Olf-1/early B cell factor is a regulator of glut4 gene expression in 3T3-L1 adipocytes.

A negative regulatory element in the 5'-flanking region of the murine glut4 gene mediates chronic insulin- and cAMP-induced repression in 3T3-L1 adipocytes. Previous work demonstrated that members of the nuclear factor 1 (NF1) family of transcription factors and an unidentified factor bind to and mediate repression from this regulatory element. By using a yeast one-hybrid screen, Olf-1/Early B cell factor (O/E-1) was isolated as a candidate for this unidentified factor. A protein complex from 3T3-L1 adipocyte nuclear extract that bound the negative regulatory element was recognized by O/E-specific antiserum, and binding activity was competed effectively by distinct O/E-binding sequences. O/E binding activity was also detected in nuclear extracts from insulin-responsive, GLUT4-expressing tissues including adipose, skeletal muscle, and heart. Mutations within the negative regulatory element that abolish binding of O/E proteins concomitantly blocked insulin-induced repression in reporter gene assays. These results suggest that one or more members of the O/E transcription factor family function as important regulators of glut4 gene expression and therefore may play a heretofore unanticipated role in glucose homeostasis and insulin signaling.

Facilitative glucose transport across the cell membrane of mammalian cells is mediated by the glucose transporter (GLUT) 1 family of integral membrane proteins (1). One member of the GLUT family, GLUT4, mediates insulin-stimulated glucose uptake primarily in muscle and adipose tissue (2). Insulin triggers glucose uptake by promoting rapid translocation of GLUT4 from intracellular storage sites to the plasma membrane (3). Circulating glucose then enters the cell and is subsequently metabolized. Although skeletal muscle appears to be the primary depository for whole body glucose disposal, adipose tissue functions as an additional storage site and plays a critical role in maintaining both insulin responsiveness and glucose homeostasis. Recent studies in mice demonstrate that adipose-selective diminution of GLUT4 leads to decreased glu-cose tolerance and to impaired in vivo insulin responsiveness in adipose tissue, muscle, and liver (4). Conversely, transgenic mice overexpressing GLUT4 in adipose tissue exhibit enhanced glucose clearance and insulin sensitivity (5). Furthermore, GLUT4 expression is decreased in adipose tissue of humans suffering from obesity or type 2 diabetes (6,7). Decreased insulin sensitivity and impaired glucose homeostasis characterize these metabolic disturbances. Thus, the expression level of GLUT4 in adipose tissue is positively correlated with insulin sensitivity and whole body glucose disposal.
The 3T3-L1 preadipocyte cell line (8) has been used extensively as an in vitro model for studying adipogenesis and adipocyte biology (9). Adipocytes derived from 3T3-L1 preadipocytes faithfully recapitulate many of the molecular, metabolic, and morphological characteristics of tissue adipocytes. 3T3-L1 adipocytes express GLUT4, and insulin acutely stimulates GLUT4 translocation and glucose uptake in these cells (10,11). However, chronic insulin treatment decreases GLUT4 mRNA expression in 3T3-L1 adipocytes (10). Decreased GLUT4 expression in adipose tissue and hyperinsulinemia are associated with obesity and type II diabetes in humans (6,7). Elucidating the molecular mechanisms of glut4 gene regulation in 3T3-L1 adipocytes may prove useful to understanding similar biological events that may also occur in vivo.
Previously we identified (12,18) a response element from the 5Ј-flanking region of the glut4 gene promoter that mediates insulin-and cAMP-induced repression of glut4 gene expression in 3T3-L1 adipocytes. Nuclear factor 1 (NF1) was identified as one of the trans-acting factors that mediates repression from this element (13). A second non-NF1 trans-acting factor was also identified during our studies (12,13,18). In this article, we present evidence to support the identification of Olf-1/early B cell factor (O/E) as the second trans-acting factor that regulates GLUT4 expression from this response element.

EXPERIMENTAL PROCEDURES
Cell Culture-3T3-L1 preadipocytes (8) were maintained in Dulbecco's modified Eagle's medium (DMEM) containing 10% calf serum (Intergen Co., Purchase, NY). Differentiation was induced as described (14) by incubating 2-day postconfluent cells (designated day 0) in DMEM supplemented with 10% fetal bovine serum (Invitrogen) and a hormonal mixture composed of 520 M 3-isobutyl-1-methylxanthine, 1 M dexamethasone, and 167 nM insulin for 48 h. Cells were then incubated in DMEM containing 10% fetal bovine serum and 167 nM insulin for another 48 h, after which they were maintained in DMEM containing 10% fetal bovine serum with a media change every 48 h. All cell culture medium was supplemented throughout with 62.5 g/ml penicillin, 100 g/ml streptomycin, and 8 g/ml biotin.
