|Home | About | Journals | Submit | Contact Us | Français|
Adaptation of cancer cells to their microenvironment is an important driving force in the clonal selection that leads to invasive and metastatic disease. O2 concentrations are markedly reduced in many human cancers compared to normal tissue and a major mechanism mediating adaptive responses to reduced O2 availability (hypoxia) is the regulation of transcription by hypoxia-inducible factor 1 (HIF-1). This review summarizes the current state of knowledge regarding the molecular mechanisms by which HIF-1 contributes to cancer progression, focusing on (i) clinical data associating increased HIF-1 levels with patient mortality; (ii) preclinical data linking HIF-1 activity with tumor growth; (iii) molecular data linking specific HIF-1 target gene products to critical aspects of cancer biology; and (iv) pharmacological data demonstrating anti-cancer effects of HIF-1 inhibitors in mouse models of human cancer.
Human cancers frequently contain areas of necrosis in which cancer cells have died due to inadequate oxygenation (Harris, 2002; Brahimi-Horn et al., 2007). Cells closest to a perfused blood vessel are exposed to relatively high O2 concentrations, which decline as distance from the vessel increases. Although such gradients exist in normal tissues, in cancers the gradients are much steeper and O2 concentrations drop to near zero in areas of necrosis. In addition to physical gradients, temporal fluctuations in oxygenation also commonly occur within tumors (Dewhirst et al., 2008). Most physiological functions of cells are modulated according to the cellular O2 concentration. A major mechanism mediating adaptive responses to reduced O2 availability (hypoxia) is the regulation of transcription by hypoxia-inducible factor 1 (HIF-1) (Semenza, 2009a). These adaptive responses are co-opted by cancer cells, in which normal feedback mechanisms have been disrupted by somatic mutation and epigenetic changes. As a result, the adaptation to hypoxia promotes many key aspects of cancer progression that ultimately lead to patient mortality (Harris, 2002).
HIF-1 is a heterodimeric protein that is composed of a constitutively expressed HIF-1β subunit and an O2-regulated HIF-1α subunit (Wang and Semenza, 1995; Wang et al., 1995). HIF-1α is subjected to O2-dependent hydroxylation on proline residue 402 and/or 564 by prolyl hydroxylase domain protein 2 (PHD2) and this modification creates an interface for interaction with the von Hippel-Lindau tumor suppressor protein (VHL), which recruits an E3 ubiquitin-protein ligase that catalyzes polyubiquitination of HIF-1α, thereby targeting it for proteasomal degradation (Kaelin and Ratcliffe, 2008). Under hypoxic conditions, hydroxylation is inhibited and HIF-1α rapidly accumulates, dimerizes with HIF-1β, binds to the core DNA binding sequence 5′-RCGTG-3′ (R, purine [A or G]) in target genes, recruits coactivators, and activates transcription. O2-dependent hydroxylation of asparagine-803 by factor inhibiting HIF-1 (FIH-1) blocks interaction of HIF-1α with the coactivators P300 and CBP under normoxic conditions (Lando et al., 2002). Both PHD2 and FIH-1 utilize O2 and α-ketoglutarate as substrates and generate CO2 and succinate as by-products of the hydroxylation reaction. HIF-2α is a protein with extensive sequence similarity to HIF-1α that is also regulated by proline and asparagine hydroxylation, dimerizes with HIF-1β, and activates transcription of a group of target genes that overlaps with, but is distinct from, those regulated by HIF-1α (Lau et al., 2007). HIF-3α is an inhibitor of HIF-1 that may be involved in feedback regulation because its expression is transcriptionally regulated by HIF-1 (Makino et al., 2007).
Immunohistochemical analysis of human cancer biopsies revealed increased levels (relative to surrounding normal tissue) of HIF-1α or HIF-2α protein (or both) in the majority of primary human cancers and their metastases (Zhong et al., 1999; Talks et al., 2000). Intratumoral hypoxia is a major mechanism underlying the increased levels of HIF-1α and HIF-2α in cancer and stromal cells. For example, the median PO2 level measured in breast cancers was 10 mm Hg, as compared to 65 mm Hg in normal breast tissue (Vaupel et al., 2004). Other inducers of HIF-1α in the tumor microenvironment include reactive oxygen and nitrogen species, which also inhibit proteasomal degradation of HIF-1α (Quintero et al., 2006; Gao et al., 2007; Li et al., 2007; Dewhirst et al., 2008). Complementing these mechanisms for blocking HIF-1α degradation, activation of the phosphatidylinositol-3-kinase and MAP kinase pathways (either as a result of oncogenic mutation or increased signaling from receptor tyrosine kinases and G-protein coupled receptors) increases HIF-1α synthesis, primarily through the action of mTOR (Laughner et al., 2001).
