|Home | About | Journals | Submit | Contact Us | Français|
The basic helix-loop-helix-Per-ARNT-Sim–proteins hypoxia-inducible factor (HIF)-1α and HIF-2α are the principal regulators of the hypoxic transcriptional response. Although highly related, they can activate distinct target genes. In this study, the protein domain and molecular mechanism important for HIF target gene specificity are determined. We demonstrate that although HIF-2α is unable to activate multiple endogenous HIF-1α–specific target genes (e.g., glycolytic enzymes), HIF-2α still binds to their promoters in vivo and activates reporter genes derived from such targets. In addition, comparative analysis of the N-terminal DNA binding and dimerization domains of HIF-1α and HIF-2α does not reveal any significant differences between the two proteins. Importantly, replacement of the N-terminal transactivation domain (N-TAD) (but not the DNA binding domain, dimerization domain, or C-terminal transactivation domain [C-TAD]) of HIF-2α with the analogous region of HIF-1α is sufficient to convert HIF-2α into a protein with HIF-1α functional specificity. Nevertheless, both the N-TAD and C-TAD are important for optimal HIF transcriptional activity. Additional experiments indicate that the ETS transcription factor ELK is required for HIF-2α to activate specific target genes such as Cited-2, EPO, and PAI-1. These results demonstrate that the HIF-α TADs, particularly the N-TADs, confer HIF target gene specificity, by interacting with additional transcriptional cofactors.
Low levels of O2, or hypoxia, are encountered by cells within rapidly growing tissues, such as developing embryos or solid tumors (Semenza, 2001 ). In response to hypoxic stress, mammalian cells activate hypoxia-inducible transcription factors (HIFs) to enhance the transcription of genes involved in glycolysis, angiogenesis, and cell survival to maintain oxygen homeostasis (Semenza, 2001 ). Therefore, hypoxic responses are essential for embryonic development and tumor progression (Semenza, 2001 ; Giaccia et al., 2004 ). HIF consists of heterodimers of α- and β-subunits (also known as aryl hydrocarbon receptor nuclear translocator [ARNT]). Both subunits are basic-helix-loop-helix (bHLH)-Per-Arnt-Sim (PAS) domain containing proteins. The HLH and PAS domains are necessary for α/β subunit dimerization, whereas basic regions from both subunits mediate binding to the hypoxia response element (HRE) of HIF target genes (CACGTG) (Wang and Semenza, 1995 ; Kinoshita et al., 2004 ).
HIF-1α/ARNT and HIF-2α/ARNT dimers are the primary factors regulating hypoxic transcriptional responses in most mammalian cells. Initial studies suggest that they have similar functions, consistent with the fact that HIF-1α and HIF-2α proteins are closely related (Ema et al., 1997 ; Flamme et al., 1997 ; Hogenesch et al., 1997 ; Tian et al., 1997 ; Wiesener et al., 1998 ; O'Rourke et al., 1999 ) (Figure 1). For example, HIF-1α and HIF-2α exhibit similar functional domain structures, containing DNA binding and dimerization domains at their N termini, and transactivation domains at their C termini (Jiang et al., 1997 ; Pugh et al., 1997 ; O'Rourke et al., 1999 ). The C termini of HIF-1α and HIF-2α are also similarly subdivided into a N-terminal activation domain (N-TAD, overlapping with the oxygen-dependent degradation domain), an inhibitory domain, and a C-terminal transactivation domain (C-TAD) (Jiang et al., 1997 ; Pugh et al., 1997 ; O'Rourke et al., 1999 ). Most importantly, HIF-1α and HIF-2α exhibit significant homology in several regions: they share 83 and 70% sequence identities in their DNA binding and dimerization domains, respectively. Furthermore, amino acids surrounding the two oxygen-sensitive proline residues (30 amino acids [aa] each) are highly conserved between HIF-1α and HIF-2α (70% similarity). In addition, their C-TADs are also very similar (67% similarity). These homologies provide the molecular basis for several common properties of HIF-1α and HIF-2α. For example, both use ARNT as a common binding partner, their protein stabilities are similarly regulated in an oxygen-dependent manner, and the transcriptional activity of both C-TADs is regulated by “factor inhibiting HIF” (FIH)-mediated hydroxylation of a conserved asparagine amino acid that blocks the recruitment of transcriptional coactivators p300 and CBP (Ema et al., 1997 ; Flamme et al., 1997 ; Hogenesch et al., 1997 ; Tian et al., 1997 ; Wiesener et al., 1998 ; O'Rourke et al., 1999 ; Mahon et al., 2001 ; Lando et al., 2002 ). Based on these properties, it has been suggested that HIF-1α and HIF-2α play related roles in hypoxic responses, and the unique phenotypes observed in Hif-α mutant mice are due to their distinct expression patterns during development (Tian et al., 1998 ; Peng et al., 2000 ; Compernolle et al., 2002 ).
More recently, we and others have shown that HIF-1α and HIF-2α actually regulate both unique and common target genes in vivo and in multiple cell lines (Hu et al., 2003 ; Grabmaier et al., 2004 ; Warnecke et al., 2004 ; Rankin et al., 2005 ; Raval et al., 2005 ; Wang et al., 2005 ; Covello et al., 2006 ; Gruber et al., 2007 ). For example, HIF-1α specifically regulates glycolytic genes (including phosphoglycerate kinase [PGK] and lactate dehydrogenase A [LDHA]) (Hu et al., 2003 ; Rankin et al., 2005 ; Wang et al., 2005 ), as well as carbonic hydrase-9 (CA IX) (Grabmaier et al., 2004 ) and BNIP3 (Raval et al., 2005 ), whereas HIF-2α exclusively regulates the Pou transcription factor Oct-4, cyclin D1, and transforming growth factor α (TGF-α) (Raval et al., 2005 ; Covello et al., 2006 ). Other hypoxia-inducible genes, such as vascular endothelial growth factor (VEGF), facilitated glucose transporter-1 (GLUT-1), adipose differentiation-related protein (ADRP), adrenomedullin (ADM), and N-myc downstream regulated 1 (NDRG-1) are regulated by both HIF-1α and HIF-2α (Hu et al., 2003 ; Raval et al., 2005 ; Hu et al., 2006 ). These results are in agreement with functional studies indicating that HIF-1α and HIF-2α exhibit distinct roles during both development and tumor progression (Maranchie et al., 2002 ; Kondo et al., 2003 ; Scortegagna et al., 2003 ; Covello et al., 2005 , 2006 ; Raval et al., 2005 ). By generating embryonic stem (ES) cells expressing HIF-2α at the Hif-1α locus, we directly compared HIF-1α and HIF-2α functions in embryos and tumors, and we showed that HIF-2α more potently promotes teratoma growth in nude mice (Covello et al., 2005 , 2006 ).
HIF-1α and HIF-2α exhibit unique target genes; however, the molecular mechanism(s) providing target gene specificity remain unclear. Differences in their DNA binding and dimerization domains could be important for target gene specificity by binding to related but not identical cis-elements; alternatively, unique transcriptional activation and inhibitory domains may be responsible for different functions via interaction with distinct promoter-specific transcription factors. In this study, the relative contributions of DNA binding and transcriptional activation domains of HIF-α to target gene specificity were investigated. Binding and activation of HIF-1α target genes by HIF-2α protein was compared with HIF-1α by using both plasmid-based reporters and endogenous genes. A number of deletion and domain swap mutants were constructed and their ability to activate reporter and endogenous target genes was analyzed. The results revealed how target gene specificity is achieved by the closely related HIF-1α and HIF-2α proteins.
