Interplay of Glucagon-Like Peptide-1 and Transforming Growth Factor-β Signaling in Insulin-Positive Differentiation of AR42J Cells

Exendin-4 was obtained from Sigma-Aldrich (St. Louis, MO). RNeasy Mini Kit, Sensiscript Reverse Transcriptase Kit, and QIAquick Gel Extraction Kit were all from Qiagen (Valencia, CA). AmpliTaq Gold with GeneAmp 10× PCR buffer and MgCl₂ solution was from Applied Biosystems (Foster City, CA). F-12K Nutrient Mixture (Kaighn’s modification) was from Gibco/Invitrogen (Carlsbad, CA).

Normal donkey serum and Rhodamine (tetrarhodamine isothiocyanate)-conjugated AffiniPure donkey anti-sheep IgG antibodies were from Jackson ImmunoResearch Laboratories (West Grove, PA). Sheep anti-human insulin (IgG fraction) was from the Binding Site (Birmingham, U.K.). In additional experiments, a pan-specific TGF-β neutralizing antibody (AB-100-NA; R&D Systems) was added at a final concentration of 30 μg/ml, which has been established by the manufacturer to inhibit 80% of TGF-β activity.

Antisense morpholino oligos complementary to smad 2 (5′-CCCCGCCCACTCACCTTGGTATGGT-3′), smad 3 (5′CAAGCTGGTCACATTCATAGCGGCT-3′), and missense control oligos were obtained from Gene Tools (Philomath, OR). Scrape delivery method was performed, and the scraped cells along with morpholino oligos were incubated over the weekend in a CO₂ incubator, which gives the best results in our system.

AR42J cells, purchased from American Type Culture Collection (Manassas, VA), were grown in Kaighn’s modification of Ham’s F12K medium with 2 mmol/l l-glutamine, 250 μg/ml Amphotericin, 100 units/ml penicillin, 100 μg/ml streptomycin, and 20% fetal bovine serum at 37°C under a humidified condition of 95% air and 5% CO₂. Cells were plated at a density of ∼10⁵ cells/ml in 12-well plates. Morpholino antisense or missense control was added separately to culture media at 20 μmol/l. Cells were then cultured with exendin-4 at doses of 1, 5, and 10 pmol/l for 3 days.

Cells cultured on noncoated glass coverslips (Fisher Scientific, Pittsburgh, PA) at a density of ∼10⁵ cells/ml were fixed with 4% phosphate-buffered paraformaldehyde for 15 min at room temperature. After fixation, the cells were permeabilized with 0.2% Triton X-100 (Sigma, St. Louis, MO) for 5 min at room temperature, blocked with normal donkey serum for 30 min, and then incubated with primary sheep anti-human insulin antibody in a moist chamber for 2 h at room temperature. The cells were rinsed with PBS and incubated with secondary rhodamine-conjugated donkey anti-sheep IgG for 2 h at room temperature in a dark cabinet. After several washes, coverslips containing cells were mounted onto slides in aqueous mounting medium with anti-fading agents (Biomeda, Foster City, CA). Fluorescence digital images were captured using an Olympus BX60 (Melville, NY) microscope attached with an Olympus U-PMTVC camera adaptor. Optronics DEI-750 (Goleta, CA) software was used to acquire and analyze images.

Total RNA was extracted from cells and treated with DNase. RNA was subjected to reverse transcription. cDNA was then amplified by PCR for 40 cycles (Table 1). All PCR products were separated by electrophoresis in 2% agarose gel. PCR cycles were as follows: initial denaturation at 95°C for 10 min, followed by 40 cycles of 94°C for 30 s, 60°C for 30 s, 72°C for 30 s, and final extension at 72°C for 10 min.

PCR amplifications were performed using a Bio-Rad iCycler (Hercules, CA) sequence detection system. Reactions were performed in a 50-μl reaction mixture including 10× Gold Buffer, 25 mm MgCl₂, 2.5 mmol/l dNTPs, 10× SYBR green, TaqGold polymerase, dH₂O, DNA template, and 10 μmol/l of each primer. Amplication was performed by initial polymerase activation for 10 min at 95°C, 40 cycles of denaturation at 95°C for 15 s, annealing at 60°C for 20 s, and elongation for 30 s at 72°C.

Proteins were separated on a 10% Tris-Hcl Ready Gel (Bio-Rad, Hercules, CA), transferred onto nitrocellulose membranes, and incubated with either β-actin antibody (at a dilution of 1:5,000) or anti–smad 3 antibody (at 4 μg/ml) overnight at 4°C. After incubation, the membranes were washed twice for 15 min in washing buffer (PBS 0.05% Tween 20) and incubated with a secondary anti-mouse (β-actin) or anti-rabbit (smad 3) antibody coupled to horseradish peroxidase (Vector Labs, Burlingame, CA) for 1 h at room temperature. Then, the membranes were washed three times for 15 min in washing buffer, and immunoreactivity was normalized by chemiluminescence (ECL+Plus kit, RPN2132; Amersham) according to the manufacturer’s instructions.

