PET radiopharmaceuticals for imaging of tumor hypoxia: a review of the evidence

Abstract

Hypoxia is a pathological condition arising in living tissues when oxygen supply does not adequately cover the cellular metabolic demand. Detection of this phenomenon in tumors is of the utmost clinical relevance because tumor aggressiveness, metastatic spread, failure to achieve tumor control, increased rate of recurrence, and ultimate poor outcome are all associated with hypoxia. Consequently, in recent decades there has been increasing interest in developing methods for measurement of oxygen levels in tumors. Among the image-based modalities for hypoxia assessment, positron emission tomography (PET) is one of the most extensively investigated based on the various advantages it offers, i.e., broad range of radiopharmaceuticals, good intrinsic resolution, three-dimensional tumor representation, possibility of semiquantification/quantification of the amount of hypoxic tumor burden, overall patient friendliness, and ease of repetition. Compared with the other non-invasive techniques, the biggest advantage of PET imaging is that it offers the highest specificity for detection of hypoxic tissue. Starting with the 2-nitroimidazole family of compounds in the early 1980s, a great number of PET tracers have been developed for the identification of hypoxia in living tissue and solid tumors. This paper provides an overview of the principal PET tracers applied in cancer imaging of hypoxia and discusses in detail their advantages and pitfalls.

Introduction

Hypoxia is a pathological condition arising in living tissue when the oxygen supply does not adequately cover the cellular metabolic demand. This phenomenon is also present in the vast majority of solid tumors and has been associated with a tendency toward poor prognosis [1]. The first to demonstrate the presence of hypoxia in human tumors were Tomlinson and Gray in the early 1960s [2]. So far we have evidence that up to 60% of locally advanced solid tumors are characterized by areas of reduced (hypoxia) or almost absent oxygen supply (anoxia) [3]. Detection of this phenomenon in tumors is of the utmost clinical relevance, because tumor aggressiveness, metastatic spread, failure to achieve tumor control, increased rate of recurrence, and ultimate poor outcome are all associated with hypoxia [4].

Onset of hypoxia in tumors is often the result of abnormal perfusion, which is typical of tumor-related neoangiogenesis and predominantly causes a transient hypoxia (acute hypoxia). In other cases hypoxia is caused by insufficient oxygen diffusion due to increased distance between the involved tissue and the blood supply (chronic hypoxia) or, to be more specific, a distance exceeding 100 μm from the nearest blood vessel, this being the diffusion distance of soluble oxygen [2]. Another mechanism of hypoxia induction is altered oxygen transport, such as occurs in disease- and/or treatment-related anemia [1,3,5-7].

The hypoxia epiphenomenon is translated into a downstream cascade of cellular adaptation mechanisms and is associated with various changes in gene expression, mostly mediated by the hypoxia-inducible factors 1 and 2 (HIF-1α and HIF-2) [5]. As reported by Post and Van Meir, the level of HIF gene activation is a function of oxygen concentration and increases exponentially when O₂ levels fall below 5% [8]. In general the median pressure of oxygen (pO₂) at which living tissues experience hypoxia is cited as around 8-10 mmHg [9,10]. At these oxygen levels, HIFs will trigger activation of genes involved in glycolysis, cell proliferation, cell survival, angiogenesis, and metastatic invasion [5,11]. This pattern of gene expression alters the malignant potential of tumors, following which cancer cells can become resistant to radiation treatment and chemotherapy [12,13].

Consequently, in recent decades there has been increasing interest in developing methods for measurement of the levels of oxygen in tumors. These methods can be invasive, such as the polarographic O₂ sensor (Eppendorf GmbH, Hamburg, Germany), or non-invasive, mainly based on imaging techniques [12]. Imaging modalities are undoubtedly more appealing for the assessment of tumor hypoxia because they guarantee all-encompassing visualization of the neoplastic tissue and can identify the phenomenon even at sites inaccessible to invasive procedures. Among the many techniques now available are optical-based methods, magnetic resonance imaging (MRI), and nuclear medicine techniques [14,15]. Some of their principal characteristics and limitations are summarized in Table 1, although an in-depth understanding of the value of each modality would require a more extensive report, which is beyond the scope of this review [14-18].