Yeast One-hybrid Screening and 3T3-L1 Adipocyte Library Construction-The plasmid ⌬L1 (15, 16) containing a yeast GAL1 minimal promoter upstream of a lacZ reporter gene and a URA3 selectable marker was kindly provided by A. Kassam and R. Rachubinski (University of Alberta). The 2 origin was removed from ⌬L1 by digestion with SpeI followed by religation. ⌬L1 lacking the 2 origin was then digested with Ecl136II, which cuts in the middle of the lacZ gene, and a PCR-amplified yeast imidazole glycerol-phosphate dehydratase gene (HIS3) was ligated into the Ecl136II site in frame with the lacZ gene thereby generating ⌬L1/HIS3. ⌬L1/HIS3 encodes a LacZ-HIS3 fusion protein that retains HIS3 activity but does not retain ␤-galactosidase activity. Complementary M2 oligonucleotides (see Fig. 1A for sequence of top strand) with XhoI-compatible overhangs were annealed, and three copies of the annealed duplex DNA were inserted into the XhoI site of ⌬L1/HIS3 upstream of the GAL1 minimal promoter. The resultant reporter plasmid, 3XM2⌬L1/HIS3, was linearized with ApaI and transformed into Saccharomyces cerevisiae strain YCB436 (MATa, ade2⌬::hisG, his3⌬200, leu2⌬1, lys2⌬202, trp1⌬63, ura3- 52,Ref. 17). Stable genomic integrants were selected on synthetic medium lacking uracil and verified by Southern blotting thus generating the yeast reporter strain YCB436::3XM2⌬L1/HIS3.
The adipocyte library was constructed using a commercially available kit as per the manufacturer's instructions (HybriZAP 2.1 XR, Stratagene). Briefly, DNase-treated poly(A) ϩ mRNA isolated from 3T3-L1 adipocytes at day 8 of the differentiation protocol was used as starting material for library construction. Both random primed cDNA and oligo(dT) primed cDNA were generated, ligated into a predigested vector (HybriZAP-2.1), packaged, and subsequently excised to yield a LEU2-selectable plasmid library of 3T3-L1 adipocyte cDNAs fused to a sequence encoding the activation domain of GAL4 (backbone vector, pAD-GAL4 -2.1). The plasmid library was transformed into S. cerevisiae strain BY4734 (MAT␣, his3⌬200, leu2⌬0, met15⌬0, trp1⌬63, ura3⌬0, Ref. 17), and transformants were selected on synthetic medium lacking leucine. A large number of transformants were scraped from plates, pooled, resuspended in liquid medium containing 10% Me 2 SO (v/v), and frozen in aliquots at Ϫ80°C.
The yeast one-hybrid screen was initiated by growing a 100-ml culture of the yeast reporter strain YCB436::3XM2⌬L1/HIS3 (see above) at 30°C to an A 600 Ͼ0.8. An aliquot of the frozen yeast strain BY4734 containing the adipocyte library (see above) was thawed and mated with YCB436::3XM2⌬L1/HIS3 in a 50-ml volume of 2ϫ YPDA ϩ 50 g/ml kanamycin for 20 h at 30°C with gentle shaking (50 rpm). The resulting diploid cells were then plated on 50 100-mm culture plates containing synthetic medium lacking uracil (to select for the reporter strain), leucine (to select for the library plasmid), and histidine (to select for HIS3 reporter activity). Library plasmids were recovered from colonies that grew on the screening plates, shuttled into Escherichia coli, and sequenced.
Reporter Constructs-The wild type GLUT4 reporter (-785GLUT4/ CAT) contains the 785 bases of 5Ј-flanking sequence of the murine glut4 gene, the GLUT4 transcription initiation site, 171 bp of GLUT4 5Јuntranslated sequence, and the coding sequence for the bacterial chloramphenicol acetyltransferase (CAT) gene (13). Reporter genes M3 and M4 contained the indicated mutations (see Fig. 1A) in the context of Ϫ785GLUT4/CAT (13,18).
