For most organisms oxygen is essential for life. When oxygen levels drop below those required to maintain the minimum physiological oxygen requirement of an organism or tissue it is termed hypoxia. To counteract possible deleterious effects of such a state, an immediate molecular response is initiated causing adaptation responses aimed at cell survival. This response is mediated by the hypoxia-inducible factor-1 (HIF-1), which is a heterodimer consisting of an α- and a β-subunit. HIF-1α protein is stabilized under hypoxic conditions and therefore confers selectivity to this response. Hypoxia is characteristic of tumors, mainly because of impaired blood supply resulting from abnormal growth. Over the past few years enormous progress has been made in the attempt to understand how the activation of the physiological response to hypoxia influences neoplastic growth. In this review some aspects of HIF-1 pathway activation in tumors and the consequences for pathophysiology and treatment of neoplasia are discussed.
Hypoxia occurs when oxygen availability/delivery is below the level required to maintain physiological O2 tensions of a particular tissue, i.e., when the tissue demand exceeds its O2 supply. Low O2 partial pressure (Po2) is a characteristic of almost all types of solid tumors, including brain, colon, breast, cervical, prostate, and head and neck (26). It is now well established that most tumors have lower median Po2 than their tissues of origin, and recurring tumors face reduced oxygen supply compared with their corresponding primary tumors. Oxygenation measurements using microelectrodes have shown that the mean oxygen tension of normal tissues lies between 40 and 50 mmHg, whereas tumors reach values between 5 and 10 mmHg or lower (96, 134, 190, 235). In addition, clinical studies show that in many cases invasive growth and metastatic spread are associated with the degree of tumor hypoxia (24, 93, 195).
Tumor hypoxia is a result of the abnormal process of neoplastic growth and depends crucially on oxygen and nutrients from the host. This supply is mainly achieved through neoangiogenesis, a process by which new blood vessels are formed from preexisting ones (84). Therefore, newly formed blood vessels are central for tumor growth. Hypoxia upregulates the most important molecule mediating neoangiogenesis, namely the vascular endothelial growth factor (VEGF), the most potent angiogenic substance known so far. Nevertheless, neoplastic vascularization occurs uncoordinatedly, resulting in functionally poor vasculature incapable of meeting tumoral demands of oxygen, thereby giving rise to hypoxic areas within tumors (33).
Hypoxic conditions elicit cellular responses designed to improve cell oxygenation/survival through several mechanisms such as neoangiogenesis promotion, improved glycolytic flux enhancing energy production, and upregulation of molecules related to cell survival/apoptosis (208). The most important protein regulating the mammalian molecular response to hypoxia is the heterodimer hypoxia-inducible factor-1 (HIF-1). HIF-1α protein is stabilized under hypoxic conditions and therefore confers selectivity to the hypoxic response. This response is initiated by the dimerization of HIF-1α with its partner HIF-1β [also known as ARNT (aryl hydrocarbon receptor nuclear translocator)] forming the transcription factor HIF-1 that in turn upregulates genes involved in adaptation responses. HIF-1α is overexpressed in neoplastic tissues (266) because of the hypoxic nature of tumors but also due to oncogene activation. This expression enables tumors to take advantage of the physiological response mechanism to hypoxia to improve their own survival. The HIF pathway and some its consequences for neoplastic diseases are discussed below.
STRUCTURE OF HIF-1
HIF-1 is one of the most important molecules involved in the mammalian response to hypoxia. It is a heterodimer consisting of two subunits: an α-subunit, protein levels of which increase with decreasing oxygen levels, and an oxygen-independent β-subunit. HIF-1α and HIF-1β share two structural characteristics: both contain basic helix-loop-helix (bHLH) and PAS domains [PAS is an acronym referring to the first proteins, PER, ARNT, SIM, in which this motif was first identified (see Fig. 1)]. The basic domain and the COOH-terminal half of PAS are specifically required for DNA binding of HIF-1, whereas the HLH domain and the NH2-terminal half of the PAS domains are responsible for the formation of the HIF-1α/HIF-1β heterodimer capable of DNA binding (123). The specific DNA sequence to which HIF-1 binds is termed “hypoxia response element” or HRE and is present in promoter or enhancer sequences of HIF-1 target genes. Binding of HIF-1 to HRE leads to the upregulation of HIF-1 target genes. It is important to note that not only the amount of HIF-1 protein but also additional factors such as coactivators or protein modifications determine the binding of HIF-1 to HRE and are important for the activation of target genes (34, 193). For instance, it has been shown by in vitro experiments that 100-fold overexpressed HIF-1α protein translocates into the nucleus under normoxic and hypoxic conditions but is not able to increase further either the HIF-1 binding capacity or the mRNA levels of HIF-1 target genes beyond the levels found in control cells exposed to hypoxia (98).
HIF-1α also contains two transactivation domains (TADs). The main function of the TADs is to recruit and interact with coactivators, which are crucial for the transcriptional activation of target genes. These domains are also important because HIF-1α undergoes posttranslational regulation mediated through hydroxylation, phosphorylation, acetylation, and/or redox modifications of these two TAD domains (28, 103, 119, 193). Cells cultured under hypoxic conditions (typically 1% O2) increase HIF-1α protein levels without concomitant elevation in mRNA expression. This suggests that the main regulation pathways rely on oxygen-dependent protein stabilization (76, 103). HIF-1α also comprises a sequence termed oxygen-dependent degradation domain (ODD domain) that allows regulation of protein stability as a function of the O2 concentration (104).
On stabilization, HIF-1α accumulates in the nucleus of hypoxic cells (40, 121) due to two nuclear localization signals (127). Hypoxic HIF-1α translocation and nuclear accumulation still occur in HIF-1β/ARNT-deficient hypoxic cells, demonstrating that these events are ARNT independent. This result was substantiated by the observation that HIF-1β/ARNT is a nuclear protein (40).
