Measurement of tumor hypoxia is required for the diagnosis of tumor and the evaluation of therapeutic outcome. Currently, invasive and noninvasive techniques being exploited for tumor hypoxia measurement include polarographic needle electrodes, immunohistochemical (IHC) staining, magnetic resonance imaging (MRI), radionuclide imaging (positron emission tomography [PET] and single-photon emission computed tomography [SPECT]), optical imaging (bioluminescence and fluorescence), and hypoxia imaging endoscopy. This review provides a summary of the modalities available for assessment of tissue oxygenation as well as a discussion of current arguments for and against each modality, with a particular focus on noninvasive hypoxia imaging with emerging agents and new imaging technologies intended to detect molecular events associated with tumor hypoxia.
- Hypoxia imaging endoscopy
- innovation of endoscopy
In the 1950s, hypoxia research began, and many clinical trials have been reported. Hypoxia of tumor affects outcomes after radiotherapy. But hypoxia has also been shown to be a poor prognostic factor after chemotherapy and surgery. These findings are attributed to chronic hypoxia. Hypoxic tumors are more likely to recur loco-regionally than well-oxygenated tumors regardless of whether surgery or radiation therapy is the primary local treatment. However, the common oxygen measurement used in these reports was polarographic needle electrodes inserted directly into specific sections of tumor tissue. In this method, hypoxia was measured in only pinpointed area for the tumor. In other words, there was no modality used in which the hypoxia imaging results were visible in real time and which reflected the hypoxic state in the whole tumor. Therefore, hypoxia imaging is expected to allow direct visualization of the biological and functional changes in cancer.
Hypoxia is a histopathological condition in which cells in tissues suffer from lack of oxygen for their normal metabolism. An oxygen saturation (StO2) of arterial blood is almost 100% and that of venous blood is approximately 70%. In contrast, the StO2 in half of cancers is 50–60% at the highest. Hypoxia takes hold as a tumor becomes large enough to disrupt the balance of oxygen supply and consumption in the area. Approximately, 50–60% of advanced cancer forming solid tumor may show hypoxic and/or anoxic conditions exhibiting heterogeneous distribution in the inside of tumor . Hypoxia proliferates rapidly in solid tumors, and their intratumoral vessels with significantly structural abnormalities are distributed spatially with dilated, tortuous, saccular, and heterogeneous figures. As a result, this distribution leads to perfusion-limited delivery of O2 . There are mainly two types of hypoxia regarding solid tumor and tissue around tumor. One is perfusion-limited O2 delivery type, the so-called acute hypoxia, which leads to ischemic condition, however, it is often transient. Another type is diffusion-limited hypoxia, the so-called chronic hypoxia, which can also be caused by an increase in diffusion distances, so that cells far away (>70 μm) from a nutritive blood vessel receive less oxygen (and nutrients) than needed . Regarding hypoxia-induced proteome and/or genome changes, cell cycle arrest, differentiation, apoptosis, and necrosis are found in solid tumor. In contrast, hypoxia-induced changes of proteome may progress tumor growth because of mechanisms enabling cells to overcome nutritive deprivation, to escape from the hostile environment and to favor unrestricted growth. Furthermore, continuous hypoxia can also bring cellular changes as a more aggressive phenotype . Since the presence of hypoxic status in solid tumors was first reported in 1953 to be among the factors associated with treatment failure following radiation therapy , tumor hypoxia has drawn attention as a pivotal event in tumor invasion, angiogenesis, apoptosis, metastasis , resistance to chemotherapy , surgery, and resistance to radiotherapy . In tumor diagnosis and treatment planning, it is crucial to have a grasp of the degree and extent of tumor hypoxia involved prior to the start of treatment.
