Pancreatic β-Cell-specific Repression of Insulin Gene Transcription by CCAAT/Enhancer-binding Protein β

Chronic exposure of β-cells to supraphysiologic glucose concentrations results in decreased insulin gene transcription. Here we identify the basic leucine zipper transcription factor, CCAAT/enhancer-binding protein β (C/EBPβ), as a repressor of insulin gene transcription in conditions of supraphysiological glucose levels. C/EBPβ is expressed in primary rat islets. Moreover, after exposure to high glucose concentrations the β-cell lines HIT-T15 and INS-1 express increased levels of C/EBPβ. The rat insulin I gene promoter contains a consensus binding motif for C/EBPβ (CEB box) that binds C/EBPβ. In non-β-cells C/EBPβ stimulates the activity of the rat insulin I gene promoter through the CEB box. Paradoxically, in β-cells C/EBPβ inhibits transcription, directed by the promoter of the rat insulin I gene by direct protein-protein interaction with a heptad leucine repeat sequence within activation domain 2 of the basic helix-loop-helix transcription factor E47. This interaction leads to the inhibition of both dimerization and DNA binding of E47 to the E-elements of the insulin promoter, thereby reducing functionally the transactivation potential of E47 on insulin gene transcription. We suggest that the induction of C/EBPβ in pancreatic β-cells by chronically elevated glucose levels may contribute to the impaired insulin secretion in severe type II diabetes mellitus.

Insulin is a hormone essential for the control of mammalian glucose homeostasis and is produced predominantly in pancreatic ␤-cells of adult animals (1). The expression of the insulin gene occurs to a large extent at the level of transcription. Control elements residing in the 5Ј 350-base pair sequence flanking exon 1 of the rat insulin I gene are sufficient to direct ␤-cell-specific expression (2) (Fig. 1A). Arrays of A and E elements 1 (Far-FLAT, Nir-P1) constitute symmetrical enhansons that cooperatively account for Ͼ90% of the transcriptional activity of the insulin gene promoter (3). The E elements are recognition motifs for transcription factors in the basic helixloop-helix (bHLH) 2 family, such as E12 and E47, which activate the insulin promoter in close synergism with A element binding homeobox transcription factors, such as IDX-1.
Chronic hyperglycemia may contribute to the pancreatic ␤cell dysfunction observed in patients with type II diabetes, a phenomenon attributed to the concept of glucose toxicity (4). Studies using in vivo animal models and in vitro ␤-cell lines have demonstrated that a reduction of insulin gene transcription by glucose toxicity is associated with the loss of transactivator proteins such as IDX-1/IPF-1/STF-1 and RIPE3b1-binding protein (5)(6)(7)(8)(9)(10). Because insulin gene transcription is both positively and negatively regulated, we sought to identify repressors that might also mediate the effects of glucose toxicity on insulin gene transcription. In this report we describe CCAAT/enhancer binding protein-␤ (C/EBP␤) as a glucoseinduced repressor of insulin gene transcription.
C/EBPs are a family of transcription factors that regulate genes of the acute phase response, cell growth, differentiation, and the expression of cell type-specific genes (11)(12)(13)(14)(15)(16). The C/EBPs consist of the activators C/EBP ␣, ␤, ␥, ␦, and ⑀ and the repressors CHOP, LIP, and C/EBP-30; the latter two repressors arise by alternative downstream translation of the mRNAs (17). The C/EBPs bind to DNA exclusively as dimers and contain a conserved C-terminal basic region-leucine zipper domain that is characterized by a DNA-contacting basic region linked to a leucine zipper dimerization motif (18). They bind preferentially to a consensus DNA sequence T(T/G)NNGNAA(T/G) (19,20). The founding member of the family of C/EBP transcription factors, C/EBP␣, is expressed during terminal differentiation of cells such as adipocytes (13) and keratinocytes. C/EBP␤ is abundant in liver, is expressed in response to stressactivated signaling pathways, and activates the expression of genes involved in the acute phase response such as cytokine genes. It has been shown that the expression of C/EBP␤ transactivates the transcription of genes encoding the insulin receptor and glucose transporter-2 (21,22), suggesting that C/EBP␤ may play an important role in glucose homeostasis and the metabolic stress associated with diabetes mellitus. The promoters of both of the rat insulin I and II gene, as well as the human insulin gene, contain sequence elements that closely resemble the consensus C/EBP-binding site (see Fig. 1). The sequence similarities among the elements imply that the C/EBP family of DNA-binding proteins may regulate the expression of the insulin gene. In addition, the activity of the insulin gene promoter is regulated by glucose and hormones, which elevate ␤-cell [Ca 2ϩ ] and cAMP levels and possibly protein kinase C activity (2,23). C/EBP␤ may mediate the effects of multiple second messengers on insulin gene expression, since its activity can be influenced by Ca 2ϩ , cAMP, and protein kinase C signaling pathways (24 -26). In the present study, we find that C/EBP␤ is expressed in pancreatic ␤-cells and is up-regulated by supraphysiologic glucose concentrations in the culture media of pancreatic ␤-cell lines. C/EBP␤ inhibits insulin promoter activity in ␤-cell lines, but not in the non-␤-cell HeLa and BHK cell lines. In pancreatic ␤-cells C/EBP␤ specifically interacts with a heptad leucine repeat sequence within activation domain 2 (AD2) of the basic helix-loop-helix transcription factor E47, thereby inhibiting the DNA binding activity and the transactivation potential of E47.

EXPERIMENTAL PROCEDURES
Reagents-DNA-modifying enzymes were purchased from New England Biolabs (Beverly, MA) or Boehringer Mannheim; radioactive compounds were from NEN Life Science Products; D-luciferin-potassium was from the Analytical Luminescence Laboratory (San Diego, CA); RPMI 1640 and DMEM medium and fetal bovine serum (FBS) were purchased from Life Technologies, Inc. Nucleotides were obtained from Pharmacia Biotech Inc. All other reagents were purchased from Sigma.
