p38 MAPK is involved in activin A- and hepatocyte growth factor-mediated expression of pro-endocrine gene neurogenin 3 in AR42J-B13 cells.

Neurogenin3 (ngn3) is a transcription factor that is essential for the differentiation of pancreatic endocrine cells. To investigate the signaling pathway that regulates ngn3 expression, we used AR42J-B13 cells as a model of the differentiation of pancreatic islets. In these cells, treatment with activin A and hepatocyte growth factor (HGF) induced the expression of ngn3. Reporter gene analysis using human ngn3 gene (NEUROG3) promoter fragments of various lengths identified the region between -402 and -327 bp of NEUROG3 as an activin A- and HGF-responsive DNA sequence. This DNA sequence normally functions as a repressor in AR42J-B13 cells, but treatment with activin A and HGF negates the repressor activity. Interestingly, function of the activin A- and HGF-responsive sequence was not influenced by the overexpression of the Smad inhibitory factor, Smad7. Instead, activin A and HGF activation was inhibited by overexpression of a dominant-negative mutant of transforming growth factor-beta-activated kinase 1 (TAK1), or mitogen-activated protein kinase kinase 3 (MKK3), or by treatment with a p38 MAPK-specific inhibitor, SB203580. Activin A and HGF function through the TAK1-MKK3-p38 MAPK pathway to relieve transcription repressors located between -402 and -326 bp on the NEUROG3 promoter, and consequently activate ngn3 expression and endocrine differentiation of AR42J-B13 cells.

Neurogenin3 (ngn3) is a transcription factor that is essential for the differentiation of pancreatic endocrine cells. To investigate the signaling pathway that regulates ngn3 expression, we used AR42J-B13 cells as a model of the differentiation of pancreatic islets. In these cells, treatment with activin A and hepatocyte growth factor (HGF) induced the expression of ngn3. Reporter gene analysis using human ngn3 gene (NEUROG3) promoter fragments of various lengths identified the region between ؊402 and ؊327 bp of NEUROG3 as an activin Aand HGF-responsive DNA sequence. This DNA sequence normally functions as a repressor in AR42J-B13 cells, but treatment with activin A and HGF negates the repressor activity. Interestingly, function of the activin Aand HGF-responsive sequence was not influenced by the overexpression of the Smad inhibitory factor, Smad7. Instead, activin A and HGF activation was inhibited by overexpression of a dominant-negative mutant of transforming growth factor-␤-activated kinase 1 (TAK1), or mitogen-activated protein kinase kinase 3 (MKK3), or by treatment with a p38 MAPK-specific inhibitor, SB203580. Activin A and HGF function through the TAK1-MKK3-p38 MAPK pathway to relieve transcription repressors located between ؊402 and ؊326 bp on the NEUROG3 promoter, and consequently activate ngn3 expression and endocrine differentiation of AR42J-B13 cells.
During development, formation of the pancreas and its subsequent differentiation into various types of exocrine and endocrine cells are controlled by the activation and extinction of a large number of genes. The expression of these genes is regulated by a well-organized cascade of transcription factors (1,2), and the basic helix-loop-helix factor neurogenin 3 (ngn3) 1 is one of the essential transcription factors expressed early in the differentiation of islet cells (3)(4)(5).
In mice, the pancreas forms from evaginated buds generated by fusion of the dorsal and ventral gut endoderm at the foregut/ midgut junction on embryonic days 9.5-10.5. At this stage, most of the cells located in the buds have the potential to differentiate into any of the mature pancreatic cell types (6). The signal pathway from Notch and notch ligand delta-like 1 participates in determination of the fate of cells in the exocrine pancreas (4). It is believed that cells eluding those signals start to express ngn3 and become endocrine cells. However, expression of ngn3 is not observed after the cells begin to express pancreatic hormones like glucagon or insulin (5,6). Thus, expression of ngn3 is restricted to immature endocrine cells and is considered to be a marker of endocrine progenitor cells during development of the pancreas.
