The role of the aryl hydrocarbon receptor nuclear translocator (ARNT) in hypoxic induction of gene expression. Studies in ARNT-deficient cells.

Hypoxia-inducible factor-1 (HIF-1), a DNA-binding complex implicated in the regulation of gene expression by oxygen, has been shown to consist of a heterodimer of two basic helix-loop-helix Per-AHR-ARNT-Sim (PAS) proteins, HIF-1alpha, and HIF-1beta. One partner, HIF-1beta, had been recognized previously as the aryl hydrocarbon receptor nuclear translocator (ARNT), an essential component of the xenobiotic response. In the present work, ARNT-deficient mutant cells, originally derived from the mouse hepatoma line Hepa1c1c7, have been used to analyze the role of ARNT/HIF-1beta in oxygen-regulated gene expression. Two stimuli were examined: hypoxia itself and desferrioxamine, an iron-chelating agent that also activates HIF-1. Induction of the DNA binding and transcriptional activity of HIF-1 was absent in the mutant cells, indicating an essential role for ARNT/HIF-1beta. Analysis of deleted ARNT/HIF-1beta genes indicated that the basic, helix-loop-helix, and PAS domains, but not the amino or carboxyl termini, were necessary for function in the response to hypoxia. Comparison of gene expression in wild type and mutant cells demonstrated the critical importance of ARNT/HIF-1beta in the hypoxic induction of a wide variety of genes. Nevertheless, for some genes a reduced response to hypoxia and desferrioxamine persisted in these mutant cells, clearly distinguishing ARNT/HIF-1beta-dependent and ARNT/HIF-1beta-independent mechanisms of gene activation by both these stimuli.

Oxygen is an important regulator of gene expression in many organisms. In mammalian cells a number of different transcription factors can be induced by hypoxia or related metabolic changes, and may play a role in oxygen-regulated gene expression. One important system has been identified through studies of the regulation of erythropoietin, a hormone that controls red cell production in accordance with blood oxygen availability (1). The oxygen-regulated function of the erythropoietin 3Ј en-hancer in hepatoma cells was found to be dependent on binding of an inducible factor termed hypoxia-inducible factor-1 (HIF-1) 1 (2). Subsequently, transient transfection studies have demonstrated this activity in non-erythropoietin producing cells (3)(4)(5) and have now implicated HIF-1 binding sites in the regulation of a number of hypoxically inducible genes. Examples include genes encoding glycolytic enzymes (6 -8), glucose transporters (9), vascular endothelial growth factor (10), tyrosine hydroxylase, a key enzyme in catecholamine biosynthesis (11), and nitric oxide synthase (12). These hypoxic responses can be mimicked by particular transition metals including cobalt (13), and iron-chelating agents such as desferrioxamine (14,15), findings that have led to the proposal that these agents interact closely with the mechanism of oxygen sensing.
Following affinity purification, HIF-1 has recently been cloned and shown to be a heterodimer of two basic helix-loophelix PAS proteins termed HIF-1␣ and HIF-1␤ (16,17). HIF-1␤ was identical to the previously identified aryl hydrocarbon receptor nuclear translocator (ARNT), a molecule essential for the transcriptional response to certain environmental hydrocarbons known as the xenobiotic response (for recent review, see Ref. 18). In this capacity ARNT/HIF1␤ forms a heterodimeric complex with another basic helix-loop-helix PAS protein, the ligand-binding subunit termed the aryl hydrocarbon receptor (AHR) (19,20). This complex binds the xenobioticresponsive element (21,22), a control sequence for genes such as the CYP1A1 gene, encoding a cytochrome P450 with aryl hydrocarbon hydroxylase activity that can convert aryl hydrocarbons to toxic or carcinogenic metabolites. An important step in the analysis of this mechanism of gene regulation was the derivation of mutant cells selected by survival in the presence of benzo(a)pyrene (23). One group of mutants was shown to be defective in the nuclear translocation of the ligand-binding complex, as assessed by conventional subcellular fractionation procedures (24), and provided the basis for cloning of the gene encoding ARNT by complementation (25). These ARNT-deficient (ARNT Ϫ ) cells lack ARNT/HIF-1␤ protein expression and function (26) and provide an opportunity to determine the importance of ARNT/HIF-1␤ in other aspects of gene regulation.