Table 1.

Examples of non-invasive methods for hypoxia determination in living tissues [14-18]

Modality		Technique	Limitations
Optical-based	Phosphorescence	Infusion of water-soluble phosphor probes into the vasculature.	The measurement represents the vascular pO2, not tissue pO2.
	Near-infrared spectroscopy (NIRS)	Non-invasive assessment of hemoglobin (Hb) saturation.	The measurement provides information on vascular oxygenation, but not on tissue pO2.
MRI-based	Blood oxygen level-dependent magnetic resonance imaging (BOLD MRI)	BOLD images reveal the changes in the amount of oxygen bound to hemoglobin in blood owing to deoxyhemoglobin, which is a paramagnetic substance.	The measurement provides information on changes in blood oxygenation, but not on the absolute oxygen concentration in tissue.
	19F-MRI or NMR (nuclear magnetic resonance)	Perfluorocarbons (PFCs) are injected intravenously and their 19F spin lattice relaxation rate (R1) varies linearly with the dissolved oxygen concentration.	The relaxation rate of 19F may depend on other physiological factors present in the tissue and not only on O2 concentration.
	Electron paramagnetic resonance imaging (EPRI)	Use of implantable paramagnetic particulates or soluble probes, intravenously injected, that physically interact with oxygen.	The molecules may predominantly distribute in the vasculature, thus biasing in part measurements of tissue oxygenation.
	Proton–electron double resonance imaging (PEDRI)	Injection of an external probe that has unpaired electrons and use of a strong EPR impulse.	The molecules may predominantly distribute in the vasculature, thus biasing in part measurements of tissue oxygenation.
	DCE-MRI (dynamic Gd-DTPA-enhanced MRI)	Injection of contrast agent and determination of vasculature perfusion/permeability.	Low specificity, because the measurement provides information on both vascular and tissue oxygenation.
Nuclear-based	Single-photon emission computed tomography (SPECT)	Injection of gamma (γ) emitting radiopharmaceuticals selective for hypoxic tissue. High specificity	Limited resolution dependent on voxel-based distribution of hypoxia.
	Positron emission tomography (PET)	Injection of positron (β+) emitting radiopharmaceuticals selective for hypoxic tissue. High specificity.	Limited resolution compared to MRI and optical methods, but superior to SPECT.

Among the image-based modalities for hypoxia assessment, positron emission tomography (PET) is one of the most extensively investigated based on the various advantages it offers: (a) a broad assortment of radiopharmaceuticals; (b) good intrinsic resolution (5 mm); (c) three-dimensional (3D) tumor representation; (d) possibility of semiquantification/quantification of the hypoxic tumor burden; (e) overall patient friendliness, and (f) ease of repetition [19]. Compared with the other non-invasive techniques, however, the biggest advantage of PET is that it displays the highest specificity for hypoxic tissue [20].

The object of the current paper is therefore to provide an overview of the principal PET radiopharmaceuticals applied in cancer imaging of hypoxia and to discuss in detail their advantages and pitfalls.

PET imaging of hypoxia

Starting with the 2-nitroimidazole family of compounds in the early 1980s [15,21], a great number of PET tracers have been developed for the identification of hypoxia in living tissues and solid tumors (Table 2). The driving force behind this development has been the need for highly specific imaging “probes” able to overcome the inconsistent correlation between findings on other imaging modalities, including PET with ¹⁸F-fluorodeoxyglucose (¹⁸F-FDG), and the hypoxia levels determined in tumor tissue [3,15,22].

Table 2.