The formation of the HIF-1 heterodimer depends on the stability of its α-subunit. By performing tonometry experiments, HIF-1α protein was detected in the nucleus after <2 min of exposure to hypoxia or anoxia (120). Further experiments using HeLa cells showed that the levels of HIF-1α protein and DNA binding of HIF-1 vary exponentially over a physiologically relevant range of O2 tension (124). Reoxygenation reduced HIF-1 DNA binding and nuclear HIF-1α protein levels within 4-8 min, suggesting a protein half-life of ∼5 min (120). Thus, under hypoxic conditions, HIF-1α protein is stabilized, translocates into the nucleus, and accumulates within a very short period of time (minutes) to allow rapid responses to lowered oxygen concentrations (see below). When oxygen levels rise, HIF-1α is quickly degraded, thereby becoming undetectable under normoxic conditions.
In summary, HIF-1α protein belongs to the bHLH/PAS superfamily. It has a defined domain architecture in which the NH2-terminal domains are involved in dimerization/DNA binding, whereas the COOH-terminal domains are concerned with stabilization and transactivation regulation of the heterodimer.
HIF-1α DEGRADATION AND OXYGEN SENSING
The nature of the oxygen sensor mechanism remained unknown for many years and gave rise to a series of theories discussed elsewhere (244). The breakthrough in this field was achieved by two groups simultaneously (112, 115). They discovered an oxygen-dependent enzymatic modification of the ODD domain of HIF-1α. Three new prolyl-4-hydroxylases [named prolyl-hydroxylase domain-containing protein (PHD 1, 2, and 3); also known as HIF (prolyl hydroxylase 3, 2, and 1, respectively)] were discovered that are able to hydroxylate two HIF-1α prolin residues (Pro402 and Pro564) in the presence of oxygen, ultimately leading to its degradation (28, 115). PHDs hydroxylate specific prolines recognizing a strongly conserved LXXLAP motif (where X indicates any amino acid and P indicates the hydroxyacceptor proline) (102), requiring oxygen and 2-oxoglutarate as cosubstrates as well as iron and ascorbate. Ascorbate is an alternative oxygen acceptor in uncoupled decarboxylation sites (137; see Fig. 2). Hypoxia-mimicking elements such as iron chelators or transition metals such as cobalt also suppress the hydroxylation of the proline residues and thus stabilize HIF-1α (260; for detailed review, see Ref. 151). Inhibition of HIF-1α hydroxylation under hypoxic conditions confirmed the key role of these PHDs in the oxygen-sensing mechanism (54). It would be therefore interesting to learn which of the three enzymes is mostly involved in the degradation of HIF-1α. PHD2 was shown to have the highest specific activity toward HIF-1α's main hydroxylation site (102). Recent studies also revealed the cellular localization and hypoxia dependency of these PHDs (102, 175). PHD1 is localized exclusively in the nucleus, PHD2 is mostly localized in the cytoplasm, and PHD3 is found in both the cytoplasm and in the nucleus, with cytoplasmic predominance. Only the mRNAs from PHD2 and PHD3 were shown to be hypoxia inducible, whereas PHD1 mRNA expression was hypoxia independent (175) and could be part of a negative feedback mechanism for the degradation of HIF-1α. It is noteworthy that HIF-1α ubiquitination has been reported to be strictly nuclear (81).
The prolyl hydroxylation of HIF-1α is of central importance in its degradation pathway. HIF-1α is constitutively expressed, but the protein levels are usually not detectable under normoxic conditions (74, 120). The degradation pathway starts with the binding of the von Hippel-Lindau tumor suppressor protein (pVHL) to the hydroxylated ODD domain of HIF-1α (112, 115). pVHL is part of the E3 ubiquitin-ligase complex that targets key regulatory proteins for ubiquitin-mediated proteolysis in the proteasome (144). Proteasomal inhibitors or mutation of the activating enzyme E1 stabilize HIF-1α, showing that under normoxic conditions HIF-1α is degraded by ubiquitination and proteasomal degradation (104, 125, 188, 200, 216, 222). The loss or mutation of pVHL in vivo stabilizes the HIF-1α protein under normoxic conditions and may lead to the VHL hereditary cancer syndrome (41, 130, 145, 181, 257). It is of note that VHL is also involved in the degradation of other HIF family members such as HIF-2α (145) and of three of the six splice variants of the HIF-3α locus (171) (see below).
A recent report (119) identified the acetylation of Lys532 within the ODD domain of HIF-1α by the acetyltransferase ARD1. The authors demonstrate that ARD1 inhibits HIF-1α transcriptional activation and protein stability and stimulates its degradation. They suggest that acetylation is, in concert with hydroxylation, critical for the proteasomal degradation of HIF-1α because it increases the interaction of HIF-1α with pVHL and consequently the pVHL-mediated ubiquitination.
Very recently, a third hydroxylation site, the asparagyl-residue Asn803 in HIF-1α and Asn851 in HIF-2α (152, 201) was also determined. Hydroxylation of the asparagyl-residue, however, does not lead to HIF-1α degradation and is therefore not directly involved in the oxygen-sensing mechanism. The hydroxylation of the Asn803 residue leads to a steric inhibition of the interaction between HIF-1α and its coactivator CBP/p300 (47, 62), interfering with its recruitment. This recruitment is critical for HIF-1α activation. This asparagyl hydroxylase was first described as factor inhibiting HIF-1 (FIH-1) (89, 162). FIH-1 is mainly localized in the cytoplasm and its mRNA is not hypoxia inducible (175). Apart from this, hydroxylase-independent mechanisms regulating HIF-1 stability also exist. As such, the molecular chaperone Hsp90 has been identified as a novel VHL- and oxygen-independent regulator of the HIF-1α protein stability (35, 110, 133). Implications in possible tumor therapies are discussed later.