Here we describe the effects of this phenotype on hypoxic gene regulation. The ARNT Ϫ and wild type Hepa-1 cells were compared with respect to the induction of HIF-1 DNA binding activity, expression of transfected plasmids bearing oxygenregulated cis-acting elements, and the induction of endogenous genes by hypoxia. Deletion mutants of ARNT/HIF-1␤ were analyzed to determine the domains that are critical for the functional activity of HIF-1.

EXPERIMENTAL PROCEDURES
Cell Lines and Culture-The murine hepatoma line Hepa-1 clone Hepa1c1c7 (hereafter referred to as Hepa-1) and derivatives c4, c39, vT{2}, Rc4 and c31 have been described previously (23,25,27,28). Briefly, the c4 mutant clone was obtained from a mutagenized culture of Hepa-1 cells that was selected for loss of aryl hydrocarbon hydroxylase activity in a single-step maneuver by survival in the presence of benzo(a)pyrene (23). c4 lacks ARNT/HIF-1␤ function (29) and also fails to express the ARNT/HIF-1␤ protein, as assessed by Western blot analysis using an antibody to the carboxyl-terminal half of the protein (26). c39 is a similarly but independently derived clone from the same complementation group. Both c4 and c39 express ARNT mRNA of normal size (25). vT{2} cells are c4T cells (a hypoxanthine phosphoribosyl transferase-deficient derivative of c4) stably transfected with pBM5/Neo-M1-1 containing a 2.5-kilobase pair insert of ARNT/HIF-1␤ cDNA and then selected for reacquired aryl hydrocarbon hydroxylase activity (25). Rc4 is a revertant line, derived from c4 by the same "reverse selection" procedure, which possesses wild type levels of ARNT activity (27). c31 is a mutant selected in the same way as c4 and c39, which expresses a dominantly acting repressor preventing transcriptional activation of the xenobiotic response (28).
All cells were grown in minimal essential medium-␣ without nucleosides (Life Technologies, Inc.) supplemented with 10% fetal calf serum (Globepharm), L-glutamine (2 mM), penicillin (50 IU/ml), and streptomycin sulfate (50 g/ml). For vT{2} cells, 400 g/ml G418 (Life Technologies, Inc.) was added to this medium. In all cases, experiments on derivative clones were conducted in parallel with identical assays on the wild type Hepa-1. Studies of the regulation of endogenous genes were performed on cells approaching confluence and incubated in humidified air with 5% CO 2 (normoxia), or exposed to hypoxia or 100 M desferrioxamine mesylate (Sigma). Hypoxic conditions were generated in a Napco 7001 incubator (Precision Scientific, Chicago, IL) with 1% oxygen, 5% CO 2 , and 94% N 2 .
Plasmids-p(Epo1-25) 3 SV40GH contained 3 copies of a 25-base pair sequence containing the HIF-1 site from the mouse erythropoietin 3Ј enhancer adjacent to the SV40 promoter (a 202-base pair NsiI-HindIII fragment) linked to growth hormone (31). p(PGK24) 3 TKGH contained 3 copies of the oligonucleotide P24, containing the HIF-1 site from the mouse phosphoglycerate kinase-1 5Ј enhancer located 10 base pairs 5Ј to the TATA box of the herpes simplex virus thymidine kinase promoter linked to growth hormone (6). Plasmid pBS␣ Ϫ contained a full-length ␣ 1 -globin gene with 1.4 kilobase pairs of 5Ј flanking sequence. Plasmid pcDNA1/Neo/mARNT and mutant derivatives were based on the pcDNA1/Neo vector (Invitrogen), which contains the cytomegalovirus early gene enhancer/promoter and were as described in Ref. 32. The deletions tested were as follows: ⌬b, lacking the basic region (amino acids 86 -102); ⌬HLH, lacking the helix-loop-helix domain (amino acids 103-142); ⌬AB, lacking the PAS domain (both A and B parts) (amino acids 173-458); and a truncated ARNT, bHLHAB, missing both the carboxyl and amino termini and containing only amino acids 70 -474.