Stable Transfections-Stable 3T3-L1 cell lines expressing promoterreporter gene constructs were prepared as described (12,13,18). For each construct, two or more independent pools of 20 -50 foci were analyzed. Cells were propagated as preadipocytes and induced to differentiate into adipocytes as described (12,13,18). On day 9 after induction the fully differentiated adipocytes were treated with 1 M insulin (Roche Molecular Biochemical) as indicated. Total RNA and nuclear extracts were purified as described (12,13,18). The level of expression of endogenous glut4 mRNA was quantitated by Northern analysis using a 1.7-kb murine glut4 cDNA probe (12,13,18). Reporter gene mRNA levels were quantitated using a competitive reverse transcriptase-PCR assay as described previously (12,13,18).
Western Immunoblot-30 g of nuclear extract was boiled in an SDS sample loading buffer (2% SDS, 10% glycerol, 100 mM dithiothreitol, 50 mM Tris-HCl, pH 6.8) for 10 min. The samples were loaded on a 10% denaturing SDS-polyacrylamide gel and processed as described previously (13). Blots were probed with a rabbit polyclonal antibody that recognizes all known O/E family members (19). The O/E-1 positive control was generated by transiently transfecting one 60-mm dish of preconfluent 293T cells with 1 g of cytomegalovirus promoter-driven O/E-1 expression plasmid (pCIS-O/E-1(8), Refs. 19 and 20) using the calcium phosphate co-precipitation method (21). Mock-transfected cells were treated in an identical manner but did not receive any expression plasmid. 24 h after transfection, the cells were lysed in an SDS-containing buffer (1% SDS, 60 mM Tris-HCl, pH 8.0) and boiled, and 0.5 g of protein from the cell lysates were loaded onto the gel. S. Wang and R. Reed (The Johns Hopkins University School of Medicine) generously provided the O/E antibody and pCIS-O/E-1 expression plasmid.
Gel Mobility Shift Assays and Preparation of Partially Purified 1a Binding Activity-DNA binding assays were conducted as described previously (12,13). 1a binding activity was purified from differentiated 3T3-L1 adipocytes using an affinity chromatography technique as described previously (22). Briefly, an oligonucleotide affinity column was prepared by linking concatenated M2 oligonucleotide to CNBr-activated Sepharose 4B (Amersham Biosciences). 30 mg of nuclear extract from 3T3-L1 adipocytes was bound to DNA-cellulose in buffer containing 100 mM KCl and then eluted as a single peak at 280 mM KCl by gradient elution. Fractions containing 1a binding activity were identified by gel mobility shift assay using the M2 oligonucleotide as the probe, desalted on a PD-10 column (Amersham Biosciences), and then bound to the oligonucleotide affinity column in the presence of poly(dI-dC)⅐poly(dI-dC) and 150 mM KCl. 1a binding activity was eluted as a single peak at 280 mM KCl and was purified over 200-fold that of crude nuclear extract.

Identification of O/E-1 as a 3T3-L1 Adipocyte Factor That
Binds the GLUT4 Insulin-response Element-Previously, we isolated a regulatory element from the glut4 promoter that mediates chronic insulin-and cAMP-induced repression of glut4 gene expression in 3T3-L1 adipocytes (12,18). We also identified two distinct factors from adipocyte nuclear extract that bind to and mediate repression from this negative regulatory element (13). One of these factors is represented by one or more members of the NF1 family of DNA-binding proteins (13). A yeast one-hybrid screen was undertaken to identify the second unknown factor referred to as 1a. Three copies of duplex DNA containing the sequence of the M2 oligonucleotide were linked to a minimal yeast promoter and placed upstream of a HIS3 gene to generate a reporter plasmid (Fig. 1, A and B, also see "Experimental Procedures" for details). Preliminary DNA binding assays indicated that the unknown factor bound the M2 oligonucleotide strongly ( Fig. 2 and Refs. 12, 13, and 18). In contrast, NF1 family members and other 3T3-L1 adipocyte nuclear proteins did not bind the M2 oligonucleotide (12,13,18). The reporter plasmid was integrated into the genome of an S. cerevisiae strain, and a plasmid library of 3T3-L1 adipocyte cDNAs fused to a sequence encoding a GAL4 activation domain was screened. cDNAs that encode a protein that binds the M2 oligonucleotide are presumed to tether the activation domain of GAL4 to the reporter gene promoter thereby activating HIS3 gene expression and enabling the yeast to grow in the absence of histidine (Fig. 1B). Seven reproducible clones emerged from this screen. The GAL4 activation domain alone did not activate reporter gene expression as expected (Fig. 1C). All seven clones exhibited specific activation of the M2-containing reporter plasmid and did not activate the parental reporter plasmid or a reporter plasmid containing an unrelated oligonucleotide in place of M2 ( Fig. 1C and data not shown). Sequence analysis revealed that all seven clones encoded the same previously identified protein termed Olf-1/Early B cell factor (O/E-1, Refs. 19, 20, and 23). Two clones contained cDNAs encoding the entire O/E-1 protein, and the remaining five clones encoded slightly truncated proteins beginning at Ser 11 of the O/E-1 amino acid sequence.