Cell Culture-The pancreatic ␤-cell line HIT-T15 (27) at passage 64 and COS-7 cells were purchased from the ATCC. Ins-1 (28) cells at passage 99 were a gift from Dr. Claes B. Wollheim (University of Geneva, Switzerland). HIT-T15 cells were maintained in RPMI 1640 medium (Life Technologies, Inc.) with 10% FBS at 37°C in a 5% CO 2 , 95% air atmosphere as described (7). Ins-1 cells were grown in RPMI 1640 medium with 10% FBS, 50 M ␤-mercaptoethanol, 1 mM sodium pyruvate, and 10 mM HEPES as reported (28). Both cell lines were passaged weekly. For the model of long term exposure, HIT cells were cultured from passage 64 to passage 82 in 11.1 mM D-glucose or 0.8 mM D-glucose with adjustment of osmolality by the addition of mannitol to the low glucose medium. Glucose concentrations for HIT-T15 cells were chosen according to the left-shifted insulin response curve as previously reported (29). Ins-1 cells were grown in 25 mM or 5.6 mM D-glucose, respectively, with mannitol adjustment as reported previously (30). The ␤TC-6 cell line (31) was a gift from Dr. Shimon Efrat (Albert Einstein University College of Medicine, New York). The ␤TC-6 cells were cultured in Dulbecco's modified Eagle's medium (25 mM glucose) supplemented with 10% FBS. Passages from 23 to 33 were used for transfection experiments.
Islet Isolation-Male Sprague-Dawley rats (150 -200 g) were anesthetized with 100 mg/kg intraperitoneal pentobarbital sodium. Islets were isolated from the pancreata using an adaptation for rat islets of the method of Gotoh et al. (32). Briefly, after cannulation of the common bile duct and instillation of 10 ml of a prewarmed (37°C) solution containing 1 mg/ml Collagenase P and 0.5 mg/ml DNase I, the pancreas was removed and digested for 30 min at 37°C in a shaking water bath followed by dilution and washing of the digest and hand picking of the released islets under a dissecting microscope. Liver nuclei were prepared by the method of Gorski et al. (33).
Antisera and Western Immunoblot-Polyclonal rabbit antisera for C/EBP␤ and E47 were obtained from Santa Cruz Biotechnology (Santa Cruz, CA). The IDX-1 antiserum was described previously (34). Western immunoblot analysis was performed on nuclear extracts prepared from the cell lines according to standard techniques (35). Extracts of pancreatic islet whole cells and liver nuclei were prepared by lysing isolated rat islets and liver nuclei in SDS-PAGE sample buffer (36). In each lane, a sample containing 100 g of protein was loaded.
Purification of Proteins Expressed in Bacteria-A C/EBP␤ protein fragment containing the basic region-leucine zipper binding domain that starts at an internal methionine site (12) was made by insertion into the BamHI-XhoI site of the pRSET-A vector (Invitrogen Inc., Carlsbad, CA). The protein was transcribed and translated and was purified with a nickel-chelating resin column and eluted by a pH gradient.
DNase I Footprinting-C/EBP␤ isolated from bacteria was incubated with a fragment of the 5Ј-flanking region of the rat insulin I gene (nucleotides Ϫ280 to ϩ1). The fragment was end-labeled on the coding strand by the polymerase chain reaction. The DNase I footprinting analysis was carried out as described previously (38).
Plasmids, DNA Transfection, and Luciferase Assays-The plasmid Ϫ410INS-LUC contains a fragment of the rat insulin I gene promoter from base pairs Ϫ410 to ϩ49 cloned into the pXP2 vector containing the coding sequence of firefly luciferase cDNA. The luciferase-reporter plasmid DNA containing the CEB mutation within the rat insulin I gene promoter was created by mutation of the CEB box from Ϫ128 TG-TAAT Ϫ133 to Ϫ128 CTCGGC Ϫ133 using oligonucleotide-directed mutagenesis (39). The luciferase plasmid containing 190 bp of the promoter of IDX-1 (Ϫ190IDX-LUC) was a gift from Dr. Marc Montminy (The Salk Institute, La Jolla, CA) and has been described previously (40). The luciferase reporter construct containing the proximal 180 bp of the promoter sequence of the human islet sulfonylurea receptor (Ϫ180SUR-LUC) 3 was a gift from Dr. Jorge Ferrer (Harvard Medical School, Boston, MA). All insulin promoter deletions were generated by polymerase chain reaction mutagenesis and subsequently sequenced. For expression of C/EBP␤ in COS-7 cells and in vitro transcription/translation, the plasmid C/EBP␤-pcDNA I was used (41). For bHLH factors, the plasmid shPanI.pBAT14 (42) (hamster homolog of E47; a gift from Dr. M. German, University of California, San Francisco, CA) was used for expression in COS cells. Pan I (rat homolog of E47) was produced by in vitro transcription/translation from the plasmid pARP5/P2 (43) (gift from C. Nelson, University of California, San Francisco). To produce the leucine zipper minus mutant form of Pan I, two point-mutations were introduced into the plasmid pARP5/P2 by polymerase chain reactionbased site-directed mutagenesis (QuikChange® site-directed mutagenesis kit, Stratagene, La Jolla, CA) to replace the second and the third leucines within the heptad leucine repeat of the AD2 of Pan I/E47 by phenylalanines. The mutated sequences were confirmed by sequencing using the dideoxy chain termination method (44) with Sequenase version 2.0 (U.S. Biochemical Corp.). The plasmid for bacterial expression of the GST-C/EBP␤-fusion protein (pGEX-KG) has been described (41).
At 50% confluence (10-cm culture dishes), HIT-T15 or ␤TC-6 cells were transfected with 3 g of the rat insulin I promoter-luciferase plasmid DNA using the DEAE-dextran method in cell suspension (45). HeLa cells were transfected using CaPO 4 (46) with 3 g of the Ϫ410INS-LUC (or deletion constructs) and 0.5 g of C/EBP␤-pcDNA I expression plasmid. After a 48-h incubation, the cells were harvested, and the luciferase activity was determined as described previously (47).
The plasmids containing fusions between the GAL4 DNA-binding domain and either AD1 or AD2 of E47 (48) were a gift from Dr. Roland Stein (Vanderbilt University, Nashville, TN). At 50% confluence (10-cm culture dishes), HIT-T15 cells were cotransfected with 10 g of a luciferase reporter-construct containing a multimerized Gal4-binding site ((GBS) 3 -p59RLG) and one of each of the GAL4 constructs with a C/EBP␤ expression plasmid (C/EBP␤-pcDNA I) or empty vector (pcDNA I) using the DEAE-dextran method in cell suspension (45). Rous sarcoma virus-CAT was used as an internal control for monitoring of transfection efficiency. After a 48-h incubation, the cells were harvested, and the luciferase activity was determined as described previously (47). Values were expressed as means Ϯ S.E. of three independent experiments.
