Repression of α-Fetoprotein Gene Expression under Hypoxic Conditions in Human Hepatoma Cells: Characterization of a Negative Hypoxia Response Element That Mediates Opposite Effects of Hypoxia Inducible Factor-1 and c-Myc¹ Free

Hypoxia, a low cellular oxygen tension, is an important regulator of gene expression in the liver both during normal life and under pathophysiological conditions. Evidence has been obtained recently showing a link between hypoxia and cirrhosis (1). Because hypoxia can influence the expression of genes that enhance malignancy (2), these results reflect the importance of viewing hypoxia as part of hepatocarcinogenesis.

It is now well established that the HIF-1³ regulates the transcription of several genes involved in oxygen homeostasis in response to reduced oxygenation. These include the genes encoding erythropoietin and VEGF (reviewed in Refs. 3, 4, 5, 6). HIF-1 is a heterodimeric basic-helix-loop-helix-PAS protein composed of the two subunits: the hypoxia response factor HIF-1α and the constitutively expressed HIF-1β, also known as ARNT. Mice homozygous for a targeted deletion in the gene encoding HIF-1α die around mid-gestation, mainly because of defective vascularization, heart malformations, and failure of the neural tube to close (7, 8). Similarly, mice lacking HIF-1β suffer from abnormal angiogenesis and responses to oxygen deprivation, underscoring the importance of HIF-1 as a critical hypoxia response factor (9).

Because the hypoxic induction of genes generally involves the binding of HIF-1 to its consensus DNA sequence RACGTGV (the so-called HRE), we postulated that the transcription of liver genes with HIF-1-like binding sites in their regulatory region should be modulated under hypoxic conditions. We therefore determined whether there was a consensus HRE in the regulatory region of 85 liver-specific genes regulated by the liver-enriched hepatocyte nuclear factor-1 (10). We found 35 genes, the transcription of which was likely to be controlled under hypoxic conditions because they have a putative HRE in their regulatory sequence. One gene was particularly striking, the afp gene. AFP is a M_r 70,000 specific oncofetal protein that is abundantly synthesized in the yolk sac, the liver, and the gut of the fetus but not in normal adult tissues (reviewed in Ref. 11). Its synthesis can resume during liver regeneration and in hepatocellular carcinomas. This protein is not only a marker for cancer and fetal disorders. AFP has multiple pleiotropic activities affecting cell differentiation, growth regulation, and tumorigenesis. Although the precise function of AFP is not fully understood, it is a carrier/transport molecule for several ligands, including polyunsaturated fatty acids. It might play an indirect or direct role in apoptosis, has immunoregulatory functions, and acts as an antioxidant. The afp gene is also a good model for studying molecular mechanisms that govern gene expression in development and carcinogenesis. It has been studied extensively by several groups, including ours (reviewed in Refs. 12, 13, 14, 15; see also Refs. 16, 17). The expression of the afp gene is mainly regulated at the transcriptional level. In rat and mice, the afp gene is under control of a promoter and three enhancers that are preferentially active in the liver cells.

We have now shown that expression of the afp gene is specifically down-regulated, less mRNA and less protein, in HepG2 human hepatoma cells when they are cultured under hypoxic conditions. Transcriptional repression of the afp gene in response to hypoxia is mediated by a 480-bp fragment located ∼3.6 kb upstream of the transcription initiation site in the rat. A DNA sequence within this fragment, which resembles the binding site for HIF-1, was shown to be involved in down-regulation triggered by hypoxia. We have also demonstrated that this sequence is a specific target for a balanced mechanism in which HIF-1 down-regulates and c-Myc up-regulates afp gene transcription.

Human hepatoma HepG2 cells (American Type Culture Collection HB 8065) were cultured as monolayers at 37°C in a well-humidified 95% air/5% CO₂ incubator. The culture medium used was a 1:1 mixture of DMEM and Ham-F12 medium with Glutamax-I, supplemented with 10% fetal bovine serum, 100 μg/ml gentamicin, and 2.5 μg/ml Fungizone (all from Life Technologies, Inc.). HepG2 cells were 70–80% confluent before treatments.

