Cross-talk between Bone Morphogenetic Protein and Transforming Growth Factor- (cid:1) Signaling Is Essential for Exendin-4-induced Insulin-positive Differentiation of AR42J Cells

A key goal of cellular engineering is to manipulate progenitor cells to become (cid:1) -cells, allowing cell replacement therapy to cure diabetes mellitus. As a paradigm for cell engineering, we have stud-iedthemolecularmechanismsbywhichAR42Jcellsbecome (cid:1) -cells. Bone morphogenetic proteins (BMPs), implicated in a myriad of developmental pathways, have not been well studied in insulin-pos-itive differentiation. We found that the canonical intracellular mediatorsofBMPsignaling,Smad-1andSmad-8,weresignificantly elevated in AR42J cells undergoing insulin-positive differentiation in response to exendin-4 treatment, suggesting a role for BMP signaling in (cid:1) -cell formation. Similarly, endogenous BMP-2 ligand and ALK-1 receptor (activin receptor-like kinase-1; known to activate Smads 1 and 8) mRNAs were specifically up-regulated in exendin-4-treated AR42J cells. Surprisingly, Smad-1 and Smad-8 levels were suppressed by the addition of BMP-soluble receptor inhibition of BMP ligand binding to its receptor. Here, insulin-pos-itive differentiation was also ablated. BMP-2 ligand antisense also strongly inhibited Smad-1 and Smad-8 expression, again with the abolition of insulin-positive enteffectsatdifferentconcentrations,arescue-typeexperimentwasperformed with three different doses of exogenous BMP-2 after BMP-2 antisense treat-menttodeterminethepotentialforeitherarescueeffectorasupraphysiologic effect and to confirm the specificity of the BMP antisense experiments. The rescuedoseof10pg/mlBMP-2showedacompleterescueevenbeyondthatof the missense controls, suggesting that endogenous BMPs are likely to be important. However, the lowest tested dose of 1 pg/ml BMP-2 does not get a rescue effect (data not shown). Interestingly, higher concentrations did not have an effect either. The loss of effect at higher and lower doses does suggest thattheBMP-2effectisaspecificandimportantendogenouseffectratherthan asimplepharmacologiceffect(Fig.6).

Type 1 diabetes is an insulin deficiency state due to pancreatic destruction of ␤-cells caused by autoimmunity. Several approaches to treat diabetes are being pursued, such as islet cell transplantation, pancreas transplantation, and genetic manipulation. However, a key alternative strategy is cellular engineering to manipulate progenitor cells to become ␤-cells, allowing cellular therapy to cure diabetes. As a paradigm for cell engineering, we have used exendin-4 treatment of AR42J cells, a fairly plastic acinar cell carcinoma-derived cell line, as a model for studying the role of bone morphogenetic protein (BMP) 2 signaling in the induction of insulin-positive differentiation. Exendin-4, a peptide from Helodermatidae venom, is a novel insulinotropic agent and a long acting analogue of glucagon-like peptide-1 (GLP-1). It interacts with endocrine pancreatic islet GLP-1 receptors, inducing a stimulatory effect on insulin secretion. Over the past few decades significant progress has been made in our understanding of the biological function of BMPs, which have been found to regulate a myriad of developmental and differentiation process in the embryo, including epithelial-mesenchymal interactions, cell fate specification, dorsoventral patterning, and apoptosis as well as the secretion of extracellular matrix components (1)(2)(3)(4)(5).
BMPs are one of the multifunctional cytokines from the transforming growth factor-␤ (TGF-␤) superfamily. The TGF-␤ isoforms proper (TGF-␤1, -␤2, and -␤3) are also included in this superfamily. The canonical pathway for the pleiotropic biological effects of BMPs is signaling through to the nucleus via ligand-induced hetero-oligomerization of type I (activin receptor-like kinases ALK-2, ALK-3, and ALK-6) and type II serine/threonine kinase receptors (BMP receptor type II and the activin receptors ActR-IIA and ActR-IIB) and their downstream effectors, known as Smad-1, Smad-5, and Smad-8 (6 -9). Upon ligand stimulation, these receptor-regulated Smads become phosphorylated by activated type I receptor kinases and form heteromeric complexes with the shared common mediator Smad-4. Subsequently, these Smad complexes translocate into the nucleus, where they regulate the transcription of target genes (10,11).