Transfection of COS-7 cells for protein expression was performed by liposomal transfer (Lipofectin®, Life Technologies, Inc.) according to the manufacturer's manual.
In Vitro Transcription/Translation-Recombinant proteins were produced by in vitro transcription/translation with the TNT-coupled reticulocyte lysate system (Promega, Madison, WI) according to the manufacturer's instructions using T7 polymerase for all plasmids transcribed. Each translation reaction was performed in duplicate with and without inclusion of [ 35 S]methionine, and protein identity was confirmed by autoradiography of products separated by SDS-polyacrylamide electrophoresis or Western immunoblot analysis.
In Vivo Labeling, Immunoprecipitation, and GST Pull Down Assays-Proteins from transfected COS-7 cells were labeled in vivo by incubation in L-methionine/L-cysteine-free Dulbecco's modified Eagle's medium (Life Technologies) containing 200 Ci/ml [ 35 S]methionine/ [ 35 S]cysteine (NEN Life Science Products) and 10% dialyzed FBS for 4 h. Nuclear extracts from COS-7 cells and from the ␤-cell lines HIT-T15 and Ins-1 were prepared as described previously (35). For immunoprecipitation, nuclear extracts and in vitro translated protein solutions were adjusted to 200 mM NaCl, 0.1% Nonidet P-40, 50 mM HEPES, 1 mM phenylmethylsulfonyl fluoride, 5 mM EDTA, 0.5 mM dithiothreitol and precleared with protein A-Sepharose (Pharmacia Biotech, Uppsala, Sweden). After the addition of the respective antisera for C/EBP␤ and E47 (Santa Cruz Biotechnology) and incubation at 4°C for 15 h, the immune complexes were precipitated with protein A-Sepharose, washed, and subjected to SDS-PAGE followed by autoradiography or Western immunoblot analysis. For GST pull down analysis, the GST fusion proteins of C/EBP␤ were prepared as described by Ron and Habener (41), except the proteins were not eluted from the glutathione-Sepharose. Interacting proteins were precipitated with the glutathione-Sepharose-coupled fusion proteins in the same buffer used for immunoprecipitation and analyzed by SDS-PAGE autoradiography and Western immunoblot as described.
Statistics-All values were expressed as means Ϯ S.E. Statistical analysis was performed via Student's t test for paired and unpaired values (49).

Potential C/EBP␤-binding Sites in the Promoters of the Rat and Human
Insulin Genes-Inspection of the promoters of the rat insulin I, rat insulin II, and human insulin genes reveals nucleotide sequence elements that resemble the consensus motif that binds the C/EBP family of transcription factors (Fig. 1, A  and B).
Expression of C/EBP␤ in Rat Islets and Cultured ␤-Cell Lines and Regulation by High Glucose Levels-Isolated rat pancreatic islets and several ␤and non-␤-cell lines were assayed for C/EBP␤ expression by Western immunoblot using liver nuclei extracts as a positive control, since C/EBP␤ was originally defined as the liver-enriched activator protein, LAP (14). The 32-kDa C/EBP␤ protein was detected in isolated rat islet whole cell extracts ( Fig. 2A), although the antiserum also recognized a more abundant protein with an apparent molecular mass of 42 kDa. C/EBP␤ was also detected in the nuclear extracts from Ins-1, ␤TC-6, and HIT-T15 cells, which are islet ␤-cell lines derived from rat, mouse, and hamster, respectively.
To determine whether the expression of C/EBP␤ is altered in ␤-cells during chronic or short term exposure to supraphysiologic concentrations of glucose, we used standard in vitro glucose desensitization models (6,30). HIT-T15 cells were serially passaged in RPMI 1640 medium containing 11.1 mM glucose or 0.8 mM glucose for 16 weeks. Since the EC 50 for glucose-stimulated insulin secretion is left-shifted to 1 mM in HIT-T15 cells rather than about 8 mM in normal islets (50), 11.1 mM glucose was chosen as a supraphysiological concentration, and 0.8 mM was considered a physiological concentration of glucose for the HIT-T15 cells. As shown in Fig. 2B, the expression of IDX-1 decreased from week 4 to 16, confirming the published observations (9). In contrast, the expression of C/EBP␤ was markedly enhanced from week 8 to 16. These observations indicate that the level of C/EBP␤ in HIT-T15 cells is up-regulated by prolonged exposure to high glucose concentrations and that C/EBP␤ might serve as a repressor of insulin gene transcription. The increased expression of C/EBP␤ in HIT-T15 cells after chronic exposure to high glucose was prevented by culturing HIT-T15 cells in the RPMI 1640 medium containing 0.8 mM glucose (Fig. 2B). To validate the in vitro long term model and to show an effect of high glucose concentrations on the regulation of insulin secretion, we measured the insulin concentration in the culture medium in response to increasing glucose concentrations in HIT-T15 cells at different passages in 2-h static incubation intervals. Whereas HIT-T15 cells cultured in high glucose displayed a passage-dependent decrease in glucose-responsive insulin secretion (50 Ϯ 7.7% after 16 weeks in 11.1 mM glucose), no such decrease was seen in cells cultured in low glucose (data not shown). The findings of long term high glucose exposure on C/EBP␤ expression were confirmed using Ins-1 cells, which respond to glucose similarly to isolated islets. C/EBP␤ expression was enhanced in Ins-1 cells cultured at 25 mM but not 5.6 mM glucose, whereas IDX-1 levels were decreased in Ins-1 cells cultured at 25 but not at 5.6 mM glucose (data not shown).
To more precisely test the regulation of C/EBP␤, we sought to examine the role of supraphysiologic glucose concentrations in a short term model, which has recently been used to examine IDX-1 expression in response to high glucose concentrations (30). Exposure of Ins-1 cells to 25 mM glucose for 72 h and then reversing the high glucose concentration back to normal resulted in a marked up-regulation of C/EBP␤ after 24 h. The up-regulation of C/EBP␤ was reversible by a subsequent 24-h period in 5.6 mM glucose (Fig. 2C). As a control, the expression of IDX-1 was examined to confirm the down-regulation of IDX-1 protein in this system described previously. We also measured the insulin content of the cells to validate the effect of high glucose on this parameter. A decrease in insulin content during the 72-h high glucose period was observed (from 98 Ϯ 4.3 milliunits/mg protein before, to 37.4 Ϯ 6.1 milliunits/mg protein after 72 h in 25 mM glucose), which was partially reversible by the subsequent culture period of 24 h in low glucose (59.4 Ϯ 6.2 milliunits/mg protein) (data not shown).
