Activity of Hypoxia-inducible Factor 2α Is Regulated by Association with the NF-κB Essential Modulator*

The hypoxia-inducible factors 1α (HIF-1α) and 2α (HIF-2α) are key regulators of the transcriptional response to low oxygen and are closely related in domain architecture, DNA binding, and activation mechanisms. Despite these similarities, targeted disruption of the HIF-α genes in mice results in distinctly different phenotypes demonstrating nonredundancy of function, although the underlying mechanisms remain unclear. Here we report on the novel and specific interaction of HIF-2α, but not HIF-1α, with the NF-κB essential modulator (NEMO) using immunoprecipitation, mammalian two-hybrid, and in vitro protein interaction assays. Reporter gene assays demonstrate that this interaction specifically enhances normoxic HIF-2α transcriptional activity, independently of the HIF-2α transactivation domain, consistent with a model by which NEMO aids CBP/p300 recruitment to HIF-2α. In contrast, HIF-2α overexpression does not alter NF-κB signaling, suggesting that the functional consequence of the HIF-2α/NEMO interaction is limited to the HIF pathway. The specificity of NEMO for HIF-2α represents one of the few known differential protein-protein interactions between the HIF-α proteins, which has important implications for the activity of HIF-2α and is also the first postulated NF-κB-independent role for NEMO.

Sim partner protein, the aryl hydrocarbon nuclear translocator ARNT (also called HIF-1␤). The heterodimer then binds hypoxic response elements (HREs) within the regulatory regions of a variety of target genes to up-regulate their expression (1). These genes typically function to both increase oxygen supply to tissues and facilitate metabolic adaptation to hypoxia.
HIF-1␣ and HIF-2␣ possess 48% amino acid sequence similarity, overlapping expression patterns, and a common requirement for ARNT to bind DNA and up-regulate expression of a largely shared group of target genes. In addition, the activity of both proteins is controlled through hydroxylation of key proline and asparagine residues in the conserved oxygen-dependent degradation domain and transactivation domain (TAD), respectively, regulating protein stability and transactivation potential (2)(3)(4). However, despite these similarities, HIF-1␣-and HIF-2␣-deficient mice manifest distinct phenotypes and hence have nonredundant functions. HIF-1␣-deficient mice display defective vasculature and mesenchymal cell death (5,6). In contrast, HIF-2␣-deficient mice are reported to manifest defective vascular remodelling (7) or decreased catecholamine (8), lung surfactant (9), or hematopoietic cell production (10). Most recently, HIF-2␣-deficient mice have been reported to exhibit a multiple organ pathology, consistent with mitochondrial disease (11). Concomitant with these phenotypes, there is a growing body of evidence demonstrating different responses between the HIF-␣ proteins in relation to cellular localization, hypoxic responsiveness, and target gene expression (12)(13)(14). Despite this information, however, little is known of the mechanisms underlying these functional differences.
In an effort to further understand differential HIF-␣ regulation, a yeast two-hybrid screen was performed using as bait a C-terminal portion of HIF-2␣, which represents the region of greatest sequence divergence between the HIF-␣ proteins (mHIF-2␣ residues 565-820). The NF-B essential modulator (NEMO, also called IKK-␥ and IKK complex-associated protein-1) was identified as a HIF-2␣-interacting protein. Mammalian two-hybrid, GST pull-down, and co-immunoprecipitation assays demonstrated that the interaction of NEMO is specific for HIF-2␣ and not HIF-1␣. Furthermore, we demonstrate that this physical interaction has a functional effect, specifically increasing the normoxic activity of HIF-2␣, but not HIF-1␣, in a dose-dependent manner, consistent with a model by which NEMO aids CBP/p300 recruitment to HIF-2␣.
Yeast Two-hybrid Screening-Yeast two-hybrid screening of 1.5 ϫ 10 6 transformants from a E10.5 whole mouse cDNA library (Invitrogen) was performed using the Proquest Matchmaker two-hybrid system (Invitrogen). Initial selection was performed on ϪLeu/ϪTrp/ϪHis ϩ 30 mM 3-aminotriazole plates, prior to more stringent analysis with the two additional reporters.