Recently, we have reported that exendin-4-induced differentiation of the ␤-cell-like phenotype in AR42J cells requires TGF-␤ isoform (TGF-␤1, -␤2, and -␤3) signaling initially in the form of Smad-2 and followed by Smad-3. Smad-3 appears to play a secondary role in suppressing further increases in insulin mRNA and may facilitate ␤-cell-like maturation (12). Given the frequent interplay of function between TGF-␤ isoform signaling and BMP signaling, here we studied the possible role of BMP signaling in exendin-4-induced differentiation of AR42J cells into insulin-positive cells.
We observed a hierarchy of BMP signaling to TGF-␤ isoform signaling. Taken together with previous findings, our results suggest that GLP-1 stimulation of insulin-positive differentiation in AR42J cells acts first via BMP ligands and Smads, followed by TGF-␤ isoform signaling.

MATERIALS AND METHODS
Reagents and Kits-Exendin-4 was obtained from Sigma-Aldrich. The RNeasy mini kit, the Sensiscript reverse transcriptase kit, and the QIAquick gel extraction kit were all from Qiagen (Valencia, CA). Am-pliTaq Gold with GeneAmp 10ϫ PCR Buffer and MgCl 2 solution was from Applied Biosystems (Foster City, CA). The F-12K nutrient mixture (Kaighn's modification) was from Invitrogen.
Antibodies/Ligands-BMP-soluble receptor 1B (BMP-R1B) antibodies from Sigma-Aldrich were at a concentration of 3 g/ml, which is defined as the effective concentration for inhibiting alkaline phosphatase production. In additional experiments, BMP-2 (obtained from R&D Systems, Minneapolis, MN) was measured by its ability through three different doses (10 pg/ml, 100 pg/ml, and 1 ng/ml) to rescue or augment the system. TGF-␤1, purchased from Sigma-Aldrich, was added at a final concentration of 10 ng/ml, which was chosen with reference to the previous synergistic effect data as those required to enhance TGF-␤ activity.
Cell Culture and Treatment-AR42J cells, purchased from American Type Culture Collection (Manassas, VA), were grown in Kaighn's modification of Ham's F-12K medium with 2 mM L-glutamine, 250 g/ml amphotericin, 100 units/ml penicillin, 100 g/ml streptomycin, and 20% fetal bovine serum at 37°C under a humidified condition of 95% air and 5% CO 2 . Cells were plated at a density of ϳ10 5 cells/ml in 12-well plates. Morpholino antisense or missense control was added separately to culture media at 20 M. Cells were then cultured with exendin-4 at doses of 1, 5, and 10 pM for 3 days. . Simple PCR to screen for BMP ligands and potential type I receptors. ALK-1, a type I receptor that activates BMP Smad-1 and Smad-5, was not present in unstimulated AR42J cells but was strongly present after 3 days of exendin-4 stimulation in AR42J cells. However, ALK-2, ALK-3, and ALK-6 were clearly present in the base-line AR42J cells, but mRNA levels appeared to diminish after exendin-4 stimulation. In addition, BMP-2 was absent in AR42J cells but became present with exendin-4 stimulation. BMP-4 and BMP-7 were present in both unstimulated and exendin-4-stimulated AR42J cells.
Reverse Transcription PCR (Non-quantitative)-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. All PCR products were separated by electrophoresis in 2% agarose gel. The 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.
SYBR Green Real Time Quantitative PCR-PCR amplifications were performed using a Bio-Rad iCycler (Hercules, CA) sequence detection system. Reactions were performed in a 50-l reaction mixture that included 10ϫ AmpliTaq Gold buffer, 25 mM MgCl 2 , 2.5 mM dNTPs, 10ϫ SYBR Green, AmpliTaq Gold polymerase, distilled H 2 O, DNA template, and 10 M each primer. Amplification was performed by initial polymerase activation for 10 min at 95°C and 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.