Bacterially Expressed C/EBP␤ Binds to Insulin Promoter Sequences-Mapping of the C/EBP␤-binding sites within the rat insulin I promoter was carried out using DNase I protection footprinting assays on a 280-base pair fragment of the 5Јflanking region of the rat insulin I gene (nucleotides Ϫ280 to ϩ1) labeled on the coding strand. Incubation of this labeled fragment of DNA with bacterially expressed C/EBP␤ resulted in three regions of DNase I protection and additional hypersensitive sites, indicating that C/EBP␤ interacts with specific sequence regions in the promoter (Fig. 3A). Counting from the transcription initiation site (ϩ1), the first protected region from nucleotide Ϫ70 to Ϫ86 contains the previously characterized A1 (P1) element (between nucleotides Ϫ64 and Ϫ85). The second region, from nucleotide Ϫ107 to Ϫ121, corresponds to the E1 (Nir) box (between nucleotides Ϫ104 and Ϫ112). DNase I hypersensitive sites flank both the A1 (P1) and E1 (Nir) elements. The third region, from nucleotide Ϫ126 to Ϫ147, however, contains the CEB box and has not been described previously. Notably, the bacterially expressed C/EBP␤ also gave a hypersensitive digestion pattern at the boundaries of the A3/4 (FLAT) element (between nucleotides Ϫ207 and Ϫ227), suggesting that C/EBP␤ may distort the DNA in this region without completely protecting the FLAT element from DNase I digestion. These data confirm that C/EBP␤ binds to the CEB box and suggest that the recombinant protein also binds to the A1 (P1) element and the E1 (Nir) box and may interact with the A3/4 (FLAT) element. C/EBP␤ does not appear to bind to the rat insulin I gene cAMP-response element (CRE, Ϫ184 TGACGTCCAAT Ϫ174 ), although the rat insulin I gene promoter CRE contains a nearly canonical TGACGTCC core sequence for binding of C/EBPs and cAMP-response element-binding protein (CREB).
To examine whether the bacterially expressed C/EBP␤ binds to the insulin promoter sequence element (CEB box), predicted to contain the consensus binding site (Fig. 1A), EMSAs with oligonucleotides corresponding to this region were performed (Fig. 3B). C/EBP␤ forms DNA-protein complexes with a 51-bp oligonucleotide probe comprising the CEB box of the rat insulin I promoter. Binding of C/EBP␤ to this element was also demonstrated by using the corresponding sequences of the rat insulin II and using the human insulin gene promoters as probes (data not shown). These data indicate that the bacterially expressed C/EBP␤ binds to the CEB boxes within the promoters of these three insulin genes as predicted by sequence comparison (Fig. 1B).
To assess the relative affinity of C/EBP␤ interaction with the A1 (P1) element, the E1 (Nir) box, the CEB box, and the A3/4 (FLAT) element, EMSA was performed using the CEB probe in the presence or absence of 30-or 300-fold excesses of the unlabeled oligonucleotides containing the CEB box, mutated CEB box, P1 element, Nir box, or FLAT element. The oligonucleotides containing the Nir box, the P1 element, and the FLAT element competed with the CEB probe for binding to the bacterially expressed C/EBP␤, but with at least 10-fold less efficiency than the unlabeled oligonucleotide containing the CEB box (Fig. 3B). These data suggest that C/EBP␤ interacts with the P1, Nir, and FLAT elements with a relatively lower efficiency compared with the CEB box.
Endogenously Expressed C/EBP␤ in HIT-T15 Cells Binds to the CEB Box-EMSAs were used to characterize the DNAbinding properties of endogenously expressed C/EBP␤ in HIT-T15 cells (Fig. 3C). The DNA probe was the CEB consensus oligonucleotide. Binding reactions were performed with or without the addition of C/EBP␤ antiserum or preimmune serum. One of the slowest migrating complexes for the CEB box probe (lane 2) was disrupted by the addition of C/EBP␤ antiserum (lane 4) but not by the preimmune serum (lane 3). Moreover, the C/EBP␤ antiserum resulted in the appearance of The binding specificities of the protein complex that binds to the probe containing the CEB box were examined by EMSA under conditions of competition with unlabeled oligonucleotides containing the wild type and mutated CEB box (Fig. 3C). The addition to the binding reaction of a 30-or 300-fold unlabeled wild type oligonucleotide containing the CEB box resulted in the abolishment of the C/EBP␤ complex (Fig. 3C,  lanes 7 and 8). In contrast, the unlabeled oligonucleotide containing the mutated CEB box (from 5Ј-agcTGTAAT-3Ј to 5Ј-agcCTGCCG-3Ј) was a much weaker competitor for the binding to the labeled CEB box probe (Fig. 3C, lanes 10 and 11). These observations suggest that the endogenously expressed C/EBP␤ in HIT-T15 cells binds specifically to the CEB box.
