Critical Reviews in Biochemistry and Molecular Biology
`
`ISSN: 1040-9238 (Print) 1549-7798 (Online) Journal homepage: http://www.tandfonline.com/loi/ibmg20
`
`Hypoxia, Clonal Selection, and the Role of HIF-1 in
`Tumor Progression
`
`Gregg L. Semenza
`
`To cite this article: Gregg L. Semenza (2000) Hypoxia, Clonal Selection, and the Role of HIF-1 in
`Tumor Progression, Critical Reviews in Biochemistry and Molecular Biology, 35:2, 71-103
`
`To link to this article: http://dx.doi.org/10.1080/10409230091169186
`
`Published online: 29 Sep 2008.
`
`Submit your article to this journal
`
`Article views: 795
`
`View related articles
`
`Citing articles: 5 View citing articles
`
`Full Terms & Conditions of access and use can be found at
`http://www.tandfonline.com/action/journalInformation?journalCode=ibmg20
`
`Roxane Labs., Inc.
`Exhibit 1017
`Page 001
`
`

`
`Critical Reviews in Biochemistry and Molecular Biology, 35(2):71–103 (2000)
`
`Hypoxia, Clonal Selection, and the Role of
`HIF-1 in Tumor Progression
`
`Gregg L. Semenza
`
`Institute of Genetic Medicine, Departments of Pediatrics and Medicine, The Johns
`Hopkins University School of Medicine, Baltimore, Maryland 21287
`
`* Address for correspondence: Gregg L. Semenza, M.D., Ph.D. Johns Hopkins Hospital,
`CMSC-1004, 600 North Wolfe Street, Baltimore, MD 21287–3914, TEL: 410-955-
`1619 FAX: 410-955-0484, E-mail: gsemenza@jhmi.edu
`
`Referee: Peter M. Glazer, M.D., Ph. D., Yale University School of Medicine
`
`Abstract: Tumor progression occurs as a result of the clonal selection of cells in which
`somatic mutations have activated oncogenes or inactivated tumor suppressor genes leading
`to increased proliferation and/or survival within the hypoxic tumor microenvironment.
`Hypoxia-inducible factor 1 (HIF-1) is a transcription factor that mediates adaptive re-
`sponses to reduced O2 availability, including angiogenesis and glycolysis. Expression of the
`O2-regulated HIF-1α subunit and HIF-1 transcriptional activity are increased dramatically
`in hypoxic cells. Recent studies indicate that many common tumor-specific genetic alter-
`ations also lead to increased HIF-1α expression and/or activity. Thus, genetic and physi-
`ologic alterations within tumors act synergistically to increase HIF-1 transcriptional activ-
`ity, which appears to play a critical role in the development of invasive and metastatic
`properties that define the lethal cancer phenotype.
`
`Key Words: angiogenesis, cancer, glycolysis p53; Ras, vascular endothelial growth factor.
`
`I. INTRODUCTION: GENETIC
`ALTERATIONS IN TUMOR CELLS
`
`A major focus of oncology research over
`the last quarter-century spanning the era of
`molecular biology has been the identifica-
`tion of genetic alterations arising as a result
`of somatic mutation within tumor cells. A
`large body of research data established the
`existence of several groups of genes that,
`when mutated, contribute to the process of
`tumor progression (for review see Vogelstein
`
`1040-9238/00/$.50
`© 2000 by CRC Press LLC
`
`and Kinzler, 1998). The first group that was
`identified consists of oncogenes, in which
`gain-of-function mutations result in in-
`creased expression of a gene product and/or
`expression of a mutant product with
`constitutive activity that is not responsive
`to normal molecular mechanisms of
` downregulation. The second group consists
`of tumor suppressor genes, in which loss-
`of-function mutations eliminate the expres-
`sion or activity of a gene product. A third
`group consists of mutator genes involved in
`DNA repair, in which loss-of-function mu-
`
`71
`
`
`
`Roxane Labs., Inc.
`Exhibit 1017
`Page 002
`
`

`
`tations increase the rate at which mutations
`occur within oncogenes and tumor suppres-
`sor genes.
`The high frequency of mutations involv-
`ing specific genes within specific tumor cell
`types implies that these genetic alterations
`play a key role in the oncogenic transforma-
`tion of that cell type. Tumor-specific muta-
`tions do not result from targeted mutagen-
`esis, but instead represent the effect of
`selection: cells in which a mutation occurs
`by chance within a specific gene that results
`in an increased rate of survival and/or divi-
`sion (that together determine the rate of cell
`proliferation) become overrepresented in the
`tumor relative to cells lacking the mutation.
`Along with the discovery of oncogenes,
`tumor suppressor genes, and mutator genes,
`the concept of clonal selection (Nowell,
`1976) represented a critical advance in our
`understanding of tumor biology.
