Regulation of Insulin Gene Transcription by ERK1 and ERK2 in Pancreatic β Cells*

We show that the mitogen-activated protein kinases ERK1/2 are components of the mechanism by which glucose stimulates insulin gene expression. ERK1/2 activity is required for glucose-dependent transcription from both the full-length rat insulin I promoter and the glucose-sensitive isolated E2A3/4 promoter element in intact islets and β cell lines. Dominant negative ERK2 and MEK inhibitors suppress glucose stimulation of the rat insulin I promoter and the E2A3/4 element. Overexpression of ERK2 is sufficient to stimulate transcription from the E2A3/4 element. The glucose-induced response is dependent upon ERK1/2 phosphorylation of a subset of transcription factors that include Beta2 (also known as NeuroD1) and PDX-1. Phosphorylation increases their functional activity and results in a cumulative transactivation of the promoter. Thus, ERK1/2 act at multiple points to transduce a glucose signal to insulin gene transcription.

Circulating insulin is produced by the ␤ cells in the pancreatic islets of Langerhans in adult mammals. Insulin regulates glucose metabolism and in turn glucose regulates synthesis and secretion of insulin by the ␤ cells. Insulin synthesis is stimulated by glucose at several steps, including transcription of the insulin gene (1)(2)(3)(4). Insulin mRNA is extremely stable with a half-life of more than 24 h (5). Thus, many studies have focused not on regulation of insulin gene transcription but on understanding the translation of its mRNA and its processing to insulin (6,7). Nevertheless, increased initiation of insulin gene transcription occurs within 10 min of an elevation in blood glucose concentration (8). Thus, insulin gene transcription, which is obviously essential for homeostasis of mRNA stores, is also sensitive to acute stimuli.
The organization of the insulin promoter is complex, allowing both tissue-restricted transcription of the insulin gene and levels of control for contextual regulation by nutrients, hor-mones, neurotransmitters, and other agents (9 -11). The proximal promoter, the sequences within a few hundred base pairs of the transcription start site, has been studied in detail for the rat and human insulin genes (for a review see Ref. 10). The proximal promoter contains multiple sequence elements capable of responding to glucose (12)(13)(14)(15)(16)(17), the best studied being the juxtaposed E and A elements, which together can function as a glucose-responsive minienhancer (13,14). Mutating or deleting these E and A elements from the promoter greatly reduces its glucose responsiveness (12), consistent with the idea that glucose regulates transcription of the insulin gene through its effects on E and A elements. How glucose signals to regulate insulin gene transcription through this glucose-sensitive element is unknown. E elements are binding sites for heterodimeric complexes formed by the neuroendocrine basic helix-loop-helix (bHLH) 1 protein Beta2 (18,19) and a ubiquitous bHLH protein such as E47 (20 -23). Binding of the bHLH heterodimer to the E element increases in response to prolonged glucose stimulation (12). The molecular signals underlying this increase in binding are unknown. Animals lacking the Beta2 gene develop diabetes and die within 3-5 days of birth (24). The development of the ␤ cells is impaired, and the insulin content of the residual cells is low.
A elements bind any of several homeodomain transcription factors found in ␤ cells, the most abundant being PDX-1 (25)(26)(27)(28). In contrast to the relatively slow glucose-induced increase in bHLH heterodimer binding to the E element, PDX-1 binding increases acutely in response to a rise in glucose concentration (26,29). The glucose-stimulated increase in DNA binding by PDX-1 is reportedly due to phosphorylation that is dependent on phosphatidylinositol 3-kinase and the p38 MAP kinase (26,30), although neither kinase was found to directly phosphorylate PDX-1. Glucose also causes PDX-1 to shift into the nucleus (31,32) and increases its transcriptional activation capacity (33). Maturity onset diabetes of the young type 4 (MODY4) has been linked to PDX-1 (34). One PDX-1 mutation identified in a MODY4 patient causes a frameshift in the activation domain. Heterozygotes for this mutation are predisposed to noninsulindependent diabetes mellitus.
