Hypoxia Regulates β-Enolase and Pyruvate Kinase-M Promoters by Modulating Sp1/Sp3 Binding to a Conserved GC Element*

The transcription rates of glycolytic enzyme genes are coordinately induced when cells are exposed to low oxygen tension. This effect has been described in many cell types and is not restricted to species or phyla. In mammalian cells, there are 11 distinct glycolytic enzymes, at least 9 of which are induced by hypoxia. Recent reports described a role for the hypoxia-inducible factor-1 (HIF-1) in the transcriptional activation of lactate dehydrogenase A, aldolase-A, phosphoglycerate kinase, and enolase-1 genes. It is not known whether the HIF-1 factor acts exclusively to regulate these genes during hypoxia, or how the other genes of the pathway are regulated. In this paper, we describe analyses of the muscle-specific pyruvate kinase-M and β-enolase promoters that implicate additional mechanisms for the regulation of glycolytic enzyme gene transcription by hypoxia. Transient transcription of a reporter gene directed by either promoter was activated when transfected muscle cells were exposed to hypoxia. Neither of these promoters contain HIF-1 binding sites. Instead, the hypoxia response was localized to a conserved GC-rich element positioned immediately upstream of a GATAA site in the proximal promoter regions of both genes. The GC element was essential for both basal and hypoxia-induced expression and bound the transcription factors Sp1 and Sp3. Hypoxia caused the progressive depletion of Sp3 determined by DNA binding studies and Western analyses, whereas Sp1 protein levels remained unchanged. Overexpression of Sp3 repressed expression of β-enolase promoters. It is concluded that hypoxia activates these glycolytic enzyme gene promoters by down-regulating Sp3, thereby removing the associated transcriptional repression.

The transcription rates of glycolytic enzyme genes are coordinately induced when cells are exposed to low oxygen tension. This effect has been described in many cell types and is not restricted to species or phyla. In mammalian cells, there are 11 distinct glycolytic enzymes, at least 9 of which are induced by hypoxia. Recent reports described a role for the hypoxia-inducible factor-1 (HIF-1) in the transcriptional activation of lactate dehydrogenase A, aldolase-A, phosphoglycerate kinase, and enolase-1 genes. It is not known whether the HIF-1 factor acts exclusively to regulate these genes during hypoxia, or how the other genes of the pathway are regulated. In this paper, we describe analyses of the musclespecific pyruvate kinase-M and ␤-enolase promoters that implicate additional mechanisms for the regulation of glycolytic enzyme gene transcription by hypoxia. Transient transcription of a reporter gene directed by either promoter was activated when transfected muscle cells were exposed to hypoxia. Neither of these promoters contain HIF-1 binding sites. Instead, the hypoxia response was localized to a conserved GC-rich element positioned immediately upstream of a GATAA site in the proximal promoter regions of both genes. The GC element was essential for both basal and hypoxia-induced expression and bound the transcription factors Sp1 and Sp3. Hypoxia caused the progressive depletion of Sp3 determined by DNA binding studies and Western analyses, whereas Sp1 protein levels remained unchanged. Overexpression of Sp3 repressed expression of ␤-enolase promoters. It is concluded that hypoxia activates these glycolytic enzyme gene promoters by down-regulating Sp3, thereby removing the associated transcriptional repression.
Glycolysis is induced in most cell types by anaerobic or hypoxic conditions when oxidative metabolism is repressed. In mammalian cells, the glycolytic pathway has 11 separate enzymes, some with multiple tissue-specific isoforms, each encoded by separate genes and mostly situated on unlinked chromosomal loci (reviewed in Ref. 1). We reported previously that the transcription rates of glycolytic enzyme genes are coordinately induced in muscle cells subjected to hypoxia (2,3). Consistent with the appearance of elevated steady state mRNA levels, the transcription rates increased gradually over 1-2 days (2). Similar effects have been described in endothelial cells exposed to hypoxia (4). The signaling pathways and genetic elements that control this response are probably complex. In recent reports, the hypoxia-inducible factor 1 (HIF-1), 1 originally identified as a factor in the regulation of the erythropoietin gene (reviewed in Ref. 5), has been shown to play a role in the transcriptional activation of a number of genes by hypoxia (6 -10). HIF-1 binds to an enhancer element containing the core sequence ACGTGC, which is an obligatory and minimal component of HIF-1-responsive genes.