`Because it is much easier to manipulate
`tumor cells in plastic dishes than in living
`animals, a great deal of research investigat-
`ing the biological function of oncogenes
`and tumor suppressor genes has been per-
`formed in cultured cells. In this context,
`studying the rate at which cells divide is
`accomplished easily. For this reason, the
`functions attributed to the protein products
`of oncogenes and tumor suppressor genes
`have primarily related to the control of
`cellular proliferation in tissue culture:
`oncogenes promote cell proliferation
`whereas tumor suppressor genes counteract
`this effect. Likewise, the influence of anti-
`and proapoptotic proteins, such as members
`of the BCL2 family, on tumor cell survival
`has been analyzed primarily in the tissue
`culture milieu.
`The goal of this review is to focus on
`adaptive mechanisms by which selected
`tumor cells are able to survive the harsh
`microenvironmental conditions that they
`impose on themselves in vivo. The adapta-
`tions that are described have in common the
`
`72
`
`fact that they are mediated, at least in part,
`by the transcriptional activator hypoxia-in-
`ducible factor 1 (HIF-1). Important proper-
`ties of HIF-1 and common characteristics of
`solid tumors are reviewed first, followed by
`a summary of recent data that have delin-
`eated specific functional relationships be-
`tween genetic alterations, HIF-1, and tumor
`progression. Finally, the clinical implica-
`tions of these results are considered.
`
`II. HIF-1 STRUCTURE AND
`FUNCTION
`
`HIF-1 is a heterodimer composed of HIF-
`1α and HIF-1β subunits (Wang and Semenza,
`1995). Both subunits contain basic-helix-
`loop-helix (bHLH)-PAS domains (Wang et
`al., 1995a). Whereas the bHLH domain de-
`fines a superfamily of eukaryotic transcrip-
`tion factors (reviewed by Semenza, 1998),
`the PAS domain, which was first identified
`in the PER, ARNT, and SIM proteins, de-
`fines a subfamily of bHLH proteins that is
`unique to metazoans (reviewed by Crews
`and Fan, 1999). The HIF-1β subunit is also
`known as the aryl hydrocarbon nuclear
`translocator (ARNT), as it was first shown to
`dimerize with the aryl hydrocarbon receptor
`(Hoffman et al., 1991). Recently, proteins
`with sequence similarity to HIF-1α (HIF-2α
`and HIF-3α) and HIF-1β (ARNT2 and
`ARNT3) have been identified, and all HIF-
`1α-like polypeptides appear able to dimerize
`with all HIF-1β-like polypeptides, at least in
`vitro (reviewed by Semenza, 2000). In knock-
`out mice, homozygosity for a targeted loss-
`o
`f
`-
`f
`u
`n
`c
`t
`i
`o
`n
`mutation in the Hif1a, Epas1, or Arnt gene
`encoding HIF-1α, HIF-2α, or HIF-1β, re-
`spectively, results in embryonic lethality (Iyer
`et al., 1998; Kozak et al., 1998; Maltepe et
`al., 1997; Tian et al., 1998). The develop-
`mental and physiological roles of these fac-
`
`
`
`Roxane Labs., Inc.
`Exhibit 1017
`Page 003
`
`

`
`tors have been reviewed elsewhere recently
`(Crews and Fan, 1999; Semenza, 1999, 2000).
`The HLH and PAS domains of HIF-1α
`and HIF-1β are required for the formation
`of a dimer in which conformation of the
`basic domains allows their recognition of
`specific DNA binding sites containing the
`core sequence 5’-RCGTG-3’ (Jiang et al.,
`1996a; Semenza et al., 1996; Wood et al.,
`1996). HIF-1 functions as a sequence-spe-
`cific transcriptional activator (Forsythe
`et al., 1996; Jiang et al., 1996a; Semenza et
`al. 1996). Both the expression of HIF-1α
`and its ability to activate transcription
`are regulated by the cellular O2 concentra-
`tion (Huang et al., 1996; Jiang et al., 1996b,
`1997b; Pugh et al., 1997; Wang et al.,
`1995a). HIF-1α has a half-life of < 5
`min under posthypoxic conditions, both in
`cultured cells and in vivo (Wang et al., 1995a;
`Yu et al., 1998). Under nonhypoxic condi-
`tions, HIF-1α is subject to ubiquitination
`and proteasomal degradation (Huang et al.,
`1998; Kallio et al., 1999; Salceda and
`Caro, 1997). Under hypoxic conditions,
`ubiquitination of HIF-1α is dramatically re-
`duced (Sutter et al., 2000). The introduction
`of specific missense mutations and dele-
`tions into the HIF-1a sequence results in
`decreased ubiquitination and increased
`expression
`of HIF-1a
`protein
`under nonhypoxic conditions (Sutter et al.,
`2000). Transactivation domain function is
`also under negative regulation in nonhypoxic
`cells and deletions within these domains
`also increase their activity in nonhypoxic
`cells (Jiang et al., 1997b; Pugh et al., 1997).