Cell Culture and Reporter Assays-HEK 293T and Jurkat cells were routinely grown at 37°C, 5% CO 2 in Dulbecco's modified Eagle's medium (Invitrogen) or RPMI (Invitrogen), respectively, supplemented with 10% fetal calf serum. Transient transfections into 40 -60% confluent 293Ts or 2 ϫ 10 5 Jurkat cells/well (24 well tray format) were performed using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. The amounts of DNA transfected are given within the figure legends. The total amount of DNA transfected into each well was normalized with the addition of appropriate empty vector (pEF-IRES-puro or pEF-BOS). Luciferase activity was determined using the dual luciferase reagent assay system (Promega) according to the manufacturer's instructions, using RLTK-luciferase as a transfection control.

Identification of NEMO as a HIF-2␣-interacting Protein-To
identify novel HIF-2␣-interacting proteins, amino acids 565-820 of mouse HIF-2␣ fused to the Gal4 DNA-binding domain (Gal4DBD) were used as bait in a yeast two-hybrid screen of a mouse E10.5 cDNA library. This region of HIF-2␣ represents the area of greatest amino acid sequence divergence between the HIF-␣ proteins (Fig. 1A). The screening of 1.5 ϫ 10 6 separate transformants yielded 33 clones positive for each of the Ura, LacZ, and His reporter genes. Among these, full-length NEMO was identified as a candidate-interacting protein.
NEMO is required for NF-B activation by virtually all known stimuli, acting as a nonenzymatic molecular scaffold to assemble the IKK complex containing both IKK-␣ and IKK-␤ kinases (21)(22)(23)(24)(25)(26)(27)(28). This complex is then able to phosphorylate and initiate the ubiquitylation and degradation of the inhibitory IB proteins in response to NF-B-activating stimuli. It is postulated NEMO may have additional scaffolding roles, binding upstream NF-B activators and recruiting them to the IKK complex.
To confirm the HIF-2␣/NEMO interaction in a mammalian system, two-hybrid assays were performed in HEK 293T cells after switching the identity of the fusion partners relative to the yeast system (Fig. 1B). Results demonstrate both HIF-2␣(565-820)-VP16 and full-length HIF-2␣ interact with Gal4DBD-NEMO under normoxic and hypoxic conditions. Strikingly, mammalian two-hybrid assays indicate that NEMO does not interact with the corresponding region of HIF-1␣(597-775) and interacts only very weakly with full-length protein.
NRP, or FIP2, is a Golgi-associated, cytokine-inducible protein with 53% amino acid similarity to NEMO (29,30). To confirm the specificity of the NEMO/HIF-2␣ interaction, NRP-GalDBD fusions were also analyzed by mammalian two-hybrid assay. The results demonstrate that the interaction between HIF-2␣ and NEMO is specific because NRP does not interact with HIF-2␣ (Fig. 1C).
The interaction of HIF-2␣ and NEMO within cells was confirmed using GST pull-down experiments with protein extracts from cells expressing Myc-tagged NEMO and HIF-2␣-GST fusion proteins. Using glutathione-agarose, HIF-2␣ (565-820)-GST, but not GST itself, effectively pulled down full-length Myc-tagged NEMO ( Fig. 2A, upper right panel). An unrelated Myc-tagged protein, mouse adiponectin receptor 1, was included as a control and was not pulled down by either HIF-2␣(565-820)-GST or GST alone. The specific pull-down of NEMO-Myc, but not adiponectin receptor-Myc, by HIF-2␣(565-820)-GST is not due to differences in expression levels of Myctagged proteins ( Fig To further characterize this interaction, pull-down experiments were performed using in vitro transcribed/translated HIF-␣ protein and mammalian cell extracts containing GST-NEMO. As anticipated, HIF-2␣(565-820) was pulled down strongly by NEMO(1-358)-GST relative to GST alone, whereas NEMO(1-358)-GST did not efficiently pull-down the equivalent region of HIF-1␣ (Fig. 2B, middle panel). In this experiment, HIF-1␣ was translated less efficiently than HIF-2␣; however, an overexposure of the autoradiogram demonstrates that the very low levels of HIF-1␣ pulled down by GST is nonspecific and not enhanced by NEMO (Fig. 2B, right panel). Furthermore, in multiple other experiments with varying levels of in vitro transcribed and translated HIF-1␣, only very low levels of nonspecific binding were ever observed. This parallels Fig. 1 in that NEMO specifically associates with HIF-2␣ but not HIF-1␣. The percentages of amino acid similarities between corresponding domains are indicated, as is the HIF-2␣ region used as bait in the yeast two-hybrid screening. B and C, 293T cells were transiently transfected with a Gal4 response element-driven reporter (5ϫGalRE-luciferase, 150 ng), pEF-bos expression vectors (100 ng), and the RLTK-luciferase internal control plasmid (10 ng), to which luciferase activity was normalized. 10 h post-transfection, the cells were exposed to hypoxia (Ͻ1% O 2 ) or retained at normoxia for an additional 14 h. The results, shown in fold induction over GalDBD alone, are in triplicate Ϯ S.D. and representative of three independent experiments.