Despite exhaustive study of the insulin promoter and control of insulin gene transcription by factors that bind to defined promoter elements, the link between glucose sensing and insulin gene transcription has remained enigmatic. We and others have shown that glucose activates the MAP kinases ERK1/2 in islet-derived cells (35)(36)(37)(38). MAP kinases, also known as ERKs, are components of highly conserved kinase cascades important for transmitting extracellular information to coordinate cellular responses. MAP kinases have been implicated in many physiological events ranging from cellular proliferation and differentiation to cell survival (reviewed in Refs. 39 and 40). ERK1/2 regulate functions throughout the cell, but among the most significant function is the control of gene transcription. ERK1/2 are stimulated by glucose in INS-1, MIN6, and ␤TC3 pancreatic ␤ cell lines (35)(36)(37)(38). ERK1/2 are activated over the same glucose concentrations, from 2 to 20 mM, as those that elicit insulin secretion. Potentiators of insulin secretion potentiate ERK1/2 activation. Although ERK1/2 are not required for glucose-stimulated insulin secretion, glucose increased the amount and activity of ERK1/2 in the nucleus of ␤ cells, suggesting that ERK1/2 may regulate gene transcription in these cells (36).
In this report, we show that dominant negative ERK2 and inhibitors of the ERK1/2 activators, the MAP/ERK kinases, MEK1/2, suppress glucose-stimulated transcription of the rat insulin I promoter in INS-1 cells, ␤TC3 cells, and primary islets. ERK2 stimulates transcription from reporter constructs in the absence of glucose via the glucose-responsive element. This process is activated at least in part because ERK2 phosphorylates bHLH transcription factors, E47/E12 and Beta2, and the homeodomain-containing transcription factor PDX-1. Phosphorylation of Beta2 and PDX-1 regulates their transactivation activities. In addition, phosphorylation of E47/E12 regulates its heterodimerization with Beta2 and subsequent DNA binding. These findings suggest that ERK1/2 are important components of the mechanism of glucose-responsive insulin gene transcription.

MATERIALS AND METHODS
Isolation of Islets-Islets from male adult Sprague-Dawley rats were isolated as described (41), incubated in glucose-free Krebs-Ringer bicarbonate HEPES buffer for 2 h, and treated as indicated for 30 min at 37°C. The islets were lysed in lysis buffer (36).
Northern Analysis-INS-1 cells were cultured as described in Ref. 36. They were infected with recombinant adenoviruses expressing either wild type or dominant negative (K52R) ERK2 at a multiplicity of infection of 20 for 48 h. Adenoviruses expressing wild type and K52R ERK2 were constructed as in Ref. 42. The cells were cultured without glucose for 24 h and then with 20 mM glucose for another 24 h. Total RNA was isolated using TRI reagent (Molecular Research Center, Inc.). 10 -20 g of total RNA was separated on a denaturing formaldehyde gel of 1.4% agarose. Prehybridization was at 45°C for 2 h, and hybridization was performed at 45°C overnight with [␣-32 P]dCTP-labeled cDNA probe at 10 6 cpm/ml. The membranes were washed twice in 2ϫ SSC, 0.1% SDS, once in 1ϫ SSC, 0.1% SDS, and twice in 0.2ϫ SSC, 0.1%SDS for 30 min each at 55°C. The bands were quantitated using a PhosphorImager.
Transfection Studies in INS-1 Cells-The rat insulin I promoter (Ϫ410/ϩ1 bp) was subcloned into pGL3-Basic (Promega) (pGL3-rINSp) containing a luciferase reporter gene. The minienhancer containing five copies of E2A3/A4 (Ϫ247/Ϫ198 bp) from the rat insulin I promoter was linked to a minimal rat insulin I promoter upstream of a chloramphenicol acetyltransferase (CAT) or luciferase reporter gene. INS-1 cells were cultured to a confluency of ϳ60 -70% and transfected using N-(1-(2,3-dioleoyloxy)propyl)-N,N,N-trimethyl ammonium methylsulfate (DOTAP) Lipofectin (Roche Applied Science). 24 h after transfection of 60-mm dishes with 10 g of pGL3-rINSp, pCMV5-␤-galactosidase, and either pCMV5 ERK2 wild type or K52R, the cells were placed in medium without glucose, or in some cases with 2 mM glucose, for 24 h before adding nothing or 15 mM glucose for 2 h. The lysates were assayed for luciferase (Promega) and ␤-galactosidase activities. 10 g each of either rINSp-CAT or E2A3/4-CAT were cotransfected with 10 g of vector alone, pCMV5 ERK2 K52R, pEF-Ras V12 or A15, or pCMV5 Raf BXB or C4B as described above. CAT assays were as described (43). PD98059 (New England Biolabs) and U0126 inhibitors were added to the cells after transfection for 2-24 h. U0126 was generously provided by J. Trzaskos (DuPont-Merck). All of the inhibitors were dissolved in Me 2 SO and used at a final concentration of 10 M for U0126 and 50 M for PD98059. 5 g of E2A3/4-luciferase was cotransfected with 5 g of pCMV5myc-ERK2-MEK1 or pCMV5myc-ERK2-MEK1LA and 2 g of CMV-PRL (Renilla luciferase; Promega) into INS-1 cells as above. In these cases, the dual luciferase system (Promega) was used to assay and normalize the luciferase activities.