Wild-type (WT), Hif-1α−/−, Hif-2α−/−, and Hif-1α−/− mouse ES cells stably transfected with WT HIF-1α or HIF-2α cDNA have been described previously (Hu et al., 2006 ). Human embryonic kidney (HEK)293 Tet-on HIF-1αDM (double proline mutation) or HIF-2αDM cells have also been described previously (Hu et al., 2003 ). The WT8/HIF-1αFlag or -HIF-2αFlag cells were established by stable transfection of pcDNA3-HIF-1αFlag or pcDNA3-HIF-2αFlag plasmid into 786-O/WT8 renal clear cell carcinoma cells and hygromycin selection. BpRc-1 cells (ARNT-deficient) stably transfected with control or ARNT-expressing plasmid were described previously by Arsham et al., 2002 . HEK293, Hep3B, and the cells described above were cultured in DMEM containing 4.5 g/ml glucose, 10% fetal calf serum (15% for ES cells), and 25 mM HEPES, under 1.5% O2 for hypoxia or 21% O2 for normoxia experiments.
Control or siRNAs specific for human HIF-1α, HIF-2α, and ELK-1 mRNAs were synthesized by QIAGEN (Valencia, CA). Hep3B cells were transfected with siRNAs at 60% confluence using HiPerFect Transfection Reagent according to the manufacturer's protocol. 24 h posttransfection, cells were cultured at 21 or 1.5% O2 for 12 h, and they were collected to analyze HIF-α or ELK-1 RNA, protein, or target genes.
pcDNA3 vectors expressing full-length normoxia-stable mouse HIF-1α (P402A/P577A), or HIF-2α (P405A/P530A) proteins were described previously (Hu et al., 2003 ). To generate normoxia-active HIF-α protein, N813 of HIF-1α or N851 of HIF-2α was mutated to alanine by amplifying the full-length plasmid by using mutation-incorporated primers and Pfu enzyme as described previously (Hu and Gupta, 2000 ). These constructs were called “HIF-α triple mutants (TMs).” Full-length pcDNA3 HIF-αTM DNA served as a template to generate several deletion mutants that lacked the N-TAD (ΔN-TAD), or inhibitory domain (ΔIH), or C-TAD (ΔC-TAD) for both HIF-1α and HIF-2α. HIF-1α/HIF-2α hybrid constructs “122,” “211,” “112,” and “221” TMs were generated by ligation of polymerase chin reaction (PCR) DNA fragments from pcDNA3 HIF-1αTM with PCR DNA fragments from pcDNA3 HIF-2αTM. Full-length cDNA from pcDNA3 HIF-1αTM and HIF-2αTM were cloned into the pcDNA 3.1-Flag vector by using BamHI/BstXI and KpnI/BstXI sites, respectively, to produce HIF-1αFlag and HIF-2αFlag plasmids. HIF-1αNFlag or HIF-2αNFlag plasmids were generated by PCR-mediated deletion of amino acids 365–836 from full-length HIF-1αFlag or 367–874 from full-length HIF-2αFlag. Addition of the VP16 transactivation domain to HIF-1αN or HIF-2αN produced HIF-1αN/VP16 or HIF-2αN/VP16. All newly constructed plasmids were further analyzed by sequencing to confirm that they encoded the desired HIF-α proteins.
RNA isolation and Northern blot analysis were performed using standard protocols. Human DNA fragments for Northern probes (PGK, LDHA, NDRG-1, ADRP, and α-tubulin) have been described previously (Hu et al., 2003 ). Mixed primer/probes sets for human or mouse HIF-1α, HIF-2α, VEGF, NDRG-1, ADM, PGK, LDHA, GLUT-1, PAI-1, Cited-2, IGFBP-1, and 18S rRNA (endogenous control) were used to measure the levels of these transcripts by the 7900HT Sequence Detection System (Applied Biosystems, Foster City, CA) according to the manufacturer's instructions.
Nuclear extracts (NEs) and cytoplasmic fractions were prepared in the presence of protease inhibitors as well as 200 μM deferoxamine (DFX) as described previously (Hu et al., 2003 ). Western blot analysis was performed using standard protocols with the following primary antibodies: anti-HIF-1α monoclonal antibodies (mAbs) (NB 100-105; Novus Biologicals, Littleton, CO, for mouse HIF-1α protein in Figures 3A and and10B),10B), anti-HIF-1α mAb (610959; BD Biosciences Transduction Laboratories, Lexington, KY, for human HIF-1α protein in Figures 2B and and12A),12A), anti-HIF-2α polyclonal antibodies (pAbs) (NB 100-122; Novus Biologicals), anti-ARNT mAb (NB 100-124; Novus Biologicals), anti-Flag M2 mAb (F-3165; Sigma-Aldrich, St. Louis, MO), anti-Myc mAb (clone 9E10; Roche Diagnostics, Indianapolis, IN), and anti-ELK mAb (H00002002-M01; Novus Biologicals).
ChIP assays were performed as described previously (Hu et al., 2006 ). Anti-HIF-1α mAb (NB 100-105; Novus Biologicals) was used for HIF-1α protein precipitation with mouse IgG2b immunoglobulin at the same concentration as a control. Anti-HIF-2α pAb (NB 100-122; Novus Biologicals) was used for HIF-2α precipitation, whereas rabbit preimmune serum served as a control. DNA from input (1:20 diluted) or immunoprecipitated samples was assayed using regular PCR in the presence of [α-32P]dCTP. The PCR products were separated by acrylamide gel electrophoresis and detected by PhosphorImager analysis. Alternatively, DNA from input or immunoprecipitated samples was quantified by SYBR Green-based quantitative (Q)-PCR. All PCR products were compared with input amounts to normalize for variations in the input signal that could arise from variable chromatin preparation. The results were plotted as -fold changes relative to their individual control antibody in individual cells. Primer pairs were pretested to amplify target genomic DNA in a linear manner. The following forward (F) and reverse (R) primers were used to detect HRE-containing LDHA genomic DNA in Hot-PCR: LDHA F, 5′-TGGCCTTTCTTTGGGGTGTCGCAGC-3′; and LDHA R, 5′-GGGGCCCAACCGTACCGCTAGATGC-3′.
The following were the primers to detect HRE-containing PGK and LDHA genomic using SYBR-Green Q-PCR: PGK F, 5′-GGCATTCTGCACGCTTCAA-3′; PGK R, 5′-GAAGAGGAGAACAGCGCGG-3′; LDHA F, 5′-ATCGATGCATTTGGGCTC-3′; and LDHA R, 5′-CAACCCGACATGCTCCTCA-3′.
WT HRE-Luciferase reporter, as well as mutant HRE-Luciferase reporter, were described by Hu et al. (2003) . The human PGK promoter reporter was constructed by inserting a 910-bp human PGK promoter fragment into the pGL3Basic vector at the BglII/HindIII sites. The control plasmid PGKdHRE contained a 50-bp deletion covering all three HIF binding sites in the PGK HRE. The human GLUT-1 reporter was similarly made, with a 700-bp HRE-containing enhancer and a 1.5-kb GLUT-1 promoter in the pGL3Basic vector. The control GLUT-1dHRE plasmid exhibited a 30-bp GLUT-1 HRE deletion. Transient transfection of cells in a 35-mm dish with HRE-dependent reporters (100 ng) and HIF-α expression plasmids (200 ng) was performed with FuGENE 6 (for HEK293 and HEK293 Tet-on; Roche Diagnostics) or Lipofectamine Plus Reagent (for BpRc-1 and Hep3B cells; Invitrogen, Carlsbad, CA) according to the manufacturer's instructions. β-Galactosidase (β-Gal) activity was analyzed to monitor transfection efficiencies in reporter gene experiments.
