, we hypothesized that FGF21 production was induced by GLP-1 in α-cells, and FGF21 might play a role in the generation of new β-cells by GLP-1. We found that exendin-4 increased the expression of FGF21 mRNA and protein in αTC1-9 cells, β-cell–ablated islets, and rAd-GLP-1–injected mouse islets. We then investigated whether FGF21 plays a role in GLP-1–induced new β-cell generation. Exendin-4 treatment of β-cell–ablated islets increased the insulin⁺glucagon⁺ bihormonal cell population, and inhibition of FGF21 by siRNA FGF21 reduced this exendin-4–induced bihormonal cell population. In vivo studies using FGF21 KO mice revealed that the insulin⁺ cells in islets were significantly reduced compared with wild-type mice after treatment with rAd-GLP-1.

PDX-1 is an important transcription factor for both pancreatic development and the differentiation of progenitor cells into the β-cell phenotype (35). Forced PDX-1 expression induces α-to-β-cell conversion (36), whereas adult β-cell–specific removal of PDX-1 results in the reverse, a rapid transdifferentiation to α-cells (37). Ngn3 is also a critical gene for pancreatic development of endocrine cells, and the combined expression of PDX-1 with Ngn3 improves the differentiation efficiency of embryonic stem cells into insulin-producing cells (38–40). As well, the ectopic expression of Ngn3 induces conversion of other cells into endocrine cells (41), and the expression of Ngn3 is necessary for the transdifferentiation of α-cells to β-cells (22). Therefore, we examined whether FGF21 can induce PDX-1 or Ngn3 expression in α-cells. We found that FGF21 treatment of β-cell–ablated islets increased both PDX-1 and Ngn3 expression and augmented PDX-1 intensity in glucagon-producing cells. We speculate that FGF21 may stimulate β-cell formation by inducing Ngn3 and PDX-1 expression. The percentage of α-cells in the islets of the rAd-GLP-1–injected FGF21 KO mice was higher than wild-type mice. In addition, it was reported that constant infusion of FGF21 increased β-cell mass without β-cell proliferation in mice (15). Therefore, FGF21 probably contributes to the increase of α-cell transdifferentiation into β-cells. Further studies are needed to determine the detailed mechanisms by which FGF21 induces the expression of these transcription factors. One possibility is that FGF21 might increase PDX-1 expression by increase of PPAR-γ, as it is known that FGF21 promotes PPAR-γ activity and expression (42), and PPAR-γ enhances the expression of PDX-1 and Nkx6.1 in INS-1 cells (43).

In summary, we found that GLP-1 contributes to new β-cell generation from α-cells. Circulating GLP-1 and GLP-1 produced from pro–α-cells could induce FGF21 and subsequently increase β-cell transcription factors, such as PDX-1 and Ngn-3. These results suggest that GLP-1 may act on α-cell transdifferentiation to β-cells via FGF21 induction.

Acknowledgments. The authors thank Dr. Ann Kyle (University of Calgary, Calgary, Alberta, Canada) for editorial assistance.

Funding. This research was supported by a grant from the Innovative Research Institute for Cell Therapy Project, Ministry of Health and Welfare, Republic of Korea (A062260); the Korea Health Technology R&D Project through the Korea Health Industry Development Institute, funded by the Ministry of Health and Welfare, Republic of Korea (grant HI14C1135); and the Basic Science Research Program through the National Research Foundation of Korea, funded by the Ministry of Education (2015R1D1A1A01060232).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. Y.-S.L. designed the study, performed experiments and data analysis, and wrote the manuscript. C.L. and J.-S.C. performed experiments and data analysis. H.-S. Jung designed the study. H.-S. Jun designed the study, performed data analysis, and wrote the manuscript. Y.-S.L. and H.-S. Jun are the guarantors of this work and, as such, had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.