O/E Proteins Are Present in 3T3-L1 Adipocyte Nuclear Extract and Bind the GLUT4 Insulin-response Element-DNA binding assays were conducted to support the identification of O/E-1 as the unknown adipocyte protein. A radioactively labeled oligonucleotide containing the murine GLUT4 negative regulatory element was used as a probe (see Fig. 1A for sequence). Nuclear extract was prepared from 3T3-L1 adipocytes, and the extract was incubated with the probe, and protein-DNA complexes were resolved by nondenaturing gel electrophoresis. As described in our previous studies (12,13,18), three major complexes formed (Fig. 2, lane 1, also see Fig. 6). Proteins contained within complexes 1b and 2 have been identified as NF1 family members (13). Consistent with those findings, an unlabeled competitor oligonucleotide containing a consen-sus NF1-binding site (oligonucleotide N) prevented visualization of the two faster migrating NF1-containing complexes leaving only the 1a complex (Fig. 2, lane 3). Upon incubation of the binding reaction with antiserum that recognizes O/E proteins, the mobility of the 1a complex was specifically retarded (Fig. 2, lanes 2 and 4). In contrast, incubation with nonimmune serum or an NF1 antibody did not affect the mobility of the 1a complex (Fig. 2, lanes 5 and 6; note that the NF-1 antibody retarded the mobility of the complexes other than the 1a complex, data not shown and Ref. 13). The O/E binding sequence from the regulatory region of the mb-1 gene (see "Discussion") specifically competed away the 1a complex as did the M2 oligonucleotide, the response element used in the yeast one-hybrid screen (Fig. 2, lanes 8, 10, and 13). In contrast, oligonucleotides M1 and M4 did not affect the 1a complex (Fig. 2, lanes  11 and 12). When the mutation-containing oligonucleotides were radioactively labeled and used as probes, the results confirmed the findings of the competition experiments, i.e. oligonucleotide M2 was bound by nuclear proteins to form the 1a complex, and oligonucleotides M1 and M4 were not (data not shown).
The consensus binding sequence for O/E proteins consists of the palindromic core sequence 5Ј-CCCnnGGG-3Ј (19, 23-25). Thus, the sequence contained within the GLUT4 regulatory element (5Ј-CCCTTGGG-3Ј, see Fig. 1A) is a perfect core O/Ebinding sequence. The M1 and M4 oligonucleotides, which are not bound by nuclear proteins to form the 1a complex, contain mutations that disrupt each of the core half-sites of the O/Ebinding sequence. In contrast, the M2 oligonucleotide, which is bound by nuclear proteins to form the 1a complex, contains a mutation that does not disrupt the core O/E-binding sequence. These results are consistent with the identification of an O/E family member as at least one component of the unknown adipocyte protein complex formerly termed 1a.