Interestingly, ngn3 is not just a marker of endocrine progenitor cells. When targeted disruption of ngn3 was performed in mice, the animals failed to generate pancreatic endocrine cells and did not express most of the transcription factors that are essential for endocrine differentiation (7). Conversely, the overexpression of ngn3 in pancreatic bud cells promoted differentiation into endocrine cells (4,5). These findings indicate that ngn3 is an essential and sufficient factor for the differentiation of endocrine cells. Thus, investigation of the mechanisms regulating the expression of ngn3 is important to better understand the process of endocrine cell differentiation and probably also to understand the process of ␤ cell regeneration.
In addition to the Notch-delta signaling pathway, several studies have suggested the importance of activins, which are members of the transforming growth factor-␤ (TGF-␤) family, in signaling for development of the pancreas (8 -10). Activins have diverse actions in a variety of mammalian tissues, including the pancreas, at many stages during development. It was recently reported that activin receptor type II (ActRII) A ϩ/Ϫ B Ϫ/Ϫ mice showed a severe reduction of insulin and glucagon expression, but only a slight reduction of carboxypepti- HNF-1␣, hepatocyte nuclear factor-1␣; CREB, cAMP-responsive element-binding protein; TGF-␤, transforming growth factor-␤; TAK1, TGF-␤-activated kinase 1; Erk, extracellular signaling-regulated kinase; MAPK, mitogen-activated protein kinase; MKK, MAPK kinase; JNK, c-Jun N-terminal kinase; ActRI, activin receptor type I; ActRII, activin receptor type II; STAT, signal transducer and activator of transcription; RE1, repressor element 1; NRSE, neuron restrictive silencer element; PAI-1, plasminogen activator inhibitor-1; EMSA, electrophoretic mobility shift assay; DMEM, Dulbecco's modified Eagle's medium; PVDF, polyvinylidene difluoride. dase A (a marker of exocrine pancreas), as early as embryonic day 13.5 (8). These results suggest that activin signals might be involved in determining the fate of cells in the pancreatic primordium, but the signaling pathways downstream of the activin receptor that are involved in the differentiation of endocrine cells have not been investigated.
AR42J cells were originally derived from a chemically induced pancreatic tumor and possess both exocrine and neuroendocrine properties. Upon treatment with activin A, AR42J cells stop growing and start to extend neurites. In the presence of betacellulin or hepatocyte growth factor (HGF) plus activin A, the cells differentiate into insulin-expressing cells (11). Thus, AR42J cells provide a model system for studying the molecular mechanisms involved in ␤ cell differentiation. Probably reflecting the physiological differentiation process, activin A and HGF can also provoke the expression of ngn3 (12).
In the present study, we investigated the mechanism by which activin A and HGF induce the expression of ngn3 using AR42J-B13 cells as a model for the differentiation of pancreatic islets. Our data revealed that the TGF-␤-activated kinase 1 (TAK1)-mitogen-activated protein kinase kinase 3 (MKK3)-p38 mitogen-activated protein kinase (p38 MAPK) pathway regulates the function of a critical region on the human ngn3 gene (NEUROG3) promoter and changes the binding of various transcription factors to the region.

EXPERIMENTAL PROCEDURES
Reporter Gene Constructs and Expression Plasmids-To generate reporter plasmids, fragments of the 5Ј region of NEUROG3 obtained by restriction digestion or PCR using the pFOX Luc1 NEUROG3-5.7kb plasmid (13) as a template were ligated upstream of the luciferase gene FIG. 1. Activin A-and HGF-induced transactivation of the NEUROG3 promoter. A, reporter plasmids containing a fragment of Ϫ6 kbp of NEUROG3 promoter linked to the luciferase gene (14) were transfected into AR42J-B13 cells. Then these cells were incubated without activin A or HGF, with activin A alone, with HGF alone, and with both activin A and HGF. At 36 h after transfection, cells were harvested and luciferase assays were performed. The luciferase activity was standardized by the protein concentration. Relative luciferase activity was calculated by setting as 1 the activity of cells transfected with the pFOX Luc1 ngn3-6kbp plasmid, without exposure to activin A and HGF. B, reporter plasmids constructed with the various indicated NEUROG3 fragments inserted upstream of the luciferase gene were transfected into AR42J-B13 cells. Then these cells were incubated with (solid bars) or without (open bars) activin A and HGF. Relative luciferase activity was calculated by setting as 1 the activity of cells transfected with the pFOX Luc1 (pFLO) plasmid alone, without exposure to activin A and HGF. All data are shown as the means Ϯ S.E.

