Tumor necrosis factor-alpha regulation of glucose transporter (GLUT1) mRNA turnover. Contribution of the 3'-untranslated region of the GLUT1 message.

In the current study we report on the contribution of the GLUT1 3'-untranslated region (UTR) to the stability of the GLUT1 mRNA. To facilitate these investigations, a hybrid construct was prepared by insertion of the GLUT1 3'-UTR into a normally stable reporter gene coding for preproinsulin. The GLUT1 3'-UTR conferred lability to the otherwise long lived construct and transferred an ability to be stabilized in response to treatment with the cytokine, tumor necrosis factor-alpha (TNF). The destabilizing element has been mapped to a region located between bases 2242 and 2347 of the GLUT1 3'-UTR; this same region also mediates the stabilization response to TNF. In vitro RNA-protein binding assays using protein extracts from control and TNF-treated cells demonstrated that two proteins, one of 37 kDa and the other of 40 kDa, recognized sequence elements within the stability-determining region and were up-regulated in response to TNF treatment. The RNA-binding activity of these proteins coincides with the stabilization of the GLUT1 message, suggesting that they may be involved in regulation of the turnover of this message.

Glucose transport across the plasma membrane is the ratelimiting step for subsequent glucose metabolism and energy production within the cell (1). This process is catalyzed by a family of tissue-specific integral membrane proteins known as the glucose transporters (GLUTs) 1 (reviewed by Bell et al. (2)). Transporter activity can be regulated by hormones and growth factors through the redistribution of transporter proteins from intracellular vesicles to the plasma membrane (translocation), modulation of transporter intrinsic activity, and stimulation of new protein synthesis through increased transcription and/or alteration in transcript stability (reviewed by McGowan et al. (3)).
Previously, we have demonstrated that treatment of quiescent 3T3-L1 preadipocytes with the cytokine tumor necrosis factor-␣ (TNF) produces a 3-4-fold stabilization of the GLUT1 (HepG2/brain/erythrocyte glucose transporter) mRNA that leads to a parallel increase in glucose transport activity within these cells (4). Coincident with the stabilization of the message was the up-regulation of an RNA binding protein(s) exhibiting specificity for an in vitro synthesized 80-base transcript containing four back to back copies of the destabilizing motif, AUUUA. During this same period of message stabilization, proteins with molecular masses of 37 and 40 kDa, with apparent binding specificity for the GLUT1 3Ј-UTR were also upregulated (5). While the GLUT1 3Ј-UTR is AU-rich, the actual sequence to which these proteins bound was not localized, and identity between the AU-rich binding protein and the GLUT1 3Ј-UTR binding proteins could not be established. However, their binding activity increased coordinately with the t1 ⁄2 of the message, consistent with a hypothesis of involvement in stabilization of the GLUT1 transcript (5).
To determine if sequences within the GLUT1 3Ј-UTR regulate decay, we stably transfected 3T3-L1 preadipocytes with a set of chimeric reporter constructs. These constructs contained the 5Ј-UTR and coding region of preproinsulin as a reporter gene, attached to the normal and various truncated versions of the GLUT1 3Ј-UTR. The preproinsulin gene was selected as a reporter for the specific reason that, similar to the GLUT4 gene, the resultant transcript would be translated on ribosomes associated with the endoplasmic reticulum. We wished to preserve this interaction, since it provides a critical structure that may influence stability. Our results indicated that sequences within bases 2246 and 2347 mediate the rapid decay of this message in the quiescent fibroblasts. This analysis also suggested that the same region controlled the TNF-induced accumulation of the GLUT1 mRNA.
RNA gel mobility shift assays utilizing the full-length 3Ј-UTR and the various truncated forms as binding probes have detailed the occupancy of the GLUT1 3Ј-UTR by RNA-binding proteins. This analysis identified two RNA-binding proteins that bound to the stability-determining region of the GLUT1 3Ј-UTR. The RNA binding activity of these proteins increased coordinately with the stabilization of the GLUT1 message following TNF stimulation.

EXPERIMENTAL PROCEDURES
Reagents-Recombinant human TNF-␣ was provided by Biogen (Cambridge, MA). The specific activity was 9.6 ϫ 10 6 units/mg of protein, based on a cytotoxicity assay using L929 cells. The endotoxin contamination was 0.12 ng/mg of protein, based on a Limulus amebocyte lysate assay (Sigma). Dulbecco's modified Eagle's medium was purchased from Hazelton Research Products (Lenexa, KS). Fetal calf serum was purchased from Hyclone (Logan, UT). All radiolabeled compounds were purchased from DuPont NEN. Hybond-N nylon filter paper was purchased from Amersham Corp. All restriction enzymes, DNA polymerases, and molecular biology enzymes were purchased from Life Technologies, Inc. Actinomycin D was purchased from Calbiochem. Protogel premixed acrylamide solution was purchased from National Diagnostics (Atlanta, GA).