Glycolytic pathway components such as PGK and LDHA are classic hypoxia inducible genes; however, none of the 13 glycolytic genes are induced by hypoxia in 786-O/WT8 renal carcinoma cells, a line that exclusively expresses HIF-2α (Hu et al., 2003 ). Induction of glycolytic genes is restored in 786-O/WT8 cells upon HIF-1α introduction, indicating that HIF-1α, not HIF-2α, activates glycolytic gene expression (Hu et al., 2003 ). Consistent with results from 786-O/WT8 cells, glycolytic genes are induced solely by HIF-1α in HEK293 Tet-on cells where addition of doxycycline induces normoxically stable HIF-1α or HIF-2α protein expression (Hu et al., 2003 ). Similarly, recombinant adenovirus-mediated expression of HIF-1α, but not HIF-2α, stimulates glycolytic gene expression in HEK293 cells (Wang et al., 2005 ). These results indicate that the glycolytic genes are HIF-1α-specific targets.
However, it is still unclear whether glycolytic genes are only regulated by HIF-1α in a cell type that expresses functional HIF-1α and HIF-2α proteins endogenously. Hep3B cells were selected to address this question, because Hep3B cells have been shown to express both HIF-1α and HIF-2α proteins, and they exhibit high levels of hypoxic induction of HIF targets such as the glycolytic genes PGK and LDHA (Hu et al., 2003 ; Warnecke et al., 2004 ). Hep3B cells were transfected with siRNAs specific to HIF-1α, HIF-2α, or both HIF-α subunit mRNAs. HIF-α target gene induction was assessed using real-time Q-PCR on RNAs prepared from siRNA-targeted cells cultured at 21% O2 (N) or 1.5% O2 (H). HIF-1α siRNA reduced HIF-1α mRNA expression to ~10% of the levels in nontargeted cells (Figure 2A, left), whereas HIF-2α siRNA dramatically decreased HIF-2α mRNA expression to ~15% of the levels in nontargeted cells (Figure 2A, right). Furthermore, combination of both siRNAs significantly diminished mRNA levels of both subunits (Figure 2A). Consistent with Q-PCR assays of HIF-α mRNA, Western blot analysis indicated that HIF-1α or HIF-2α protein levels were greatly reduced in hypoxic Hep3B cells targeted with HIF-1α or HIF-2α siRNA, whereas cells targeted with both siRNAs exhibited a significant reduction (90%) of both HIF-α proteins (Figure 2B). However, transfection of control siRNA had no effect on HIF-α mRNA (Figure 2A, con) or protein levels (Figure 2B, con), and no cross-reactivity between HIF-1α and HIF-2α siRNAs was observed (Figure 2, A and B).
In accordance with our previous Northern blot analysis (Hu et al., 2003 ), Q-PCR assays indicated that several hypoxia-inducible genes exhibited a significant induction in hypoxic Hep3B cells, with NDRG-1 displaying the highest induction at 16-fold (Figure 2C). HIF-2α siRNA significantly reduced hypoxic induction of VEGF, NDRG-1, and ADM expression, whereas HIF-1α siRNA decreased their induction much less than that of HIF-2α siRNA (Figure 2C). Interestingly, the greatest reduction was observed in Hep3B cells targeted with both HIF-1α and HIF-2α siRNAs (Figure 2C). These results confirmed that VEGF, NDRG-1, and ADM were common target genes of HIF-1α and HIF-2α, although they were preferentially regulated by HIF-2α in Hep3B cells. Several hypoxia-inducible genes involved in glucose metabolism (PGK, LDHA, and GLUT-1) were also analyzed. In contrast to VEGF, NDRG-1, and ADM, knockdown of HIF-1α seemed to be sufficient to ablate hypoxic induction of PGK, LDHA and GLUT-1 in Hep3B cells (Figure 2C). We concluded that in Hep3B cells exhibiting both HIF-1α and HIF-2α proteins, endogenous HIF-1α specifically regulates the glycolytic genes PGK and LDHA, whereas both HIF-1α and HIF-2α stimulate VEGF, NDRG-1, and ADM expression. However, HIF-2α seems to be the critical HIF-α regulator of these genes in the Hep3B hepatocytes.
We showed previously that introduction of HIF-1α into 786-O/WT8 cells restores glycolytic gene induction (Hu et al., 2003 ). It is of interest to see whether overexpressed HIF-2α activates glycolytic gene expression. To test this, we generated 786-O/WT8 clones that expressed high levels of HIF-2α protein by stably transfecting them with WT HIF-2α cDNA (Figure 3A, right). Similarly, 786-O/WT8 cells expressing WT HIF-1α protein were created to serve as a control (Figure 3A, left). Both HIF-1α and HIF-2α cDNAs were Flag-tagged at their C termini, allowing detection of HIF-α proteins by using anti-Flag antibody. As shown in Figure 3A, left, parental 786-O/WT8 cells lacked HIF-1α expression, but two independent 786-O/WT8/HIF-1αFlag clones expressed HIF-1α protein under hypoxia as detected by anti-Flag and anti-HIF-1α antibodies. Although parental 786-O/WT8 cells exhibited endogenous HIF-2α protein expression under hypoxia (Figure 3A, right), HIF-2α protein levels were increased in 786-O/WT8/HIF-2αFlag clones to a level that allowed detection of HIF-2α protein under normoxia (Figure 3A, right). Q-PCR analysis of HIF target genes was performed in parental 786-O/WT8 cells, and cells expressing high levels of either HIF-1α or HIF-2α. Consistent with our previous results (Hu et al., 2003 ), parental 786-O/WT8 cells (expressing HIF-2α, but not HIF-1α) showed hypoxic induction of ADM and GLUT-1, but not PGK and LDHA. Of note, LDHA exhibited weak HIF-independent hypoxic induction (Figure 3B, WT8) as we showed previously in mouse ES cells (Hu et al., 2006 ). As expected, 786-O/WT8/HIF-1αFlag cells revealed increased hypoxic induction of ADM and GLUT-1 genes, and restored hypoxic induction of PGK and LDHA genes (Figure 3B, HIF-1α), as HIF-1α has been shown to regulate all four genes in multiple cell types (Hu et al., 2003 ; Hu et al., 2006 ). Interestingly, 786-O/WT8/HIF-2αFlag cells exhibited higher levels of hypoxic expression of ADM and GLUT-1 genes than that of 786-O/WT8/HIF-1αFlag cells. Higher levels of ADM and GLUT-1 expression in normoxic 786-O/WT8/HIF-2αFlag cells likely reflected HIF-2α protein expression in these cells under normoxia (Figure 3B, HIF-2α). In contrast to HIF-1α, overexpressed HIF-2α protein did not activate PGK and LDHA gene expression in 786-O/WT8/HIF-2αFlag cells (LDHA induction was changed from 1.3-fold in parental cells to 1.5-fold in HIF-2α–overexpressed cells). These data indicated that overexpressed HIF-2α in 786-O/WT-8 cells has no significant effect on the expression of the HIF-1α unique genes PGK and LDHA, despite that HIF-2α regulates ADM and GLUT-1 efficiently.
The effect of overexpressed HIF-2α on glycolytic gene expression was further investigated in mouse ES cells (Figure 3C). ADM, GLUT-1, PGK, and LDHA were hypoxically induced in WT ES cells, but not in Hif-1α−/− ES cells (Figure 3C). This is consistent with our previous findings that HIF-2α is not functional in mouse ES cells (Hu et al., 2006 ). As expected, stable reintroduction of WT HIF-1α into Hif-1α−/− ES cells restored hypoxic induction of all four genes (Figure 3C, HIF-1α). Overexpression of WT HIF-2α into Hif-1α−/− ES cells by stable transfection restored hypoxic induction of ADM and GLUT-1 (Figure 3C, HIF-2α), but not PGK and LDHA (LDHA induction was changed from 1.5-fold in parental cells to 1.8-fold in HIF-2α–overexpressed cells) (Figure 3C, HIF-2α). Thus, the HIF-1α target genes PGK and LDHA were not stimulated by overexpressed HIF-2α in both 786-O/WT8 and mouse ES cells, whereas HIF-2α activated the HIF-1α/HIF-2α common target genes GLUT-1 and ADM. These data further confirmed target gene specificity between HIF-1α and HIF-2α. In addition, endogenous target gene selectivity for PGK and LDHA is also essentially maintained with overexpressed HIF-α protein.