Western Blot Analysis-Proteins were separated on a 10% Tris-HCl Ready Gel (Bio-Rad), transferred onto nitrocellulose membranes, and incubated with a ␤-actin antibody (Abcam, Cambridge, MA) at a dilution of 1:5000, a Smad-1 antibody (Upstate Biotechnology) at 4 g/ml, an ALK-1 antibody (Abcam) at 0.2 g/ml, or a BMP-2 antibody (Abcam) at 2 g/ml overnight at 4°C. After incubation, the membranes were washed twice for 15 min in washing buffer (phosphate-buffered saline and 0.05% Tween 20) and incubated with a secondary anti-mouse (␤-actin/BMP-2), anti-rabbit (Smad-1), or anti-goat (ALK-1) antibody coupled to horseradish peroxidase (Vector Laboratories, Burlingame, CA) for 1 h at room temperature. The membranes were then washed three times for 15 min in washing buffer, and immunoreactivity was normalized by chemiluminescence (Amersham Biosciences ECL Plus kit, RPN2132) according to the manufacturer's instructions.

RESULTS AND DISCUSSION
Involvement of BMP Smads in Insulin-positive Differentiation-We found previously (12) that TGF-␤ isoforms and their Smads (-2 and -3) were essential for mature ␤-cell formation from AR42J cells treated with exendin-4 (12). Thus, we also wished to study a potential role for BMP signaling in the induction of this insulin-positive differentiation. Because Smad-1, Smad-5, and Smad-8 are receptor-regulated Smads in the BMP signaling pathway, we first examined the possible involvement of those Smads in the insulin-positive differentiation of AR42J cells induced by exogenous exendin-4. Interestingly, Smad-1 and Smad-8 mRNA levels were greatly elevated in response to exendin-4 treatment, whereas Smad-5 mRNA levels dropped precipitously from the base line (Fig. 1). These changes are reminiscent of those seen for Smad-3 and Smad-2, respectively (12).
The key result was that ALK-1 was up-regulated from undetectable at base line to strongly positive in the exendin-4-treated AR42J cells. ALK-2, ALK-3, and ALK-6 were present at base line and then decreased with the exendin-4. BMP-2 was negative at base line and then up-regulated with the exendin-4 treatment. BMP-4 and BMP-7 were expressed in both base-line and exendin-4-treated cells. It is possible that BMP-2, which turns on in response to the stimulus, could then form a heterodimer with BMP-4 or BMP-7 and become stimulatory, possibly signaling through ALK-1. BMP-4 or BMP-7 alone may be inhibitory. (Fig. 2) We then used BMP-soluble receptors as BMP ligand inhibitors to study a potential role for endogenous BMP signaling in exendin-4-induced insulin-positive differentiation of AR42J cells. BMP soluble receptors inhibited insulin expression and also blocked PDX-1 and Pax-4 expression similarly as in our previous findings with TGF-␤-neutralizing antibodies (Fig. 3A-C) (12). Then, mRNA levels of Smad-1, Smad-2, Smad-3, Smad-5, and Smad-8 in the BMP-soluble receptortreated cells were tested (Fig. 3, D-H). There was suppression of Smad-1 and Smad-8 mRNA, implying that the soluble receptors had blocked BMP downstream signaling (Fig. 3, D and H). Prevention of the normal rise in Smad-3 mRNA that occurs in response to TGF-␤ isoform signaling (Fig. 3F) suggests that BMP signaling is acting upstream of TGF-␤ isoform signaling. In Smad-2 and Smad-5, the decrease that normally occurs with exendin-4 was not affected (Fig. 3, E and G).