DNA sequence and functional analyses have revealed the presence of active HIF-1-binding sites in the non-coding regions of mammalian lactate dehydrogenase A (LDH-A), aldolase-A, enolase-1, and phosphoglycerate kinase (6,7,9). They have not been described in a number of the other mammalian glycolytic enzyme genes including hexokinase, glyceraldehyde-3-phosphate dehydrogenase, glucose-phosphate isomerase, ␤-enolase, or some of the other tissue-specific isogenes. The absence of appropriately located sites indicates that there may be other mechanisms for regulating these genes. An additional consideration is that hypoxia response elements distinct from HIF-1 binding sites have been described in plant alcohol dehydrogenase, LDH, and aldolase gene promoters (11)(12)(13). Therefore, there is a precedent and perhaps a requirement for additional hypoxia-regulated mechanisms for mammalian glycolytic enzyme gene expression.
Promoter elements with the core sequence GGGC/T/AGG bind a number of transcription factors including members of the Sp family (14 -17). The factors Sp1 and Sp3 appear to be ubiquitous in mammalian cells where they compete with similar binding affinities for common target sequences (18). Sp1 is typically a positive-acting transcription factor and may interact directly with the basal TFIID transcriptional complex (16). In contrast, Sp3 appears to possess both transcriptional repressor and activating properties, the relative strengths of which depend on the promoter context as well as the cellular background (16 -21 Here, we present evidence that changes in the relative abundances of Sp1 and Sp3 contribute to the positive regulation of the glycolytic enzyme genes PK-M and ␤-enolase by hypoxia. Exposure of C 2 C 12 myocytes to hypoxia caused depletion of Sp3 from nuclear extracts without affecting Sp1. The loss of Sp3 binding to the conserved GC element in these promoters correlated with transcriptional activation. Overexpression of Sp3 caused transcriptional repression of cotransfected ␤-enolase promoters in support of a regulatory mechanism involving the hypoxia-mediated alleviation of Sp3 repression.

MATERIALS AND METHODS
Cell Culture, Transfections, and Hypoxia-Our culture conditions and transfection of C 2 C 12 myocytes have been described previously (22,23). Cells were transfected at about 30% confluence and maintained in 20% fetal bovine serum, for the duration of the experiment (myoblasts) or the medium was replaced with low mitogen medium (0.5% horse serum) and incubated for an additional 4 -6 days (serum-starved, myotubes). Myocytes in 60-mm culture plates were transfected with 9.5 g of the test plasmid (␤-enolase or PK-M) and 0.5 g of pRL-TK (Promega Corp., Madison, WI) as the internal control to correct for variations in transfection efficiency. For overexpression/cotransfection analyses, the indicated amounts of reporter and interfering plasmid (Sp1 or Sp3) were transfected, again with 0.5 g of pRL-TK. All values are expressed as ratios of the test:internal control reporter expression. Our methods for exposing cells to hypoxia have been described previously (24,25). Equal amounts of protein were assayed for expression of luciferase using the Promega Corp. dual-luciferase reporter assay system. Protein was assayed using a Bio-Rad assay kit.
