Hypoxia-inducible Factor and Its Biomedical Relevance*

Oxygen homeostasis is a cornerstone of both the physiology and intermediary metabolism systems. Despite the explosion of knowledge brought about by recombinant DNA technology, links between classic physiology and molecular biology are often fragmentary and tenuous. However, during the past decade, there has been enormous progress in understanding adaptation to hypoxia at the molecular level. A growing number of physiologically relevant genes are up-regulated by falling intracellular oxygen tension via a novel mechanism for oxygen sensing and signaling that triggers the activation of the hypoxia-inducible transcription factor HIF. The importance of this pathway is suggested by its presence in virtually all cells within virtually all higher eukaryotes from flies and worm to man. In this Minireview we summarize a large body of recent information on the oxygen-dependent regulation of the subunit of HIF, by both ubiquitin-proteasomal degradation and by transcriptional activation. In addition we review some of the biomedical aspects of HIF-dependent gene expression, particularly those that impact angiogenesis and tumor biology.

Oxygen homeostasis is a cornerstone of both the physiology and intermediary metabolism systems. Despite the explosion of knowledge brought about by recombinant DNA technology, links between classic physiology and molecular biology are often fragmentary and tenuous. However, during the past decade, there has been enormous progress in understanding adaptation to hypoxia at the molecular level. A growing number of physiologically relevant genes are up-regulated by falling intracellular oxygen tension via a novel mechanism for oxygen sensing and signaling that triggers the activation of the hypoxia-inducible transcription factor HIF. 1 The importance of this pathway is suggested by its presence in virtually all cells within virtually all higher eukaryotes from flies and worm to man. In this Minireview we summarize a large body of recent information on the oxygen-dependent regulation of the ␣ subunit of HIF, by both ubiquitin-proteasomal degradation and by transcriptional activation. In addition we review some of the biomedical aspects of HIF-dependent gene expression, particularly those that impact angiogenesis and tumor biology.

HIF Family and Structural Domains
HIF-1 is a transcription factor that binds specifically in hypoxia to a 5Ј-RCGTG-3Ј hypoxia-responsive element (HRE) in the promoter or enhancer of various hypoxia-inducible genes (1), which include erythropoietin, vascular endothelial growth factor, glucose transporters, and glycolytic enzymes, as well as genes involved in iron metabolism and cell survival (1)(2)(3).
HIF-1 is a heterodimer composed of a 120-kDa HIF-1␣ subunit and a 91-94-kDa HIF-1␤/ARNT subunit, both of which are members of the basic helix-loop-helix (bHLH)-PAS family (1,3). PAS is an acronym for the three members first recognized (Per, ARNT, and Sim). Accordingly, HIF-1␣ and HIF-1␤ each contain a bHLH domain near the N terminus preceding the PAS domain (Fig. 1). Whereas the basic domain is essential for DNA binding, the HLH domain and N-terminal half of the PAS domain are necessary for heterodimerization and DNA binding (1). Moreover, there are two transcriptional activation domains in HIF-1␣; one is referred to here as N-terminal activation domain (NAD) and the other as C-terminal activation domain (CAD). In contrast, HIF-1␤ contains only one transcriptional activation domain at the C terminus. Furthermore, HIF-1␣ possesses a unique oxygen-dependent degrada-tion domain (ODD) that controls protein stability. A portion of the ODD overlaps with the NAD.
In addition to the ubiquitous HIF-1␣, the HIF-␣ family contains two other members, HIF-2␣ (also called EPAS1, MOP2, or HLF) and HIF-3␣, both of which have more restricted tissue expression (3). HIF-2␣ and HIF-3␣ contain domains similar to those in HIF-1␣ and exhibit similar biochemical properties, such as heterodimerization with HIF-1␤ and DNA binding to the same DNA sequence in vitro (1). Despite these similarities, neither HIF1␣ Ϫ/Ϫ nor HIF2␣ Ϫ/Ϫ embryos can survive (see below), suggesting the lack of functional complementation in vivo within the family. In addition, several HIF-1␣ variants have been detected (3). Of particular interest are splice variants HIF-1␣ 516 , HIF-1␣ 557 , and HIF-1␣ 735 that terminate respectively at codons 516, 557, and 735, resulting in the absence of both NAD and CAD or of CAD only. However, the biological significance of these isoforms is unclear. Mouse HIF-3␣ has a splicing variant, termed as inhibitory PAS domain protein (IPAS, Fig. 1), that contains only bHLH and PAS domains (4,5). IPAS dimerizes preferentially with HIF-1␣ instead of HIF-1␤, thereby forming an abortive complex that is unable to bind to the HRE. As a result, strong expression of IPAS in the hypoxic corneal epithelium of the eye accounts for the suppression of HIF-mediated expression of angiogenic genes and consequently an avascular cornea (4). By contrast, in the mouse heart and lung tissues IPAS mRNA is hypoxia-regulated, indicating a negative feedback mechanism that controls HIF-1␣ activity (5).