Transfection Studies in Primary Islets-Adult mouse islets were picked by hand from collagenase-digested adult female CD-3 mice and cultured overnight in RPMI medium 1640 with 10% fetal bovine serum. The islets were transfected using a modification of the adenovirusassisted transfection technique previously described (16). Aliquots of 100 islets were placed in 12 ϫ 75-mm polystyrene culture tubes and washed three times with 1 ml of OPTI-MEM 1 medium (serum-free; Invitrogen). Plasmid DNA (4 g) in 0.1 ml of medium was mixed with 4 g of 25-kDa polyethylenimine (Aldrich) in 50 l of medium at room temperature for 15 min. Replication-deficient adenovirus 5 dI-342, as previously described (44), was adjusted to yield a final concentration of 10 11 virus particles/ml. 0.1 ml of the adenovirus was added to the plasmid DNA/polyethylenimine mixture and incubated at room temperature for an additional 5 min. The mixture was added to the islets at 37°C for 30 min, after which the islets were washed three times with 1 ml of RPMI medium 1640 containing 10% fetal bovine serum and incubated at 37°C for 36 h. Protein concentration in islet extracts was determined with the NanoOrange Protein Quantification Assay (Molecular Probes). Approximately 2 g of the protein extract were assayed for CAT activity (45).
Kinase Assays and Phosphoamino Acid Analysis-In vitro kinase assays were performed with purified active ERK2, the stress-activated protein kinases (SAPK) ␣ (also called c-Jun N-terminal kinase), or p38␣ MAP kinases (49) in 30 l of 20 mM HEPES, pH 8.0, 10 mM MgCl 2, 100 M ATP ([␥-32 P]ATP; 5-15 cpm/fmol) with the active kinase and substrates at 30°C for 30 min. The samples were analyzed on polyacrylamide gels in SDS that were either stained in Coomassie Blue before autoradiography or transferred onto nitrocellulose or polyvinylidene difluoride membranes. Phosphoamino acid analysis was carried out as described (50).
Transactivation Assays-Wild type and the triple phosphorylation mutant of human E12 AD2 were subcloned into pM (SV40 promoter) to make GAL4-DNA-binding fusion proteins. Wild type and mutated forms of Beta2 (156 -355) and PDX-1 (1-149) were subcloned into pGAL4 (CMV promoter) and cotransfected with GAL4-binding domains linked to a luciferase reporter gene (G5E1bLuc) with the CMV promoter-driven Renilla luciferase gene (CMV-PRL; Promega) into ␤TC3 cells using the SuperFect transfection system (Qiagen). The cells were harvested 48 h later. The cells were exposed to PD98059 and SB203580 for 24 h as indicated.
Recombinant Retroviruses-Recombinant retroviruses expressing Myc-tagged wild type and S274A Beta2 (91-355) were made as described in Ref. 51. The cDNAs were subcloned into pBabe-Puromycin and then transfected into the amphotropic packaging cell line using the CellPhect transfection kit in 100-mm dishes (Amersham Biosciences). After 48 h, media containing recombinant retroviruses were harvested and filtered through 0.45-m filters. An infection mixture of 3 ml of recombinant retrovirus supernatant and 2 ml of INS-1 growth medium plus polybrene (Sigma) (final concentration, 4 g/ml) was used to infect INS-1 cells. After 6 h, 5 ml of INS-1 medium was added for 72 h. 0.5 g/ml puromycin (Sigma) was added to select cells expressing recombinant retroviruses for 3-5 days.
