Dual control of glut1 glucose transporter gene expression by hypoxia and by inhibition of oxidative phosphorylation.

glut1 gene expression and glucose transport are stimulated in a variety of cells and tissues in response to hypoxia. glut1 is also up-regulated by inhibitors of oxidative phosphorylation (such as azide) in the presence of oxygen. Here, we test the hypothesis that hypoxia stimulates glut1 gene expression independent of its inhibitory effect on oxidative phosphorylation. We examined the effect of cobalt chloride, a known stimulator of genes responsive to reduced oxygen concentration per se, on GLUT1 expression under normoxic conditions and compared the results with the response to azide. Exposure of a rat liver cell line (Clone 9) to 250 μM cobalt chloride increases GLUT1 mRNA content, which becomes evident at 2 h, reaches a maximal value of ∼12-fold at 8 h, and remains elevated at ∼8-fold at 24 h. GLUT1 mRNA was the only GLUT isoform expressed in control cells and in cells exposed to cobalt chloride or azide. The induction of GLUT1 mRNA by cobalt chloride is associated with a ∼10-fold stimulation of cytochalasin B-inhibitable 3-O-methyl-D-glucose transport at 24 h. In contrast to the rapid decrease in cell ATP levels and the stimulation of glucose transport in response to azide, cell ATP content and glucose transport remained unaltered during the initial 1-h period of exposure to cobalt chloride. The effect of cobalt chloride on GLUT1 mRNA content is mimicked by Ni(II) or Mn(II) but not by Fe(II). Employing actinomycin D, we found no increase in the ∼1.5-h half-life of GLUT1 mRNA in cobalt chloride-treated cells, suggesting that the effect of cobalt chloride on GLUT1 mRNA content is largely mediated at the transcriptional level; in contrast, GLUT1 mRNA half-life increased to >8 h in azide-treated cells. In transient transfections we found that ∼6 kilobase pairs (kbp) of 5′-flanking region of the rat glut1 promoter confers both cobalt chloride- and azide-inducibility to a reporter gene. Deletion of ∼2,500 base pairs (bp) from the 5′ end of the ∼6-kbp DNA fragment results in a reduction of the response to cobalt chloride and a complete loss of the response to azide. A 666-bp DNA segment located ∼6.0 kbp upstream of the transcription start site was found to be necessary for the increase in reporter gene expression in response to azide, whereas a 480-bp segment located at approximately −3.5 kbp mediated the response to cobalt chloride. The 480-bp segment is highly homologous to the previously reported mouse glut1 enhancer and contains several potential regulatory elements, including a hypoxia-inducible element; an additional hypoxia-inducible element is present in the 666-bp segment. Our results suggest that glut1 gene expression is regulated in a dual fashion by hypoxia per se and in response to inhibition of oxidative phosphorylation.

sponse to cobalt chloride. The 480-bp segment is highly homologous to the previously reported mouse glut1 enhancer and contains several potential regulatory elements, including a hypoxia-inducible element; an additional hypoxia-inducible element is present in the 666-bp segment. Our results suggest that glut1 gene expression is regulated in a dual fashion by hypoxia per se and in response to inhibition of oxidative phosphorylation.
The expression of a number of genes in mammalian cells is regulated by a decrease in the concentration of oxygen in the extracellular environment. For example, the erythropoietin (EPO 1 ) gene and the gene encoding vascular endothelial growth factor are each stimulated in response to hypoxia, effectively resulting in the activation of processes that enhance blood oxygen-carrying capacity (erythropoiesis) and oxygen delivery (neovascularization), respectively (1)(2)(3)(4)(5)(6)(7). Equally important and of fundamental significance to the maintenance of cellular homeostasis is the enhancement of glucose transport by hypoxia (8 -11), a process that is largely mediated by glut1, a member of the family of facilitative Na ϩ -independent transporters (12,13). glut1, which is expressed in a variety of cells and tissues, is thought to mediate a significant fraction of non-insulin-dependent transport of glucose in the organism (13,14), and its expression is augmented by exposure to serum and growth factors and following transformation (15)(16)(17)(18)(19). In addition, GLUT1 mRNA and transporter expression are rapidly and markedly up-regulated in response to hypoxia (9 -11). Interestingly, in recent studies employing Clone 9 cells (a nontransformed rat liver cell line), we have shown that GLUT1 expression is also markedly up-regulated by inhibitors of oxidative phosphorylation such as azide or cyanide (20 -24). These observations raise the question of whether the induction of GLUT1 mRNA in response to hypoxia results from the attendant inhibition of oxidative phosphorylation or whether reduced oxygen concentration per se, operating through distinct oxygensensing mechanisms, also stimulates glut1 gene expression.
