Hypoxia-inducible Mammalian Gene Expression Analyzed in Vivo at a TATA-driven Promoter and at an Initiator-driven Promoter*

We have analyzed protein-DNA interactionsin vivo at transcriptional control elements for two hypoxia-inducible genes in mouse hepatoma cells. The promoter for the phosphoglycerate kinase 1 (PGK1) gene contains an initiator element, but no TATA sequence, whereas the promoter for the glucose transporter 1 (Glut1) gene contains a TATA element but no initiator sequence. Our findings reveal hypoxia-inducible, Arnt-dependent occupancy of DNA recognition sites for hypoxia-inducible factor 1 (HIF-1) upstream of both target genes. The conserved recognition motif among the five recognition sites is 5′-CGTG-3′. The PGK1 promoter exhibits constitutive occupancy of a binding site for an unknown protein(s); however, we detect no protein-DNA interaction at the initiator element, in either uninduced or induced cells. The Glut1 promoter also exhibits constitutive protein binding; in addition, the TATA element exhibits partial occupancy in uninduced cells and increased occupancy under hypoxic conditions. We find no evidence for hypoxia-induced changes in chromatin structure of either gene. Time-course analyses of the Glut1 gene reveal a temporal relationship between occupancy of HIF-1 sites and TATA element occupancy. Our findings suggest that the promoters for both hypoxia-responsive genes constitutively maintain an accessible chromatin configuration and that HIF-1 facilitates transcription by recruiting and/or stabilizing a transcription factor(s), such as TFIID, at both promoters.

Perturbations in the environment present a continual challenge to cellular systems responsible for maintaining homeostasis. In some cases, cells utilize transcriptional pathways for adapting to change; such responses present opportunities to understand better the mechanisms by which environmental signals alter the expression of specific sets of genes. The bHLH/ PAS 1 class of proteins mediate several important transcrip-tional responses to environmental stimuli. These proteins interact with each other via their HLH (helix-loop-helix) domains to form transcription factors (usually heterodimers and usually activators) that bind to DNA via their basic (b) regions (1)(2)(3). The PAS domains contribute to protein-protein interactions and to other functions (4 -6). The designation PAS stands for the three proteins first discovered to contain the domains: Per, a Drosophila protein that helps regulate circadian rhythmicity; Arnt, a protein that heterodimerizes with the mammalian aromatic hydrocarbon receptor (AhR) in mediating transcriptional responses to xenobiotics like dioxin; and Sim, a protein implicated in Drosophila central nervous system development (7).
Induction of microsomal cytochrome P4501A1 by the environmental contaminant 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) is a well studied example of an adaptive transcriptional response mediated by a bHLH/PAS heterodimer. TCDD enters the cell and binds AhR; subsequently, the liganded AhR heterodimerizes with Arnt, and AhR/Arnt binds to enhancer DNA upstream of the CYP1A1 gene, thereby inducing its transcription (1-3, 8, 9). The induced cytochrome P4501A1 enzyme catalyzes the initial oxygenation step in a metabolic pathway that converts lipophilic substrates to water-soluble derivatives, facilitating their elimination from the cell. Metabolism by cytochrome P4501A1 primarily facilitates detoxification; thus, enzyme induction represents adaptation to a chemical exposure. However, the oxygenation reaction can also generate electrophilic reactants that are potentially toxic or carcinogenic, particularly when detoxification pathways become saturated (10).
bHLH/PAS proteins also mediate transcriptional responses to physiological perturbations, such as decreased oxygen tension. Hypoxia stabilizes the bHLH/PAS protein known as hypoxia-inducible factor-1␣ (HIF-1␣) thereby causing it to accumulate within the cell; HIF-1␣ heterodimerizes with Arnt to form a DNA-binding transcription factor known as hypoxiainducible factor-1 (HIF-1). HIF-1 mediates adaptive responses to low oxygen tension, such as the induction of genes encoding erythropoietin and some glycolytic enzymes (11)(12)(13).
