Review | Published:

Investigating hypoxic tumor physiology through gene expression patterns

Oncogene 22, 5907–5914 (01 September 2003) | Download Citation

Supplemental data is available at: http://cbrl.stanford.edu/hypoxia/welcome.htm

Subjects

Abstract

Clinical evidence shows that tumor hypoxia is an independent prognostic indicator of poor patient outcome. Hypoxic tumors have altered physiologic processes, including increased regions of angiogenesis, increased local invasion, increased distant metastasis and altered apoptotic programs. Since hypoxia is a potent controller of gene expression, identifying hypoxia-regulated genes is a means to investigate the molecular response to hypoxic stress. Traditional experimental approaches have identified physiologic changes in hypoxic cells. Recent studies have identified hypoxia-responsive genes that may define the mechanism(s) underlying these physiologic changes. For example, the regulation of glycolytic genes by hypoxia can explain some characteristics of the Warburg effect. The converse of this logic is also true. By identifying new classes of hypoxia-regulated gene(s), we can infer the physiologic pressures that require the induction of these genes and their protein products. Furthermore, these physiologically driven hypoxic gene expression changes give us insight as to the poor outcome of patients with hypoxic tumors. Approximately 1–1.5% of the genome is transcriptionally responsive to hypoxia. However, there is significant heterogeneity in the transcriptional response to hypoxia between different cell types. Moreover, the coordinated change in the expression of families of genes supports the model of physiologic pressure leading to expression changes. Understanding the evolutionary pressure to develop a ‘hypoxic response’ provides a framework to investigate the biology of the hypoxic tumor microenvironment.

Introduction

Evolution of the transcriptional ‘hypoxic response’

Molecular oxygen is beneficial to many organisms due to its role in efficient energy production. Organisms that rely upon oxygen-dependent energy production suffer when oxygen falls below a certain threshold level. However, too much oxygen can also result in the generation of toxic, damaging radical byproducts. The net effect of these opposing pressures necessitates the evolution of molecular mechanisms to regulate the cellular availability and usage of oxygen.

The response to low-oxygen conditions has been identified in both procaryotes and eucaryotes. Although the mechanisms for gene regulation in response to low-oxygen conditions are diverse, parallel systems have evolved in different organisms that are transcriptionally responsive to low-oxygen. This convergent evolution suggests that the ability to respond to low-oxygen conditions offers a competitive Darwinian advantage. Although the magnitude of the ‘hypoxic’ response may differ between unicellular and multicellular organisms, they both share the principal goal of maintaining the critical energy levels required for homeostasis, while extracellular oxygen concentrations decrease. While this review is focused on the transcriptional changes in the pathophysiologic environment of hypoxic tumor cells, the reason why hypoxic gene regulation evolved is obviously in response to a more physiologic condition of hypoxia such as occurs during embryonic development (Chen et al., 1999a) or during wound healing (Warren et al., 2001).

Mechanisms of gene regulation in response to hypoxia

Hypoxia/anoxia-responsive gene expression has been well characterized in procaryotic organisms such as bacteria as well as in lower and higher eucaryotes such as yeast and mammals, respectively. Unicellular organisms have less control over environmental oxygen, so their hypoxia response is designed to utilize efficiently the decreased levels of oxygen. The facultative anaerobic bacterium Escherichia coli utilizes the fumarate, nitrate reduction (FNR) protein, which is a bifunctional protein that acts both as a hypoxic sensor and a hypoxia-responsive transcription factor. Under aerobic conditions, FNR exists as an inactive apoprotein, but under hypoxia, it forms an active homodimer that is dependent on a redox-sensitive 4Fe–4S cluster (Green et al., 2001). In contrast, one mechanism that yeast Saccharomyces cerevisiae uses for hypoxic gene expression is a heme-regulated repressive system based on the ROX1 (repressed in oxygen) molecule. Under aerobic conditions, abundant heme activates the HAP1 complex that in turn transactivates the ROX1 gene. ROX1 protein blocks expression of a large set of hypoxia-responsive genes (Becerra et al., 2002). Under hypoxic conditions, heme-deficient HAP1 represses ROX1 transcription, and the hypoxia-responsive genes are derepressed (Zitomer et al., 1997). These are two examples of the diverse mechanisms that have evolved in single-cell procaryotes and lower eucaryotes to respond to low-oxygen concentrations.