Miscellaneous Materials and Methods-Site-directed mutagenesis was carried out with the Stratagene QuikChange kit. Unless otherwise indicated, the protein concentrations were measured using BCA reagent (Pierce) with albumin as standard. Bacterial expression vectors encoding GST-c-Jun (1-221) and GST-ATF (1-254) were kindly provided by Michael Karin (University of California, San Diego). A bacterial expression vector encoding GST-Mnk was generously provided by Tony Hunter (Salk Institute). The glucokinase promoter (Ϫ280/ϩ4 bp) was generously provided by Mark Magnuson (Vanderbilt University).

RESULTS
Glucose Activates ERK1/2 in Primary Islets of Langerhans-In previous studies, we showed that glucose activated ERK1/2 in INS-1 insulinoma cells (36). Stimulation was observed in cells preincubated without glucose and in cells preincubated in 3 mM glucose. Before examining effects of ERK1/2 on insulin gene transcription, we first examined the effect of glucose on the activity of ERK1/2 in primary islets and in another glucose-sensitive cell line ␤TC3. Intact islets were preincubated in the absence of glucose for 2 h. Under these conditions, small amounts of activated ERK1/2 were detected using antibodies that selectively recognize the phosphorylated forms of the kinases. This is consistent with studies in INS-1 cells; in these cells complete inactivation of ERK1/2 by removing glucose required from 1 to 4 h. In isolated islets, increasing the glucose concentration to 15 mM caused a marked activation of ERK1/2, which is potentiated by forskolin (Fig. 1A). Similar findings were observed in ␤TC3 cells, except that ERK1/2 were activated at the lowest glucose concentration tested, consistent with the left shift in glucose-stimulated insulin secretion in these cells (Fig. 1B).
ERK1/2 Regulate Proinsulin mRNA Levels in Pancreatic ␤ Cells-To determine whether ERK1/2 regulate insulin gene transcription, we first measured the effect of blocking ERK1/2 on insulin mRNA content in INS-1 cells. Wild type or dominant negative (kinase inactive) K52R ERK2 were expressed from adenoviruses in INS-1 cells, and the insulin mRNA content was measured by Northern analysis. K52R ERK2 caused a small but significant reduction in the amount of insulin mRNA after 24 h (Fig. 1C). However, no glucose-stimulated increase in insulin mRNA was observed in these cells, which is most likely due to the high stability and low turnover rate of insulin mRNA in INS-1 cells.
ERK1/2 Regulate Transcription from the Rat Insulin I Promoter in ␤ Cells-Depression of insulin mRNA levels by K52R ERK2 could be through an effect on the stability of insulin mRNA or on the initiation of insulin gene transcription. To determine whether there were effects on transcription, we measured glucose-dependent changes in the expression of a reporter gene coupled to the rat insulin I promoter (Ϫ410/ϩ1 bp) with and without blockade of ERK1/2 activity. K52R ERK2 blocked glucose-stimulated transcription from the promoter ( Fig. 2A). Neither the CMV nor the glucokinase promoter (52) was substantially affected by K52R ERK2 (data not shown). PD98059 and U0126, two chemically distinct inhibitors of MEK1/2, blocked glucose activation of ERK1/2 in INS-1 cells and suppressed glucose-stimulated rat insulin I promoter activity (Ref. 36 and Fig. 2, B and C). These observations are consistent with the results obtained with K52R ERK2, suggesting that ERK1/2 regulate rat insulin I transcriptional activity in INS-1 cells.
To demonstrate a regulatory role of ERK1/2 in insulin gene transcription in islets, the ability of K52R ERK2 to affect tran- scription from a rat insulin I promoter construct was tested in freshly isolated adult rat islets. In these experiments the unstimulated condition was 2 mM glucose. The stimulation of transcription from the rat insulin I promoter induced by 16 mM glucose was strongly inhibited by K52R ERK2 (Fig. 2, B and C). Glucose-stimulated transcription from the promoter in rat islets was also suppressed by PD98059 (Fig. 2, B and C). These results support findings in rat insulinoma INS-1 cells suggesting that ERK1/2 are important regulators of insulin gene transcription in pancreatic ␤ cells.