The AhR/Arnt and HIF-1␣/Arnt heterodimers constitute useful models for analyzing the action of bHLH/PAS proteins. The study of such prototypes is likely to generate general insights because bHLH/PAS proteins contribute to diverse biological processes required both for normal development and for adaptation to the environment (14 -22). We have previously developed techniques for studying AhR/Arnt function in intact cells, and we have analyzed TCDD-induced alterations in protein-DNA interactions and chromatin structure at the CYP1A1 gene in mouse hepatoma cells in vivo (23)(24)(25). These experimental approaches permit us to study the action of bHLH/PAS proteins at chromosomal target genes under physiological conditions. Here, we have studied the response to hypoxia in vivo, and we have analyzed protein-DNA interactions at two genes, Glut1, which encodes a glucose transporter, and PGK1, which encodes phosphoglycerate kinase 1, an enzyme in the glycolytic pathway. We chose to study the induction of these two genes because one (Glut1) has a transcriptional promoter that contains a TATA element but no initiator sequence, while the other (PGK1) has a transcriptional promoter that contains an initiator element but no TATA sequence (26,27). Therefore, our findings provide new insights not only into the action of the bHLH/PAS heterodimer HIF-1, but also into the function of two different types of mammalian promoters in vivo. Cell Culture-Wild-type (Hepa 1c1c7), Arnt-defective (BP r C1), and Arnt-reconstituted mouse hepatoma cells were routinely cultured in an incubator (NAPCO, model 5430) maintained at 37°C and 95% air, 5% CO 2 (normoxic conditions) to 95% confluence (about 10 7 cells/100-mm plate) as described previously (28). Cells were then subjected to hypoxic conditions by placing them in an automatic CO 2 /O 2 incubator (Forma Scientific, model 3159) maintained at 37°C and 1% O 2 , 5% CO 2 , 94% N 2 for 18 h (unless otherwise noted). Uninduced cells remained in normoxic conditions. ⌽-NX cells were cultured as described previously (29).

Materials-DNase
Retroviral Expression of Arnt cDNA-Reconstitution of Arnt-defective cells with full-length Arnt cDNA was as described previously (30) with the exception that the ⌽-NX cell line was used rather than the BOSC23 cell line.
Analysis of Gene Expression-Total RNA was isolated from cells subjected to normoxic or hypoxic conditions using RNeasy spin columns according to the manufacturer's directions. Ten g of total RNA was electrophoresed through a 1% agarose gel containing formaldehyde and transferred to a nitrocellulose membrane. Blots were hybridized with 32 P-labeled PGK1, Glut1 or actin cDNA probes, washed, dried, and autoradiographed as described previously (31).
DMS Modification of Genomic DNA-Cells subjected to normoxic or hypoxic conditions were removed from the incubator, and DMS was immediately dispersed into the culture medium (5 l of DMS/ml of culture medium). The cells were incubated at 22°C for 2.5 min under normoxic conditions, the culture medium was removed, and the cell monolayer was rinsed three times with phosphate-buffered saline. 2 ml of stop buffer (20 mM Tris-HCl (pH 8.0), 20 mM NaCl, 20 mM EDTA, 1% sodium dodecyl sulfate, 600 g/ml proteinase K) was then added to each plate and the viscous solution was collected into a 15-ml conical polypropylene tube containing 2 ml of dilution buffer (150 mM NaCl, 5 mM EDTA), and incubated at 37°C for 3 h. The genomic DNA was then purified and analyzed by ligation-mediated PCR (LMPCR) as described previously (23).
DNase I Digestion of Genomic DNA-Cells subjected to normoxic or hypoxic conditions were removed from the incubator, and their culture medium was immediately removed and replaced with 3 ml of digestion buffer (150 mM sucrose, 80 mM KCl, 35 mM HEPES (pH 7.4), 5 mM K 2 HPO 4 , 5 mM MgCl 2 , 2 mM CaCl 2 , 0.01% lysolecithin, 20 g DNase I/ml) at 22°C. After a 2.5-min incubation under normoxic conditions, the digestion buffer was removed and was replaced by 2 ml of stop buffer. The viscous solution was collected into a 15-ml conical polypropylene tube containing 2 ml of dilution buffer, and incubated at 37°C for 3 h. The genomic DNA was then purified and analyzed by LMPCR as described previously (23).

