Clinical Biomarkers for Hypoxia Targeting

Abstract

Tumor hypoxia or a reduction of the tissue oxygen tension is a key microenvironmental factor for tumor progression and treatment resistance in solid tumors. Because hypoxic tumor cells have been demonstrated to be more resistant to ionizing radiation, hypoxia has been a focus of laboratory and clinical research in radiation therapy for many decades. It is believed that proper detection of hypoxic regions would guide treatment options and ultimately improve tumor response. To date, most clinical efforts in targeting tumor hypoxia have yielded equivocal results due to the lack of appropriate patient selection. However, with improved understanding of the molecular pathways regulated by hypoxia and the discovery of novel hypoxia markers, the prospect of targeting hypoxia has become more tangible. This chapter will focus on the development of clinical biomarkers for hypoxia targeting.

Introduction

Hypoxia is a common phenomenon in solid neoplasms. It arises when tissue oxygen demands exceed the oxygen supply from the vasculature. Hypoxic regions develop within solid tumors due to aberrant blood vessel formation, fluctuations in blood flow and increasing oxygen demands from rapid tumor expansion.¹ That hypoxia exists in human tumors was first demonstrated by Thomlinson and Gray in 1955.² It was subsequently noted that hypoxia limits tumor cell response to radiation and chemotherapy and predisposes them to metastasis; these findings resulted in substantial laboratory and clinical efforts to overcomethis microenvironmental effect.^{1, 3–5} Unfortunately, most clinical trials targeting hypoxia have yielded inconclusive results to date.^6–10 The lack of improved outcomes from hypoxia targeting could be partially attributed to poor patient selection for hypoxia targeted therapies.¹¹ Therefore, considerable efforts have been devoted to identify clinical markers for tumor hypoxia. These hypoxic markers could be used to identify patients most likely to benefit from a hypoxia-sensitizing treatment regimens. Finally it has been proposed that measurement of hypoxia may also be a method to monitor treatment efficacy.

At the present time, there exist several clinical approaches for detecting tumor hypoxia. However, none of these approaches represents a clear “gold standard” as agreed by the experts in a recent hypoxia workshop that was convened by the National Cancer Institute.¹² A reason for the lack of an ideal biomarker is that there exist extreme spatial and temporal heterogeneities in tissue oxygen levels due to the complex nature of blood supplies and cellular oxygen consumption, and none of the current methods can completely capture this heterogeneity. Existing methods for assessing hypoxia differ from one another in several aspects, including sampled tissue volumes (macroscopic versus microscopic), time intervals (seconds to hours), compartment (intracellular versus interstitial) and type of hypoxia (chronic versus acute). Despite their differences, these approaches or biomarkers can be categorized into 2 groups: direct and indirect. Their advantages and disadvantages are detailed below and summarized in Table 1.

Table 1.

Advantages and disadvantages for different approaches in assessing tumor hypoxia

Method	Examples	Measure	Spatial resolution	Advantages	Disadvantages
PO2 Histography	Eppendorf electrode OxyLite fiber optic probe	pO2	0.5 mm (thousands of cells)	Direct Rapid real-time measurements Validated in human tumors	Invasive Tumor inaccessibility Pressure dependence Inter-observer variability Readings affected by necrosis No spatial information
Direct imaging	19F-MRI BOLD-MRI EPRI	pO2 or deoxy- Hb	0.2–1mm	Some approach (BOLD- MRI) non invasive Spatial information	Some approaches (19F-MRI or EPRI) require injection of contrast materials locally or systemically Minimal clinical data available
Exogenous Markers	EF5 Pimonidazole	Chronic hypoxia	1.0 um (single cell)	Highly sensitive Reproducible	Require drug injection Require extra biopsies Sampling bias
Endogenous hypoxia marker	HIF-1 CA IX Glut-1	Biologic hypoxia	1.0 um (single cell)	Apply to archival tissues No drug injection No extra biopsies	Less hypoxia specific Variability in staining & interpretation Sampling bias Antibody specificity
Secreted markers	OPN VEGF	Biologic hypoxia	N/A	Non invasive Inexpensive No biopsy or drug injection Serial measurements	Less hypoxia specific Less tumor specific Specimen processing critical Antibody specificity
PET-based hypoxia imaging	18F-MISO 18F-FAZA 18F-EF5 18FETNIM 60CuATSM 24I-IAZGP	Chronic hypoxia	2–10 mm	Spatial resolution Serial measurements RT directed targeting Reproducible	Requires dedicated equipments (tracer generation & imaging) Expensive Radiation exposure Tracer synthesis expertise Low sensitivity for certain tracers