HIF-1α and HIF-2α protein levels can also be increased in cancer cells due to loss of function of many different tumor suppressors, which results in either increased HIF-1α synthesis or decreased HIF-1α degradation (Table 1). Remarkably, a large number of proteins encoded by transforming viruses that cause tumors in humans also induce HIF-1 activity (Table 2). Most remarkable is the case of Kaposi sarcoma, a highly vascularized tumor that is caused by a infection with a herpesvirus, the genome of which encodes three different proteins that together increase HIF-1α protein half-life, nuclear localization, and transactivation under non-hypoxic coinditions, thereby mimicking the effect of hypoxia. The finding that, in addition to intratumor hypoxia, major genetic and epigenetic alterations resulting in oncogene gain of function or tumor suppressor gene loss of function lead to increased HIF-1 activity suggests that increased HIF-1 activity represents a final common pathway in cancer pathogenesis, i.e. clonal selection favors those cancer cells in which HIF-1 activity is increased (Semenza, 2000; Vogelstein and Kinzler, 2004; Gillies and Gatenby, 2007; Gatenby and Gillies, 2008).
Taken together, the observed effects of tumor suppressor loss of function and transforming virus protein expression provide compelling evidence that HIF-1 activation promotes oncogenesis and/or cancer progression. In support of this hypothesis, a large body of clinical data demonstrates an association between increased levels of HIF-1α or HIF-2α protein, as determined by immunohistochemistry of standard formalin-fixed, paraffin-imbedded, tumor biopsy sections, with increased patient mortality in many different human cancers (Table 3). The data from studies of early-stage breast, cervical, and endometrial cancers are particularly striking. Early stage disease is usually associated with a good prognosis, but the subset of patients whose tumors contained high levels of HIF-1α had a significantly increased mortality rate. Thus, HIF-1α immunohistochemistry has the potential to identify patients for whom more aggressive therapy may be indicated. Testing this hypothesis by clinical trials is warranted, especially now that anti-cancer drugs that target HIF-1 are available, as described below.
Complementing the clinical data is a large body of experimental data demonstrating that HIF-1α loss of function (LOF) results in decreased tumor growth, vascularization, and metastasis, whereas HIF-1α gain of function (GOF) has the opposite effects (Table 4). Whereas the clinical data are by their nature correlative, the experimental data demonstrate causation, using many different cancer cell types and many different techniques to modulate the levels of HIF-1α or HIF-2α. Whereas the overall conclusion of these experiments is that HIF-1α and HIF-2α promote cancer progression, the association is not absolute. For example, whereas HIF-1α LOF decreased the growth of SW480 colorectal carcinoma xenografts, HIF-2α LOF increased xenograft growth and, remarkably, immunohistochemical analysis of human colon cancer biopsies revealed a significant association between loss of HIF-2α protein expression and advanced tumor stage (Imamura et al., 2009). In contrast, HIF-2α GOF increased the growth of 786-O renal cell carcinoma xenografts whereas HIF-1α GOF decreased xenograft growth (Raval et al., 2005).
The xenograft studies have many limitations, not the least of which is that LOF is achieved in the tumor cells but not in the host stromal cells, in which hypoxia-induced HIF-1 activity contributes to tumor progression. For example, tumor growth is impaired in mice with mutant p53 and germline heterozygosity for a HIF-1α knockout allele as compared to mutant p53 mice that are wild type for HIF-1α (Bertout et al., 2009) and in mice with conditional knockout of HIF-1α in vascular endothelial cells (Tang et al., 2004). There are no published examples of combined HIF-1α + HIF-2α inhibition in tumor + stroma resulting in increased tumor growth. Thus, inhibitors that target HIF-1α and HIF-2α will have the highest probability of therapeutic efficacy across a broad range of human cancers.
HIF-1α and HIF-2α exert their effects on cancer progression by binding to and activating the transcription of target genes with cis-acting hypoxia response elements containing the consensus binding site 5′-RCGTG-3′. The hundreds of genes that are induced by hypoxia in a HIF-1-dependent manner encode proteins from A to Z that play key roles in every critical aspect of cancer biology (Table 5), including angiogenesis (Liao and Johnson, 2007), cell survival, chemotherapy and radiation resistance (Moeller et al., 2007), genetic instability (Bindra et al., 2007), immortalization, immune evasion (Lukashev et al., 2007), invasion and metastasis (Chan and Giaccia, 2007; Sullivan and Graham, 2007), proliferation, metabolism and pH regulation (Swietach et al., 2007; Chiche et al., 2009; Semenza, 2009b), and stem cell maintenance (Barnhart and Simon, 2007). Among these functions, the most fundamental may be angiogenesis and glucose/energy metabolism, which control oxygen delivery and utilization, respectively. Angiogenesis has been a major focus of cancer biology and therapy over the last decade, while tumor metabolism is likely to attract similar attention over the next decade.