Studies performed on the regulation of the EPO gene by oxygen have suggested that the response to hypoxia is mediated by oxygen-sensing mechanisms (1,(25)(26)(27). More specifically it has been shown that exposure to hypoxia leads to the induction of a hypoxia-inducible factor that, when bound to its cognate hypoxia-inducible DNA element (HIE) located in the 3Ј-flanking region of the EPO gene, stimulates EPO gene transcription (2, 26 -29). Although agents such as cobalt chloride employed in the presence of oxygen mimic the stimulatory effects of hypoxia on EPO gene expression (1, 5, 26 -28, 30, 31), chemical inhibitors of oxidative phosphorylation such as cyanide or rotenone do not enhance EPO expression (25,32). The induction of hypoxia-inducible factor binding activity by hypoxia and cobalt chloride in a variety of cell lines (33)(34)(35) and the recent identification of hypoxia-inducible factor binding sites in a number of other genes responsive to hypoxia (5,30,31,36) suggest that this same oxygen-sensing regulatory mechanism may also be involved in the regulation of the glut1 gene by hypoxia.
The present study is focused on the question of whether the induction of the glut1 gene by hypoxia is mediated only in response to the attendant inhibition of oxidative metabolism or whether distinct oxygen-sensing mechanisms also mediate the response. To differentiate between mechanisms mediating the response to lowered oxygen concentration per se from those mediating the response to the associated inhibition of oxidative phosphorylation, we examined the effect of cobalt chloride employed under normoxic conditions on glut1 gene expression. We found that the enhancement of GLUT1 expression by cobalt chloride occurred in the absence of any inhibition of oxidative phosphorylation and that the induction was mediated by qualitatively different mechanisms and pathways than those mediating the stimulation of GLUT1 mRNA expression in response to azide. A preliminary report of some of these findings has been presented (37 Quickhyb was obtained from Stratagene (La Jolla, CA). pGL2-Basic plasmid, pRL-TK plasmid, the Dual-luciferase reporter assay system, and the Profection transfection kit were purchased from Promega (Madison, WI). Luciferase assay kits and Monolight 2010 luminometer were obtained from Analytical Luminescence Laboratory (San Diego, CA).
Cell Culture-Clone 9 cells were passaged in DMEM with 10% calf serum at 37°C with 9% CO 2 and reached confluence after ϳ4 days. Cells were used between passages 27 and 55. C 2 C 12 and 3T3-L1 cells were grown in DMEM containing 10% fetal calf serum and were used immediately upon reaching confluence prior to differentiation. 24 h prior to initiation of all experiments, the medium was changed to serum-free DMEM.
RNA Isolation and Northern Blots-Cytoplasmic RNA was isolated as described previously employing a Nonidet P-40-containing buffer (21,38). RNA samples were fractionated in 1% formaldehyde-agarose gels, and the resulting blots were probed with ϳ50 ϫ 10 6 cpm of full-length rat glut1 cDNA (39) that had been 32 P-labeled by the random priming method. To ensure equal RNA loading of the lanes and to control for completeness of RNA transfer, ethidium bromide staining of ribosomal 28S and 18S bands on the gels and on the nitrocellulose paper was monitored throughout. Measurement of [ 3 H]3-OMG Uptake-Triplicate confluent cells grown on 60-mm culture plates were employed as described previously (20,21). Cytochalasin B-inhibitable [ 3 H]3-OMG uptake was calculated as the difference in uptake in the absence and presence of cytochalasin B assayed in parallel.
DNA Constructs-phage EN11 (18) was replicated as described (38). The ϳ6-kbp BglII/XhoI DNA fragment representing the 5Ј-flanking region of the rat glut1 gene promoter spanning from ϳϪ6,000 to ϩ138 bp of the major transcription start site (18) was cloned into the promoterless Photinus (firefly) luciferase reporter vector pGL2-Basic (Promega) and used as a template to generate various deletions using mapped restriction enzyme sites (constructs A-J). To generate constructs K and L, the 480-bp PstI/SacI fragment (spanning ϳϪ3,500 to ϳϪ3,000 bp), to which flanking SalI sites had been added, was cloned into the SalI site immediately upstream of the minimal c-fos promoter vector expressing the luciferase reporter gene (40). The sequences of the 480-and 666-bp fragments were determined by the dideoxy method (38). pRL-TK (Promega) is a neutral constitutive plasmid containing the herpes simplex virus-thymidine kinase promoter expressing Renilla (sea pansy) luciferase.