`Thus, under hypoxic conditions, HIF-1 tran-
`scriptional activity increases rapidly due to
`synergistic effects on HIF-1a protein ex-
`pression and transactivation domain func-
`tion. The HIF-1 transactivation domains
`have been shown to interact with the
`coactivator proteins CBP, p300, SRC-1, and
`TIF2, and these interactions are promoted
`by the redox regulatory proteins thioredoxin
`
`and Ref-1 (Arany et al., 1997; Carrero et al.,
`2000; Ema et al., 1999).
`To date, approximately 30 target genes
`that are transactivated by HIF-1 have been
`identified (Table 1 and data not shown).
`These genes encode proteins that are re-
`quired for angiogenesis, regulation of blood
`vessel tone, and vascular remodeling; cell
`proliferation and viability; erythropoiesis and
`iron metabolism; glucose transport and gly-
`colysis. Remarkably, the vast majority of
`these gene products have been shown to be
`overexpressed in human tumor cells. Each
`HIF-1 target gene contains a hypoxia re-
`sponse element, a cis-acting transcriptional
`regulatory sequence. The presence of a HIF-
`1 binding site is necessary but not sufficient
`to constitute a functional hypoxia response
`element (Forsythe et al., 1996; Semenza et
`al., 1996; Semenza and Wang, 1992).
`
`III. COMMON BIOCHEMICAL
`AND PHYSIOLOGICAL
`CHARACTERISTICS OF SOLID
`TUMORS
`
`A. The Warburg Effect
`
`Seventy years ago, Warburg demon-
`strated that compared with normal cells, tu-
`mor cells are characterized by a marked
`increase in glycolytic metabolism, even
`when cultured in the presence of high O2
`concentrations (Warburg, 1930). Glucose
`transport into tumor cells is also markedly
`increased in order to provide increased
`amounts of the substrate that is ultimately
`converted to lactate by the glycolytic en-
`zymes. Indeed, increased uptake of labeled
`glucose derivatives is utilized in clinical
`diagnostic tests to identify occult tumors in
`patients. Furthermore, glycolytic metabo-
`lism is correlated with disease progression,
`
`73
`
`
`
`Roxane Labs., Inc.
`Exhibit 1017
`Page 004
`
`

`
`Table 1.
`Known HIF-1 Target Genes
`Gene product
`Adenylate kinase 3
`α1B-Adrenergic receptor
`Adrenomedullin
`Aldolase A
`
`Aldolase C
`Endothelin-1 (ET-1)
`Enolase 1 (ENO1)
`Erythropoietin (EPO)
`Glucose transporter 1
`
`Glucose transporter 3
`Glyceraldehyde phosphate dehydrogenase
`Heme oxygenase 1
`Hexokinase 1
`Hexokinase 2
`Insulin-like growth factor 2 (IGF-2)
`IGF binding protein 1
`IGF factor binding protein 3
`Lactate dehydrogenase A
`
`Nitric oxide synthase 2 (NOS2)
`p21
`p35srj
`Phosphofructokinase L
`Phosphoglycerate kinase 1
`
`Plasminogen activator inhibitor-1
`Pyruvate kinase M
`Transferrin
`Transferrin receptor
`Vascular endothelial growth factor (VEGF)
`
`VEGF receptor FLT-1
`
`74
`
`Ref.
`
`Wood et al., 1998
`Eckhart et al., 1997
`Cormier-Regard et al., 1998
`Iyer et al., 1998; Ryan et al., 1998;
`Semenza et al.,1996
`Iyer et al., 1998
`Hu et al., 1998
`Semenza et al., 1996; Iyer et al., 1998
`Semenza and Wang, 1992; Jiang et al.,1996a
`Iyer et al., 1998; Ryan et al., 1998;
`Wood et al., 1998
`Iyer et al., 1998
`Iyer et al., 1998; Ryan et al., 1998
`Lee et al., 1997
`Iyer et al., 1998
`Iyer et al., 1998
`Feldser et al., 1999
`Tazuke et al., 1998
`Feldser et al., 1999
`Iyer et al., 1998; Ryan et al., 1998;
`Semenza et al., 1996
`Melillo et al., 1995; Palmer et al., 1998
`Carmeliet et al., 1998
`Bhattacharya et al., 1999
`Iyer et al., 1998
`Carmeliet et al., 1998; Iyer et al., 1998; Ryan et al.,
`1998
`Kietzmann et al., 1999
`Iyer et al., 1998
`Rolfs et al., 1997
`Lok and Ponka, 1999; Tacchini et al., 1999
`Carmeliet et al., 1998; Forsythe et al., 1996;
`Iyer et al., 1998; Ryan et al., 1998
`Gerber et al., 1997
`
`
`
`Roxane Labs., Inc.