Hypoxic treatment was performed on HepG2 cells plated on 60-mm-diameter glass dishes that were placed in stainless steel hypoxia chambers. The chambers were evacuated until reaching the desired oxygen partial pressures deduced by measuring the pressure using an electronic pressure sensor (SMC digital pressure sensor, Tokyo, Japan) and then refilled with a gas mixture containing 5% CO₂ and 95% N₂. Final oxygen concentrations of 10, 2, or 0.1% were used. The chambers were then placed in an incubator at 37°C for 24 h. The cells were still viable under these conditions (trypan blue exclusion, 98% viability after 24 h at 0.1% O₂).

Total RNA was extracted from cultured cells with the “RNA plus” kit (Quantum Appligene, Illkirch, France), based on the protocol of Chomczynski and Sacchi (18). Total RNA (5–10 μg/lane) was fractionated by electrophoresis through 1% agarose gel containing 0.7 m formaldehyde, transferred to a nylon membrane filter (Hybond-N membrane; Amersham, Orsay, France), and fixed by exposure to UV light. The integrity of the blotted RNA was evaluated by the methylene blue staining of the 28S and 18S rRNA subunits.

Total RNA (2 μg) from HepG2 cells was added to a 20-μl reverse transcription-PCR reaction containing 20 units of RNasin (Promega), 200 units of Moloney murine leukemia virus reverse transcriptase (Life Technologies, Inc.), 10 mm DTT (Life Technologies, Inc.), 0.5 mm of each deoxynucleotide triphosphate, 12.5 ng/μl pd(T)12–18 primers (Pharmacia), and 1× reverse transcription buffer (Life Technologies, Inc.) and incubated at 37°C for 1 h. To use a fragment of the human afp cDNA as probe, afp primers were used for a PCR (25 cycles, 94°C/48°C/72°C). The sequences of the afp primers were 5′-AAATACATCCAGGAGAGCCA-3′ (sense strand) and 5′-CTGAGCTTGGCACAGATCCT-3′ (antisense strand). We used the EcoRI/ClaI fragment of plasmid pMycECl, which contains part of the human c-Myc cDNA (a kind gift from Dr. O. Brison, Institut Gustave Roussy, Villejuif, France). The probes for Glut-1 and 18S, used as controls, were provided by Prof. J. Girard (Centre National de la Recherche Scientifique, Hôpital Cochin, Paris, France). Total RNA from HepG2 cells that had been exposed to hypoxia was also hybridized with a human VEGF₁₆₅ cDNA probe (19) as a control of the stimulation of the vegf gene under hypoxia.

All of the cDNA probes were labeled with [α-³²P]dATP using the Priming It II random labeling kit (Stratagene). The synthetic 21-mer oligonucleotide used to reveal the 18S ribosomal subunit was labeled using [γ-³²P]ATP (3000 Ci/mmol) and T4 polynucleotide kinase (New England Biolabs, Inc.).

The RNA blots were prehybridized for 20 min at 68°C in Quick-Hyb buffer (Promega) following the manufacturer’s recommendations and then with the ³²P labeled probe for 2 h at 68°C. The hybridized filters were washed twice at room temperature for 15 min each with 2× SSC, 0.1% SDS and once for 30 min with 0.1× SSC, 0.1% SDS at 65°C. Radioactive signals were detected by exposing the filters to Hyperfilms MP (Amersham, Orsay, France). Several probes were used on the membranes in succession; the previous probe was removed by dipping the membrane in 0.1× SSC, 0.1% SDS solution at 100°C and allowing it to cool to room temperature.

Quantitative measurements were obtained by measuring the intensity of the radioactive signals with an Instant Imager (Packard). They were usually normalized to the signal for 18S rRNA.

HepG2 cells were grown on culture plates containing 5 ml of medium. Medium that had been in contact with cells for 24 h at 20, 10, 2, or 0.1% O₂ was removed, cooled to 2°C, and assayed for AFP. The amounts of endogenous AFP protein were measured by rinsing the cells on the plate twice with PBS, scraping them off, and suspending them in 200-μl Reporter Lysis buffer (Promega). Protein was determined by the Bradford method with bovine immunoglobulin (Bio-Rad) as standard (20). The amounts of AFP were determined by microparticule enzyme immunoassay (AFP AXSYM Abbott, Abbott Park, IL).