ARNT immunoprecipitation was also able to co-precipitate full-length Myc-tagged NEMO when co-transfected with HIF-2␣ but not HIF-1␣ (Fig. 2C). This not only provides further evidence for the interaction of NEMO and HIF-2␣ in vivo but also suggests the possibility of NEMO being present within transcriptionally active HIF-2␣/ARNT heterodimeric complexes. Significantly, the overexpression of NEMO did not significantly affect the ability of ARNT to immunoprecipitate HIF-␣.
Finally, these experiments were all performed using overexpressed proteins. Therefore, it was critical to demonstrate the interaction between endogenous HIF-2␣ and NEMO. Because normoxic HIF-2␣ was not detectable in any cell line tested, includingmouseembryonicfibroblasts(12),weutilizeddipyridyl-treated HepG2 cells, which expressed considerable amounts of HIF-2␣ protein. The mammalian two-hybrid assays have previously demonstrated that the HIF-2␣/NEMO interaction can occur equally well in hypoxic or normoxic conditions (Fig.  1B). Immunoprecipitation experiments using ␣-NEMO antibodies clearly show that endogenous HIF-2␣ co-immunoprecipitates with endogenous NEMO, and this is specific because the use of a nonspecific antibody has no effect (Fig. 2D). In this experiment, immunoprecipitated NEMO and HIF-2␣ ran slightly higher in size to that observed in protein extracts, presumably because of the abundance of protein within immunoprecipitated samples. In conjunction with Fig. 1, these results indicate that in both yeast and mammalian cells, NEMO physically interacts with HIF-2␣ but not HIF-1␣.  1 and 4) represent 5% input. FIG. 3. HIF-2␣(565-820) interacts with NEMO(50 -358). A, schematic representation of the NEMO protein with predicted coiled coil (CC), leucine zipper (LZ), and zinc finger (ZnF) domains. The numbers indicate amino acids for mouse NEMO. B, 293T cells were transiently transfected with 5ϫGalRE-luciferase (150 ng), expression vectors (100 ng), and the RLTK-luciferase internal control plasmid (10 ng) to which luciferase was normalized. 10 h post-transfection cells were either exposed to hypoxia (Ͻ1% O 2 ) or retained at normoxia, for an additional 14 h. The results are given in triplicate Ϯ S.D. and representative of three independent experiments. C, whole cell extracts of 293T cells transiently transfected with GalDBD-NEMO deletion mutants (100 ng) were separated by SDS-PAGE (10% gel), transferred to nitrocellulose, and Western blotted with ␣-Gal antibodies.
Defining a Minimal HIF-2␣-binding Site on NEMO-Based upon the predicted domain architecture of NEMO from its primary amino acid sequence (Fig. 3A), a series of N-and C-terminal NEMO deletion mutants were fused to the Gal4 DNA-binding domain and analyzed for HIF-2␣ interaction via mammalian two-hybrid assays. Full interaction was maintained with removal of both the N-terminal 49 amino acids of NEMO and the C-terminal zinc finger domain (Fig. 3B, deletion mutant 50 -358). Further deletion at either end resulted in a dramatic loss of HIF-2␣ binding. Western analysis confirmed that the loss of reporter activity was not a result of decreased protein expression, because Gal4DBD-NEMO deletions (for ex-ample amino acids 1-302 and 196 -358), which bind HIF-2␣ poorly, expressed better than the 1-358 and 50 -358 constructs that bind HIF-2␣ with the same affinity as full-length NEMO (Fig. 3C). These data demonstrate that amino acids 50 -358 of NEMO are required for high affinity interaction with amino acids 565-820 of HIF-2␣.