ALK-1 is an activator of BMP Smads, although it can bind TGF-␤ isoforms and Müllerian inhibiting substances. We used ALK-1 morpholino antisense and blocked insulin-positive differentiation and the accompanying PDX-1 and Pax-4 up-regulation (Fig. 4, A-C). Elevations in Smad-1 and Smad-8 mRNA were again inhibited (Fig. 4, D and H), whereas Smad-5 mRNA levels were barely affected (Fig. 4G). Interestingly, Smad-2 suppression was unaffected (Fig. 4E), but elevations in Smad-3 mRNA levels were blocked (Fig. 4F). These results suggest that ALK-1 morpholino blocks BMP Smad activation, which then leads secondarily to inhibition of the elevation of Smad-3 mRNA levels. These results would again imply that TGF-␤ isoform signaling is downstream of BMP signaling. To confirm the effects of the morpholino ring antisense on Smad levels, Western blotting was performed in mor-pholino ring antisense-treated and sense-treated controls, confirming sequence-specific effects in target Smads by antisense (Figs. 4I, 5I,  and 7H).
To test a potential role for the endogenous ligand BMP-2 in insulinpositive differentiation as suggested by the screening PCR in Fig. 2, BMP-2 morpholino antisense was used. Here again we saw complete suppression of insulin II, PDX-1, and Pax-4 mRNA (Fig. 5, A-C), consistent with the soluble BMP receptor data shown earlier. For the Smads, there was a strong suppression of Smad-1 and Smad-8 mRNA levels (Fig. 5, D and H), but no effect on Smad-2 and Smad-5 suppression (Fig. 5, E and G). Smad-3 was suppressed back to base line (Fig. 5F), which is again consistent with TGF-␤ isoform signaling being downstream of BMP signaling.
Rescue Effects of Exogenous BMP-2 on Insulin-positive Differentiation-Because BMP is known to act through different receptors and can exert differ- enteffectsatdifferentconcentrations,arescue-typeexperimentwasperformed with three different doses of exogenous BMP-2 after BMP-2 antisense treatment to determine the potential for either a rescue effect or a supraphysiologic effect and to confirm the specificity of the BMP antisense experiments. The rescue dose of 10 pg/ml BMP-2 showed a complete rescue even beyond that of the missense controls, suggesting that endogenous BMPs are likely to be important. However, the lowest tested dose of 1 pg/ml BMP-2 does not get a rescue effect (data not shown). Interestingly, higher concentrations did not have an effect either. The loss of effect at higher and lower doses does suggest that the BMP-2 effect is a specific and important endogenous effect rather than a simple pharmacologic effect (Fig. 6).
Smad-1 Is Necessary in the Hierarchy of ␤-Cell Differentiation-Initially, ALK-1 mRNA was found to be up-regulated with exendin treatment in AR42J cells (Fig. 2). ALK-1 mainly works through the phosphorylation of Smad-1 and Smad-5, but not Smad-8. Because Smad-1 mRNA was up-regulated and Smad-5 down-regulated (Fig. 1), Smad-1 seemed to be a likely candidate in mediating ALK-1 effects. We found that there was a strong inhibition of insulin II, PDX-1, and Pax-4 mRNA with Smad-1 antisense (Fig. 7, A-C). In all cases, when insulin mRNA levels were reduced a concomitant block of the key pro-insulin transcription factors PDX-1 and Pax-4 was also seen. Interestingly, the typical changes in Smad-2 and Smad-3 mRNA levels seen with exendin-4 treatment were not seen with Smad-1 antisense (Fig. 7, D and E), consistent with our theory that BMP signaling is upstream of TGF-␤ isoform signaling. Additionally, even with TGF-␤ blockage, Smad-2 was always down-regulated previously whenever AR42J cells were exposed to exendin-4. The failure of Smad-2 mRNA expression levels to go down despite exendin-4 suggests that the mechanism for Smad-2 down-regulation is through BMP-mediated Smad-1 activation. With the Smad-1 antisense, although the Smad-8 mRNA levels were blocked (Fig. 7G) the Smad-5 mRNA levels did not go down (Fig. 7F), consistent with the possibility that there is a direct competition between Smad-1 and Smad-5 similar to that suggested earlier for Smad-2 and Smad-3 (12).