Having confirmed HIF target gene specificity in several cell lines, we wanted to use PGK (HIF-1α unique target) and GLUT-1 (HIF-1α/HIF-2α common target) as models to investigate which domain of HIF-α is important for target gene specificity. Although HIF-1α and HIF-2α proteins share highly conserved N-terminal DNA binding and dimerization domains, they exhibit differences in these regions that could be important for target gene specificity. We hypothesized that HIF-2α is unable to bind to the PGK promoter; therefore, it does not activate PGK expression. To test this hypothesis, we first assessed the ability of HIF-2α to activate a luciferase reporter gene under the control of a synthetic promoter containing three copies of the mouse PGK HRE (WT-HRE) (Arsham et al., 2002 ). Normoxic-stable HIF-2α stimulated the cotransfected WT-HRE reporter as efficiently as normoxic-stable HIF-1α in HEK293 cells under normoxia (Figure 4A, left). HIF-α proteins activated reporter gene expression via HRE binding, because no reporter gene induction was observed for Mut-HRE where HIF binding sites (HBSs) in HREs are mutated (Figure 4A, left). To better control HIF protein levels in reporter gene assays, activation of the WT-HRE reporter was analyzed in HEK293 Tet-on HIF-αDM (double proline mutation) cells in which physiological levels of HIF-1α and HIF-2α proteins were expressed upon doxycycline addition (Hu et al., 2003 ). Interestingly, HIF-1α and HIF-2α exhibited similar levels of WT-HRE reporter induction in HEK293 Tet-on HIF-αDM cells (Figure 4A, right). These data are in agreement with previous reports that HIF-2α has similar transactivation capability to that of HIF-1α in PGK reporter gene assays (Wiesener et al., 1998 ), suggesting HIF-2α can bind to and activate the HRE isolated from PGK.
The WT-HRE reporter is an artificial plasmid constructed by linking three copies of 18 nt PGK HREs in tandem with the thymidine kinase (TK) promoter (Arsham et al., 2002 ). To test whether HIF-2α induces a more natural PGK promoter, we generated a PGK reporter derived from a human genomic DNA fragment containing the PGK promoter including its HREs (Figure 4B). As stated above, GLUT-1 represents a gene that is regulated by both HIF-1α and HIF-2α. Its promoter and HRE-containing enhancer were isolated and used to generate an additional reporter construct (Figure 4B). For controls, HREs were deleted from PGK and GLUT-1 plasmids, resulting in the PGKdHRE-Luc and GLUT-1dHRE-Luc constructs shown (Figure 4B). As expected, both HIF-1α and HIF-2α stimulated the WT GLUT-1 reporter to similar levels in the transient transfection experiments, and no induction was observed for the GLUT-1dHRE construct (Figure 4C, right). Interestingly, HIF-2α protein clearly induced the WT PGK reporter, albeit less efficiently than the same amount of transfected HIF-1α plasmid (Figure 4C, left). The results suggest that reporter gene assays fail to detect different responses to HIF-1α and HIF-2α clearly displayed by endogenous gene assays (Figures 2C and and3,3, B and C). However, the reporter gene studies suggest that HIF-2α can bind to the PGK HRE in the context of an artificial or more natural episome.
Whereas reporter gene studies indicated that HIF-2α has access to PGK HREs in a plasmid, it is still possible that HIF-2α is unable to bind to the chromosomal PGK promoter. ChIP experiments were therefore performed to directly investigate whether HIF-2α could bind to the PGK HRE in vivo. ES cells were selected due to the availability of Hif-1α−/− and Hif-2α−/− ES cells, which served as useful negative controls for ChIP. In addition, we reported previously that HIF-2α binds to its target genes GLUT-1 and VEGF in WT and Hif-1α−/− ES cells (Hu et al., 2006 ). Anti-HIF-1α or HIF-2α antibodies were used to precipitate cross-linked HIF-1α or HIF-2α protein, and the amount of coprecipitated HRE-containing genomic DNA fragments from PGK and LDHA promoters was assessed by SYBR Green-based Q-PCR (Figure 5A), as well as regular PCR by using 32P-labeled dCTP (Figure 5B). Isotype-matched control antibodies generated similar background signals from WT, Hif-1α−/−, Hif-2α−/−, and Hif-1α−/−/HIF-2αFlag ES cells, whereas anti-HIF-1α antibody enriched the HRE-containing PGK and LDHA genomic DNA fragments only from WT and Hif-2α−/− ES cells. This enrichment was not observed in HIF-1α-deficient ES cells, including Hif-1α−/− and Hif-1α−/−/HIF-2αFlag ES cells (Figure 5A), suggesting that HIF-1α antibodies specifically immunoprecipitated HIF-1α–associated genomic DNA fragments, and more importantly that HIF-1α protein interacted with the HREs of PGK and LDHA in hypoxia-treated HIF-1α–expressing cells. This was consistent with the fact that HIF-1α protein stimulated PGK and LDHA gene expression in WT and Hif-2α−/− ES cells (Figure 3C). ChIP with anti-HIF-2α antibody coprecipitated HRE-containing genomic DNA fragments from PGK and LDHA in HIF-2α expressing hypoxia-treated WT, Hif-1α−/−, and Hif-1α−/−/HIF-2αFlag cells, but not in Hif-2α−/− ES cells (Figure 5A). Hif-1α−/−/HIF-2αFlag cells exhibited more HIF-2α binding to the HREs in comparison with that in Hif-1α−/− ES cells, which was consistent with higher levels of HIF-2α protein expression in these cells (Figure 5A). In agreement with SYBR Green Q-PCR detection, primers covering the entire LDHA HRE also detected HIF-1α or HIF-2α binding to the LDHA promoter in cells with appropriate HIF-α protein expression (Figure 5B). The ChIP results indicated that although HIF-2α does not activate the expression of the HIF-1α target genes PGK and LDHA, endogenous HIF-2α, like HIF-1α, binds to PGK and LDHA HREs. The ChIP data are also consistent with reporter gene studies showing that HIF-2α binds to the HREs and activates the PGK reporter. Moreover, these results clearly eliminated the possibility that failure of HIF-2α to induce PGK expression is due to the inability of HIF-2α to bind to endogenous PGK HREs.
The ChIP and reporter assays described above demonstrated that HIF-2α binds the promoters of HIF-1α unique genes, suggesting that the N-terminal bHLH and PAS domains of HIF-1α and HIF-2α are functionally interchangeable. To directly compare the N termini of HIF-1α and HIF-2α, we deleted the C-terminal transactivation domains from full-length (Figure 6A, HIF-1αTM and HIF-2αTM) to generate HIF-1αN (aa1–aa364) and HIF-2αN (aa 1–aa366) mutants of HIF-α that contained intact DNA binding and dimerization (bHLH and PAS) domains, but lacked the transactivation domains. HIF-1αN/VP16 and HIF-2αN/VP16 proteins were also generated by the addition of a potent transactivation domain from VP16 at the C termini of HIF-αN (Figure 6A), allowing us to investigate HIF-αN–mediated transcriptional activation. The HIF-αN and HIF-αN/VP16 constructs were Flag-tagged at their C termini to facilitate their detection. We first tested the expression and subcellular localization of the proteins produced from these constructs in normoxic HEK293 cells. Anti-Flag Western blot analysis detected their expression in HEK293 cells transfected with the indicated plasmids (Figure 6B). Although HIF-2αN and HIF-2αN/VP16 had similar numbers of amino acids to their HIF-1α counterparts, HIF-2αN and HIF-2αN/VP16 migrated more slowly, likely due to different posttranslational modification between HIF-1α and HIF-2α proteins. All proteins were detected in nuclear fractions under normoxia (Figure 6B), consistent with these constructs containing a nuclear localization signal in their bHLH domains and lacking an oxygen-dependent degradation domain (Kallio et al., 1998 ). All constructs were then tested for transcriptional activity in transient transfection assays by using a WT-HRE reporter gene in BpRc-1 cells with or without ARNT function. As expected, both full-length normoxia-functional HIF-α proteins (HIF-1αTM and HIF-2αTM) activated the reporter construct in an ARNT-dependent manner (Figure 6C, 1αTM and 2αTM), whereas HIF-1αN and HIF-2αN proteins exhibited no transcriptional activities (Figure 6C, 1αN and 2αN), likely due to the lack of a transcriptional activation domain. Importantly, HIF-1αN/VP16 and HIF-2αN/VP16 exhibited strong ARNT-dependent transcriptional activities (Figure 6C, 1αN/VP and 2αN/VP). These results indicated that HIF-αN/VP16 and ARNT dimers, not HIF-α/VP16 itself, activate WT-HRE reporter gene expression, providing strong evidence that HIF-αN proteins dimerize with ARNT and bind to the HREs of HIF target genes.