3T3-L1 Fibroblast Cell Culture-Murine 3T3-L1 fibroblasts used in these studies were the gift of the late Dr. Ora Rosen (Memorial Sloan-Kettering Cancer Center (New York, NY)). These cells were plated and grown to confluence in Dulbecco's modified Eagle's medium supplemented with 10% calf serum as described previously (4). Four days postconfluence, the medium was replaced with Dulbecco's modified Eagle's medium supplemented with 0.5% essentially fatty acid-free bovine serum albumin. Cells were characterized as quiescent 24 h after this final medium change when the aphidicolin-sensitive component of DNA synthesis was at a minimum. All experiments were initiated within 24 h after reaching quiescence. TNF was dissolved in Krebs-Ringer-Hepes buffer containing 0.1% bovine serum albumin, and a 1:100 dilution was added to the cell culture medium, producing a final concentration of 5.0 nM. The duration of all treatments was 10 h. cDNAs and Plasmids-The GLUT1 cDNA used in these studies is the 3T3-L1 clone (accession number M23384) as reported by Kaestner et al. (6). All position numbering is derived from that report. The ␤-actin probe was donated by Dr. Don W. Cleveland (Johns Hopkins University, Baltimore, MD). The pBC12BI expression vector (7) was purchased from Dr. Bryan Cullen (Duke University, Durham, NC).
Transfections-Transfections were performed using calcium phosphate DNA co-precipatation method as described by Sambrook et al. (8). The pBC12BI expression vector and derivatives produced for these studies do not contain a selectable marker and required co-transfection with the vector pSV 2 neo, which conferred neomycin resistance. All studies reported represent the results of two independent stable transfections.
Cytosolic Extract Preparation and Polysome Isolation-The S10, S100, and polysomal protein fractionation procedure was performed as described by Brewer and Ross (9). Cells were harvested by washing with ice-cold phosphate-buffered saline (two times), scraping in 2 ml of phosphate-buffered saline/plate, combining the scraped cells, and pelleting them by centrifugation at 500 ϫ g for 10 min at 4°C. The cell pellet was resuspended in 3.5 ml of buffer A (10 mM Tris-HCl, pH 7.6, 1 mM potassium acetate, 1.5 mM magnesium acetate, 2 mM dithiothreitol) by gentle pipetting. Cells were lysed by applying 20 strokes with a Teflon pestle homogenizer, and the crude lysate was spun at 10,000 ϫ g for 10 min at 4°C to remove nuclei. Small amounts of the supernatant from this spin were aliquoted and designated as the S10 fraction. The remaining supernatant was removed and layered onto a 1.5-ml cushion of buffer B (30% (w/v) sucrose solution in buffer A) in a Beckman SW55 tube (Palo Alto, CA). These samples were centrifuged at 130,000 ϫ g for 3 h at 4°C. The high speed supernatant from this spin was designated S100 and was carefully removed from the top of the sucrose cushion and immediately frozen at Ϫ80 degrees. The remaining sucrose pad was removed by aspiration; the pellet containing the polysomes was washed twice with buffer A and harvested by repeated pipetting with 500 l of buffer A. Once in solution, the polysomes were transferred to a microcentrifuge tube and immediately frozen. The protein yields were determined by BCA protein analysis (Pierce), and the polysome yields were quantified by measuring absorbances at 260 and 280 nM.
Plasmid Construction-An 876-base SacI-EcoRI fragment (bases 1668 -2544 relative to the transcription initiation site) of the GLUT1 cDNA was ligated into identical sites within the vector pBluescript (Stratagene) to generate the plasmid pGT1UTR. Site-directed mutagenesis was performed on this plasmid, according to the method of Kunkel (10), to introduce a SpeI restriction enzyme site at base 1902. This new plasmid was named pGT1:SDM1902 and was used as the starting material for producing truncation deletion mutants using exonuclease III digestion as described below. The plasmids that resulted from the exonuclease III digestion procedure were assigned names according to the portion of the GLUT1 3Ј-UTR that remained (i.e. TD2088 contained the segment between base 2088 and the 3Ј terminus). These plasmids were named pTD1902, pTD2088, pTD2242, and pTD2347. These plasmids as well as the pGT1UTR construct were still contained within the original pBluescript vector. This allowed for the production of radiolabeled transcripts derived from these constructs that were used in the RNA binding studies.