Principal radiopharmaceuticals applied in PET imaging of tumor hypoxia

Uptake mechanism	Tracer	Tumors imaged	Benefits	Limitations
Pasteur effect (anaerobic glycolysis) [25]	18F-FDG (18F-fluorodeoxyglucose)	NSCLC [23,27,32,37]	Good correlation with tumor aggressiveness and prognosis	Overlap between uptake in normoxic (Warburg effect) [26] and hypoxia tumor tissue
		Head and neck tumors [31]	Easily reproducible and broad availability
		Oral squamous cell carcinoma [40,41]
		Gastric cancer [39]
Nitroimidazole-like uptake: reduction into RNO2 radicals and RNHOH compounds in hypoxic conditions. Then covalent binding to macromolecules [21,59]	18F-MISO (18F-fluoromisonidazole)	Head and neck tumors [35,42-45]	Broadest evidence of value as a hypoxia tracer.	Lack of correlation in all tumors
		Locally advanced HNSCC [35,46]	Good correlation with immunohistochemistry and prognosis in most cases.	Low tumor-to-background ratio
		Glioblastoma multiforme (GBM) [37,47,48]	Good availability	Variable reproducibility
		Breast cancer [49]
		NSCLC [32,33,50]
		Renal cell carcinoma [51]
	18F-FAZA (18F-fluoroazomycin-arabinozide)	Head and neck tumors [52,53]	Good correlation with immunohistochemistry and prognosis in most cases.	More limited evidence compared to 18F-MISO.
		Cervical cancer [54]	Faster diffusion and clearance with slightly higher tumor-to-background ratio than 18F-MISO.
		Prostate cancer [55]
		NSCLC [56,57]
		Rectal cancer [58]
	18F-FETNIM (18F-fluoroerythronitroimidazole)	NSCLC [60]	Promising tracer with possible correlation with outcome.	Limited evidence compared to 18F-MISO.
		Esophageal cancer [61]	Slightly higher tumor-to-background ratio than 18F-MISO.
	18F-EF5 (18F-2-nitroimidazol-pentafluoropropyl acetamide)	Brain tumors [62]	Promising tracer with possible correlation with outcome	Limited evidence.
		Soft tissue sarcoma [63]
		Head and neck tumors [64]
	18F-EF3 (18F-2-nitroimidazol-trifluoropropyl acetamide)	Rats bearing syngeneic rhabdomyosarcoma tumours [65]	Promising tracer.	Very limited evidence, mostly preclinical.
		Head and neck tumors [66]
	18F-FETA (18F-fluoroetanidazole)	Mice bearing MCF-7, RIF-1, EMT6, HT1080/26.6, and HT1080/1-3C xenografts [67,68]	Promising tracer with better biodistribution than 18F-MISO.	Preclinical evidence
	124I-IAZG (124I-iodoazomycin galactopyranoside)	Hepatocellular carcinoma [69]	Promising tracer	Preclinical evidence
	68Ga-labeled nitroimidazole analogs (68Ga-NOTA-nitroimidazole, 68Ga-DOTA-nitroimidazole, 68Ga-SCN-NOTA-nitroimidazole)	Tumor xenografted mice [70,71]	Promising tracer	Preclinical evidence
Reduction of Cu(II)-ATSM complex into Cu(I)-ATSM and dissociation of Cu(I) in hypoxic conditions: then Cu(I) nuclide binding to intracellular proteins [77]	60,61,62,64Cu-ATSM (60,61,62,64Cu-diacetyl-bis(N4-methylthiosemicarbazone)	NSCLC [34]	Good correlation with immunohistochemistry and prognosis.	Evidence more limited compared to 18F-MISO.
		Head and neck tumors [72,73]	Early uptake of the tracer with high tumor-to-background ratio.	Less clear mechanism of uptake in tumor hypoxia compared to nitroimidazole-like compounds.
		Cervical cancer [74,75]	Possibility for late acquisition with 64Cu-ATSM.
		Rectal cancer [76]	Possibility for radionuclide therapy with 64Cu-ATSM.
		Brain tumors [78]
Recognizes carbonic anhydrase IX (CA IX) [80]	124I-cG250 (124I-chimeric mAb G250)	Renal cell carcinoma [79]	Promising tracer	Preclinical evidence
	89Zr-cG250-F(ab′)2 (89Zr-chimeric G250 F(ab′)2)	Head and neck tumors [81]	Promising tracer	Preclinical evidence

Abbreviations: HNSCC, head and neck squamous cell carcinoma; NSCLC, non-small cell lung cancer.