To expand on previous observations that the GLP-1 agonist exendin-4 was able to induce insulin-positive differentiation (14), we first performed a broad dose-response of GLP-1 in the AR42J cells and found that exendin-4 in the 1- to 10-pmol/l range gave maximal insulin II mRNA levels, as determined by semiquantitative RT-PCR (Fig. 1A). Further dose-responses over a narrower range showed that 5 pmol/l specifically gave the highest levels of insulin II mRNA expression in our hands (Fig. 1B). To further analyze the nature of this effect, we studied levels of pdx-1 mRNA, a transcription factor that is expressed at low levels in early pancreatic progenitor epithelial cells but at high levels in differentiated β-cells. Like insulin, pdx-1 was found to have the highest expression with the same doses of exendin-4, implying a similarly affected population of cells and implying that the insulin-positive cells may be becoming mature β-cells (Fig. 1C). Similarly, IAPP mRNA also was maximal at this dose (Fig. 1D), again indicating mature β-cell differentiation. However, HNF-3β mRNA, a marker of endoderm, was suppressed by exendin-4, suggesting that HNF-3β enhancement is neither necessary nor responsible for the increased pdx-1 (though the pdx-1 gene has an HNF-3β responsive element) (Fig. 1E). Furthermore, immunocytochemistry for insulin revealed that only a few untreated AR42J cells were insulin positive, whereas many insulin-positive cells were seen after exendin-4 treatment of the AR42J cells (Figs. 1F–H).

In light of the known role for activin in inducing an endocrine phenotype in AR42J cells (1,2,16), we next determine whether the activin/TGF-β isoform-specific smads (smad 2 and 3) may mediate the effect of exendin-4. We first analyzed levels of smad mRNAs. Interestingly, smad 2 mRNA was found to be decreased significantly upon stimulation of AR42J cells with exendin-4 at all doses tested (Fig. 2A). Conversely, smad 3 was found to be strongly upregulated after exendin-4 treatment (Fig. 2B). To determine whether levels of these two smads may be secondarily dependent on levels of smad 4, which serves as a common binding partner for smad 2 or 3 and can potentially serve as a rate-limiting factor, we found minimal change of smad 4 levels with or without exendin-4 treatment (Fig. 2C). The changes in the levels of these receptor-specific smads (2 and 3) may be a key point of regulation. Similarly, by Western blotting, levels of the proteins were found to correlate with the mRNA.

To determine whether the observed changes in smad 2 and 3 levels mediate the insulin-positive differentiation seen with exendin-4 treatment, we used a morpholine ring antisense approach to inhibit production of smad 2 and/or smad 3 protein in the AR42J cells. In the presence of smad 2 antisense, no increase in insulin II mRNA or pdx-1 mRNA was seen, suggesting that smad 2 is necessary for β-cell differentiation (Figs. 3A and B). In an effort to better define the differentiation state of the endocrine cells that were present, levels of pax 4 and pax 6 mRNA were assayed. Levels of both of these mRNAs were reduced to baseline by smad 2 antisense (Figs. 3C and D), suggesting a complete absence of endocrine cells. In addition, smad 3 mRNA levels were decreased as a result of smad 2 inhibition, suggesting that smad 2 in AR42J cells may be necessary both for insulin-positive differentiation as well as for full smad 3 upregulation (Fig. 3E). Immunocytochemistry confirmed the decrease in insulin, though the number of positive cells was not as dramatically decreased as the insulin II mRNA levels, suggesting that the number of mRNA molecules per insulin-positive cell was inhibited in addition to the decreased number of insulin-positive cells (Figs. 3F–H). To confirm the inhibiting effects of the morpholino ring antisense on smad levels, Western blotting was performed in morpholino ring antisense-treated and sense-treated controls, confirming sequence-specific decreases in target smads by antisense (Fig. 3I).