Basal Expression, Inducible Expression, and Arnt-Two
bHLH/PAS proteins, HIF-1␣ and Arnt, form a heterodimer, designated as HIF-1, which regulates responses to low oxygen tension, such as adaptive increases in PGK1 and Glut1 transcription (30,(32)(33)(34). As a prelude to our analyses of hypoxiainducible PGK1 and Glut1 gene expression in vivo, we verified that, in mouse hepatoma cells, the induction mechanism for both genes involves HIF-1.
We determined the requirement for Arnt by analyzing gene expression in Arnt-defective cells (Fig. 1). Studies of mRNA levels in uninduced cells reveal that both wild-type and Arntdefective cells express the PGK1 and Glut1 genes constitutively; therefore, neither gene requires Arnt for basal expression. In contrast, low oxygen tension induces PGK1 and Glut1 mRNA in wild-type cells but not in Arnt-defective cells. Reconstitution of Arnt-defective cells with Arnt cDNA restores the response of both genes to hypoxia. Therefore, for both genes, hypoxia-responsiveness requires Arnt. In addition, the regulation of inducible expression differs from that of basal expression, because the former requires Arnt while the latter does not. Because induction requires Arnt, it also presumably re- quires Arnt's hypoxia-responsive heterodimerization partner, HIF-1␣. However, mouse hepatoma cells defective in HIF-1␣ are not available, so we cannot perform genetic reconstitution experiments to document the need for HIF-1␣. Therefore, we used other approaches to implicate HIF-1␣ in the induction mechanism, as described below.
Protein-DNA Interactions at HIF-1 Binding Sites in Vivo-Previous electrophoretic mobility shift studies suggested that the HIF-1␣/Arnt heterodimer binds to a "consensus" DNA sequence, 5Ј-(A/G)CGTG-3Ј (12,35); this sequence occurs in the 5Ј flanking region of both the Glut1 and PGK1 genes and confers hypoxia-responsiveness upon a reporter gene in transient transfection experiments (36 -38). Therefore, if HIF-1␣/Arnt mediates inducible gene expression in vivo, we envisioned that these consensus sequences would exhibit hypoxia-inducible, Arnt-dependent occupancy. To test this idea, we used an LMPCR footprinting method (39) to analyze the pattern of protein-DNA interactions at the potential HIF-1 binding sites in vivo. Our findings (Fig. 2, A and B) reveal that, in uninduced wild-type cells, the DNA encompassing the potential HIF-1 binding sites upstream of the PGK1 and Glut1 genes exhibits a DNase I digestion pattern similar to that of naked DNA. These findings imply that HIF-1 sites are unoccupied in uninduced cells in vivo. This result is consistent with the observation that basal expression of both genes occurs in Arnt-defective cells ( Fig. 1) and, therefore, does not require the HIF-1␣/Arnt heterodimer. In contrast, hypoxia-induced cells exhibit two neighboring DNase I footprints upstream of the PGK1 gene and one DNase I footprint upstream of the Glut1 gene in vivo. The footprints are Arnt-dependent, because they are absent in Arnt-defective cells but occur in Arnt-defective cells reconstituted with Arnt. These findings reveal hypoxia-inducible, Arntdependent protein-DNA interactions in vivo within regions that contain DNA sequences known to bind HIF-1 in vitro.
