If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Division of Cancer Sciences, The University of Manchester, Manchester, UKSydney Medical School, The University of Sydney, Sydney, New South Wales, Australia
Division of Cancer Sciences, The University of Manchester, Manchester, UKCRUK Manchester Institute and Manchester Cancer Research Centre, Manchester, UK
The hallmarks of cancer underpin current understanding and perspectives of cancer biology all of which are associated with tumour hypoxia.
•
There is level 1a evidence that giving hypoxia-targeting treatments improves locoregional control and survival; yet there are no biomarkers in routine use to aid management decisions.
•
The biological cellular adaptation and transcriptional response to hypoxia underlies the development of gene expression signatures.
•
Gene expression signatures address past limitations of hypoxia biomarkers and could progress biologically optimised radiotherapy.
•
The clinical oncology community should standardise practice to build multiple accessible cohorts for validating signatures.
Abstract
The history of radiotherapy is intertwined with research on hypoxia. There is level 1a evidence that giving hypoxia-targeting treatments with radiotherapy improves locoregional control and survival without compromising late side-effects. Despite coming in and out of vogue over decades, there is now an established role for hypoxia in driving molecular alterations promoting tumour progression and metastases. While tumour genomic complexity and immune profiling offer promise, there is a stronger evidence base for personalising radiotherapy based on hypoxia status. Despite this, there is only one phase III trial targeting hypoxia modification with full transcriptomic data available. There are no biomarkers in routine use for patients undergoing radiotherapy to aid management decisions, and a roadmap is needed to ensure consistency and provide a benchmark for progression to application. Gene expression signatures address past limitations of hypoxia biomarkers and could progress biologically optimised radiotherapy. Here, we review recent developments in generating hypoxia gene expression signatures and highlight progress addressing the challenges that must be overcome to pave the way for their clinical application.
]. Hypoxia promotes tumour development and progression and is associated with not only the hallmarks of cancer but also the established and new factors that influence response to fractionated radiotherapy (Fig. 1). Oxygen is a potent radiosensitiser due to its high electron affinity and ability to stabilise the free radicals produced when sparsely ionising radiations interact with tissue. Interventions targeting hypoxia can improve the therapeutic ratio of radiotherapy, but to date, none have been adopted globally. The reasons for this failure to change widespread clinical practice are multifactorial but include poor trial designs, limited availability of proven hypoxia-modifying therapeutics without commercial backing, the continually changing background of clinical practice and the lack of appropriately validated biomarkers to select patients who would benefit the most. Many approaches for measuring tumour hypoxia have been and continue to be studied. The rapid advancement in genomic technologies and accessibility of big data repositories paved the way for developing gene signatures that are relatively easy to validate for clinical application in comparison with other approaches. Hypoxia gene expression signatures derived over the last decade provide an opportunity to address past limitations and progress the use of biomarkers of hypoxia for biologically optimising radiotherapy. This overview covers the importance of hypoxia in linking traditional concepts in radiobiology with the hallmarks of cancer, strategies for measuring and modifying the hypoxic tumour microenvironment and the potential role of hypoxia signatures in improving the efficacy of hypoxia modification.
Table 1Summary of oxygen levels in normal tissue, tumours and reference markers.
Fig. 1Hypoxia, the hallmarks of cancer and the foundation principles of fractionated radiotherapy: The development of the radioresistant phenotype. Repair, reactivation of immune response, radiosensitivity, repopulation, reoxygenation and redistribution (collectively known as the 6 Rs of radiobiology) provide the rationale for fractionated radiotherapy shown here linked to Hannahan and Weinberg's hallmarks of cancer [
]. Each hallmark has an association with tumour hypoxia, a condition that is dynamic and fluctuating in the tumour microenvironment. Some links are well established, such as induction of angiogenesis and deregulated cellular energetics. Others are less obvious, such as the association between hypoxia and evading growth suppressors or replicative mortality.
1.1.1 Inducing angiogenesis
Sustained proliferative signalling allows cancer cells to progress through the cell cycle without adhering to the strict regulations and checkpoints enforced in normal cells. The rapidly growing cancer cells quickly exhaust the available oxygen and outgrow the vasculature. Cancer cells must induce angiogenesis and establish new vasculature to supply the nutrients and oxygen needed for tumour growth to exceed 1–2 mm [
]. Hypoxia induces a cell stress response that is mediated by multiple transcription factors, in particular the key heterodimeric hypoxia-inducible factors (HIFs). The transcription factors regulate genes which encode proteins that allow cells to survive in hypoxia. The most important HIF-inducible pro-angiogenic factors are the vascular endothelial growth factor (VEGF) family of proteins that stimulate the growth of endothelial cells and construction of blood vessels in and around the tumour. Angiogenesis in cancers is associated with the formation of substandard vasculature lacking supporting pericytes that is leaky and prone to collapse. This fragility leads to acute and transient hypoxia. VEGF-targeted therapy aims to normalise tumour vasculature to improve perfusion and oxygen and drug delivery. Trials of anti-VEGF agents as monotherapy have been largely negative and a phase III study in combination with radiotherapy in an unselected cohort with glioblastoma multiform demonstrated no benefit [
Angiogenesis provides the vasculature for disseminating cancer cells, which is also promoted by hypoxia via its role in epithelial-to-mesenchymal transition (EMT). EMT is a multiple step process that enables cell motility and migration. Activation of HIF1α promotes EMT by suppressing E-cadherin to promote metastasis through the lysyl oxidase (LOX)-Snail pathway [
]. Hypoxia is immunosuppressive via its promotion of macrophage and regulatory T-cell infiltration and upregulation of programmed death-ligand 1 (PD-L1) on tumour and myeloid derived suppressor cells [
]. The disorganised tumour vasculature does not express proteins that are essential to enable T-cell extravasation from the blood vessels into the tumour [
]. The consequence for radiotherapy is unclear, but it may interrupt the abscopal effect, a rare phenomenon where local radiotherapy is associated with distant tumour regression, which provides a rationale for targeted immunomodulation in combination with radiotherapy [
Local radiotherapy and granulocyte-macrophage colony-stimulating factor to generate abscopal responses in patients with metastatic solid tumours: a proof-of-principle trial.