`Exhibit 1017
`Page 005
`
`

`
`because lactate concentrations of > 20 µmol/
`g are 6.5-fold more common in primary
`cervical cancers with metastatic spread com-
`pared with non-metastasizing primary tu-
`mors (Schwickert et al., 1995). A correla-
`tion between lactate production and
`metastasis was also reported for head and
`neck tumors (Walenta et al., 1997). In addi-
`tion to aerobic glycolysis (the Warburg ef-
`fect), the existence of intratumoral hypoxia
`provides an additional stimulus for glycoly-
`sis, as described below.
`Even though the Warburg effect is one
`of the most universal characteristics of solid
`tumors, only recently has insight been gained
`into the molecular mechanisms by which
`energy metabolism is shifted from oxida-
`tive to glycolytic pathways. Quiescent thy-
`mocytes cultured in the absence of mito-
`gens have been shown to derive > 80% of
`their ATP from oxidative phosphorylation,
`whereas mitogen-stimulated thymocytes
`derive > 80% of their ATP from glycolysis
`(Brand and Hermfisse, 1997). It has been
`proposed that the switch from oxidative to
`glycolytic metabolism occurs in order to
`reduce the generation of reactive oxygen
`species that might otherwise damage repli-
`cating DNA. If this switch is hard-wired
`into the program of cellular proliferation,
`then it is not surprising that virtually all
`tumor cells would manifest the Warburg
`effect despite their thorough disregard for
`genomic integrity.
`Hypoxia response elements have been
`identified in the promoters of genes encod-
`ing aldolase A (ALDA), enolase 1 (ENO1),
`glucose transporter 3 (GLUT3), lactate de-
`hydrogenase A (LDHA), phosphoglycerate
`kinase 1, and phosphofructokinase L (Ebert
`et al., 1995; Firth et al., 1994, 1995; Semenza
`et al., 1994, 1996) and HIF-1-dependent
`transactivation of the ALDA and ENO1 pro-
`moters has been demonstrated (Semenza et
`al., 1996). A 68-bp fragment of the ENO1
`promoter, which is the most powerful hy-
`
`poxia response element identified to date,
`contains three HIF-1 binding sites (Semenza
`et al., 1996). Analysis of mouse embryonic
`stem cells that are homozygous for a tar-
`geted loss-of-function mutation in the Hif1a
`gene encoding HIF-1α revealed decreased
`expression of 13 different genes encoding
`glucose transporters and glycolytic enzymes
`(Iyer et al., 1998; Ryan et al., 1998). Thus,
`HIF-1 mediates coordinate transcriptional
`activation of the entire pathway from glu-
`cose uptake to lactate production. Further-
`more, expression of HIF-1α is induced when
`tumor cells are cultured under conditions
`associated with cellular proliferation such
`as low cell density and growth factor stimu-
`lation (Feldser et al., 1999; Zhong et al.,
`1998; Zhong et al., 2000). C-MYC, a tran-
`scription factor whose primary role is to
`stimulate cellular proliferation, also
`transactivates several genes encoding gly-
`colytic enzymes, including LDHA (Dang
`and Semenza, 1999; Shim et al., 1997). HIF-
`1 (consensus binding site sequence, 5’-
`RCGTG-3’) and C-MYC (consensus
`binding site sequence, 5’-CACGTG-3’) rec-
`ognize overlapping binding sites in the
`LDHA gene (Dang and Semenza, 1999;
`Semenza et al., 1996; Shim et al., 1997).
`
`B. Angiogenesis
`
`A large body of experimental evidence
`indicates that primary tumors and metastases
`will not grow beyond a volume of several
`mm3 without establishing a blood supply
`(reviewed in Fidler and Ellis, 1994;
`Folkman, 1971, 1990, 1992; Hanahan and
`Folkman, 1996; Zetter, 1998). In the ab-
`sence of vascularization, cell viability is lim-
`ited by the diffusion of O2 and glucose from
`host vessels. Under these circumstances, cell
`division and cell death occur at equal rates
`and no net tumor growth occurs. In
`
`75
`
`
`
`Roxane Labs., Inc.
`Exhibit 1017
`Page 006
`
`

`
`RIP1Tag2 transgenic mice, every pancre-
`atic-islet β cell expresses SV40 T antigen,
`approximately 50% of the islets demonstrate
`β cell hyperplasia, but only a
`subset of hyperplastic islets demonstrate
`neovascularization; an even smaller subset
`of islets progress to carcinoma, but, nota-
`bly, all of these manifest neovascularization
`(Folkman et al., 1989). The onset of angio-
`genesis permits rapid tumor growth, thus
`setting the stage for subsequent develop-
`ment of the invasive and metastatic proper-
`ties that define the lethal cancer phenotype.