The mammalian transcriptional response to hypoxia is considerably more complicated, relying on multiprotein complexes to regulate several transcription factors, the most well studied being (hypoxia-inducible factor 1 (HIF-1)). General stress-responsive transcription factors such as AP-1, NF-κB and Egr1 have also been reported to be regulated by hypoxia and/or reoxygenation (Faller, 1999; Ten and Pinsky, 2002). However, the sensitivity of these factors to mild hypoxia, and the duration of their transcriptional response is much less than that of HIF-1. The ‘general stress responsive’ transcription factors are therefore not well suited to regulating gene expression in the chronically hypoxic tumor microenvironment. Furthermore, the number of genes that are upregulated by non-HIF mechanisms in response to chronic hypoxia seems to be small compared to the HIF-responsive genes.

In contrast to gene induction by hypoxia, several mechanisms of gene repression by hypoxia have also been reported. Repression can contribute to the hypoxic response by downregulating such genes as the antiangiogenic thrombospondin (TSP) genes (Tenan et al., 2000). The TSP genes function to block new blood vessel formation, so they are presumably downregulated by hypoxia in order to stimulate angiogenesis (Laderoute et al., 2000). Recent in vitro data have suggested that hypoxic downregulation of gene expression can be through the induction of the (negative cofactor 2 (NC2)) transcriptional repressor (Denko et al., 2003). NC2 has been shown to block gene expression by regulating core promoter action through binding to the TATA-associated factor TBP (Kamada et al., 2001). In addition, the P53 tumor suppressor gene has been found to associate with the transcriptional corepressor mSin3A under hypoxia, and may also be involved in repressing a subset of genes in these conditions (Koumenis et al., 2001). Further study is needed to define all the genes that are repressed in response to hypoxia, and that are regulated through the activity of either P53, NC2, or other mechanisms (Yun et al., 2002).

HIF-1 regulation in mammalian cells

HIF1 was identified through the investigation of the hypoxia-responsive gene erythropoetin (EPO) (Goldberg et al., 1988). EPO is the growth factor responsible for stimulating the bone marrow to produce new red blood cells, especially in times of systemic hypoxia, such as during travel to high altitudes (Caro, 2001). The enhancer element responsible for EPO induction under hypoxia was identified (Imagawa et al., 1991), and the protein that bound to this hypoxia-responsive element was purified, cloned, and found to be a heterodimer (Wang and Semenza, 1993). One subunit of the dimer was constitutively expressed in cells (HIF-1β or ARNT), and the other component was oxygen labile and stabilized in cells exposed to low oxygen (HIF-1α) (Jiang et al., 1996). The molecular mechanism of HIF activation has been recently elucidated, largely because of the functional interaction between HIF-1α and the tumor suppressor protein von Hippel–Lindau (VHL).

Under normoxic conditions, HIF-1α is rapidly degraded, with a half-life of a few minutes. In contrast, under hypoxic conditions, HIF-1α becomes stabilized (Maxwell et al., 1999), associates with the HIF-1β subunit, translocates to the nucleus, and binds in a sequence-specific manner to a hypoxia-responsive element in target genes to activate transcription. The regulation of HIF-1α stability has been shown to rely heavily upon the VHL tumor suppressor protein (Maxwell et al., 1999). Under normoxia, HIF-1α is rapidly hydroxylated at proline 564 (Ivan et al., 2001; Jaakkola et al., 2001) by a novel family of proline hydroxylases (Bruick and McKnight, 2001; Epstein et al., 2001). The hydroxylated form of HIF-1α binds to the VHL protein that is part of the multiprotein complex containing elongins B and C and CUL1 (Hon et al., 2002; Min et al., 2002). This complex acts to ubiquitinate HIF-1α and target it to the proteasome for degradation (Maxwell et al., 1999). Under hypoxia, HIF-1α is not hydroxylated (Chan et al., 2002), does not bind to VHL, and becomes stabilized.