We also performed in vivo footprinting using DMS as a probe; this technique is more precise than DNase I footprinting because it can delineate the specific guanine residues that interact with proteins bound in the major DNA groove. Our findings reveal that in uninduced cells, the pattern of guanine methylation by DMS is similar to that of naked DNA, again implying that proteins do not bind to these regions under normoxic conditions (Fig. 2, A and B). In contrast, induced cells exhibit a pattern of methylation protection that indicates occupancy of three HIF-1 binding sites upstream of the PGK1 gene and two HIF-1 sites upstream of the Glut1 gene (Fig. 2, A-C). Their DNA sequences reveal that each of the five protein binding sites contains a 5Ј-CGTG-3Ј sequence, with considerable variability in the neighboring nucleotides; thus, there is no obvious consensus sequence in vivo beyond the tetranucleotide region (Fig. 2D). At each site, the guanine residues protected in vivo are those that interact with HIF-1 in vitro, as measured previously in a methylation interference analysis (Ref. 40; Fig.  2D). These findings imply that the hypoxia-inducible protein-DNA interactions identified in vivo reflect the binding of HIF-1 to cognate recognition motifs upstream of the PGK1 and Glut1 genes. Note that our in vivo analyses reveal the existence of two additional HIF-1 binding sites (centered at Ϫ218 upstream of the PGK1 transcription start site and at Ϫ3008 upstream of the Glut1 transcription start site) that were not previously identified using reporter gene or mobility shift assays.
We also performed several in vitro experiments to further analyze the DNA sites identified in vivo. Electrophoretic mobility shift studies reveal that nuclear extracts derived from mouse hepatoma cells contain a hypoxia-inducible factor, presumably HIF-1, that binds to the DNA sites occupied in vivo. Furthermore, mutation of the guanines known to interact with protein abolishes the inducible protein-DNA interaction in vitro (data not shown). Also, HIF-1␣ and Arnt cDNAs expressed together by in vitro transcription/translation interact with the wild-type, but not with the mutant, HIF-1 sites (data not shown). Finally, reporter gene studies reveal that the wildtype HIF-1 sites, but not the mutant sites, confer hypoxia responsiveness upon a luciferase gene after transient transfection into wild-type cells, but not after transfection into Arntdefective cells (data not shown). These in vitro experiments impute a functional role to the inducible protein-DNA interactions identified in vivo.
In our previous studies of dioxin-induced CYP1A1 gene expression in vivo, we observed a substantial increase in the nuclease susceptibility of DNA in the vicinity of chromosomal binding sites for the AhR/Arnt heterodimer; such findings signify a change in chromatin structure from a relatively inaccessible to an accessible configuration (23,24). Therefore, it is notable that we do not detect analogous hypoxia-inducible increases in the nuclease susceptibility of DNA surrounding the HIF-1 binding sites in vivo (data not shown); this finding implies that the inducible occupancy of HIF-1 binding sites upstream of the PGK1 and Glut1 genes occurs without an increase in DNA accessibility or a substantial change in chromatin structure. Therefore, we surmise that the hypoxiainducible protein-DNA interactions take place within a chromatin structure that is constitutively accessible to HIF-1.
We envision that occupancy of HIF-1 binding sites upstream of the PGK1 and Glut1 promoters is necessary for induction of gene expression and that the binding of HIF-1 activates both promoters by a similar mechanism. It is of interest that the two promoters have different organizations; the Glut1 promoter contains a TATA element but no initiator (Inr) sequence, whereas the PGK1 promoter is TATA-less, but contains an Inr element. Therefore, as an initial step in understanding promoter function in a native chromosomal context, we analyzed the occupancy of both promoters in vivo in both uninduced and induced cells.
Protein-DNA Interactions at the PGK1 and Glut1 Promoters in Vivo-The PGK1 promoter does not contain a TATA sequence, and transcription starts at an Inr element (27). Analyses of the PGK1 promoter in vivo (Fig. 3.) reveal that uninduced cells exhibit a DNase I footprint upstream of the Inr element, indicative of constitutive protein binding in the region spanning Ϫ20 to Ϫ80; this result agrees with observations made on the human PGK1 promoter (41,42). We detect no in vivo protein-DNA interaction at the Inr site in uninduced cells, even though the PGK1 gene exhibits constitutive expression. An 18-h period of hypoxia produces no detectable changes in the pattern of protein-DNA interactions at the PGK1 promoter; notably, there is no evidence of inducible protein binding at the Inr region even in the face of increased gene expression. We assume that PGK1 transcription requires a protein-DNA interaction(s) at the Inr (43, 44); our data imply that the interaction in vivo is not strong enough to be detected by DNase I footprinting.  5 and 6). Naked genomic DNA treated with these agents in vitro was also analyzed (lanes 1 and 4). Numbers indicate distances (in base pairs) from the major PGK1 transcription initiation site. The bracket indicates a DNA region protected from DNase I digestion in normoxic and hypoxic cells. Arrows indicate guanine residues protected from DMS modification. The line indicates the initiator element.