The expression vectors used in the analysis of chimeric transcript half-lives were derived from the vector pBC12BI. This vector contains the gene for preproinsulin under the control of a Rous sarcoma virus long terminal repeat promoter. The plasmids used for this study were constructed through ligation into the pBC12BI using a SmaI site at base 1207 and a BstEII site at base 1215. These sites are located 50 bases after the translation stop sequence such that ligation into this position will replace the 3Ј-UTR of the preproinsulin gene, provided that the inserted DNA contains a polyadenylation sequence. Polymerase chain reaction with the primers (SmaI, 5Ј-ACAACCCGGGACTC-CCAAGT-3Ј; BstEII, 5Ј-GTGGTCACCCTCCAGCCATTTA-3Ј) was used to incorporate SmaI and BstEII sites into the 5Ј and 3Ј termini, respectively, of products amplified from the vectors pGT1UTR, pTD1902, pTD2088, pTD2242, and pTD2347. The polymerase chain reaction products were purified, digested with SmaI and BstEII, and ligated into the pBC12BI vector to generate the plasmids pBC:GT1UTR, pBC: TD1902, pBC:TD2088, pBC:TD2242, and pBC:TD2347. These vectors produced transcripts that contained the 5Ј-UTR and coding region of the preproinsulin gene with either the full-length or truncated forms of the GLUT1 3Ј-UTR.
Exonuclease III Truncation Deletions-The plasmids pTD1902, pTD2088, pTD2242, and pTD2347 were constructed through the unidirectional digestion of the plasmid pGT1:SDM1902. This procedure was modified from the method previously described by Henikoff (11). The plasmid pGT1:SMD1902 was digested with SacI and SpeI, purified, and resuspended in exonuclease III digestion buffer (60 mM Tris-HCl, pH 8.0, 0.6 mM MgCl 2 ). Unidirectional digestion occurred in a 5Ј 3 3Ј direction from the SpeI site but did not proceed in the 3Ј 3 5Ј direction past the SacI site of the GLUT1 3Ј-UTR. Exonuclease III digestion was performed with 500 units of exonuclease III at 22°C. To stop the reaction, aliquots were removed at 2-min intervals, placed into S1 nuclease buffer (40 mM potassium acetate, 300 mM NaCl, 1.5 mM ZnSO 4 , 5% (v/v) glycerol), and digested with 2 units of S1 nuclease/ aliquot for 30 min at room temperature. The extent of truncation was assessed by agarose gel electrophoresis of a portion of each aliqout. Those samples that visually represented significant deletions (ϳ100 -200 bases) were ligated and used to transform competent bacteria. The extent of truncation was determined through dideoxy sequencing of miniprep DNA.
Restriction Digest Deletion Mutant PBC:⌬2246 -2347-The GLUT1 3ЈUTR contained in the PBC:GLUT1 construct was cleaved at position 2246 with MmeI and at 2347 with BbvII; the ends were first blunted with T4 DNA polymerase and then ligated with T7 DNA ligase. The resulting deletion was confirmed by restriction mapping and dideoxysequencing of the miniprep DNA.
Preparation of a Construct Containing Bases 2246 -2347 of the GLUT1 3Ј-UTR under the Control of the T7 Promoter-The region was isolated using polymerase chain reaction. The forward primer for the reaction was 5Ј-GCCTAATACGACTCACTATAGGGGACCAGTTGGA-AGCACTGG-3Ј, while the 3Ј-primer was 3Ј-AGTTAAGTTCTCAGCC-TC-5Ј. Bases 4 -21 of the forward primer comprise the T7 RNA polymerase promoter, which was contained in the amplified fragments, permitting production of sense RNA transcripts following in vitro transcription. Bases 23-41 (boldface type) correspond to positions 2242-2257 of the 3Ј-UTR. The 3Ј-primer corresponds to positions 2352-2370 of the 3Ј-UTR.
RNA Isolation and Northern Blot RNA Analysis-RNA isolation, Northern analysis, and mRNA half-life determinations were performed as described previously (12) using either radiolabeled cDNA probes or cRNA probes (described below). All cDNA probes were labeled with [␣-32 P]-dATP (3000 Ci/mmol) by the random priming method described by Feinberg and Vogelstein (13) and used at an activity of 1 ϫ 10 7 cpm/ml. cRNA probes were prepared as described under "In Vitro Transcription Reactions" (see below). Hybridization with cRNA probes was performed at 65°C, and the probes were used at a concentration of 1 ϫ 10 6 cpm/ml of hybridization solution. The hybridization solutions and washes were the same as described previously (14).