NEMO Dose-dependently Increases HIF-2␣ Activity-To characterize the functional effect of the HIF-2␣/NEMO interaction, NEMO was overexpressed in HEK 293T cells co-transfected with a HRE luciferase reporter gene together with either HIF-2␣ or HIF-1␣ (Fig. 4A). HIF-␣ overexpression increased HRE activity at normoxia, presumably through titration of inhibitory hydroxylases. NEMO overexpression dose-dependently increased normoxic HIF-2␣-dependent reporter gene activity but had no significant effect with HIF-1␣. At the highest levels of NEMO, HIF-2␣-driven reporter expression increased to a maximum of ϳ3-fold. As a result of either a nonspecific NEMO overexpression effect or the potential weak interaction between NEMO and full-length HIF-1␣, transfection of highest NEMO levels also resulted in a small increase in HIF-1␣-dependent reporter activity. Hypoxia, a major stimulus of HIF-␣ activation, dramatically increased reporter activity, overriding the positive influence of NEMO at the lower HIF-2␣ levels in normoxia (Fig. 4A). NEMO expression similarly increased nor-moxic HIF-2␣, but not HIF-1␣, reporter activity in Jurkat cells in a dose-dependent manner, demonstrating that this effect is not specific for 293T cells (Fig. 4B).
To support the notion that the effect of NEMO overexpression on HIF-2␣ activity is a result of physical interaction, Gal4DBD-tagged NEMO deletion mutants were also assayed for their affect on HRE-driven reporter activity (Fig. 4C). In agreement with data in Fig. 3B, GalDBD-NEMO deletion mutants capable of HIF-2␣ binding up-regulated HIF-2␣ activity to similar extents, whereas an interaction-deficient NEMO deletion mutant (GalDBD-NEMO 1-302) did not significantly alter HIF-2␣ activity. As anticipated, none of the NEMO were transiently co-transfected with a luciferase reporter gene controlled by four NF-B response elements (4ϫNRE-luciferase, 100 ng) and expression plasmids for HIF-1␣ or HIF-2␣ (as indicated), together with RLTK-luciferase internal control (10 ng). The cells were maintained at normoxia for 18 h, and the results are given in triplicate Ϯ S.D. and representative of three independent experiments. C, 293T cells were transfected with 4ϫNRE-luciferase (100 ng), HIF-␣ expression plasmid (200 ng), and RLTKluciferase internal control (10 ng). 14 h post-transfection, the cells were treated with tumor necrosis factor ␣ (20 ng/ml) or phorbol 12-myristate 13-acetate (10 ng/ ml) and assayed for luciferase activity 8 h later.
constructs had a significant affect on HIF-1␣ activation. Taken together, these data demonstrate that NEMO is specifically able to up-regulate HIF-2␣ transcriptional activity on a HRE-driven reporter gene in a cellular context and that this is dependent upon its high affinity interaction with HIF-2␣.
HIF-2␣ Does Not Affect NF-B Activation-Because NEMO is essential for NF-B activation in response to virtually all known stimuli (21, 23), we assessed the effect of transient overexpression of HIF-␣ on a NF-B-responsive reporter gene. In both 293T (Fig. 5A) and Jurkat (Fig. 5B) cells, expression of neither HIF-1␣ nor HIF-2␣ significantly changed NF-B reporter activity under these assay conditions. Similarly, HIF-␣ expression did not have a major effect on NF-B activation by the known inducers tumor necrosis factor ␣ or phorbol 12myristate 13-acetate in 293T cells (Fig. 5C).
Endogenous NEMO Increases HIF-2␣ Activity-To investigate the influence of endogenous NEMO on HIF-2␣ activity, the cDNA for NEMO was subcloned into an expression vector in the reverse orientation to produce an antisense transcript, enabling down-regulation of cellular NEMO levels in 293T cells. As expected, expression of antisense NEMO resulted in the decreased activity of a HRE reporter gene driven by HIF-2␣ but not HIF-1␣ (Fig. 6A). The reduction in HIF-2␣ activity, although significant, was not complete. This is likely attributable to the incomplete knockdown of NEMO expression, as shown with co-expression of the antisense transcript with trace amounts of Myc-tagged NEMO (Fig. 6B), or factors other than solely NEMO being responsible for the higher intrinsic activity of HIF-2␣, compared with HIF-1␣, on a HRE reporter. For example, differential activity of the N-terminal TAD. In addition to NEMO overexpression in Fig. 4, these data implicate endogenous NEMO in the regulation of the normoxic activity of HIF-2␣.