Studies of the VHL tumor suppressor have identified a role for hypoxia-regulated genes in modifying malignant transformation in human tumors (Kondo et al., 2002). VHL has been identified as a classic tumor suppressor gene that requires the loss of both alleles to generate disease (Friedrich, 1999). The von Hippel–Lindau syndrome is a result of the germline loss of one VHL allele, with the loss of the second allele resulting in high-frequency renal cell carcinoma, pheochromocytomas and hemangioblastomas (Friedrich, 1999). In the absence of VHL, HIF-1α exhibits increased stability in normoxic conditions and stimulates constitutive expression of HIF-responsive genes under normoxic conditions (Wykoff et al., 2000). Thus, it is thought that one factor contributing to VHL's role as a tumor suppressor is its ability to downregulate the expression of hypoxia-responsive genes under normoxic conditions (Kondo et al., 2002).

The importance for HIF-1-regulated genes in driving tumor development is supported by studies on murine tumor cell lines that are deficient in either HIF-1α (Ryan et al., 1998) or HIF-1β (Maxwell et al., 1997). Both tumors that are derived from HIF-1α and HIF-1β-deficient cells exhibit reduced aggressiveness in allografted tumor formation in immune-deficient mice (Maxwell et al., 1997; Ryan et al., 1998). The reduced tumor formation of HIF-deficient tumor cells has been attributed to several factors, including decreased levels of vascular endothelial growth factor (VEGF) secretion that leads to decreased levels of angiogenesis in vivo (Grunstein et al., 1999).

Physiologic versus pathophysiologic hypoxia

Mice in which HIF-1α has been deleted fail to develop beyond day 10 (Iyer et al., 1998; Ryan et al., 1998). This observation suggests that HIF-1α or hypoxia-regulated genes are necessary for normal embryonic development. One target hypoxia-responsive gene that has been hypothesized to be in large part responsible for these effects is the vascular endothelial growth factor A (VEGFA) gene. Embryos deficient in a single copy of VEGFA do not develop (Carmeliet et al., 1996; Ferrara et al., 1996). The direct link between hypoxia and VEGF is found when the HRE of VEGF is specifically mutated, and mice also show a partial embryonic lethality at day 10 (Oosthuyse et al., 2001). These animal studies reinforce an essential role for hypoxia-responsive genes during embryonic development. Physiologic hypoxia occurs during development, wound healing, exposure to increased elevation, or other transient vascular alterations (Elson et al., 2000). These examples are conditions in which the hypoxic state elicits a response and that response in turn ‘cures’ the hypoxic insult. Wounding causes vascular damage that leads to the induction of VEGF, the reestablishment of blood flow through new vessels and a return to a normoxic state (Figure 1).

Figure 1
Figure 1

Model showing comparison of physiologic stress response to pathophysiologic stress response

Full size image

In contrast to the physiologic hypoxia described above, tumor hypoxia is an example of a chronic, pathophysiologic condition. The difference lies in the fact that the response is unable to resolve completely the hypoxic insult (Dvorak, 1986) (Figure 1). Despite hypoxia-induced VEGF production in the tumor, the formation of new blood vessels is unable to keep pace with growing tumor cells. The result is a constant production of hypoxia-dependent changes that are never able to establish a uniform normoxic environment. The continuous subversion of the normal hypoxic response is what gives the hypoxic tumor its unique phenotype (Denko and Giaccia, 2001). Understanding the normal hypoxic response can therefore give insight into how it can lead to the pathophysiologic effects.