Cytoskeleton proteins also regulate HIF-1α levels. Small GTPases such as Rac1, Rho, and Cdc42 are upregulated by hypoxia and important for HIF-1 activation (91, 232). Under hypoxic conditions an increase in reactive oxygen species (ROS) would follow hypoxic ATP depletion, which would activate Rho GTPases affecting HIF-1α levels (232). Data on other cytoskeletal proteins such as microtubules are not unanimous. This is important because many microtubule-disrupting substances are used for chemotherapy. In one report (125) microtubule disruption led to HIF-1α stabilization through a nuclear factor (NF) κB-dependent pathway. In another, the microtubule disrupting substance 2ME2 was shown to inhibit tumor growth and angiogenesis possibly due to reduction of HIF-1α and VEGF expression (159).
OTHER MEMBERS OF THE HIF FAMILY AND CANCER INVOLVEMENT
Two other proteins belonging to the bHLH-PAS superfamily and showing striking homology to HIF-1α have been discovered (see Fig. 1). These proteins share several characteristics with HIF-1α, such as hypoxic protein stabilization, heterodimerization with ARNT(s), DNA-recognition/binding, and reporter gene transactivation. The best-studied protein was first termed endothelial PAS protein (EPAS-1) (228). Later, it was proposed this novel protein be named “HIF-2α” (246) because of its similar domain architecture to HIF-1α (see Fig. 1). Recent studies (247) showed that HIF-2α is expressed in a complementary but not overlapping pattern to HIF-1α in specific cells of most organs after systemic hypoxic exposure. HIF-2α expression has also been involved in tumorigenesis. HIF-2α is expressed in a variety of tumors in different levels, but mostly its expression is associated with stromal cells, especially with tumor-associated macrophages (155, 217, 227). This confirms the observation that epithelial tumor cell lines usually have higher levels of HIF-1α compared with HIF-2α, but in macrophages and endothelial cell lines exactly the opposite is observed (249). This has consequences for the activation of target genes because relative levels of either HIFα subunit together with cell-specific posttranslational modification may govern gene regulation (79). HIF-2α levels have been correlated with tumor grade in non-Hodgkin lymphoma (155, 217) and bladder cancers (183). This correlation is of importance because HIF-2α-positive macrophage clusters within invasive breast cancer are directly connected with increased tumor angiogenesis and inversely with survival (155). In a mouse mammary tumor model, VEGF expression has been observed only in stromal cells but not in tumor cells, demonstrating the central importance of the stromal cells for tumor vascularization (66). A recent report showed that whereas HIF-1α has similar expression levels in high- and low-grade astrocytomas, high-grade astrocytomas overexpress HIF-2α to a much greater extent than low-grade ones (136). Using differential expression profiles the authors showed that this HIF-2α overexpression is related to a dysregulation of the epidermal growth factor/PI3K pathway. These results show that growth factor receptor activation can, in addition to HIF-1α, also lead to increased activity of HIF-2α (see below).
HIF-2α may also upregulate a different set of target genes involved in neoangiogenesis than HIF-1α. For instance, the upregulation of the VEGF receptor 2 (131) and angiopoietin receptor tie-2 (229) are upregulated exclusively by HIF-2α in vitro. Interestingly, although HIF-1α and HIF-2α bind to the same HREs (31), the VEGF promoter sequence is more inducible by HIF-2α than by HIF-1α (249). In addition, HIF-2α expression has been correlated with VEGF to a greater extent than HIF-1α in a series of different cancers (57, 60, 71, 183, 254).
A third protein, called HIF-3α, also shares considerable sequence homology with HIF-1α and HIF-2α (83) but lacks a TAD-C (see Fig. 1). This characteristic accounts for its inhibitory effects over HIF-mediated transcription (86). The expression of HIF-3α was found in the distal tubules of the kidney and led to the suggestion that it could act as a negative regulator of the HIF pathway in this tissue (86). In accordance with this suppression effect, a dominant negative regulator of the HIFαs [the inhibitory PAS domain protein (IPAS)] was identified as a splice variant of the HIF-3α locus (164), which is also degraded in a VHL-dependent way (171). This protein is expressed in Purkinje cells and in the corneal epithelium of the eye, where it is thought to play a role in maintenance of the avascular phenotype of this tissue by forming nonfunctional complexes with the HIFs (164). Whether it plays a role in tumorigenesis still remains to be determined.
In analogy to HIF-1α, sequence homologues of HIF-1β (ARNT) have also been discovered in recent years that may play physiological roles as β-class partners of the HIF-α subunits. The ARNT2 protein (50, 90) has been shown to substitute ARNT in heterodimerizing with HIF-1α, HIF-2α, and HIF-3α before HRE binding in DNA-binding assays (90). Furthermore, forced expression of ARNT2 in ARNT-deficient cells rescued hypoxic gene induction (165) and partial redundancy with ARNT has also been shown in vivo (135). A striking difference between these proteins is that the pattern of expression of ARNT2 is restricted primarily to brain and kidney (50, 90, 118), whereas HIF-1β is ubiquitously expressed. In contrast to ARNT2, a third ARNT homologue, MOP3 [member of PAS 3, also called BMAL1 (brain and muscle ARNT-like protein 1) or simply ARNT3] (99, 108, 226) is a weak dimerization partner of the HIF-α(s) (99, 118) and was shown not to participate in the hypoxic response (43). The multiple possibilities given by the amount of dimerization partners and diverse expression patterns add to the complexity of the hypoxic signaling response and the precise role of these molecules still remains to be elucidated.