`Indeed, for many tumor types there is a
`statistically significant correlation between
`tumor angiogenesis (as measured by blood
`vessel density) and patient survival (re-
`viewed in Fox, 1997; Jekunen and Kairemo,
`1997; Zetter, 1998).
`Tumor angiogenesis occurs as a result
`of increased expression of angiogenic
`factors and decreased expression of
`antiangiogenic factors. The angiogenic fac-
`tor that has been shown to have the most
`important role in mediating blood vessel
`formation in response to developmental,
`physiological, or oncogenic stimuli is vas-
`cular endothelial growth factor (VEGF) (re-
`viewed in Ferrara and Davis-Smyth, 1997).
`There is a strong correlation between VEGF
`expression and blood vessel density and
`clinical outcome in many tumor types
`(Ferrara and Davis-Smyth, 1997). Inhibi-
`tion of VEGF expression or binding to its
`receptor on endothelial cells has a dramatic
`effect on tumor growth, invasion, and
`metastasis in animal models (Benjamin et
`al., 1999; Benjamin and Keshet, 1997;
`Borgstrom et al., 1996; Cheng et al., 1996;
`Grunstein et al., 1999; Im et al., 1999; Jain
`et al., 1998; Kim et al., 1993; Millauer et al.,
`1994, 1996; Saleh et al., 1996; Shaheen et
`al., 1999; Skobe et al., 1997; Yuan et al.,
`1996). Somatic mutations resulting in
`oncogene activation (e.g., H-Ras and v-Src)
`and tumor suppressor gene inactivation (e.g.,
`
`76
`
`p53 and VHL) are associated with increased
`VEGF gene expression (reviewed in Ferrara
`and Davis-Smyth, 1997). In addition, in sev-
`eral tumor types, most notably glioblastoma
`multiforme (Shweiki et al., 1992), VEGF
`expression is correlated with intratumoral
`hypoxia (Ferrara and Davis-Smyth, 1997).
`The role of HIF-1 in activating hypoxia-
`induced transcription of the VEGF gene is
`well established. A hypoxia response ele-
`ment is located approximately 1 kb 5’ to the
`transcription start site of the human VEGF
`gene (Forsythe et al., 1996; Liu et al., 1995)
`and is also present in the mouse and rat Vegf
`genes (Levy et al., 1995; Shima et al., 1996).
`A reporter gene containing a 48-bp VEGF
`hypoxia response element upstream of an
`SV40 promoter and luciferase coding se-
`quences is expressed in a hypoxia-inducible
`manner. Forced overexpression of HIF-1α
`and HIF-1β results in reporter gene expres-
`sion under nonhypoxic conditions and a su-
`perinduction in response to hypoxia
`(Forsythe et al., 1996). A 3-bp mutation that
`disrupts the HIF-1 binding site in the hy-
`poxia response element resulted in a loss of
`reporter gene expression in response to hy-
`poxia or coexpressed HIF-1α and HIF-1β
`(Forsythe et al., 1996). In mouse embryonic
`stem cells homozygous for a targeted muta-
`tion in the Hif1a gene encoding HIF-1α
`there is no induction of VEGF mRNA ex-
`pression in response to hypoxia (Carmeliet
`et al., 1998; Iyer et al., 1998; Ryan et al.,
`1998), although VEGF expression is still
`induced by glucose deprivation (Iyer et al.,
`1998).
`In a series of mouse hepatoma subclones,
`there is a strong correlation between the level
`of hypoxia-induced HIF-1 DNA-binding ac-
`tivity and VEGF mRNA expression (Forsythe
`et al., 1996; Salceda et al., 1996; Wood et al.,
`1996). Furthermore, when these subclones are
`injected into nude mice, there is a strong cor-
`relation between the levels of hypoxia-induced
`HIF-1 and VEGF expression ex vivo and tu-
`
`
`
`Roxane Labs., Inc.
`Exhibit 1017
`Page 007
`
`

`
`mor xenograft growth and angiogenesis in
`vivo (Jiang et al., 1997a; Maxwell et al., 1997).
`In situ hybridization analysis of tumors de-
`rived from HIF-1expressing subclones has
`demonstrated that viable cells surrounding
`areas of necrosis express high levels of VEGF
`and GLUT3 mRNA, whereas such expression
`is not seen at similar locations in tumors de-
`rived from HIF-1 nonexpressing subclones
`(Maxwell et al., 1997), suggesting that physi-
`ologic induction of hypoxia-inducible genes
`mediated by HIF-1 is important for tumor
`growth and vascularization. Vascularization
`of teratomas derived from mouse embryonic
`stem cells is also dramatically affected by loss
`of HIF-1α expression (Carmeliet et al., 1998;
`Ryan et al. 1998).
