Tumor hypoxia
Regions of profound hypoxia and necrosis are common in solid tumors and their presence correlates with an aggressive clinical course. Classical studies relating necrotic regions of tumor to blood vessel disposition suggested that tumor necrosis was, at least in part, driven by hypoxia [
17]. More recently, this has been supported by the demonstration of striking up-regulation of HIF target genes in regions immediately adjacent to the necrotic areas [
18]. This pattern of gene expression is absent in experimental tumors that are genetically defective for HIF, clearly indicating that it is driven by this pathway [
19].
Given that HIF target genes include many of those underpinning dysregulated tumor metabolism [
20] and that both tumor hypoxia [
21] and the extent of dysregulated metabolism [
2,
5] show clear correlations with aggressive cancer phenotypes, it is tempting to conclude that activation of HIF pathways by tumor hypoxia itself is the major cause of dysregulated tumor metabolism. However, a number of observations reveal that the interconnections are more complex. First, within solid tumors, regions of hypoxia (assessed either by
in vivo imaging or by use of histochemical markers) show less co-incidence with regions of HIF up-regulation than might be expected [
22‐
24]. Second, hematological malignancies that do not involve solid tissue masses also manifest up-regulation of HIF [
25,
26]. Taken together these studies indicate that although micro-environmental hypoxia clearly contributes to the activation of HIF in cancer, other factors must also be important.
Multiple metabolic pathways impact on the regulation of HIF, raising the possibility that in addition to HIF activation driving dysregulated metabolism in cancer, dysregulated metabolism promotes the activation of HIF. One possibility is that altered 2-OG availability modulates HIF hydroxylation. In addition to its role in the Krebs cycle, 2-OG also serves as a co-substrate/product in reductive amidation/oxidative deamidation by glutamate dehydrogenase and is the major amino group acceptor for transaminases. Thus, 2-OG would be well placed to act as a metabolic sensor regulating HIF hydroxylase activity. In keeping with this, reduced intracellular 2-OG in cultured cells depleted of amino acids has recently been reported to reduce PHD activity [
27]. Unexpectedly, in this case, the regulation of PHD activity had no effect on HIF-α protein levels, but was reported to effect mTORC1 activation by amino acids. The PHD enzymes were suggested to play a role as metabolic sensors linking amino acid availability with the mTORC1 pathway, although whether limiting 2-OG availability modulates HIF hydroxylation in this setting remains unclear.
Other Krebs cycle intermediates and endogenous organic acid metabolites may also alter HIF hydroxylase activity by competing with 2-OG at the catalytic site (fumarate and succinate) or by product inhibition (succinate) [
28,
29]. In addition, in different assays, citrate, isocitrate, malate and oxaloacetate have been reported to bind or inhibit recombinant HIF hydroxylases. Different HIF hydroxylases are differentially sensitive to these inhibitors. For instance, fumarate is a more potent inhibitor of the PHDs than FIH, whereas citrate is a more potent inhibitor of FIH than the PHDs [
28,
30]. Both fumarate and succinate reach very high levels in hereditary cancers associated with the inactivation of fumarate hydratase (FH, hereditary leiomyomatosis and papillary renal cell carcinoma) and succinate dehydrogenase (SDH, hereditary paraganglioma) respectively [
31‐
33]. In these settings, fumarate and succinate clearly induce HIF, at least in part by inhibiting the PHD enzymes [
30]. However, whether and under what circumstances the Krebs cycle and other metabolic intermediates reach the levels required to inhibit the HIF hydroxylases in common cancers is less clear.
Interestingly, a number of studies has demonstrated that provision of glucose in cell culture medium and/or on-going glucose metabolism is necessary for the induction of HIF-1α by hypoxia [
34]. Different investigators have provided evidence for a range of mechanisms. Provision of lactate or pyruvate has been shown to stabilize HIF-1α in glucose depleted cell cultures and it has been suggested that this effect is mediated by inhibition of the PHDs by pyruvate produced from lactate by lactate dehydrogenase [
35]. Lactate often accumulates to high levels (10 mM) in tumors and has also been proposed to activate HIF and vascular endothelial growth factor (VEGF)-mediated angiogenesis in cancer [
36]. Oddly, neither pyruvate nor lactate was found to compete with 2-OG or inhibit purified recombinant PHDs under standard (ascorbate containing) reaction conditions [
28,
29]. Some insight into this paradox may be provided by recent work suggesting a different mode of inhibition, whereby pyruvate and oxaloacetate inactivate the PHDs by oxidation, which is reversed by ascorbate [
37]. Thus, metabolic intermediates have the potential to inhibit the HIF hydroxylases by at least two mechanisms; competition with 2-OG and oxidation.