Cells (70–80% confluent) were lysed in 2× Laemmli sample buffer. The cell protein extracts were electrophoresed through 12.5% SDS-polyacrylamide gels and transferred by semidry blotting to nitrocellulose membranes (Amersham) using standard procedures. The membranes were stained with Ponceau red (Sigma Chemical Co.) to confirm equal protein loading and transfer. They were then incubated overnight at 4°C with blocking solution (PBS, 0.1% Tween 20, and 5% nonfat milk), followed by specific antibodies to HIF-1α (1:1000 dilution; kindly provided by Dr. C. Bradfield, McArdle Laboratory for Cancer Research, Madison, WI) or c-Myc (1:100 dilution; kindly provided by Dr. A. Durrbach, Equìpe en Restructuration 1984, Villejuif, France) in blocking solution. The membranes were treated with horseradish peroxidase-coupled secondary antibody (1:5000, in blocking solution; Jackson Immunoresearch Laboratories, Inc.) and ECL+Plus substrate (Amersham) was used for detection. The intensity of the bands on the film (Hyperfilm; Amersham) was determined by scanning the film and using computer software (NIH-Image).

The parent plasmid pAFP-CAT containing the CAT reporter gene under the control of the −324 to +6 fragment of the rat afp promoter has been described previously (21). The construction of plasmid pPO1,2,3-AFP-CAT, which contains the whole rat afp 7-kb regulatory region, has also been reported (21). Conventional cloning procedures were used to insert the −7040 to −3849 (EcoRI-HindIII) fragment, the −3849 to −2531 (HindIII-HindIII) fragment, or the −2531 to −323 (HindIII-HindIII) fragment of the rat afp gene 5′ extragenic region into the HindIII site of pAFP-CAT, yielding pPO1-AFP-CAT, pPO2-AFP-CAT, and pPO3-AFP-CAT, respectively. Plasmids pPO21-AFP-CAT and pPO22-AFP-CAT were obtained by subcloning the blunted −3849 to −3369 (HindIII-SacI) and −4261 to −3672 (SpeI-SacI) fragments into the blunted PstI site of the plasmid pUC19. Both SalI-HindIII fragments were then cloned into the SalI-HindIII sites of the plasmid pAFP-CAT.

Mutation of the afp/HRE sequence in plasmid pPO21-AFP-CAT gave the plasmid pPO21mut AFP-CAT in which the CG in the CACGTG motif was replaced by an A. It was obtained by site-directed mutagenesis using the QuickChange Site-Directed Mutagenesis kit from Stratagene as indicated by the manufacturer. The point mutation was verified by sequencing.

Plasmid pVEGF-LUC contains a 385-bp deletion fragment, derived from the 1511-bp fragment from the VEGF promoter, inserted into pGL₂ basic (19, 22).

Plasmid containing HIF-1α coding sequence was generously provided by Dr. G. Semenza (Johns Hopkins University School of Medicine, Baltimore, MD). Plasmid pHcmyc, which contains the whole human c-myc gene under the control of its own regulatory sequence, was generously provided by Dr. C. Cremisi (Institut de Recherche sus le Cancer, Villejuif, France).

Transfection used 5 μg/dish reporter plasmid. HepG2 cells were transfected using the calcium phosphate method (21). Two or 5 μg of HIF-1α or c-Myc expression vectors were cotransfected with 5 μg of pPO21-AFP-CAT, pPO21mut-AFP-CAT, or 385-bp VEGF-LUC plasmids. The total amount of transfected DNA was kept at 10 μg by adding pBluescript plasmid. The plates were washed twice in PBS and then incubated in complete medium under 5 or 20% O₂ for 24 h. Cells were then washed twice with cold PBS and CAT or LUC activity was assayed (21). CAT and LUC activities were measured and normalized to the protein content.

We found that the regulatory region of the rat afp gene contains a sequence, at −3625 to −3618, which is very similar to the HIF-1 binding sites in the genes encoding human aldolase A (23), human VEGF (24), human/rat/mouse insulin-like growth factor binding protein-I (25) and mouse heme oxygenase (26) shown in Table 1. This afp motif is an HIF-1 consensus binding site with the core sequence ACGTG. We therefore determined whether expression of the afp gene was altered under hypoxia.