HIF-1αN and HIF-2αN proteins lack transcriptional activation domains, but they are capable of interacting with ARNT and binding to HREs. If HIF-1αN and HIF-2αN are interchangeable, HIF-2αN (like HIF-1αN) should be able to inhibit full-length HIF-1α protein-mediated induction of HIF-1α unique genes (e.g., PGK), as well as HIF-1α/HIF-2α common targets (e.g., GLUT-1) by occupying the HREs in a nonproductive way. As expected, full-length HIF-1α induced both PGK and GLUT-1 reporter genes in HEK293 cells, whereas HIF-1αN or HIF-2αN exhibited no effect on reporter gene activities (Figure 7A, top and middle). Introduction of increasing amounts of HIF-1αN decreased full-length HIF-1α–mediated activation of PGK and GLUT-1 reporters proportionally (Figure 7A, top and middle), demonstrating the feasibility of our experimental design. Interestingly, HIF-2αN also decreased full-length HIF-1α–mediated PGK and GLUT-1 reporter expression in a dose-dependent manner (Figure 7A, top and middle), suggesting that HIF-2αN and HIF-1αN performed similar functions in these reporter assays. Cotransfection of ARNT did not relieve HIF-αN–mediated repressive effects on full-length HIF-1α, suggesting that HIF-αN repressed full-length HIF-1α through competitive binding to the HREs of reporter genes, but not via competitive interaction with ARNT protein (unpublished data). This was further confirmed by results demonstrating that HIF-1αmbHLH and HIF-2αmbHLH proteins fail to inhibit full-length HIF-1α function in identical reporter assays (Figure 7A, bottom), because these mutants maintain their ARNT binding capability, but they are unable to bind DNA (Hu et al., 2006 ). We concluded that HIF-2αN behaved identically to HIF-1αN in transient reporter gene assays.
Next, HIF-1αN/VP16 and HIF-2αN/VP16 proteins were tested for their ability to stimulate HIF reporter genes (Figure 7B) as well as endogenous target genes in HEK293 cells (Figure 7C). As expected, both HIF-αN/VP16 proteins exhibited similar activities in promoting GLUT-1 reporter expression in HEK293 cells (Figure 7B, right). Interestingly, HIF-1αN/VP16 and HIF-2αN/VP16 proteins also activated the PGK reporter gene to similar levels (Figure 7B, left), although weaker PGK reporter induction by full-length HIF-2α protein was observed (Figure 7B, left). Consistent with stable transfection in 786-O/WT8 and ES cells (Figure 3, B and C), full-length HIF-1α activated all four endogenous HIF target genes in HEK293 cells, whereas HIF-2α induced the expression of ADM and GLUT-1, but not PGK and LDHA (Figure 7C, 1αTM and 2αTM). However, HIF-2αN/VP16 protein (like HIF-1αN/VP16) activated the endogenous HIF-1α/HIF-2α common target genes ADM and GLUT-1, as well as the HIF-1α unique target genes PGK and LDHA (Figure 7C, 1αN/VP and 2αN/VP). This was in direct contrast to our finding that full-length HIF-2α was unable to induce PGK and LDHA (Figure 6C, 2αTM), suggesting that regions outside of the DNA binding and dimerization domains are responsible for the inability of HIF-2α to regulate PGK and LDHA. The identical repressive functions of HIF-1αN and HIF-2αN, and similar positive activities of HIF-1αN/VP16 and HIF-2αN/VP16, suggest that the N termini of HIF-1α and HIF-2α (bHLH and PAS domains) play no critical roles in HIF target gene specificity by using reporter gene and endogenous gene assays.
The above-mentioned results indicate that differences in DNA binding and dimerization domains are not sufficient for HIF-α target gene specificity, suggesting that other regions of the protein must be responsible, such as the more diverse transcriptional activation domains. The C-termini of HIF-1α and HIF-2α contain two TADs (N- and C-TADs) and an inhibitory region flanked by the two TADs (Figure 1) (Jiang et al., 1997 ; Pugh et al., 1997 ; Maemura et al., 1999 ; O'Rourke et al., 1999 ). The N-TAD of HIF-1α is located approximately from residues 360–600 (Jiang et al., 1997 ) (Pugh et al., 1997 ), whereas the HIF-2α N-TAD spans aa500 to aa582 (O'Rourke et al., 1999 ) or aa500 to aa639 (Maemura et al., 1999 ). The C-TADs located at the C termini of both proteins (aa786–aa826 for HIF-1α and aa828–aa870 for HIF-2α) are 69% identical (Jiang et al., 1997 ; Pugh et al., 1997 ; Maemura et al., 1999 ; O'Rourke et al., 1999 ). Although HIF transcriptional activity depends on these domains, the relative contribution of the N-TAD and C-TAD to HIF function is unclear. To address this question, a number of HIF-α deletion mutants (lacking the N-TAD, C-TAD, or inhibitory domain) were generated, and their activities in activating HRE-reporter genes and several endogenous HIF-α target genes were analyzed. HIF-1α deletion mutants were derived from a full-length, normoxia-active HIF-1α construct (Figure 8A, left, HIF-1αTM Myc tagged) by deleting aa365–aa587 (ΔN-TAD-S, S for small), aa365–aa659 (ΔN-TAD-L, L for large), aa660–aa782 (ΔIH, IH for inhibitory domain), or aa783–aa836 (ΔC-TAD). HIF-2α deletion mutants were generated by deleting aa366–aa541 (Δ-N-TAD-S), aa366–aa618 (Δ-N-TAD-L), aa619-aa820 (Δ-IH), or aa821–aa874 (Δ-C-TAD) from full-length, normoxia-functional HIF-2α (HIF-2αTM) (Figure 8A, right). Two versions of N-TAD deletion mutants (small or large) were created to control for a possible inconsistency for HIF-2α N-TAD as reported previously (Maemura et al., 1999 ; O'Rourke et al., 1999 ). We initially checked the expression and cellular localization of these deletion constructs by transfecting them into HEK293 cells. All plasmids produced normoxia-stable proteins with the anticipated size, as detected by anti-Myc Western blot analysis (Figure 8B). Although the amount of transfected plasmids remained constant, ΔIHs and ΔC-TADs of HIF-1α and HIF-2α consistently produced more protein than other constructs (Figure 8B). Importantly, all HIF-α deletion mutants translocated to the nucleus as they were readily detected in the nuclear fractions of transfected cells (Figure 8B).