2. Clinical importance for measurement of tumor hypoxic state
A variety of techniques are being proposed to assess tumor hypoxia, which can be broadly categorized into direct measurements and indirect measurements according to different principles and the ability to quantify tissue oxygenation. Direct measurements, including polarographic needle electrode, phosphorescence imaging, near-infrared spectroscopy (NIRS), blood oxygen level dependent (BOLD) and 19F magnetic resonance imaging (MRI) and electron paramagnetic resonance (EPR) imaging, can detect oxygen partial pressure (pO2), oxygen concentration, or oxygen percentage. Recently, a hypoxia imaging endoscopy that can derive the oxygen saturation (StO2) was developed in endoscopic fields. Indirect measurements, including measuring exogenous and endogenous hypoxia markers, can provide parameters related to oxygenation.
Many clinical trials have been performed using direct and indirect measurement methods. It is now known that hypoxia affects outcome after radiotherapy, with poor prognosis in hypoxic cancers. Next, hypoxia has also been shown to be a poor prognostic factor after chemotherapy and surgery. Furthermore, hypoxic tumors are more likely to recur loco-regionally than well-oxygenated tumors regardless of whether surgery or radiation therapy was the primary local treatment.
3. Direct measurements for hypoxia
3.1. Polarographic needle electrodes for direct tumor tissue
The invasive polarographic needle electrodes have been widely employed since the 1990s to assess tumor oxygen status and to measure pO2 in both human and animal studies [7, 8]. As the gold standard modality, their use has been extended not only to lymph node metastases but to more accessible tumors, which include head and neck cancer, cervical cancer, soft tissue sarcomas of the extremities, astrocytic brain tumors, lung cancer, pancreatic cancer, prostate cancer, and lymph node metastases [8–12]. With the average median pO2 before treatment of 11.2 mmHg (range 0.4–60 mmHg) , these measured values help prediction of the tumor response to treatment  and tumor metastatic potential . The polarographic needle electrode is currently available under CT guidance for evaluating tumor pO2 in deep-seated organs as well as for assessing overall tumor oxygen status  with the caveat, however, that insertion of an electrode into the tumor leads to disruption of tissues, thus rendering it difficult to distinguish the necrotic areas and to establish the patterns of hypoxia involved. Furthermore, the use of the modality not only calls for great expertise but is associated with large interobserver variability.
4. Noninvasive imaging of hypoxia
While the polarographic needle electrode and immunohistochemical (IHC) staining can provide a relatively accurate estimation of tumor oxygenation, being subject to selection bias, provide only a partial, but not complete, picture of the entire tumor site . This has led to an increasing interest in the use of noninvasive functional and molecular imaging modalities, which is capable of yielding a large amount of high-quality experimental data per protocol by increasing the number of quantitative data collections and by guiding tissue sampling and allowing a rapid and effective combination of analyses to be conducted .
Several imaging modalities have been developed, to date, to allow direct or indirect measurement of tumor oxygenation, with a few of these remaining less mature for clinical application. Of these, EPR spectroscopy, which involves the use of unpaired electron species to obtain images and spectra, is currently being explored in animals as a means to provide a quantitative measure of tissue oxygenation . Although this modality has considerable potential to be developed as a tumor oximeter, i.e., in monitoring changes after tumor oxygenation , a suitable paramagnetic marker with low toxicity for human remains yet to become available. The need for appropriate EPR instrumentation in the clinical setting also prevents this promising modality from becoming widespread . Photoacoustic tomography (PAT) is also available for imaging blood oxygenation using the differential optical contrast between O2Hb and dHb. PAT has been implemented for imaging cerebral blood oxygenation of rats
4.1. Magnetic resonance imaging
BOLD-MRI is shown to have potential as a diagnostic modality for tumor hypoxia . Hemoglobin occurs as deoxyhemoglobin in oxygen-deficient states, where not oxyhemoglobin but paramagnetic deoxyhemoglobin can increase the transverse relaxation of the surrounding protons . BOLD-MRI employs deoxyhemoglobin-derived endogenous signals as image contrast to depict changes in oxygenation in blood. Decreased oxygenation in blood results in decreased signal intensity in T2-weighted images, and this correlation between the BOLD-MRI signal and vascular oxygenation allows pO2 to be directly estimated. This has indeed led to numerous studies being conducted to investigate carbogen breathing in mice, oxygenation in tumor models , and kidney function in patients [22, 25, 26] using the modality as a noninvasive technique with high spatial and temporal resolution . As with phosphorescence and near-infrared fluorescence imaging, the major disadvantage of BOLD-MRI is that it reflects change in oxygen tension in vasculature but not those in tissues. Again, not being a quantitative method, it may easily be affected by multiple factors such as flow effects, hematocrit, pH, and temperature .