Analysis of the PGK1 promoter using DMS as a probe reveals constitutive protection of several guanine residues within the constitutive DNase I footprint; therefore, the constitutive protein binding occurs in the major DNA groove. We observe no alteration in the pattern of guanine modification after induction. Together, our findings reveal that induction of PGK1 gene expression is not associated with detectable changes in occupancy of the PGK1 promoter using in vivo footprinting techniques. Therefore, we infer that the protein-DNA interactions associated with hypoxia-inducible transcription from an Inrdriven promoter are relatively weak, at least when compared with those at a TATA-driven promoter (see below). In addition, we observe no increase in the susceptibility of the PGK1 promoter to DNase I during induction (data not shown). These findings contrast with our previous observations at the CYP1A1 promoter, which undergoes a substantial increase in accessibility during induction of gene expression by dioxin (23,24,45). The lack of an induced increase in nuclease susceptibility, together with constitutive protein binding, implies that the PGK1 promoter assumes an accessible chromatin structure in both uninduced and induced cells. Therefore, we envision that communication of the induction signal from HIF-1 binding sites to the PGK1 promoter does not involve substantial remodeling of chromatin.
The Glut1 promoter contains a TATA element about 30 bp upstream of the transcription start site and has no initiator element. Analysis of the Glut1 promoter in vivo (Fig. 4) reveals that uninduced cells exhibit a DNase I footprint spanning Ϫ84 to Ϫ73, implying that a protein(s) binds constitutively to a region upstream of the TATA element. Analyses using DMS imply multiple guanine residues participate in the constitutive protein-DNA interaction and that binding occurs in the major DNA groove. Thus, like the PGK1 promoter, a region of the Glut1 promoter exhibits constitutive in vivo occupancy; studies in hypoxic cells indicate that occupancy persists during induction of gene expression. In vivo DNase I footprinting reveals partial occupancy of the TATA element in uninduced cells; this finding is consistent with constitutive expression of the Glut1 gene. Hypoxia-induced cells exhibit a stronger DNase I footprint at the TATA element, implying that induction increases occupancy of this site in vivo (Fig. 4, lanes 1-3). (Note that DMS is not useful for analyzing the TATA element, because the site contains no guanines and because the TATA-binding protein occupies the minor DNA groove (46,47).). Thus, our findings indicate that induction of Glut1 gene expression is associated with increased occupancy of the TATA element. However, the increase in TATA occupancy is not accompanied by an increase in nuclease susceptibility of the Glut1 promoter in induced cells (data not shown). This finding, together with constitutive protein binding, suggests that the Glut1 promoter is in an accessible configuration in uninduced cells and does not undergo a major change in chromatin structure during induction.
Time Course of Inducible Protein-DNA Interactions on the Glut1 Gene in Vivo-Both the PGK1 and the Glut1 promoters appear to assume a chromatin structure that is constitutively accessible. Therefore, we infer that the primary role of HIF-1 binding at upstream sites is to facilitate promoter occupancy by increasing the availability of a transcription factor(s) at the promoter rather than by relieving a repressive chromatin structure. HIF-1 could increase the availability of transcription factors by recruiting them to the promoter. HIF-1 could also stabilize protein binding at the promoter. Recruitment and stabilization imply the existence of a temporal relationship between upstream binding of HIF-1 and downstream promoter occupancy. Therefore, we analyzed the time course of HIF-1 occupancy and TATA element occupancy in vivo during induction and de-induction of the Glut1 gene. (We could not perform analogous studies for PGK1, because its promoter exhibits no inducible changes.) Our findings (Fig. 5A) reveal that HIF-1 binding in vivo is maximal after 2 h of hypoxia and remains maximal for at least 18 h. Induced occupancy of the TATA element exhibits a similar time course (Fig. 5B). The coincidence between HIF-1 occupancy and TATA occupancy during induction is consistent with a recruitment and/or a stabilization mechanism.