Mechanisms of HIF-␣ Activation
Hypoxia is the physiologic trigger that activates HIF. However, HIF is also up-regulated by certain transition metals (Co 2ϩ , Ni 2ϩ , Mn 2ϩ ) and by iron chelation. In addition, as discussed below, certain growth factors and cytokines can also activate HIF.
Both HIF-1␣ and HIF-1␤ mRNAs and proteins are constitutively expressed. Only HIF-1␣ responds to changes in oxygen tension, although HIF-1␤, despite its apparent insensitivity to hypoxia, is required for HIF-1 activity. The process of HIF-1␣ activation, including enhanced protein stability and transcriptional activity, has been studied extensively in the past few years. Mechanisms for HIF-2␣ regulation are similar, unless otherwise noted.
Because VHL contains a single, conserved hydroxyproline-binding pocket for HIF-␣, independent recognition of Pro-402 and Pro-564 by VHL (21,28) constitutes functional redundancy within the ODD for mediating proteolysis (9). Mutation of either proline alone only partially stabilizes HIF-1␣, whereas mutation of both prolines markedly increases its stability (21). VHL-mediated degradation is arguably the most critical mechanism for physiological regulation of HIF-1␣, although how these hydroxyprolines are selected for VHL binding remains to be elucidated.
HIF-1␣ Transcriptional Activation-Transcriptional activation is another key step that regulates HIF-1␣ activity. The CAD is essential for transcriptional activation of HIF-1 target genes, whereas the NAD appears to be functionally dispensable (9,35).
The highly conserved CAD (36 -38) functions via interaction of its leucine-rich interface with the CH1 domain of p300/CBP (39), a widely employed transcriptional co-activator (40 -43). Structural studies show that unbound CAD is intrinsically disordered and that the CH1 domain serves as a scaffold for CAD folding through extensive hydrophobic and polar interactions (44,45). The CAD per se is stable, but its transcriptional activity is hypoxia-inducible. This response is, at least in part, attributable to hypoxia-induced p300/CBP binding, which is governed by hydroxylation of Asn-803 in HIF-1␣ (46). The asparaginyl hydroxylation is catalyzed by another Fe(II)-and O 2 -dependent enzyme (47,48), FIH-1, which interacts with the C-terminal HIF-1␣ and represses HIF-1␣ transcriptional activity (49). Moreover, Asn-803, as part of an ␣ helix, is buried in the molecular interface, and therefore hydroxylation of Asn-803 would preclude or diminish p300/CBP binding (44, 45) (Fig. 2). Thus, in normoxia hydroxylated Asn-803 prevents p300/ CBP binding, whereas hypoxia inhibits the activity of asparaginyl hydroxylase, thereby enhancing p300/CBP interaction and up-regulating target gene expression.
However, this hypothesis is difficult to reconcile with the observation that a stable HIF-1␣ mutant under normoxia is capable of transcriptionally activating target genes in both cell culture and animal models (9,35). Furthermore, VHL-deficient cells constitutively express genes that are otherwise hypoxia-inducible (11). The possible requirement for VHL in FIH-1 function (49) may account for this apparent paradox so that the loss of VHL binding or activity would impair FIH-1 activity. However, in vitro FIH-1 alone seems to bind strongly to HIF-1␣, and more interestingly, overexpression of FIH-1 inhibits HIF-1␣ transcriptional activity under both normoxic and hypoxic conditions (49). Furthermore, replacement of the adjacent Cys-800 with hydrophobic residues, such as valine, markedly increases CAD activity in normoxia (39), indicating that such mutations either inhibit Asn-803 hydroxylation or bypass the steric hindrance presumably imposed by hydroxylated Asn-803.