Hypoxia can induce inflammation, but it can also dampen the inflammatory response in tissue as both states of the tumour microenvironment intersect along several pathways. Pro-inflammatory responses to hypoxia promote tumour progression. Pro- and anti-inflammatory responses also play a role in the development of early and late toxicities of radiotherapy. Under hypoxia, these effects are orchestrated by the transcription factors HIF1 and nuclear factor kappa-light-chain-enhancer of activated B cells (NF-kB) [
]. NF-kB is frequently activated under pro-inflammatory conditions, and HIFs can also induce pro-inflammatory cytokines. There is extensive cross talk between the two signalling pathways in hypoxia-driven inflammation [
]. Hypoxia can instigate an anti-inflammatory response that is similar to the mechanism used to protect normal tissue from an overactive immune response after inflammation [
]. However, hypoxic tumour regions are generally associated with chronic inflammation, which may diminish the phenomena of the radiotherapy bystander effect [
]. Balancing and modulating the opposing pro- and anti-inflammatory mechanisms induced by hypoxia and radiotherapy to enhance treatment response remains an ongoing challenge given that inflammation is an integral component of the cancer hallmarks.
1.1.5 Genomic instability and mutation
Genomic instability is another facilitator of the hallmarks of cancer. Hypoxia-induced inhibition of DNA repair promotes microsatellite, chromosomal and genomic instability [
]. In both in vitro and in vivo models, hyper-mutation rates have been shown under both acute and chronic hypoxia due to decreased mismatch and homologous recombination DNA damage repair [
]. In comparison, radiation-induced genomic instability is promoted by reactive oxygen species and defective cellular responses to DNA double-strand breaks and non-targeted bystander effects [
]. Telomerase is expressed in 85–90% of human tumours but not in most healthy tissues. Pre-clinical studies showed radiosensitisation after inhibition of telomerase [
]. Treatment with MAPK inhibitors alone or by dual targeting with PI3K in combination with radiotherapy is of interest in pre-clinical and early phase studies [
Long-term outcome by risk factors using conformal high-dose-rate brachytherapy (HDR-BT) boost with or without neoadjuvant androgen suppression for localized prostate cancer.
The high rate of glycolysis in tumour cells facilitates DNA repair by rejoining of radiation-induced DNA strand breaks by activating both non-homologous end joining and homologous recombination (HR) pathways [
]. Phosphatidylinositol 3-kinase (PI3K) along with mammalian target of rapamycin complex activation of HIFs drives the high rate of oxygen-independent glycolysis (Warburg effect) [
]. Glycolytic metabolism may impact on radiotherapy in particular in the setting of hypoxia whereby aerobic cells use lactate transported into cells by monocarboxylate transporters for oxidative phosphorylation, whereas hypoxic cells use glucose for growth [
]. Under hypoxia, HIF1α upregulates expression of SLC2A1 (encoding glucose transporter 1, GLUT1) along with glycolytic enzymes to assist with metabolic adaptation of the tumour cell; however, the exact molecular mechanism of this metabolic shift is unclear. In addition, upregulation of carbonic anhydrase 9 by HIF1α under hypoxia promotes the secondary pathway of glycolysis (pentose phosphate pathway) resulting in extracellular matrix acidification and production of reducing species protecting DNA from radiation-induced damage [
]. Therefore, inhibition of glycolysis is another possible mechanism to overcome radiation resistance under hypoxia and lies behind the strategy of reprogramming metabolism by repurposing metformin as a radiosensitiser [
]. Retrospective series of patients with prostate cancer receiving radiotherapy who were on metformin for diabetes mellitus have shown conflicting results [
]. A randomised phase II trial in non-small-cell lung cancer of chemoradiotherapy alone or with metformin showed no benefit in terms of improved progression-free or overall survival [
Initial reporting of NRG-LU001 (NCT02186847), randomized phase II trial of concurrent chemoradiotherapy (CRT) +/- metformin in locally advanced Non-Small Cell Lung Cancer (NSCLC).
1.1.8 Sustaining proliferative signalling, evading growth suppressors and resisting cell death
Evading growth suppressors and—the ultimate hallmark of cancer—resisting cell death involve complex interconnected pathways that occur at the genetic, epigenetic and tumour microenvironment level. Hypoxia enhances resistance to apoptosis through upregulation of key regulators such as MDM2, ARC and PI3K [
The role hypoxia plays in moderating and inhibiting the therapeutic effect of radiotherapy within the conceptual framework of the hallmarks of cancer provides a basis for the discovery of clinically relevant gene signatures and therapeutic targets. The aim is to bridge the gap between the exponential growth of genomic knowledge and the advanced radiotherapy technology resulting from a well-established solid foundation in physics, anatomy and pathophysiology. The genomic revolution is providing an opportunity to realise the vision of personalised and precision radiotherapy through patient selection and treatment optimisation.
1.2 Strategies for targeting hypoxia
There is level 1a evidence that giving hypoxia-targeted treatments with radiotherapy improves locoregional control and survival without compromising late side-effects. A meta-analysis of 10,108 patients recruited into 86 trials showed that hypoxia modification of radiotherapy improved locoregional control (odds ratio [OR]: 0.77; 95% confidence interval (CI): 0.71–0.81; P < 0.001) and overall survival (OR: 0.87; 95% CI: 0.80–0.95; P < 0.01) [
]. The evidence is particularly strong for squamous cell carcinoma of the head and neck where a meta-analysis of 4805 patients recruited into 32 randomised trials favoured hypoxia modification (OR: 0.71; 95% CI: 0.63–0.80, P < 0.001) [
]. These trials involved breathing normobaric or hyperbaric oxygen or the use of hypoxic cell radiosensitisers. Other approaches have also been studied, and the various methods can be categorised broadly into physiological modification by increasing oxygen delivery, oxygen mimetic radiosensitisation, pharmacological modulation, hyperthermia and radiotherapy intensification.