To examine the effect of hypoxia on afp gene expression, human HepG2 hepatoma cells were exposed to various levels of O₂ (20, 10, 2, and 0.1%) for 24 h. We then measured the steady-state concentrations of afp mRNA using the Northern procedure. Hypoxia inhibited afp mRNA levels in a dose-dependent manner. Culturing HepG2 cells under 2 and 0.1% O₂ reduced the steady-state concentrations of afp mRNA to 50 and 33% of the controls (20% O₂), respectively, whereas 10% O₂ did not significantly alter afp mRNA (Fig. 1,A). Conversely, the vegf mRNA amounts were increased (up to 5-fold) in HepG2 cells exposed to hypoxia (Fig. 1 A), indicating the specificity of the down-regulatory effect of hypoxia on afp gene expression.

We then examined the time course of the hypoxic action on afp mRNA concentration using HepG2 cells cultured under 0.1% O₂ for 6, 12, and 24 h. The amount of afp mRNA decreased with time. The concentration of afp mRNA began to decrease after 12 h incubation under 0.1% O₂ and remained low until the end of the period examined (24 h; Fig. 1,B). The amount of afp mRNA was reduced by ∼25% after 12 h of hypoxia and by ∼50% after 24 h of hypoxia. By opposition, the amounts of vegf and glut-1 mRNA increased throughout the culture under hypoxic conditions (by 4- and 3-fold, respectively, at 24 h; Fig. 1 B). This further demonstrated that hypoxia specifically inhibited afp gene expression and also confirmed the viability of the cells.

We determined whether the amount of AFP protein was reduced by hypoxic treatment using a microparticle enzyme immunoassay to quantify intracellular and secreted AFP protein. Exposure to hypoxia resulted in a 61% ±13 (2% O₂) and 73% ±4 (0.1% O₂) reduction in the intracellular concentrations of AFP protein in the whole cell lysate after 24 h (Fig. 2,A). Hypoxia also decreased the amounts of AFP protein secreted into the culture medium in a concentration-dependent manner. They were 2.9 ± 1.1 μg (2% O₂) and 2.3 ± 1 μg (0.1% O₂) compared with 8.5 ± 3.4 μg with normoxia (Fig. 2,A). But Western blotting showed that the amount of HIF-1α protein increased markedly in HepG2 cells cultured for 24 h at 0.1% O₂ (Fig. 2 B). These results showed the specificity of the concentration-dependent inhibition of AFP protein amount in HepG2 cells by hypoxia.

Experiments using inhibitors of protein and RNA synthesis indicated that newly synthesized proteins are necessary to mediate the decrease in afp mRNA by hypoxia, and that stability of the afp mRNA is not affected under hypoxic conditions (data not shown). Taken together, these results strongly suggest that the drop in afp mRNA in response to hypoxia is attributable mainly to a decrease in the rate of afp gene transcription.

We used transient transfection to analyze further the effect of hypoxia on the transcriptional activity of the whole rat afp regulatory region and on some of its subfragments. We measured the activities of plasmids bearing the CAT gene under the control of the whole 7-kb regulatory region of the rat afp gene (pPO1,2,3-AFP-CAT) or individual regions (PO1, PO2, and PO3) in front of the afp promoter in HepG2 cells cultured under normoxic and hypoxic conditions. Hypoxia inhibited the activity of plasmid pPO1,2,3-AFP-CAT, which contains the full rat afp regulatory region, to 28% of its value under normoxic conditions (Fig. 3). It also inhibited that of the plasmid pPO2-AFP-CAT, which contains the region from −3849 to −2513 in which lies the putative HRE, to 16% of its value under control conditions. The activities of plasmids pPO1-AFP-CAT and pPO3-AFP-CAT, which contain regions from −7040 to −3849 and from −2513 to −323, respectively, were not reduced significantly by hypoxia (data not shown).

We divided the PO2 region into three regions, PO21, PO22, and PO23, to better define the HRE. These were cloned in front of the afp promoter. The activity of plasmid pPO21-AFP-CAT, which contains the putative HRE, was drastically affected by hypoxia. It dropped to 15.4% of its control value (Fig. 3). The activities of the plasmids containing the adjacent region PO22 in front of the afp promoter or the afp promoter alone (pAFP-CAT) were much less affected (48.5 and 60% of their control values; Fig. 3). These experiments indicate that the PO21 region, from −3849 to −3369, mediates the down-regulatory effect of hypoxia on afp transcription. This effect is specific because the activity of a LUC plasmid bearing the 385-bp of the vegf gene containing the HIF-1 response element (19) resulted in a 10.8-fold induction when HepG2 cells were cultured under hypoxia.