Transcriptional activation may be amplified by the presence of two or more nearby functional HREs. Alternatively the HRE may be flanked by a cis-acting DNA element, the cognate transcription factor of which also binds to p300/CBP. Examples include the erythropoietin enhancer in which HIF cooperates with HNF-4 in recruiting p300/CBP and the promoter of LDH-A in which HIF and CREB/ATF-1 synergistically activate transcription by the same mechanism (42).
Role of Protein Kinase Pathways-Phosphorylation is crucial in controlling activities of myriad proteins including transcription factors. It has been long known that HIF-1␣ is a phosphoprotein, yet no functionally relevant phosphorylation sites have been identified. In fact, it has become evident that de novo phosphorylation may not be required for stabilization of HIF-1␣ under hypoxia (19, 50 -52). However, in certain cell types, the phosphatidylinositol 3-kinase (PI3K)-Akt-FKBP12-rapamycin associated protein (FRAP)/mammalian target of rapamycin (mTOR) pathway and the MAP kinase pathway do appear to mediate growth factor-or cytokine-induced HIF-1␣ activation by stimulating HIF-1␣ translation or enhancing HIF-1␣ transcriptional activity under non-hypoxic conditions (for a review see Ref. 53).
Increased HIF-1␣ Translation-In contrast to regulation via the ubiquitin-proteasome pathway, both heregulin and insulin-like growth factor 1 stimulate HIF-1␣ synthesis by inducing phosphorylation of the translational regulatory proteins 4E-BP1, p70 S6 FIG. 1. Schematic representation of HIF-1␣, HIF-1␤, and IPAS. Relevant domains are indicated in abbreviations and specified in the text, and the sites of Pro-402, Pro-564, and Asn-803 (P402, P564, and N803) are marked. The NAD shown in red lies within the ODD represented in yellow.

FIG. 2. Oxygen-dependent regulation of HIF depends on site-specific hydroxylation of the HIF-␣ subunit.
Above a threshold level of O 2 and Fe 2ϩ , Pro-402, Pro-564, and Asn-803 in HIF-1␣ are hydroxylated, catalyzed by prolyl 4-hydroxylase (PH) and asparaginyl hydroxylase (FIH or factor that inhibits HIF), respectively. The hydroxylated prolines (P-OH) permit VHL binding, thereby triggering ubiquitin conjugation and proteasomal degradation of HIF-1␣. The hydroxylated asparagine (N-OH) prevents HIF-1␣ binding to p300/CBP, thereby inhibiting transcriptional activation. The prolyl hydroxylase activity is blocked not only by low oxygen tension but also by desferrioxamine (DFO) or cobalt ion, all of which result in stabilization of HIF-1␣, enabling it to interact with HIF-1␤. The heterodimer recruits p300/CBP and binds to the HRE in the promoter or enhancer of target genes, allowing interaction with the basal transcription machinery (BTM). Oxygen tension is indicated by intensity of the red background that decreases from left to right. kinase, and eukaryotic initiation factor 4E (53)(54)(55). These proteins are substrates of the PI3K-Akt-FRAP/mTOR cascade, which can be activated by a variety of growth factors and cytokines including epidermal growth factor, heregulin, insulin, insulin-like growth factors, and interleukin-1␤, etc. to induce HIF-1␣ accumulation and target gene expression (53). Consistently inhibitors targeting PI3K or FRAP/mTOR prevent growth factor-and cytokine-induced HIF-1␣ accumulation. Overexpression of PTEN, which dephosphorylates the PI3K products, results in decreased HIF-1␣ expression in glioma cells (56). Moreover, the 5Ј-untranslated region of HIF1␣ gene contains an internal ribosome entry site that enables efficient translation under both normoxia and hypoxia (57,58). This region seems to be responsible for the increased translation by the growth factors (54). It should be noted, however, that the induction of HIF-1␣ by hypoxia is far greater than by growth factors and cytokines (51,52), and effects of the two stimuli are additive (55).