NEMO Regulation Is Not Dependent upon ARNT Heterodimerization, the HIF-2␣ C-TAD, or Protein Stability-Be-
cause the C-TAD of HIF-2␣ is hypoxically regulated, crucial to HIF-2␣ activity, and adjacent to the NEMO-interacting region, we investigated whether the C-TAD is required for the ability of NEMO to up-regulate HIF-2␣ activity. The minimal NEMOinteracting region of HIF-2␣ (565-820), or a larger region including the C-TAD (565-874), were fused to the Gal4 DNAbinding domain, as were the corresponding regions of HIF-1␣. These constructs were transfected into 293T cells in the presence or absence of NEMO and analyzed for transcriptional activity on a Gal4-responsive reporter gene. NEMO expression specifically increased the activity of both HIF-2␣ constructs by similar amounts (ϳ2-fold), although total reporter activity mediated by HIF-2␣ 565-820 is greatly reduced because of the absence of an endogenous transactivation domain (Fig. 7A). Thus, up-regulation of HIF-2␣ activity by NEMO is not dependent upon the HIF-2␣ C-TAD. Consistent with previous data, NEMO had no effect on HIF-1␣ constructs, regardless of C-TAD presence or absence. Western analysis of whole cell lysates from the same transfected cells demonstrated that the up-regulation of HIF-2␣ activity by NEMO is not attributable to changes in HIF-2␣ protein levels, because no significant difference was observed with NEMO overexpression (Fig. 7B). Thus, NEMO/HIF-2␣ interaction does not appear to regulate HIF-2␣ activity by altering protein stability. This is supported by pulse-chase experiments indicating that NEMO does not specifically increase HIF-2␣ stability compared with HIF-1␣ (data not shown). Therefore, NEMO appears to specifically increase the transcriptional activity of the HIF-2␣ C terminus in a protein stability and C-TAD-independent manner.
Another possible mechanism of NEMO action is increasing HIF-2␣/ARNT heterodimerization. Co-immunoprecipitation experiments, however, demonstrate that the ability of HIF-2␣ to heterodimerize with ARNT is not affected by NEMO overexpression (Fig. 2C).
NEMO Increases HIF-2␣ Activity via p300 Recruitment-NEMO is predominantly localized in the cytoplasm (Ref. 24 and data not shown). However, recent data has indicated a potential nuclear role. Genotoxic stress has been found to direct the sumoylation of a non-IKK-associated portion of NEMO, resulting in nuclear accumulation (31), whereas tumor necrosis factor ␣ has been shown to induce recruitment of NEMO to the IB␣ promoter (32). Furthermore, NEMO is reported to shuttle between the nucleus and cytoplasm and compete with IKK␣ and p65 for CBP binding (33).
Because CBP/p300 is recruited to the HIF-␣ C-TAD for transcriptional activation under hypoxia (4, 34 -42) and is also capable of binding NEMO (33), we hypothesized the C-TADindependent association of CBP/p300 with HIF-2␣ via NEMO at normoxia as a potential mechanism by which NEMO specifically up-regulates HIF-2␣ activity. To test this hypothesis, HIF-2␣(565-820) and HIF-1␣(597-775)-GalDBD fusion proteins, which lack the endogenous transactivation domains and hence would not be expected to independently bind p300, were expressed with NEMO-Myc and full-length p300-VP16 (Fig.  8A). Expression of NEMO-Myc or p300 alone increased HIF-2␣ but not HIF-1␣, reporter activity. However, co-expression of NEMO and p300 with HIF-2␣-GalDBD led to a striking and specific increase in reporter activation that was more than the additive for the expression of NEMO or p300 individually, demonstrating synergistic activity.