FIG. 2.
A, structure of pFLO-TK Ngn3 from Ϫ402 to Ϫ305 bp is shown. B, pFLO-TK (open bars) and pFLO-TK Ngn3 from Ϫ402 to Ϫ305 bp (solid bars) were co-transfected with the pRL-TK plasmid into NIH3T3 cells, ␤TC cells, and AR42J-B13 cells. Relative luciferase activity was calculated by setting as 1 the activity of cells transfected with the control plasmid, without exposure to activin A and HGF. All data are shown as the mean Ϯ S.E. in pFOX Luc1 plasmids or upstream of the thymidine kinase (TK) minimal promoter in pFOX Luc1TK plasmids (14). Mutagenesis of the reporter gene constructs was performed using a Quick Change sitedirected mutagenesis kit according to the manufacturer's instructions (Stratagene, La Jolla, CA). All constructs were confirmed by sequencing. An expression plasmid for dominant-negative MKK6 (SRalpha-HA-MKK6AA) was kindly provided by Dr. E. Nishida (Kyoto University, Kyoto, Japan). Expression plasmids for dominant-negative MKK3 (pCS2ϩDN MKK3) and dominant-negative MKK4 (pCS2ϩDN MKK4) were kindly provided by Dr. K. Sugiyama (Nippon Boehringer Ingelheim Co., Kawanishi, Japan) and Dr. M. Hibi (Center for Developmental Biology, RIKEN, Kobe, Japan). Expression plasmids for TAK1 (pCMV-HA-TAK1) and dominant-negative TAK1 (pCMV-HA-TAK1 (K63W)) were kindly provided by Dr. K. Matsumoto (Nagoya University, Nagoya, Japan).
Cell Culture and Transfection-AR42J-B13 cells were cultured in low glucose Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% FCS, 20 mM HEPES, and 5 mM NaHCO 3 . To cause conversion into insulin-producing cells, we treat AR42J-B13 cells with 2 nmol/l activin A (a generous gift from Dr. A. Okano; Ajinomoto Co. Kawasaki, Japan) and 100 pmol/l HGF (R&D Systems, Minneapolis, MN). NIH3T3 cells were cultured in DMEM supplemented with 10% calf serum. ␤TC3 cells (a generous gift from Dr. D. Hanahan, University of California, San Francisco, CA) were cultured in DMEM supplemented with 2.5% fetal bovine serum and 15% horse serum. For reporter gene assays, cells were plated in six-well tissue culture plates at 24 h before transfection. AR42J-B13 cells, with or without exposure to activin A and HGF, were transfected with 1.8 g of each luciferase reporter plasmid using Effectene reagent (Qiagen, Tokyo, Japan) according to the manufacturer's protocol. To assess the effect of the Smador MAPK-related signal pathways on the NEUROG3 promoter, we co-transfected 1 g of each expression vector with 1.8 g of the reporter gene vector. Controls for the experiments using dominant-negative expression vectors always contained equal amounts of the empty vector (pcDNA3). At 36 h after transfection, the cells were harvested and the luciferase assay was performed as described previously (15). Luciferase activity was standardized by the protein concentration, which was measured using a Bio-Rad protein assay kit (Bio-Rad Laboratories, Tokyo, Japan). NIH3T3 cells and ␤TC3 cells were transfected with 1.8 g of each luciferase reporter plasmid and 0.2 g of pRL-TK plasmid using Trans Fast transfection reagent (Promega, Tokyo, Japan). At 36 h after transfection, cells were harvested and analyzed for firefly luciferase activity, which was standardized by Renilla luciferase activity. All transfection experiments were performed in triplicate on at least three separate occasions.