To check whether the putative afp/HRE, located at −3625 to −3619, is the target of repression by hypoxia, we introduced a point mutation (5′-CAATGGC-3′ instead of CACGTGGC) in this sequence in plasmid pPO21mut-AFP-CAT. Previous studies, including ours, have demonstrated that similar mutations within the HIF-1 binding site of the hypoxia-responsive EPO enhancer abolish both the binding and transcriptional function of HIF-1 (19, 22, 26). The activity of plasmid pPO21mut-AFP-CAT, which bears the mutation, was much less affected by hypoxia than that of plasmid pPO21-AFP-CAT containing the wild-type sequence (Fig. 3). This shows that the sequence covering the consensus HIF-1 binding site is involved in the down-regulation of afp gene transcription by hypoxia.

HepG2 cells were transiently cotransfected with HIF-1α expression vector and CAT reporter plasmids containing either the wild-type afp/HRE (pPO21-AFP-CAT) or the mutated afp/HRE (pPO21mut-AFP-CAT) sequences to determine whether the afp/HRE sequence, which mediates the down-regulation of afp gene expression under hypoxia, is a target for HIF-1. The activity of pPO21-AFP-CAT was repressed to 35% of its control value by overproduction of HIF-1α, whereas that of pPO21mut-AFP-CAT was unaffected (Fig. 4 A). In contrast, the activity of the LUC reporter plasmid bearing the 385-bp VEGF fragment was increased 12-fold. These results strongly suggest that HIF-1 is implicated in the repression of the afp gene under hypoxia. The involvement of HIF-1 in the repression of afp gene expression under hypoxia is also in agreement with the fact that the down-regulation of afp mRNA is blocked by cycloheximide (data not shown) and thus requires the synthesis of protein.

The activity of plasmid pPO21mut-AFP-CAT was 30–50% lower than that of plasmid pPO21-AFP-CAT (data not shown) under basal conditions. This suggests that the mutation in the afp/HRE prevented the action of a stimulatory transcription factor targeted at this element. The sequence of the afp/HRE is similar to that of the c-Myc consensus binding site 5′-CACGTG-3′ (Ref. 27; Fig. 4 B).

HepG2 cells were transiently cotransfected with plasmids pPO21-AFP-CAT and pPO21m-AFP-CAT with or without a c-Myc expression vector to determine whether the afp/HRE is a functional c-Myc site. Overproduction of c-Myc caused a reproducible 3.5-fold increase in the activity of the plasmid pPO21-AFP-CAT, whereas its effect was much less (1.5-fold induction) on the mutated plasmid pPO21m-AFP-CAT (Fig. 4 C). c-Myc overexpression had no effect on the LUC plasmid bearing the 385-bp of the vegf gene containing the HIF-1 response element. These results strongly suggest that c-Myc proteins specifically activate afp gene transcription through the afp/HRE sequence. We therefore postulate that HIF-1 and c-Myc could functionally compete to modulate the activity of the PO21 fragment.

These observations prompted us to test the effect of hypoxia on the expression of the c-myc gene by measuring mRNA and protein concentrations. HepG2 cells were exposed to various levels of O₂ (20, 10, 2, and 0.1%) for 24 h (Fig. 5 A). There was a marked decrease in the steady-state concentration of c-myc mRNA at 2 and 0.1% O₂. The amounts of c-myc mRNA began to decrease after 6 h of incubation under 0.1% O₂ (data not shown).

Western blot experiments were performed to determine whether the amount of c-Myc protein was also reduced by hypoxia. HepG2 cells cultured under low oxygen conditions contained almost no c-Myc protein (Fig. 5,B). We used cycloheximide to examine the requirement for ongoing protein production for the decrease in c-myc mRNA by hypoxia. Cycloheximide had no effect on the amount of c-myc mRNA in hypoxic cells, indicating that de novo synthesis of protein is not required (Fig. 5 C). Hence, the hypoxic repression of c-myc gene expression could be independent of HIF-1.