Transient Transfection-Cells were transfected by the method of calcium phosphate-mediated DNA transfer using Profection (Promega) according to a modification of the manufacturer's protocol. For each construct, 10 g of test plasmid was transfected into each of four subconfluent (90%) 100-mm culture plates containing Clone 9 cells passaged the day before. Approximately 1 h after the addition of DNA, the plates were washed twice with Hanks' balanced salt solution and incubated for ϳ5-8 h in DMEM containing 10% calf serum. Cells were then recovered by trypsinization, pooled, and divided equally into 12 10-mm culture wells. After overnight incubation, the medium was changed to DMEM devoid of serum, and after ϳ6 -8 h, azide or cobalt chloride was added and the incubation was continued for 24 h. All experiments were performed in triplicate culture plates with parallel control plates receiving DMEM as diluent.
To compare the basal activity of the various promoter constructs, equimolar amounts of the test vectors were transfected into Clone 9 cells in parallel experiments. 10 g (per 100-mm plate) of the largest construct was transfected each time. Nonspecific plasmid DNA (pBluescript, Stratagene) was mixed with the smaller test vectors to adjust to a total of 10 g of DNA. To control for the efficiency of transfection, 10 g of a chloramphenicol acetyltransferase reporter construct containing an SV40 promoter/enhancer (pCAT-Control, Promega) was also cotransfected. Reporter activity was assayed 48 h posttransfection.
For hypoxic experiments, 6 100-mm culture plates were each cotransfected with 5 and 1 g of test vector (see Fig. 6A, construct A) and control vector (pRL-TK), respectively. ϳ8 h posttransfection, cells were pooled and split into 12 60-mm plates and incubated overnight with medium containing 10% calf serum. 8 h prior to exposure to hypoxia, the medium was changed to minimal essential medium devoid of serum. One set of 6 plates was incubated in a normoxic chamber (20% O 2 , 5% CO 2 ) and the other was exposed to hypoxia (nominal 3% O 2 , 5% CO 2 ) for 24 h.
Luciferase Assay-Luciferase assay was performed by a minor modification of the manufacturer's protocol (Analytical Luminescence Laboratory). In brief, cells were washed five times with magnesium-and calcium-free phosphate-buffered saline and lysed directly on each plate by the addition of 200 l of lysis buffer. The plates were then agitated gently on a gyrator shaker at 4°C for 20 min. Lysates were centrifuged at 14,000 ϫ g for 1 min, and a 100-l aliquot was used for measurement of luciferase activity in relative light units. Dual luciferase assay was performed according to the manufacturer's protocol (Promega).
Chloramphenicol Acetyltransferase Assay-Standard chloramphenicol acetyltransferase assays were performed using the thin-layer chromatography method. In brief, cells were washed five times with magnesium-and calcium-free phosphate-buffered saline and scraped in 1 ml of magnesium-and calcium-free phosphate-buffered saline. Onefourth of the cells were used for luciferase assay. The remainder of the cells were spun at 4°C and resuspended in 150 l of 0.25 M Tris-HCl, pH 8.0, and chloramphenicol acetyltransferase assays were performed as described (38). Chloramphenicol acetyltransferase activity was quantitated as the percentage of butyrylated products.
Statistical Analysis-Values are expressed as means Ϯ S.E. Student's unpaired two-tailed t test was used, and a p value of less than 0.05 was considered to be significant (41).
GenBank TM /EMBL-The nucleotide sequences of the 666-kbp (Bg-lII/PstI) segment and the 480-bp (PstI/SacI) segment of the rat glut1 5Ј-flanking region have been submitted to GenBank TM /EMBL.