Electrophoretic Mobility Shift Assay-Nuclear extracts from AR42J-B13 cells seeded in 10-cm dishes were prepared according to the procedure described by Sadowki and Gilman (16). Single-stranded oligo-nucleotides corresponding to the sequences in the NEUROG3 promoter were 5Ј-end-labeled with [␥-32 P]ATP using T4 polynucleotide kinase. The labeled oligonucleotide was purified on a column and annealed to an excess of the complementary strand. The EMSA buffers and electrophoresis conditions have been described previously (17). One microliter of in vitro reaction mixture or 1 g of nuclear extract were used for each 10-l binding reaction. The mixture was incubated for 15 min at room temperature before polyacrylamide gel electrophoresis, and radioactive bands were visualized using Fuji BAS2500 (Fujifilm, Tokyo, Japan). The oligonucleotides used in these experiments are as follows. Region 1, 5Ј-CAA AGT CCC CTC CAG GAC AGA TGC TAA GGA AAG-3Ј; region 2, 5Ј-CTA AAT TTC CCC ATG TGT AAC GTG CAG GCA TTG-3Ј.
Western Blot Analysis-AR42J-B13 cells seeded in 10-cm dishes were starved in serum-free medium with or without SB203580 (Calbiochem, San Diego, CA) for 1 h, followed by incubation with or without activin A and HGF stimulation for different periods of time. Then 30 g of total cell extract prepared from these cells was separated on SDS-polyacrylamide gel, transferred to a PVDF membrane (Millipore, Tokyo, Japan), and incubated overnight at 4°C with the specific antibody (Erk1/2, phospho-Erk1/2, p38 MAPK, phospho-p38 MAPK, phospho-MKK3/6, JNK, and phospho-JNK (Cell Signaling, Tokyo, Japan)). Bound primary antibodies were detected with peroxidase-coupled anti-mouse or antirabbit secondary antibodies, and immunoreactivity was visualized by chemiluminescence (Western Lightning, PerkinElmer Life Sciences, Tokyo, Japan) according to the manufacturer's instructions.

RESULTS
Mapping the Activin A-and HGF-responsive DNA Sequence in the NEUROG3 Promoter-To analyze the activin A-responsive cis-element of the NEUROG3 promoter, we constructed a series of progressive 5Ј deletions of the promoter (each extending ϩ261 bp on the 3Ј-end) linked to the firefly luciferase gene and transfected these constructs into AR42J-B13 cells. It has been shown already that activin A plus HGF can induce the expression of ngn3 in this cell line (12). Although HGF alone, but not activin A, induced a modest increase of NEUROG3 promoter activity, the addition of activin A plus HGF induces a marked increase of its activity (Fig. 1A). Similar activation was observed with the reporter gene constructs containing Ϫ2.6 kbp, Ϫ461 bp, and Ϫ402 bp NEUROG3 promoter fragments (Fig. 1B). However, deletion of the fragment between Ϫ402 and Ϫ327 bp caused an increase in basal NEUROG3 promoter activity and a loss of responsiveness to activin A and HGF. These findings demonstrated that the region between Ϫ402 and Ϫ327 bp functions as a repressor and that activin A and HGF can negate its repressor effect.
To investigate whether the DNA sequence between Ϫ402 and Ϫ327 bp was sufficient to function as a repressor and an activin A-and HGF-responsive element, we investigated the promoter activity of reporter constructs containing the DNA fragment upstream of a heterologous promoter, the TK minimal promoter (Fig. 2). In every cell lines that we tested, this fragment acted as a repressor, but activin A-and HGF-responsive transactivation was not observed. These data suggest that this fragment alone was sufficient to function as a repressor, but the basal promoter sequence of NEUROG3 seems to be essential for activity as an activin A-and HGF-responsive cis-element.
Several Transcription Factors Are Possibly Involved in the Activin A-and HGF-induced Transactivation of ngn3-The activin A-and HGF-responsive region that we mapped had the following elements: one E-box, one repressor element 1/neuronrestrictive silencer element (RE1/NRSE)-like sequence (18), and one AT-rich sequence (Fig. 3A). To identify the cis-elements required for activin A-and HGF-mediated transactiva-tion of ngn3, we engineered mutations within the possible cis-elements located in the activin A-and HGF-responsive region (Ϫ402 to 327 bp), to investigate the function of each DNA sequence as a cis-element. When mutations were introduced into either the E-box or the RE1/NRSE-like sequence, basal NEUROG3 promoter activity was enhanced with reducing activin A and HGF responsiveness. On the other hand, mutations of the AT-rich sequence only had a modest effect on basal NEUROG3 promoter activity or on activin A-and HGF-responsive promoter activity (Fig. 3B).