GROWTH FACTORS, ONCOGENES, AND HIF-1α ACTIVATION PHOSPHORYLATION CASCADES
Recent studies have shown that HIF-1α protein can be stabilized already under normoxic conditions in vivo (219). Although little is known about mechanisms regulating these processes, posttranslational modifications, especially phosphorylation, seem to play a major role in HIF-1α activation. Many tumor suppressor genes/oncogenes influence or are constituents of phosphorylation cascades and thus are able to affect HIF-1α expression levels independent of oxygen levels. Interestingly, these cascades can be initiated after growth factor receptor binding to receptor tyrosine kinases, which in turn activate the downstream targets. Examples of growth factors and cytokines shown to upregulate HIF-1α are epidermal growth factor, basic fibroblast growth factor, heregulin, interleukin-1β, insulin-like growth factors (IGF) 1 and 2, and insulin (58, 88, 153, 264). This is important because different levels of growth factors and their receptors can modulate/activate the HIF pathway through stimulation of phosphorylation cascades.
Two main phosphorylation cascades involved in HIF-1α activation exist (see Fig. 3; for review see Ref. 14). The three broad subfamilies of the mitogen-activated protein kinase (MAPK) cascades, namely the c-Jun NH2-terminal kinases (JNKs), p38 MAPKs, and the extracellular signal-regulated kinases (ERKs) have been shown to regulate the HIF pathway (4, 68, 178, 193, 210).
One of the best studied MAPK pathways involved in HIF-1α regulation leads to the activation of ERK1-2 (also called p44/42) after activation of the upstream molecules Ras/Raf-1/MEK-1/ERK1-2. Initially, it was suggested that the MAPK pathway would be involved only in transactivation activity through direct phosphorylation and not influence the stabilization or DNA-binding ability of HIF-1α (105). However, exposure of HIF-1α-overexpressing HeLa cells to the specific inhibitor PD-98059 (see Fig. 3, left) completely abolished trans-activation activity of both normoxic and hypoxic overexpressed HIF-1α without compromising stabilization, demonstrating the importance of the MAPK pathway for the functionality of HIF-1α (98). Interestingly, although MAPK phosphorylation does takes place (202, 214), phosphorylation of HIF-1α and HIF-2α residues considered central for transactivation activity were shown not to be phosphorylated by this pathway (77). It has been suggested therefore that MAPK signaling affects the transactivation activity of p300, possibly regulating the interaction between p300 and TAD-C (202). Activation of the MAPK pathway was also shown to increase HIF-1α protein synthesis (64) through its effects on eIF-4E (see below). Moreover, both stress-activated serine/threonine protein kinases JNK and p38 MAPK have also been shown to increase activity under hypoxic conditions in certain cell lines (42, 68, 75, 206, 207).
The phosphatidylinositol (PI)-3-kinase (PI3K) signaling cascade represents another important HIF-1α phosphorylation pathway. The downstream kinase AKT has several targets involved in apoptosis, cell cycle, and growth as well as translation (236). One of these targets is FKBP12/rapamycin-associated protein [FRAP; also known as mTOR (mammalian target of rapamycin)]. FRAP is in turn an activator of p70 ribosomal protein S6 kinase (p70s6k) (25, 205, 265), a kinase that enhances the translation of mRNAs that have 5′-polypyrimidine tracts as can be found in HIF-1α (114). In addition, it phosphorylates the translational regulatory protein 4E-binding protein (4E-BP). Its hyperphosphorylation increases protein translation rates (72, 191). This pathway is also negatively regulated by the tumor suppressor gene PTEN (phosphatase and tensin homologue deleted on chromosome 10, see Fig. 3; for review, see Ref. 236). New studies suggest that glycogen synthase kinase-3 (GSK3; Fig. 3) is also involved in HIF-1α protein stabilization under hypoxic conditions by still unknown mechanisms (179). A myriad of mechanisms involving almost every step of activation of PI3K-cascade has been shown to increase HIF-1α expression in tumor cells: activation through growth factors such as epidermal growth factor and insulin, increased growth factor receptor number, increased activation of PI3K or AKT (for example by oncoproteins), and/or loss of PTEN (122, 153, 170, 231, 265, 268). Interestingly, in a study comparing a panel of breast cancer cell lines, HIF-1α alone, but not its homologue HIF-2α, was regulated by a PI3K-dependent pathway, demonstrating that the effects of this pathway may be HIF-1α specific (19). Interestingly, despite these data, the involvement of the PI3K pathway in the regulation of HIF-1α [especially in cancer (64, 153, 265)], has been challenged by two recent reports (5, 9).
Ras is a multifunctional oncogene that can stimulate both the MAPK- and the PI3K-pathways (Fig. 3; Refs. 187, 194). H-Ras transformation has been shown to increase HIF-1α protein levels and target gene activation independent of oxygen through the PI3K pathway (37, 173). An important effect of Ras-mediated upregulation of VEGF seems to depend on an intact HIF-1 binding site within the 5′-flanking region of the VEGF promoter (173). Active Src also enhances HIF-1α levels under normoxia through a PI3K-dependent (in some cell lines also MAPK dependent) mechanism resulting in an enhanced cap-dependent protein translation (132). This increase in HIF-1α protein levels could be a consequence of an increased overall rate of protein synthesis rather than a direct effect of the PI3K pathway on the HIF-1α promoter (132). In one report, Ras- and v-Src-mediated HIF-1α upregulation under normoxic conditions are attributed to prolyl-hydroxylase inhibition (35), suggesting that malignant transformation could directly stabilize HIF-1α independent of VHL.
Growth factors may upregulate HIF-1α through activation of one or both phosphorylation cascades. Signaling from the epidermal growth factor receptor HER2 increased HIF-1α protein synthesis through a PI3K-dependent pathway (153). Colon cancer cells stimulated with IGF-1 or PGE2 induced HIF-1α protein through MAPK- and PI3K-dependent mechanisms (64, 65). Interestingly, in normal prostate epithelial cells, although both phosphorylation pathways were stimulated by IGF-1 and resulted in increased levels of HIF-1, VEGF promoter HRE mutation did not inhibit the effect of IGF-I on VEGF expression, suggesting that PI3-K signaling does not increase VEGF transcription through transactivation by HIF-1 in normal cells (30).