The deletion constructs were then tested for their transcriptional function in transient transfection assays by using a WT-HRE-Luc in HEK293 cells (Figure 8C). In comparison with full-length normoxia-active HIF-1α protein, deletion of the N-TAD (both small and large) reduced transcriptional activity (50% of the full-length), suggesting that the N-TAD is an important activator for full-length HIF-1α; however, ΔN-TAD mutants (retaining the C-TAD) exhibited substantial activity, suggesting that the C-TAD alone is still functional and contributes to full-length HIF-1α activity. In agreement with previous reports (Jiang et al., 1997 ; Pugh et al., 1997 ), inhibitory domain deletion (ΔIH) enhanced the activity of HIF-1α (150% of the full length) (Figure 8C, WT-HRE, HIF-1α). HIF-2α deletion mutants exhibited a similar pattern of change in activity, although deletion of HIF-2α N-TAD seemed to have less effect than HIF-1α N-TAD deletion (75% of the full length of HIF-2α) (Figure 8C, WT-HRE, HIF-2α), indicating that either the HIF-2α N-TAD or C-TAD was sufficient to maintain almost full activity of the full-length HIF-2α. The activities of these mutants were further tested using the above-described GLUT-1 and PGK reporters (Figure 4B). Deletion of N-TAD alone, but not the C-TAD of HIF-α, reduced HIF activity, whereas IH deletion increased their activity (Figure 8C, Glut-1 and PGK). Thus, in the reporter assays, the N-TAD is sufficient to maintain full-length activity without the C-TAD, whereas the C-TAD alone exhibits a 25–50% reduction in the absence of N-TAD, suggesting both the N-TAD and C-TAD can function independently and the N-TAD likely contributes more than the C-TAD to gene activation.
To further study the relative contribution of HIF-α C-terminal functional domains to HIF-α activity, all deletion mutants were individually transfected into HEK293 (Figure 9, A and B) and Hep3B cells (Figure 9C). Endogenous HIF target gene expression in transfected cells was measured by Q-PCR and presented as relative levels compared with cells transfected with an empty plasmid (Figure 9). HIF-1α target genes LDHA and PGK, HIF-2α targets PAI-1 and Cited-2, and HIF-1α/HIF-2α common targets ADM and NDRG-1 were selected for their broad expression patterns and differential response to HIF-1α or HIF-2α activity. ADM was similarly activated by full-length HIF-1α or HIF-2α in HEK293 cells (Figure 9A). Interestingly, deletion of either the C-TAD or N-TAD (less effect than C-TAD) of HIF-α proteins reduced the capability of HIF-α in ADM induction in HEK293 cells, suggesting either N-TAD or C-TAD can activate ADM expression. However, no significant activity enhancement was observed when the IH domain was deleted (Figure 9A, ΔIH). Another HIF-α common target, NDRG-1, was also similarly activated by either HIF-1α or HIF-2α (Figure 9A, NDRG-1). However, deletion of the N-TAD, (but not the C-TAD) reduced NDRG-1 expression, whereas IH deletion seemed to elevate NDRG-1 levels. Of note, N-TAD deletion mutants (retaining the C-TAD) still exhibited >50% activity in comparison with full-length HIF-α. These results suggested that both the N-TAD and C-TAD are important for HIF activity in activating HIF-1α/HIF-2α common targets NDRG-1 and ADM. The glycolytic genes LDHA and PGK were exclusively induced by HIF-1α in HEK293 cells (Figure 9B). Interestingly, induction of these HIF-1α target genes was totally dependent on the HIF-1α N-TAD, because N-TAD deletion (but not the C-TAD or IH) eliminated the ability of HIF-1α to activate these genes (Figure 9B).
HIF-2α has been shown to specifically activate Cited-2 and PAI-1 expression in HEK293 (Wang et al., 2005 ; Aprelikova et al., 2006 ) and A549 adenocarcinoma cells (Sato et al., 2004 ). However, HIF-2α induced Cited-2 expression 1.5-fold in HEK293 cells, making it difficult to assess the relative activity of HIF-2α mutants here (unpublished data). Hep3B cells were selected to assess the relative contribution of HIF-2α N- and C-TADs to HIF-2α activity, because these cells exhibit high levels of HIF target gene inducibility (Hu et al., 2003 ). HIF-1α or HIF-2α deletion mutants were transfected into Hep3B cells, and PAI-1 and Cited-2 expression levels were analyzed. Full-length HIF-2α activated PAI-1 (8.0-fold) and Cited-2 (2.5-fold) expression to higher levels than HIF-1α (2.0-fold for PAI-1 and 1.3-fold for Cited-2), consistent with the idea that these genes are primarily regulated by HIF-2α (Figure 9C). In contrast to the results described above for ADM, NDRG-1, LDHA, and PGK, ΔN-TAD-S and ΔN-TAD-L exhibited a significant difference in PAI-1 and Cited-2 activation, suggesting the region between aa541 and aa618 of HIF-2α is critical for regulation of these HIF-2α targets (Figure 9C). Deletion of the HIF-2α C-TAD did not have a significant effect on the expression of the two HIF-2α–specific targets (Figure 9C), suggesting that the N-TAD plays a more important role in regulating HIF-2α target genes. In summary, analysis of HIF-α deletion mutants suggested that both the N-TAD and C-TAD are important for HIF-α common target genes; however, the N-TAD of HIF-1α or HIF-2α plays a dominant role in activating HIF-1α– or HIF-2α–specific genes.
The data described above indicated that HIF-α inhibitory domains inhibit HIF-α activity and that they are not involved in target gene specificity. Therefore, we divided the C-terminal halves of HIF-α into two parts, named “N-TAD” for proteins containing aa415–aa659 for HIF-1α and aa418–aa619 for HIF-2α, and “C-TAD” for proteins containing aa660–aa836 for HIF-1α and aa620–aa874 for HIF-2α. Here, the C-TAD also contains the IH domains of HIF-1α and HIF-2α (Figure 10A). To further characterize the domain important for target gene specificity, several HIF-1α/HIF-2α domain swap constructs were made, as diagrammed in Figure 10A. The “122” protein contains the HIF-1α N-terminal DNA binding and dimerization domains (HIF-1α aa1-aa415) and HIF-2α C terminus (HIF-2α aa419-aa874), whereas the “211” protein contains the HIF-2α N-terminus (HIF-2α aa1-aa418) and HIF-1α C-terminus (HIF-1α aa416-aa836). These two constructs were designed to test the contribution of N-terminal or C-terminal halves of HIF-α to target gene specificity. If HIF-α C-terminal halves were critical, 122 should behave like HIF-2α. To evaluate the relative roles of N-TAD versus C-TAD in target gene regulation, 112 (HIF-1α aa1-aa659 and HIF-2α aa620-aa874) and 221 (HIF-2α aa1-aa619 and HIF-1α aa660-aa836) hybrids were generated (Figure 10A). All constructs were normoxia-functional, due to mutations in all three O2-sensitive amino acids (TM). Western blot analysis using anti-HIF-1α or HIF-2α antibodies demonstrated that all constructs produced proteins with the expected sizes (Figure 10B). Although both anti-HIF-1α and anti-HIF-2α antibodies used here recognized construct 112, these antibodies were not expected to recognize construct 221, which was the case (Figure 10B). Detection of these proteins in both cytoplasmic and nuclear fractions under normoxia was likely due to overexpression, because full-length HIF-1α and HIF-2α also exhibited similar subcellular localization (Figure 10B). All hybrid proteins were tested in transfection assays using the above-described WT-HRE reporter gene in HEK293 cells (Figure 10C). In general, the hybrid and full-length proteins exhibited similar transactivation activities with the exception being 112, which was the most active protein in these reporter assays. Although no correlation was observed between specific domains and relative activity in reporter gene assays, the results suggested that they bound the promoter and activated reporter gene expression. In addition, construct 221 was functional in reporter gene assays, suggesting that this construct also produced active protein (Figure 10C).