Tumor hypoxia and clinical correlates

Tumor oxygenation can be directly measured in vivo in patients by microelectrodes or PET imaging through hypoxic marker binding (Hockel and Vaupel, 2001). Numerous clinical studies have demonstrated that the pretreatment oxygenation status of solid tumors can be used to stratify patients prospectively. These studies indicate that mean oxygen status below 10 mmHg in the tumor predicts for poor survival in patients with tumors of the head and neck (Nordsmark et al., 1996), or cervix (Hockel et al., 1996) or in soft tissue (Brizel et al., 1996). The poor prognosis of hypoxic tumors is independent of treatment modality, with patients treated surgically suffering a similarly poor outcome as patients treated with radiotherapy (Hockel et al., 1996). Most importantly, the clinical data suggest that hypoxic tumors represent a biologically more aggressive tumor, in addition to one that is more resistant to oxygen-dependent therapies such as radiation (or some chemotherapies) (Teicher, 1994).

Target genes regulated by hypoxia

What are the specific genes that are responsible for the hypoxic tumor phenotype, and how do they give the hypoxic tumor its increased predisposition to invade, metastasize and apoptose? Several groups have used genomic approaches to identify gene expression profile changes in hypoxia (Denko et al., 2000a; Koong et al., 2000; Wykoff et al., 2000; Scandurro et al., 2001). Thus far, the different cell types examined in vitro, the relative level and duration of hypoxia used, and the resultant gene profiles have underscored the heterogeneity of the induced genes. An analysis of these various studies indicates that there is a core set of genes that are induced consistently by hypoxia and a large number of genes that exhibit cell-type-specific induction. The differential response of various cell types emphasizes the specialized roles for the different cellular components of the solid tumor. Some of the more specific examples of cell-type-dependent hypoxia-responsive genes include the production of tyrosine hydroxylase (TH) by cells of the carotid body (Millhorn et al., 1996), or the production of EPO by cells of the juxtaglomerular apparatus (Fandrey and Bunn, 1993). However, some genes are induced by hypoxia in a large number of diverse cell types, such as VEGF or glucose transporter 3 (GLUT3). Members of the family of glycolytic enzymes are also induced in almost all cell types, even if the specific members may not be induced in all (Seagroves et al., 2001). The cellular need to generate energy seems to be an obvious universal response to hypoxia, while production of TH or EPO is required in a more specialized cellular context.

One example of expression profiling cells treated in vitro with hypoxia is shown in Figure 2. The hypoxic profile of six cell types, representing normal cervical and dermal keratinocytes (NCK, NDK), normal stromal fibroblasts (NCF) and transformed keratinocytes (Siha, C33a, FaDu), is shown. In this cDNA array of 6800 genes, 110 genes were found to be hypoxia-responsive, with 84 being induced, 24 being repressed and three showing a mixed response of induction in some cells with repression in others. Extrapolating from these data, approximately 1.5% of the genome is transcriptionally responsive to hypoxia. This frequency of hypoxia-responsive genes agrees with previous findings using a much smaller array (Koong et al., 2000).

Figure 2
Figure 2

Expression profile changes in response to long-term hypoxia. The subset of genes with robust expression changes was selected as described in the text. Expression ratios were converted into log base 2 values; these are plotted in matrix form, with cell lines listed across the top and gene name and accession number on the side. Boxes are colored to indicate relative expression levels. The red values indicate induction, while green values indicate repression with maximal changes of approximately 10-fold

Full size image

The hypoxia-responsive genes identified in this experiment are ordered in Figure 2 by both cell type and induction pattern for display purposes. A red signal represents induction, and green represents repression, with maximal changes of approximately 10-fold. Using this set of genes, cell types were ordered for the similarity of hypoxic response by principle component analysis (Raychaudhuri et al., 2000). PCA orders the cell lines as depicted in Figure 2 in the order of similarity and shows that the normal keratinocytes are most closely related, the normal fibroblasts more distantly related and the tumor lines even more distant from the normal keratinocytes. These two axes, the similarity of cell type in the x-dimension, and the level of induction in the y-dimension describe the representation of the ‘hypoxiatome,’ as depicted in Figure 2. Figure 3 shows a panel of selected genes used as probes for Northern blot comparison of techniques.