`In addition to hypoxia, VEGF expres-
`sion is also induced by exposure of both
`transformed and nontransformed cells to a
`variety of growth factors, including epider-
`mal growth factor, basic fibroblast growth
`factor (FGF-2), insulin-like growth factor 1
`(IGF-1), interleukin 1β, platelet-derived
`growth factor, and tumor necrosis factor α
`(TNF-α) (Jackson et al., 1997; Ryuto et al.,
`1996; Stavri et al., 1995a, 1995b; Warren et
`al., 1996). Remarkably, all of these
`cytokines/growth factors also induce HIF-
`1α protein expression and/or HIF-1 DNA-
`binding activity (Feldser et al., 1999;
`Hellwig-Burgel et al., 1999; Zelzer et al.,
`1998; Zhong et al., 2000).
`
`C. Hypoxia
`
`Given the strong correlation between
`angiogenesis and tumor progression de-
`scribed above, one might predict that that
`there is a positive correlation between tu-
`mor oxygenation and tumor progression.
`Remarkably, the opposite is true: first, in
`the majority of human cancers that have
`been analyzed, the mean pO2 is markedly
`
`lower than that of normal tissue in the same
`organ (reviewed in Brown and Giaccia,
`1998; Vaupel, 1996; Vaupel et al., 1989).
`Second, clinical studies in which oxygen-
`ation of cervical or head and neck tumors
`has been measured in situ have demonstrated
`that an intratumoral pO2 of < 10 mmHg is
`associated with a significantly increased
`frequency of tumor invasion and metastasis
`and of patient death (Brizel et al., 1996;
`Hockel et al., 1996). In both cervical and
`head and neck cancers, a correlation be-
`tween elevated lactate levels and metastasis
`has also been reported (Schwickert et al.,
`1995; Walenta et al., 1997).
`What is the basis for the apparent para-
`dox regarding the effects of tumor hypoxia
`and angiogenesis on clinical outcome? The
`blood vessels induced by tumors are struc-
`turally and functionally abnormal, resulting
`in marked regional heterogeneity in tumor
`perfusion (reviewed in Brown and Giaccia,
`1998; Gillies et al., 1999; Vaupel, 1996;
`Vaupel et al., 1989). Analysis of tumor xe-
`nografts has revealed that although there is
`an inverse correlation between mean pO2
`and distance of a tumor cell from the nearest
`capillary (reaching a mean pO2 of 0 at > 200
`µm), individual tumor cells with a pO2 of 0
`can be identified immediately adjacent to a
`blood vessel, indicating an absence of blood
`flow (Helmlinger et al., 1997). These data
`suggest that the induction of angiogenesis is
`necessary but not sufficient to ensure tumor
`progression.
`Intratumoral hypoxia is also associated
`with genetic instability and resistance to
`chemotherapy and radiation, thus providing
`a basis for tumor recurrence (reviewed by
`Brown and Giaccia, 1994, 1998; Yuan and
`Glazer, 1998). For chemotherapy, it is has
`been hypothesized that hypoperfused tumor
`tissue would be deprived both of O2 and any
`chemotherapeutic agent administered
`parenterally; alternatively, cell cycle arrest
`under hypoxic conditions may provide pro-
`
`77
`
`
`
`Roxane Labs., Inc.
`Exhibit 1017
`Page 008
`
`

`
`tection against agents that induce apoptosis
`in proliferating cells. For radiation, it has
`been hypothesized that decreased O2 con-
`centrations would result in decreased for-
`mation of reactive oxygen species, which
`are believed to trigger radiation-induced cell
`death. Whereas these proposed mechanisms
`may indeed contribute to the survival of
`hypoxic tumor cells following cancer
`therapy, it is also possible that hypoxic tu-
`mor cells express survival factors that me-
`diate protection against any apoptosis-in-
`ducing stimulus. Exposure of cultured
`glioma cells to hypoxia results in increased
`survival when the cells are subsequently
`exposed to chemotherapeutic agents (Liang,
`1996), an effect that is obviously indepen-
`dent of vascularization/perfusion. Insulin-
`like growth factor 2 (IGF-2) represents an
`example of a hypoxia-induced survival fac-
`tor (Kim et al., 1998), and its expression
`appears to be regulated by HIF-1 (Feldser et
`al., 1999).