ERK1/2 Regulate Insulin Gene Transcription via a Glucoseresponsive Element-To determine the portion of the rat insulin promoter that is regulated by ERK1/2, we first examined the effects of blocking ERK1/2 on a well described glucoseresponsive element E2A3/4 (Ϫ247/Ϫ198 bp). K52R ERK2 and the MEK inhibitor PD98059 each suppressed the glucose-dependent transcriptional activity of E2A3/4 in INS-1 cells (Fig.  3A), suggesting that regulation of insulin promoter activity by ERK1/2 occurs at least in part through this glucose-responsive portion of the promoter. Similar results were also found in ␤TC3 cells (not shown). We also determined whether ERK1/2 regulate insulin gene transcription via the glucose-sensitive element in primary mouse islets. K52R ERK2 had negligible effect on the glucose-sensitive element at basal glucose, whereas it blocked the ability of glucose to stimulate transcription promoted by E2A3/4 at 16 mM glucose (Fig. 3A). These results are consistent with our observations in ␤ cell lines suggesting that ERK1/2 regulate insulin gene transcription at least in part through E2A3/4.
To determine whether ERK1/2 are sufficient for initiating transcription from the insulin gene promoter independently of the glucose stimulus, constitutively active mutants of Raf and MEK1(DE), upstream activators of ERK1/2 were introduced; they stimulated expression of the reporter gene driven by E2A3/4 (not shown). To examine sufficiency of ERK2 itself, independent of its upstream activators, we used ERK2-MEK1 fusion proteins. Fusion of wild type ERK2 to the wild type allele of its upstream activator MEK1 or to MEK1 with a mutated nuclear export signal, ERK2-MEK1 and ERK2-MEK1LA, respectively, yields constitutive ERK2 activity without activating endogenous ERK proteins (53). In the absence of activated ERK2, glucose doubled transcription from this reporter. Both forms of active ERK2 stimulated transcription from this E2A3/4-driven reporter by ϳ3.5-4-fold above the basal glucose control (Fig. 3B). Thus, ERK2 is sufficient to promote gene transcription via a glucose-responsive element as well as or better than glucose alone. These results support the conclusion that ERK-sensitive motifs are contained within the glucose-responsive element.
ERK2 Phosphorylates Beta2, E47/E12, and PDX-1 in Vitro-To begin to elucidate the mechanism of induction of the glucose-responsive element, we tested the ability of ERK2 and other MAP kinase family members to phosphorylate several factors known to bind to the A and E boxes. E box elements bind to factors belonging to the bHLH family. bHLH transcription factors bind to DNA as homodimers or as heterodimers formed between the ubiquitously expressed class A members of the bHLH family, such as E47/E12 (22), and the tissue-restricted class B bHLH members. The currently known class B bHLH protein in ␤ cells is Beta2 (18,19). Five copies of the A motif in the rat insulin promoter are recognition sites for homeodomain proteins. Homeodomain-containing transcription factors that bind these A motifs from ␤ cells include PDX-1 and Lmx1.1 (28,54,55). These proteins, E47, Beta2, PDX-1, and Lmx1.1, were used as in vitro substrates for activated MAP kinases (49). ERK2 phosphorylates E47, Beta2, and PDX-1 but not Lmx1.1 in vitro (Fig. 4, B-D, and data not shown). E47 and PDX-1 were also phosphorylated by activated SAPK␣ and p38␣ (Fig. 4, B-D). On the other hand, Beta2 and Lmx1.1 were not phosphorylated by either of these kinases.
Phosphoamino acid analysis of Beta2 showed that ERK2 phosphorylated only serine residues (Fig. 5B). The amino acid sequence of Beta2 revealed four potential serine residues, 162, 259, 266, and 274, that fit a consensus ERK2 phosphorylation motif and that were mutated to alanine (Fig. 5A). Mutation of Ser 274 , which lies in the most typical ERK2 phosphorylation motif, reduced the phosphorylation of Beta2 to less than onethird of the wild type (Fig. 5C), suggesting that Ser 274 is the predominant site for ERK2. Mutation of the other three sites individually reduced Beta2 phosphorylation by one-third or less, suggesting that each is a minor ERK2 phosphorylation site (Fig. 5C). Mutation of all four residues reduced phosphorylation to less than 20% of the wild type protein, indicating that there are one or more additional minor sites.