Transactivation of the Rat Insulin I Gene Promoter by C/EBP␤ in Non-␤-cells-Because HeLa and BHK-21 cells lack certain ␤-cell-specific transcription factors required for the ex-pression of the insulin gene, such as IDX-1 and BETA-2 (51, 52), they have been widely used to assess the transactivation of the insulin gene by recombinant proteins. The effect of C/EBP␤ on the rat insulin I gene promoter was examined by cotransfection experiments using HeLa and BHK-21 cells with a C/EBP␤ expression vector (pcDNA I) and a reporter construct containing portions of the rat insulin I gene 5Ј-flanking region in the plasmid pXP2 (Ϫ410INS-LUC). The Ϫ410INS-LUC consists of nucleotides Ϫ410 to ϩ49 of the rat insulin I gene linked to the gene encoding the firefly luciferase (LUC) and was activated 22-fold by C/EBP␤ in HeLa cells (Fig. 4A). Under the same conditions, C/EBP␣ also stimulated Ϫ410INS-LUC expression, but the effect was weaker than that of C/EBP␤ (4.2fold; data not shown). The empty expression vector for C/EBP␤, pcDNA I, had no significant effect on Ϫ410INS-LUC expression. In addition, C/EBP␤ did not affect the truncated thymidine kinase promoter of herpes simplex virus (Ϫ81 to ϩ52 bp, pTK81-LUC) that lacks a C/EBP␤-binding site. To examine whether the transactivation of Ϫ410INS-LUC by C/EBP␤ was mediated through interactions with the CEB box, nucleotide FIG. 3. Binding of recombinant and endogenous C/EBP␤ to insulin promoter elements. A, 32 P-labeled probe of the rat insulin I promoter was incubated with recombinant C/EBP␤ before DNase I digestion. Base numbers and control elements are indicated at the right. The arrows denote hypersensitive patterns. B, 32 P-labeled CEB box oligonucleotide was incubated with recombinant C/EBP␤. Unlabeled oligonucleotides comprising the rat insulin I promoter wild type CEB box, mutated CEB box, Nir box, P1 box, and FLAT element were used as competitors. C, 32 P-labeled CEB box oligonucleotide was incubated with nuclear extracts (Nu. Ex.) from HIT-T15 cells. C/EBP␤-specific antiserum (␣C/EBP␤) and preimmune and normal rabbit serum (NRS) were used in supershift experiments; unlabeled wild type or mutant CEB box probe was used as competitor. The asterisk denotes a major DNA-probe complex that probably consists of a C/EBP isoform other than C/EBP␤. substitution mutations were introduced into this site in the rat insulin I promoter from Ϫ125 agcTGTAAT Ϫ133 to Ϫ125 agcCT-GCCG Ϫ133 (CEB mutation). The same mutations of CEB that were introduced into the oligonucleotide probes of the gel shift assays depicted in Fig. 3B resulted in a marked decrease in the binding affinity of C/EBP␤. Mutation of the CEB box significantly reduced the transactivation by C/EBP␤ on the rat insulin I gene promoter (from 22-to 2-fold), suggesting that C/EBP␤ transactivates the rat insulin I gene promoter mainly through the newly identified CEB box in non-␤-cells (Fig. 4A). Other gene promoters containing the C/EBP␤-binding sites, such as an angiotensinogen gene promoter construct, containing four copies of its C/EBP␤-binding site (APRE-LUC), the rat glucagon gene promoter (Ϫ350Glu-LUC) (53) were also activated by C/EBP␤ in HeLa cells, with 190 Ϯ 11-fold (n ϭ 3) and 20 Ϯ 2-fold (n ϭ 8) stimulation, respectively (Fig. 4B). Moreover, the promoters of the transcription factor IDX-1 gene (Ϫ190IDX-LUC) and the sulfonylurea receptor gene (Ϫ180SUR-LUC) were also activated in non-␤-cells (Fig. 4B). Co-transfection of C/EBP␤ and Ϫ410INS-LUC into BHK21 cells (a baby hamster kidney cell line that was used to characterize the effects of the homeodomain proteins Lmx-1 and Cdx-3 on rat insulin I gene transcription (54)) confirmed the findings of the HeLa cell transfection experiments (data not shown). These data suggest that C/EBP␤ is a positive regulatory factor for the rat insulin I gene in non-␤-cells.
Repression of the Rat Insulin I Gene Promoter by C/EBP␤ in ␤-Cells-To examine whether C/EBP␤ also transactivates the

FIG. 4. Transactivation and repression potential of C/EBP␤ in non-␤-and ␤-cells. HeLa cells (A and B) or HIT-T15 cells (C, D, E, and
F) were transiently cotransfected with C/EBP␤ expression plasmid (ϩC/EBP␤) or empty pcDNA I vector (Control) and several reporter-gene constructs as described under "Experimental Procedures." Wild type rat insulin I promoter (Ϫ410INS-LUC), CEB box mutated rat insulin I promoter (CEB-Mutation), truncated thymidine kinase promoter of herpes simplex virus (pTK81-LUC), multimerized C/EBP␤-binding sites of the angiotensinogen promoter (APRE-LUC), glucagon promoter (Ϫ350GLU-LUC), rat IDX-1 promoter (Ϫ190IDX-LUC), human ␤-cell sulfonylurea receptor promoter (Ϫ180SUR-LUC), rat insulin II promoter CAT reporter (Ϫ410INS-II-CAT), and 5Ј-deletion mutant luciferase reporter constructs of the rat insulin I promoter from Ϫ410 to Ϫ90 are shown. rat insulin I gene promoter in ␤-cells, the C/EBP␤ expression vector and Ϫ410INS-LUC were co-transfected into HIT-T15 cells. Surprisingly, C/EBP␤ markedly inhibited the rat insulin I gene promoter activity without affecting the 81 bp of the thymidine kinase promoter (pTK81-LUC) (Fig. 4C). In contrast, both the APRE-LUC and glucagon promoter (Ϫ350Glu-LUC) were activated by C/EBP␤ in HIT-T15 cells, with 23.8 Ϯ 2.1-and 2.3 Ϯ 0.1-fold stimulation, and the promoters of IDX-1 and SUR were stimulated 5.8 Ϯ 0.2 and 2.1 Ϯ 0.3-fold, respectively (Fig. 4D). These findings indicate that the C/EBP␤-mediated repression of the rat insulin I gene promoter activity in HIT-T15 cells is unique to the rat insulin I gene inasmuch as C/EBP␤ activates the glucagon, APRE, IDX-1, and SUR promoters in HIT-T15 cells. C/EBP␣ also inhibited the rat insulin I gene promoter activity in HIT-T15 cells (data not shown). The rat insulin II gene promoter (Ϫ410INS-II-CAT), which also contains a putative C/EBP␤-binding site (Fig. 1B), was also inhibited by C/EBP␤ (Fig. 4E). In contrast to the transactivation of the rat insulin I gene promoter in the non-␤ HeLa and BHK cells, mutation of the CEB box within the 410-bp rat insulin I gene promoter did not affect the inhibition of the mutated rat insulin I gene promoter activity by C/EBP␤ in HIT-T15 cells (Fig. 4C, CEB-Mutation). These observations suggest that C/EBP␤ inhibits the rat insulin I gene promoter through cis-elements or their corresponding transcription factors other than the CEB box and its binding proteins. Similar results were obtained from co-transfection experiments using ␤TC-6 and Ins-1 cells (data not shown).