1.2.1 Increasing oxygen delivery
The first hypoxia-targeting attempts in the 1950s used hyperbaric oxygen with the aim of increasing oxygen flow to tumours. Patients breathed 100% hyperbaric oxygen at pressures between 2 and 4 atm. Several multicentre randomised trials were conducted with mixed results. A meta-analysis of 19 trials (n = 2286) demonstrated statistically significant improvements in local control and disease-free survival in head and neck (risk ratio [RR]: 0.83, 95% CI: 0.70 to 0.98, P = 0.03) but not cervical (RR: 0.88, 95% CI: 0.69 to 1.11, P = 0.27) cancer [
]. However, there was evidence for an increased risk of severe toxicity during the course of radiotherapy with hyperbaric oxygen (RR: 2.35, 95% CI: 1.66 to 3.33). Hyperbaric oxygen therapy is unsuitable for routine practice due to logistical issues and patient discomfort over a protracted fractionation schedule. The use of normobaric carbogen (a mixture of 95–98% oxygen and 2–5% carbon dioxide), however, is more feasible clinically. Recent trials gave carbogen with the vitamin B3 analogue nicotinamide. The nicotinic acid amide is incorporated in vivo into nicotinamide adenine dinucleotide and nicotinamide adenine dinucleotide phosphate, coenzymes in redox metabolism. It is a PARP-1 inhibitor, and the radiation sensitising ability is attributed to overcoming perfusion-limited acute hypoxia through inhibiting intermittent vascular constriction [
]. Only one study has compared the new generation PARP-inhibitor (olaparib) with nicotinamide in vivo, which showed a significantly higher nicotinamide dose is needed for the same perfusion effect as olaparib. This would suggest that olaparib is more potent than nicotinamide [
]. In vivo studies with Olaparib and radiotherapy combinations have shown evidence of hypoxia selection and improvement in the setting of breast and lung cancer, and this area of research needs to be progressed [
]. Recent randomised trials showed the efficacy of giving carbogen plus nicotinamide with radiotherapy. A phase III trial in muscle-invasive bladder cancer with carbogen and nicotinamide (CON) reported a 13% (P = 0.04) improvement in overall survival compared with radiotherapy alone with minimal additional toxicity [
]. The phase III trial of accelerated radiotherapy plus CON (ARCON) in cT2-T4 laryngeal cancer (n = 345) showed CON improved 5-year locoregional control (93% versus. 86%, P = 0.04) [
]. The benefit was maintained in patients with pre-treatment anaemia for locoregional control (79% versus 53%, P = 0.03) and disease-free survival (68% versus 45%, P = 0.04). No difference was observed in metastases-free survival between the arms irrespective of haemoglobin status [
]. A recently completed phase II trial of external beam radiotherapy and hypoxia modification with CON (PROCON) in 50 patients with high-risk prostate cancer reported no increased toxicity associated with treatment and similar tolerability was reported in cervical cancer [
]. The systemic toxicity from chemotherapy can be avoided with CON making it a particularly attractive radiosensitising option for older/frail patients with cancer and in low-resource settings.
The concept of increasing oxygen delivery to tumours by increasing haemoglobin levels is intuitive. Approaches include red blood cell transfusion and erythropoietin. Studies showed poor outcomes with low levels of haemoglobin, which was supported by pre-clinical evidence [
]. Clinical trials examining the effect of transfusion, however, have reported mixed results, and the use of erythropoietin (EPO) with radiotherapy was shown to be detrimental in three meta-analyses [
Association of hemoglobin level with survival in cervical carcinoma patients treated with concurrent cisplatin and radiotherapy: a Gynecologic Oncology Group Study.
]. The mechanism of EPO resistance is unclear but is thought to be driven by hypoxia and not anaemia with the upregulation of the EPO receptor under hypoxia in cancer cells, leading to enhanced tumour growth and invasion in the presence of exogenous EPO [
Targeting hypoxic cells with oxygen mimetic drugs is an approach that is of ongoing interest and involved initially repurposed antibiotic drugs from the nitroimidazole family. The agents have high affinity for electrons and work by mimicking the oxygen effect when given with radiation. The first agent studied was metronidazole, followed by misonidazole and then several third-generation nitroimidazoles (e.g. nimorazole, etanidazole). The benefit of nimorazole was shown in a randomised trial in the head and neck of patients for locoregional control (OR: 1.97, 95% CI: 1.33–2.93, P = 0.002) and disease-specific survival (OR: 1.92, 95% CI: 1.30–2.84, P = 0.002), and it has been adopted as standard of care in some Scandinavian countries [
A randomized double-blind phase III study of nimorazole as a hypoxic radiosensitizer of primary radiotherapy in supraglottic larynx and pharynx carcinoma. Results of the Danish Head and Neck Cancer Study (DAHANCA) Protocol 5-85.
]. Ongoing randomised trials are addressing the efficacy of nimorazole added to the current standard of care which uses cisplatin chemoradiotherapy with a phase II study reporting comparable toxicity [
Locally advanced head and neck cancer treated with accelerated radiotherapy, the hypoxic modifier nimorazole and weekly cisplatin. Results from the DAHANCA 18 phase II study.
]. The largest trial involved tirapazamine. A phase III trial of 861 patients with head and neck cancer failed to show a survival benefit for giving tirapazamine with chemoradiotherapy [
Tirapazamine, cisplatin, and radiation versus cisplatin and radiation for advanced squamous cell carcinoma of the head and neck (TROG 02.02, HeadSTART): a phase III trial of the Trans-Tasman Radiation Oncology Group.
]. In 693 protocol-compliant patients who received a minimum of 60 Gy to the tumour volume, there was a borderline improvement in time to locoregional failure for radiotherapy with cisplatin and tirapazamine versus cisplatin (hazard ratio (HR) = 0.74, 95% CI: 0.54 to 1.02, P = 0.067). A phase III study in cervical cancer did not reach target accrual; however, limited analysis showed no difference in the primary end-point of progression-free survival (63% versus 64%, HR = 1.05, 95% CI: 0.75 to 1.47, P = 0.79) [
Phase III randomized trial of weekly cisplatin and irradiation versus cisplatin and tirapazamine and irradiation in stages IB2, IIA, IIB, IIIB, and IVA cervical carcinoma limited to the pelvis: a Gynecologic Oncology Group study.
]. Other bioreductive agents include evofosfamide and banoxantrone, but these have not yet been tested in phase III trials in combination with radiotherapy.
Drugs that decrease oxygen consumption in combination with radiotherapy have also entered early phase trials, namely metformin and atovaquone (a quinone commonly used as an anti-protozoal drug) [
]. Both drugs have shown to reduce cellular respiration in vivo through inhibition of the mitochondrial complexes at pharmacologically achievable concentrations.