Oncogenes Related to HIF-1α Degradation
VHL. VHL disease is a familial cancer syndrome. According to Knudson's (139, 189) “two-hit” hypothesis, individuals inheriting a single mutated gene can develop neoplastic disease in affected tissues in association with the loss of expression or inactivation of the nonmutated VHL allele. The first hint about the target of VHL tumor suppressor protein was the observation that pVHL-related tumors, such as hemangioblastomas and clear cell renal carcinoma (CCRC), are typically hyper-vascular and occasionally associated with polycythemia (44, 230, 248), conditions tightly related to upregulation of the hypoxia-inducible genes VEGF and erythropoietin (146). The proof for the crucial role of VHL in the VHL syndrome was tumor suppression caused by reintroduction of wild-type pVHL into VHL-/- cells (109).
pVHL plays a central role in the degradation of HIF-1α by binding through its β-domain to the prolyl-hydroxylated ODD of HIF-1α (Fig. 2; Ref. 181), ultimately leading to the polyubiquitination and degradation of HIF-1α through the proteasome (for reviews, see Refs. 126, 189). Because the loss of VHL results in HIF-1α accumulation (41, 181), it was tempting to propose that the HIF pathway forms the basis for etiology of the VHL syndrome. Indeed, inhibition of the HIF pathway was required for tumor suppression by VHL (140) and, in a study involving CCRCs, 77% of all tumors (26 of 34) expressed HIF-1α (233). VHL inactivation has also been correlated with stimulation of the HIF pathway in early kidney lesions (166). Nevertheless, neither loss of pVHL by itself (despite stabilization of HIF-1α and activation of target genes resulting in increased angiogenesis) (161) nor expression of an HIF-1α mutant that is not able to bind to VHL was shown to promote tumor growth (167). This led to further studies considering the role of the second VHL target, namely HIF-2α, which still remains to be better defined. Unlike HIF-1α, reintroduction of HIF-2α does rescue the tumorigenic phenotype in VHL-transfected VHL-/- tumor cells, suggesting that it might be more important than HIF-1α in the pathogenesis of this disease. HIF-1α activation was also observed in later but not in early kidney lesions caused by VHL loss (166). Other studies (60, 233) demonstrated increased protein levels of the HIFs in CCRCs and central nervous system hemangioblastomas but only HIF-2α (not HIF-1α) mRNA upregulation. This is an unusual regulation mechanism of this protein, suggesting that VHL-independent mechanisms are involved in the etiology of this disease. The authors also remarked that in CCRC, HIF subunit expression is biased toward HIF-2α (233). Interestingly, in a further study (167), recapitulation of the tumorigenic phenotype could only be achieved through competitive inhibition of the active site of VHL, demonstrating that HIFs are not the crucial components of VHL-mediated tumorigenesis. In summary, HIF-1α seems to play a role in VHL disease but the tumorigenic phenotype requires HIF-dependent and -independent mechanisms in addition to other oncogenic mutations.
p53. The loss of the p53 tumor suppressor is a central event in the genesis and pathophysiology of the vast majority of tumors. The relation between p53, HIF-1α, and apoptosis still needs to be better defined. Many of the studies showing p53-induced apoptosis under hypoxic conditions were performed under oxygen concentrations close to 0% (anoxia) (6, 78, 204) that are not optimal for hypoxic induction of HIF-1α, because HIF-1α induction starts at 5% and reaches its optimum at 0.5% (124). Although one group (78) could only show p53-dependent apoptosis at 0.02% O2 but not at 0.2% O2, others (246) demonstrated that HIF-1α upregulation is not sufficient for hypoxic/anoxic p53 induction at 1% O2. It has been suggested that under anoxic conditions, the appearance of dephosphorylated HIF-1α correlates with p53 induction and that this is the major form binding to p53 (223). This binding would lead to p53 stabilization and subsequent activation of the p53-dependent apoptotic pathway during anoxia. This reinforces the idea that p53-mediated hypoxic responses in association with HIF-1α are more important in anoxic than in hypoxic conditions.
Despite the differences in the O2 concentrations necessary to optimally induce HIF-1α and p53, several reports demonstrated their interaction. As such, one group showed that hypoxic/anoxic p53 induction is HIF-1α dependent and that HIF-1α coprecipitates together with p53, suggesting a direct HIF-1α/p53 interaction stabilizing p53 (6). However, it is also of note that p53 (10, 82, 156) and HIF-1α (8, 13, 34, 52, 53, 127) bind to distinct regions of the transcriptional coactivator CBP/p300, and this could lead to coimmunoprecipitation without direct HIF-1α/p53 protein interaction. Nevertheless, an immobilized peptide assay showed binding of the ODD domain of HIF-1α and the core domain of p53 (85). One group suggested that the interaction is actually heterotrimeric involving p53, HIF-1α, and mouse minute double 2 (Mdm-2) (192). In this heterotrimeric complex, HIF-1α rather than p53 should be the preferential target of Mdm-2 for ubiquitination and degradation and would, in turn, contribute to the stabilization of p53. This model also offers an alternative mechanism for the degradation of HIF-1α not involving direct interaction with the pVHL tumor suppressor protein. The latest report studying the interaction between p53, HIF-1α, and Mdm-2 confirms these results but finds a strong interaction only between HIF-1α and Mdm-2 and not between p53 and HIF-1α (38). Therefore, the authors postulate that HIF-1α regulation of p53 activity occurs solely through its interaction with Mdm-2, blocking Mdm-2-mediated p53 ubiquitination and nuclear export and abrogating transcriptional repression.