Hb: Hemoglobin

Direct oxygen measurements in tissues

Needle Electrode

Direct approaches can be applied to tissue (needle electrodes, fiberoptic probes) or blood (measurements or imaging of oxyhemoglobin saturation and oxygen diffusion). Polarographic needle electrodes (pO₂ histograph, Eppendorf, Hamburg, Germany) provided the first convincing evidence that hypoxia existed in human solid tumors.^{13, 14} The sensing electrode, mounted on the tip of a needle, is advanced via a step motor through the tissue, taking rapid measurements (1.4 s) to avoid spurious readings from pressure artifacts caused by the needle. A histogram of oxygen tensions (pO₂) can then be obtained from multiple sampling points along different tracks. Normal tissues typically show a Gaussian pO₂ distribution with the median value between 40–60 mm Hg; whereas tumors invariably show lower pO₂ measurements (Figure 1). Clinical investigations with the microelectrodes have illustrated that regions of hypoxia can be found in a wide range of human tumors, including cancers of the brain, head and neck (HN), lung, breast, rectum, pancreas, cervix and prostate.¹⁵ Several studies have showed that low tumor pO₂, defined by either the median value or the hypoxic fraction (% readings below 2.5 or 5 mm Hg), correlated with poor treatment outcomes in HN, cervical, prostate and lung cancers.^16–23 One study also found that tumor pO₂ predicted for pathologically persistent neck nodes in patients undergoing a neck dissection for clinical N2–3 necks after chemoradiation treatment.²⁴ Pooled data from several institutions in 397 head and neck cancer (HNC) patients provided strong evidence that tumor pO₂ is an independent predictor for survival.²⁵ In cervical cancers and sarcomas, lower pO₂ has been associated with increased risk of nodal and distant metastasis, respectively.^{21, 26, 27}

Figure 1.

An example of the pO₂ distribution as measured by the polarographic electrode in (a) Control normal subcutaneous tissue, (b) Involved neck node and (3) Primary head and neck cancer in the same patient with a head and neck squamous cell carcinoma.

Although the microelectrode technique can directly measure tumor pO₂, it does suffer from several drawbacks that make it difficult for general use. These include high machine cost, invasiveness, tumor inaccessibility, pressure dependence, inter-observer variability, failure to distinguish necrosis from hypoxia, and the lack of spatial information on hypoxia. In addition, it primarily measures extracellular oxygen level at a low resolution of 500 cells or greater. Therefore, although the microelectrode has been the most studied approach for assessing hypoxia to date, it is unlikely that this technique can be used routinely to select for patients with hypoxic tumor in large phase III clinical trials.²⁸

Direct imaging

Different techniques can be used to directly image oxygen levels in tissues or blood based on known properties of paramagnetic agents. One such technique is ¹⁹F MRI, which employs injectable perfluorocarbons (PFC), whose ¹⁹F nuclear magnetic resonance spin lattice is highly sensitive to oxygen, hence allowing measurement of vascular and tissue oxygenation.¹² This approach, however, is limited by the requirement of local injection of PFC compounds directly into the tumor for imaging.

Another approach is the blood oxygen level-dependent magnetic resonance imaging (BOLD MRI), whose image contrast rests on the balance between paramagnetic deoxyhemoglobin and diamagnetic oxygehemoglobin and the effect of the latter on MR signals.¹² Although BOLD MRI does not require injection of an exogenous contrast agent, its signal can be influence by factors other than hypoxia including blood flow, CO₂ tension, hematocrit, pH and biphosphoglycerate. ²⁹ In a recent study of 24 patients with prostate cancer undergoing radical prostatectomy, BOLD MRI was performed preoperatively and correlated with pimonidazole staining, an indirect marker for hypoxia (see below), using a co-registered histologic and imaging grid map of whole mount sections. R2 (MR relaxivity parameter) maps from BOLD MRI yielded high sensitivity but low specificity for defining hypoxic tumor regions stained with pimonidazole.³⁰

EPR or electron paramagnetic resonance imaging detects species with unpaired electrons.¹² Molecular oxygen, which has 2 unpaired electrons, can be imaged when a biologically compatible, inert free radical is introduced directly into the tumor. This approach is currently still in preclinical models and is limited by the requirement for direct tumoral injection or implantation of particulate paramagnetic materials.¹² It does however, have a theoretical advantage of allowing repeated measurement over a long period of time. To date, clinical experience with these direct imaging approaches, specifically with regards to predicting prognosis is limited and their utility as biomarkers for tumor hypoxia needs to be validated in large clinical trials.