Nuclear Extracts and Gel Mobility Shift Assay-Nuclear extracts were prepared from confluent plates of C 2 C 12 myocytes grown under a normal aerobic environment or under hypoxia as described previously (28,29). For hypoxic cell extracts, cell lysis was performed with the cells still under hypoxia to avoid reoxygenation effects (30). Sequences of the oligonucleotide probes used were as follows: SRE: 5Ј-CAACACCCAAA-TATGGCT-3Ј; ␤-ENO-GC WT GC box: 5Ј-AAAGAGAGGCGGGGCTG-GCTGGG-3Ј. Gel-purified double-stranded oligonucleotides were endlabeled with [ 32 P]ATP using T4 polynucleotide kinase (Promega) and [␥-32 P]ATP (NEN Life Science Products). Equal amounts of radioactive probe (1.5-2.5 ϫ 10 4 cpm) were added to binding reactions that contained 8 g of nuclear extract protein in 20 l of a buffer containing 4 mM Tris (pH 7.8), 12 mM Hepes (pH 7.9), 60 mM KCl, 30 mM NaCl, 0.1 mM EDTA, 1 g of poly(dI-dC) (Amersham Pharmacia Biotech). Reactions were incubated for 20 min at 22°C before separating on nondenaturing 5% polyacrylamide gels at 4°C (29). Where indicated, an-FIG. 1. Regulation of PK-M and ␤-enolase promoters by hypoxia. The constructs indicated were cotransfected with pRL-TK (Promega Biotech) into C 2 C 12 myoblasts as described under "Materials and Methods." The medium was replaced after 18 h, and, after an additional 24 -48 h, the cells were transferred to hypoxia (0.5% O 2 ) or maintained under an aerobic (21% O 2 ) atmosphere. After an additional 20 h, cultures were harvested for luciferase assays. Data are expressed as the ratio of test plasmid luciferase activity/internal control RN-luciferase/g of protein.
Results show means Ϯ S.E. from at least four separate experiments for each plasmid. The numbers to the left of each construct indicate the length of the promoter numbered 5Ј from the transcription start site, for pyruvate kinase M (PK-M) and ␤-enolase (␤-Eno) promoters. pGL HRE Epo has been described previously (26), pRSV-luc contains the murine Rous sarcoma virus promoter inserted upstream of the luciferase gene, and pRL-TK (Promega Biotech) and p␣MHC 86 (both negative controls) have also been described previously (26,44). In the promoter diagrams, open squares represent GATAA elements, filled ovals PK-M/␤-Eno GC elements, open ovals Sp1 sites, and filled squares HIF-1 binding sites; Luc in the open box represents the luciferase gene. tibodies (2 or 4 g/reaction) were added to the binding reactions before adding the probe and the reactions were incubated for an additional 30 min at 22°C; competitor (unlabeled oligonucleotides) were added immediately before the probe, at 100-fold molar excess.
Northern and Western Blots-Northern blot procedures were exactly as described previously (22,28,29). For Sp1, the insert cDNA was excised from pSVSp1F; for Sp3, the insert was excised from pCMVSp3; the ␤-actin probe has been described previously (29). Our Western blot procedures have also been described previously (30). Anti-Sp3, Sp1, and actin antibodies were from Santa Cruz Biotechnology, Inc. Duplicate gels were stained with Coomassie Blue and filters were stained with Ponceau S to check for protein loading and transfer.
Statistics-Numerical data are expressed as means Ϯ S.E., and p values were determined by t test using Macintosh InStat 2.00 software (GraphPad, La Jolla, CA). Fig. 1 shows the effects of hypoxia on the transient expression directed by 5Ј regions of the PK-M and ␤-enolase promoters in C 2 C 12 myoblasts. Solid circles shown in the figure represent the positions of GC elements that are putative binding sites for the Sp family of transcription factors. Both promoters were induced by approximately 4-fold under hypoxia. Deletions of ␤-enolase from Ϫ628 to Ϫ80 had a minimal effect on hypoxic induction, but basal expression was reduced by ϳ 70%, indicating that positive regulatory sites, not involved in hypoxic induction, are located between Ϫ80 and Ϫ628 (see Fig. 2B). When the promoter was truncated to Ϫ45, basal expression was decreased further, and hypoxia-inducible expression was lost. Positive and neutral controls are also shown in Fig. 1. The SV40 early promoter containing 4 copies of the erythropoietin hypoxia response element (pGL HRE Epo) was induced 5.3 Ϯ 1.3-fold, and expression from the Rous sarcoma virus, truncated ␣-myosin heavy chain, and thymidine kinase promoters (pRSV-Luc, p␣MHC 86 Luc, pRL-TK) were minimally affected by hypoxia. Therefore, the PK-M, ␤-enolase, and SV40hypoxia response element promoters were specifically induced by hypoxia and the levels of induction were similar, implicating the presence of hypoxia-responsive elements in all three promoters.