When a given cell type is exposed to hypoxia, increased expression of several hundred mRNAs is induced and expression of a roughly equal number are repressed in a HIF-1-dependent manner (Manalo et al., 2008). The hypoxia-induced binding of HIF-1 to many target genes that are activated by hypoxia has been demonstrated by chromatin immunoprecipitation assays, whereas binding of HIF-1 to hypoxia-repressed genes is not observed, suggesting an indirect, but HIF-1-dependent, mechanism for repression (Mole et al., 2009). For example, loss of E-cadherin expression is essential for cancer cell invasion and metastasis (Cavallaro and Christofori, 2004). In the renal cell carcinoma line RCC4, HIF-1 activates transcription of the genes encoding TCF3 (also known as E12/E47), ZFHX1A (also known as δEF1 or ZEB1), and ZFHX1B (also known as SIP1 or ZEB2), which are known to bind to the promoter of the gene encoding E-cadherin to repress its transcription (Krishnamachary et al., 2006). HIF-1 may also promote genetic instability by repressing transcription of the MSH2 and MSH6 genes, which encode subunits of the DNA mismatch repair protein MutSα, by blocking the interaction of MYC with SP1, an effect that does not require direct DNA binding or even the presence of HIF-1β (Koshiji et al., 2005).
Within any given cancer, only a subset of these genes will be activated by HIF-1, which increases the transcription of genes that are already active within a cell type. In addition, subsets of these genes are differentially regulated by HIF-1α and HIF-2α. Thus, the biological consequences of HIF-1 activation in a cancer will vary depending upon the battery of target genes that respond. This heterogeneity is in no way unique to HIF-1 (Yu et al., 1999) and represents one of the greatest obstacles to improving cancer therapy. The major conclusion to be drawn from these data is that HIF-1 plays an incredibly broad and important role in cancer biology and that the delineation of the HIF-1 transcriptome has provided a molecular basis for the association between intratumoral hypoxia and invasion, metastasis, and patient mortality (Vaupel et al., 2004).
Many of the novel anti-cancer drugs that target specific pathways have been shown to have anti-angiogenic effects that appear to be due in large part to their inhibition of HIF-1 activity (Table 6), including the BCR-ABL inhibitor imatinib/Gleevec (Mayerhofer et al., 2002); epidermal growth factor receptor inhibitors gefitinib/Iressa, erlotinib/Tarceva, and cetuximab/C225 (Luwor et al., 2005; Pore et al., 2006); HER2neu inhibitor trastuzumab/Herceptin (Laughner et al., 2001); and the mTOR inhibitors rapamycin, temsirolimus/CCI-779, and everolimus/RAD-001 (Laughner et al., 2001; Majumder et al., 2004; Thomas et al., 2006), which all act by inhibiting mTOR-dependent translation of HIF-1α mRNA into protein (Laughner et al., 2001). Inhibition of HIF-1 activity leads to decreased VEGF expression. Drugs that inhibit phosphatidylinositol-3-kinase and ERK signal transduction inhibit HIF-1 and SP1, both of which bind to the VEGF gene promoter (Pagès and Pouysségur, 2005).
Another novel class of molecularly targeted anti-cancer agents consists of inhibitors of heat shock protein 90 (HSP90) such as geldanamycin, 17-allylaminogeldanamycin (17-AAG), and 17-dimethylaminoethylamino-17-demethoxygeldanamycin (17-DMAG), which target HIF-1a for proteasomal degradation (Isaacs et al., 2002). In the absence of inhibitors, HSP90 binds to HIF-1α and promotes its stability by preventing the binding of RACK1, whereas the inhibitors block HSP90 binding, thereby promoting the binding of RACK1, which triggers ubiquitination and proteasomal degradation of HIF-1α (Liu et al., 2007). Histone deacetylase inhibitors are another novel class of anti-cancer drugs that promote HIF-1α degradation (Kong et al., 2006; Qian et al., 2006) and inhibit HIF-1α transactivation domain function (Fath et al., 2006) at clinically relevant concentrations. Finally, the proteasome inhibitor bortezomib inhibits HIF-1 by blocking its transcriptional activity (Kaluz et al., 2006). Bortezomib may promote binding of the asparagine hydroxylase FIH-1 to HIF-1α (Shin et al., 2008) although this finding has been disputed (Kaluz et al., 2008). In addition to these drugs, many other chemical compounds have been shown to inhibit HIF-1 activity and are in preclinical development (Melillo, 2007).