Figure 3
Figure 3

Northern analysis of hypoxic induction in several test genes. The same series of cells lines were treated with hypoxia in vitro, and Northern blot was used to determine hypoxic gene regulation. Representative genes are shown for several of the functional classes (GAPDH, glyceraldehye 3-phosphate dehydrogenase; G6PI, glucose 6-phosphate isomerase; GBE, glycogen branching enzyme; VEGF, vascular endothelial growth factor; PGF, placental growth factor; EphA1, ephrin A1; PLOD2, lysyl hydroxylase 2; PAI-1, plasminogen activator inhibitor 1; TF, tissue factor; IGFBP3, insulin-like growth factor binding protein 3; BNIP3, 19 K interacting protein; DEC1, differentiated in embryonic chondrocytes; BTG1, B-cell translocation breakpoint gene)

Full size image

Interestingly, the differences in hypoxia-responsive profiles cannot be simply explained by the fact that the cells were genetically unstable tumor cells. The normal stromal and normal epithelial cells also show differences in their hypoxic profiles. These data underscore the importance of studying hypoxic gene expression in a cell-type-dependent context. From this sample, we can see that some genes are widely inducible, such as NIP3L, IGFBP3 and Dec1, but many genes are inducible only in certain cellular contexts, such as placental growth factor or ephrin A1. Identifying the additional factors regulating hypoxic gene responsiveness will continue to be a major point of investigation.

Coordinated regulation suggests functional categories of response

Based on the published function(s) of the various hypoxia-responsive genes, we grouped the most robustly induced genes into functional categories. Table 1 is a compilation of data from our lab as well as data from the literature. The six largest functional groups are shown and represent genes involved in metabolism/transport, angiogenesis, tissue remodeling, apoptosis, proliferation/differentitation and gene expression. In addition to the five major categories, there are many interesting genes that fall into smaller functional categories such as mRNA processing (RNAse L, CDC-like kinase 1, CCR4-associated protein), or genes with unknown functions (HIG2, RTP, RAIG3, TGFBind, VitDind68). Some genes, such as PAI-1, can be placed in multiple categories of tissue remodeling and angiogenesis.

Table 1: Categorization of selected hypoxia-regulated genes by function
Full size table

Hypoxia and metastasis

The physiological demands of hypoxia and the response to those demands in the various cell types result in different gene expression patterns. For example, lysyl hydroxylase (PLOD2, Figure 2 and Figure 3) is normally required for collagen maturation and is therefore normally only expressed in fibroblasts (Wang et al., 2000). However, under hypoxia, both primary and transformed keratinocytes induce high levels of PLOD2 expression, while fibroblasts exhibit little change in their PLOD2 mRNA levels. The same environmental signal elicits two different responses in the two cell types. While the reason for PLOD2 upregulation in hypoxic epithelial cells is unclear, we hypothesize that it is related to the wounding/re-epithelialization response of the keratinocytes. The intimate interaction between stromal and epithelial cells in the hypoxic solid tumor combines to regulate PLOD2 expression and, presumably, collagen maturation. The alterations in the extracellular matrix of hypoxic tumors could in turn contribute to hypoxia-induced metastasis (Denko & Giaccia, 2001; De Jaeger et al., 2001).

Furthermore, the hypoxic profiles of the untransformed cells are more closely related to each other than to the tumor cells. For example, while tumor cells exhibit a significant downregulation in the basal level of PAI-1 expression (Figure 2), hypoxic upregulation is found both in the untransformed cells and in the tumor cells (Figure 2 and Figure 3). Likewise, hypoxia can upregulate another member of the plasmin pathway, urokinase-type plasminogen activator receptor, which has also been implicated in metastasis (Postovit et al., 2002). The induced response, irrespective of the transformed phenotype, suggests that the physiological demands upon the cells have a dominant effect on their expression patterns. Thus, one can see how alteration of the extracellular protease plasmin by PAI-1 by both hypoxia and transformation could also combine to contribute to hypoxia-induced metastasis (Denko and Giaccia, 2001; De Jaeger et al., 2001).