¹⁸F-fluorodeoxyglucose (¹⁸F-FDG)

Undoubtedly ¹⁸F-FDG PET remains a cornerstone for tumor evaluation, response assessment, and disease prognostication, but it requires careful handling when trying to depict hypoxic tissue. The fact that tumor hyperglycolysis due to up-regulation of glucose transporters (GLUTs) and glycolytic enzymes can be driven by HIF-1α [22,23] offers some justification for the use of ¹⁸F-FDG as a surrogate marker of hypoxia [24]. Moreover, we know that under reduced levels of oxygen (↓pO₂), living cells switch their metabolic pathway for ATP production to anaerobic glycolysis, also known as the Pasteur effect [25].

However, in the case of hypoxic tumor cells, a wide overlap exists between ¹⁸F-FDG uptake due to aerobic glycolysis, the so-called Warburg effect [26], and anaerobic glycolysis [25,27] (i.e., normoxic and hypoxic conditions, respectively) (Figure 1). The fact that HIF-1α expression can be observed also in non-hypoxic tumor regions [28,29] suggests that other factors can indirectly influence glucose metabolism and ¹⁸F-FDG uptake in those areas [22]. It therefore appears comprehensible why, in many experiments, the correlation between ¹⁸F-FDG uptake and the level of tumor hypoxia has not been confirmed or conflicting results have been obtained [3,30]. These shortcomings apply to the imaging of a variety of tumor types, including head and neck carcinoma, lung cancer, sarcomas, breast cancer, and brain tumors [22,31-38]. For instance, in two different studies of, respectively, 24 and 36 patients with head and neck squamous carcinoma [31,33], direct comparison of ¹⁸F-FDG uptake and hypoxia determination using a polarographic O₂ sensor documented a lack of correlation. Similarly, no correlation of glucose metabolism on ¹⁸F-FDG PET and hypoxia was observed in non-small cell lung cancer (NSCLC) patients [32-34]. These data are not to be considered absolutely negative, because ¹⁸F-FDG has been documented to be capable of defining more aggressive tumor types, also correlated with HIF-1α expression, in patients with gastric carcinoma [39] or tongue cancer [40], as well as in those with both the above-mentioned neoplasia, i.e., NSCLC [23,27] and oral squamous cell carcinoma [40,41].

Figure 1.

Overview of the uptake and retention mechanisms of FDG (A), F-MISO (B), and Cu-ATSM (C) in living cells under hypoxic conditions. For FDG there is a wide overlap between the cellular uptake in normoxic (Warburg effect) and hypoxic conditions (Pasteur effect). For the other two tracers, after passive diffusion through the membrane, the radiopharmaceutical is retained according to the oxygen tension (pO₂) present in the intracellular environment: in the presence of reduced pO₂, F-MISO undergoes progressive reduction by the nitroreductase enzyme (NTR); also, Cu(II)-ATSM nuclide is reduced to copper (I) by the intracellular thiols, making the Cu-ATSM complex less stable. Both processes are reversible in the presence of sufficient O₂, and the molecules (F-MISO and Cu (II)-ATSM) are free to leave the cell. Conversely, in hypoxic conditions the Cu(I)-ATSM complex is progressively dissociated, with the formation of H₂-ATSM and free Cu(I), which is very rapidly incorporated into intracellular proteins. In contrast, the reduced F-MISO is covalently bound to the intracellular proteins [59,84,122,128]. GLUT, glucose transporter; HK, hexokinase; G-6-P, glucose-6-phosphate).

In summary, the limitations on the specific application of ¹⁸F-FDG for the detection of hypoxia persist, and in the case of tumor imaging it is advisable to combine this tracer with other hypoxia-avid ones in order to achieve a comprehensive assessment of the tumor characteristics [25].

Nitroimidazole family of compounds

¹⁸F-fluoromisonidazole (¹⁸F-FMISO)

The fluorinated nitroimidazole derivative ¹⁸F-fluoromisonidazole, or ¹⁸F-FMISO, is the most widely studied PET tracer for hypoxia imaging. It was first developed for this purpose in 1986 [82,83] and since then has been extensively used for the detection of many tumor types in both the preclinical and the clinical context [3,15]. Like the other compounds in the nitroimidazole family, this tracer is passively diffused through the cell membrane owing to its lipophilicity, and once within the intracellular environment it is reduced into R-NO₂ radicals by the nitroreductase enzyme (NTR) (Figure 1). This process is still reversible and when the cell is well oxygenated, the tracer is not entrapped and can freely flow back into the extracellular environment. Conversely, in the presence of reduced levels of oxygen (pO₂ <10 mmHg) the process of ¹⁸F-FMISO reduction continues slowly; the consequence is the progressive production of R-NHOH compounds that bind covalently to intracellular molecules, and ultimately entrapment of the tracer within the cell [59,84,85].