Induction of GLUT1 mRNA by Hypoxia, Azide, and Cobalt
Chloride-In previous studies we have shown that inhibition of oxidative phosphorylation by azide in Clone 9 cells results in the induction of GLUT1 mRNA (21,22,24). To determine whether GLUT1 mRNA is induced by hypoxia as well as by azide, we incubated Clone 9 cells under normoxic conditions in the absence and presence of azide and under hypoxic conditions (6.5 and 1.2% O 2 ) in the absence of the inhibitor. Compared with control cells incubated under normoxic conditions, GLUT1 mRNA content in cells incubated at 6.5 and 1.2% O 2 increased 4-and 10-fold, respectively (Fig. 1A). In accordance with previous observations (21,23), exposure to 5 mM azide also led to an increase in GLUT1 mRNA content (Fig. 1A), suggesting that the response to hypoxia, in part or in its entirety, may be secondary to inhibition of oxidative phosphorylation.
To determine whether induction of GLUT1 mRNA by hypoxia can occur by mechanisms distinct from the attendant inhibition of oxidative phosphorylation, we exposed cells to 250 M cobalt chloride (Fig. 1B). It is evident that both cobalt chloride and azide enhanced GLUT1 mRNA content in the presence of oxygen. In dose-response studies we found that GLUT1 mRNA increased ϳ4-fold after 16 h of exposure to 50 M cobalt chloride, that the response was maximal (ϳ12-fold) following exposure to 250 M cobalt chloride, and that higher concentrations of cobalt chloride were associated with cell toxicity. Based on these results, cobalt chloride at a concentration of 250 M was employed in all subsequent studies. It should be noted that mRNAs encoding GLUT2, GLUT3, or GLUT4 isoforms were not detected in Clone 9 cells under basal conditions or after exposure to cobalt chloride or azide (data not shown).
We also measured the effect of azide and cobalt chloride on GLUT1 mRNA content of C 2 C 12 and 3T3-L1 cells (Table I); these cell lines were studied before their differentiation to myotubes and adipocytes, respectively. GLUT1 mRNA content was clearly increased in response to azide and cobalt chloride in both cell lines. We next monitored the time course of GLUT1 mRNA induction by 250 M cobalt chloride over a 24-h interval. We found that upon exposure to cobalt chloride, relative GLUT1 mRNA content remained unchanged at 30 min and 1 h, increased significantly by 3-Ϯ 0.5-fold at 2 h, reached a maximum value of 12-Ϯ 1.0-fold at 8 h, and remained elevated at 8-Ϯ 0.8-fold at 24 h (data not shown). Hence, although both azide and cobalt increased GLUT1 mRNA content after a "delay" period, the effect of cobalt chloride was somewhat more rapid (2 versus 3 h) and was of greater magnitude (12-versus 8-fold) than that of azide (21,24).
The effect of cobalt chloride on the induction of mRNAs of other genes responsive to hypoxia has been previously shown to be mimicked by Ni(II) and Mn(II), presumably by interacting with a putative heme-containing oxygen-sensing molecule (1). Similarly, an induction of GLUT1 mRNA content in Clone 9 cells was also observed in response to Ni(II) and Mn(II) but not to Fe(II) (all employed at 250 M) (data not shown). Moreover, induction of GLUT1 mRNA by cobalt chloride required de novo protein synthesis because the increase could be blocked by the simultaneous presence of anisomycin; it should be noted that anisomycin and other protein synthesis inhibitors added alone also induce GLUT1 mRNA expression (Fig. 2). 2 If the mechanism of induction of GLUT1 mRNA in response to azide is different from that of cobalt chloride, then it might be expected that the stimulatory effect of the two agents employed together would be additive. Cells were exposed to maximally effective concentrations of the agents, either singly or in combination (Fig. 3). Exposure to 250 M cobalt chloride and 7.5 mM azide resulted in 5.8-Ϯ 0.6-fold and 2.7-Ϯ 0.4-fold increases in GLUT1 mRNA content, respectively, whereas the presence of both agents increased GLUT1 mRNA content by 8.1-Ϯ 0.6-fold.