Identification of the Hypoxia-responsive Region of the ␤-Enolase Promoter-
In Fig. 2, the DNA sequences of the PK-M and ␤-enolase promoters downstream of Ϫ90 are aligned; they contain identical 16-bp conserved sequence regions centered at Ϫ72 and Ϫ61 bp, respectively, that include a central GC box (highlighted in Fig. 2A). Neither promoter contains a classical TATA box site, but the ␤-enolase gene has a GATAA sequence at position  Fig. 1. The expression of each construct relative to p␤Eno 628 was also calculated as described above. Results are means Ϯ S.E. from at least six separate experiments. C, response to hypoxia. Constructs were transfected, exposed to hypoxic or aerobic conditions, and analyzed as described in the legend for Fig. 1. Results are means Ϯ S.E. from at least six experiments.
Ϫ34, and PK-M has the same sequence at Ϫ8. This combination of GATA-like element and upstream GC box is reminiscent of the erythroid-specific PK promoter as well as other erythroid and myeloid-specific genes (31,32). To examine the possible contributions of these elements in the response to hypoxia, we mutated each site as indicated in the ␤-enolase promoter, and assayed for transient expression.
The relative strengths of the different promoter deletions and mutations under aerobic conditions are shown in Fig. 2B. p␤-Eno 628 expression was about the same as the positive control pGL HRE Epo. Deletion to Ϫ101 or Ϫ80 reduced basal (aerobic) expression by 45% and 65%, respectively. Mutation of the GC element in Ϫ101 (M1) or deletion of the promoter to Ϫ45 bp, reduced basal expression by Ͼ90%, but in both cases the basal expression was still 2-3-fold greater than the promoterless vector (pGLBV). The GATAA element contributed significantly to basal promoter activity. Replacement of this element with a classical TATAA box (M3) almost doubled basal expression, whereas replacement with a random sequence reduced expression to 25% of the wild type. These results indicate that the GATAA element contributes to basal promoter activity, but may be less efficient than a classical TATA box sequence.
The responses of these mutations to hypoxia is shown in Fig.  2C. The Ϫ101 and Ϫ80 truncations were still responsive to hypoxia (see also Fig. 1). Mutation of the GC element in Ϫ101-Eno (M1) caused a precipitous loss of both aerobic and hypoxic activity. This finding suggests that the loss of basal and hypoxia-inducible activity of the Ϫ45 bp truncation is due to the deletion of the GC element and indicates that an intact GC box is required for efficient basal expression and induction by hypoxia. Mutation of the GAGA sequence (M2) that flanks the GC element of both ␤-enolase and PK-M promoters did not interfere with basal expression or induction by hypoxia. Similarly mutation of the proximal GATAA element to a TATAA element (M3) or to a nonsense sequence (M4) did not reduce the -fold induction by hypoxia. Therefore, the GATAA element probably does not contribute directly to the activation of the promoter by hypoxia. Instead, the results of both deletion and mutation analyses support a critical role for the GC element in basal expression and induction by hypoxia.
Protein Binding Studies-Gel mobility shift assays were used to determine whether protein-DNA binding to the GC elements correlated with the function of the promoter. In a recent report, we demonstrated that purified Sp1 binds to the PK-M/␤-enolase GC element described in Fig. 2A (29). In the same study, we demonstrated that an oligonucleotide containing the GC region mutation M1 (described in Fig. 2) did not react with purified Sp1 or nuclear extract proteins. This is consistent with the requirement for an intact Sp-binding site for basal function and hypoxia-inducibility of these promoters. Since a number of proteins bind GC elements, including other members of the Sp family, specific antibodies were used to identify individual components in the complexes. These results are shown in Figs. 3 and 4. In the absence of antibody, three specifically shifted bands can be identified, including one fast and two slower migrating bands. The highest mobility band disappeared after 24 h in hypoxia (Fig. 3A, second lane). Preincubation with an anti-Sp1-specific antibody primarily supershifted the lowest mobility band (Fig. 3A, lane 5, and B, lane 3). This is in agreement with previous reports (17,18,21) and indicates that this complex is composed mostly of Sp1. Preincubation of extracts with Sp2, Sp4, or Egr-1 antibodies (Fig.  3A, lanes 6, 8, and 9) did not change the gel shifts. To confirm this, in Fig. 3B, the same samples were separated on 3.5% polyacrylamide to resolve the low mobility complexes. Again, there was no apparent interaction of Sp2, Sp4, or Egr-1 antibodies with any of the specific complexes. In contrast, preincubation with the Sp3 antibody eliminated the rapidly migrating band and almost eliminated the intermediate complex (Fig. 3A, lane 7, and B, lane 5). Reaction with the Sp3 antibody generated two faint supershifts. These results are in agreement with previous reports and identify Sp3 as the main component of the two slower migrating complexes (17,18,21). Panel C of Fig. 3 indicates that binding of serum response factor (SRF) to its recognition site (SRE) was not affected by hypoxia.