Since the cytoplasmic ␤-actin gene was not inducible in these experiments, it was used as an internal control for RNA extraction efficiency and sample processing. To accommodate the large difference in mRNA abundance between ␤-actin and the test genes, separate hybridizations of aliquots of the same RNA were set up in parallel with riboprobes to the test gene and ␤-actin. The samples were amalgamated after RNase digestion and thereafter processed and electrophoresed together. After quantitation of both species, the abundance of the test mRNA relative to ␤-actin mRNA was calculated. Hypoxic induction is the ratio of gene expression in hypoxia to that in normoxia.
To measure the expression of transfected plasmids, 10 g of RNA from transfected cells was assayed by simultaneous hybridization with probes for the growth hormone reporter and ␣ 1 -globin control as described (3). After quantitation, correction was made for transfection efficiency and sample processing using expression of ␣ 1 -globin. The ratio of the corrected growth hormone reporter gene expression in hypoxia to that in normoxia was calculated to determine the inducible response in each experiment. The statistical significance of differences in inducible responses between parallel experiments on Hepa-1 and c4 cells was tested using Student's paired t test with logarithmic transformation.

RESULTS
HIF-1 DNA Binding Activity-To determine whether ARNT/ HIF-1␤ was a necessary component of the HIF-1 DNA-binding complex, we compared HIF-1 activity in wild type Hepa-1 cells and the ARNT Ϫ derivative, c4. In the wild type cells, the inducible species, HIF-1, was clearly observed in cells after exposure to hypoxia (1% oxygen), whereas it was undetectable in hypoxic c4 cells. No new species were observed in the c4 cells, and no consistent changes were observed in the constitutive species which bind at this site. We also tested induction of HIF-1 by DFO. In wild type cells, HIF-1 was induced by DFO but had a slightly greater mobility than the complex induced by hypoxia. Again, this complex was absent in nuclear extract from c4 cells exposed to DFO. Similar results were obtained using two different HIF-1-binding oligonucleotides, derived from the mouse erythropoietin 3Ј enhancer (E24) and the mouse phosphoglycerate kinase-1 5Ј enhancer (P24) (Fig. 1).
Regulation of Hypoxia-responsive Elements-To analyze the role of ARNT/HIF-1␤ in the function of hypoxically responsive cis-acting elements, wild type and c4 cells were transfected with plasmids bearing such elements linked to reporter genes. Two plasmids were tested: p(Epo1-25) 3 SV40GH and and p(PGK24) 3 TKGH, which contain three copies of the HIF-1 site from the mouse erythropoietin 3Ј enhancer and from the mouse phosphoglycerate kinase-1 5Ј enhancer, respectively. Results are summarized in Table II. In normoxic Hepa-1 and c4 cells, the reporter gene expression was similar. In wild type Hepa-1 cells, each of the plasmids supported an approximately 6-fold induction of reporter gene by hypoxia. In the c4 cells, induction by hypoxia was virtually absent, although a low level of inducibility was observed occasionally using each of the reporter plasmids.
Further transfections were performed to compare the inducible responses to hypoxia and to DFO in wild type and mutant cells. Cells in these experiments were split into three aliquots after transfection with p(Epo1-25) 3 SV40GH; one aliquot was maintained in normoxia, one in hypoxia, and the third exposed to DFO (100 M). Results are shown in Table II. Wild type cells showed similar responses to hypoxia and DFO. In c4 cells, responses to both stimuli were severely reduced. The small and inconsistent response to DFO was similar to that observed with hypoxia.
Restoration of Hypoxia-inducible Responses by Transfection with ARNT Expression Plasmids-In order to determine whether the ARNT Ϫ c4 cells could regain inducible responses to hypoxia following expression of a functional ARNT/HIF-1␤ gene, cells were co-transfected with the mouse ARNT/HIF-1␤ expression plasmid pcDNA1/Neo/mARNT and the hypoxically inducible reporter plasmid p(Epo1-25) 3 SV40GH. After electroporation, cells were divided into two aliquots, which were incubated in normoxia for 24 h to allow expression of ARNT/ HIF-1␤ before one of the aliquots was exposed to hypoxia for 16 h. Cells were harvested a total of 40 h after transfection. Each experiment was controlled positively (wild type cells transfected with p(Epo1-25) 3 SV40GH alone) and negatively (c4 mutant cells transfected with p(Epo1-25) 3 SV40GH and the empty vector pcDNA1/Neo). In cells transfected with the ARNT/HIF-1␤ expression plasmid, a substantial inducible response to hypoxia was restored, which was not seen in cells co-transfected with pcDNA1/Neo; the inducible response in the complemented c4 cells was not quite as great as in wild type Hepa-1 cells, being 2.5-fold versus 4.2-fold in the parallel experiments (Fig. 2). However, under these experimental conditions, the normoxic level of reporter gene expression was slightly higher than in the wild type cells.