`
`IV. IMMUNOHISTOCHEMICAL
`ANALYSIS OF HIF-1α
`EXPRESSION IN TUMORS
`
`A. Increased Expression of HIF-
`1ααααα in Common Human Cancers
`and Their Metastases
`
`The analysis of tumor-derived cell lines
`either in tissue culture or xenograft assays
`provides a means to investigate molecular
`mechanisms underlying tumor biology. Yet,
`such investigations must rest on a firm foun-
`dation of histopathological studies utilizing
`clinical material that provide the basis for
`generating hypotheses to be tested in model
`systems. Although clinical investigation is
`often limited by the lack of optimal con-
`trols, in the case of cancer, the patient’s
`
`78
`
`normal tissue often can be utilized effec-
`tively for this purpose. With these principles
`in mind, a mouse monoclonal antibody spe-
`cific for HIF-1α was generated, and a sen-
`sitive immunohistochemical assay was de-
`veloped to detect HIF-1α expression in
`formalin-fixed and paraffin-imbedded hu-
`man tumor biopsy specimens (Zhong et al.,
`1999). Whereas HIF-1α expression was not
`detected in the corresponding normal tissue
`(either adjacent to the tumor in the biopsy or
`in separate autopsy specimens), HIF-1α ex-
`pression was detected in two-thirds of the
`primary tumors of the brain, breast, colon,
`lung, ovary, and prostate that were analyzed
`(Table 2). Although HIF-1α expression was
`frequently detected in many, but not all,
`types of malignant tumors, it was not de-
`tected in benign tumors such as breast fi-
`broadenoma and uterine leiomyoma (Zhong
`et al., 1999). In contrast, HIF-1α expression
`was detected in early neoplastic lesions
`(breast comedo-type ductal carcinoma
`in situ, colon adenoma, and prostatic
`intraepithelial neoplasia) detected inciden-
`tally in tumor specimens, suggesting that
`increased HIF-1α expression can occur early
`in tumor progression (Zhong et al., 1999).
`In animal models, induction of angiogen-
`esis has been demonstrated during the tran-
`sition from hyperplasia to neoplasia
`(Folkman et al., 1989). HIF-1α expression
`in early neoplastic lesions may promote
`vascularization mediated by VEGF and
`possibly other HIF-1-regulated angiogenic
`factors.
`
`B. HIF-1ααααα Expression in Brain
`Tumors
`
`Immunohistochemical analysis of brain
`tumors was particularly instructive (Zagzag
`et al., 2000). Several studies have shown
`that compared with low-grade gliomas, the
`
`
`
`Roxane Labs., Inc.
`Exhibit 1017
`Page 009
`
`

`
`Table 2
`Expression of HIF-1a in Human Cancers Compared with Corresponding Normal Tissuea
`
`Organ
`
`Normal Tissue
`
`Benign Tumors
`
`Primary Cancers
`
`Metastases
`
`Brain
`Breast
`Colon
`Lung
`Ovary
`Prostate
`Uterus
`
`0/10
`0/18
`0/24
`0/10
`0/10
`0/12
`0/3
`
`—
`0/10b
`—
`—
`—
`—
`0/2e
`
`5/9
`25/52
`22/22
`3/3
`2/2
`9/11
`—
`
`—
`9/13c
`9/10c
`—
`—
`5/10c,d
`—
`
`aData from Zhong et al., 1999.
`bFibroadenomas
`cLymph node metastases
`dBone metastases
`eLeiomyomas
`
`high-grade glioblastoma multiforme (GBM)
`has
`significantly
`higher levels of VEGF expression and
`neovascularization (Leung et al., 1997;
`Pietsch et al., 1997; Plate et al., 1992; Plate
`and Mennel, 1995; Plate and Warnke, 1997;
`Takano et al., 1996; Takekawa and Sawada,
`1998). Even in low-grade astrocytomas, pa-
`tient survival is inversely correlated with
`VEGF expression and vascular density
`(Abdulrauf et al., 1998). GBM is also char-
`acterized by extensive areas of necrosis, rep-
`resenting tumor cells located beyond the
`effective diffusion distance of O2 from the
`nearest blood vessel. VEGF mRNA is highly
`
`expressed in the pseudopalisading cells sur-
`rounding necrotic areas, suggesting that hy-
`poxia is a stimulus for VEGF expression by
`these cells (Plate et al., 1992; Shweiki et al.,
`1992). Remarkably, the expression of HIF-
`1α is correlated with tumor grade and, in
`particular, with vascularity (Table 3 and Fig-
`ure 1). Expression of HIF-1β was also ana-
`lyzed by immunohistochemistry, and its ex-
`pression is also correlated with tumor grade,
`although it is more widely expressed than
`HIF-1α and is not as strongly correlated
`with tumor vascularity. Most striking is the
`pattern of HIF-1α protein expression in
`GBM which are identical to that described
`
`Table 3
`Vascularity and Expression of HIF-1ααααα and HIF-1βββββ in Human Brain Tumorsa
`
`Tumor
`Low grade astrocytoma
`Low grade mixed glioma
`Anaplastic astrocytoma
`Anaplastic oligodendroglioma
`Anaplastic mixed gliomad
`Glioblastoma multiforme
`Hemangioblastoma
`
`n
`2
`2
`2
`3
`9
`14
`10
`
`Vascularityb
`1.0
`2.5
`3.5
`3.7
`2.4
`3.9
`3.7
`
`HIF-1αααααc
`1.5
`2.5
`3.5
`3.0
`2.8
`4.0
`3.2
`
`HIF-1βββββc
`1.0
`0.5
`2.5
`2.3
`2.4
`3.2
`3.2
`
`aData from Zagzag et al., 2000.