Recently, interest has also focused on another 'oncometabolite’, 2-hydroxyglutarate (2-HG), which has the potential to competitively inhibit 2-OG dioxygenases. Specific mutations in the genes encoding isocitrate dehydrogenases (IDH) 1 and 2 have been observed at a high rate in low and medium grade gliomas, secondary glioblastoma, acute myeloid leukemia, and at a lower rate in other malignancies, including myelodysplastic syndromes, T-cell lymphoma, chondrosarcoma and cholangiocarcinoma [
38,
39]. In affected cells, 2-HG is formed as a result of reduction of 2-OG by the abnormal enzyme and accumulates to very high levels [
40]. However, 2-HG, particularly the 'R’ enantiomer that is formed by the mutant IDH enzymes, is a poor inhibitor of the PHD enzymes [
41]. It is therefore unlikely to contribute to any up-regulation of HIF that is observed in these settings and has even been reported to activate PHD2, resulting in a reduction in HIF [
42].
In addition to effects of metabolites on the 'oxygen sensing’ hydroxylation reaction, multiple interactions of metabolic and HIF signaling pathways have been defined at other levels. For instance, HIF-α levels are subject to complex translational controls operating through nutrient and cellular energy-sensing mTOR complexes, with HIF-1α being regulated through mTORC1 and 2, and HIF-2α principally by mTORC2 [
43]. In normal cells, several mechanisms exist whereby hypoxia can reduce translation either through mTOR pathways or via regulation of eIF2α or eEF2. These pathways are themselves independent of HIF and involve activation of AMPK or PERK in response to metabolic changes arising from hypoxia [
44]. However, translational control can also be HIF-dependent. For instance, HIF activates transcription of REDD1 [
45], which activates the tuberous sclerosis TSC1/2 tumor suppressor complex [
46], an upstream inhibitor of mTORC1. Cancer cells can evade this down-regulation of HIF-α translation at least in part through oncogenic dysregulation of mTOR complexes.
Yet another interface with metabolism is mediated by reversible acetylation at specific sites in HIF-α polypeptides. HIF-1α can be acetylated at multiple lysine residues by acetyltransferases, such as p300/CBP associated factor (PCAF), and acetylation can be reversed by several classes of enzyme, including classical histone deacetylases (HDACs) and sirtuins [
47‐
49]. Given their sensitivity to another key parameter of energy status, the cellular NAD+/NADH ratio, the interface between sirtuins and HIF has attracted widespread interest. The mammalian sirtuin family (SIRT1-SIRT7) of lysine deacetylases couple deacetylation with NAD + hydrolysis. SIRT1, 3 and 6 have all been implicated in the regulation of HIF activity [
48,
50‐
53], although there is disagreement as to the exact nature of the interconnections. In one study, deacetylation of HIF-1α at K674 by SIRT1 was reported to block p300 recruitment by HIF-1α, with the inactivation of SIRT1 in hypoxia (by decreased NAD + levels) releasing this negative control [
48]. Another study has reported SIRT1 activity to be required for full activity of HIF-1α in hypoxia [
53], while a third reports a specific functional interaction between SIRT1 and HIF-2α leading to up-regulation of HIF-2α transactivation in hypoxia [
52]. SIRT1 itself is also reported to be a HIF target gene in some [
54], but not all, settings [
53].
Iron, ascorbate and oxidant stresses
The binding of Fe(II) at the catalytic center of the HIF hydroxylases, like that of other 2-OG dioxygenases, is relatively labile. The enzymes also require ascorbate for maintenance of an active Fe(II) catalytic center. These properties render them susceptible to modulation by redox signals and iron availability, raising questions as to whether abnormalities in redox status and/or iron availability provide another link between abnormal metabolism in rapidly dividing cancer cells and activation of HIF.
In tissue culture, supplementation with either iron or ascorbate promotes HIF hydroxylase activity and suppresses basal HIF levels in oxygenated cells [
55]. Cancer patients are often poorly nourished and systemic iron deficiency is common [
56]. Furthermore, rapid growth and/or poor blood supply may exacerbate cellular iron and ascorbate deficiencies within the tumor. Somewhat surprisingly, the possibility that iron and ascorbate deficiency may be important contributors to HIF activation in clinical cancer has not been intensively investigated. In scorbutic rodents, ascorbate supplementation did not affect physiological measures of HIF activation, such as the production of erythropoietin, suggesting that tissue culture studies may not be representative of effects in the intact organism [
57]. Nevertheless, low cellular ascorbate has recently been associated with increased HIF and aggressive phenotype in clinical endometrial cancer [
58].
The ability of both iron deficiency and redox stresses to up-regulate HIF in tumors is strongly supported by experimental studies. For instance, the suppression of iron uptake by shRNA-mediated knockdown of transferrin receptor-1 has been shown to activate HIF and enhance angiogenesis in a breast cancer cell line xenograft model [
59]. In tumors derived from Ki-Ras transformed fibroblasts, activation of the antioxidant response by junD has been reported to enhance PHD activity, reduce HIF and impair angiogenesis [
60]. In another xenograft model, both ascorbate and the anti-oxidant N-acetylcysteine were found to suppress HIF activation and growth of human lymphoma cells through increased hydroxylation [
61]. Kinetic studies on purified recombinant enzyme have indicated that reducing agents other than ascorbate can only partially substitute for ascorbate in activating the HIF hydroxylases [
62], and it is possible that in cells, these agents are acting indirectly on processes that affect cellular Fe(II) or ascorbate levels.