Determination of Chimeric mRNA Half-lives-The half-life data presented in the text represent the mean Ϯ S.E. from a total of four independent experiments performed on each of two cell lines derived from two independent transfections. Quantitation was performed by densitometric analysis using ImageQuant software on a Molecular Dynamics PhosphorImager. Half-lives were determined through linear regression analysis for the best fit line on logarithmic plots of control and TNF-treated data points, as described previously (5).
In Vitro Transcription Reactions-In vitro transcription was performed according to the manufacturer's procedures for the MAXI-script transcription kit (Ambion Inc., Houston, TX). These probes are radiolabeled through the inclusion of [␣-32 P]UTP (800 Ci/mmol) in the reaction. The resulting transcripts had a specific activity of approximately 1 ϫ 10 8 cpm/g as determined by trichloroacetic acid precipitation.
RNA Gel Mobility Shift Assay-RNA band shift analysis was performed as described by Nagy and Rigby (15). Radiolabeled RNA (10 5 cpm, ϳ1.0 ng) was prepared as described above and incubated with 30 g of S10, S100, or polysomal protein fractions. The incubations were performed at 37°C for 30 min in a final buffer concentration of 10 mM HEPES, pH 7.9; 15 mM KCl; 3 mM MgCl 2 ; 0.2 mM dithiothreitol; 1 g/ml yeast tRNA; 25 g/ml poly(I) and 10% glycerol. The reactions were then placed on ice and cross-linked using a Stratalinker 1800 (Stratagene) for 5 min at 3000 microwatts/cm 2 . This was followed by RNase digestion (10 units of RNase T 1 , 10 g of RNase A) for 30 min at room temperature. The cross-linked complexes were analyzed by 12% SDS-polyacrylamide gel electrophoresis, dried, and subjected to autoradiography.

GLUT1 3Ј-UTR Contains a Destabilizing and a TNF-respon-
sive Element-3T3-L1 preadipocytes were stably transfected with an expression vector containing the preproinsulin gene or a construct where the 3Ј-UTR of the preproinsulin gene has been replaced with the GLUT1 3Ј-UTR. All transfectants required cotransfection with the vector pSV 2 -neo that contained a neomycin resistance gene to allow for antibiotic selection. Following transfection, G418-resistant colonies were pooled to ensure a heterogeneity of integration sites, and permanent transfected cell lines were established. These transfectants along with a transfection control, which contained only the pSV 2 neomycin resistance vector, were analyzed for GLUT1 mRNA turnover under control and TNF-treated (5 nM for 10 h) conditions; a typical analysis is shown in Fig. 1. The following summary half-life data represent the mean Ϯ S.E. from a total of four independent experiments performed on each of two cell lines derived from two independent transfections. The results of the study demonstrate that the cells transfected with parental preproinsulin vector produce preproinsulin mRNA that is extremely stable under control conditions and does not respond to TNF (t1 ⁄2 values: control, 675 Ϯ 20 min; TNF-treated, 682 Ϯ 20 min). Replacing the preproinsulin 3Ј-UTR with the GLUT1 3Ј-UTR results in a transcript with a much shorter half-life (control t1 ⁄2 ϭ 51 Ϯ 9 min) that is stabilized in response to TNF treatment (TNF t1 ⁄2 ϭ 181 Ϯ 18 min). The turnover rate for this chimeric transcript resembled those of the GLUT1 mRNA, suggesting that the GLUT1 3Ј-UTR conferred mRNA decay characteristics consistent with the GLUT1 message to the normally stable preproinsulin reporter gene. Control transfectants that contained only the pSV 2 -neo vector were also analyzed for GLUT1 mRNA turnover. The half-life of GLUT1 within these cells was calculated to be 39 Ϯ 8 min under control conditions and 168 Ϯ 20 min following TNF treatment, indicating that the transfection procedure does not alter the behavior of the GLUT1 transcript.