To identify the domain important for target gene specificity, HIF-1α/HIF-2α hybrid constructs were transfected into HEK293 cells and endogenous HIF target genes (HIF-1α unique genes PGK and LDHA and HIF-1α/HIF-2α common genes NDRG-1 and ADRP) were analyzed by Northern blot hybridization of RNA isolated from HEK293 cells 36 h posttransfection (Figure 11A). In comparison with control GFP plasmid, WT HIF-1α protein induced all four analyzed genes, particularly LDHA, NDRG-1, and ADRP (Figure 11A, lane 1). Consistent with the idea that HIF-1α DM (double proline mutation) is more stable than WT HIF-1α, higher levels of target gene induction were observed in HIF-1α DM-expressing cells (Figure 11A, lane 2). However, no significant difference was observed between HIF-1α DM and HIF-1α TM (normoxia-active, triple mutations) (Figure 11A, lane 3), likely due to any FIH inhibitory effects of being overcome in these overexpression assays (Mahon et al., 2001 ; Lando et al., 2002 ). Interestingly, HIF-2α WT, DM, and TM exhibited similar activity during target gene induction (Figure 11A, lanes 4, 5, and 6), likely due to WT HIF-2α being more stable than WT HIF-1α under normoxia (Hu et al., 2003 ). Importantly, although HIF-2α activated the expression of the HIF-1α/HIF-2α common targets NDRG-1 and ADRP, HIF-2α was unable to induce PGK gene expression in HEK293 cells (Figure 11A). Moreover, HIF-2αTM exhibited limited induction of LDHA compared with HIF-1αTM (Figure 11A). Therefore, PGK, LDHA, and ADRP, but not NDRG-1 (similarly induced by HIF-1α and HIF-2α), were good candidate genes to study HIF target gene specificity in HEK293 cells.
All normoxia-active HIF-1α/HIF-2α hybrid constructs were introduced into HEK293 cells to test their ability to activate endogenous HIF target gene expression. 122 and 211 proteins exhibited similar transcriptional properties to full-length HIF-2α and HIF-1α, respectively (Figure 11A, lane 8 for 122, and lane 10 for 211). This demonstrated that the C-terminal, not the N-terminal halves of HIF-α proteins, played a critical role in target gene specificity. Furthermore, 112 and 221 proteins behaved equally to HIF-1α and HIF-2α respectively, suggesting that the N-TAD, not the C-TAD, provided HIF target gene specificity (Figure 11A, lane 9 for 112 and lane 11 for 221). In addition, plasmids containing the HIF-1α N-TAD such as 112 and 211 exhibited HIF-1α specificity, indicating that the HIF-1α N-TAD determines HIF-1α target gene specificity.
Most HIF target genes analyzed in HEK293 cells are either regulated similarly by both HIF-1α and HIF-2α, or preferentially by HIF-1α (Figure 11A). Therefore, we used Hep3B cells for HIF-2α preferential genes (Figure 11B). Q-PCR was used to analyze several HIF-2α preferential target genes in Hep3B cells in response to HIF-1αTM, HIF-2αTM, or HIF-1α/HIF-2αTM hybrids (Figure 11B). In addition, the HIF-1α target genes PGK and LDHA were also investigated (Figure 11B). Consistent with results from HEK293 cells (Figure 11A), PGK was exclusively activated by proteins containing HIF-1α N-TAD (HIF-1α, 112, and 211), and not by proteins lacking HIF-1α N-TAD (HIF-2α, 122, and 221) in Hep3B cells (Figure 11B). Similarly, the HIF-1α target LDHA was stimulated by HIF-1α and HIF-1α N-TAD containing hybrids 112 and 211 (Figure 11B). These results confirmed that the N-TAD of HIF-1α is required for HIF-1α unique gene expression in another cell type.
PAI-1 and ADM were highly responsive to hypoxia and primarily regulated by HIF-2α in Hep3B cells as determined by siRNA and transfection studies (Figures 2B and and9C).9C). Consistently, these genes were also preferentially stimulated by HIF-2α upon transient transfection of Hep3B cells (Figure 11B). Importantly, hybrids containing HIF-2α N-TAD (122 and 221) also exhibited stronger induction of PAI-1 and ADM than hybrid constructs containing HIF-1α N-TAD (112 and 211) (Figure 11B), indicating that HIF-2α N-TAD was important for HIF-2α target gene specificity. Thus, by analyzing endogenous HIF target genes in two cell lines, we concluded that the N-TAD, and not the DNA binding domain or C-TAD, correlated well with differential transcriptional activity of HIF-α proteins.
The above-mentioned data indicated that the HIF-α N-TAD is critical for HIF target gene specificity, although both the N-TAD and C-TAD are required for maximal HIF activity for some HIF common target genes. Transcriptional activation domains frequently mediate interaction with other transcriptional cofactors. A coactivator, the ETS transcription factor ELK, has recently been shown to interact with HIF-2α protein, and its expression is required for hypoxic induction of Cited-2, Wisp2, and IGFBP3 in MCF7 breast cancer cells (Aprelikova et al., 2006 ). Of note, endogenous ELK protein interacts with endogenous HIF-2α (and not HIF-1α) in both MCF7 and 786-O cells. To gain further insight into HIF-1α/HIF-2α target gene specificity, we evaluated the effects of ELK reduction (by siRNA) on HIF-1α or HIF-2α target genes. Hep3B cells were transfected with siRNAs to HIF-1α, HIF-2α, or ELK and levels of several HIF-1α or HIF-2α target gene under normoxia and hypoxia were analyzed. siRNAs to HIF-1α, HIF-2α, or ELK specifically decreased their protein levels (Figure 12A). As expected, hypoxic induction of HIF-2α specific targets Cited-2, Epo, and PAI-1 was reduced when Hep3B cells were targeted with HIF-2α siRNA, but not HIF-1α siRNA, whereas the levels of HIF-1α targets Glut-1, LDHA, and PGK were diminished in hypoxic Hep3B cells transfected with HIF-1α siRNA (Figure 12B). Interestingly, ELK knockdown reduced the levels of HIF-2α specific targets Cited-2, Epo, and PAI-1, but not the HIF-1α targets including Glut-1, LDHA, and PGK (Figure 12B). Furthermore, ELK siRNA also failed to reduce HIF-2α–dependent expression of IGFBP-1 and several HIF-2α preferential genes such as ADM and NDRG-1 (unpublished data), suggesting ELK is a promoter-dependent, not general regulator for HIF-2α. These results indicated that HIF-2α can achieve target gene specificity, by specifically interacting with ELK or other transcriptional coactivator or repressor proteins.
All transcription factors, including HIF-1α and HIF-2α, have DNA binding domains recognizing specific sequences in the promoters of target genes and transcriptional activation domains interacting with cofactors to stimulate transcription initiation and/or elongation (Blau et al., 1996 ). Related transcription factors are often grouped based on their similar DNA binding domains; thus, members of the same transcription factor family often regulate similar target genes by binding to conserved cis-elements on their promoters or enhancers. The bHLH-PAS transcription factors HIF-1α and HIF-2α have been shown to similarly regulate multiple reporter genes via their conserved bHLH-PAS domains (Ema et al., 1997 ; Flamme et al., 1997 ; Tian et al., 1997 ; Wiesener et al., 1998 ). In spite of these significant similarities in their DNA binding and dimerization domains, we and others demonstrated that HIF-1α and HIF-2α have unique as well as common target genes when endogenous HIF target genes were analyzed (Hu et al., 2003 ; Grabmaier et al., 2004 ; Warnecke et al., 2004 ; Raval et al., 2005 ; Wang et al., 2005 ; Covello et al., 2006 ), suggesting that some differences in their DNA binding and/or transcriptional activation domains determine target gene selection. This report demonstrates that the N-TAD confers target gene specificities of HIF-1α and HIF-2α, whereas the C-TAD promotes the expression of HIF-1α/HIF-2α common target genes.
Our results identify several important features controlling HIF-α target gene specificity. We found that some HIF target gene activity is highly cell type dependent. For example, GLUT-1 is specifically regulated by endogenous HIF-1α in Hep3B cells (Figure 2C) and mouse ES cells (Figure 3C), but regulated by endogenous HIF-2α in 786-O/WT8 cells (Figure 3B). We propose that genes such as GLUT-1 are HIF-1α/HIF-2α common targets because both proteins are capable of regulating GLUT-1, although in different cell types.