1.2.4 Hyperthermia
Another method to radiosensitise hypoxic cells involves hyperthermia at temperatures between 41 and 43 °C for 30–60 min (weekly or biweekly) during radiotherapy [
]. Although the exact mechanism is unclear, hyperthermia is thought to target hypoxic cells by inhibiting radiation-induced DNA repair or indiscriminant cytotoxicity [
]. A meta-analysis of 19 randomised trials involving 1519 patients with oesophageal cancer showed that adding hyperthermia to chemoradiotherapy improved overall survival (OS) (OR: 1.91, 95% CI: 1.27–2.87, P = 0.002) [
]. A network meta-analysis of 13 trials recruiting 1000 patients with cervix cancer reported improved complete response rates with hyperthermia and chemoradiotherapy compared with radiotherapy (OR: 4.52, 95% CI: 1.93–11.78) or chemoradiotherapy (OR: 2.91, 95% CI: 1.97–4.31) alone [
Hyperthermia and radiotherapy with or without chemotherapy in locally advanced cervical cancer: a systematic review with conventional and network meta-analyses.
]. A meta-analysis of 451 patients with head and neck cancer from six trials showed that thermoradiotherapy improved locoregional control rates compared with radiotherapy alone (62.5% versus 39.6%; OR: 2.92, 95% CI: 1.58–5.42, P = 0.001) [
]. A meta-analysis of 23 trials with 2052 patients showed that giving hyperthermia with radiotherapy improved locoregional control in multiple tumour sites with the exception of lung cancer (OR: 1.97, 95% CI: 1.63–2.37, P = 0.48) [
Radiotherapy-based approaches to targeting hypoxia such as ‘dose painting’ or selective radiation dose escalation have been proposed to overcome hypoxia. For radiotherapy, the identification of hypoxic sub-volumes in tumours based on imaging or by dose prescription at a voxel level raises the possibility of dose escalation to these regions and is feasible in pre-clinical studies [
]. The main issue with this method is identifying the hypoxic region to target, adapting to any changes in hypoxia during treatment and the effect of additional concurrent therapies. At present, positron emission tomography (PET) tracers for imaging hypoxia or magnetic resonance imaging (MRI) are under investigation in clinical trials. In ESCALOX, patients with head and neck cancer will be assessed with 18F-fluoromisonidazole (8F-FMISO) before commencing radiotherapy [
Do selective radiation dose escalation and tumour hypoxia status impact the loco-regional tumour control after radio-chemotherapy of head & neck tumours? The ESCALOX protocol.
]. The RETEP5 phase II trial in non-small cell lung cancer (NSCLC) using 8F-FMISO to identify hypoxic regions to dose paint found that high 8F-FMISO uptake was associated with a poor prognosis, which was not reversible with dose painting to the hypoxic regions suggesting the increment in radiation dose achievable in the trial was insufficient to overcome radioresistance [
Phase II study of a radiotherapy total dose increase in hypoxic lesions identified by (18)F-misonidazole PET/CT in patients with non-small cell lung carcinoma (RTEP5 study).
1.3 Approaches for measuring hypoxia in patients with cancer
Although there is level 1a evidence that hypoxia modification of radiotherapy is beneficial, there has been little global impact on clinical practice [
]. Criticism has been levelled at the quality of radiotherapy, patient selection, inadequate follow-up, unselected target population, clinician equipoise and irrelevance in the face of changing standards of care. Despite the many methods and trials that have been developed to target the effects of hypoxia, there has been very little success in translating this to the clinic with the exception of nimorazole in head and neck cancer in Scandinavia and CON in bladder cancer which is a standard of care option in the UK. However, the limited use of CON in the UK in bladder cancer despite a 13% improvement in overall survival highlights the challenges. Another trial in bladder cancer reported at a similar time showed that adding chemotherapy (5-fluorouracil plus mitomycin-C) to radiotherapy also improved overall survival by 13% [
]. It is convenient in oncology departments to administer the chemotherapy rather than sorting out the logistics for giving CON, and so, concurrent chemoradiotherapy is more widely used. The development of hypoxia biomarkers that are suitable for routine clinical use has the potential to realise the therapeutic benefits of hypoxia-modifying agents that have not been widely incorporated into clinics. These approaches are broadly summarised as direct, exogenous, endogenous and radiomic markers.
1.3.1 Direct measurements of oxygen levels
Initial studies measuring oxygenation in human tumours involved inserting large electrodes into cervical cancers in the 1960s [
]. The limitation of using large electrodes, associated with tissue compression and bleeding artefacts, was addressed with the development of fine needle microelectrodes and the Eppendorf pO2 histograph with its automated stepper motor. Eppendorf studies showed that hypoxia is associated with a poor prognosis after radiotherapy in patients with cancers of the head and neck, cervix, prostate and soft tissue sarcoma [
]. While hypoxia measured using oxygen microelectrodes is considered by some to be the gold standard, the technique is limited by accessibility, not only of some tumours but also equipment, and sampling error when measured along a very fine track. Insertion of microelectrodes is also an invasive diagnostic procedure. Electrodes measure interstitial pO2, which might not necessarily reflect intracellular hypoxia, and they do not distinguish hypoxic from necrotic tissue [
]. For these reasons, it can be impossible to validate other hypoxia markers against this gold standard which highlights the need for establishing a new benchmark. Alternative methods for direct measurements of oxygen levels have been explored but not progressed for clinical application in tumours. Several other approaches have been developed with each method having advantages and disadvantages. To date, no method has been adopted into routine clinical practice.
1.3.2 Exogenous hypoxia markers
Bioreductive agents can be used in combination with immunohistochemistry or immunofluorescence to characterise tumour hypoxia. Pimonidazole and pentafluropropyl (EF5) are nitroimidazoles that covalently bind to thiol-containing proteins in hypoxic cells due to the reducing nature of the hypoxic microenvironment [
]. Chemically reduced adducts irreversibly bind to cellular macromolecules and can be detected using antibodies. The in vivo reductive activation occurs in both acute and chronic hypoxic microenvironments but is more sensitive at severe hypoxia compared with microelectrode measurements possibly due to the extent of necrosis present [
]. Both agents are used in animal and experimental medicine studies but arguably have limited applicability in routine clinical practice due to the need for administration intravenously (although a pimonidazole oral formulation is now available) around 16 h before tissue biopsy. The latter is a limitation because most diagnostic biopsies are taken at referral hospitals rather than radiotherapy centres. Both have been assessed in small patient cohorts, and pimonidazole staining was shown to predict benefit from hypoxia modification [
The prognostic value of pimonidazole and tumour pO2 in human cervix carcinomas after radiation therapy: a prospective international multi-center study.