To confirm this mechanism with full-length, untagged HIF-␣ protein, HRE reporter gene activity was analyzed with NEMO and p300 co-expression at normoxia (Fig. 8B). As anticipated, FIG. 6. Antisense NEMO specifically reduces normoxic HIF-2␣ activity. A, 293T cells were transiently co-transfected with a 4ϫHREluciferase reporter (10 ng), RLTK-luciferase internal control (10 ng), HIF-␣ expression (10 ng), and antisense NEMO (or control) vectors (500 ng). Luciferase activity was assayed 24 h post-transfection, and the results are presented in triplicate Ϯ S.D. and representative of three independent experiments. B, 293T cells were transiently co-transfected with Myc-tagged NEMO (5 ng) and antisense NEMO or the empty vectors as controls. 18 h post-transfection, the whole cell extracts were separated by SDS-PAGE (10% gel), transferred to nitrocellulose, and Western blotted with an ␣-Myc antibody. the expression of NEMO or p300 with HIF-2␣, but not HIF-1␣, yielded a 2-fold increase in reporter activity, which was dramatically increased upon NEMO and p300 co-expression. Similar co-expression with HIF-1␣ gave a far smaller increase in activity, consistent with the small increase observed in the absence of additional HIF-␣. This is at least partially attributable to a small HIF-independent effect of NEMO and p300 overexpression on reporter activity because increases were seen with the coexpression of NEMO and p300 on both HRE (Fig. 8B) and GalRE (Fig. 8A) reporter constructs. Similar results were also seen with CBP overexpression (data not shown).
These data indicate that NEMO and CBP/p300 act synergistically to increase HIF-2␣, but not HIF-1␣, activity, suggesting a model by which NEMO facilitates CBP/p300 recruitment to HIF-2␣. Such a model is further supported by data in Fig. 2C, demonstrating the co-immunoprecipitation of NEMO by ␣-ARNT antibodies, suggesting that NEMO is present in transcriptionally active HIF-2␣/ARNT heterodimeric complexes. This is consistent with previously published data showing that NEMO can shuttle between the nucleus and cytoplasm and compete with IKK␣ and p65 for CBP binding (33). DISCUSSION In this study, we present mammalian two-hybrid and GST pull-down data demonstrating that NEMO differentially interacts with amino acids 565-820 of HIF-2␣, but not HIF-1␣. Data are also presented in which endogenous HIF-2␣ is co-immunoprecipitated with endogenous by NEMO. Furthermore, we show this interaction has a functional consequence. First, normoxic activity of HIF-2␣ is increased in a dose-dependent manner upon NEMO overexpression, and second, transfection of an antisense NEMO construct decreases HIF-2␣ dependent reporter gene activity, indicating that endogenous NEMO contributes to HIF-2␣ activity. This is consistent with a model by which NEMO is able to bind and recruit co-activators (CBP/ p300) to HIF-2␣, supported by reporter gene assays where HIF-2␣, but not HIF-1␣, activity increases synergistically upon NEMO and p300 co-expression. The ability of NEMO to bind CBP has been recently reported (33).
These data also show that NEMO up-regulates HIF-2␣ activity at normoxia but not hypoxia where increased HIF-2␣ activity overrode any discernable effect of NEMO. Importantly, the HIF-␣ proteins have been shown to be functional in the absence of hypoxia, as demonstrated in HIF-1␣ Ϫ/Ϫ macrophages and embryonic stem cells, which express reduced levels of HIF-␣ target genes at normoxia (5,43). The interaction of NEMO with HIF-2␣ may therefore contribute to a physiologically important normoxic role, given that reduction in endogenous NEMO by antisense expression specifically reduced HIF-2␣ activity.
Several lines of evidence also implicate stimuli other than hypoxia as physiological HIF-2␣ inducers. The finding that HIF-2␣ Ϫ/Ϫ embryonic stem cells are only resistant to hypoglycemic, but not hypoxic, induced apoptosis, implies that hypoglycemic stress may represent such a stimuli (13), as may cellular acidosis, stabilizing HIF-␣ protein via nucleolar sequestration of the Von Hippel-Lindau protein (44). Multiple growth factors have also been implicated in HIF-␣ induction (reviewed in Ref. 45). Furthermore, specific cell lines and tissues such as fibroblasts, macrophages, neutrophils, arterial endothelia, and neonatal rat brains display strong endogenous, normoxic immunoreactivity of HIF-2␣ (12, 46 -50). Given a number of these studies report cytoplasmic localization of HIF-2␣, and the implication that hypoxia may not be the sole stimulus for HIF-2␣ activity in vivo, our data demonstrate a mechanism by which NEMO may lead to transcriptionally active HIF-2␣ at normoxia. Despite the high similarity of HIF-1␣ and HIF-2␣ with respect to amino acid sequence, domain architecture, DNA binding, and activation pathway, HIF-1␣-and HIF-2␣-deficient mice exhibit dramatically different phenotypes, indicating nonredundant protein functions. HIF-1␣ Ϫ/Ϫ embryos die by E11 displaying defective vascularization, cardiovascular malformation, and failure of neural tube closure because of mesenchymal cell death (5,6). A different vascular phenotype has been reported for HIF-2␣ Ϫ/Ϫ embryos, with death by E12.5 because of insufficient vascular remodelling (7). Interestingly, other studies using HIF-2␣-deficient mice report different phenotypes including bradycardia from impaired catecholamine production (8), respiratory distress syndrome from reduced lung surfactant (9), impaired survival of bone marrow hematopoietic cells (10), and abnormalities in metabolism and reactive oxygen species homeostasis (11).