Apoptotic response to hypoxia

P53 independent, hypoxia-induced apoptosis is thought to rely upon the hypoxic upregulation of certain proapoptotic genes, such as BH3-containing BNIP3 and BNIP3L (Bruick, 2000). While the kinetics of BNIP3 induction correlate with apoptosis, there are several characteristics of its expression pattern that still need to be reconciled. For example, VHL-negative cells are able to withstand constitutive expression of these supposedly apoptogenic molecules (Sowter et al., 2001), and normal fibroblasts can induce high levels of BNIP3 expression in the absence of apoptosis (Figure 2 and Figure 3 and data not shown). Normoxic expression of BNIP3/BNIP3L in the VHL-negative tumor cells could lead to the selection of cells with a second site mutation somewhere downstream in the apoptotic signaling pathway (Graeber et al., 1996).

Alternatively, it is possible that BNIP3/BNIP3L have different functions under hypoxia from those reported in transfection studies under normoxia (Chen et al., 1999b). We could therefore gain new insight into the function(s) of BNIP3/BNIP3L by this supposed discrepancy between hypoxic induction and apoptosis. While BNIP3/BNIP3L induced by hypoxia could still be targeted to the mitochondria just as the normoxic experiments report (Chen et al., 1999b), their biochemical roles under hypoxia may be different. It is possible that the mitochondrion has physiological pressures under hypoxia that require BNIP3/BNIP3L expression, and it is only under normoxic conditions that its overexpression is apoptogenic.

Additional levels of regulation

Finally, there are genes that are induced by hypoxia in some cell types, while they are repressed by hypoxia in other cell types with unknown factors controlling this regulation, such as ornithine decarboxylase (ODC) and glucose-regulated protein 78 (GRP78). These examples make it clear that the definition of a hypoxia-responsive gene is cell-type dependent. There are also multiple levels of gene regulation within the same family. For example, lactate dehydrogenase isoform A (LDHA) is strongly induced in all cell lines, while LDH B is repressed in all the cell lines tested (Figure 2). Thus, we have examples of added complexity to hypoxic gene expression, multiple levels of control of hypoxic response between cells of different origins and multiple levels of control between members of a single gene family.

Conclusions

We can infer that physiologic changes in the hypoxic tumor drive gene expression changes, and these in turn result in the different cellular effects within that tumor. For example, low oxygen stimulates angiogenesis, and the establishment of new vessels requires tissue remodeling. The genes involved in tissue remodeling may then be part of the explanation as to why hypoxic tumors are more likely to be locally invasive or distantly metastatic. The expression changes in the lysine hydroxylase gene PLOD2 in keratinocytes and tumor cells is an example of how gene expression profiling can allow us to infer new insights about tumor physiology in hypoxic tumors. Keratinocytes are able to play an active role in collagen maturation in the hypoxic solid tumor instead of the fibroblast that is traditionally thought to serve this function. The interaction between hypoxia, fibroblasts and tumor cells in tissue remodeling represents a cell-context-dependent regulation of gene expression to result in a coordinated tissue response. Understanding that tumor cells are an active participant in this process may influence the development of targeted antimetastatic chemotherapeutics in the future.

Furthermore, we may learn how the biochemical characterization of hypoxic gene products may also be influenced by the cellular physiology. For example, seeing that the potentially apoptogenic genes BNIP3 and BNIP3L are robustly induced in all hypoxic cells regardless of their apoptotic potential makes one question their role in hypoxia-induced apoptosis. The targeting of BH3-only molecules to the mitochondria may represent a survival response to impaired mitochondrial function under hypoxic conditions. The function of the BH3 proteins may be very different in this environmental context than under control conditions. We therefore need to design our functional assays to take into account the conditions in which the gene products are expressed.

References

  1. , , , , and . (2002). Mol. Microbiol., 43, 545–555.

  2. , , , , and . (1999). Genes Dev., 13, 64–75.

  3. , , , , , and . (1996). Cancer Res., 56, 941–943.