]. The most consistently induced are studied as endogenous markers of hypoxia with the advantage that they are detectable using immunohistochemistry on archival diagnostic tissue and hence can be validated in multiple retrospective cohorts. The most widely studied proteins are HIF-1α and two of its downstream gene products: GLUT-1 and CAIX [
]. A hypoxia marker independent of this pathway that has been studied is osteopontin (a secreted phosphoglycoprotein) with the advantage that it can be measured in tissue and plasma [
Intrinsic markers of tumour hypoxia and angiogenesis in localised prostate cancer and outcome of radical treatment: a retrospective analysis of two randomised radiotherapy trials and one surgical cohort study.
Hypoxia and angiogenic biomarkers in prostate cancer after external beam radiotherapy (EBRT) alone or combined with high-dose-rate brachytherapy boost (HDR-BTb).
]. Endogenous markers have been studied in multiple cancers of which a detailed analysis is outside the scope of this review. Although there are some conflicting data, high tumour marker expression tends to be associated with a poor prognosis. A meta-analysis of CAIX studies in head and neck cancer highlighted the challenges in developing a validated biomarker based on endogenous hypoxia-associated gene expression with issues around reporting of staining localisation, different quantification methods and varying thresholds to stratify patients [
]. The prognostic value of HIF-1α has been analysed in seven meta-analyses in nasopharyngeal (n = 1476), oesophageal (n = 1566), head and neck (n = 1474), lung (n = 2056), prostate (n = 1342), connective tissue (n = 942) and glial (n = 603) tumours [
]. These studies encountered similar limitations related to threshold cut-off, antibody specificity, reproducibility of reporting and interpretation bias. Comparative studies between endogenous and exogenous markers in clinical samples have at best demonstrated weak correlations [
]. While endogenous markers are useful for large retrospective studies, they have limited application in routine clinical practice. The advantages of gene expression signatures over immunohistochemistry are their better reproducibility and elimination of reporting bias [
Imaging methods provide a non-invasive and repeatable measure of tumour hypoxia. They also assess the entire tumour volume and can be used in tumours that are not readily accessible and where obtaining enough biopsy material remains challenging (e.g. pancreas, lung). The most widely studied methods are PET and MRI. The most common PET tracers are nitroimidazole analogues. These drugs undergo different intracellular metabolism depending on the availability of oxygen in tissue. In a hypoxic state, additional reductive processes occur ensuring that the drugs are trapped in the cell although the process is not entirely irreversible as the agents will eventually clear from the cell [
]. The PET radiopharmaceuticals commonly studied are 18F-FMISO and 18F-fluoroazomycin-arabinofuranoside (18F-FAZA). There are more than 100 clinical studies using common PET tracers across a number of tumour types published with a meta-analysis demonstrating a trend towards better outcome in tumours with low tracer activity (OR: 0.25, 95% CI: 0.16–0.39), although significance was not reached in all tumour subtypes [
]. There are a few ongoing challenges with using PET including tracer uptake due to limited perfusion which may complicate interpretation, overlap with uptake in normoxic tissue, reproducibility and appropriate validation methods [
MRI has also been used to assess tumour hypoxia. Most studies used blood oxygen level-dependent (BOLD)in combination with dynamic contrast-enhanced (DCE) imaging [
]. The imaging methods have been around for over a decade yet there is limited validation for these methods as hypoxic biomarkers to inform clinical practice. BOLD MRI measures regional differences in deoxygenated haemoglobin levels. DCE imaging uses a gadolinium contrast agent that tracks through the microvasculature, and the measured signal change over time can be used to fit models and estimate haemodynamic parameters such as perfusion. Newer MRI biomarkers are emerging such as dynamic susceptibility contrast MRI and tumour oxygen level dependent (TOLD) MRI that show promise in pre-clinical experiments, and further studies are required [
]. There is some evidence to suggest BOLD MRI can be used to map chronically hypoxic regions in prostate cancer, and the approach has been validated against pimonidazole staining [
The advantage of imaging approaches is the ability to make whole tumour measurements, use for inaccessible cancer sites and the opportunity for repeat scanning during a course of radiotherapy (raising the possibility of their use as early response biomarkers). There are several obstacles to overcome before imaging biomarkers can be translated into clinical practice including cross-centre standardisation of image acquisition, image analysis and reporting, as well as validation as a predictive biomarker. Combining hypoxia radiomics with transcriptomic data to address the limitations of both methodologies is a growing area of interest [
]. Integration of multi-omic data including correlation with histology and wet lab validation to determine the link between the hypoxic phenotype and predicting treatment outcome however is still in its infancy.
Although the literature is saturated with prognostic hypoxic biomarkers including those in the setting of radiotherapy, very few have been tested for their predictive capability when treating with hypoxia modification in combination with radiotherapy. Table 2 lists the studies that have evaluated the hypoxia markers discussed earlier for their ability to predict benefit from hypoxia modification of radiotherapy within phase III clinical trials. The heterogeneity of outcome is the very justification for seeking a robust and practical biomarker for clinical use to ensure future work focusses on populations with hypoxia and are not diluted by chance or as a property of a particular tumour model with subjects that will never show an altered outcome with hypoxia modification.
Table 2Clinical trials evaluating hypoxia biomarkers (excluding gene signatures) as predictors of from hypoxia targeting.
Cancer
Pts
Method
Marker
Randomisation
Hypoxia criteria
Benefit from hypoxia targeting in hypoxic tumours
P
Pubmed ID
HNSCC∗
578
Endogenous (plasma)
Osteopontin
70 Gy in 35# with CIS/5FU or CIS/TPZ
High >711 ng/mL Middle 407–710 ng/mL Low<407 ng/mL
Pts = number of patients; HNSCC = head and neck squamous cell carcinoma; RT = radiotherapy; CON = carbogen and nicotinamide; HIF-1α = hypoxia-inducible factor -1α; # = number of radiotherapy fractions; MIBC = muscle-invasive bladder cancer; SCC = squamous cell carcinoma; TPZ = tirapazamine; CRT = chemoradiotherapy, HTV = hypoxic tumour volume (at least one voxel with a FAZA signal greater than the calculated threshold); HPV = human papilloma virus; 18F-FMISO PET = [18F]-fluoromisonidazole positron emission tomography scans; HF = hypoxic fraction (the area positive for pimonidazole relative to the total tumour area); ARCON = accelerated radiotherapy with carbogen and nicotinamide); ψ = cohorts have been validated with transcriptomic signatures (Table 3). LRF = locoregional failure; LRFS = local relapse-free survival. H-score: the product of intensity (0–3) and estimated percentage labelling of viable tumour cells in cores (100× magnification), giving a range of 0-300 RC = regional control rate at 5 years.