The O/E-binding Site within the glut4 Promoter Is Required for Insulin-induced Repression-A previously described assay (see "Experimental Procedures" and Refs. 12, 13, and 18) was used to examine the contribution of the identified O/E-binding site to insulin-induced repression of glut4 gene expression. A portion of the glut4 gene promoter containing the negative insulin-response element was placed upstream of a CAT reporter gene and stably integrated into 3T3-L1 preadipoctyes. When these cells were differentiated into adipocytes and then treated with insulin, both the CAT reporter gene and endogenous glut4 gene were repressed by ϳ50% when compared with untreated controls (see wild type, Fig. 3). Mutations within the insulin-response element that do not affect either NF1 or O/E binding activity did not affect the ability of insulin to repress FIG. 1. Use of a yeast one-hybrid screen to isolate a candidate for factor 1a. A, sequence of the insulin-response element located in the murine glut4 gene promoter (glut4 IRE). Shown are the top strands of the wild type glut4 IRE, glut4 IRE mutations (M1, M2, M3, and M4), NF1-binding sequence (N), and the O/E-binding site from the mb-1 gene promoter (mb-1). A summary of the proteins that bind each sequence as determined previously (12,13) is also shown. B, schematic representation of the yeast one-hybrid screen. Shown is the integrated reporter containing three copies of the M2 oligonucleotide (see A for sequence) upstream of a GAL1 minimal promoter (GAL1 P) linked to a HIS3 reporter gene (see "Experimental Procedures" for details). The unknown adipocyte factor 1a is illustrated as a protein fused to the GAL4 activation domain (GAL4AD). Binding of 1a to one or more M2 sites tethers the GAL4AD in proximity to the GAL1 promoter. Reporter gene expression is then activated (shown as an arrow) allowing the S. cerevisiae strain to grow on synthetic medium lacking histidine. C, a series of S. cerevisiae strains containing different integrated reporter constructs (see below) were mated with a second S. cerevisiae strain transformed with either a GAL4 activation domain-encoded vector alone (GAL4AD) or a GAL4AD vector containing the cDNA of clone 1 from the screen (GAL4ADϩ1a.1). Note that clone 1 was one of seven clones isolated during the screening process, and all seven clones gave similar results (data not shown). Three independent mated diploids from each group were plated on synthetic medium (SM) lacking the indicated amino acids. Synthetic medium lacking uracil and leucine (SM-Ura-Leu) was used to select for the reporter and GAL4AD-encoding plasmid. Synthetic medium lacking uracil, leucine, and histidine (SM-Ura-Leu-His) was used to select for HIS3 reporter activity. Integrated reporters are indicated as follows: the parental reporter vector (⌬L1/HIS3), the M2-containing reporter vector (3XM2⌬L1/HIS3, illustrated in B above), and a reporter vector containing an oligonucleotide unrelated to M2 (ARE7⌬L1/HIS3).  (see M4, Fig. 3). These results indicate that the O/E-binding site located within the negative regulatory element of the glut4 5Ј-flanking sequence is required for insulin-mediated repression.
3T3-L1 Preadipocytes, White Adipose Tissue, and Other Insulin-responsive, GLUT4-expressing Tissues Express O/E Proteins-Experiments were conducted to characterize the expression pattern of O/E proteins in the 3T3-L1 cell line and in white adipose tissue. Nuclear extracts from 3T3-L1 preadipocytes and adipocytes were subjected to immunoblot analysis using antiserum that recognizes all known O/E family members. Immunoreactive bands were detected in both preadipocytes and adipocytes (Fig. 4A, lanes 3 and 4). There is a modest induction of O/E protein during the course of 3T3-L1 differentiation (Fig. 4B). Similar but less intense bands were also observed in mouse white adipose tissue nuclear extract (Fig.  4A, lane 5). As expected, a similar immunoreactive band was detected in 293T cells transfected with an O/E-1 expression plasmid but was not detected in mock-transfected cells (Fig.  4A, lanes 1 and 2).
The presence of O/E proteins in 3T3-L1 nuclear extracts was confirmed further by using DNA binding assays. When incubated with the glut4 insulin-response element probe, extracts from both preadipocytes and adipocytes formed similar com-

FIG. 2. O/E-like binding activity is present in 3T3
-L1 adipocyte nuclear extract. 3 g of nuclear extract from 3T3-L1 adipocytes was incubated with a radioactively labeled double-stranded oligonucleotide corresponding to the glut4 insulin-response element (see Fig. 1A). Where indicated, a 100-fold excess of one or more unlabeled doublestranded oligonucleotides was added as competitor (see Fig. 1A for oligonucleotide sequences; mb in Fig. 2 corresponds to mb-1 in Fig. 1A). Also where indicated, 4 l of the following antibodies were added to the binding reaction: anti-O/E (O), anti-NF1 (N), or nonimmune control (C). The three adipocyte nuclear protein-DNA complexes (1a, 1b, and 2) are indicated with arrows.

FIG. 3. Effect of insulin on glut4 promoter/CAT reporter gene expression in 3T3-L1 adipocytes.