We used the wild-type hepatoma cell line Hepa1 c1c7 and its mutant derivative Hepa1 c4 mouse hepatoma cells to further investigate the role of HIF-1 in the repression of c-myc gene expression by hypoxia. Hepa1 c4 cells lack HIF-1β (28) and consequently cannot form HIF-1 heterodimer susceptible to bind DNA (29). These cells were cultured under 20 or 0.1% O₂ for 24 h, and the concentrations of c-myc and glut-1 mRNAs were analyzed by Northern blotting. Hypoxia repressed c-myc gene expression in both the parental Hepa1 c1c7 and deficient HIF-1-β Hepa1 c4 cell lines (Fig. 5 D). As expected, the amount of glut-1 mRNA increased in the parental Hepa1 c1c7 cell line, but hypoxia had no effect on this mRNA in the Hepa1 c4 cell line (data not shown). These results show that HIF-1, critically involved in the hypoxic responsiveness of glut-1, was lacking in the Hepa1 c4 cell line (5). They further suggest that HIF-1 is not involved in the mechanism by which hypoxia repressed c-myc gene expression.

Hypoxia is an essential developmental and physiological stimulus but is also a component of many disorders, including heart attack, stroke, and cancer (7). Liver cells are physiologically exposed to different oxygen tensions and are more exposed to hypoxia under pathological conditions, as shown recently for cirrhosis (1). The HIF-1 transcription factor is most important in regulation by hypoxia and was originally defined by its capacity to bind to a site required for the hypoxic induction of transcription of many genes (reviewed in Refs. 3, 4, 5, 6).

We postulated that hepatic genes containing putative HIF-1 binding sites in their regulatory regions are potential target genes for hypoxia. A computer search for HIF-1 binding sites in regulatory regions revealed 35 liver genes with such a site and have examined the sensitivity of the afp gene to hypoxia.

Our results indicate that the expression of the afp gene in the HepG2 human hepatoma cells is specifically down-regulated under hypoxia and that HIF-1 is involved in this process by decreasing the activity of the afp gene regulatory region. The target for both the negative regulatory effect of hypoxia and HIF-1 on afp gene transcription was located between −3625 and −3618 in the rat afp gene regulatory region. An interesting feature of the negative hypoxia response element of the afp gene is the presence of two possible HIF-1 consensus binding sites: a CACGTGGG putative binding site on the sense strand from −3625 to −3618, and a similar CACGTGGC sequence on the antisense strand from −3620 to −3627. This architecture has not yet been found in any other HRE. This core sequence 5′-CGCACGTGGG-3′ conforms well with the published consensus HRE [5′-(C/G/T)ACGT(G/C)C(G/T)-3′; Ref. 30)] or BACGTGSK, where B is C or G or T, S is C or G, and K is G or T (23). However, a C has been found at the last position in genes encoding aldolase A (31), enolase 1 (31), and phosphoglycerate kinase 1 (32). Our mutational analysis focused on this regulatory region of the afp gene and confirmed that this sequence is necessary for repression by hypoxia and is a negative functional target for HIF-1. Thus, the afp/HRE site is essential for hypoxic repressibility and the HIF-1 transcription factor is involved in this process.

It is reasonable to think that HIF-1 binds to the negative afp HRE with a low affinity because, using gel shift assays with different conditions, we did not observe the formation of a stable complex between HIF-1 from nuclear extracts of hypoxic HepG2 cells and a ³²P-labeled oligonucleotide containing the afp HRE (data not shown). However, the oligonucleotide containing the afp/HRE competed with the oligonucleotide covering the vegf/HIF-1 binding site, used as a probe, in the formation of the specific HIF-1/DNA complex (data not shown). This displacement was specific, because there was no longer competition with the mutated afp/HRE (data not shown). It may also well be that an accessory protein at a neighboring site, or that a specific chromatin structure, is required to recruit HIF-1 to this negative HIF-1 response element.

Several transcription factors can act as either activators or repressors, depending on protein-protein interactions dictated by the promoter architecture and the physiological context (33, 34). Our data provide the first evidence that HIF-1 may also act as both an activator and a repressor of transcription. Although HIF-1α is known to be one of the binding partners of CBP/p300, a transcriptional coactivator (4), the mechanism by which hypoxia down-regulates gene expression is not known. Future studies might investigate whether corepressors of transcription play a role in down-regulating the transcription of some genes. It also remains to be elucidated how the interaction of HIF-1 with corepressors is regulated in the context of hypoxia.