Localization of the DNA Sequences Important for C/EBP␤ Repression of Rat Insulin I Gene Promoter Activity in ␤-Cells-
The DNA sequences within the rat insulin I 5Ј-flanking region essential for the C/EBP␤-mediated repression of promoter activity was investigated by introducing a series of 5Ј-deletions into a rat insulin I promoter-LUC fusion gene (Fig. 4F). Deletion of the DNA sequence between bp Ϫ410 and Ϫ282 did not affect the C/EBP␤-mediated inhibition of the rat insulin I gene promoter activity in HIT-T15 cells, indicating that this region is not important for the C/EBP␤ action. Subsequent deletions suggest that there are two regions critical for the negative regulation of the rat insulin I gene promoter activity by C/EBP␤. The removal of the sequence between Ϫ282 and Ϫ190 that contains the E2 (Far) box and the A3/4 (FLAT) element resulted in a decreased basal promoter activity but also in a significantly decreased repression by C/EBP␤. The involvement of the sequence proximal to bp Ϫ120, which contains the E1 (Nir) box and the A1 (P1) element, in C/EBP␤-mediated repression of the rat insulin I gene promoter activity was also suggested, although not unequivocally established, because basal promoter activity dropped to background levels upon deletion of the E1 (Nir) element. More important, however, the deletion of the promoter region between Ϫ190 and Ϫ120, which contains the CEB box, did not completely abolish C/EBP␤mediated inhibition of the rat insulin I gene promoter, suggesting that this element is not exclusively mediating the C/EBP␤ effects on this promoter in pancreatic ␤-cells. However, deletion to Ϫ90, which eliminates the E1 (Nir) element eliminates inhibition by C/EBP␤. The results of these experiments furthermore suggest that the rat insulin I gene promoter contains multiple negative regulatory DNA elements or their corresponding DNA-binding protein factors as targets for the C/EBP␤-mediated repression in ␤-cells. The co-transfection of the deletion constructs and the C/EBP␤ expression vector into ␤TC-6 cells gave similar results (data not shown).
Interaction of C/EBP␤ with Basic Helix-Loop-Helix Transcription Factors-Because deletion or mutation of the CEB box from an insulin-promoter-reporter plasmid did not abolish repression of reporter gene activity by C/EBP␤ in ␤-cells (but did abolish activation in non-␤-cells), we reasoned that the molecular mechanism for the inhibition of insulin promoter activity in ␤-cells may be different from direct binding to the CEB box. One possibility is an inhibitory interaction of C/EBP␤ with activating transcription factors of the insulin gene. One family of transcription factors that has been shown to transactivate the insulin gene promoter consists of bHLH proteins, such as E12/E47 and BETA-2. We examined by co-immunoprecipitation experiments whether C/EBP␤ would interact with the bHLH factor E47. Indeed, endogenous E47 immunoreactivity, co-migrating with E47-protein expressed in COS cells (Fig.  5A, lane 2), was readily co-immunoprecipitated from nuclear extracts of the ␤-cell line Ins-1 using an antiserum directed against C/EBP␤ (Fig. 5A, lane 4), providing evidence for a direct protein-protein interaction of these two transcription factors. This molecular interaction was further confirmed in a ␤-cell-independent cell system by co-immunoprecipitation of E47 with C/EBP␤ antiserum from extracts of COS-7 cells, transfected with expression plasmids for C/EBP␤ and E47 and labeled in vivo with [ 35 S]methionine for visualization of the immune complexes by autoradiography (Fig. 5B, upper panel). No specific immune complexes were precipitated from cells, transfected with empty vectors (Mock, lanes 1 and 2). After transfection with E47 expression plasmid alone, the expressed E47 protein could be immunoprecipitated with the E47 antiserum (lane 4), but not with the C/EBP␤ antiserum (lane 3).
Only when both C/EBP␤ and E47 were expressed in COS-7 cells, co-immunoprecipitation of C/EBP␤ and E47 proteins was accomplished by C/EBP␤ antiserum (lane 5). Of note, C/EBP␤ could not be co-immunoprecipitated with the antiserum directed to E47 (Fig. 5B, lane 6, and Fig. 5A, lane 3), a finding we attribute to a possible masking effect of the antigenic epitope in E47 upon interaction with C/EBP␤. The specificity of the C/EBP␤ antiserum is demonstrated by immunoprecipitating only C/EBP␤ protein from extracts of COS-7 cells, transfected with the C/EBP␤ expression vector only (Fig. 5B, lanes 7 and  8). The identity of the bands in the autoradiography was confirmed by Western immunoblotting with specific antisera for C/EBP␤ and E47 (Fig. 5B, lower panels).
Interference of C/EBP␤ with the Leucine Zipper Domain of E47-The bHLH proteins of the E2A family, E12 and E47, contain two structurally and functionally distinct transactivation domains, AD1 and AD2 (55). In contrast to AD1, which is equally active in all cells, the AD2 transcriptional activation domain functions preferentially in pancreatic ␤-cells. Analysis of the amino acid sequence of E47 reveals a heptad leucine repeat within AD2, which is conserved in the E47 proteins of different animal species. Therefore, we reasoned that an interaction of the leucine zipper transcription factor C/EBP␤ with the leucine repeat within AD2 of E47 could underlie the demonstrated physical interaction of both proteins. To further characterize the molecular basis of this interaction, we introduced two point mutations into the leucine repeat domain of E47, changing the first two leucines to phenylalanines (Fig. 5C) and yielding the protein E47-LZ Ϫ . In vitro translated and labeled proteins of C/EBP␤, E47, and E47-LZ Ϫ were subjected to immunoprecipitation, and immune complexes were visualized by autoradiography after fractionation by SDS-PAGE (Fig. 5D,  top). Aliquots of the reticulocyte lysates before immunoprecipitation were included as controls in the electrophoresis (Fig. 5D,  lanes 1-3). In reactions containing only single proteins, both wild type and E47-LZ Ϫ proteins were immunoprecipitated by the E47 antiserum (lanes 4 and 5), and in vitro translated C/EBP␤ protein was precipitated by C/EBP␤ antiserum (lane 6). When mixed together, E47 antiserum precipitated again only wild type E47 (lane 7) or mutated E47 (lane 9) alone. Both wild type E47 and C/EBP␤, however, were co-immunoprecipitated by the C/EBP␤ antiserum (lane 8). In contrast, only C/EBP␤ alone was precipitated by the C/EBP␤ antiserum when used with the mutated E47-LZ Ϫ (lane 10). These findings provide evidence for the notion that C/EBP␤ may interact with E47 via the leucine repeat within AD2. This protein-protein interaction was also confirmed by an antiserum-independent assay, namely the GST pull down procedure. Labeled and in vitro translated wild type E47 specifically bound to GST-C/EBP␤ (Fig. 5E, lane 4). In contrast, the E47-LZ Ϫ mutant protein showed no interaction with the GST-C/EBP␤ fusion protein (lane 5). GST-C/EBP␤ was able to precipitate both wild type E47 and C/EBP␤ (lane 7) but not mutant E47-LZ Ϫ together with C/EBP␤ (lane 8). C/EBP␤ was included in the reactions in lanes 6 -8 to ensure that the GST-C/EBP␤ was functional (that it dimerized with C/EBP␤).