Truncation Analysis of the GLUT1 3Ј-UTR-The initial findings suggested that the GLUT1 3Ј-UTR contained elements that could regulate the turnover of the normally stable preproinsulin transcript. Exonuclease III digestion was used to pro-duce a series of constructs that were truncated from the 5Ј-end of the 3Ј-UTR (Fig. 2A). These constructs were transfected into 3T3-L1 preadipocytes, selected for resistance to G418, and pooled as described above. The expression of these constructs following transfection was verified by Northern blot analysis using a cRNA (riboprobe) generated against the terminal 250 bases of the GLUT1 message (Fig. 2B).
To investigate whether the removal of any portion of the GLUT1 3Ј-UTR altered the behavior of the pBC:GLUT1 chimera, the transfectants containing the truncation deletions were analyzed for the decay of the chimeric transcripts under control and TNF-treated conditions (Fig. 3). Under control conditions, transcripts encoded by PBC:TD1902, PBC:TD2088, and PBC:TD2242 possessed decay rates of 56 Ϯ 11, 51 Ϯ 10; and 52 Ϯ 8 min, respectively. Treatment of the cells with TNF resulted in a significant (p Ͻ 0.05, Student's t test) stabilization of the chimeric mRNA with values determined to be 176 Ϯ 20 min for PBC:TD1902; 184 Ϯ 9 min for PBC:TD2088; and 180 Ϯ 9 min for PBC:TD2242. These values were comparable with that observed for the half-life for endogenous GLUT1 in control and TNF-treated cells. However, the construct PBC:TD2347 was remarkably stable in untreated quiescent cells with a half-life of 477 Ϯ 30 min, a value that approximated the t1 ⁄2 of the preproinsulin mRNA (675 Ϯ 20 min) encoded by the parental PBC12BI vector more than the PBC:GLUT1 chimera FIG. 1. GLUT1 3-UTR confers lability and TNF-responsive stabilization to the preproinsulin reporter construct. A, autoradiogram of mRNA half-lives for endogenous GLUT1, the PBC:GLUT1 construct, and the PBC12BI vector alone. Total RNA was isolated from control and TNF-treated (5 nM for 10 h) 3T3-L1 cells at 0, 1, 3, and 5 h following treatment with actinomycin D (6 g/ml). Twenty micrograms of total RNA from each time point were analyzed by Northern blot with radiolabeled cDNA (a random primed probe made from a template corresponding to the full-length GLUT1 3Ј-UTR for GLUT1 and PBC: GLUT1) or cRNA (prepared from a 292-base SmaI/EcoRI fragment generated from the coding sequence of the preproinsulin clone for PBC12BI) probes as described under "Experimental Procedures." Quantification and half-life determinations were performed as described under "Experimental Procedures." (51 Ϯ 8 min). TNF treatment yielded a half-life of 497 Ϯ 35 min for the PBC:TD2347 chimera, a value that was not statistically different from the value obtained under control conditions ( p Ͼ 0.05, Student's t test). These data indicate that the ability to destabilize the PBC:GLUT1 transcript resides within the region defined by bases 2242 and 2347 of the GLUT1 3Ј-UTR. Transcripts that contained this segment of the 3Ј-UTR remained unstable, while those that do not have this segment are inherently stable. The dramatic increase in stability produced by the removal of this region (2242-2347) complicated the identification of a TNF-responsive element. Prior to the removal of this portion of the GLUT1 3Ј-UTR, all of the chimeras exhibited significant stabilization in response to TNF (Fig. 3 and see above). This suggests that the TNF-responsive element resides somewhere between base 2242 and the 3Ј terminus.

Stabilization of the Message and Loss of TNF Response upon Deletion of the Region between 2246 and 2347 of the PBC:
GLUT1 Chimera-The findings described above were consistent with the hypothesis that the region between 2242 and 2347 contained an element essential for the determination of the stability of this message. In addition, the region potentially contained a TNF-responsive element. Restriction sites at positions 2246 and 2347 of the construct permitted removal of the region to examine these possibilities (detailed under "Experimental Procedures"). The deletion mutant PBC:GLUT1 ⌬2246 -2347 was prepared (Scheme I) and stably transfected into the 3T3-L1 cells as described above.
Expression and TNF regulation of the mutant chimeric RNA were then examined. As shown in Table I, deletion of this region of the GLUT1 3Ј-UTR resulted in a significant stabilization of the message in control cells (2.4-fold over the PBC: GLUT1 mRNA, p Ͻ 0.05). These data were consistent with the deleted region containing an element essential for the rapid turnover of the mRNA. Moreover, the half-life of the message did not change when the cells were exposed to TNF, supporting a role for the region in mediating the stabilizing effects of this cytokine.