The GCCACGTGGG sequence of the HRE in the afp gene regulatory region is of particular interest, because it could be identified as a functional motif for c-Myc. The nucleotide composition of the sequences flanking the CACGTG core influences the DNA binding affinity of the Myc/Max/Mad network and of several other transcriptional regulators that are also known to bind to the Myc E-box core sequence (35, 36). Myc target genes show a preference for 5′ GC, 5′ CG, or 5′ AG immediately preceding the core sequence. Several lines of evidence indicate that Myc proteins regulate many target genes by binding to E-box DNA elements located in control regions of the respective genes (37). Our data indicate that the afp gene belongs to this growing list of genes controlled by c-Myc. These results are in agreement with recent work using antisense oligodeoxynucleotide against c-myc mRNA transferred into human hepatoma cells (38). The authors found a reduction in c-Myc protein, followed by inhibition of cell proliferation and decreased afp gene expression. The fact that the afp regulatory region could be specifically activated by the overexpressed c-Myc (Fig. 4 D) strongly suggests that afp gene expression is partially controlled by this oncogenic protein. It is in agreement with the fact that c-myc expression is up-regulated very early in the liver regeneration process (39, 40), preceding afp reexpression (41).

Our finding that the region from −3625 and −3619 is a target for two transcription factors with opposite effects led us to think that a balance in the relative concentrations of these proteins is involved in controlling the activity of this afp regulatory region under hypoxic conditions. The concentrations of c-Myc mRNA and protein are down-regulated under hypoxic conditions, whereas the production of HIF-1 protein is increased. This would favor the repression of afp gene expression by HIF-1. Our working hypothesis concerning the dual HIF-1/c-Myc regulation of afp gene expression is summarized in Fig. 6. A large amount of c-Myc regulates the transcription of the afp gene under normoxic conditions, when human HepG2 hepatoma cells are proliferating. However, the disappearance of c-Myc and stabilization of HIF-1α protein under hypoxic conditions, when most of the cells are proliferating slowly or not at all (42), allows the HIF-1 complex to block afp gene transcription (Fig. 6). The negative afp/HRE thus provides an excellent model with which to examine the interplay between c-Myc and HIF-1 in the negative regulation of gene expression under hypoxia.

Our study with Hepa1 c4 cells, which do not produce functional HIF-1β and have no HIF-1 DNA binding activity, strongly suggested that HIF-1 is not necessary for the repression of c-myc gene expression. Thus, our data indicate that there are at least two mechanisms by which hypoxia can repress gene expression, HIF-1 dependent (afp mechanism) and HIF-1 independent (c-myc mechanism) repression pathways.

There seems to be an inverse relationship between hypoxia and cell proliferation. Hypoxia alone can cause cell cycle arrest in cultured cells without affecting viability, at least initially, even when the oxygen is very low (0.1–2% O₂; Ref. 42). This arrest is generally in the late G₁ and early S phases. Our working model, using HepG2 hepatoma cells, fits perfectly with these observations. Viable cells proliferate slowly or are resting under our hypoxic conditions. Most (95–98%) of HepG2 cells were viable after 24 h at 0.1% O₂. Preliminary experiments also showed an accumulation of HepG2 cells in the G₁-S phases (data not shown). The expression of c-myc is correlated with growth in normally growing adult tissues, and evidence indicates that c-Myc moves cells from the G₁ to S-phase (43). The down-regulation of c-myc expression under extremely low oxygen concentrations may modify the regulation of transcription of many of Myc target genes, especially those involved in cell cycle/growth arrest. For example, c-Myc represses gadd45 expression (37), which can cause cell cycle arrest when overexpressed. Whether the down-regulation of c-myc expression affects the arrest or simply results from the arrest is unknown.

In conclusion, our investigations of afp gene expression in the HepG2 human hepatoma cells have provided several insights into how hypoxia mediates gene repression. They suggest that HIF-1 can be associated with both up- and down-regulation under hypoxic conditions. Further investigations on the nature of the partners of HIF-1 involved in down-regulation are crucial to elucidate the mechanisms involved in the repression of afp and other oxygen-regulated genes. Our findings also point to a novel “time share” mechanism using a E-box consensus sequence occupied by two different tenants, HIF-1 and c-Myc, the relative concentrations of which vary in response to hypoxia. This opens the way for potentially fruitful future research.

We are grateful to the researchers already cited in “Materials and Methods” for their generous gifts of cell lines, antibodies, and plasmids and to Prof. O. Parkes for checking our English. We are also grateful to Dr. M. Pontoglio (Institut Pasteur, Paris, France) for supplying us with the list of hepatocyte nuclear factor-1 target genes.