19F MRI involves the use of two types of markers, i.e., perfluorocarbons (PFCs) and fluorinated nitroimidazoles as contrast agents, which are not used in conventional T1-weighted MRI. While being highly hydrophobic, PFCs are highly oxygen soluble . Due to the linear relationship between the 19F spin lattice relaxation rate of PFCs and the dissolved oxygen concentration, the 19F-based oximetry allows vascular oxygenation to be measured
“Vessel architectural imaging” (VAI) has recently been proposed as a new paradigm in MRI providing a basis for vessel caliber estimation  by incorporating an overlooked temporal shift in the MR signal, thus generating, unlike any other noninvasive imaging modality, new information on vessel type and function. Indeed, this new modality allowed an oral pan-vascular endothelial growth factor (pan-VEGF) receptor kinase inhibitor to be evaluated for its therapeutic efficacy in glioblastoma patients , demonstrating using VAI that anti-VEGF therapy not only normalizes tumor vasculature and alleviates edema but also prolongs survival in these patients.
4.2. Positron emission tomography
Efforts have recently been directed toward developing contrast agents for noninvasive hypoxia imaging with positron emission tomography (PET) and single-photon emission computed tomography (SPECT). Organic molecular markers labeled with positron-emitting radioisotopes are employed in PET imaging to allow the extent of tumor hypoxia to be measured. Commonly used radioisotopes include 18F, 124I, and 60/64Cu and the molecular markers to be labeled with these isotopes include 2-nitroimidazoles, e.g., fluoromisonidazole (FMISO), EF5, and fluoroetanidazole (FETA), nucleoside conjugates, e.g., iodoazomycin arabinoside (IAZA), and Cu(II)-diacetyl-bis (N4-methylthiosemicarbazone) (Cu-ATSM) [38–40]. These markers are shown not only to bind maximally to severely hypoxic cells to form such stable adducts as are detectable with a PET scanner but to provide a clear demarcation of hypoxic cells
Of the first-generation nitroimidazoles, 18F-labeled misonidazole (18F-FMISO) is the most commonly used as being sensitive only to the presence of hypoxia in viable cells . It is reported that a hypoxic state defined as <10 mmHg is required to induce significant 18FMISO uptake . 18F-FMISO uptake is shown to vary widely depending on the type of patients and tumors, whereas 18 F-FMISO is shown to allow hypoxia to be detected in various tumors such as glioma, head and neck cancer, renal tumor, and non-small cell lung cancer [42–44]. A clinical trial of glioblastoma multiforme patients  demonstrated increased 18F-FMISO uptake and retention on both post-treatment FMISO and FDG images, suggesting that reoxygenation did not take place. It is reported that the distribution of oxygen and hypoxia was increased and decreased, respectively, in non-small cell lung carcinomas following treatment, as assessed by sequential FMISO imaging . Given that no correlation is shown between patient diagnosis and degree of decrease in FMISO uptake and retention, in selectively boosting the radiation dose to hypoxic subvolumes, there appears to be a larger role for serial imaging during treatment than for baseline volume measurement. Again, pretreatment FMISO uptake/retention and survival has been shown to be correlated and allow treatment failure to be predicted [45, 47]. However, 18F-FMISO may not be readily available for use in other cancers [42, 48].