RNA Gel Mobility Shift Analysis with S10, S100, and Polysomal Protein Fractions-The occupancy of the GLUT1 3Ј-UTR by RNA-binding proteins was assessed through gel mobility shift analysis. Postnuclear (S10), cytosolic (S100), and polysomal protein extracts were prepared from control and TNFtreated quiescent 3T3-L1 cells. These proteins were incubated with a radiolabeled transcript corresponding to the full-length GLUT1 3Ј-UTR. A total of five RNA-protein complexes from the S10, S100, and polysomal protein extracts with molecular masses of approximately 67, 45, 40, 37, and 26 kDa were detected by the label transfer (Fig. 4). The postnuclear S10 fraction displayed all five bands, and the RNA binding activities at 67, 40, and 37 kDa appeared to increase with TNF treatment. The 67-, 46-, and 37-kDa proteins partitioned into the postmitochondrial S100 fraction, but no differences in their RNA binding activities between control and TNF-treated extracts were observed. The TNF-treated polysomal protein fraction contained all five RNA binding activities including the 40and 37-kDa bands that were not observed within the polysomal preparations from control cells.
Comparison of Gel Mobility Shifts with the Full-length 3Ј-UTR and the TD2347 Truncation Deletion-A comparison study was carried out with the S10, S100, and polysomal fractions using radiolabeled transcripts prepared from the fulllength GLUT1 3Ј-UTR and the various truncation deletions. The primary goal of this study was to determine if any of the RNA-protein interactions observed in Fig. 4 were specific for the destabilizing region located between bases 2242 and 2347 of the GLUT1 transcript. The sequential removal of regions within the GLUT1 3Ј-UTR did not alter the binding pattern within any of the control extracts (Fig. 5, A and C). However, significant alterations in the RNA binding activity were found within the postnuclear (S10) and polysomal protein fractions obtained from TNF-treated cells. Removing the portion of the GLUT1 3Ј-UTR located between bases 2242 and 2347, the stability-determining element, reduced the label transfer to both the 37-and 40-kDa binding activities found within these extracts (Fig. 5, A and C). The TNF-treated S100 fraction demonstrated no alteration in binding activity when either transcript was used in the assay.
Protein Binding to the Region Located between Bases 2242 and 2347 of the GLUT1 3Ј-UTR-As described above, deletion of the region between 2246 and 2347 of the GLUT1 3Ј-UTR resulted in stabilization of the message and loss of the TNF response. Moreover, it has been identified by deletion analysis (Figs. 4 and 5) as the site of TNF-inducible protein binding. Thus, it was of interest to determine if this region upon isolation remained a ligand for the TNF-inducible protein. A riboprobe corresponding to the region was prepared as described under "Experimental Procedures," and RNA gel mobility shift  The data for endogenous GLUT1 and PBC:GLUT1⌬2246 -2347 were obtained from RNA extracted from the same cell lines at the same time and visualized on the same Northern blots. The data for PBC: GLUT1 mRNA was obtained from a different stably transfected cell line.
assays were performed. The data displayed in Fig. 6 indicate that TNF treatment resulted in the up-regulation of binding activity in the S10/polysome-containing fraction of proteins of 37 and 40 kDa as well as a doublet at 67 kDa. These data, similar to those obtained with the 3Ј-UTR truncations described in Figs. 4 and 5A, support the identification of the region between 2246 and 2347 as the site for TNF-inducible protein binding. Interestingly, use of this "binding site-specific riboprobe" demonstrated decreased binding in both the 67-and 37-kDa regions of the gel when the TNF-treated S100 fraction was used as a source of ligand. This was not evident in the data presented in Figs. 4 and 5, where larger portions of the 3Ј-UTR were used as probes and may reflect an alteration in affinity of the proteins for the smaller probe after TNF treatment or the ability to detect the partitioning of the binding proteins from soluble to particulate fractions of the cell. These possibilities are currently under investigation. However, the ability to use the specific binding site probe to demonstrate TNF-inducible binding provides a convenient assay for the purification and characterization of the proteins involved.

DISCUSSION
Regulation of transcript stability plays a major role in controlling the expression of the HepG2/erythrocyte/brain glucose transporter, GLUT1 (3). GLUT1 is the predominant glucose transporter within several biologically important tissues including blood-tissue barriers such as placenta and the bloodbrain barrier (16,17). Several laboratories have reported that conditions such as glucose deprivation (18,19), experimental diabetes (20), and inhibition of oxidative phosphorylation (21) as well as cytokine and hormone stimulation (4,(22)(23)(24) alter the stability of the GLUT1 transcript. The mechanisms that produce this effect, including the role of cis-acting sequence elements and trans-acting factors, have remained unclear.