The exact interaction of p53 and HIF-1α in the activation of target genes such as VEGF is also still unclear. Transient cotransfection experiments (6) showed that HIF-1α potentiates p53 reporter gene expression by increasing p53 protein. On the other hand, high levels of p53 have been shown to inhibit HIF-1-dependent reporter gene activation (18, 240). Some reports showed p53 as a repressor of the hypoxia-inducible VEGF transcription (23, 61, 180), a finding challenged by one group (2) who warned about the interpretation of studies concerning repressor functions based on transient p53 overexpression. Nevertheless, a third group (192) reported that loss of p53 promotes neoangiogenesis and tumor growth in nude mice. These findings were attributed to increased HIF-1-dependent transcriptional activation of the VEGF gene. Indeed, loss of p53 activity is associated with tumor angiogenesis in human tumors (225, 259). It can promote neoangiogenesis through increased expression of VEGF (192) and/or through the decrease in the expression of the anti-angiogenic factor thrombospondin-1 (46, 234). Thus the specific connection between VEGF expression and the activity of HIF-1α together with p53 still has to be elucidated.
p14ARF. The Ink4a/Arf locus (CDKN2A) encodes two distinct tumor suppressor proteins, p16INK4a and p14ARF (p19ARF in the mouse) (for review, see Ref. 211). p16INK4a is an inhibitor of the cyclin D-dependent kinases, CDK4 and CDK6, whereas p14ARF antagonizes the function of the p53 negative regulator HDM2 (human homolog of Mdm-2 in the mouse). Recently, it has been shown that p14ARF is a strong HIF-1 inhibitor in a p53-independent fashion. p14ARF forms a complex with HIF-1α promoting its relocalization and sequestration in the nucleolus (56).
Wt1 (Wilms' tumor gene). HIF-1 was shown to upregulate the Wt1 gene that has central functions during development and in the pathogenesis of Wilms' tumor (also known as nephroblastoma) (238). Wt1 gene is a tumor suppressor gene, which is homozygous disrupted in ∼15% of sporadic Wilms' tumors (49). Nephroblastoma arises from the failure of mesenchymal blastema to differentiate into metanephric structures in the embryonic kidney, thereby exemplifying Wt1's dual role in development and tumorigenesis. In addition, Wt1 is a transcriptional activator of the antiapoptotic protein Bcl-2, which could positively contribute to tumor development (172).
HIF-1 TARGET GENES INVOLVED IN TUMORIGENESIS
The key role of HIF-1α in the response to physiological and pathological hypoxia is reflected in an exponential increase of HIF-related reports. To date, several dozen HIF-1 target genes have been identified, and the list grows steadily. Some of the most important HIF-1 target genes involved in tumorigenesis are discussed below.
VEGF is one of the most studied HIF-1 target genes due to its involvement in a series of important human pathologies involving hypoxic states such as ischemia-reperfusion and cancer (51). VEGF is the most potent promoter of neoangiogenesis, a process that improves tissue oxygenation through capillary density and is crucial for tumor development (59). The idea of having an avascular tumor mass depending on neoangiogenesis (and therefore on angiogenic factors such as VEGF to allow the tumor's growth) brought about a concept called the “angiogenic switch” (33, 73, 84, 209). This concept is based on the hypothesis that there is a balance between inducers and inhibitors of angiogenesis in normal tissues. Thus an imbalance toward induction either by reducing the inhibitor concentration, for example, because of loss of a tumor suppressor gene [such as thrombospondin-1 in the absence of p53 (46, 87)], or by increasing activator levels such as by VEGF induction, will lead to neoangiogenesis (84). Hypoxia is an important environmental factor directing the angiogenic switch, with HIF-1α playing a critical role in “flipping the switch” through direct upregulation of VEGF mRNA. Apart from hypoxia, activation or enhancement of HIF-1-mediated gene upregulation through several oncogenes/tumor suppressor genes under normoxic conditions may also contribute to the switch. Two recent reports confirmed the importance of HIF-1α in the angiogenic switch in in vivo experiments in mice tumor xenografts (55, 107).
Another important effect of VEGF is the induction of vascular permeability (241, 260) contributing significantly to high interstitial fluid and oncotic pressure found in solid tumors (22, 218). These characteristics are important because they may influence drug delivery into tumor tissue (116). Indeed, inhibition of VEGF in solid tumors improves drug delivery (117).
HIF-1 also upregulates a number of other molecules involved in angiogenesis and vascular tone regulation (31, 154, 174, 184, 255), including the multifunctional peptide adrenomedullin, which has been shown to be involved in tumor progression through several mechanisms including angiogenesis (63, 69, 267).
The upregulation of oxygen-independent metabolic pathways, such as glycolysis, to enhance the production of energy under reduced Po2 is a hallmark of the hypoxic response. Almost every enzyme of the glycolytic pathway, as well as glucose transporters crucial for glucose uptake, are upregulated by HIF-1 (113, 244). Also the mRNA level of an enzyme that ultimately leads to the allosteric activation of the rate-limiting enzyme of the glycolytic pathway (phosphofructokinase) is induced by hypoxia in an HIF-1-dependent manner (177). The final outcome of HIF-1 upregulation is a high glycolytic flux. Another important factor to consider, apart from the energetic point of view, is that precursors of the pyrimidine/purine pathways are produced during glycolysis. As such, stimulation of the glycolytic pathway, for instance in hypoxic cancer cells, may indirectly facilitate proliferation by enhancing the supply of DNA precursors (80). This would add an important anabolic role to the already well-known energetic role of the glycolytic pathway and could explain why tumors have high glycolytic rates even in the presence of oxygen, an effect historically known as the Warburg effect (242). Reinforcing this idea, a positive feedback loop between pyruvate and lactate (both being end products of glycolysis) and HIF-1 activation has been shown in glioma cells (158).