Stable 3T3-L1 cell lines expressing glut4 promoter/CAT reporter genes containing the wild type glut4 IRE, M3 mutation, or M4 mutation (see Fig. 1A for sequence) were differentiated into adipocytes and treated with insulin (or not for controls), and cellular RNA was harvested. Endogenous glut4 mRNA and CAT reporter mRNA were measured (see "Experimental Procedures" for details). Results are expressed as a ratio of the expression level in insulin- Nuclear extract was prepared from cells every 24 h over the course of a 6-day differentiation protocol. C, DNA binding assays using 3 g of nuclear extract from 3T3-L1 preadipocytes (Pre) or adipocytes (Ad) were conducted as described in Fig. 2. Where indicated, unlabeled competitor oligonucleotides and antiserum were added to the binding reaction also as described under Fig. 2. N, NF1 competitor; mb, mb-1 competitor; O, O/E antibody. The 1a complex is indicated with an arrow. plexes (Fig. 4C, lanes 1 and 5). As shown previously, addition of an unlabeled NF1 competitor oligonucleotide prevented visualization of the faster migrating complexes but did not affect the slower migrating O/E-containing complexes (Fig. 4C, lanes 2  and 6). In contrast, an unlabeled oligonucleotide containing the O/E-binding site from the mb-1 gene promoter competed away the slower migrating complexes (Fig. 4C, lanes 3 and 7). Furthermore, addition of O/E antibodies to the DNA binding reaction retarded the migration of the slower migrating complexes (Fig. 4C, lanes 4 and 8).
Nuclear extracts prepared from several insulin-responsive tissues were incubated with the glut4 insulin-response element probe to examine the presence of O/E proteins. As shown previously, extract from 3T3-L1 adipocytes formed several complexes of which the slowest migrating O/E-containing complex was not competed away by addition of an unlabeled NF1binding oligonucleotide (Fig. 5, lanes 1 and 2). In contrast, the O/E-binding site from the mb-1 gene promoter effectively prevented visualization of the O/E-containing complex (Fig. 5, lane  3). Similar results, albeit to different degrees, were obtained when examining nuclear extracts from white adipose tissue (Fig. 5, lanes 4 -6), cardiac muscle (Fig. 5, lanes 7-9), skeletal muscle (Fig. 5, lanes 13-15), and brain and spleen (data not shown for the last two tissues). Nuclear extract from liver did form complexes with the glut4 probe, but the NF1 competitor oligonucleotide alone prevented visualization of the majority of these complexes suggesting that this tissue does not contain appreciable amounts of O/E binding activity. These results strongly suggest that one or more O/E proteins are expressed in several insulin-responsive, GLUT4-expressing tissues including white adipose tissue, cardiac muscle, and skeletal muscle.
O/E and NF1 Proteins Compete for Binding to the glut4 Insulin-response Element-Results from previous studies (12,13,18) in combination with those presented here suggest that both NF1 and O/E transcription factor family members mediate repression of glut4 gene expression in 3T3-L1 adipocytes in response to insulin. The core binding elements for both O/E and NF1 proteins are palindromic sequences (see Fig. 1A). In the glut4 gene, these elements overlap so that the 3Ј half-site of the O/E-binding site and the 5Ј half-site of the NF1-binding site are contained within the same DNA sequence (see Fig. 1A). This would suggest that O/E and NF1 proteins compete for binding to this region. Therefore, DNA binding assays were carried out to examine whether O/E and NF1 proteins influence each other's interaction with the glut4 insulin-response element. 3T3-L1 adipocyte nuclear extract was incubated with the glut4response element probe in the presence of increasing amounts of competitor M4 oligonucleotide which is bound by NF1 proteins but is not bound by O/E proteins. Binding of O/E proteins to the GLUT4 probe increased in proportion to the amount of competitor M4 oligonucleotide added (Fig. 6A, lanes 1-6). Conversely, as increasing amounts of O/E proteins are bound to the GLUT4 probe, less NF1-containing complexes are formed (Fig.  6B, lanes 1-4). These results suggest that NF1 and O/E family members compete for binding to the GLUT4 insulin-response element.