`bVascularity: 0, vascular hyperplasia not detected; 1, minimal; 2, mild; 3, moderate; 4, marked
`increase in tumor vascularity compared with surrounding normal brain tissue
`cHIF-1α or HIF-1β: 0, nuclear staining not detected by immunohistochemistry; 1, staining
`in less than 1% of nuclei; 2, 1–10% of nuclei; 3, 10–50%; 4, greater than 50%.
`dastrocytoma/oligodendroglioma
`
`79
`
`
`
`Roxane Labs., Inc.
`Exhibit 1017
`Page 010
`
`

`
`Figure 1. Correlation between brain tumor grade, vascularity, and HIF-1a expression. Based on an
`analysis of 42 human brain tumors (Zagzag et al., 2000). See Table 3 for definitions.
`
`high
`for VEGF mRNA, with
`of HIF-1α
`in
`levels
`detected
`pseudopalisading cells surrounding areas of
`necrosis (Zagzag et al., 2000). Taken to-
`gether, these data are consistent with the
`hypothesis that HIF-1 mediates hypoxia-
`induced VEGF expression in GBM.
`Hemangioblastoma is another extremely
`vascular brain tumor. The vascularity of
`hemangioblastomas is so great that, as the
`name implies, they were originally believed
`to be tumors of hemangioblasts, the stem
`cells from which hematopoietic and endo-
`thelial cells are derived. However, it is now
`apparent that the “stromal” cells (as op-
`posed to the blood vessels) are in fact tumor
`cells that produce extremely high levels of
`VEGF (Flamme et al., 1998; Krieg et al.,
`1998; Stratmann et al., 1997; Wizigman-
`Voos et al., 1995). Unlike GBM, hemangio-
`blastomas are so richly vascularized that
`areas of necrosis are unusual. Yet, immuno-
`histochemistry revealed that in those he-
`mangioblastomas with the greatest degree
`
`80
`
`of vascularization, HIF-1α was expressed
`in the majority of tumor cells, whereas ex-
`pression was not detected in vascular cells
`(Zagzag et al., 2000; Zhong et al., 1999).
`The detection of high levels of HIF-1α in
`tumor cells immediately adjacent to most
`blood vessels suggests that such expression
`is not hypoxia induced.
`A similar pattern of expression was de-
`tected in another extremely vascularized
`tumor, clear cell renal carcinoma (Zhong et
`al., 1999). Hemangioblastoma and clear cell
`renal carcinoma share in common inactiva-
`tion of the VHL gene encoding the von
`Hippel-Lindau tumor suppressor protein
`(Gnarra et al., 1994; Herman et al., 1994;
`Kanno et al., 1994; Shuin et al., 1994). In
`renal carcinoma cell lines lacking VHL ex-
`pression, the normal O2-regulated expres-
`sion of HIF-1α and HIF-2α is lost; these
`proteins are constitutively expressed at high
`levels under nonhypoxic conditions and
`activate transcription of downstream target
`genes, including VEGF (Maxwell et al.,
`
`
`
`Roxane Labs., Inc.
`Exhibit 1017
`Page 011
`
`

`
`1999). Analysis of multiple renal carcinoma
`cell lines with VHL loss-of-function revealed
`that HIF-2α was overexpressed in all,
`whereas in several lines HIF-1α expression
`was completely absent. In contrast, HIF-1α
`was highly expressed in all of the heman-
`gioblastoma (n = 10) and clear cell renal
`carcinoma (n = 1) biopsies that were ana-
`lyzed by immunohistochemistry (Zagzag et
`al., 2000; Zhong et al., 1999), suggesting
`that there may be a selection against HIF-
`1α overexpression in cultured renal carci-
`noma cells that is not relevant to their bio-
`logical behavior in vivo. However, primary
`clear cell renal carcinomas express an
`antisense HIF-1α RNA species (Thrash-
`Bingham and Tartof, 1999). Thus, the role
`of HIF-1α expression in these tumors re-
`quires further analysis.
`
`C. HIF-1ααααα and Tumor Invasion
`
`GBM is one of the most highly inva-
`sive forms of human cancer and patients
`with this tumor have a life expectan