The glycolytic shift has other important consequences for tumor pathophysiology. The end product of anaerobic glycolysis is lactate. The enhanced production of this metabolite, possibly in association with the HIF-1-mediated upregulation of carbonic anhydrases [CA; specifically the isoforms CAIX and CAXII (253)], could be a major cause for the acidic pH usually found in tumors (220). CA expression and acidic pH have been related to tumor invasiveness (168, 185, 196) and can be used as a marker of tumor hypoxia (12, 157, 182). Nevertheless, CAIX expression as a prognostic factor is still controversial (29, 39, 70, 224). Also of interest is the fact that CAIX is upregulated by HIF-1-independent mechanisms in mild hypoxic states (1-5% O2), where HIF-1α is not optimally expressed (129).
Is HIF-1 a Proapoptotic Factor?
The major role of HIF-1 in the hypoxic response is to promote cellular adaptation through various mechanisms. The majority of these adaptation mechanisms promote cell survival in adverse environments. Intriguingly, despite all possible adaptation advantages HIF-1 activity brings, direct mutations leading to HIF-1 gain-of-function in tumors have not been found. Mutations in HIF-1α ODD have only been found in human hormone-refractory prostate cancers, but the importance of these mutations for tumor progression is still unclear (7). In addition, VHL deletion by itself (the hallmark of the VHL disease) does not promote teratocarcinoma growth despite HIF-1α accumulation (161). One possible explanation is that the consequences of HIF-1 activation, such as neoangiogenesis promotion, are not clonally selected. Another possibility may rely on the fact that HIF-1 also upregulates proapoptotic proteins. An example is Nip3, which is upregulated after prolonged exposure to hypoxia (only expressed after 4 days of exposure to 0.5% O2) (27). It is speculated that, due to their slow response and modest apoptotic activity, hypoxic cells/tissues have a critical opportunity to adapt to oxygen deprivation by means of activation of an initial HIF-1-dependent protective response. However, persistent O2 deprivation could cause Nip3 accumulation and promote cell death.
Two studies using HIF-1α+/+ and HIF-1α-/- cells support the idea of HIF-1α as a proapoptotic protein during tumor progression. It has been shown that parameters concerning perfusion as well as oxygenation and VEGF expression levels were reduced in wild-type tumors, whereas threefold more hypoxic areas were found in HIF-1α-/- tumors (32). Surprisingly, despite all the differences mentioned above, HIF-1α-/- tumors grew at the same rate as the HIF-1α+/+ tumors during the first 3 wk of observation and clearly outpaced wild-type growth in the following weeks. This outcome was explained by an increased proliferation rate of the HIF-1α-/- cells after 3 wk of growth, accompanied by diminished apoptosis and increased cellular stress. The authors concluded that the loss of HIF-1α affects tumor growth by progressive selection of intensely proliferating cells, as occurs in the absence of p53 (78). Interestingly, this was the only publication showing enhanced growth of HIF-1α-/- tumors, whereas all the other reports claimed that loss of either HIF-1α or HIF-1β leads to retardation of tumor growth compared with wild-type tumors (80, 101, 121, 169, 198, 199, 251). In accordance, the ratio of apoptosis between HIF-1α-/- and HIF-1α+/+ tumors in these studies were similar (101, 169, 251) or even increased in the HIF-1α-/- tumors (198). Thus these data were in striking contrast to the study mentioned above (32). A second report using the same cells as in Ref. 32 describes the expression of HIF-1α as a function of the distance to the blood vessels in mixed HIF-1α+/+ and HIF-1α-/- tumors and found that HIF-1α+/+ cells are preferentially located near blood vessels, whereas HIF-1α-/- cells had preferentially a distal localization (258). These authors speculated that the HIF-1α+/+ cells undergo marked apoptosis at more distal locations where nutrients and oxygen are more limited. Note that in one of the few reports showing HIF-1α expression as a marker of good prognosis (Ref. 237; see below), HIF-1α also correlated with apoptosis and proapoptotic factors such as caspase-3, Fas, and Fas ligand.
Nevertheless, there are other reports supporting HIF-1's antiapoptotic properties. Increased HIF-1 DNA binding and target gene expression (such as erythropoietin, p21, and glycolytic enzymes) conferred protection against oxidative stress-induced apoptosis in cortical neurons (263). HIF-1α inhibition caused p53-independent apoptosis in glioblastoma cells (45) and HIF-1α constitutive expression rendered pancreatic cancer cells apoptosis resistant (3).
Thus, although HIF-1 activation mainly results in adaptation aimed at cell survival, one should bear in mind that depending on the cellular background HIF may also lead to augmented levels of apoptosis.
HIFs AND MALIGNANCY
The possibility of using polarographic electrodes in clinical settings resulted in a series of clinical studies correlating hypoxia and tumor malignant progression (93-95, 97). In addition, HIF-1α overexpression was detected in a variety of human cancers and their metastases (141, 227, 266). HIF-1α expression levels correlate with tumor grade and invasion in gliomas (262) and pathological stage in breast cancer with the levels of HIF-1α being higher in poorly differentiated lesions than in the well-differentiated ones (21). Additionally, in these tumors, HIF-1α expression is associated with proliferation and VEGF and estrogen receptor expression, which potentially correlate with malignancy. In patients with advanced-stage and lymph node negative breast cancer, HIF-1α protein overexpression was associated with an overall significantly shorter and disease-free survival time (20, 203). Others (16) found it to be a negative prognostic parameter for survival in oligodendriomas. HIF-1α was indicated as a novel predictive and prognostic parameter in the radiotherapy of different cancers (1, 11, 142).
The association of HIF-1α expression with tumor suppressor gene/oncogenes may even be more valuable as a prognostic factor. In ovarian cancer (17), the combination between mutant p53 expression and HIF-1α overexpression was associated with a highly significant increase in risk of patient mortality. In addition, combination of HIF-1α overexpression with low levels of the antiapoptotic protein Bcl-2 in early-stage esophageal cancer is associated with treatment failure (143).