Next, to try to identify the factors mediating activin A-and HGF-responsive transactivation of ngn3, we investigated the nuclear protein that binds to each mutation-sensitive cis-element before and various times after treatment with activin A and HGF using EMSA (Fig. 4). EMSA was able to identify factors that bound to each site and gradually decreased in binding affinity after treatment with activin A and HGF (C1 and C3 in Fig. 4) as well as factors (C2 and C4 in Fig. 4) that bound to each cis-element and gradually increased in affinity after treatment. Antiserum directed against RE1-silencing transcription factor/neuron-restrictive silencer factor antibody, FIG. 4. Activin A and HGF alter the DNA-binding activity of several factors in AR42J-B13 cells. The cells were incubated with activin A and HGF for 30 min, 1, 6, 12, or 24 h or without them. Then nuclear extracts were harvested, and 1 g of nuclear protein was used for EMSA. Upper and lower panels show the results obtained using 32 P-labeled oligonucleotide probes corresponding to the sequences for regions 1 and 2 (the sequences are shown below), respectively. basic helix-loop-helix protein E47, and c-Myc antibody did not recognize any of these protein-DNA complexes (data not shown). These results suggested that several factors with positive or negative effect on NEUROG3 promoter activity might be involved in the activin A-and HGF-induced transactivation of ngn3.

Smad-dependent Signaling Is Not Required for Activin Aand HGF-induced Transactivation of ngn3-Activin
A is a member of the TGF-␤ superfamily, and signaling for this family is mediated by a heteromeric complex of two types of transmembrane serine/threonine kinase receptors. Binding of a ligand to the receptor complex leads to the phosphorylation of type II receptor kinase and thereby activates type I receptor kinase (19). The activated type I receptor then phosphorylates receptor-activated Smads (R-Smads), such as Smad2 or Smad3, which bind to the related factor Smad4 and move into the nucleus. Once inside the nucleus, this Smad complex associates with other transcription factors to activate the transcription of target gene (19).
To investigate the involvement of a Smad-dependent pathway in activin A-and HGF-induced transactivation of ngn3, we transfected NEUROG3 promoter reporter plasmids along with a plasmid expressing the inhibitory-Smad (Smad7) (20) (Fig.  5). In AR42J-B13 cells, as in other TGF-␤-responsive cells, treatment with activin A and HGF activated a reporter plasmid containing the Smad-binding site of the PAI-1 promoter (20), and this activation was inhibited by Smad7. However, we could not find any influence of Smad7 on activin A-and HGF-induced stimulation of NEUROG3 promoter activity. These findings demonstrated that Smad signaling is not required for the induction of ngn3 by activin A and HGF.
P38 MAPK Is Required for Activin A-and HGF-induced Transactivation of ngn3-In addition to Smad activation, members of the TGF-␤ family activate other signaling cascades, including MAPK pathways (19). TGF-␤ is known to activate the extracellular signaling-regulated kinase (Erk), c-Jun N-terminal kinase (JNK), and p38 MAPK pathways (19,21). To investigate whether the MAPK pathway was involved in the signal transduction pathway for ngn3 induced by activin A and HGF, we tested for phosphorylation of Erk, JNK, and p38 MAPK after stimulation with activin A and HGF. As shown in Fig. 6, p38 MAPK and Erk were phosphorylated after treatment with FIG. 5. Smad-dependent signaling is not required for activin A-and HGF-induced transactivation of the NEUROG3 promoter. AR42J-B13 cells, with or without exposure to activin A and HGF, were transfected with 1.8 g of p3TP Luc plasmid containing the PAI-1 promoter (20) or 1.8 g of pFOX Luc1 ngn3 Ϫ461 bp plasmid, and were co-transfected with 1 g of Smad7 expression plasmid or 1 g of pcDNA3 plasmid. At 36 h after transfection, cells were harvested and luciferase activity standardized by the protein concentration was measured. Relative luciferase activity was calculated by taking as 1 the activity of cells transfected with the p3TP Luc plasmid or with pFOX Luc1 ngn3 Ϫ461 bp plasmid and co-transfected with the pcDNA3 plasmid, without exposure to activin A and HGF. All data are shown as the mean Ϯ S.E.