To determine which domains of ARNT/HIF-1␤ were required for functional restoration of the hypoxically inducible response, a series of deletional mutants of ARNT/HIF-1␤ was tested in this co-transfection system. Transfection of plasmids lacking either the basic region (⌬b), the helix-loop-helix domain (⌬HLH), or the PAS domain (⌬AB) all failed to restore function. In contrast, a plasmid containing only the basic, helix-loophelix, and PAS domains but lacking the carboxyl and amino termini (bHLHAB) restored hypoxic inducibility almost as effectively as the full coding sequence of ARNT/HIF-1␤ (Fig. 2).
Regulation of Endogenous Gene Expression by Hypoxia and DFO-Since ARNT/HIF-1␤ was necessary for the functional responses to hypoxia and DFO conveyed by transfected cisacting sequences which contained HIF-1 binding sites, we compared hypoxically inducible endogenous gene expression in the wild type Hepa-1 cells and the ARNT Ϫ c4 mutant cells using RNase protection assays for a variety of mouse genes. The results are summarized in Fig. 3. Normoxic and hypoxic Hepa-1 cells were assayed for erythropoietin mRNA expression, but no signal was detectable from 100 g of RNA.
Exposure of wild type cells to hypoxia (1% O 2 for 16 h) led to induction of mRNA for phosphoglycerate kinase-1 (PGK-1) and lactate dehydrogenase-A (LDH-A) by 4-fold and 5-fold, respectively. This response was virtually abolished in the ARNT Ϫ c4   3 -TKGH. Expression of the non-inducible plasmid pBS␣ Ϫ was used to correct for transfection efficiency. Reporter gene expression in normoxia was similar in the two cell types (being 1.5 and 1.6 for p(Epo1-25) 3 SV40GH and 0.2 and 0.3 for p(PGK24) 3 TKGH in Hepa-1 and c4 cells, respectively). The inducible response is the ratio of stimulated to normoxic reporter gene expression. Values are mean Ϯ S.D.; the number of experiments is shown in parentheses. * indicates p Ͻ 0.05 (statistical analysis of differences in inducible responses between the two cell types by Student's t test). Both the Epo and PGK-1 elements conveyed inducible responses in wild type cells, but this response was markedly diminished in ARNT Ϫ c4 cells. The inducible responses to hypoxia and DFO were of similar magnitude in wild type cells and were similarly abrogated in ARNT Ϫ c4 cells. cells, although, as with the transient transfection studies, a small response to hypoxia was observed occasionally. The mRNA for the glucose transporter-1 (Glut-1) was induced almost 10-fold by hypoxia in wild type cells. This response was much reduced in the c4 mutant cells. However, in contrast with LDH-A and PGK-1, a 2-3-fold induction of Glut-1 mRNA persisted in the mutant cells. The expression of two hypoxically inducible growth factor genes was compared in wild type Hepa-1 cells and the c4 mutants. Both vascular endothelial growth factor (VEGF) and platelet-derived growth factor B chain (PDGF-B) were expressed in an hypoxically inducible manner in the wild type cells; after 16 h of hypoxia, steady-state mRNA levels were increased approximately 7-fold for VEGF and approximately 4-fold for PDGF-B. In the c4 mutant cells, no hypoxic induction of PDGF-B mRNA was observed; the induction of VEGF mRNA was reduced to approximately 2-fold.