Indirect approach – Injectable markers

Immunohistochemical staining of 2- nitroimidazole compounds on tissue sections

Indirect approaches use injectable molecular reporters of oxygen as endpoints. These reporters include 2-nitroimidazole compounds such as misonidazole (1-(alpha-methoxymethyl ethanol)-2-nitroimidazole)³¹, pimonidazole (1-(2-nitro-1-imidazolyl)-3-N-piperidino-2-propanolol)³², and EF5 (nitroimidazole[2-(2-nitro-1H-imidazol-1-yl)-N-(2,2,3,3,3-pentaflouropropyl) acetamide).³³ These compounds form stable adducts with intracellular macromolecules and this binding is prohibited at higher oxygen levels.³⁴ Detection of these adducts with antibodies can then provide quantitative information on the relative oxygenation at cellular resolution.^{35, 36} In general, 2-nitroimidazole markers stain for areas of chronic hypoxia, at pO₂ levels below 10 mm Hg and are more sensitive at severe hypoxic conditions than the microelectrode.^{36, 37} Although these compounds share a common ring structure and the nitro group, which confer oxygen dependency, their behaviors are not the same due to the different side chains, which significantly affect the pharmacokinetics, accumulation rates and tissue distributions of these compounds. For examples, EF5 is significantly more lipophilic than other 2-nitroimidazole markers, resulting in more even biodistribution in the body with slower whole body drug elimination.³⁸ Pimonidazole accumulation rate is more dependent on pH than other compounds, resulting in more pimonidazole binding on the vessel side in transient hypoxia.³⁵ The differences in the pharmacokinetics and tissue distribution of these agents have been exploited to analyze the hypoxic fluctuation in individual tumors in experimental animal models. Double hypoxia marker assays, in which consecutive injections of two different 2-nitroimidazole compounds were administered before and after treatment, have been used to visualize spatial and temporal changes from carbogen breathing and hydralazine infusion.^{39, 40} For an excellent review of injectable markers, please see Ljungkvist et al.³⁵

Though widely used in animal studies, there is minimal clinical data regarding the prognostic significance of this approach in cancer patients. In a small study evaluating pimonidazole, microvessel density count and carbonic anhydrase IX (CA IX) binding in 42 HNC, pimonidazole staining was more pronounced at distance > 100 μm from blood vessels than CA IX, suggesting that it is more specific for chronic hypoxia.⁴¹ High pimonidazole staining correlated with a higher risk of locoregional relapse in patients treated with radiotherapy (RT) alone but not in those treated with RT plus carbogen and nicotinamide, which were used to modulate tumor hypoxia. In a small study of 16 sarcoma patients, severe hypoxia, defined as ≥ 20% of EF5 binding in the primary tumors, correlated with increased risk of distant metastasis.⁴² In 18 patients with supratentorial gliomas, increasing EF5 binding was associated with higher tumor grade and shorter time to recurrence.⁴³

This immunohistochemical (IHC) approach, though informative and elegant, is limited by the requirement for exogenous drug administration, additional tumor biopsies for staining and expertise in staining quantification. Because of these limitations, it has not been widely used clinically and needs to be validated in larger studies. An on-going phase III trial is evaluating the use of pimonidazole as a predictive marker. Patients with larynx cancer are randomized to receive accelerated radiation therapy alone or the same treatment with carbogen and nicotinamide.³⁵ The result of this study will help to elucidate the role of pimonidazole as a clinical hypoxia biomarker.