A number of molecules involved in invasion such as urokinase plasminogen activator receptor (uPAR), matrix-metalloproteinase 2, cathepsin D, and fibronectin 1 among others have been shown to be upregulated at the mRNA level only in HIF-1α+/+ cells but not in the HIF-1α-/- ones (147). This in association with advantages brought about by the upregulation of other target genes such as glycolytic enzymes and VEGF would provide a molecular basis for the association between bad prognosis and HIF-1α seen in many tumors (147). It has been pointed out that invasion-related genes are activated by HIF-1 in a cell-specific fashion but that HIF-1α overexpression can affect multiple aspects of the invasion process.
HIF-1 AND THERAPY
The importance of the HIF pathway in tumorigenesis is well established based on great amounts of experimental data. Many of the enzymes involved in this pathway (for example prolylhydroxylases) are possible therapeutic targets. As such, the idea of interfering with this pathway to reduce/impair tumor growth is very attractive. Accordingly various compounds (36, 111, 148, 150, 160, 176, 256), anti-sense approaches (45, 221), and peptides interfering with the interaction between HIF-1 and the transcriptional coactivator CBP/p300 (149) have been shown to negatively interfere with the expression of HIF-1α, leading to angiogenesis inhibition and reduced tumor growth. However, it still remains to be defined whether and which of these approaches will lead to success in clinical settings.
The fact that tumors have lower Po2 than the corresponding original tissues can also be explored to improve specificity in cancer therapy. In the case of gene therapy, the problem of achieving selectiveness is accessed through transcriptional targeting using inducible enhancer/promoter units activated only under the conditions existent in neoplastic but not in normal cells/tissues. A number of hypoxia-regulated vectors have been developed to explore the HIF-1/HRE system in gene therapy (15, 186, 213). The HRE copy number seems to be central for the enhanced gene expression under hypoxic conditions (197, 212). This strategy has been often used in gene-directed prodrug therapy and has the advantage of not being restricted to a single type of tumor.
The discovery of the PHDs and of FIH-1 also revealed possible therapeutic targets for ischemia and/or cancer treatment. For instance, in ischemia treatment one could envision a pharmacological inhibition of one or more PHDs, thereby triggering an enhanced hypoxic response that could be beneficial after ischemic injuries (243, 250, 252). An important aspect in oncology is that HIF-1α is highly expressed in the great majority of tumors and thus contributes to cancer progression (266). Therefore acceleration of its degradation might improve therapy outcome. At least one prolyl-hydroxylase activating compound has been identified so far (48). Because prolyl-hydroxylase activity is reversibly saturable, HIF-1α overexpression leads to HIF-1α accumulation (98) due to saturation. On the other hand, supplementation of physiological concentrations of cofactors necessary for the prolyl-hydroxylase activity such as ascorbate and iron have great negative effects on the activation of HIF-1 target genes because of increased HIF-1α degradation (92, 138).
The Hsp90 inhibitor geldanamycin promotes the dissociation of HIF-1α from Hsp90 and enhances its ubiquitination (shown in VHL-/- cells) and proteosomal degradation, thereby demonstrating the importance of Hsp90 for the stability of HIF-1α protein. In addition, geldanamycin stimulated degradation of HIF-1α in a panel of cancer cell lines (160). Radicicol, another Hsp90 inhibitor had a greater impact on HIF-1 transcriptional activity than on HIF-1α protein stability itself (106). Most interestingly, geldanamycin (67) and an oxime derivative of radicicol termed KF58333 (150) inhibited the growth of xenografts due to antiproliferative and antiangiogenic effects that could be attributed to HIF-1 inhibition.
The importance of HIF-1 in physiological as well as pathological conditions is quite recent. HIF-1α was first discovered in 1995 (239), and the first study showing that the lack of HIF-1β retarded tumor growth was performed in 1997 (121). This means that in less than a decade a whole new field of intensive investigation with many important therapeutic consequences has become a major issue. HIF-1α has been shown to be an adverse prognostic factor and knowledge about the pathway has brought about a myriad of possible therapeutic targets such as the HIF-prolyl hydroxylases as well as new therapy opportunities such as hypoxia-directed gene therapy. In addition, the interplay between tumor suppressor genes/oncogenes, growth factors, and phosphorylation cascades and the HIF-pathway has helped to further clarify the mechanism of their involvement in tumorigenesis. But yet, despite compelling evidence such as overexpression in the majority of cancers and metastasis (266), progressive expression levels correlated with increased tumor malignancy (21), and interference with the pathway leading to diminished xenograft tumor growth (149), some doubt still remains about the contribution of HIF-1 to tumor development. Conversely, HIF-1α is a marker of good prognosis in some tumors (237): it upregulates proapoptotic proteins (215) and loss of VHL causing HIF dysregulation and does not promote tumor growth (161). These facts reflect the multifactorial nature of cancer and the fact that, in cells derived from different tissues and exposed to different amounts/types of growth factors, the modulation of the hypoxic response and the relative expression of hypoxic regulated genes may be different and therefore result in different tumor progression outcome. Adding to the already inherent complexity of the HIF-1-mediated hypoxic response, the discovery of further HIFαs boosts the difficulty of the analysis.
In summary, the HIF pathway has attracted much attention in the last years due to its implication in tumorigenesis, and many of these efforts have led to the discovery of promising potential therapeutic targets. In addition, the definition of which type of cell preferentially expresses either subunit and how this leads to different hypoxic responses (79) could have important consequences for tumor research. However, some aspects are still under debate, and the exact contribution of HIF-1α and HIF-2α for tumor progression remains unclear and requires more rigorous study.
This work was supported by the Swiss National Science Foundation (31000A0-100214) and the 6th Framework Programme of the European Commission (Project EUROXY).
- Copyright © 2004 the American Physiological Society