Effect of Cobalt Chloride on the Rate of Glucose Transport and Cell ATP Content-Incubation of cells in the presence of cobalt chloride for 24 h resulted in a ϳ10-fold stimulation in the rate of cytochalasin B-inhibitable 3-O-methyl D-glucose transport (Fig. 4). In marked contrast to the early stimulation in response to azide (21,22), there was no change in the rate of glucose transport following a 1-h exposure to cobalt chloride (Fig. 4). Moreover, exposure of Clone 9 cells to 5 mM azide or 0.5 mM cyanide is associated with a rapid Ͼ50% fall in cell ATP, with ATP levels increasing to normal levels by 1 h despite the continued presence of the inhibitors (20,22). Again in contrast to inhibitors of oxidative phosphorylation, cell ATP content 2 A. Behrooz and F. Ismail-Beigi, unpublished observations.  Transcriptional Enhancement of GLUT1 mRNA Expression in Response to Cobalt Chloride-To examine the potential role of GLUT1 mRNA stabilization in the induction of GLUT1 mRNA in response to cobalt chloride, we measured the half-life of GLUT1 mRNA in control and cobalt chloride-treated cells; azide-treated cells served as a positive control. Employing actinomycin D, we found that the ϳ1.5-h half-life of GLUT1 mRNA in control cells was unchanged in cobalt chloridetreated cells; in contrast, GLUT1 mRNA half-life was increased to Ͼ8 h in azide-treated cells, as has been described previously (23) (Fig. 5). These results are consistent with the possibility that the stimulatory effect of cobalt chloride on GLUT1 mRNA content is largely mediated at the transcriptional level.
To identify cis-regulatory elements involved in the transcriptional stimulation of the glut1 gene in response to cobalt chloride, we analyzed the rat glut1 promoter in a series of transient transfection assays. In addition, we extended this analysis to include our previously reported stimulation of glut1 gene transcription by azide (23). A series of constructs spanning ϳ6 kbp of genomic DNA from the 5Ј-flanking region of the rat glut1 gene was fused to a promoterless luciferase reporter vector (pGL2-Basic) and transfected into Clone 9 cells. To determine whether the construct containing ϳ6 kbp of 5Ј-flanking region is capable of being stimulated by hypoxia, we transiently transfected Clone 9 cells with the construct and exposed the cells to nominal 3% oxygen for 24 h. Results showed a 2.3-Ϯ 0.3-fold increase in luciferase activity of cells exposed to hypoxia compared with control cells incubated under normoxic conditions (p Ͻ 0.05; Table II). In parallel experiments, the same hypoxic stimulus led to a 4.0-Ϯ 0.6-fold increase in the level of endogenous GLUT1 mRNA.
We next determined the luciferase activity of transfected . At the indicated times thereafter cytoplasmic RNA was isolated from duplicate plates for analysis of GLUT1 mRNA content. In each experiment, GLUT1 mRNA content at various times was expressed as the fraction of GLUT1 mRNA content present at time 0. The experiment was repeated twice and the results averaged. Regression lines are drawn by the method of least squares (means Ϯ S.E.). The rate of disappearance of GLUT1 mRNA was not significantly different in control and cobalt chloride-treated cells, whereas the rate was significantly decreased in azide-treated cells. cells exposed for 24 h to diluent, 250 M cobalt chloride, or 5 mM azide, and the results are expressed relative to luciferase activity in diluent-treated transfected cells (Fig. 6A). Basal expression of constructs A, F, and H was 20-fold greater than that of construct J, which does not contain the glut1 TATA box and extends only 15 bp upstream of the glut1 major transcription start site. Luciferase expression of the construct containing ϳ6 kbp of the 5Ј-flanking region (construct A) and the construct containing a deletion of ϳ2 kbp within the 6-kbp region (construct B) was enhanced 6-and 3-fold by cobalt chloride and azide, respectively. Moreover, treatment of cells containing either construct A or B with cobalt chloride and azide together resulted in luciferase expression that was larger than the re-sponse to either agent alone, with the effect being nearly additive for both constructs (ϳ8-fold; n ϭ 12 for both constructs). Deletion of ϳ2,500 bp from the 5Ј end of the ϳ6-kbp DNA (construct C) results in a complete loss of the response to azide. In contrast, the response of construct C to cobalt chloride, although reduced to approximately one-half that of constructs A or B, is still present (Fig. 6A). The positive response of construct C to cobalt chloride maps to a 480-bp PstI/SacI fragment between ϳϪ3.5 and ϳϪ3.0 kbp from the transcription start site; deletion of this region (constructs D-H) is marked by a loss of the response to cobalt chloride, and inclusion of the 480-bp fragment upstream to 200 bp of proximal glut1 promoter reconstitutes the response to cobalt chloride (compare constructs C, H, and I). Interestingly, fusion of the 666-bp BglII/PstI segment, located between ϳϪ6.0 and ϳϪ5.3 kbp of the transcription start site, upstream to construct C reconstitutes the response to azide (compare construct B with C) and augments the response to cobalt chloride.