When smad 3 antisense was used instead of smad 2, there was a surprising fivefold increase in insulin II mRNA levels (Fig. 4A) and a doubling of the number of insulin-positive cells by immunocytochemistry (Figs. 4G–I). Unlike the previous elevations of insulin II mRNA, however, there was a decrease in pdx-1 and IAPP mRNA levels (Figs. 4B and C), implying that the elevated levels of insulin expression in the absence of smad 3 could represent an immature, possibly proliferative β-cell progenitor. In addition, levels of nestin mRNA expression were also elevated (data not shown), suggesting that these smad 3 antisense-treated cells may be frozen in a predifferentiated state, with low levels of IAPP and pdx-1. Pax6 mRNA showed a 10- to 15-fold increase (Fig. 4D), again suggesting amplification of a precursor population of relatively immature β-cells. Minimal effects were seen on pax4 mRNA levels (Fig. 4E). Together, these results suggest that smad 3 antisense induced an increased number of endocrine precursor cells. Smad 2 mRNA levels were only modestly increased (Fig. 4F). Immunocytochemistry correlated with the RT-PCR results, except to a less dramatic degree, suggesting that either the number of insulin mRNA molecules per cell was lower or that there were numerous insulin-positive cells below the threshold of detection of immunocytochemistry.

Given the opposing effects of smad 2 or 3 antisense inhibition, we then blocked both smad 2 and 3 simultaneously, using double antisense treatment of AR42J cells (Fig. 5). Here, the results were essentially identical to those obtained with smad 2 antisense alone. Thus, it appears that smad 3 effects are probably downstream or dependent on smad 2 effects. It also suggests that smad 2 may initially be necessary for elevation of insulin levels and then subsequently may lower insulin levels, possibly as part of β-cell maturation.

Given the apparent necessity for smad 2 and 3 in properly regulated, exendin-4–induced insulin-positive differentiation in AR42J cells, we presumed that activin and/or TGF-β isoform signaling may be playing an important role in mediating the exendin-4–induced effects. Thus, we performed a screen for differentially expressed TGF-β isoform ligands and activin ligands, as well as potential type I receptors for those ligands. The screen was performed in AR42J cells, with and without exendin-4 treatment (Fig. 6). From this screen, it appeared that the activin monomers (inhibin A and B) were minimally, if at all, expressed in exendin-treated AR42J cells. TGF-β1 and -β2, however, were expressed, and TGF-β1 was upregulated with exendin-4 treatment. (It was essentially not expressed in untreated AR42J cells.) In addition, the TGF-β type I receptor (alk5), which phosphorylates and activates both smad 2 and 3, was clearly expressed in AR42J cells, either with or without exendin-4 treatment. Activin-binding type I receptors were present (data not shown), but TGF-β isoforms do not typically bind to activin type I receptors (4).

Therefore, we next tested the possibility that TGF-β isoform ligands were mediating the exendin-4–induced insulin-positive differentiation. A specific, well-characterized TGF-β pan-neutralizing antibody was able to almost completely abolish exendin-4–induced insulin-positive differentiation (Figs. 6B–G), as measured by insulin II mRNA, pdx-1 mRNA, and pax4 mRNA, suggesting that TGF-β isoform signaling was necessary for mature β-cell formation. Interestingly, both smad 2 and 3 mRNA levels were decreased, suggesting that exendin-4 inhibits smad 2 independent of TGF-β isoform signaling but that the secondary elevation in smad 3 levels seen with exendin-4 treatment, as well as the insulin-positive response, are TGF-β isoform dependent. Bone morphogenic protein (BMP) smads (smad 1, 5, and 8) were unaffected in the TGF-β pan-neutralizing antibody-treated cells. An irrelevant antibody against the transcription factor HNF-3β (data not shown) had no effect on any of these parameters.

In an effort to both confirm the specific role of TGF-β isoform signaling and better delineate the overall signaling mechanisms, we used antisense against the TGF-β type I receptor (alk5). The alk5 antisense had a dramatic inhibitory effect on exendin-4 induction of insulin, pdx-1, and pax4 mRNA (Fig. 7). Alk5 antisense specifically blocked the upregulation of smad 3, suggesting that alk5 is a type I receptor necessary for inducing smad 3 transcription. No apparent effect on the inhibition of smad 2 was seen, suggesting that smad 2 may not be critically regulated by alk5.

To determine whether the TGF-β isoforms may play a synergistic role with exendin-4, and perhaps with GLP-1, in the induction of insulin-positive differentiation and insulin expression, exogenous TGF-β1 was added to AR42J cells with and without exendin-4 treatment (Fig. 8). Here there was a dramatic increase in insulin mRNA levels when 10 ng/ml TGF-β1 ligand was added with 5 pmol/l exendin-4. TGF-β1 synergistically increased the insulin II mRNA levels by 10- to 15-fold over exendin-4–induced insulin alone. Pax4, IAPP, and pdx-1 were similarly elevated. TGF-β1 ligand alone essentially had no proinsulin effect on the AR42J cells, and higher doses of TGF-β1 ligand (100 ng/ml) with exendin-4 were inhibitory. Smad 3 levels were elevated with 10 ng/ml TGF-β1 treatment in the exendin-treated AR42J cells, suggesting that smad 3 may be a key mediator in regulating the maturation of insulin-positive cells.