Inhibition of E47 Binding Activity by Interaction with C/EBP␤-To determine whether interactions of C/EBP␤ with E47 affect the binding activity of the E47 homodimer to the E-box DNA response elements within the insulin promoter, we tested the binding of in vitro translated proteins in electrophoretic mobility shift assays on the E1 elements Nir and RIPE3 of the rat insulin I and II promoters, respectively (Fig.  6, A and B). On both elements, the specific complex containing the E47 homodimer (lanes 1), as determined by competition FIG. 5. Physical interaction of C/EBP␤ with E47. A, nuclear extracts from COS cells transfected with empty vector (Mock) or an E47 expression vector (E47) and nuclear extracts from untransfected Ins-1 cells were immunoprecipitated with the indicated antisera (IP␣), and immunoprecipitates were subjected to Western immunoblotting. B, COS cells transfected with empty vector (Mock) or expression plasmids for E47 and/or C/EBP␤ were labeled in vivo with [ 35 S]methionine, and extracts were immunoprecipitated with the indicated antisera (IP␣). Labeled immunoprecipitates were visualized by autoradiography after SDS-PAGE (top). The identity of bands was confirmed by Western immunoblotting (lower parts). The asterisks denote nonspecific bands. C, point mutations changing leucine to phenylalanine introduced within the AD2 of E47. D, 35 S-labeled in vitro translated wild type E47 (E47), leucine zipper mutated E47 (E47-LZ Ϫ ), and C/EBP␤ proteins were co-precipitated with the indicated antisera (IP␣). Aliquots of translation products (lanes 1-3) or immunoprecipitates (lanes 4 -10) were visualized by autoradiography after SDS-PAGE (top). The identity of the bands was confirmed by Western immunoblotting (lower parts). E, 35 S-labeled in vitro translated wild type E47 (E47), leucine zipper mutated E47 (E47-LZ Ϫ ), and C/EBP␤ proteins were precipitated with a GST-C/EBP␤ fusion protein (GST-pull down), and pure translation products or precipitates were visualized by autoradiography after SDS-PAGE (top). The identity of the bands was confirmed by Western immunoblotting (lower parts).
with an excess of cold probe (lanes 2) and antiserum supershift analysis (lanes 3), is either diminished (Nir) or displaced (RIPE3) by the addition of C/EBP␤ (lanes 4). In contrast, the E47-LZ Ϫ mutant protein, although capable of forming a binding complex on the DNA elements (lanes 5), is resistant to interference by C/EBP␤ (lanes 8). This observation further supports the involvement of the leucine repeat sequence within activation domain 2 of E47 in the interaction with C/EBP␤.
Functional Inhibition of the Transcriptional Activation Potential of E47 by C/EBP␤-The question whether the proteinprotein interaction between C/EBP␤ and E47 would also lead to a functional inhibition of the transactivation potential of E47 was assessed by transient transfection into HIT-T15 cells of expression plasmids, encoding fusion proteins of the yeast Gal4-DNA binding domain with either AD1 or AD2 of E47, together with a luciferase reporter construct containing a multimerized Gal4-binding site (GBS) linked to 59 bp of the angiotensinogen gene promoter (Fig. 7). The Gal4 constructs containing AD1 and AD2 were equally active in transactivating the luciferase reporter gene in HeLa cells (not shown), whereas the AD2 construct (Gal4-DBD:E47-(329 -436)) was significantly more active than the AD1 construct (Gal4-DBD:E47-(1-99)) in HIT-T15 cells. Cotransfection of an expression plasmid for C/EBP␤ repressed the activity of Gal4-DBD:E47-(329 -436) (AD2) on the luciferase reporter gene, leaving the transactivation potential of Gal4-DBD:E47-(1-99) (AD1) unaffected (Fig.  7). These findings provide evidence for a functional inhibitory interaction of C/EBP␤ with the AD2 domain of E47 leading to a reduced transactivation potential of E47.

DISCUSSION
Chronic hyperglycemia in patients with type II diabetes mellitus may contribute to defective glucose-induced insulin secretion, a phenomenon that has been attributed to glucose toxicity (4). After culture in high glucose concentrations for 7 days, human islets contain markedly reduced insulin content, a change that can be partially reversed by subsequent culture in lower glucose concentrations (56). Using immortalized ␤-cell lines, it was found that chronic exposure to supraphysiologic glucose concentrations is associated with decreased insulin gene transcription and decreased expression of the insulin gene transactivators IDX-1 and RIPE3b1-binding proteins (5-9). It was recently reported, however, that the chronic glucotoxic alterations of insulin gene expression in the pancreatic ␤-cell line HIT-T15 are only partially reversible upon subsequent lowering of the high glucose levels, although the expression of IDX-1 and RIPE3b1 factors was readily restored (57). These findings imply that an additional inhibitory factor, which reg-FIG. 6. Inhibition of DNA binding of E47 by C/EBP␤. In vitro translated unlabeled wild type E47 (E47) and leucine zipper-mutated E47 (E47-LZ Ϫ ) were used in EMSA with 32 P-labeled doublestranded oligonucleotides comprising the E1 elements of the rat insulin I gene promoter (Nir) (A) and the rat insulin II gene promoter (RIPE3) (B). Supershift and competition experiments were performed with E47 antiserum (␣E47) or a 100-fold excess of unlabeled probe (100x). ulates insulin gene transcription under these conditions, may be involved. In this study, we have identified C/EBP␤ as a transrepressor of insulin gene transcription, which is up-regulated by supraphysiological glucose levels in pancreatic ␤-cell lines. A high affinity binding site for C/EBP␤ in the rat insulin I gene promoter, the CEB box, and several relatively lower affinity sites, namely the A1(P1), the E1 (Nir box), and the A3/4 (FLAT) elements, were identified. DNase I footprint analysis using recombinant C/EBP␤ indicates that C/EBP␤ binds to the CEB box, the Nir box, and the P1 element and may interact with the FLAT element. DNA-protein binding assays using short oligonucleotides indicated that the CEB box is the high affinity binding site, whereas the other sites interact with C/EBP␤ with at least a 10-fold lower affinity compared with the CEB box. Although C/EBP␤ is capable of binding to the CREs of the phosphoenolpyruvate carboxykinase and somatostatin genes (58,59), it did not interact with the rat insulin I gene CRE as indicated by both DNase I footprint analysis and EMSA. C/EBP␣ was shown not to bind to the glucagon CRE (53,58), suggesting that the CRE alone is not sufficient for the binding of C/EBP proteins, and the flanking sequences may play a critical role for the converged binding of C/EBP proteins and CREB.