  4. . (2000). Proc. Natl. Acad. Sci. USA, 97, 9082–9087.

  5. and . (2001). Science, 11, 11.

  6. , , , , , , , , , , , , , , and . (1996). Nature, 380, 435–439.

  7. . (2001). High Alt. Med. Biol., 2, 145–154.

  8. , , and . (2002). J. Biol. Chem., 277, 40112–40117.

  9. , and . (1999a). Teratology, 60, 215–225.

  10. , , , , , , , and . (1999b). J. Biol. Chem., 274, 7–10.

  11. , and . (2001). Br. J. Cancer, 84, 1280–1285.

  12. , , , , and . (2000a). Clin. Cancer Res., 6, 480–487.

  13. , , , , and . (2003). J. Biol. Chem., 278, 5744–5749.

  14. and . (2001). Cancer Res., 61, 795–798.

  15. , , and . (2000b). Exp. Cell Res., 258, 82–91.

  16. . (1986). N. Engl. J. Med., 315, 1650–1659.

  17. , and . (1995). J. Biol. Chem., 270, 29083–29089.

  18. , , , , and . (1996). Biochem. J., 313 (Part 3), 809–814.

  19. , , , and . (2000). Cancer Res., 60, 6189–6195.

  20. , , , , , , , , , , and . (1997). Oncogene, 14, 2475–2483.

  21. , , , , , , , , , , , , , , , , , , and . (2001). Cell, 107, 43–54.

  22. . (1999). Clin. Exp. Pharmacol. Physiol., 26, 74–84.

  23. and . (1993). Blood, 81, 617–623.

  24. , , , , , , , and . (1996). Nature, 380, 439–442.

  25. . (1999). Cancer, 86, 2478–2482.

  26. , , and . (1995). Am. J. Physiol., 268, C1362–C1368.

  27. , and . (1988). Science, 242, 1412–1415.

  28. , , , , , and . (1996). Nature, 379, 88–91.

  29. , and . (1998). Blood, 91, 3300–3307.

  30. , and . (2001). Adv. Microb. Physiol., 44, 1–34.

  31. , , , and . (1999). Cancer Res., 59, 1592–1598.

  32. , , , , and . (1996). Cancer Res., 56, 4509–4515.

  33. and . (2001). J. Natl. Cancer Inst., 93, 266–276.

  34. , , , , , , , , and . (2002). Nature, 417, 975–978.

  35. , , and . (1991). Blood, 77, 278–285.

  36. , , , , , , , , and . (2001). Science, 292, 464–468.

  37. , , , , , , , , , , , , , and . (2001). Am. J. Pathol., 158, 905–919.

  38. , , , , , , , , , and . (1998). Genes Dev., 12, 149–162.

  39. , , , , , , , , , , , and . (2001). Science, 292, 468–472.

  40. , , , and . (1996). J. Biol. Chem., 271, 17771–17778.

  41. , , , , , , and . (2001). Cell, 106, 71–81.

  42. , , , and . (2002). Cancer Cell, 1, 237–246.

  43. , , , , , , , , , and . (2000). Cancer Res., 60, 883–887.

  44. , , , , , , , , , and . (2001). Mol. Cell Biol., 21, 1297–1310.

  45. , , and . (1993). J. Clin. Invest., 92, 99–104.

  46. , , , , , , , and . (2000). Clin. Cancer Res., 6, 2941–2950.

  47. and . (1998). Circ. Res., 83, 852–859.

  48. , , , , , , , and . (1997). Proc. Natl. Acad. Sci. USA, 94, 8104–8109.

  49. , , , , , , , , and . (1999). Nature, 399, 271–275.

  50. , , , , and . (1995). J. Exp. Med., 182, 1683–1693.

  51. , , , , , , , and . (1996). Adv. Exp. Med. Biol., 410, 135–142.

  52. , , , , and . (2002). Science, 296, 1886–1889.

  53. , , , and . (1998). Biochem. Biophys. Res. Commun., 243, 514–517.