The second-generation nitroimidazoles include 18F-fluorerythronitroimidazole (FETNIM) [49, 50], FETA , and EF5 [52, 53], which are more water soluble and not readily susceptible to degradation by most oxidizing mechanisms in place in humans. 18F-EF5 was tested in clinical trials for its feasibility as an imaging agent for hypoxia  and was shown to be hypoxia-specific, with its increased uptake shown to be correlated with the extent of tumor and high risk of metastasis in cancer patients , suggesting its usefulness in identifying high-risk candidates for clinical trials evaluating the influence of early chemotherapy on the occurrence of metastasis . 18F-FAZA has great promise as an imaging agent for tumor hypoxia due to its faster diffusion into cells and faster clearance from normal tissues than 18F-FMISO . PET imaging using 18F-FMISO demonstrated very high tracer uptake in all seven patients with high-grade gliomas evaluated, showing the potential of 18F-FMISO as an imaging agent in assessing hypoxia in this tumor type .
4.3. Phosphorescence imaging
Phosphorescence imaging with injection of porphyrin complex (Oxyphor) into the vasculature also allows tumor vascular pO2 to be measured [58, 59]. Recently, a general approach has been proposed through which to construct phosphorescent nanosensors with tunable spectral characteristics, varying degrees of quenching, and a high oxygen selectivity . These probes are shown to exhibit excellent performance in measuring vascular pO2 in the rat brain with
4.4. Visible light spectroscopy
In the search for noninvasive, continuous modalities for monitoring ischemia, electrical bioimpedance cardiac output monitoring has been proposed but shown to be incompatible with the thermodilution methods [65, 66]. Again, while near-infrared spectroscopy (NIRS)  is shown to respond to both hypoxemia [68, 69] and ischemia [70–72], its clinical use has been limited to large organs, such as the brain [73, 74, 85–87] with its broad normal ranges reported to be between 48% and 88% [75, 76]. Similarly, wide normal ranges are reported for sublingual capnography [77–79]. Also available, albeit invasive are polarographic oximetry probes  and fiber-enabled pulmonary catheters.
Visible light spectroscopy (VLS) appears to be similar to NIRS on some counts  with its mean VLS StO2 shown to be not significantly different from NIRS StO2 reported in human studies [67–76]. Again, the fractional contribution of venous blood to the cerebral NIRS signal has been reported to be 0.84 ± 0.21 ranging from 0.60 to 1.00 [82–84]. Using central venous and pulse oximetry saturation as estimates for local venous and arterial saturation, it is shown to be not significantly different at 0.89 ± 0.04. It is suggested that the two modalities cover similar microvascular compartments.
At the same time, VLS is shown to be superior to NIRS in monitoring tissues that lend themselves to monitoring, thus suggesting a more versatile role for VLS in patient treatment . The NIRS light sources and detectors require to be spaced 2–5 cm apart or more to illuminate and monitor a large, homogenous tissue volume (>30 ml), thus making NIRS with its long and bulky sensors unsuitable for monitoring tissue regions, e.g., thin tissues such as gastrointestinal mucosa or small tumors. In contrast, the visible light used in VLS is shown to be strongly absorbed by tissue and VLS measurement to be highly localized thus making VLS unsuitable for transcranial use or use over thick skin dominated by surface tissue properties. Using VLS, a rapid real-time drop in tumor oxygenation was detected during local ischemia following clamping or epinephrine administration , with the tissue oximetry performed during endoscopy demonstrating a significantly lower tissue oxygenation (StO2) in tumors (46% ± 22%) than in normal mucosa (72% ± 4%) (
5. Hypoxia imaging endoscopy with no phosphor
Kaneko et al.  reported hypoxia imaging endoscopy equipped with a laser light source. In this system, signals from the laser light passed through the processor were calculated as StO2. The measurement range of StO2 was from 0% to 100% in contactless of tumor or normal mucosa under endoscopic observation. Display imaging was performed with the use of laser light alone without phosphor, provided a display of overlay and pseudocolor images. The laser light used was not near-infrared but ranged within visible light wavelengths. In principle, this utilized the difference in absorption coefficient between oxyhemoglobin and deoxyhemoglobin. Two challenges were identified, however, in deriving the StO2 of tissue in alimentary tracts from differences in absorption spectra between oxyhemoglobin and deoxyhemoglobin using small numbers of wavelengths. First, there is not only a small difference in optical absorption spectra in the visible light region but also a narrow bandwidth between isosbestic points. Second, the reflectance of a tissue depends on hematocrit (Hct) as well as StO2, given that light absorption increases as hemoglobin density increases.