Underlying these knock-out phenotypes, there is a growing body of evidence demonstrating the differences between the molecular responses of HIF-1␣ and HIF-2␣. In mouse embryonic fibroblasts for example, HIF-2␣, in contrast to HIF-1␣, is reported to be cytoplasmically expressed at normoxia, unresponsive to hypoxia, and displays no endogenous transactivation. Furthermore, HIF-2␣ is reported to be nonresponsive to the primary mechanisms of HIF regulation involving proline and asparagine hydroxylation in this cell type (12). Glycolytic enzymes appear to be specifically up-regulated by HIF-1␣ (14), whereas the Rta gene of Kaposi's sarcoma-associated herpesvirus (51), the tyrosine kinase Tie2 (52), the vascular endothelial growth factor receptor Flt-1 (53), and erythropoietin (54) are reported HIF-2␣ targets. As mentioned previously, HIF-2␣ Ϫ/Ϫ, unlike HIF-1␣ Ϫ/Ϫ, embryonic stem cells are only resistant to hypoglycemic, although not hypoxic, induced apoptosis (13), whereas the renal cell carcinoma cell line 786-0 has been used to implicate HIF-2␣, and not HIF-1␣, in the tumorgenicity of subcutaneously injected immunocompromised mice (55).
In contrast to this growing body of evidence demonstrating clear differences between HIF-1␣ and HIF-2␣ function, few mechanisms are known to account for this nonredundancy. A reducing activity supplied by the redox factor Ref-1 is required for HIF-2␣ DNA binding, although not for HIF-1␣, which constitutively binds DNA (56). HIF-2␣ has also been reported to specifically associate with the Ets1 transcription factor in Flk-1 gene expression (57). Lastly, PI3K inhibitors are reported to specifically inhibit HIF-1␣ hypoxic induction (58).
The identification of NEMO as a HIF-2␣, but not HIF-1␣, interacting protein in both yeast and mammalian cells, repre- sents one of the few characterized protein interactions with specificity between the HIF-␣ proteins. Furthermore, the fact that both overexpressed, and antisense, NEMO modulates HIF-2␣ activity identifies NEMO as a novel and specific activator of HIF-2␣ at normoxia. NF-B signaling plays a critical role in the regulation of apoptotic and immune responses, and the NEMO protein is required for the activation of this pathway by virtually all known stimuli, acting as a nonenzymatic scaffold to assemble the IKK complex, a focal point for NF-B regulation (21)(22)(23)(24)(25)(26)(27)(28). This is reflected by the reported signal-regulated, post-translational modifications of NEMO, including IKK-␤-and protein kinase C␣-dependent phosphorylation (59,60), sumoylation (31), and ubiquitylation (61,62). NEMO-deficient mice die early in embryogenesis (E12.5) from massive liver apoptosis, a similar phenotype to that reported for RelA and IKK-␤ deficiency (63,64). The physiological importance of NEMO is further underscored by the distinct X-linked human diseases incontinentia pigmenti and anhidrotic ectodermal dysplasia that have been linked to NEMO mutation (reviewed in Ref. 65).
Other non-IKK NEMO-binding factors have been identified, including both viral and nonviral proteins (23, 66 -74). In each of these cases, the interaction with NEMO regulated NF-B activity. Previous reports have demonstrated a secondary pool of NEMO that is not IKK-associated (33). Furthermore, this pool is able to shuttle into the nucleus and interact with the co-activator CBP. The fact that HIF-2␣ overexpression does not alter NF-B activity demonstrates the first NF-B independent role for NEMO, consistent with a model by which NEMO functions as a scaffold, recruiting CBP/p300 to HIF-2␣ and thereby increasing HIF2␣ normoxic activity. Future work to characterize the role of NEMO in the regulation of HIF-2␣ activity will require the identification and analysis of endogenous HIF-2␣specific target genes.