Modulation of Transcriptional Activity-Both the p42/p44 and p38 MAP kinases phosphorylate HIF-1␣ in vitro. Transfection with active forms of p42/p44 MAP kinase stimulates HIF-1␣ transcriptional activity without affecting HIF-1␣ stability (50,53). Moreover, inhibitors of MEK1 (which is upstream of p42/p44) or p38 block HIF-1␣-mediated reporter gene expression. Therefore, all of these findings support the role of MAP kinases in augmenting HIF-1␣ transcriptional activity. However, whether the MAP kinases play a direct role in hypoxia-induced HIF-1␣ transactivation still remains to be determined. Because HIF-1␣ transactivation is a complex process involving multiple proteins, it is likely that these MAP kinases act on HIF-1␣ interacting proteins rather than HIF-1␣ itself to stimulate transactivation. In fact, there is no direct evidence that hypoxia induces HIF-1␣ phosphorylation or that any phosphorylation sites confer transcriptional activity. Although Thr-796 in CAD has been shown to be necessary for CAD function and p300/CBP binding, apparently it is not a target of p42/p44 MAP kinase or casein kinase II (59). Furthermore, the fact that replacement of Thr-796 with Ser maintains transcriptional activity suggests that the hydroxyl group is required not phosphorylation.

Biomedical Relevance
Development-Both HIF-1 and HIF-2 play essential roles in embryonic development. Targeted disruption of HIF1␣, HIF2␣, and HIF1␤ results in embryonic lethality. In normal embryos HIF-1␣ expression increases between E8.5 and E9.5 (60). HIF1␣ Ϫ/Ϫ embryos die by E11 (60,61). From E8.5 there was lack of blood vessel formation in the brain along with reduction in the number of somites and defective formation of the neural fold. In addition, these embryos had multiple defects in cardiovascular development. Use of the nitroimidazole marker EF5 demonstrated global hypoxia in HIF1␣ Ϫ/Ϫ embryos (61). HIF2␣ Ϫ/Ϫ embryos do not survive beyond E16.5. They have normal systemic vasculature but impaired fetal lung maturation (62) and a very slow heart rate because of decreased production of catecholamine (63).
Additional insights into the in vivo significance of HIF have come from experiments in which either HIF-1␣ is inactivated in specific tissues or its overall expression is reduced. The cartilaginous growth plate at the end of long bones is avascular and therefore hypoxic. When targeted inactivation of HIF-1␣ is limited to cartilage, there is cell death within the interior of the growth plate and, surprisingly, up-regulation of vascular endothelial growth factor in a HIF-1 independent manner (64). HIF-1␣ deficiency in Rag-2 Ϫ/Ϫ chimeric mice results in markedly defective maturation of B lymphocytes (65). Heterozygous mice carrying a single HIF-1␣ gene have normal development and no apparent phenotypic abnormalities. However, when challenged by hypoxia, HIF1␣ ϩ/Ϫ mice have impaired cardiovascular (66) and neural (67) adaptational responses.
Tumor Pathogenesis-The critical role that HIF plays in cancer biology can be inferred from immunohistochemistry data demonstrating normal levels of HIF-1␣ protein in benign tumors, elevated levels in a variety of primary malignant tumors, and even more marked elevations in tumor metastases (68). Unless or until a tumor gains adequate blood vessels, the interior of the tumor mass becomes progressively hypoxic as its size increases. Thus the overexpression of HIF-1␣ is in part because of its induction through the ubiquitous pathway of oxygen sensing and signaling.
In addition, cancer-inducing mutations often further enhance HIF expression. The p53 tumor suppressor gene is inactivated in the majority of human cancers. Loss of wild type p53 in tumors enhances HIF-1␣ levels and augments HIF-dependent transcription (29). Increased activity of the HER2 receptor tyrosine kinase is a prevalent and important genetic alteration in breast cancer, correlating with tumor aggressiveness and decreased patient survival. Enhanced HER2 signaling increases the rate of synthesis of HIF-1␣ (54). The PTEN tumor suppressor gene is sometimes deleted in the brain tumor glioblastoma multiform. Loss of PTEN function activates the Akt/protein kinase B signaling cascade, resulting in increased HIF-1␣ levels (56). The transforming potential of the v-Src oncogene is due in part to its induction of HIF (69).