To determine whether the 480-bp fragment behaves as an enhancer, we tested whether it can increase the function of a heterologous promoter and confer cobalt chloride-inducibility to it. For these studies, the fragment was cloned upstream of a minimal c-fos promoter (Fig. 6B). Inclusion of the 480-bp fragment led to enhanced basal expression of the minimal c-fos promoter by Ͼ20-fold (see legend to Fig. 6). Moreover, this fragment confers cobalt chloride-inducibility on the heterologous c-fos promoter in an orientation-independent manner (constructs K and L). Interestingly and unexpectedly, the 480-bp fragment also confers azide-inducibility on the minimal c-fos promoter (Fig. 6B).
FIG. 6. Analysis of the 5-flanking region of the rat glut1 gene. A, regions upstream of the rat glut1 gene mediating the effects of azide and cobalt chloride on glut1 gene transcription were mapped using various deletion constructs. Transient transfection and luciferase assays were performed as described under "Experimental Procedures." -Fold induction is calculated as the ratio of luciferase activity in cobalt chloride-or azide-treated cells to the activity present in diluent-treated cells measured in parallel plates. Experiments were performed at least twice in triplicate or quadruplicate culture plates and the results averaged; values are expressed as means Ϯ S.E. (n Ն 6). 1, after normalization of luciferase activity of the test vectors to the activity of the control chloramphenicol acetyltransferase reporter (SV40-chloramphenicol acetyltransferase), basal activities were expressed relative to the activity of construct H, which was given a value of 1.0. Construct J did not contain the glut1 TATA box and extended only 15 bp upstream of the glut1 major transcription start site. Basal luciferase activity varied by only 2-fold in constructs shown and averaged ϳ10 5 relative light units/mg of protein. B, the 480-bp PstI/SacI fragment was cloned upstream to the c-fos minimal promoter driving the luciferase gene. 2, basal luciferase activity (ϳ10 3 relative light units/mg of protein) was increased ϳ20to 40-fold by inclusion of the 480-bp fragment. The effect of cobalt chloride and azide on luciferase activity was determined as described above.

TABLE II
Effect of hypoxia on GLUT1 mRNA content and GLUT1 promoter activity in Clone 9 cells Clone 9 cells were transfected as described under "Experimental Procedures" and pooled into two sets of six plates. In parallel hypoxic experiments, one set was used for GLUT1 mRNA content analysis and the other set was used to measure the activity of a firefly luciferase reporter construct containing ϳ6.0 kbp of GLUT1 5Ј-flanking sequence (Fig. 6A, construct A). The experiment was performed in triplicate and the results averaged (mean Ϯ S.E.). GLUT1 mRNA content and luciferase activity is expressed as -fold increase in cells exposed to nominal 3% oxygen (hypoxic) for 24 h relative to control cells incubated in 20% oxygen (normoxic). In each plate, luciferase activity was calculated as the ratio of firefly luciferase activity of the test vector (construct A) to Renilla luciferase activity of the control vector (pRL-TK). Similar results were obtained in a second, independent experiment.
Sequence Analysis of the Rat glut1 Enhancers-The 480-bp SacI/PstI segment and the 666-bp BglII/PstI segment were sequenced and analyzed for the presence of potential binding sites for transcriptional regulatory elements (Fig. 7, A and B,  respectively). Interestingly, a region in the 3Ј end of the 480-bp fragment that shares a high degree of homology to the HIE of the human or mouse EPO gene was identified (2,26,27,35). In addition, this fragment contains two phorbol ester response elements (42), two SP-1s (43), an SRE (44), and three core p53 binding sites (45). A search of the GenBank TM data base revealed considerable sequence homology between the 480-bp fragment identified in this study and the Enhancer-1 region of the mouse glut1 gene located ϳ3.1 kbp upstream of the transcription start site (46). By deletion analysis, the SRE homology region has been shown to be critical to the stimulation of the mouse glut1 Enhancer-1 by serum or platelet-derived growth factor (46).