The current study provides evidence that the GLUT1 3Ј-UTR regulates the decay of the GLUT1 message. Insertion of the GLUT1 3Ј-UTR into the otherwise long lived preproinsulin transcript is sufficient to destabilize the resulting mRNA. Additionally, this chimeric transcript can also be stabilized by treatment with TNF, a characteristic of the endogenous GLUT1 mRNA. The 3Ј-UTR of a number of short-lived mRNAs such as oncogenes and growth factors has also been shown to promote the degradation of heterologous RNAs (25,26). The 3Ј-UTRs of these early response genes contain a destabilizing motif known as an adenylate-uridylate-rich element that mediates their rapid decay. The destabilizing element within the GLUT1 3Ј-UTR was mapped to a portion located between bases 2242 and 2347. Importantly, when this region was deleted, the message was markedly stabilized (Table I). The -fold stabilization was not as large as we would have predicted based on the truncation mutations. As opposed to those experiments, the effect was observed in the context of the nearly intact 3Ј-UTR (Scheme I), suggesting that other regions may facilitate the stabilization of the message on their own or interact with the 2242-2347 region to affect stability. This region is predominantly GC-rich (ϳ60%) and was not homologous to either a GC-rich stability element reported for the transforming growth factor-␤1 gene (27) or to a TNF-responsive stability element identified within the surfactant-B protein mRNA (28).
A comparison of this 105-base sequence against sequences within GenBank TM revealed conservation of this element among GLUT1 isoforms across species lines (human, mouse, and chicken) conferring a degree of evolutionary significance to the reported observations. In addition to the homology found within 3Ј-UTRs, significant homology was also found to a portion of the GLUT1 coding region for both mice and humans beginning at base 290. The redundant expression of destabilizing sequences has been observed within the c-myc and c-fos protooncogene mRNAs. These mRNAs have been demonstrated to contain destabilizing sequences within their open reading frames in addition to elements within their 3Ј-UTRs (29 -31). These observations have led to the suggestion that multiple destabilizing elements may be employed in order to prevent the overexpression of a protein within normal cells. It is also possible that this sequence may affect other aspects of the posttranslational processing of the GLUT1 message such as translational initiation (32). This said, we did not observe any TNFinducible binding to probes prepared from this coding region element 2 and have not as yet investigated whether this element actually functions as a stability-determining sequence.
Our chimeric transcript analysis did not directly identify sequences that mediate the stabilizing effect of TNF that was observed with several of the chimeric transcripts as well as the endogenous GLUT1 message. However, when the region from 2246 to 2347 was deleted, the ability of TNF to stabilize the message was lost. These data suggest that the same region responsible for the inherent instability of the GLUT1 transcript also functions to stabilize this message in response to TNF.
The production of hybrid transcripts demonstrated that specific sequences within the GLUT1 3Ј-UTR regulate the decay of this message. To gain some insight into how these sequences may function to control GLUT1 mRNA turnover, we searched for trans-acting factors, such as RNA-binding proteins, using gel mobility shift analysis. We observed several binding activities within extracts prepared from control and TNF-treated cells (Fig. 4). Two of these proteins, with apparent molecular masses of 37 and 40 kDa, demonstrated enhanced RNA-binding activity following TNF treatment. These proteins were contained within the polysomal protein fraction and were also detected within the postnuclear S10 fraction (Figs. 4 and 5). The involvement of these RNA-binding activities in regulating the turnover of the hybrid constructs was investigated by comparing the RNA binding patterns obtained with the full-length FIG. 4. RNA binding analysis from S10, S100, and polysomal protein fractions. RNA gel mobility shift analysis was performed with postnuclear (S10), postmitochondria (S100), and polysomal protein fractions prepared from control (C) and TNF-treated (10 h at 5 nM) (T) 3T3-L1 cells. The 3Ј-untranslated portion of the GLUT1 message was radiolabeled by in vitro transcription and incubated with either 40 (S10, S100) or 5 g (polysomal) of protein extracts. The reactions were UV cross-linked, digested with a combination of RNases A and T 1 , and electrophoresed on 12% SDS-polyacrylamide gel electrophoresis. Molecular weights were determined through comparison with molecular weight standards (not shown) using ImageQuant software with a Molecular Dynamics PhosphorImager system. The data are representative of an experiment performed three times with similar results. GLUT1 3Ј-UTR with the various truncated versions used for the in vivo analysis ( Fig. 2A). Two RNA binding proteins, one of 37 kDa and the other of 40 kDa, previously detected within the TNF-treated