Cell-based screening assays have been performed to identify inhibitors of HIF-1 transcriptional activity. These assays have revealed that several traditional chemotherapeutic agents are potent HIF-1 inhibitors, including topoisomerase I inhibitors, such as topotecan (Rapisarda et al., 2002) and microtubule targeting drugs, such as taxotere (Escuin et al., 2005). DNA intercalating drugs, including echinomycin (Kong et al., 2005) and anthracyclines, such as doxorubicin (Adriamycin) and daunorubicin (Lee et al., 2009) inhibit HIF-1 transcriptional activity by blocking its binding to DNA. Since extensive clinical experience has already defined the use and limitations of these drugs, one might wonder why this is at all noteworthy. The key point is the use of these drugs as HIF-1 inhibitors engenders an entirely new treatment paradigm. Traditional chemotherapies have been administered to patients at maximum tolerated doses as cytotoxic agents intended to kill as many dividing cancer cells as possible. Unfortunately, these drugs also kill dividing hematopoietic cells and, as a result, they are administered episodically followed by drug-free intervals of 2–3 weeks during which bone marrow recovery occurs. Unfortunately, tumor recovery often occurs as well and eventually leads to the selection of a resistant clone of cancer cells that metastasizes and kills the patient. (Note also that cell proliferation is inhibited under conditions of reduced O2 availability, thereby protecting hypoxic cancer cells from chemotherapy.)
The laboratory of the late Judah Folkman discovered that administering the chemotherapy agent cyclophosphamide at reduced doses on a more frequent and regular dosing interval prevents bone marrow toxicity and blocks tumor growth by interfering with angiogenesis (Hahnfeldt et al., 2003), which was thought to result either from inducing apoptosis or decreasing proliferation of endothelial cells (Kerbel and Kamen, 2004). Treatment with doxorubicin for 5 days inhibited HIF-1-dependent transcription and significantly reduced the levels of mRNAs encoding the angiogenic cytokines stem cell factor (also known as Kit ligand), stromal-derived factor 1 (SDF-1), and vascular endothelial growth factor (VEGF) in human prostate cancer xenografts (Lee et al., 2009). These secreted proteins activate resident endothelial cells and also increase the numbers of circulating angiogenic cells, which home to the tumor and promote vascularization. Treatment of tumor bearing mice with doxorubicin for 5 days reduced blood SDF-1 levels and the number of circulating angiogenic cells to levels similar to those observed in non-tumor-bearing mice and significantly reduced tumor vascularization, leading to growth arrest (Lee et al., 2009).
It has long been known that the hypoxic fraction of human cancers is resistant to radiation therapy due to reduced generation of oxygen radicals (Gray et al., 1953; Moeller et al., 2007). A major paradigm shift occurred with the discovery that the tumor vasculature represents an important target of radiation therapy and a major determinant of the clinical response (Garcia-Barros et al., 2003). Radiation was shown to induce HIF-1 activity, leading to the production of VEGF and other angiogenic cytokines that protect the endothelial cells of the tumor vasculature from radiation-induced death (Moeller et al., 2004). Treatment of tumor-bearing mice with the HIF-1 inhibitor YC-1 (3-(5′-hydroxymethyl-2′furyl)-1-benzyl indazole) dramatically increased radiation-induced vessel destruction and tumor control (Moeller et al., 2004). The HIF-1 inhibitor PX-478 (S-2-amino-3-[4′-N, N,-bis (chloroethyl) amino] phenyl propionic acid N-oxide dihydrochloride) also blocked radiation-induced VEGF expression and thereby sensitized tumor xenografts to radiation therapy (Schwartz et al., 2009). Among patients with oropharyngeal cancer, those whose tumor biopsies showed the highest HIF-1α protein levels by immunohistochemistry had a significantly increased incidence of failure to achieve complete remission after radiation therapy (Aebersold et al., 2001). Taken together, these experimental and clinical data delineate an important role for HIF-1 in mediating radiation resistance and provide a compelling rationale for clinical trials that combine radiation therapy with HIF-1 inhibitors.
It has only been ten years since the first report of HIF-1α overexpression in primary human cancers and their metastases (Zhong et al., 1999), but during that time a large body of scientific data has been amassed delineating the mechanisms and consequences of increased HIF-1 activity during human cancer progression. Most recently, a growing number of drugs that inhibit HIF-1 have been identified and validated as anti-cancer agents. The challenge now is to identify on a patient-by-patient basis: (i) those cancers in which HIF-1 is playing a critical role in disease pathogenesis; (ii) those combinations of drugs or other treatment modalities to which addition of a HIF-1 inhibitor will have additive or even synergistic effects; and (iii) clinical assays that will demonstrate an effect of HIF-1 inhibitors (or lack thereof) on critical aspects of cancer biology such as tumor metabolism and vascularization. Close collaborations between basic scientists and clinical oncologists will be needed to devise, evaluate, and refine strategies for translating research knowledge into safe and effective cancer therapies.
Conflict of Interest
The author declares no conflict of interest.