An imaging system equipped with laser diodes of 445 and 473 nm and a white fluorescent pigment body was therefore developed. Hypoxia imaging with this system rendered visible an alimentary tract tumor in real time and allowed the whole tumor to be visualized. With the tumor surface and normal mucosa rendered visible, no heterogeneity was seen with the use of this system. In the first-in-human clinical trial, early cancers of the esophagus, stomach, and colorectum were detected as hypoxic areas (Figure 2). Furthermore, colorectal adenomas with histologically low-grade atypia were also detected as hypoxic areas and no complications were reported in the patients with visualization of these tumors in real-time hypoxia imaging which involved only laser light without injection or oral administration of phosphor. As mentioned above, it will be expected that the hypoxia imaging endoscopy is shown to be superior to VLS or NIRS in measuring StO2 of surface of tumor and normal mucosa.
6. Indirect hypoxia evaluation
Proteins and genes whose expression is associated with hypoxia have potential as endogenous molecular markers of hypoxia and have been explored over the years; meanwhile, hypoxia-specific agents have also been explored and shown to be useful in monitoring hypoxia . Immunohistochemical (IHC) staining for hypoxia marker adducts in situ is also available to provide indirect quantitative information on the relative oxygenation of tissue at a cellular resolution. IHC approaches have a role to play particularly
6.1. Hypoxia-inducible factor 1
Optical imaging has had an important role to play in evaluating hypoxia, especially in biopsy specimens. With the introduction of transgenes with the hypoxia responsive element as promoter sequences coupled to reporter genes, e.g., luciferase reporter gene [90, 91] or green fluorescent protein (GFP) , a number of modalities have been developed to allow HIF-1 activity to be directly measured. Of these, a HIF-1-dependent promoter-regulated luciferase reporter gene, shown to produce a 100-fold increased luciferase response to hypoxia, has been used to evaluate anti-hypoxia therapy for its efficacy in animals . Again, an imaging probe has been developed for HIF-1-active cells using a PTD-ODD fusion protein. given that, being involved in the same ODD control as HIF-1α, PTD-ODD fusion proteins are thought likely to be co-localized with HIF-1α [93–96]. First developed as a model probe, PTD-ODD-enhanced GFP-labeled with near-infrared fluorescent dye Cy5.5 was shown to permeate cell membrane with high efficiency, with its stability controlled in an oxygen concentration-dependent manner; to accumulate in hypoxic tumor cells with HIF-1 activity, thus allowing the hypoxic tumor cells with HIF-1 activity to be imaged in contrast to the surrounding cells under aerobic conditions . Bioluminescence imaging has also been used to noninvasively depict HIF-1α as it is upregulated
6.2. Carbonic anhydrase IX
Downstream of HIF-1, carbonic anhydrase 9 (CA IX), a member of the CA family known to exist in cytosolic, membrane-associated, mitochondrial, and secreted carbonic anhydrases (CAs), may represent an alternative target . A membrane-associated enzyme involved in the respiratory gas exchange and acid-base balance, CA IX is shown to be found less abundantly in normal tissue and only in gastric mucosa, small intestine, and muscle. Under hypoxic conditions, CA IX is shown to be overexpressed in different types of cancer , with the staining pattern shown to be more generalized in VHL-associated tumors and focal-perinecrotic in non-VHL-associated tumors .
CA IX has been imaged with fluorescent-labeled sulfonamides in a tumor xenograft model to allow hypoxic and (re)-oxygenated cells to be distinguished , which demonstrated that CA IX required exposure to hypoxia for its binding and retention—a finding confirmed by an
Furthermore, hypoxia markers have been identified and shown to be induced by hypoxia and expressed in human tumors, including VEGF and GLUTs, both of which are upregulated by increased activity of HIF-1 under hypoxic conditions . Imaging strategies targeting these proteins have also been explored for their ability to assess tumor vasculature and proliferation, while the relationship between pO2 values and protein expression levels remains unclear .