Understanding the mechanisms by which precursor cells can be induced to differentiate into insulin-positive β-cells is critical if we are to successfully engineer cells for the treatment of diabetes. Over the last few years, two extracellular signaling pathways have been shown to play a role in the induction of insulin-positive differentiation in several cell systems. GLP-1–induced insulin-positive differentiation in certain transformed cell lines (7–15) and some TGF-β superfamily signaling molecules have been found to control endocrine and insulin-positive differentiation in some pancreatic cell lines and tissues (7–15,17,18). Of particular interest, both GLP-1 and a combination of activin and either HGF or β-cellulin have the capacity to induce the pancreatic exocrine cell line, AR42J, to differentiate into insulin-positive endocrine cells (7–15). Here we found that there is an interaction of the two signaling pathways to induce insulin-positive differentiation. The interaction was demonstrated by the fact that inhibition of TGF-β signaling at any of several levels blocked GLP-1–induced insulin-positive differentiation, whereas TGF-β ligands alone were unable to induce insulin-positive differentiation in the absence of exendin stimulation. Further interactions were implied by the synergistic effect of low-dose exogenous TGF-β ligands with exendin-4.

The canonical TGF-β signaling pathway entails phosphorylation of receptor-specific smads and subsequent translocation to the nucleus of a smad heteromeric complex that directly activates gene expression (4,5). However, several studies have documented complex intracellular interactions of other pathways on smad signaling pathways. In particular, the G-protein signaling pathways (of which GLP-1 receptor is included) have been shown to significantly interact with smad signaling (19). Also, notch signaling (20) and the JAK/STAT pathway (21) have significant effects on smad pathways. In many cases, these pathways can act synergistically with the smads, as may be occurring here in AR42J cells.

We found that smad 2 expression was necessary for exendin-4–induced insulin-positive differentiation but also that smad 2 expression was strongly downregulated with exendin-4. The necessity for smad 2 is consistent with earlier findings by Zhang et al. (6) that smad 2 was necessary for the endocrine commitment of AR42J cells induced by activin alone. Therefore, it seems plausible that while smad 2 expression is necessary for endocrine lineage commitment, further differentiation requires downregulation of smad 2. Zhang et al. found that with activin stimulation alone, there was persistent expression and accumulation of phosphorylated smad 2, resulting in apoptosis.

Smad 2 and 3 are well known to act competitively in several differentiation systems (22–24). We found that smad 3 antisense resulted in a greatly increased number of insulin-positive cells and in increased insulin mRNA but without a concomitant increase in expression of β-cell–specific markers, such as IAPP. Based on these data, it appears that smad 3 may support maturation of β-cells and inhibition of proliferation but only after smad 2 has initiated a proendocrine differentiation program. This possibility is supported by the fact that the smad 3 antisense had no effect if cells were also treated with smad 2 antisense, suggesting that smad 2–mediated effects are a necessary preparation for smad 3 actions. A decrease in smad 2 levels in the absence of any change in smad 4 may be necessary to allow smad 3 binding to smad 4. Also, it appears that the drop in smad 2 levels may be directly related to GLP-1 signaling, since the drop occurred with exendin-4 treatment alone, even in the presence of TGF-β neutralizing antibodies. The mechanism by which exendin-4 may lead to smad 2 downregulation is unclear.

One logical area of interplay between the GLP-1 and TGF-β signaling systems may be the induction of TGF-β expression and secretion by AR42J cells to act by either autocrine or paracrine pathways to initiate secondary aspects of insulin-positive differentiation. This role for TGF-β would be consistent with known effects of TGF-β in cultured embryonic pancreas (17). It seems unlikely, however, that GLP-1 plays an important role in the normal endogenous differentiation of early endocrine cells because GLP-1 is not expressed significantly in the embryonic pancreas until later in gestation (10).

The observed synergistic effect of low-dose TGF-β ligands with exendin-4 is of interest, not only with respect to potential cellular engineering applications but also because of implications regarding mechanisms of differentiation. The combination of TGF-β ligand and exendin-4 gave low smad 2 and high smad 3 levels as well as high insulin and high pdx-1, suggesting the presence of mature β-cells. Smad 3 antisense, however, gave high levels of insulin but not pdx-1. From these data, we conclude that smad 3 plays a role in the maturation of β-cells with the acquisition of pdx-1 expression. GLP-1 appears to be necessary for the high levels of insulin, but the intracellular molecules downstream of TGF-β signaling necessary for this synergistic action remain to be determined.