Imaging studies using 2- nitroimidazole compounds

These 2-nitroimidazole compounds can also be labeled with ¹⁸F and employed as special tracers for hypoxia imaging using PET or SPECT imaging approaches. The most extensively investigated 2-nitroimidazole tracer is ¹⁸Fmisonidazole (F-MISO). Typically, a tumor to blood ratio of ≥ 1.2 has been used as a reasonable cut point between normoxia and hypoxia for F-MISO. Tumor to muscle ratio of F-MISO has been shown to significantly correlate with tumor hypoxic fraction as measured by the polarographic needle electrode in 16 HNC patients.⁴⁴ Preliminary clinical data suggested that hypoxia imaging with F-MISO could be used to predict treatment outcome and assess treatment response to RT and chemotherapy in solid cancers.^45–48 The largest series was in 73 HNC patients where pretreatment F-MISO imaging was an independent prognostic factor for survival when fluorodeoxyglucose standard uptake value, a measure of glucose metabolism, was removed from the model.⁴⁵ More importantly, in 45 patients with pretreatment F-MISO PET imaging, hypoxic tumors were less likely to fail when treated with a combined regimen of chemoradiation and the hypoxic cytotoxic agent Tirapazamine (TPZ) when compared to a non-TPZ regimen.⁴⁹ Serial F-MISO imaging at 3–4 weeks into the radiation course for HNC has also been performed in 2 studies, which showed either eliminated or decreased F-MISO uptake in most if not all patients.^{50, 51} However, the prognostic implication of decreased F-MISO uptake during treatment is not clear. Although all patients with persistent or increased F-MISO uptake failed, a substantial portion of those with improved uptake also recurred, making it hard to interpret the results of these studies.

Other PET tracers that have been investigated in patients are ¹⁸F-EF5, ¹⁸F-FAZA and ⁶⁴Cu-ATSM. Clinical data on the ¹⁸F-EF5 and ¹⁸F-FAZA are minimal with small series reporting the feasibility of imaging patients with solid tumor.^{12, 52} Cu-ATSM (Copper (II) diacetyl-bis(N4)-methylthiosemicarbazone) is activated under hypoxia by a different mechanism than the 2-nitroimidazole compounds and mechanistic studies suggested the involvement of the mitochondria.⁵³ It has been shown to correlate strongly with oxygen electrode measurements in animal tumors and to be feasible to use for imaging tumor hypoxia in cancer patients.⁵⁴ It also enjoys the advantage of having a short half-life (23 minutes), which makes it possible to perform serial imaging studies on the same patients.¹² In two small series of less than 20 patients each, Cu-ATSM has been shown to accumulate in tumors in cervical and non-small cell lung cancers (NSCLC) and its uptake pattern was predictive for treatment response in NSCLC and survival in cervical cancers.^{55, 56}

From the radiation-targeting standpoint, PET imaging with hypoxic tracers can theoretically be combined with intensity-modulated radiotherapy (IMRT) for dose escalation to improve local control. The ability of IMRT for dose painting provides a tantalizing possibility of delivering higher doses to hypoxic regions visualized by PET tracers in the tumor without increasing normal tissue toxicity. The premise of such dose escalation requires that hypoxic regions in the tumor remain relatively stable before and during the course of IMRT treatment over several weeks. Serial F-MISO imaging at 3–4 weeks into the radiation course for HNC suggested that the regions of persistent hypoxia, if present, were located in the pre-treatment hypoxic volumes.^{50, 51} In addition, feasibility studies suggested that dose painting can be applied to target hypoxic regions in the tumor using F-MISO or ¹⁸F-FAZA PET/CT-guided IMRT.^{51, 57, 58} The main question is whether this is clinically achievable in patients, and clinical trials for dose escalation using hypoxia PET imaging is warranted. However, since PET-based hypoxia imaging requires expensive dedicated equipments for both imaging and tracer generation, it is only available at selective academic institutions, which may limit its general use.

Indirect approach – Tissue Endogenous markers

Endogenous molecular markers for tumor hypoxia represent proteins and genes whose expressions are induced by hypoxic exposure. One of the most studied oxygen response pathways is that mediated by the hypoxia inducible factor-1 (HIF-1), which regulates genes that are involved in cell metabolism, angiogenesis, invasion, metastasis and apoptosis. HIF-1 and several of its downstream targets such as Glut-1 (glucose transporter-1), CA IX and vascular endothelial growth factor (VEGF) have been widely investigated as prognostic markers in HNC with mixed results. Table 2 summarized representative large clinical series (>40 patients) that focused on the prognostic significance of HIF-1, CA IX and Glut-1 for certain solid tumors (HN, cervix, breast and lung cancers). It is by no mean an exhaustive list but it does show that in general, elevated expression of these markers portends poorer outcomes in patients treated with non-surgical, and for certain sites, surgical therapies.