extracts using the full-length GLUT1 3Ј-UTR, could not be detected when the TD2347 probe was used (Fig. 5). We demonstrated that these two proteins bound all forms of the GLUT1 3Ј-UTR except the one in which the stability-determining region had been removed (TD2347). This suggests that the RNA binding activity of these proteins is specific for the stability-determining region (2242-2347) and can be activated by TNF. This finding further supports the hypothesis that the destabilizing element located between bases 2242 and 2347 functions to stabilize the GLUT1 message through TNF-controlled RNA-protein interactions. Interestingly, the data displayed in Fig. 5B demonstrate that a 37-kDa protein present in the S100 (cytosolic fraction) bound all forms of the 3Ј-UTR and did not appear to be regulated by TNF. One possible explanation of this observation is that the regulated binding activities measured in the S10 and polysomal fractions are not derived from the same protein found in the S100. The S100 37-kDa protein that binds to the 2347 mutant may be a nonspecific RNA binding activity, many of which coincidentally are in this molecular mass range. This hypothesis is reinforced by our recent study of dehydrogenase binding to the GLUT1 3Ј-UTR, which indicated that both lactate and glyceraldehyde 3-phosphate dehydrogenases bind to this region of the UTR (33). Additionally, both enzymes have subunit molecular sizes in the appropriate range and are particularly abundant in the cytosolic fraction of these cells. A second explanation addresses the issue that only the S10 and polysomal fractions contain the translational machinery and that specific interactions mediated by the presence of these components may be necessary to observe the regulation. These issues are currently under investigation.
Our work has documented that in the GLUT1 3Ј-UTR, a GC-rich region (bases 2242-2347) contributes significantly to the post-transcriptional regulation of this gene. The interaction between this region and two RNA-binding proteins following TNF treatment resembles a similar mechanism through which the expression of the transferrin receptor mRNA is regulated FIG. 5. Mapping of the RNA binding activity to the stabilitydetermining element. RNA gel mobility shift analysis was performed as described in the legend to Fig. 4 with radiolabeled ligands corresponding to the various truncated forms of the GLUT1 3Ј-UTR described in the legend to Fig. 2A. Analysis was performed with S10 (A), S100 (B), and polysomal protein extracts (C) from control and TNFtreated (5 nM for 10 h) 3T3-L1 cells. The molecular weights for the FIG. 6. Protein binding to the region located between bases 2242 and 2347 of the GLUT1 3-RNA gel mobility shift analysis was performed as described in the legend to Fig. 4 with a radiolabeled ligand corresponding to the region between bases 2242 and 2347 of the GLUT1 3-UTR. Analysis was performed with S10/ polysome-containing extracts (lanes 1 and 2) and S100 (soluble) extracts (lanes 3 and 4) prepared from control and TNF-treated (5 nM for 10 h) cells, respectively. The molecular weights for the various binding activities are indicated and were determined as described previously. The data are representative of an experiment performed twice with different extract preparations with similar results.
various binding activities are indicated and were determined as described in the legend to Fig. 4. The data are representative of an experiment performed three times with similar results. post-transcriptionally (34). A destabilizing element within the transferrin receptor 3Ј-UTR serves as a nuclease recognition site that normally allows for the rapid decay of this message (35). In response to changes in cellular iron concentration, the iron-responsive binding protein binds to this destabilizing sequence, blocks the activity of the nuclease, and prevents the decay of the transferrin receptor mRNA (36). The similarities between the two systems have led us to propose the model shown in Fig. 7, where in control cells, the destabilizing region is unprotected and the message exhibits a rapid turnover. Treatment of the cells with TNF leads to occupation of the region by a complex composed of at least two proteins (37 and 40 kDa), protection against nucleolytic attack, and stabilization of the mRNA. The model presented in Fig. 7 represents the simplest of all models and a good place to start. However, the data presented in this study do not rule out more complex models where the proteins may bind at the defined site but interact with accessory proteins and/or other regions of the message. Further definition of the system requires the purification and cloning of these binding activities, a task that is currently under way. However, the results presented in the current study represent a unique contribution to the understanding of glucose transporter gene expression. Additionally, these findings support previous studies suggesting that TNF works through inducible RNA-binding proteins to influence gene expression at the post-transcriptional level (5,32), a result that has important implications for the understanding of TNF involvement in numerous biological systems.