The sequence of the 666-bp segment was searched for the presence of potential cis-regulatory elements using a transcription factor data base (Fig. 7B) (43). We found regions matching binding sites for several trans-acting factors, including activator proteins 2 and 3 (AP-2, AP-3), C/EBP, p53, and nuclear FIG. 7. Nucleotide sequence of the rat glut1 5 enhancers. A, the 480-bp segment. The 5Ј end of the 480-bp rat sequence (PstI) was located ϳ3.5 kbp upstream of the glut1 major transcription start site. Potential transcription factor binding sites are delineated as follows: HIE (boxed), p53 (dotted underline), phorbol ester response element (double overline), SRE (single overline), and SP-1 (underline). The mouse Enhancer-1 sequence was located in a similar region to that of the rat gene. For optimal alignment, the rat and mouse sequences have been extended by six and five spaces, respectively. B, the 666-bp segment. The 5Ј end of the 666-bp segment (BglII) was located ϳ6.0 kbp upstream of the glut1 major transcription start site. Potential transcription factor binding sites are as follows: HIE (boxed), p53 (dotted underline), C/EBP (double overline), AP-2 (single overline), AP-3 (underline), NF-1 (star overline), NF-3 (star underline), and steroid receptor half-site (arrow). factors 1 and 3 (NF-1, NF-3). Of interest is the presence of an HIE site (2). In addition, this region contains a number of cis-elements that have been shown to mediate tissue-specific and developmental gene regulation. The functional significance of these sites remains to be determined. DISCUSSION It is well established that the expression of a number of genes is modified during the adaptive response of cells and tissues to exposure to hypoxic conditions (6,9,30,31,36,47). Since hypoxia leads to the inhibition of oxidative phosphorylation, it is important to distinguish between regulatory mechanisms that are controlled by decreased levels of oxygen per se from those that are controlled by the attendant inhibition of oxidative phosphorylation. This distinction can be made through employment of specific chemical agents that selectively mimic the actions of the different components of the hypoxic response. Specifically, cobalt chloride employed in the presence of oxygen mimics many of the effects of lowered oxygen concentration per se on the expression of several genes (1, 2, 5, 26 -28, 30, 31), and azide or cyanide lead to an inhibition of oxidative phosphorylation in the absence of hypoxia (11,20,21,32). Employing such a strategy, we have shown that glut1 gene expression is stimulated in response to hypoxia by two distinct pathways. The first pathway appears to involve oxygen-sensing regulatory molecules similar to those regulating the expression of EPO and vascular endothelial growth factor genes (1,2,5,26,27,46). This pathway is stimulated by exposure to cobalt chloride in the presence of oxygen, an effect that is shared by Ni(II) and Mn(II) (1). The second pathway leading to the induction of GLUT1 mRNA is activated in response to inhibition of oxidative phosphorylation in the absence of hypoxia.
The pathways leading to induction of GLUT1 mRNA expression in Clone 9 cells in response to hypoxia per se (i.e. by cobalt chloride) versus in response to inhibitors of oxidative phosphorylation (i.e. by azide) differ in a number of important respects. Exposure to cyanide or azide leads to an immediate inhibition of oxidative phosphorylation and a fall in cell ATP content that is associated with a rapid stimulation of glucose transport that reaches ϳ8-fold control values by 1 h (20 -22). In contrast, no change in cell ATP content or in the rate of glucose transport was observed during a 1-h period of exposure to cobalt chloride. Incubation for longer periods in the presence of either azide or cobalt chloride led to increases in GLUT1 mRNA content and a stimulation of glut1-mediated glucose transport, with the effect of cobalt chloride on GLUT1 mRNA content being greater in magnitude and becoming evident somewhat earlier. Azide and cobalt chloride also differed in their effects on GLUT1 mRNA turnover; whereas the half-life of GLUT1 mRNA increased from ϳ1.5 to Ͼ8 h in cells exposed to azide, GLUT1 mRNA half-life remained unaltered in cells incubated in the presence of cobalt chloride. This finding not only suggests that the increase in GLUT1 mRNA content in response to cobalt chloride is mediated primarily at the transcriptional level but also suggests that the previously reported stabilization of GLUT1 mRNA in cells exposed to hypoxia may be due to the attendant inhibition of oxidative phosphorylation (48).
Employing deletion analysis and site-directed mutagenesis, DNA elements in the mouse Enhancer-1 region that mediate the induction by inhibitors of oxidative phosphorylation and by hypoxia have very recently been reported (49). Results of our experiments identifying the region of the rat glut1 enhancer (480-bp fragment) necessary for the response to cobalt chloride are in agreement with this analysis. The mouse Enhancer-1 contains a 7-bp element with high homology to the HIE (49). Moreover, mutation of the core 4-bp consensus nucleotides in the mouse glut1 Enhancer-1 abolishes its positive transcriptional response to cobalt chloride (49). Interestingly, the rat glut1 sequence of this region bears greater homology than that of the mouse gene to the HIE present in either the human or mouse EPO gene (2,26,27,49), and it would be expected that deletion or mutation of this region in the rat glut1 promoter should likewise lead to the elimination of cobalt chloride responsiveness. It is worth emphasizing that although the 480-bp segment was necessary for the response to cobalt chloride, this segment alone was not sufficient for the full response to cobalt chloride. Inclusion of the 666-bp segment located upstream of this region augmented the cobalt chloride response by more than 2-fold (compare constructs C and I with constructs A and B in Fig. 6). Interestingly, within the 666-bp segment we found one region with high homology to the core HIE consensus. Although the function of this region remains to be determined, it is tempting to speculate whether it might cooperate with the downstream HIE site contained within the 480-bp segment to mediate the full cobalt chloride response.