The amount of ¹⁸F-FMISO uptake is therefore influenced by the O₂ level in tumor tissue, as is confirmed by the good correlation observed between tracer uptake and pO₂ polarography [32,33] or immunohistochemical determination of hypoxia [86,87]. However, the time line of the above-mentioned processes is rather long for an ¹⁸F-fluorine labeled tracer (T _1/2 109 min), because selective retention of ¹⁸F-FMISO in hypoxic tissue requires an uptake period of around 2-4 h after intravenous injection [11,45,88]. In addition, despite this uptake period, tracer accumulation is still low, as documented by the low tumor-to-plasma or tumor-to-muscle ratio of 1.2-1.4, used as the optimal cut-off for definition of hypoxia [15,45]. These aspects represent the main drawbacks of ¹⁸F-FMISO, and may limit the applicability of the tracer in clinical practice. To overcome the problem, a dynamic approach has been tested by Thorwarth et al. [89,90], in which a kinetic analysis is used to separate the component associated with hypoxia-specific tracer binding from that related to unbound tracer. This kinetic approach is, however, cumbersome and still restricted by the resolution limit of the technology itself.

Hypoxia imaging with ¹⁸F-FMISO has been investigated in numerous solid tumors, including gliomas [38,47,48,91], head and neck carcinoma [42-46], NSCLC [33,50], breast tumors [49], and renal carcinoma [51] (Table 2). In patients with brain tumors, for instance, Hirata et al. [48] supported a role for ¹⁸F-FMISO PET in differentiating glioblastoma multiforme (GBM) from other less malignant gliomas based on the level of tumor hypoxia. Moreover, Swanson et al. [92] reported a good correlation between the hypoxic volume determined by ¹⁸F-FMISO and the MRI-defined tumor burden, with particular interest on disrupted vasculature on gadolinium-enhanced T1-weighted sequences (T1Gd). Their data confirm that the angiogenic process is stimulated by hypoxia in GBM and indirectly anticipate the more recently reported association between tumor aggressiveness visualized on ¹¹C-methionine imaging, disrupted blood–brain barrier vasculature on contrast-enhanced-MRI, and hypoxia depicted with ¹⁸F-FMISO [93].

Additionally, PET imaging with ¹⁸F-FMISO has been shown to discriminate prognosis in GBM For example, Spence et al. [47] studied 22 GBM patients before biopsy or between resection and radiation therapy (RT) and observed both the volume and the intensity of hypoxia as determined by ¹⁸F-FMISO before therapy to be strongly correlated with time to progression and survival.

Similarly, in patients with head and neck tumors, Rajendran et al. [43] documented a prognostic role for pretherapy ¹⁸F-FMISO uptake with respect to overall survival, with hypoxic volume and nodal involvement also being predictive factors. Rischin et al. [94] demonstrated that in patients receiving- non-tirapazamine-containing chemoradiotherapy for stage III or IV head and neck tumors, hypoxia on FMISO PET was associated with a higher rate of locoregional failure. The introduction of kinetic analysis of ¹⁸F-FMISO, as reported by Eschmann et al. [42], could also predict higher risk of relapse.

In 20 postmenopausal women with stage II-IV breast cancer, Cheng et al. [49] analyzed the role of ¹⁸F-FMISO PET before and after endocrine therapy with letrozole. Tracer uptake was detected at 2 and 4 h after injection and tumor-to-background ratio was correlated to treatment outcome after 3 months. The authors observed a positive correlation between baseline ¹⁸F-FMISO uptake and response to therapy (p <0.0001) and could define a tumor-to-background ratio at 4 h of ≥1.2 as the optimal cut-off point, allowing the prediction of 88% (15/17) of cases of progressive disease. No correlation, however, was found between ¹⁸F-FMISO uptake and HIF-1α expression at immunohistochemistry.