During de-induction (i.e. after restoration of normoxia), occupancy of HIF-1 binding sites in vivo becomes undetectable within 10 min (Fig. 5A). Presumably, the loss of HIF-1 binding reflects the rapid degradation of HIF-1␣ under normoxic conditions (48). During de-induction, TATA occupancy in vivo falls to uninduced levels within 1 h (Fig. 5B). Thus, TATA occupancy is not sustained in the absence of HIF-1 binding upstream. This finding implies that HIF-1 may stabilize occupancy of the TATA element in vivo. DISCUSSION We have analyzed protein-DNA interactions at the hypoxiaresponsive PGK1 and Glut1 genes in vivo. Our findings relate constitutive and inducible occupancy of protein binding sites to basal and inducible gene expression from two different types of promoters. Our studies suggest that HIF-1 facilitates core promoter occupancy via recruitment and/or stabilization processes. Our in vivo observations provide a working model for hypoxia-inducible gene expression that can be tested in future studies of chromatin structure, interactions between HIF-1 and other transcription factors, and promoter-binding proteins.  5 and 6). Naked genomic DNA treated with these agents in vitro was also analyzed (lanes 1 and 4). Numbers indicate distances (in base pairs) from the Glut1 transcription initiation site. Brackets indicate DNA regions containing a constitutive and hypoxia-inducible DNase I footprint. Thin arrows indicate guanine residues protected from DMS modification. Bold arrows indicate guanine residues hypersensitive to DMS modification. The line indicates the TATA element.
Such experiments have the potential to reveal novel aspects of bHLH/PAS protein function.
Our studies identify five in vivo binding sites for the HIF-1␣/Arnt heterodimer. We do not observe a strong consensus recognition sequence among the five sites; the only common motif is the tetranucleotide 5Ј-CGTG-3Ј. We note that this same tetranucleotide is present in each of the DNA binding sites for the AhR/Arnt heterodimer and is required for dioxin responsiveness (49). Thus, the HIF-1␣/Arnt and AhR/Arnt heterodimers share a core DNA recognition sequence. UV crosslinking studies imply that the 5Ј-CGTG-3Ј tetranucleotide interacts largely with the Arnt component of the AhR/Arnt heterodimer (50). We envision that it also interacts with the Arnt component of HIF-1␣/Arnt. The in vivo binding sites do not exhibit other features that can clearly account for the DNA binding specificity of HIF-1␣/Arnt versus AhR/Arnt. This situation is reminiscent of that for the Drosophila bHLH/PAS proteins Trh (trachealess) and Sim (single-minded), both of which form heterodimers with Arnt; the Trh/Arnt and Sim/ Arnt heterodimers regulate different sets of genes but bind to the same DNA sequence, which includes the 5Ј-CGTG-3Ј motif. Domain swapping studies reveal that the PAS regions of Trh and Sim confer DNA recognition specificity upon the respective heterodimers (5). By analogy, we hypothesize that a similar situation exists for the hypoxia-responsive and dioxin-responsive regulatory systems and that the PAS domains of AhR and HIF-1␣ contribute to specific DNA recognition. Our in vivo footprinting experiments reveal constitutive and hypoxia-inducible occupancy of the TATA element at the Glut1 promoter. These footprints presumably reflect the binding of TATA-binding protein (TBP) to DNA. Within the cell, TBP is part of a protein complex known as transcription factor IID (TFIID), which contains additional components, known as TBP-associated factors (TAFs). In vitro experiments imply that TAFs interact with DNA, albeit more weakly than TBP does (51). The binding of TFIID to a TATA-containing promoter (via TBP) permits the formation of a preinitiation transcription complex, which includes RNA polymerase II and other factors (44,(52)(53)(54). The protein(s) that interacts with Inr-containing promoters is not yet established. In vitro experiments (e.g. DNase I footprinting, electrophoretic mobility shift studies, in vitro transcription) imply that a TAF(s) can bind and activate (some) TATA-less, Inr-containing promoters (43,55). The interaction between TFIID and Inr DNA is weaker than the TFIID-TATA interaction, as estimated in electrophoretic mobility shift assays (43). Therefore, our inability to detect protein-DNA interactions at the PGK1 Inr in vivo is not surprising and is consistent with in vitro observations. Because TFIID can mediate the formation of preinitiation complexes at both TATA-containing and TATA-less, Inr-containing promoters, we envision that the induction mechanism for both PGK1 and Glut1 reflects HIF-1-mediated recruitment and/or stabilization of TFIID at the respective promoters.