The Hierarchy between BMP and TGF-␤ Isoform Signaling-Loss of insulin-positive differentiation with Smad-1 antisense could be rescued with exogenous TGF-␤1, suggesting that BMP or Smad-1 signaling is upstream of TGF-␤ isoforms (Fig. 8A-C). Also, these results suggest that a key role of Smad-1 is to activate, directly or indirectly, TGF-␤ isoform signaling. Smad-2 and Smad-5 were elevated with the Smad-1 antisense (Fig. 7, D and F), but that effect was mostly reversed with the addition of exogenous TGF-␤1 (Fig. 8, D and F).
Conclusion-Understanding the mechanisms of the TGF-␤ superfamily signaling cascade in the possible regulation of insulin-positive differentiation has high importance for our goal of engineering glucose- regulated insulin-producing cells for the treatment of diabetes mellitus. Our results show that there is a novel synergistic activity of TGF-␤ and BMP Smad proteins as part of an intracellular signaling pathway that plays an important role during insulin-positive differentiation of AR42J cells. Based on the observed patterns of Smad protein expression, Gupta and co-workers (13) speculated that specific individual or combinations of Smads may play different roles in lineage selection during kidney development. The expression of various Smad proteins has been shown in pancreatic islets and in glucagon cells by reverse transcription-PCR and by immunostaining (14).
BMPs and TGF-␤s have been strongly implicated in many embryologic and cell differentiation pathways. Jiang et al. (15) showed that BMP induces embryonic pancreatic epithelia to form insulin-positive cell colonies. Thus, we hypothesized that BMP signaling may play a role in the insulin-positive differentiation of AR42J cells. Our results support the idea that BMP-2 ligands play an important endogenous role in this process and that BMPs sit high on a hierarchy above TGF-␤ isoforms in the initiation of insulin-positive differentiation.
Previously, we quantified BMP Smads, namely Smad-1, Smad-5, and Smad-8, in TGF-␤ neutralizing antibody-treated AR42J cells and found that there were essentially no changes in any of the BMP Smads (12), which is consistent with TGF-␤ signaling being downstream of BMP signaling. Here, with the BMP soluble receptor we showed that the Smad-2 decrease that normally occurs with exendin-4 treatment was not affected, whereas the elevation in Smad-3 that occurs in response to exendin-4 was completely blocked. These results suggest a novel overlap of BMP and TGF-␤ isoform pathways with BMP upstream of TGF-␤ isoform signaling, which, in turn, appears to be specifically responsible for Smad-3 up-regulation. Based on these findings, we hypothesize that BMP may be necessary to stimulate these precursor cells to become endocrine-committed progenitor cells that then proceed further to differentiate under the influence of Smad-3 into mature ␤-cells. Overall, there appears to be a synergistic interaction of BMP and TGF-␤ signaling to push AR42J cells toward an insulin-positive fate.
To understand the role of Smad-1 in mouse development, Robertson and co-workers (16) generated a Smad-1 null mutant mouse. They found an essential role for Smad-1-dependent signals in primordial germ cell specification (16), but because of early lethality there was no potential for analysis of pancreatic development. Our results with Smad-1 antisense and the rescue of Smad-1 antisense by exogenous TGF-␤1 suggest that BMP signaling activates Smad-2 and Smad-3 path-ways, possibly through the induction of the release of TGF-␤ isoforms from the cells in a paracrine or autocrine manner.
It was interesting that the BMP-soluble receptor did not block Smad-2 down-regulation, whereas Smad-1 antisense did. These results suggest that BMP ligand-independent pathways, which may also activate Smad-1, may play a more important role in the decrease of Smad-2 levels. We found previously that Smad-2, which is highly expressed in untreated AR42J cells, was necessary for insulin-positive differentiation.
A role for Smad-5 down-regulation in insulin-positive differentiation is also unclear. The Hammerschmidt group (17) has shown distinct roles for Smad-1 and Smad-5 during dorsoventral patterning of the zebrafish embryo. Their data suggest that Smad-1 acts later than Smad-5 and is itself a transcriptional target of Smad-5-mediated BMP-2 signaling (17). Thus, the decrease in Smad-5 may disinhibit or otherwise allow Smad-1 to up-regulate and/or become activated.