Additional experiments were carried out to determine the kinetics and dynamics of hypoxia-induced changes in Sp1 and Sp3 binding to the ␤-enolase/PK-M GC element. Fig. 4A demonstrates that hypoxia did not affect the Sp1-specific complex, but the high mobility Sp3-containing band was progressively reduced and eliminated after 8 h. The individual Sp1 and Sp3 bands can not be distinguished in these autoradiographs.
When the same extracts were treated with anti-Sp3 antibody, the supershifted bands (Fig. 4B, SS-Sp3 arrows) were also reduced during hypoxia and eliminated after 8 h. Again, there was no discernible change of the Sp1 complex. Treatment of the extracts with both Sp1 and Sp3 antibodies generated the expected three supershifted bands corresponding to two Sp3 and one Sp1 supershifts (Fig. 4C, SS arrows). Again, the Sp1 supershift did not change in the hypoxia extracts, but the supershifted Sp3 bands disappeared. These results indicate a selective depression of Sp3 but not Sp1 protein level in nuclear extracts form hypoxia-treated cells. Consistent with this, the residual Sp1 band that did not react with the antibody, did not change under hypoxia (panel C). However, a new band (X in panel C) with close mobility to the slow Sp3 complex, but antigenically distinct from Sp3, appeared within the first hour of hypoxia and increased progressively over 24 h. We do not know the identity, properties, or function of this new complex.
Western Blot Analyses-As a second approach to measure changes of Sp1 and Sp3 in response to hypoxia, nuclear proteins from aerobic and hypoxic cells were subjected to Western analyses. In Fig. 5A, anti-Sp3 antibody recognized two specific proteins migrating with apparent molecular sizes of approximately 100 and 80 kDa, respectively. This is consistent with previous reports, and the different size proteins may result from the use of alternative start sites on the Sp3 gene (17,21). Both Sp3-specific bands were reduced in the hypoxia extracts and were almost eliminated after 24 h. These results support the gel mobility shift data described above. Cells transfected with a CMV driven Sp3 cDNA were analyzed for comparison (lane 5). The same size bands were recognized, and overexpression of the Sp3 proteins is apparent. In panel B, the same blot probed with anti-actin antibody showed no change of actin, demonstrating equivalent loading and transfer of proteins. Fig.  5C shows a Western blot probed with anti-Sp1 antibody. In contrast to the effects on Sp3, there was no overall change of total Sp1 protein in the extracts from aerobic and hypoxic cells.
In agreement with previous studies, the Sp1 antibody recognized two specific bands migrating at ϳ95 and 106 kDa. These bands have been identified as non-phosphorylated and phosphorylated Sp1, respectively (33). As indicated in the figure, there was a trend toward an increase of the phosphorylated Sp1 protein in the extracts from longer hypoxia exposure times. There was a 26 Ϯ 6% increase in the p106 form of Sp1 after 16 -20 h under hypoxia (n ϭ 5, p Ͻ 0.05 by t test).
Transcript Levels of Sp1 and Sp3-The protein binding studies and Western analyses indicated decreased levels of Sp3 but not Sp1 in nuclear extracts from cells exposed to hypoxia. To determine whether hypoxia caused changes of Sp gene expression, mRNA transcript levels were measured by Northern blots. Fig. 6 shows that there was no significant change in Sp1 or Sp3 transcripts during 24 h of hypoxia.