The biological cellular adaptation and transcriptional response to hypoxia underlies the development of gene expression signatures. There are multiple ways to derive signatures and a number start with data generated in cell lines in hypoxia versus normoxia conditions. The genes that are significantly upregulated or exceed a defined fold change threshold are grouped and curated to form a group of ‘seed’ genes which have recently been streamlined to be tumour specific [
]. Further refining of gene signatures in vivo with exogenous hypoxia tracers in xenografts and clinical correlation with oxygen electrode measurements have also been used [
]. Once a signature is locked, independent cohorts are used for signature validation. There are approximately 34 published hypoxia signatures with three shown to predict benefit from having hypoxia-targeted treatments with radiotherapy [
]. An advantage for validating hypoxia signatures is that hypoxia is an adverse prognostic factor irrespective of treatment. There are far more publicly available surgical series with transcriptomic and linked outcome data but fewer radiotherapy cohorts. The lack of radiotherapy transcriptomic data is one of the limitations when deriving and validating radiotherapy biomarkers. Table 3 lists the published hypoxia gene expression signatures that have incorporated patient samples in their derivation or validation. Some examples are outlined in the following paragraphs.
Table 3Hypoxia gene signatures by tumour site derived or validated using patient samples.
There are currently three hypoxia gene signatures derived specifically for squamous cell carcinoma of the head and neck. One signature was derived using seed genes (10 genes strongly and consistently upregulated by hypoxia in cell lines in the literature) to build co-expression networks in a cohort of head and neck cancer samples profiled using gene expression arrays. The 99-gene signature derived was prognostic not only in an independent head and neck cancer cohort but also in a breast cancer series [
]. The approach was repeated in three head and neck and three breast cancer cohorts to derive hypoxia ‘metagene’ signatures (head and neck, breast, common) which were prognostic across three cancer types (head and neck, breast, lung) [
]. The top 26 genes in the head and neck metagene signature were taken forward onto a platform suitable for clinical application and the ‘reduced’ signature validated in a prospective cohort [
Prospective technical validation and assessment of intra-tumour heterogeneity of a low density array hypoxia gene profile in head and neck squamous cell carcinoma.
A 15-gene classifier was derived initially from hypoxia-induced genes in cell lines. The signature was refined using paired gene expression with 18F-FAZA imaging in xenograft tumours and head and neck clinical samples from lymph node metastasis using oxygen electrode measurements to confirm upregulation and hypoxia specificity in vivo. It was retrospectively validated to be prognostic and predictive in the phase III trial comparing radiotherapy alone versus with nimorazole [
A third signature used 64Cu-ATSM PET/CT to derive a hypoxic volume-associated gene expression signature. The hypoxic volumes correlated significantly with the two previously published signatures. The subsequent 21-gene signature was associated with poor progression-free survival [
A hypoxia signature for bladder cancer was derived using a multi-seed approach and 611 candidate genes curated from the literature. An in silico approach of building a co-expression network involved a bladder cancer training cohort of patients undergoing cystectomy from The Cancer Genome Atlas Program (TCGA). The derived 24-mRNA signature was prognostic in four independent cohorts of patients treated with cystectomy, and the predictive capacity of the signature was confirmed in the phase III trial of radiotherapy alone or with hypoxia modification with CON [
The use of the exogenous hypoxia marker pimonidazole was used to derive a 32-gene prostate hypoxia signature which was analysed against Ki67, a proliferation marker and an independent watchful waiting cohort (GSE16560) demonstrating both high correlation and prognostic significance [
]. A 28-gene signature was derived using the seed gene approach with 848 ‘seeds’ identified as being induced in hypoxia in at least two of four prostate cancer cell lines [
]. A co-expression cluster network was built using the seed genes in the TCGA prostate cancer cohort. The 28-gene signature derived was validated in 10 additional cohorts for prognostic significance. There is no gene overlap between the genes in the two signatures highlighting the large number of hypoxia-induced genes and variability in methodologies. Both signatures may play a role in identifying patients that could benefit from hypoxia-targeting treatment, but the lack of randomised trials in prostate cancer is a limitation.
1.4.5 Sarcoma signatures
Two sarcoma gene signatures have been derived. The first which used a hierarchical clustering method and published hypoxia-related genes in a derivation cohort found that the expression of 26 of 107 hypoxia-associated genes was significantly higher in sarcoma than in normal tissue; however, impact on prognosis was not reported [
]. The most recent signature was derived from seven sarcoma cell lines exposed to hypoxia with 33 upregulated protein coding genes across the cell lines selected for curating in a training cohort. A k-means clustering method was used to establish two clusters based on the phenotype of the expression of the 33 genes. The final 24-gene signature was selected from the cluster with the most significant upregulated seed genes and subsequently validated in two independent cohorts [
]. A derivation cohort of patients had pre-treatment DCE-MRI with gadopentetate dimeglumine contrast, and the uptake of the contrast measured by the relative signal intensity increase as a function of time using the Brix pharmacokinetic model (ABrix) was determined for each tumour voxel. The mean ABrix value over the 20th to 30th percentile interval for that tumour was used for analysis along with matched gene expression data in a subset of the 78 patients. The genes were ranked by their ABrix parameter with significant correlation seen with the hypoxia gene set and the ABrix. The derived 31-gene signature was further refined into a 6-gene signature and its prognostic ability validated in an independent cohort [
Integrative analysis of DCE-MRI and gene expression profiles in construction of a gene classifier for assessment of hypoxia-related risk of chemoradiotherapy failure in cervical cancer.