FIG. 6. Activin A and HGF activate p38 MAPK in AR42J-B13 cells. AR42J-B13 cells were starved in serum-free medium for 1 h and then stimulated with activin A and HGF for different periods, as indicated. To confirm the effect of SB203580, cells were serum-starved with SB203580 and stimulated with activin A and HGF for 60 min. Total extracts (30 g) prepared from these cells were separated on SDS-polyacrylamide gel, transferred to a PVDF membrane, and incubated with the specific antibodies for p38 MAPK, phospho-p38 MAPK, Erk1/2, phospho-Erk1/2, JNK, and phospho-JNK. Reaction products were visualized using chemiluminescence. activin A and HGF, although with different time courses. On the other hand, JNK was not phosphorylated.
Next, we focused on the involvement of p38 MAPK in this signal transduction pathway. SB203580 is a well known p38 mitogen-activated protein kinase (p38 MAPK)-specific inhibitor. As shown in Fig. 6, it inhibited the activation of p38 MAPK provoked by activin A and HGF, while showing no effect of the activity of Erk in this cell line. In this state, the typical morphological changes and promoter activity of NEUROG3 provoked by activin A and HGF were disturbed (Fig. 7, A and B). Furthermore, whereas C2 and C4 showed no definite changes of binding affinity after the addition of SB203580 (data not shown), SB203580 inhibited the decrease of C1 and C3 binding affinity provoked by activin A and HGF (Fig. 7C). These results suggested that signaling through p38 MAPK regulate the binding affinity of at least two transcription factors and thus may play an important role in activin A-and HGF-induced differentiation of AR42J-B13 cells.
The involvement of p38 MAPK in the signal transduction pathway suggested that TAK1 (22,23) might be involved in the TGF-␤ signaling pathway. TAK1 is a member of the MAPK kinase kinase family. TAK1 activates several members of the MAPK kinase family, including MKK6, MKK3, MKK7, and MKK4. Among them, MKK6 and MKK3 activate p38 MAPK. To investigate the possible involvement of the TAK1 pathway in the process of differentiation into pancreatic ␤cells, we examined the effect of overexpression of TAK1 and its dominant-negative mutants. Interestingly, co-transfection of TAK1 modestly augmented the activin A-and HGF-responsive transactivation of the NEUROG3 promoter and co-transfection with dominant-negative mutants of TAK1 caused a reduction in activin A-and HGF-responsiveness (Fig. 8A).
Next, we investigated the effect of dominant-negative mutants of MKK6, MKK3, and MKK4 on this pathway. In agreement with other findings, MKK3 and/or MKK6 were phosphorylated after addition of activin A and HGF to cultures of these cells. The time course of the phosphorylation of MKK3 or MKK6 resembled that of p38 MAPK, suggesting that MKK3 or MKK6 could directly activate p38 MAPK (Fig. 8B). Co-transfection with dominant-negative mutants of MKK3 reduced the activin A-and HGF-responsive transactivation of the NEU-ROG3 promoter (Fig. 8C), whereas co-transfection with dominant-negative mutants of MKK6 and MKK4 had no effect. These results suggested that induction of ngn3 expression by activin A and HGF involves the TAK1-MKK3-p38 MAPK signal transduction pathway. DISCUSSION Previous studies have revealed that signals from the notochord are essential for the differentiation of endocrine cells in the early stage of pancreatic development. In fact, removal of the notochord disrupts endocrine cell differentiation in the pancreas. Interestingly, treatment with activin A can induce endocrine cell differentiation even without the notochord (24) .   FIG. 7. A p38 MAPK inhibitor,  SB203580, inhibits the differentiation of AR42J cells treated with activin A and HGF. A, AR42J-B13 cells, with or without exposure to activin A and HGF, were transfected with 1.8 g of pFOX Luc 1 ngn3 Ϫ461 bp plasmid in the presence (solid bars) or the absence (open bars) of SB203580. Relative luciferase activity was calculated by setting as 1 the activity of cells transfected with pFOX Luc1 ngn3 Ϫ461 bp plasmid, without exposure to activin A and HGF. All data are shown as the mean Ϯ S.E. B, morphological effects of incubation with or without activin A and HGF for 24 h in the presence or absence of SB203580. C, AR42J-B13 cells were starved in serum-free medium for 1 h in the presence or absence of SB203580 and then treated with or without activin A and HGF for 12 h. Nuclear extracts were harvested from the cells for EMSA. 32 P-Labeled oligonucleotides corresponding to the sequences for regions 1 and 2 were used for the detection of C1 and C3, respectively.