Since HIF-1␣ and ARNT/HIF-1␤ mRNAs have themselves been reported to be induced by hypoxia (17), we compared the regulation of these mRNAs in wild type and mutant cells. As has been reported previously, ARNT/HIF-1␤ mRNA was present at a near normal level in c4 mutant cells (25). Neither HIF-1␣ nor ARNT/HIF-1␤ mRNA levels were greatly increased by hypoxia in wild type Hepa-1 cells; induction of HIF-1␣ was less than 2-fold, and changes in ARNT/HIF-1␤ mRNA were barely discernible. In contrast with the other genes, the response of HIF-1␣ and ARNT/HIF-1␤ to hypoxia was greater in the c4 mutant cells than in the wild type cells. These differences were clear after 16 h of exposure to hypoxia, when HIF-1␣ mRNA was induced almost 4-fold, and ARNT/HIF-1␤ was induced 2-fold. Basal levels of HIF-1␣ mRNA were also 2-fold greater in the c4 mutant cells.
To determine whether induction by DFO and hypoxia were altered in a similar manner in the c4 mutant cells, responses were compared directly in equal aliquots of cells divided for parallel incubations in normoxia, hypoxia, and DFO (100 M). Four genes were studied: LDH-A, PGK-1, VEGF, and Glut-1 (Table III). In wild type cells, these genes were induced to a similar extent by both stimuli. In the c4 mutants, both inducible responses were substantially reduced. Surprisingly, however, the response to DFO in the c4 cells was better preserved than the response to hypoxia. This difference was observed in each experiment for all of the genes assayed, but the difference was most striking for PGK-1 and LDH-A. In c4 cells, these genes were virtually unresponsive to hypoxia, whereas the response to DFO persisted at approximately 50% of that seen in wild type cells.
To ascertain that the altered pattern of hypoxically inducible endogenous gene expression resulted from a defective ARNT/ HIF-1␤ gene product, rather than some other difference in the c4 cell line, or some indirect consequence of the defective xenobiotic response itself, we tested hypoxically inducible gene expression in a series of other cell lines derived from Hepa-1. These cells were: c39, an independently derived benzo-(a)pyrene-resistant clone in the same complementation group as c4; vT{2}, a c4 derivative stably transfected with an expression plasmid pBM5/Neo-M1-1 bearing a functional human ARNT/HIF-1␤ gene; Rc4, a revertant selected from the original c4 line that expresses wild type (Hepa-1) levels of ARNT activ-  1-4). The inducible response in ARNT Ϫ c4 cells is restored by transfection with the full coding sequence of ARNT/ HIF-1␤ (lanes 5 and 6) and also by the bHLHAB truncated ARNT cDNA (lanes 7 and 8) but not the empty expression vector (lanes 9 and 10) ity; and c31, a benzo(a)pyrene-resistant clone expressing a dominantly acting repressor of the xenobiotic response. Responses of LDH-A and Glut-1 mRNAs were analyzed. Results are shown in Fig. 4. c39 cells behaved very similarly to c4 cells, whereas both vT{2} cells and Rc4 showed restored hypoxic responsiveness. Rc4 cells showed a higher normoxic expression of Glut-1 and vT{2} cells a higher normoxic expression of LDH-A than Hepa-1 cells. However, HIF-1 activity in these lines was comparable to that in Hepa-1 and was not increased in either normoxia or hypoxia (data not shown). The dominant dioxin-resistant c31 cells showed normal hypoxic gene regulation.

DISCUSSION
The hypoxically inducible transcriptional factor, HIF-1, was first recognized as a nuclear factor binding to a site that was critical for the function of the erythropoietin 3Ј enhancer in transiently transfected cells (2). The recent purification and molecular cloning of HIF-1 have demonstrated that it is a heterodimer of two basic helix-loop-helix PAS proteins (17): HIF-1␣, a newly described member of this family, and HIF-1␤, a transcription factor previously recognized as the ARNT (25). The existence of ARNT Ϫ mutant clones has now provided the opportunity to analyze further the importance of ARNT/HIF-1␤ in responses to hypoxia.