Our results, however, differ markedly from conclusions reached on the region in the mouse glut1 Enhancer-1 that is necessary for glut1 induction in response to inhibition of oxidative phosphorylation (49). Analysis of the mouse Enhancer-1 sequence revealed the presence of an SRE (44,46); an identical sequence is also present in the 480-bp enhancer segment of the rat glut1 gene (Fig. 7). Deletion or mutation of this latter region in the mouse glut1 Enhancer-1 segment cloned upstream to a heterologous promoter was associated with a loss of induction in response to azide or other inhibitors of oxidative phosphorylation (49). Although we also demonstrated that the 480-bp segment conferred inducibility of the minimal c-fos promoter to azide, we found that the segment was not sufficient to confer such inducibility to the rat glut1 promoter in the context of additional glut1 flanking sequences (Fig. 6, A and B). In keeping with this inference, we found that construct C, which contains the 480-bp segment, showed no positive response to azide. Interestingly, the addition of the 666-bp fragment to construct C reconstituted the full azide response (compare constructs A and B with C). This suggests that the 666-bp region is necessary for the response to azide, although sequences contained in the 480-bp region appear to augment the response to azide (compare constructs A and B with D).
What then is the physiological significance of the 480-bp segment's response to azide, as described previously (49) and confirmed in this study (Fig. 6B). Although the answer to this question remains to be resolved, we would point out that: 1) the response to azide is observed only in studies examining the role of this region in isolation by use of minimal heterologous promoters, and that the response is not observed in the context of the rat glut1 promoter/5Ј-flanking region; and 2) it has been reported that both Fos and Jun proteins are induced in some cell systems in response to hypoxia (50,51,52). If a similar response occurs in our cells, then increased activator protein-1 (AP-1) activity may stimulate the SRE present in the 480-bp segment and lead to the observed response to azide in the context of the heterologous promoter constructs. However, a similar series of events may be ineffective at stimulating the gene in response to inhibition of oxidative phosphorylation because of additional cis-elements present both upstream and downstream to the 480-bp segment located in its naturally occurring context. Further studies are required to clarify these issues. Taken together, our results indicate that the ϳ6-kbp 5Ј-flanking region, which includes sequences in addition to the 480-bp enhancer region, is necessary for the stimulation of glut1 gene expression in response to inhibition of oxidative phosphorylation.
Results of the present study demonstrate that glut1 gene expression is dually regulated by hypoxia, namely by a reduction in oxygen concentration per se and by the attendant inhibition of oxidative phosphorylation. Such dual control by hypoxia has not been reported for other hypoxia-responsive genes, whose expression is thought to be regulated only by decreased concentrations of oxygen (1-3, 5, 26, 30, 31, 36, 40). In the case of glut1, we can only speculate that the evolutionary advantage of its dual regulation by hypoxia may reflect the fact that glucose transport is rate-limiting for glucose metabolism in a variety of cells (11,53); dual control of GLUT1 expression would enable sustained high rates of glycolytic ATP synthesis in the face of reduced mitochondrial ATP production resulting either from exposure to low oxygen concentrations or from exposure to poisons or toxins inhibiting oxidative phosphorylation. It should be noted that the regulation of other known hypoxia-responsive genes by agents such as azide has not been examined rigorously, and it is possible that the expression of some of these genes is also controlled in a dual fashion by hypoxia. Further studies are necessary to identify and characterize other hypoxia-responsive genes that, similar to glut1, might be dually regulated. Moreover, although the sequence of events and mechanisms underlying the regulation of specific genes by hypoxia per se are becoming better understood, virtually nothing is known concerning the regulation of genes responding to inhibition of oxidative phosphorylation. Clearly, the latter constitutes an important area for future investigation.