Table 2.

Significance of HIF-1, CA IX & Glut-1 endogenous markers for selective solid cancers, including head and neck, cervical, breast & non-small cell lung cancers

HIF Markers		Head & Neck Cancer
Author	Hypoxia Marker	# Pts	Treatment	Survival
Aebersold89	HIF-1α	98	RT or RT+C	LRC, DFS, OS (Multivariate)
Koukourakis90	HIF-1α, HIF-2α	75	RT+C	LRC, OS for HIF-2α only (Multivariate)
Beasley91	HIF-1α	69	S	Improved DFS, OS (Multivariate)
Hui92	HIF-1α, HIF-2α, CA IX, VEGF	90 (NPC)	RT or RT+C	PFS for HIF-1α + CA IX & HIF-1α + VEGF but not individual marker (Multivariate)
Kyzas93	HIF-1α, VEGF	81	S	OS for VEGF, not for HIF-1α (Univariate)
Winter94	HIF-2α, CA IX	140	S	CSS & DFS for HIF-1α (multivariate); Improved with addition of HIF-2α; No significance for CAIX
Koukourakis95	HIF-2α, CA IX	198	RT (CHART vs conventional)	LRC, OS for both (Multivariate)
HIF Markers		Cervical Cancers
Bachtiary96	HIF-1α	67	RT	PFS, CSS (Multivariate)
Birner97	HIF-1α	91	S ± RT	DFS, OS (Multivariate)
Burri98	HIF-1α	78	RT ± C	LPFS (Univariate), OS (Multivariate)
Hutchison99	HIF-1α	99	RT	No (DFS, MFS, LRFS as endpoints)
Haugland78	HIF-1α	42	RT	No (DFS as endpoint)
HIF Markers		Breast Cancers
Dales100	HIF-1α	745	S	DMFS, OS (Multivariate)
Kronblad101	HIF-1α, VEGF	377, Stage II, premenopause	S + RT+ C/Tam	RFS for low grade tumors (Multivariate)
Schindl102	HIF-1α	206 (LN +)	S ± C	DFS, OS (Multivariate)
Vleugel103	HIF-1α, CA IX, Glut 1	166	S + C	DFS (Only HIF-1α evaluated & Univariate)
Trastour104	HIF-1α, CA IX	132	S ± R ± C/Tam	DFS, MFS for HIF-1α mainly (Multivariate)
HIF Markers		Non-Small Cell Lung Cancers
Giatromanolaki 105	HIF-1α, HIF-2α,VEGF	98	S	OS for HIF-2α only (Multivariate)
Swinson106	HIF-1α	172	S ± RT ± C	OS for CA IX only (Multivariate)
Kim107	HIF-1α, CA IX	74, Stage I-II	S	DFS for CA IX Only (Multivariate)
CA IX only		Head & Neck Cancer
Beasley108	CA IX	79	Surgery	Not assessed
Koukourakis109	CA IX	75	CRT	LRCS, OS (Univariate only)
Jonathan110	CAIX, Glut-1, Glut-3	58	RT + ARCON	Better LRC & MFS with stronger CA IX & Glut-3 (Univariate)
De Schutter111	CA IX, Glut-1	67	RT ± C	LRC, DFS only for CA IX + Glut-1 (multivariate)
CA IX only		Cervix
Hedley112	CA IX	102	RT ± C	No prognostic significance
Loncaster113	CA IX	130	RT	MFS, OS (Multivariate)
CA IX only		Breast Cancer
Span114	CA IX (mRNA)	253	S ± RT ± C/Tam	More resistance to adjuvant systemic treatment (Multivariate)
Brennan115	CA IX	400, Stage II, premenopause	S + R ±Tam	CSS (Multivariate)
Hussain116	CA IX	144	S	OS (Multivariate)
Chia	CA IX	‘103	S + RT ± C/Tam	OS (Multivariate)
CA IX only		Non Small Cell Lung Cancer
Giatromanolaki117	CAIX, HIF-1α, HIF-2α	107	S	OS (Multivariate)
Kon-No118	CA IX	134	S	OS, DFS (Univariate only, not multivariate)
Swinson119	CA IX	175	S	OS for perinuclear staining pattern (Multivariate)
Glut-1 only		Head & Neck Cancer
Oliver120	Glut-1	54, OC only	S	LRC, FFR, CSS (Univariate)
Kunkle121	Glut-1	118, OC only	S ± RT	OS (multivariate)
Mineta122	Glut-1	99, HP only	CRT	PFS (Multivariate)
Glut-1 only		Cervical Cancer
Airley123	Glut-1	121	RT	MFS (Multivariate)
Mayer76	Glut-1	47	S ± C or RT ± C	OS, PFS (Univariate only)
Glut-1 only		Breast Cancer
Stackhouse124	Glut-1	141	S ± RT ± C/Tam	No significance for recurrence
Glut-1 only		Non-Small Cell Lung Cancer
Minami125	Glut-1	47	S	OS (Multivariate)
Nguyen126	Glut-1	53	S ± R ± C	No significance for DFS