  54. , , , , , , and . (1999). Brain. Res. Mol. Brain Res., 74, 26–34.

  55. , and . (1996). Radiother. Oncol., 41, 31–39.

  56. , , , , and . (2001). Biochem. Biophys. Res. Commun., 289, 39–43.

  57. , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , and . (2001). Nat. Genet., 28, 131–138.

  58. , , and . (1996). Eur. J. Biochem., 241, 403–410.

  59. , , , , , and . (2001). Cancer Res., 61, 4206–4213.

  60. , , , , , , and . (1998). J. Clin. Invest., 102, 919–928.

  61. , , , and . (1993). Cancer. Res., 53, 5822–5827.

  62. , , , and . (2002). J. Biol. Chem., 277, 35730–35737.

  63. , and . (2000). Pac. Symp. Biocomput., 455–466.

  64. , , , and . (1991). Mol. Cell Biochem., 103, 141–148.

  65. , and . (1998). EMBO. J., 17, 3005–3015.

  66. , and . (2002). FEBS Lett., 530, 186–190.

  67. , and . (1996). Arch. Biochem. Biophys., 334, 389–394.

  68. , , , and . (2001). Int. J. Oncol., 19, 129–135.

  69. , , , and . (1997). J. Biol. Chem., 272, 601–608.

  70. , , , , , , and . (2001). Mol. Cell. Biol., 21, 3436–3444.

  71. , , and . (1994). J. Biol. Chem., 269, 23757–23763.

  72. , , and . (1992). Nature, 359, 843–845.

  73. , , , and . (2001). Cancer Res., 61, 6669–6673.

  74. , , and . (1998). Invest. Ophthalmol. Vis. Sci., 39, 1028–1035.

  75. , , , , , , , , and . (1998). Proc. Natl. Acad. Sci. USA, 95, 10188–10193.

  76. . (1994). Cancer Metast. Rev., 13, 139–168.

  77. and . (2002). Curr. Opin. Crit. Care, 8, 242–250.

  78. , , , , , , , and . (2000). J. Exp. Med., 191, 1789–1798.

  79. , and . (2000). DNA Cell Biol., 19, 71–77.

  80. and . (1993). Proc. Natl. Acad. Sci USA, 90, 4304–4308.

  81. , , , , , , and . (2001). J. Surg. Res., 99, 147–155.

  82. , , , and . (2000). Oncogene, 19, 6297–6305.

  83. , , , , , , , , , and . (1999). J. Biol. Chem., 274, 15030–15040.

  84. , , , , , , , , and . (1998). Proc. Natl. Acad. Sci. USA, 95, 8298–8303.

  85. , , and . (2002). Dev. Cell, 2, 331–341.

  86. , and . (1997). Kidney Int., 51, 507–513.

Download references

Acknowledgements

This work was supported in part by Varian Biosynergy and NIH Grant CA67166. LF was supported by a predoctoral fellowship from Fondazione Italiana per la Ricerca sul Cancero.

Author information

Affiliations

  1. Division of Radiation and Cancer Biology, Department of Radiation Oncology, Stanford University School of Medicine, Stanford, CA 94305, USA

    • Nicholas C Denko
    • , Lucrezia A Fontana
    • , Karen M Hudson
    • , Patrick D Sutphin
    •  & Amato J Giaccia

  2. Department of Genetics, Stanford University School of Medicine, Stanford, CA 94305, USA

    • Soumya Raychaudhuri
    •  & Russ Altman

Authors

  1. Search for Nicholas C Denko in:

  2. Search for Lucrezia A Fontana in:

  3. Search for Karen M Hudson in:

  4. Search for Patrick D Sutphin in:

  5. Search for Soumya Raychaudhuri in:

  6. Search for Russ Altman in:

  7. Search for Amato J Giaccia in:

Corresponding author

Correspondence to Nicholas C Denko.

About this article

Publication history

Published

DOI

https://doi.org/10.1038/sj.onc.1206703

Rights and permissions

To obtain permission to re-use content from this article visit RightsLink.