DISCUSSION
Data presented here show that one or more members of the O/E family bind to and regulate transcription of the glut4 gene from a bona fide regulatory element located within the glut4 gene promoter. The O/E family is composed of unique helixloop-helix-related transcription factors that bind DNA as homodimers and heterodimers (19,20,23,25). Two groups independently and simultaneously identified the first member of the O/E family of transcription factors, O/E-1. Hagman et al. (23) identified EBF as a transcriptional regulator of the mb-1 gene (encoding Ig␣), a gene specifically expressed during the early stages of B cell lymphopoiesis. O/E-1 null mice lack immunoglobulin-expressing B cells thus establishing a critical role for O/E-1 in B cell differentiation (26). Wang and Reed (19) simultaneously reported the cloning of Olf-1 as a transcrip- FIG. 5. O/E proteins are detected in nuclear extracts from insulin-responsive, GLUT4-expressing tissues. DNA binding assays using nuclear extracts from 3T3-L1 adipocytes (Ad), white adipose tissue (WAT), cardiac muscle (Ht), liver (Li), and skeletal muscle (Sk) were conducted as described in Fig. 2. Where indicated, unlabeled competitor oligonucleotides were added to the binding reaction also as described under Fig. 2. N, NF1 competitor; mb, mb-1 competitor. The 1a complex is indicated with an arrow. All extracts were obtained from mouse tissues, and the following amount of nuclear protein was added per binding reaction: WAT, 12 g; Ht, 60 g; Li, 100 g; Sk, 100 g. For comparison 8 g of nuclear extract from 3T3-L1 adipocytes was included (lanes 1-3). 6. O/E and NF1 proteins compete for binding to the GLUT4 insulin-response element. DNA binding assays using 6 g of nuclear extract from 3T3-L1 adipocytes and the wild type GLUT4 IRE probe. A, 50-5000-fold excess M4 competitor oligonucleotide was added to the binding reaction as indicated. B, 5000-fold excess M2 competitor oligonucleotide or partially purified 1a protein (see "Experimental Procedures") was added to the binding reaction as indicated. The three adipocyte nuclear protein-DNA complexes are indicated with arrows (1a is the O/E-containing complex; 1b and 2 are the NF1-containing complexes). tional regulator of a group of olfactory neuron-specific genes. At least three distinct O/E family members (O/E-1, Ϫ2, and Ϫ3) with Ͼ75% identity at the amino acid level have since been identified (20). Alternatively spliced messages encoding O/E-1 and Ϫ2 give rise to O/E-1(0), O/E-1(8), O/E-2(0S), and O/E-2(9L) (Ref. 20). The original descriptions of Olf-1 (19) and EBF (23) differed only by an 8-amino acid sequence located in EBF but missing in Olf-1. This difference arises by alternative splicing of an optionally included exon that encodes the additional 8 amino acids, and Olf-1 and EBF have been renamed O/E-1(0) and O/E-1 (8), respectively (20). The functional significance of the differences among the various O/E proteins has not been established, and the potential for functional diversity upon heterodimerization between O/E family members is not known. Homologs of mammalian O/E proteins have been reported in Drosophila melanogaster (27), Xenopus laevis (28,29), Danio rerio (30), and Caenorhabditis elegans (31) suggesting evolutionary conservation of this signaling component.
It Our results confirm many of these findings as O/E proteins and/or O/E DNA binding activity were detected in adipose tissue, skeletal muscle, heart, brain, and spleen but not readily detected in liver. We conclude that at least O/E-1 is expressed in adipose tissue, skeletal muscle, and heart and that in these tissues O/E-1 may function in vivo as a modulator of GLUT4 expression.
Our current results from studies using 3T3-L1 adipocytes indicate that O/E negatively regulates glut4 gene expression in response to insulin. Previously, we had demonstrated that the glut4 promoter containing the M4 mutation, which disrupts O/E binding, is not fully responsive to repression by cAMP (18), indicating that O/E proteins also participate in the regulation of the glut4 promoter by cAMP. Thus far, O/E family members have been shown to function as positive regulators of gene expression (19,20,23). The zinc finger protein Roaz was identified during a two-hybrid screen as an O/E-1 interacting protein (32). Roaz is a negative regulator of O/E-mediated transcriptional activation and appears to exert a negative influence by perturbing the DNA binding ability of O/E proteins (32). We considered the possibility that O/E positively regulates glut4 gene expression and that insulin promotes down-regulation of O/E protein or perturbs O/E DNA binding activity. Immunoblot analysis of and DNA binding experiments with nuclear extracts from 3T3-L1 adipocytes treated with insulin showed no obvious changes in O/E protein level or apparent mobility and no changes in DNA binding activity or protein-DNA complex mobility (data not shown). These results are consistent with our previous findings (13) that insulin does not appear to affect the O/E-DNA complex as assessed by gel mobility shift assay, although at the time the identity of the 1a-containing complex was unknown. Although we have been unable to detect an insulin-induced change in mobility of O/E proteins in either Western analysis or gel mobility shift assays, it is possible that insulin promotes changes in phosphorylation or some other form of covalent modification that may affect the transcriptional activation potential of O/E. Studies are underway to examine the mechanism through which O/E proteins mediate repression of the glut4 promoter by insulin.