7. Heterogeneity of tumor
Tissue oxygenation is shown to be highly heterogeneous due to the presence of both highly oxygenated arterial vascular regions and poorly oxygenated tissues and cells. Spatial and temporal heterogeneity also contribute to the complexity of the issue. Heterogeneity is thus a major factor in hypoxia measurement that affects our ability to stratify patients and predict outcomes using the imaging technologies available, and its biological implications need to be further explored, and effective approaches to assessing heterogeneity remain to be established. Hypoxia imaging endoscopy allowed early cancers of the pharynx, esophagus, stomach, and colorectum to be captured in whole for the first time , with no heterogeneity found in nearly all early cancers or colorectal neoplasia detected. Given that tissue heterogeneity may vary between early, advanced, and metastatic tumors, however, it remains crucial to elucidate tissue heterogeneity as it is associated with tumor progression.
8. Future of hypoxia measuring methods
Given the wide variety of techniques available for assessing hypoxia, e.g., polarographic needle electrodes, IHC staining, PET, MRI, optical imaging with NIR fluorescence or bioluminescence, visible light spectroscopy, and hypoxia imaging endoscopy, it remains critically important to determine their relative advantages and disadvantages for clinical application. Improvements in hypoxia measuring techniques will hinge primarily on which techniques are chosen and how these techniques are applied in the clinic. Clearly, the best of these are expected to be sensitive to the biological sequel of hypoxia, and the ideal one expected to be clinically safe, readily available, minimally invasive, and free from radiation exposure, while at the same time providing high resolution and ease of use. In addition, NIR over 1000 nm wavelength, the so-called biological window, will be promising, because this wavelength area is good for tissue permeability due to reducing both light scattering and infrared absorption .
In the endoscopic fields of alimentary tracts, the existing diagnosis for neoplasia is based on the morphologic features of the tumor. However, imaging of a tumor focused on its function or metabolism yields a novel set of data. Hypoxia imaging endoscope system equipped with a laser source allows oxygen saturation to be shown with two types of overlay and pseudocolor images displayed one on top of the other . Available for handling similarly to conventional endoscopy, this modality is easy to treat with and completely safe without being invasive. Of the large number of patients with cancers in the alimentary tract, such as oral cavity, esophagus, stomach, and colorectum in the world, a majority with advanced cancer patients receives chemotherapy, radiotherapy, and combination therapy. In this regard, this modality is expected to allow not only hypoxic states but also hyperoxic states of tumor to be detected in these patients, thus contributing to selection of therapy or drug as well as evaluation of their therapeutic efficacy. Furthermore, this modality will serve as a screening method facilitating detection of early cancer. Advances in research into hypoxia and intratumoral microvessels of tumor with this endoscopic modality are expected and lead to development of new drugs. Thus, the proposed laser source-equipped hypoxia endoscope system appears to have the potential to redraw the endoscopic landscape.
9. Conclusions and perspectives
Tumor hypoxia assessment allows cancer patients to be followed up early after treatment initiation and drug resistance and radioresistance to be predicted. Current insights into the molecular mechanisms of hypoxia have indeed led to novel probes being developed for noninvasive imaging of hypoxia. Again, real-time hypoxic imaging in digestive endoscopy was obtained using such laser light as remains within visible light wavelengths, with no use of any probes. For innovation of endoscopy, it was elucidated that most of all early cancers and precursor lesions have already been to hypoxic state. This is a cutting edge finding. This imaging technology highlights a novel aspect of cancer biology as a potential biomarker which may come to be widely used in cancer diagnosis and treatment effect prediction. These approaches appear to have great promise and further studies on the predictive value of hypoxia measurement in tumors may help identify independent predictive marker of hypoxia as well as optimal parameters for assessing hypoxia. It remains to be clarified whether these new agents may help reduce hypoxic disease or whether they are available for hypoxia imaging.