It has also been proposed that the increased production of mitochondrial reactive oxygen species in hypoxia contributes to HIF activation by impairing the activity of HIF hydroxylases (reviewed in [
63]). However, whether reduced activation of HIF following the application of mitochondrial inhibitors arises from a reduction in reactive oxygen species, or an increase in intracellular oxygen levels (as a result of reduced mitochondrial oxygen consumption) is controversial (reviewed in [
64]). Interestingly, HIF asparaginyl hydroxylation is much more sensitive to inhibition by hydrogen peroxide than HIF prolyl hydroxylation [
65], whereas the reverse is true for inhibition by hypoxia [
66]. This suggests that hypoxia and reactive oxygen species affect HIF signaling by distinct mechanisms. Since HIF asparaginyl hydroxylation persists under all but the most severe levels of hypoxia [
66] and specifically modulates the expression of some but not all HIF target genes [
67], these findings also suggest that interplay between hypoxia and redox signals in tumors not only activates HIF, but shapes the nature of the HIF transcriptional response.
In addition to modulation of HIF hydroxylase activity, redox signals and oxidant stress (such as metabolic dysregulation) impinge on the HIF pathway at many other levels. Effects are observed on both transcription and translation of individual HIF-α isoforms. For instance, the HIF-1α promoter contains a well characterized nuclear factor kappa B (NF-kB) binding site that conveys up-regulation by oxidant stresses [
68], while HIF-1β transcription can also be activated directly by NF-kB [
69]. Studies of the effects of nicotinamide adenine dinucleotide phosphate (NADPH) oxidases in pVHL-defective renal cancer cell lines (in which the pVHL-dependent proteolysis is disrupted) suggest several other levels of control. For instance, NADPH oxidase, specifically Nox4, was found to elevate HIF-2α mRNA levels [
70], whereas NADPH oxidase-dependent generation of reactive oxygen species has been proposed to enhance translation of HIF-2α [
71]. Taken together, these findings reveal multiple interactions with redox signals that have the potential to affect both quantitative and qualitative aspects of HIF pathway activation in cancer.
Oncogenic and tumor suppressor pathways
In addition to activation by multiple micro-environmental stimuli, the HIF system is activated by diverse tumor suppressor and oncogene pathways. The most striking of these is mutation of the von Hippel-Lindau tumor (
VHL) suppressor (reviewed in [
72]). As outlined above, pVHL is part of the ubiquitin E3 ligase complex that targets HIF-α subunits to the ubiquitin-proteasome pathway. Biallelic inactivation of
VHL thus blocks oxygen-dependent proteolysis of HIF-α and leads to constitutive activation of the HIF pathway. Interestingly, however, more detailed analysis of HIF in pVHL-associated cancer has revealed the importance of quantitative effects on HIF activation. In particular, there is a clear correlation between the quantitative effects of specific mutations on HIF dysregulation and the prevalence of different types of neoplasia in families affected by VHL disease [
73,
74]. Severe dysregulation of HIF is associated with a predisposition to renal cancer, but appears to be incompatible with pVHL-associated phaeochromocytoma, which is associated with partially inactivating mutations that lead to more modest levels of HIF pathway activation.
HIF is also activated by a range of growth factors acting through PI3K/PTEN/AKT or RAS/RAF/MAPK signaling cascades (reviewed in [
75]). Activation of these pathways by somatic mutation and gene amplification is common in many types of cancer and dysregulation of the PI3K/PTEN/AKT pathway leads to up-regulation of HIF through increased synthesis of HIF-α subunits [
76]. The AKT serine/threonine kinase has multiple downstream targets, and likely increases HIF-α translation by both mTOR-dependent and mTOR-independent mechanisms [
77]. It is also possible that AKT may increase HIF-α levels through other mechanisms. For example, another substrate of AKT, GSK3b, has been implicated in regulating HIF-1α protein degradation through a pVHL independent mechanism [
78].
The RAS/RAF/MAPK pathway has been reported to impact on HIF activity primarily through the regulation of transactivation. Phosphorylation of either HIF-1α or the co-activator p300 by different kinases (either p42/p44 MAPK or p38) activates HIF, both by promoting the formation of HIF/p300 complexes and by enhancing p300 transactivation [
79].
Diverse interactions between HIF and p53 tumor suppressor pathways have been reported (reviewed in [
80,
81]). Though not all reports are in agreement, the induction of p53 has generally been shown to suppress HIF activity. Both direct physical interactions between p53 and HIF-1α [
82] and indirect functional interactions have been described, including competition between p53 and HIF-α for the p300 co-activator [
83] and p53-dependent promotion of HIF-α degradation by the mouse double minute 2 homolog (MDM2) ubiquitin-ligase [
84].