The ARNT Ϫ c4 cells were first tested for HIF-1 activity. They failed to show inducible HIF-1 DNA binding and were unable to support the hypoxically inducible responses usually conveyed by multimerized HIF-1 binding sites from the mouse Epo and PGK-1 enhancers. The defective response was observed despite the presence of an increased level of HIF-1␣ mRNA in c4 cells and indicates that ARNT/HIF1␤ is an essential component of the complex that recognizes HIF-1 binding sites. Although the responses to hypoxia conveyed by the HIF-1-binding elements were usually absent in c4 cells, a very small response was occasionally seen. The mutation responsible for lack of ARNT function is unknown. It is possible that, in contrast with the xenobiotic response, some function in the hypoxic response is retained. Another explanation would be that HIF-1␣ has other dimerization partners, which can interact weakly with the HIF-1 binding sequence.
Regulation by hypoxia has been reported for a number of widely expressed transcription factors including NF-B (33), some members of the Fos/Jun family (34), and p53 (35)  hypoxia and desferrioxamine Responses to hypoxia and DFO were compared in aliquots of wild type Hepa-1 and ARNT Ϫ c4 cells exposed in parallel. The value given is the ratio of stimulated to normoxic expression and is the mean Ϯ S.D. of at least three independent experiments. * indicates p Ͻ 0.05 (statistical analysis of differences in inducible responses between the two cell types by Student's t test). The normoxic expression of each gene (relative to ␤-actin) in Hepa-1 and c4 cells is given in Fig. 3. In wild type Hepa-1 cells, each gene assayed was inducible by both hypoxia and DFO. The inducible responses were reduced in ARNT Ϫ c4 cells, but in all cases the response to DFO was better preserved than the hypoxic response.

FIG. 4. Comparison of the hypoxic induction of Glut-1 and LDH-A mRNA in cell lines with differing capacities for ARNT function.
A, RNase protection assay of Glut-1 mRNA. Paired aliquots of cells were incubated in normoxia (N) or hypoxia (H). The Glut-1 signal is from 20 g of RNA and the ␤-actin signal is from 1 g (see "Experimental Procedures"). B, summary of results for Glut-1 and LDH-A. Induction is the ratio of mRNA expression in hypoxic cells to that in normoxic cells (mean Ϯ S.D. of three independent experiments). The relative normoxic expression of the two genes in each cell type is calculated as a ratio of the gene expression to that of ␤-actin and then normalized to the expression in Hepa-1 cells. Cells are: Hepa-1 (wild type); c4, ARNT Ϫ ; c39, an independent benzo(a)pyrene-resistant clone from the same complementation group as c4; vT{2}, c4 bearing a stably transfected human ARNT/HIF-1␤ gene; Rc4, a c4 derivative, selected for reversion to xenobiotic responsiveness; c31, a strain expressing a dominantly acting repressor of the xenobiotic response. The presence or absence of a functional ARNT gene correlates with responses to hypoxia.
hypoxically inducible responses were conveyed by sequences lying both 5Ј and 3Ј to the human gene, of which only the 3Ј region contained a HIF-1 site (36). Goldberg and colleagues did not find these regions to be active, but defined a conserved HIF-1 site lying further 5Ј to the rat gene (10), as did Kourembanas and colleagues working on the human gene (37). In c4 cells the alteration in induction of VEGF mRNA was different from that observed for PGK-1 or LDH-A mRNA; although a consistent and substantial reduction in the level of induction was observed, an inducible response clearly persisted. A very similar pattern of altered expression was observed for Glut-1 mRNA. This indicates that both ARNT/HIF-1␤-dependent and ARNT/HIF-1␤-independent mechanisms of hypoxic induction are operating on the regulation of Glut-1 and VEGF mRNAs.
Security in assigning these altered responses to the defect in ARNT/HIF-1␤ was provided by the analysis of several related cell lines, in which hypoxic induction was clearly correlated with the presence or absence of a functional ARNT/HIF-1␤ gene product. Responses were reduced in both of the independently derived ARNT Ϫ lines c4 and c39, whereas in the ARNTexpressing stably transfected line vT{2}, the revertant Rc4, and line c31, they were essentially similar to those observed in wild type cells. The results in c31 cells were interesting. These cells possess a dominantly acting repressor, which interferes with both the transcriptional response to dioxin and the DNA-binding proteins responsible for in vivo footprints at the xenobioticresponsive element (XRE), but does not impair in vitro DNA binding of the activated receptor complex to the XRE (28). The preserved responses to hypoxia in these cells indicate that the repressor does not interfere with the function of HIF-1.