Thus, the signals triggered by activin A play an important role in endocrine cell differentiation. In this study, we investigated the signal transduction pathway of activin A-and HGF-induced expression of ngn3 using AR42J-B13 cells as a model of endocrine cell differentiation.
Our 5Ј deletion analysis of the NEUROG3 promoter identified the region between Ϫ402 and Ϫ326 bp of the NEUROG3 as an activin A-and HGF-responsive DNA sequence. This region seems to function as a repressor in AR42J-B13 cells, whereas activin A and HGF relieve this repression. When this region was isolated from the NEUROG3 basal promoter sequence and ligated upstream of a heterologous promoter, it still repressed the transcription but was no longer responsive to activin A and HGF. These results suggested that some transcriptional repressors might bind directly to the DNA sequence within this region, but the response to activin A and HGF required the basal promoter of NEUROG3. The proximal NEUROG3 promoter may contain activin A-and HGF-responsive sequences that are required for relief of the distal repression activity. Alternatively, activin A and HGF signals may influence coactivators or co-repressors that interact with factors binding to both the proximal and distal sequences.
For transmembrane signaling, activin requires two types of receptors containing the serine/threonine kinase domain, activin receptor type I (ActRI) and type II (ActRII) (25). Activin directly binds to ActRII, and this complex associates with Ac-tRI, resulting in hyperphosphorylation of ActRI by the kinase activity of ActRII (19,26). The activated serine/threonine kinase of ActRI phosphorylates Smad proteins, especially Smad2 or Smad3, which activate transcription of the target genes of activin. In AR42J-B13 cells, Smad2 is far more abundant than Smad3; and the signals from activin A can activate Smad2 (27). In our study, treatment with activin A and HGF also activated a control DNA sequence containing a TGF-␤-responsive element, showing that the activin-ActR-Smad pathway is intact in AR42-JB13 cells. However, Smad7 did not affect the transactivation of ngn3 by activin A and HGF. This demonstrated that Smad-dependent signals, although commonly considered the main signaling pathway for activin A, are not required for the transactivation of ngn3 by activin A and HGF.
Instead of Smad-dependent signaling, the TAK1-MKK3-p38 MAPK pathway was involved in this transactivation process. Several studies have recently revealed a role for p38 MAPK in TGF-␤ signaling (21,28). Our findings show that the p38 MAPK pathway can play a role in activin A signaling as well. The existence of other, non-Smad pathways for TGF-␤/activin signaling is reminiscent of Wingless/Wnt molecules, which also can signal through more than one pathway (29). Recent studies indicate that known components of the canonical Wingless/Wnt signaling pathway can function independently of ␤-catenin to establish planer cell polarity through a JNK or RhoA signaling cascade (30 -32). Like Wnt signals, diverse sets of signals might be needed to mediate the various cellular responses to activin A.