Results presented here also suggest that NF1 and O/E family members compete for binding to the GLUT4 negative-response element. The functional significance of this competition in relation to glut4 gene expression is not clear at this point. However, it is intriguing to mention that NF1 family members have been implicated as regulators of olfactory gene expression including known O/E target genes (33,34). Additional experiments will be required to more thoroughly examine cross-talk between NF1-and O/E-mediated signaling pathways.
We and others (10,35,36) have demonstrated that prolonged exposure of 3T3-L1 adipocytes to insulin decreases expression of GLUT4. How this relates to the regulation of GLUT4 expression in adipose tissue in vivo is uncertain. On the one hand, GLUT4 expression is decreased in adipose tissue in hyperinsulinemic, insulin-resistant states. This is true in animal models of insulin resistance including high fat diet-induced obesity and genetic models of obesity and insulin resistance in rodents (37), as well as in the insulin resistance of obesity and type 2 diabetes mellitus in humans (6,7). On the other hand, in vivo experiments have suggested that insulin increases GLUT4 expression in adipose tissue. Insulin depletion by fasting or streptozotocin treatment decreases adipose tissue GLUT4 levels in rodents; subsequent refeeding or insulin treatment restores GLUT4 expression levels (38 -41). In addition, insulin treatment by continuous infusion increases adipose GLUT4 expression in rats (42)(43)(44) and in nondiabetic humans (45). The in vivo experiments are clearly more complex than the treatment of cells in tissue culture, so that the effects seen with insulin treatment in vivo are likely to be both a response to the insulin and the indirect effect of changes in other hormones induced by the insulin. It is also possible, of course, that in this situation, 3T3-L1 adipocytes do not faithfully replicate the response of tissue adipocytes. However, there is evidence that the regulatory element we have been studying in the 3T3-L1 adipocyte model also mediates regulation of GLUT4 expression in vivo. Oshel et al. (39) have found that the homologous region of the human GLUT4 gene, which is 74% identical in sequence to the mouse gene, is necessary for the appropriate tissue-specific expression of GLUT4. In addition, Tsunoda et al. (46) have demonstrated that the region of the glut4 gene that mediates high fat diet-induced repression of GLUT4 in mice includes this response element.
The implications of identifying O/E proteins as mediators of insulin-regulated gene expression likely extend beyond its role in regulating GLUT4 expression in adipose tissue. As noted above, studying gene regulation by insulin in vivo is complicated by the secondary changes that insulin induces. Nonetheless, we have identified O/E activity in skeletal muscle, and prolonged insulin infusion was found to stimulate GLUT4 expression in muscle in normal humans (45) and in rats in some studies (43), although not in others (44). Other possible tissues where O/E proteins may mediate gene regulation by insulin would include the brain and developing lymphocytes. We have found that nuclear extract prepared from whole brain contains a significant amount of O/E activity (data not shown), strongly suggesting that O/E expression in the brain is more extensive than being restricted to olfactory neurons. The presence of both insulin receptors (47) and GLUT4 (48,49) in neurons in the hypothalamus and the role of insulin receptor signaling in the control of appetite (50) make this an important area of future research. Significantly, a recent report (51) demonstrated that hyperinsulinemia decreased GLUT4 expression in the hypothalamus. The effect of insulin on lymphopoiesis and lymphocyte function is not well characterized, although insulin receptor expression in activated B lymphocytes has been reported (52)(53)(54). T and B lymphocytes from hyperinsulinemic obese patients, but not from lean patients, are compromised in their ability to be stimulated by mitogenic agents (55). These findings (56) and a general impairment of immunity in states of obesity suggest that hyperinsulinemic states may compromise the immune system. When considering that insulin may regulate O/E activity and that O/E-1 is a critical regulator of B cell differentiation, it will be interesting to determine whether insulin, acting through O/E pathways, affects immune function.
A functional role for O/E family members has been established in B cell differentiation (23,26), odorant signal transduction (19), and neuronal development (27,57). O/E-1 null mice exhibit a severe defect in B cell differentiation (26). Otherwise these mice appear to be normal, and nonlymphoid tissues expressing O/E-1, such as the olfactory epithelium and adipose tissue, show no apparent signs of abnormality (26). Functional redundancy within the O/E family in certain tissues may explain the lack of additional phenotypes in the O/E-1 null mice. Alternatively, perturbation of O/E-mediated glut4 gene expression in O/E-1 null mice may result in a metabolic phenotype that is not readily observed. It may be of interest to re-evaluate O/E-1 null mice when considering the data presented here. In any case, our results suggest that the functional roles of the O/E transcription factor family may be extended to include a role in glucose homeostasis and insulin signaling.