Pt: patients; S: Surgery; RT: radiotherapy; C: chemotherapy; Tam: Tamoxifen; ARCON: Carbogen and nicotidamide; CHART: Continuous hyperfractionated accelerated radiotherapy

OC: Oral cavity cancer; NPC: Nasopharyngeal carcinoma; HP: Hypopharyngeal carcinoma; LN+: Lymph node positive; LRC: locoregional control; DFS: disease-free survival; PFS: Progression-free survival: OS: Overall survival; CSS: Cancer specific suvival; LPFS: Local progression-free survival; MFS: Metastasis-free survival; FFR: Freedom from relapse

Other endogenous tissue makers that have been studied in relation to hypoxia include VEGF, BNIP3 (Bcl-2/adenovirus E1B 19 kDA-interacting enzyme), Lysyl oxidase (LOX), Lactate Dehydrogenase isoenzyme-5 (LDH-5), Plasminogen activator inhibitor-1 (PAI-1) and Galectin-1.^59–67 At the present time, the clinical relevance of these markers is unclear since results are either conflicting such as those for VEGF ^67–69 or intriguing as for LOX and LDH-5 but would need further validation from larger, more uniform series.

The advantage of endogenous markers is that levels of these proteins can be assessed on archival materials, thereby allowing rapid correlation to treatment outcomes. In addition it requires neither the injection of foreign material nor any additional invasive procedure beyond that of a biopsy at diagnosis. A significant drawback to these approaches is that these proteins can be regulated by factors other than hypoxia. For example, HIF-1α expression can be influenced by several non-hypoxic stimuli including nitric oxide, cytokines (interleukin-β and tumor necrosis factor-α), trophic stimuli (serum, insulin, insulin-like-growth factors) and oncogenes (p53, Vsrc, PTEN, etc).^70–73 Comparison of the staining patterns between endogenous and injectable markers showed that the former, in general, stained more diffusely and closer to the blood vessels than the latter, suggesting other modes of induction and activation at a wider range of oxygen concentration.^{41, 74} In addition, there is minimal correlation between intensity of endogeneous marker staining and tumor pO_2. In the most vigorous studies where tumor biopsies were performed along the paths of the polarographic electrode and stained for HIF-1α, CA IX and Glut-1, there was no observed correlation between any staining parameter and measured pO₂.^75–77

Other major drawbacks of endogenous tissue makers include the lack of standardization for the IHC protocol and result interpretation for individual markers across different laboratories. For example, different studies used different HIF-1 antibodies with different sensitivities and binding affinities for the protein. The duration of tissue fixation is highly variable and extended fixation time has been shown to reduce HIF-1α expression.⁷⁸ Comparisons of results are further hampered by diverse image evaluation techniques and different interpretations of positive staining. While some studies employed quantitative image analysis, others used visual estimation of positive stained cell number and intensity. Moreover, while most authors assessed only nuclear staining for transcriptional factors such as HIF-1α, others also included cytoplasmic expression. Finally, since only a very small portion of tumor is assessed with endogenous markers, this approach is prone to sampling bias and does not reflect hypoxia heterogeneity in the entire tumor. These drawbacks make it less desirable to use individual endogenous markers alone to select for patients with hypoxic tumors.