We also found that there is some disconnection between results obtained from reporter gene assays and results obtained from analyses of endogenous target genes. Although HIF-2α is unable to regulate endogenous PGK gene expression in multiple cell types, HIF-2α stimulates PGK promoter-mediated reporters (Figure 4, A and C). In addition, HIF-1α/HIF-2α hybrids exhibit clear target gene specificities when endogenous target genes were analyzed (Figure 11); however, they are similarly active in reporter gene assays (Figure 10C). Furthermore, the HIF-α C-TAD seems to be disposable for reporter assays (Figure 8C), but required for optimal expression of some endogenous targets (Figure 9). Although there is a discrepancy between reporter and endogenous gene assays, reporter gene assays clearly indicate that HIF-1α and HIF-2α are capable of binding HRE DNA elements. In addition, some reporter genes can reflect physiological responses to HIF proteins if the reporter constructs are long enough to include most critical cis-acting elements or if the essential regulatory elements are concentrated in the short promoter fragments cloned into the reporter (Ema et al., 1997 ; Tian et al., 1997 ). Therefore, loss of multiple critical cis-acting elements in promoter constructs might be responsible for broader HIF-1α and HIF-2α activity in activating reporter genes. This is consistent with our observation that the synthetic WT-HRE reporter (PGK derived HRE upstream of the TK promoter) is equally stimulated by HIF-1α and HIF-2α, whereas a more natural PGK reporter is activated by HIF-1α better than HIF-2α (Figure 4C). Nevertheless, the endogenous PGK gene is exclusively regulated by HIF-1α (Figures 2, ,3,3, ,7,7, and and1111).
We have used multiple assays to show that the N-terminal halves of HIF-α are not critical for HIF target gene specificity. These results included HIF-2α binding to HIF-1α target gene promoters, similar HIF-1αN/VP16 and HIF-2αN/VP16 functions in target gene induction, and identical activities of the hybrid proteins 211 and 122 in comparison with full-length HIF-1α and HIF-2α, respectively. However, these results do not exclude the possibility that HIF-1α/HIF-2α use different HBSs in the HREs of HIF target genes. In fact, several lines of experimentation indirectly support this idea. For example, VEGF, NDRG-1, and ADM are HIF-1α/HIF-2α common target genes. If HIF-1α and HIF-2α use the same HBS in these HREs, HIF common target gene expression would not be down-regulated in cells where one HIF-α subunit is reduced or eliminated. Our siRNA experiments indicated that the function of HIF-1α and HIF-2α in regulating HIF common target gene is synergistic, not competitive (Figure 2). This is consistent with the observation that the binding of HIF-1α (or HIF-2α) to HREs is not significantly enhanced by the deletion of its counterpart α subunit in ES cells (Figure 5). Furthermore, our results did not address whether other potential difference exist between the N termini of HIF-1α and HIF-2α. For example, a recent report demonstrated that PAS-B of HIF-1α, but not HIF-2α is phosphorylated. This phosphorylation difference explains why HIF-1α (not HIF-2α) inhibits the expression of the HIF target NBS1 (To et al., 2006 ).
Both HIF-1α and HIF-2α contain two transcriptional activation domains (N-TAD and C-TAD) (Jiang et al., 1997 ; Pugh et al., 1997 ; Maemura et al., 1999 ; O'Rourke et al., 1999 ). To assess the relative contribution of the N-TAD and C-TAD in HIF-α activity and target gene specificity, we generated a number of deletion mutants of HIF-1α and HIF-2α, as well as HIF-1α/HIF-2α hybrid proteins. By analyzing a select group of genes, we determined that both TADs are required for optimal HIF common target gene induction; however, the N-TAD seems to be more important for unique HIF-α activity. Our results are consistent with several reports indicating that genes such as the HIF-1α target PGK is activated by PHD2 siRNA, but not by FIH siRNA under normoxia (Bracken et al., 2006 ; Dayan et al., 2006 ; Elvidge et al., 2006 ).
The HIF deletion mutants and HIF-1α/HIF-2α hybrid proteins are normoxia-functional due to mutagenesis of three O2-sensitive amino acids (two prolines in the N-TAD and one asparagine in the C-TAD). This allows us to analyze their function under normoxia and to avoid endogenous HIF activity when hypoxic treatment is performed. One potential problem of this approach is that mutagenesis of O2-sensitive amino acids may not be sufficient to confer full activity to the N-TAD or C-TAD under normoxia. To address this concern, we generated hybrid proteins containing the Gal4 DNA binding domain and full-length HIF-1αTM or HIF-2αTM, as endogenous HIF protein would not induce the Gal4-Luc reporter. We determined that hypoxic treatment did not increase the activity of the Gal4/HIF-TM proteins (Supplemental Figure 1). In addition, we also examined the function of HIF-αTM and their deletion mutants (N-TAD, C-TAD, or both TADs) under both normoxia and hypoxia in Hif-1−/− ES cells, as these cells lack HIF-1α protein and endogenous HIF-2α is not functional (Hu et al., 2006 ). Again, we found that hypoxic treatment did not further enhance the activity of these proteins during transactivation of cotransfected WT-HRE Luc reporters (Supplemental Figure 2). Collectively, these results demonstrate that mutation of three O2-sensitive amino acids renders the N-TAD and C-TAD fully functional under normoxia. These supplemental data support our conclusion that the N-TAD is the principle transactivation domain responsible for target gene specificity of HIF-1α or HIF-2α; the C-TAD alone activates some HIF-1α/HIF-2α common targets such as ADM and NDRG-1 (Figure 9A).
As a transcriptional activation domain, the N-TAD likely serves as an interaction site for important cofactors. Indeed, several reports indicate that transcriptional cooperation between HIF-α and other factors is required for hypoxic induction of multiple HIF target genes. For example, EPO expression is enhanced by interaction between HIF-1α and Smad3/4 (Sanchez-Elsner et al., 2004 ), whereas transcription of VEGF Receptor-2 (FLK-1) depends on association between HIF-2α and ETS-1 (Elvert et al., 2003 ). We have now determined that ELK is required for HIF-2α to activate several HIF-2α–specific targets in Hep3B cells. These results are consistent with a recent report indicating that ELK specifically interacts with HIF-2α (but not HIF-1α) at the promoters of two HIF-2α–dependent genes: Cited-2 and WISP2 (Aprelikova et al., 2006 ). In aggregate, HIF-2α interaction with multiple ETS family members such as ETS-1, ETS-2, GABP, and ELK may confer target gene selection to the HIF-2α subunit in numerous cell types (Aprelikova et al., 2006 ).
In summary, we have further delineated target gene specificity of the two closely related transcription factors HIF-1α and HIF-2α. More importantly, we have used multiple deletion and domain swap mutants to investigate the relative contribution of different HIF-α functional domains in target gene regulation. These results collectively indicate that the N-terminal transcription domains of HIF-α are largely responsible for what genes are activated by HIF-1α versus HIF-2α. Finally, specific interaction between the HIF-α proteins with coregulator(s) such as ELK is a critical determinant of selective HIF target gene activation in hypoxic cells.
We thank members of the Simon laboratory, particularly Brain Keith, for thoughtful discussions and reading the manuscript. This work was supported by National Institutes of Health grant 66130 (to M.C.S.), the Howard Hughes Medical Institute, and the Abramson Family Cancer Research Institute. M.C.S. is an investigator of the Howard Hughes Medical Institute.
This article was published online ahead of print in MBC in Press (http://www.molbiolcell.org/cgi/doi/10.1091/mbc.E06-05-0419) on September 5, 2007.
The online version of this article contains supplemental material at MBC Online (http://www.molbiolcell.org).