E47 was phosphorylated on serine and threonine residues by ERK2 (Fig. 6C). Three contiguous potential phosphorylation sites, serines 352, 355, and 359 in human E12, are present in its second activation domain (Fig. 6A). When these residues were mutated to alanine in E47, phosphorylation by ERK2 was partially reduced compared with wild type E47 (Fig. 6B). Phosphorylation of threonine was completely abolished in the triple  6. ERK2 phosphorylation sites on E47. A, a diagrammatic representation of E47/E12 indicating the mutated serine and threonine residues. B, in vitro kinase assays of His 6 -tagged recombinant wild type or the E47 triple phosphorylation site mutant with active ERK2. C, phosphoamino acid analysis of wild type and the mutant E47. mutant, whereas serine phosphorylation was only slightly reduced. Thus, E47 is phosphorylated by ERK2 on these and additional serine residues that were not identified.
Phosphoamino acid analysis showed that PDX-1 was phosphorylated on serine residues by ERK2 (Fig. 7B). The two potential serine phosphorylation sites, residues 61 and 66, were mutated to alanine. Because PDX-1 was also phosphorylated by SAPK␣ and p38␣, in vitro kinase assays were performed with active forms of all three kinases. All three kinases showed some preference for Ser 66 ; p38␣ did not phosphorylate PDX-1 with this site mutated (Fig. 7C). Mutation of both residues eliminated phosphorylation by SAPK␣ and p38␣.
Evidence for in Vivo Phosphorylation of ␤2 in INS-1 Cells-Phosphorylation of wild type Beta2 by ERK2 resulted in a decrease in its electrophoretic mobility (Fig. 8A) observed as two bands of reduced mobility on the Beta2 immunoblot. This phosphorylation-dependent mobility change was abolished in Beta2 S274A (Fig. 8A), consistent with the interpretation that phosphorylation of Ser 274 induces the shift. To determine whether Beta2 is phosphorylated in INS-1 cells, we took advantage of the changes in electrophoretic mobility of Beta2 as a measure of its phosphorylation state. Immunoblotting of lysate proteins from INS-1 cells expressing wild type Beta2 revealed five species of the protein, whereas in INS-1 cells expressing Beta2 S274A, only the three lower bands were observed (Fig.  8B). The various species of Beta2 are most likely differently phosphorylated forms of the protein, consistent with the idea that Beta2 is phosphorylated on multiple sites including Ser 274 in cells.
Phosphorylation of ␤2 Regulates Its Transactivation-ERK2 phosphorylation sites in both E47 and Beta2 are located in their respective activation domains. Thus, we examined whether phosphorylation of the ERK2 sites is required for the transactivating activity of the transcription factors using GAL4-DNA-binding domain chimeras of AD2 of E12 and the Beta2 activation domain (155-355). AD2 of E47/E12 has been previously shown to be functional in ␤ cells (20). However, the transactivating activity of the E12 triple phosphorylation mutant was unchanged compared with wild type E12 in ␤TC3 cells or INS-1 cells (data not shown).
In contrast, mutating Ser 274 significantly diminished transactivation by Beta2 in the presence of glucose (Fig. 9, A and B). Although S274A Beta2 exhibited a transactivating activity only about 30% of wild type, mutation of the other ERK2 sites individually only reduced transactivation to about 60 -70% of wild type. Mutation of all four residues depressed the transactivation activity of Beta2 to 20% of wild type (Fig. 9A), indicating that the other three sites contribute less than Ser 274 . Similar results were obtained in INS-1 cells (Fig. 9A). To determine whether Beta2 transactivation is dependent on activation of ERK1/2 in ␤ cells, we tested whether blocking ERK1/2 activation will suppress the transactivating activity of Beta2 observed in the presence of glucose. In the presence of the MEK inhibitor PD98059, the activity of wild type Beta2 was reduced to less than 20% of that in its absence. The addition of PD98059 further reduced the transactivation activities of the S274A/ S266A/S259A and S274A/S266A/S59A/S162A mutants to less than 10% of the wild type control (Fig. 9B). This suggests that Ser 274 has the largest effect on transactivating potential but that the additional sites also contribute to the glucose-enhanced transactivating activity of Beta2.