Previous work has demonstrated that the E2 (Far) box and the A3/4 (FLAT) element, and their counterparts located proximal to the transcription initiation site, termed the E1 (Nir) box and the A1 (P1) element, are the most important cis-acting DNA elements required for rat insulin I gene expression. The E2A gene products, E12, E47, and/or E2-5, bind to the Far and Nir boxes and activate the rat insulin I gene promoter synergistically with the homeodomain proteins IDX-1 (IPF-1/STF-1/ PDX-1), HNF-1␣, and Lmx-1 (52,54,60), which bind to the FLAT and P1 elements. Although C/EBP␤ activates the rat insulin I gene promoter in non-␤-cells through binding to the newly identified CEB box, this interaction does not mediate the repression of the rat insulin I promoter by C/EBP␤ in ␤-cells.
Our studies indicate that C/EBP␤ inhibits rat insulin I gene transcription through physical and functional interaction with the basic helix-loop-helix protein, E47.
The bHLH protein family of transcription factors is divided into three classes according to their DNA-binding properties, structural features, and tissue distribution (61). Factors of the E2A family (E12, E47, E2-5) are ubiquitously expressed members of class A. Class A factors of the E2A family are components of the major ␤-cell nuclear binding complex (insulin enhancer factor, IEF) of the rat insulin I and human insulin gene promoters (62). Furthermore, the class B bHLH factor BETA-2 is expressed specifically in pancreatic ␤and ␣-cells and is reported to heterodimerize with E2A proteins on the RIPE3 element of the rat insulin II promoter, an interaction that is believed to contribute to the tissue specific expression of the insulin II gene (51). Importantly, more than 90% of the overall activity of the rat insulin I promoter activity is attributable to the synergistic transactivation by E2A proteins and homeobox factors (IDX-1, Lmx-1) (3). Thus, it is conceivable that molecular interference of C/EBP␤ with E47 disrupts not only the homo-or heterodimerization of the bHLH factors themselves but also their synergistic transactivation with homeodomain proteins (Fig. 8).
Two transactivation domains (AD1 and AD2) have been identified in E47. The AD2 subdomain contains a heptad leucine repeat sequence. Mutation of this "leucine zipper" altered the transcriptional activity of E47. Interestingly, AD1 functions in a wide variety of tissues and cells, whereas the expression of AD2 activity is largely restricted to pancreatic ␤-cells, supporting a potentially important role for the AD2 domain in regulating gene transcription in ␤-cells. In addition to the ability of the AD2 domain of E47 to contribute to transactivation, we uncovered evidence that the leucine repeat serves as a domain for a direct protein-protein interaction with C/EBP␤; mutations of two of the leucines abrogated this interaction. Furthermore, C/EBP␤ inhibited binding of the E47 homodimer to E-box-containing elements within the rat insulin I and II promoters. Whether this is due to the formation of a classical leucine zipper dimerization between C/EBP␤ and E47 has not been unequivocally established by our studies. A report showing an inhibition of insulin gene transcription by the leucine zipper transcription factor c-Jun via targeting of the AD2 of E47 in ␤-cells (48, 63) parallels our findings in part. c-Jun functionally inhibited the transactivation potential of the AD2 of E47, but in contrast to our observations with C/EBP␤, c-Jun did not appear to physically interact with the AD2 of E47. Therefore, it is conceivable that leucine zipper transcription factors of different families may interact by different mechanisms with the AD2 of E47.
Recently, it has been suggested that E2A factors are not required for insulin gene transcription, based on the observation that mice with a targeted disruption of the E2A gene do not appear to have abnormalities in the morphology of the endocrine pancreas and do not develop overt diabetes (64). Because the basal and stimulated, as well as the tissue-specific, expression of the insulin gene is regulated in a complex manner, however, the absolute contribution of different bHLH proteins to insulin promoter activity within different animal species is largely unknown. E2A gene products represent only one subfamily of bHLH factors, and other ubiquitously expressed members could substitute for E2A in the mouse gene knockout model. This notion is further supported by the observation that BETA-2 knockout mice do develop diabetes (66), a circumstance that may be attributable to the tissue-specific action of this bHLH transcription factor in mice, and the E47/BETA-2heterodimer may also be a target for C/EBP␤-mediated repression of insulin II gene activity. Furthermore, it remains to be determined whether or not E47/BETA-2 heterodimers also bind to and activate E-box-containing elements within the rat insulin I promoter, a finding that would further support the FIG. 8. Model of C/EBP␤-mediated inhibition of insulin gene transcription. Proposed mechanism of inhibitory action of C/EBP␤ exemplified for the rat insulin I gene promoter. A, in the absence of C/EBP␤, E47 and IDX-1, binding to the A and E elements of the rat insulin I gene promoter, constitute two symmetrical enhansons that exhibit synergism in their transactivation potential. B, in the presence of C/EBP␤, the DNA binding of E47 homodimers is inhibited by proteinprotein interaction with C/EBP␤, which disrupts both the transactivation potential of E47 and the synergistic interaction with IDX-1. A similar mechanism is conceivable for the rat insulin II promoter.
importance of C/EBP␤ as a glucose-induced repressor of insulin gene transcription.
In conclusion, C/EBP␤ may serve as a transcription factor mediating the dysregulation of insulin gene expression under conditions of sustained supraphysiological glucose concentrations. In fact, we have extended our studies toward examination of the expression of C/EBP␤ in pancreatic ␤-cells in animal models of diabetes mellitus, with preliminary results implying an involvement of this factor in the pathophysiology of glucotoxic alterations during the development and progression of this disease. 5