A recent analysis of eight published hypoxia signatures in 8006 tumours across 19 cancer types independently validated them as consistent pan-cancer hypoxia biomarkers. There was a significant correlation between the hypoxia scores generated using the eight signatures. When analysed using the same hypoxia gene signature and sorted by the median hypoxia score, squamous cell carcinomas of the head and neck, cervix and lung were the most hypoxic, whereas adenocarcinomas of the prostate and thyroid the least hypoxic tumour types. The reported inter-tumour heterogeneity was as pronounced within a tumour type (accounting for 42% of the variance in scores) than between them based on each of the hypoxia signatures examined [
Validation is complicated by cohort heterogeneity, tissue preservation (formalin, rapid freezing, RNAlater), platform for generating transcriptome data (RNAseq, multiple microarrays, quantitative polymerase chain reaction (qPCR)), age of samples and methods used for nucleic acid extraction, quantitation and assessment of quality control [
]. The latter technical variables can affect multiplex sequencing and must be controlled or accounted for. Validation typically involves retrospective cohorts with full transcriptomic data, which might have limited clinical utility at present. Gene signatures require validation on prospectively collected tissue using an assay and cut-off/dichotomisation suitable for clinical application. Once an assay/platform is selected, there is usually a need to identify tumour site–specific endogenous control genes, select reliable reference samples and design target gene primers suitable for formalin-fixed paraffin-embedded tissue tumour biopsies. Then, there are different methods used in the derivation of a signature to summarise expression levels, and it is challenging to validate a specific cut-off across platforms. Cohort medians and quartiles are often used but do not reflect the underlying continuous variable. Clustering methods such as Κ-means algorithm to determine centroid-based classification are also commonly used. An additional issue around validating hypoxia gene expression signatures is that there is no gold standard comparator.
The issue of intra-tumour heterogeneity and determining treatment allocation based on a single biopsy from an individual's tumour might be an obstacle in clinical application and is an impediment for any tissue-based biomarker. There is a high level of hypoxia heterogeneity across solid tumours as demonstrated by oxygen electrode measurements and their correlation with immunohistochemistry markers of hypoxia with sampling bias contributing to poor biomarker validation [
]. Sampling error, tissue volume, total percentage tumour and histological parameters such as tumour necrosis may hinder the accuracy of the assay. To overcome these issues, it has been suggested that multiple biopsies or image-guided targeted biopsies should be incorporated into the qualification of the assay with an estimation of variance on 30–50 patients as a minimal requirement for validation [
]. The first studies of intra-tumour variation of hypoxia measured by a gene signature showed a lower level of intra-tumour variation from a gene signature than with single measures of hypoxia [
Prospective technical validation and assessment of intra-tumour heterogeneity of a low density array hypoxia gene profile in head and neck squamous cell carcinoma.
An additional advantage when generating hypoxia signatures over other approaches for measuring hypoxia is the potential for finding new therapeutic targets or opportunities for drug repurposing as hypoxic radiosensitisers or to individualise hypoxia-targeted therapies. There are several drug and transcriptomic connectivity mapping software packages available that may aid in identifying common pathways and their molecular targets [
]. Gene expression signatures also provide insight into the underlying pathophysiology of hypoxia and the disease processes as new avenues to explore mechanisms of disease pathogenesis and response to radiotherapy. Table 3 lists the common genes across the published hypoxia signatures where signature derivation and validation included patient cohorts. Gene frequency analysis across the published hypoxia signatures showed that no gene was common to all; however, within each tumour site where more than one hypoxia-associated gene expression signature is available, there are genes that are common such as LOX, PFKFB3, ALDOA, KCTD11, P4HA1 and SLC2A1 in head and neck cancer. Reliability in published signatures can be an issue given there is no standardisation on reporting with some publications using different nomenclature for gene symbols and not cross-referencing secondary annotations such as Affymetrix probe set or Entrez ID.
1.5 The roadmap to clinical application
Despite many approaches developed to identify and target hypoxia over 50 years, there has been very little success in translating the research into the clinic. Fig. 2 is a roadmap for the steps required to move from discovery to clinical application. Hypoxia has been a cornerstone of radiotherapy-related research, yet there is only one phase III trial targeting hypoxia modification with full transcriptomic data available – BCON. The recent NIMRAD trial that randomised patients with head and neck cancer to radiotherapy alone or with nimorazole, which completed recruitment in May 2019, provides a further opportunity as transcriptomic data will be generated [
]. The development of new multi-omic based hypoxic biomarkers that are suitable for routine clinical use has the potential to realise the therapeutic benefits of hypoxia-modifying agents that have not been widely incorporated into practice despite efficacy being established. There are several roadblocks to the widespread clinical implementation of any genomic biomarker that emerge throughout the pipeline from discovery to economic justification.
Fig. 2Bench to bedside pipeline for hypoxia gene expression signatures. The six stages of translating a companion hypoxia gene expression signature into clinical practice. The laboratory phase (the first two stages) consists of using pre-existing knowledge of known hypoxia-associated genes in combination with in vitro and in silico methodologies to derive and technically validate a signature using tumour-specific cell samples and human tumour samples (fresh frozen versus FFPE). The initial steps test the robustness and reproducibility of the biomarker and ideally ensuring that it is assay platform agnostic. The clinical phase consists of establishing the prognostic and predictive value of the signature and ideally qualification in a phase III trial (using either biomarker-strategy or biomarker by treatment-interaction stratified trial design) with incremental value demonstrated by comparison with established hypoxia biomarkers and cost effectiveness analysis. RCT = randomised controlled trial; CEA = cost effectiveness analysis; FFPE = formalin-fixed paraffin-embedded tissue; EBRT = external beam radiotherapy; R = randomise; CON = carbogen and nicotinamide; HIF = hypoxia-inducible factors; HS = hypoxia score; DSS = disease-specific survival; mRNA = messenger ribonucleic acid.