Overexpression of Sp1 and Sp3-The reciprocal correlation between Sp3 abundance and transcriptional activation of the ␤-enolase promoters by hypoxia is consistent with an alleviation of Sp3-mediated repression. Clearly, this can only be the case if Sp3 is a repressor of these promoters. To determine whether Sp3 can function as a repressor of ␤-enolase, C 2 C 12 myocytes, cotransfected with ␤-enolase promoters and pCMV-Sp3, were subjected to hypoxia and analyzed for luciferase expression. The results are shown in Fig. 7. Overexpression of Sp3 significantly repressed the hypoxia-mediated activation of both ␤-enolase constructs with intact GC elements but did not affect the expression of the M1 mutation (Sp3 coexpression did not significantly affect aerobic or hypoxic expression of the M1 mutation). The HIF-1-dependent induction of p␣MHC 86 HRE was similarly unaffected by Sp3 overexpression. Neither of the Sp1 vectors influenced ␤-enolase inducibility under the same conditions. These results indicate that Sp3 can function as a transcriptional repressor of the basal ␤-enolase promoter, probably through interaction with the GC element. DISCUSSION We show in this article that the region between Ϫ80 and Ϫ28 of the human ␤-enolase gene confers positive regulation by hypoxia. The region contains a GC-rich, Sp-family binding site and a 3Ј proximal GATAA element. A similar sequence combination is present in the same location of the hypoxia-responsive PK-M promoter in the rat (27) and is a common combination in erythroid-specific promoters, many of which are directly or indirectly regulated by oxygen availability (31,32). When the ␤-enolase GC element was mutated or deleted, basal (aerobic) expression was reduced and the response to hypoxia was lost. This indicates that the GC element is required alone or in combination with downstream element(s) to promote transcriptional activation by hypoxia.
Gel mobility shift binding studies demonstrated that mutation of the GC element eliminated the binding of Sp proteins coincident with loss in function of the promoter. We reported previously that this site was also responsible for hyperoxic regulation of ␤-enolase expression through the reversible oxidation of Sp1 cysteine sulfhydryl groups (29). These observations implicate the GC element as a critical redox control site in the ␤-enolase, PK-M, and possibly other promoters with a similar arrangement of proximal elements.
Protein binding studies identified Sp1 and Sp3 as the major components of the complexes formed between nuclear extracts and the PK-M/␤-enolase GC element. Although we cannot preclude involvement of other factors, the combination of Sp1 and Sp3 antibodies blocked or supershifted each of the specific complexes. Therefore, these proteins are implicated in the regulation. Protein binding as well as Western blots indicate that Sp1 levels did not change markedly in response to hypoxia whereas Sp3 levels fell dramatically.
The loss of Sp3 binding could account for the transcriptional activation if Sp3 functions as a repressor when bound to these promoter elements. The effects of overexpressing Sp3 indicates that Sp3 can indeed repress transcription from the ␤-enolase promoters. Previous studies, typically using Schneider SL2 cells that lacks endogenous Sp factors, have shown that Sp3 is a strong repressor of some GC-dependent promoters, and it can FIG. 6. Effects of hypoxia on transcript levels. C 2 C 12 cells were exposed to aerobic or hypoxic conditions as described under "Materials and Methods" and harvested at the indicated times for Northern blot analyses. Equal amounts of RNA were loaded, and the blot was probed sequentially with the indicated probes. Data are representative of four separate experiments.

FIG. 7.
Repression of ␤-enolase promoters by overexpressing Sp3. C 2 C 12 cells were cotransfected with equal amounts (5 g each) of reporter and Sp plasmids and 0.5 g of pRL-TK. After 24 h the medium was replaced with growth medium. After 24 -48 h the cultures were exposed to hypoxia or remained aerobic for another 20 h. Harvesting and assays were as described under "Materials and Methods" and in the legend for Fig. 1. Data are presented as mean Ϯ S.E. Repression of Ϫ628 ␤-Eno and Ϫ101 ␤-Eno by cotransfected pCMV-Sp3 was significant (p ϭ Ͻ 0.005 and Ͻ 0.001, respectively, by paired t tests, n ϭ 6). The expression of p101Eno M1 was unaffected by cotransfected pCMVSp3 (p ϭ 0.138, n ϭ 4), p␣MHC 86 HRE was not significantly affected, and cotransfections of pSVSp1F, pSVSp1FX, pCMV-c-Jun, or pCMV-SRF did not affect the hypoxia inducibility of Ϫ101␤-Eno (also by t test). antagonize multiple classes of positive-acting factors (17)(18)(19)(20)(21). Our observation that pCMV-Sp3 caused Ͼ60% repression of Ϫ101␤-Eno in hypoxic C 2 C 12 cells is substantial because these cells already have a high Sp background. The observation that overexpression of Sp3 countered the promoter induction by hypoxia provides strong evidence that Sp3 depletion is associated with gene activation. We propose that the hypoxia-mediated depletion of Sp3 alleviates a repressor activity, thereby activating transcription from the ␤-enolase promoters.