Our data clearly show the importance of p38 MAPK in the signal transduction pathway for activin A-and HGF-induced expression of ngn3. However, because transfection of the dominant-negative forms of TAK1, MKK3, and p38 MAPK specific inhibitor did not completely inhibit activation of the NEU-ROG3 promoter, other pathways that do not involve TAK1- FIG. 8. Induction of ngn3 expression by activin A and HGF involves the signal transduction pathway of TAK1-MKK3-p38 MAPK. A, AR42J-B13 cells were transfected with 1.8 g of pFOX Luc1 ngn3 Ϫ461 bp plasmid and co-transfected with 0.5 g of the expression plasmid for pCMV-HA-TAK1 or 0.5 g of the expression plasmid for pCMV-HA-TAK1(K63W). Relative luciferase activity was calculated by taking as 1 the activity of cells transfected with the pFOX Luc1 ngn3 Ϫ461 bp plasmid, without exposure to activin A and HGF. All data are shown as the mean Ϯ S.E. B, AR42J-B13 cells were starved in serum-free medium for 1 h and then stimulated with activin A and HGF for different periods, as indicated. Then total cell extract (30 g) prepared from these cells was separated on SDS-polyacrylamide gel, transferred to a PVDF membrane, and incubated with the specific antibody for MKK3/6 or phospho-MKK3/6. Reaction products were visualized using chemiluminescence. C, AR42J-B13 cells were transfected with 1.8 g of the pFOX Luc1 ngn3 Ϫ461 bp plasmids and co-transfected with 1 g of the expression plasmid for pCS2ϩDN MKK3, pCS2ϩDN MKK4, or SRalpha-HA-MKK6AA. All data are shown as the mean Ϯ S.E. MKK3-p38 MAPK may also contribute to NEUROG3 promoter activation. Previous studies have shown that RAS GTPase, Rho-like GTPase (including Rho, Rac, and Cdc42), and protein phosphatase 2A are activated by TGF-␤ (33)(34)(35)(36). In this study, we detected the activation of Erk soon after the addition of activin A and HGF to AR42J-B13 cells, although its functional role was not investigated. Thus, assessment of other signal transduction pathways affected by activin A and HGF will be needed to complete our understanding of this process.
HGF is a multifunctional cytokine that has been shown to have a variety of effects on different cells. The receptor for HGF is a heterodimeric tyrosine kinase encoded by the c-met protooncogene (37,38). c-met signaling is mediated by autophosphorylation of its own receptor, which leads to marked up-regulation of its kinase activity. Previous studies have shown that MAPK, the signal transducer and activator of transcription 3 (STAT3), and phosphatidylinositol 3-kinase are involved in HGF signaling (39 -41), but no previous data have suggested that HGF activates the TAK1-MKK3 signaling pathway. Muller et al. (42) found that HGF alone could activate p38 MAPK in MLP29 cells. We found that HGF by itself activates p38 MAPK in AR42J-B13 cells. 2 These results suggest that activation of p38 MAPK by HGF might enhance the activation from the activin A-TAK1-MKK3 signaling pathway.
Recently, Mirk/Dyrk1B was identified as a transcriptional activator of HNF-1␣ (43). Mirk binds directly to DcoHm, a newly identified DcoH family member that stabilizes HNF-1␣ dimers. Thus, Mirk, DcoHm, and HNF-1␣ form a complex together. Mirk binds to a specific region within the CREBbinding protein-binding site of HNF-1␣ and phosphorylates HNF-1␣ at a site adjacent to this binding region. Interestingly, it has been shown that MKK3 enhances Mirk kinase activity and the transcriptional activation of HNF-1␣ by Mirk. Our findings demonstrated the involvement of TAK1-MKK3-p38 MAPK in activation of the NEUROG3 promoter by activin A, but the mechanism did not involve the simple activation of transcription factors by activin A and HGF. Accordingly, a complex mechanism like Mirk-mediated HNF-1␣ transcriptional activation could be involved in the regulation of ngn3 by activin A and HGF.
This study is the first step toward clarification of the mechanism by which activin A and HGF activate ngn3 gene expression. Although our findings were obtained using a cell line, these data suggest that signaling via the ActR-TAK1-MKK3-p38 MAPK pathway might play an important role in determining the fate of early pancreatic precursor cells. Considering the fact that ngn3 controls the initiation of islet cell formation, a detailed understanding of these signaling pathways may eventually contribute to a method for generating new islet cells from pancreatic precursor cells.