Evidence around evaluating economic efficiency of genomic biomarkers is lacking, and their potential value may not be captured using generic measures such as quality-adjusted life years but do allow for comparisons across diagnostics [
]. Incorporating economic modelling at an early stage of biomarker development may be advantageous in making an early decision on likely clinical utility. The latter should underpin go/no-go decisions on the further substantial investment into crossing the current barriers to clinical application. Commercialisation is likely to be key to implementation, and the community needs to address pathways for commercialisation that include the need for high-quality documentation of experiments and protection of intellectual property. Evidence outside radiation oncology shows signatures can be translated for patient selection and are more reliable than immunohistochemistry which requires international consensus on methodology and reporting [
]. Cost effectiveness studies have demonstrated the benefit of using gene signatures to guide the use of adjuvant chemotherapy in breast cancer (EndoPredict, Oncotype DX and Prosigna) when used in conjunction with validated clinical nomograms and have facilitated widespread clinical application [
The potential to incorporate hypoxia gene signatures into the management of patients undergoing radiotherapy provides a fresh look at treatment paradigms that have been well researched but poorly implemented. There is also increasing interest in biomarker-stratified radiotherapy trials including those using gene expression signatures to de-escalate treatment in ‘biomarker low-risk’ patients [
]. A number of clinical trial designs incorporating companion predictive biomarkers for targeted drugs in oncology have been proposed which are applicable to radiotherapy trials involving hypoxia modification [
Given the importance of validation but limitation of available data sets, there is a need for the radiation oncology community to pool resources and standardise practices to build multiple accessible cohorts for testing signatures and comparing with other measures of hypoxia – much similar to the rich repository the TCGA has provided. These cohorts need to be from the main tumour sub-sites where radiotherapy is a definitive treatment option, with biopsies throughout the course of treatment or before and after resection in the case where neoadjuvant radiotherapy is indicated.
2. Summary
Hypoxia is associated with all the hallmarks of cancer, tumour development and progression and a poor prognosis. There is a high level of evidence that targeting the hypoxic microenvironment improves radiotherapy outcomes. We know patients with the most hypoxic tumours benefit most from hypoxia-targeted treatments, but, despite the many approaches developed, no biomarker has progressed to clinical application. The genomic revolution over the last two decades resulted in a surge in oncology biomarkers and a move towards a precision medicine approach in trials and routine clinical practice. There are no biomarkers in routine use for patients undergoing radiotherapy to aid management decisions, and a roadmap is needed to ensure consistency and provide a benchmark for progression to application. Gene signatures are used in oncology and show the most promise as hypoxia biomarkers. This decade we should either achieve the goal of personalised hypoxia targeting with radiotherapy or accept that overcoming the hypoxia problem is forever lost in translation.
3. Statement of search strategies
A literature search was used to examine relevant English language publications from PubMed supplemented by hand searching of abstracts from recent international meetings. Key words used alone and in combination include ‘hypoxia’, ‘radiotherapy’ and ‘biomarkers’ from 1 January 1990 to 1 January 2020. Further studies were identified by examining the reference lists of all included articles. Those discussed in this article were manually chosen at the discretion of the authors.
Funding
The authors are supported by the NIHR Manchester Biomedical Research Centre, Cancer Research UK (C147/A25254), Prostate Cancer UK (PG14-008-TR2) and the Movember Foundation as part of the Belfast–Manchester Centre of Excellence (CEO13-2-004).
Conflict of interest statement
Authors declare no conflict of interest.
References
Hammond E.M.
Asselin M.C.
Forster D.
O'Connor J.P.
Senra J.M.
Williams K.J.
The meaning, measurement and modification of hypoxia in the laboratory and the clinic.
Local radiotherapy and granulocyte-macrophage colony-stimulating factor to generate abscopal responses in patients with metastatic solid tumours: a proof-of-principle trial.
Long-term outcome by risk factors using conformal high-dose-rate brachytherapy (HDR-BT) boost with or without neoadjuvant androgen suppression for localized prostate cancer.
Initial reporting of NRG-LU001 (NCT02186847), randomized phase II trial of concurrent chemoradiotherapy (CRT) +/- metformin in locally advanced Non-Small Cell Lung Cancer (NSCLC).
Association of hemoglobin level with survival in cervical carcinoma patients treated with concurrent cisplatin and radiotherapy: a Gynecologic Oncology Group Study.
A randomized double-blind phase III study of nimorazole as a hypoxic radiosensitizer of primary radiotherapy in supraglottic larynx and pharynx carcinoma. Results of the Danish Head and Neck Cancer Study (DAHANCA) Protocol 5-85.
Locally advanced head and neck cancer treated with accelerated radiotherapy, the hypoxic modifier nimorazole and weekly cisplatin. Results from the DAHANCA 18 phase II study.
Tirapazamine, cisplatin, and radiation versus cisplatin and radiation for advanced squamous cell carcinoma of the head and neck (TROG 02.02, HeadSTART): a phase III trial of the Trans-Tasman Radiation Oncology Group.
Phase III randomized trial of weekly cisplatin and irradiation versus cisplatin and tirapazamine and irradiation in stages IB2, IIA, IIB, IIIB, and IVA cervical carcinoma limited to the pelvis: a Gynecologic Oncology Group study.
Hyperthermia and radiotherapy with or without chemotherapy in locally advanced cervical cancer: a systematic review with conventional and network meta-analyses.
Do selective radiation dose escalation and tumour hypoxia status impact the loco-regional tumour control after radio-chemotherapy of head & neck tumours? The ESCALOX protocol.
Phase II study of a radiotherapy total dose increase in hypoxic lesions identified by (18)F-misonidazole PET/CT in patients with non-small cell lung carcinoma (RTEP5 study).
The prognostic value of pimonidazole and tumour pO2 in human cervix carcinomas after radiation therapy: a prospective international multi-center study.
Intrinsic markers of tumour hypoxia and angiogenesis in localised prostate cancer and outcome of radical treatment: a retrospective analysis of two randomised radiotherapy trials and one surgical cohort study.
Hypoxia and angiogenic biomarkers in prostate cancer after external beam radiotherapy (EBRT) alone or combined with high-dose-rate brachytherapy boost (HDR-BTb).
Prospective technical validation and assessment of intra-tumour heterogeneity of a low density array hypoxia gene profile in head and neck squamous cell carcinoma.
Integrative analysis of DCE-MRI and gene expression profiles in construction of a gene classifier for assessment of hypoxia-related risk of chemoradiotherapy failure in cervical cancer.
= proliferative signalling; 
= evading cell death; 
= genome instability; 
= tumour promoting inflammation; 
= inducing angiogenesis; 
= avoiding immune destruction; 
= activating invasion & metastasis; 
= evading growth suppressors; 
=deregulating energy metabolism; 
= enabling replicative immortality00075-7&id=gr1.jpg){kind=link}
00075-7&id=gr2.jpg){kind=link}