It is unknown why some promoters utilize a TATA element and others utilize an Inr element to initiate transcription. The two elements are not interchangeable; for example, a TATA sequence cannot substitute for the native Inr element on the terminal deoxynucleotidyl transferase promoter (56). It is possible that TATA-driven promoters and Inr-driven promoters respond differentially to particular activators or inhibitors. If so, future studies may reveal that the PGK1 and Glut1 genes differ in their responsiveness to some signals, even though both genes respond to hypoxia via an apparently similar mechanism. These two hypoxia-responsive genes may constitute a useful experimental system for analyzing functional differences between mammalian TATA-containing and Inr-containing transcriptional promoters.
We envision that protein-protein interactions are important in the recruitment of TFIID. HIF-1 might interact either with FIG. 5. Time course of protein-DNA interactions on the Glut1 enhancer and promoter in vivo. Wild-type cells were exposed to hypoxic conditions (1% O 2 , 18 h) followed, in some instances, by normoxic conditions for the indicated lengths of time. The DNase I digestion pattern was visualized by LMPCR (lanes 2-13). Naked genomic DNA treated with DNase I in vitro was also analyzed (lane 1). Numbers indicate distances (in base pairs) from the Glut1 transcription initiation site. Brackets indicate constitutive and hypoxia-inducible DNase I footprints. Lines indicate binding sites for HIF-1 and the TATA element. A, LMPCR analysis of the Glut1 enhancer using primer set D. B, LMPCR analysis of the Glut1 promoter using primer set G. a component of the TFIID complex or with an intermediary (co-activator) protein. For example, others have suggested that p300/CBP functions as a co-activator during transcriptional adaptation to hypoxia (57). Thus, HIF-1 might recruit TFIID via p300/CBP. Recruitment does not require Arnt's C-terminal domain because Arnt-defective cells reconstituted with truncated Arnt cDNAs respond to hypoxia (30). We envision that a domain(s) of HIF-1␣ that contains transactivation capability is primarily responsible for recruiting TFIID to the Glut1 and PGK1 promoters. The availability of HIF-1␣-defective cells will facilitate testing this hypothesis (32).
The PGK1 and Glut1 promoters exhibit constitutive expression; a protein(s) binds constitutively to each promoter in vivo; neither promoter exhibits an increase in nuclease susceptibility following induction. These observations imply that, in vivo, promoter chromatin is accessible to DNA-binding proteins under both uninduced and induced conditions. The mechanism for generating and maintaining an accessible chromatin structure is unknown. One possibility is that the constitutively bound proteins prevent nucleosome formation at the Glut1 and PGK1 promoters. In addition, the constitutively bound proteins might interact with a component(s) of TFIID, thereby stabilizing promoter occupancy and facilitating basal and inducible gene expression. These are intriguing areas for future study.
Searches of the genomic data base indicate that the bHLH/ PAS class of proteins contains numerous members, many of which remain to be characterized. Phylogenetic analyses of proteins containing PAS domains imply that HIF-1␣, Arnt, and AhR are prototypes that have diverged from an ancient ancestor (58). Their conservation during evolution suggests that bHLH/PAS proteins play interesting and important roles in both vertebrate and invertebrate species. Therefore, knowledge of bHLH/PAS protein function obtained in studies of prototypes such as HIF-1␣, AhR, and Arnt has broad implications for understanding fundamental biological processes involving gene regulation and development.