Perhaps the most convincing and heuristic association between tumor-inducing mutation and HIF expression is the VHL protein.
As shown in Fig. 2, VHL plays a crucial role in targeting HIF-1␣ and HIF-2␣ for proteasomal degradation. von Hippel-Lindau disease is an autosomal dominant hereditary cancer syndrome in which affected family members have a germ line mutation that usually impairs either VHL binding to HIF-1␣ or the function of VHL-ubiquitin E3 ligase complex. These heterozygotes have a very high predisposition to develop vascular tumors in the kidney and less often in retina, cerebellum, adrenal medulla, and pancreas. Within these tumors, there is an acquired loss of the other VHL allele. In addition, over one-half of sporadic cases of clear cell renal carcinoma have biallelic loss of VHL genes. Because of the absence of functional VHL, HIF-1␣ is constitutively expressed. Although HIF plays an important role in tumor pathogenesis (70 -72), it may not be the critical oncogenic substrate of VHL (72).
In the Chuvash region of Russia there is a high incidence of a VHL mutation: Arg-200 3 Trp (73). Homozygotes have polycythemia (high red cell mass) because of HIF-dependent up-regulation of erythropoietin gene expression. These individuals do not appear to be at increased risk for developing tumors. Surprisingly, this mutational site on VHL is not close to the binding domains for either HIF-1␣ or the E3 ubiquitin ligase complex.
The striking up-regulation of HIF-1␣ in so many different tumors by so many different molecular mechanisms begs the question of how this phenomenon impacts tumor biology. There is evidence that HIF enhances tumor viability and growth by several different mechanisms. Oxygen supply to a tumor is limited by diffusion unless it can develop a functional network of blood vessels. HIF has a remarkably positive impact on tumor angiogenesis. When injected into immunocompromised mice, embryonic stem cells generally form highly vascular teratocarcinomas. In contrast, embryonic stem cells that are HIF1␣ Ϫ/Ϫ form smaller tumors that are less vascular and express much lower levels of HIF-dependent vascular endothelial growth factor (VEGF) (61).
HIF plays a complex role in mediating hypoxia-induced apoptosis. Some genes involved in cell cycle control, such as p53 and p21, are HIF-dependent (74,75). Anti-apoptotic Bcl-2 is down-regulated by hypoxia (74) whereas the pro-apoptotic family member Nip3 is up-regulated (75). Accordingly chronic or persistent hypoxia reduces proliferation and increases apoptosis in a HIF-dependent manner. p53 is required for hypoxia-induced apoptosis in vivo (76). Tumors that lack p53 are resistant to hypoxia-induced apoptosis within the hypoxic interior and therefore may be resistant to antiangiogenic agents.
Normal tissues adapt to hypoxia in part by switching from aerobic respiration to anaerobic glycolysis (the Pasteur effect). In the early part of the last century, Otto Warburg showed that under normoxic conditions tumors have a higher rate of glycolysis compared with normal tissues. This phenomenon helps to maintain cell viability in the hypoxic interior of the tumor. This metabolic response is mediated by HIF-1-dependent up-regulation of genes encoding enzymes of the glycolytic pathway (77,78).
Tissue Ischemia-Just as HIF is a critical determinant of cancer growth and progression, it plays an equally positive role in the response of an organ or tissue to damage following compromise of blood flow to vital organs such as the heart or brain. When patients develop acute coronary artery occlusion, there is a prompt increase in the expression of HIF-1␣ and VEGF in the myocardium (79). Effective vascular remodeling following ischemic injury depends on an integrated program of HIF-dependent gene expression. Consti-tutive tissue-specific expression of a stable HIF-1␣ mutant transgene resulted in well ordered angiogenesis, whereas overexpression of VEGF lead to the formation of leaky vessels and inflammation (35). These experimental observations are relevant to the development of therapeutic interventions to improve vascular remodeling following acute arterial occlusion. In a rabbit model of hind limb ischemia, injection of DNA expressing a fusion protein containing the N-terminal half of HIF-1␣ and VP16 was more effective than a VEGF-expressing vector in restoring blood flow (80).