To circumvent this dilemma, suggestions have been made to combine several endogenous markers together to improve hypoxia specificity. For example, gene expression analysis has been used to generate a hypoxia gene signature or a hypoxia metagene to predict treatment outcomes in several solid tumors, including HNC.^{79, 80} Chi et al described a gene array signature for hypoxia in breast cancer patients that strongly predicted for poor outcomes.⁷⁹ Using gene expression profiling of 59 HNC, Winter et al generated a hypoxia metagene by identifying genes whose expression clustered with 10 known hypoxia regulated genes.⁸⁰ They found that this metagene was able to predict recurrence-free survival in an independent HNC data set as well as overall survival in another breast cancer series. We have also used a combination of gene expression and proteomic analyses to identify novel hypoxia induced proteins. After confirming their hypoxic inducibility in cell lines and animal models, we investigated their utility in combination with CA IX to predict outcomes by staining a HNC tissue array with known tumor pO₂. These studies resulted in a panel of 4 hypoxia markers (CA IX, Lysyl oxidase (LOX), Galectin-1 and Ephrin A1) that can be used to predict treatment outcomes in terms of cancer-specific survival.⁶⁰ (Figure 2) These endogenous hypoxia signatures, though promising, need to be validated in larger independent datasets before they can be used in the clinical settings.

Figure 2.

Cancer-specific survival by hypoxia marker score comprised of Galectin-1, Ephrin A1, Lysyl Oxidase, CA IX cytoplasmic and CA IX membrane staining, where a score of 1 was assigned to strong staining for each marker and a score of 0 to negative and week staining. This has been adjusted for age and hemoglobin levels, 2 other significant factors on univariate analysis.

Indirect approach - Secreted hypoxia markers

Our laboratory has focused on identifying secreted markers of hypoxia that can be rapidly and inexpensively measured in the blood. Two markers that have been tested clinically with mixed results are VEGF and osteopontin (OPN). Although circulating VEGF levels were elevated in cancer patients^{81, 82} and in those with acute hypoxia such as obstructive apnea⁸³, the relationship between tumor hypoxia and systemic VEGF levels is unclear. Dunst et al found that serum VEGF levels independently correlated with hypoxic tumor subvolume in 56 HNC patients.⁸⁴ However, it also correlated with total tumor volume, hemoglobin level and platelet counts. They did not report on the clinical significance of serum VEGF levels in terms of treatment outcomes. In contrast, we did not find a direct relationship between plasma VEGF and tumor pO₂ in 48 HNC patients in our study (unpublished observations). We did however found a small but significant relationship between OPN level and tumor pO₂ in our patient cohort.⁸⁵ This was confirmed by Nordsmark et al.⁸⁶ In addition, plasma OPN was an independent and significant predictor for treatment outcomes in these patients and another independent group of HNC patients.⁸⁷ These results were confirmed by the DAHANCA group in a larger cohort of HNC patients treated with radiation therapy +/− nimorazole, a hypoxic cell radiosensititizer.⁸⁸ Intriguingly, only patients with high pretreatment circulating OPN levels benefited from nimorazole whereas those with low-intermediate levels did not, suggesting that OPN may be use to select patients for hypoxia targeting. Further validation of this marker is ongoing in another set of HNC patients treated with or without Tirapazamine (TPZ), a hypoxic cell cytotoxin. Advantages of secreted markers for hypoxia is that they are non-invasive, easy to measure, inexpensive and allow for serial measurements through the course of therapies. However, they do suffer the same drawbacks faced by endogenous tissue markers including the lack of method standardization and regulation by factors other than hypoxia. In addition, spatial information is lost and contributions from non-cancerous tissues and other pathological processes such as inflammation cannot be ruled out. Therefore, combining one or multiple of these markers with another approach of assessing hypoxia such as imaging is more desirable than using them alone.

Conclusion

In summary there exist presently several ways or biomarkers for assessing hypoxia. However, none of these approaches, by itself, can capture all the intricacies of tumor hypoxia and its heterogeneity. Therefore, none is currently considered the “gold standard” biomarker for hypoxia. In theory, a combination of biomarkers is more robust than a single marker; yet, there is no current clinical data to support such a hypothesis. Further work is needed to validate the utility of incorporating multiple biomarkers such as imaging plus tissue or blood markers to identify patients with hypoxic tumors for future targeting.