Phosphorylation of PDX-1 Regulates Its Transactivating Ability-We tested whether mutation of the serine residues in the activation domain of PDX-1 affected its transactivating activity in ␤ cells using GAL4-DNA-binding chimeras. Mutation of either Ser 61 or Ser 66 to alanine decreased the transactivation activity of the mutant to 30 and 50% of the wild type observed in the presence of glucose, respectively (Fig. 9C). However, mutation of both residues had no greater effect than mutation of Ser 61 alone, suggesting that phosphorylation of Ser 61 is more significant and required for the full glucoseinduced transactivation potential of PDX-1. In the presence of PD98059, transactivation is reduced by 50%, consistent with the conclusion that ERK2 does phosphorylate PDX-1 in response to glucose to increase its transactivating activity (Fig.  9C). Because PDX-1 is also phosphorylated by p38␣, we compared effects of the p38 inhibitor SB203580 on its transactivation. The p38 inhibitor had a negligible effect on PDX-1, sug- gesting that glucose does not regulate the transactivating activity of PDX-1 through p38 in ␤TC3 cells.
Heterodimerization and DNA Binding of E47 and ␤2 Are Regulated by ERK2 Phosphorylation-We also examined the effects on E47 dimerization and DNA binding by electrophoretic mobility shift assay using recombinant His 6 -tagged proteins. E47 alone bound poorly to the oligonucleotide derived from the E box. Binding was enhanced by phosphorylation of E47 in vitro by ERK2 (Fig. 10). Myc-tagged Beta2 by itself, either unphosphorylated or phosphorylated by ERK2, did not bind DNA (data not shown). When unphosphorylated forms of E47 and Myc-Beta2 were mixed together, two bands corresponding to an E47 homodimer-DNA complex and an E47-Beta2 heterodimer-DNA complex were observed (Fig. 10). Under these conditions, the heterodimer-DNA complex was weak. The compositions of the complexes were confirmed by supershifting with anti-E47 and anti-Myc antibodies. An oligonucleotide with mutations in the E box consensus sequence was not shifted by the proteins, indicating that the shifts observed with the wild type oligonucleotide are specific (data not shown). The formation of the heterodimer-DNA complex was increased when either E47 or Beta2 was phosphorylated by ERK2. Only when both E47 and Beta2 were phosphorylated by ERK2 in vitro was maximal heterodimer-DNA complex formation observed. Phosphorylation also decreased the formation of the E47 homodimer complex. DISCUSSION ERK1/2 mediate an essential step in the signaling cascade through which glucose regulates insulin gene transcription. The first suggestion that ERK1/2 lie in this pathway was pro-vided by evidence that both kinases are activated by glucose in insulinoma cell lines (35), evidence that we have now extended to islets. The demonstration that preventing ERK1/2 activation with inhibitors or blocking ERK1/2 activity with a dominant negative mutant prevents glucose stimulation of the insulin promoter establishes a necessary role for ERK1/2 in the path- way, as has been previously suggested (56). Furthermore, active ERK2 alone is sufficient to induce transcription driven by E2A3/4 in the absence of glucose.
The involvement of the glucose-responsive E2A3/4 minienhancer implicates transcription factors that act through E and A elements. Three of these factors, Beta2, PDX-1, and E47/12, are substrates for ERK1/2 and are activated upon phosphorylation by ERK1/2. ERK2 phosphorylates Beta2 at multiple sites within its activation domain that enhance its glucose-sensitive transactivating capability. Phosphorylation of Beta2 by ERK2 also enhances its capacity to heterodimerize with E47/12 and bind to DNA. The fact that blocking ERK1/2 activity using inhibitors or mutants also blocks glucose-stimulated transactivation by Beta2 strongly supports the conclusion that ERK1/2 regulate insulin transcription through effects on Beta2. Petersen et al. (57) have shown that Ser 274 influences the nuclear translocation of Beta2; use of PD98059 also implicated ERK1/2 in this event.