It is not clear whether Sp1 has an active or passive role in this regulation. Previous studies indicated that Sp3 may repress transcription independently of Sp1 activity, probably by interacting directly with other transcription factors (16,20). Sp3 may also compete with Sp1 for the common DNA binding site, thereby reducing transcriptional activation by Sp1 (17,21). The observation that overexpressing Sp1 did not augment the activity of Ϫ101␤-Eno suggests either that endogenous Sp1 is saturating, or that Sp3 repression is independent of Sp1. Indeed, we have no direct evidence to indicate that Sp3 binding is replaced by Sp1 during hypoxia. The apparent increase of Sp1 p106 during hypoxia (Fig. 5) indicates an increased proportion of phosphorylated Sp1 in hypoxic nuclear extracts, and this in turn may reflect an increased binding of Sp1 because DNA-bound Sp1 is phosphorylated by a DNA-dependent kinase (33,34). Alternatively displaced Sp3 may be substituted by the new GC-binding factor that appears to be induced by hypoxia (Fig. 4C). We have no information on this new complex, except that it is a GC-binding factor with a similar migration rate to Sp3, it accumulates under hypoxia, and it is not recognized by Sp1 or Sp3 antibodies.
Sp3 down-regulation by hypoxia appears to be by a posttranscriptional pathway since Sp3 transcript levels did not change. Hypoxia-mediated degradation, post-translational modifications, or translational mechanisms may be involved. Interestingly, a post-translational mechanism involving ubiquitin-mediated degradation has recently been described for the regulation of HIF-1 by hypoxia (35). Previous studies suggest that the repressor function of Sp3 is stronger when there are multiple Sp1/Sp3 binding sites (16,19). Therefore, positive responses to hypoxia and the regulation of gene expression by this pathway may be selective for a subset of genes with multiple GC-rich promoter elements.
Results presented here suggest that mechanisms other than, or in addition to, HIF-1 regulate the expression of glycolytic enzyme genes by hypoxia. Functional HIF-1 binding sites have been described in LDH-A, phosphoglycerate kinase, enolase-1, and aldolase-A genes, and mutational analyses as well as studies of cells from HIF-1 knockout mice support the involvement of HIF-1 in the regulation of these genes (6,10,36). Our GenBank data base analyses revealed 2 HIF-1 binding sites in the first intron of the PK-M gene but no sites in 7194 bp of the human ␤-enolase gene. HIF-1-regulated genes are induced by both hypoxia and transition metals (5), although these functions may be controlled by separate regulatory elements in other genes. GenBank screens also revealed multiple metal response element binding sequences (TGCACT) in both PK-M and ␤-enolase gene promoters (not shown). These elements bind the metal response factor MTF-1 that mediates positive transcriptional responses to metals (37)(38)(39)(40). For genes that do not contain HIF-1 sites, separate MREs and HREs could account for the dual regulation by hypoxia and metals (41).
Finally, although contributions of Sp-binding sites to the hypoxia response have not been reported previously in mammalian cells, anaerobic response elements have been described in the maize alcohol dehydrogenase, LDH, and aldolase gene promoters (11,42,43). The anaerobic response element in the alcohol dehydrogenase promoter contains a GC-rich element that constitutively binds Sp1 and is essential for activation by hypoxia (12,13). Therefore, hypoxia-mediated regulation of gene expression through the modulation of GC-binding factors may be an ancient pathway that has been conserved in plant and mammalian glycolytic enzyme genes.