Overall, these results indicate that ARNT/HIF-1␤ has a major role in oxygen-dependent gene regulation. For PGK-1, LDH-A, and PDGF-B, the loss of induction indicates that limited redundancy exists in the mechanisms underlying induction of these genes by hypoxia. We found no evidence that the XRE conveyed a response to hypoxia (data not shown). Nevertheless, ARNT/HIF-1␤-dependent responses to hypoxia are not necessarily restricted to HIF-1 itself, since it is conceivable that ARNT/HIF-1␤ has other undiscovered dimerization partners involved in the response to hypoxia. The results also provide clear evidence for ARNT/HIF-1␤-independent mechanisms of hypoxic gene induction; such processes could involve either other mechanisms of transcriptional activation or increased mRNA stability.
Since many transcription factors bind cis-acting elements within their own regulatory sequences, and HIF-1␣ and ARNT/ HIF-1␤ mRNA levels have been reported to be induced markedly by hypoxia (17), these mRNAs were examined in wild type and c4 cells. In wild type Hepa-1 cells, we did not observe a high level of induction of these mRNAs by hypoxia; an increase in mRNA was consistently observed for HIF-1␣ but was less than 2-fold. Thus, it appears that in Hepa-1 cells mRNA induction is not a major mechanism of HIF-1 activation. However, interesting changes in expression were observed in ARNT Ϫ cells. As has been reported previously, ARNT/HIF-1␤ mRNA levels were near normal (25). However, HIF-1␣ mRNA was considerably increased; the basal level was 2-fold higher than in wild type cells, and a 4-fold induction was observed after 16 h of hypoxia. This is compatible with the existence of some form of regulatory mechanism operating in compensation for the lack of a functional HIF-1 complex.
Previous studies have shown that HIF-1 is induced by DFO as well as hypoxia (14). The DNA binding and transient transfection studies in c4 cells exposed to DFO showed that, as for hypoxia, ARNT/HIF-1␤ is essential for the formation of the HIF-1 DNA-binding complex and transcriptional activation.
The analysis of endogenous gene expression in cells exposed to DFO was interesting. Inducible responses to DFO were clearly reduced in c4 cells compared with wild type cells. However, in c4 cells, an inducible response was retained for Glut-1 and VEGF, which was similar although somewhat greater than the hypoxic responses. Thus, DFO appears able to activate ARNT/ HIF-1␤-independent mechanisms of gene regulation by hypoxia in addition to those dependent on ARNT/HIF-1␤. One surprising finding was that the response to DFO was consistently better preserved in c4 cells than the response to hypoxia, particularly for PGK-1 and LDH-A. This might imply some difference in the mechanisms of gene activation by hypoxia and DFO but precise understanding will require a clearer knowledge of the mechanism of oxygen sensing and how DFO interacts with this process.
Transfection of the c4 cells with ARNT/HIF-1␤ expression plasmids also provided a means of determining which domains of ARNT/HIF-1␤ are required for HIF-1-mediated transcriptional responses to hypoxia. Function was lost when any of the basic, helix-loop-helix, or PAS domains were deleted. Assuming that similar levels of protein expression were achieved, these results indicate that all of these regions are functionally critical for the operation of HIF-1. This result is similar to that obtained for functional analysis of ARNT/HIF-1␤ deletions in the xenobiotic response (32) and suggests that the interaction of ARNT/HIF-1␤ with HIF-1␣ is similar to its interaction with AHR. The bHLHAB deletion, which lacked the 5Ј and 3Ј regions of ARNT/HIF-1␤, including the 3Ј transactivation domain (38,39), retained function. Again, this result is similar to that reported for the xenobiotic response and indicates that the 3Ј transactivation domain of ARNT/HIF-1␤ is not necessary for transcriptional response in either system.
In summary, these studies have demonstrated the critical importance of ARNT/HIF-1␤ in the hypoxic regulation of many different genes. The ARNT Ϫ mutant cells are an important tool for analyzing the extent and mechanisms of gene regulation by this system and should also prove useful in understanding its role in cell physiology.