Although phosphorylation of the sites identified on E47/12 has no apparent effect on its transactivating capacity, ERK2 phosphorylation of E47/12 increases its propensity to form heterodimers and to bind to DNA. Phosphorylation of E47 also appears to reduce the formation of E47 homodimer-DNA complexes; such complexes may be nonfunctional in promoting insulin transcription in ␤ cells. ERK1/2 phosphorylation also likely regulates dimerization of E47 in other cell types in which different heterodimers arise.
ERK2 also phosphorylates PDX in its activation domain. Phosphorylation of PDX-1 increases its DNA binding and transactivation capacity (33). Mutation of ERK1/2 phosphorylation sites on PDX-1 halves its transactivation capacity and blockade of ERK1/2 activity reduces transactivation to near that in the absence of glucose, indicating that phosphorylation by ERK1/2 enhances its glucose-dependent transactivation. Thus, ERK1/2 act at multiple loci within the machinery that controls insulin gene transcription to transduce changes elicited by glucose. Phosphorylation could stimulate the Beta2 and PDX-1 activation domains in any of several ways. The affinity for coactivators such as p300/CBP (58, 59) could be increased, and interactions with repressors could be reduced. Phosphorylation could also enhance cooperative interactions, such as the interaction between PDX-1 and the bHLH heterodimer (55).
Phosphorylation of PDX-1 is reportedly regulated by a kinase downstream in a p38-dependent pathway (30). The p38 pathway may regulate the localization or DNA binding of PDX-1 but has no apparent effect on glucose-dependent transactivation based on inhibitor studies. In vitro, PDX-1 is phosphorylated by ERK2, c-Jun N-terminal kinase/SAPK, and p38 itself, suggesting that PDX-1 may integrate signals from multiple MAP kinase pathways. p38 phosphorylates the same two sites as ERK2, indicating that it could increase PDX-1 transactivation. However, glucose-stimulated PDX-1 transactivating activity is not blocked by the p38 inhibitor, although it is reduced by preventing ERK1/2 activation. Furthermore, p38 is activated very poorly by glucose (36), consistent with the failure of the p38 inhibitor to interfere with glucose-induced transactivation. Perhaps the p38 cascade impacts PDX-1 in response to agents other than glucose.
Although we have identified functional changes in three factors that will lead to increased insulin gene transcription, other functional changes may also be caused by ERK2 phosphorylation. Regulation of bHLH transcription factors by ERK1/2 occurs at multiple steps. It has been hypothesized that heterodimer complexes that bind to E box elements may synergize with complexes bound to A elements to cause transactivation of the insulin gene. In particular, E47 has been shown to synergistically interact with PDX-1 (47). In addition, ERK1/2 may have other substrates in ␤ cells, either direct or through protein kinase targets such as Rsk.
In conclusion, ERK1/2 are viewed almost monolithically as enzymes activated during and as a necessary part of cell proliferation. In fibroblasts, activation of ERK1/2 has been associated with exit from G 0 into G 1 of the cell cycle and cell proliferation. Their roles in differentiated cells have often been overlooked. Although most of our understanding of the regulation and functions of these protein kinases comes from fibroblasts, ERK1/2 are highly expressed in most cell types including post-mitotic neurons and neuroendocrine ␤ cells. In cortical neurons, glutamate-induced changes in transcription from the serum response element, thought to be important for long term adaptive changes, are mediated in part by ERK1/2 (60 -63). ERK1/2 have been linked to long term potentiation, both directly through the induction of ERK nuclear translocation by glutamate and by inference from the deficiency of an animal lacking the calcium-sensitive Ras exchange factor in acquisition of long term memory (61,62). The tight relationship between ERK1/2 activation and glucose concentration suggested that these kinases may link glucose sensing to mechanisms that maintain insulin production both short and long term. Transcriptional control appears to be a major target for ERK signaling. We show here that ERK1/2 appear to serve a function in ␤ cells similar to that in neuronal cells by helping to integrate long and short term nutrient sensing information in the nucleus to maintain insulin homeostasis.