An important application for hypoxia imaging is undoubtedly RT planning. It is well known that the pretreatment oxygenation in cancer tissue influences response to treatment, because treatment effectiveness is strictly related to the amount of free oxygen radicals. Consequently the radiation dose necessary to achieve the same therapeutic effect is much higher for hypoxic tumors [88]. ¹⁸F-FMISO has therefore been investigated in this context. Starting with their feasibility study, Lee et al [45] reported the use of ¹⁸F-FMISO PET to increase the dose to hypoxic regions in head and neck carcinoma. In the same clinical setting, they tried to determine the reproducibility of the PET scan at two different time points prior to RT [95] and to assess the influence on dose-painting at intensity-modulated radiotherapy (IMRT) [96]. On the basis of these studies they concluded that changes in the spatial distribution of tumor hypoxia, as detected by serial FMISO PET, compromised the coverage of hypoxic tumor volumes achievable by dose-painting IMRT [96]. However, even when such changes occurred, dose-painting always increased the equivalent uniform dose of the hypoxic areas. In rectal cancer, the use of ¹⁸F-FMISO PET for target definition prior to RT [97] appears less reliable due to non-specific tracer uptake in normoxic tissue and diffusion through the bowel wall.

High reproducibility of tumor hypoxia evaluated by ¹⁸F-FMISO PET was recently reported by Okamoto et al. [98] in 11 patients with untreated head and neck cancer who were investigated twice with ¹⁸F-FMISO PET at an interval of 48 h. In this cohort the 4-h tracer uptake parameters (SUV_max, tumor-to-background, and tumor-to-muscle ratio) showed no significant difference between the scans and, except in one case, the location of the SUV_max peaks, although different in PET1 and PET2, were within the full-width at half-maximum of the PET/CT scanner.

In a prospective study by Tachibana et al. [99] a limited cohort of ten patients was studied before and during fractionated RT with ¹⁸F-FMISO PET/CT in order to determine the intratumoral hypoxic areas and their reoxygenation. The study revealed a high percentage of tumor reoxygenation (8/10) during RT, suggesting that dose escalation to the hypoxic areas on the initial PET/CT scan might be inappropriate. However, the authors suggested that if frequent imaging with ¹⁸F-FMISO PET/CT becomes available, adaptive RT for tumor hypoxia might be used clinically.

Somewhat similar findings were reported in the study by Lee et al. [100], in which resolution of tumor hypoxia on mid-treatment ¹⁸F-FMISO PET during fractionated RT, as would be expected for doses higher than 40 Gy, was consistent with the concept of reoxygenation. However, despite the promising results from the first report [45], neither the presence nor the absence of hypoxia defined by ¹⁸F-FMISO PET during mid-treatment evaluation correlated with patient outcome.

More contradictory results have been reported in NSCLC. Lack of correlation between expression of tumor markers of hypoxia and ¹⁸F-FMISO uptake was observed in a series of 17 patients with resectable NSCLC [37]. Gabel et al. [33] also found a lack of correlation between high initial tracer uptake and treatment response in NSCLC patients, although they reported that decreased ¹⁸F-FMISO uptake at post-treatment evaluation was indicative of a favorable outcome. Evidence of utility of ¹⁸F-FMISO PET in renal cell carcinoma or sarcoma is even more limited [36,51,101].

Taken together, these data raise a question mark over the use of ¹⁸F-FMISO as a “universal” tracer for hypoxia imaging.

¹⁸F-Fluoroazomycin arabinoside (¹⁸F-FAZA)

The slow uptake of ¹⁸F-FMISO in target tissue and slow clearance of unbound ¹⁸F-FMISO from non-hypoxic areas stimulated the development of other tracers with improved pharmacokinetics, including ¹⁸F-fluoroazomycin arabinoside (¹⁸F-FAZA), a second-generation 2-nitroimidazole compound developed in 1999 [102]. Compared to ¹⁸F-FMISO, the biodistribution of ¹⁸F-FAZA is improved through the addition of a sugar moiety, making it less lipophilic [102,103]. Souvatzoglou et al. [53] reported a higher contrast with non-target tissues for ¹⁸F-FAZA compared to ¹⁸F-FMISO, with an average tumor-to-muscle ratio of 2.0±0.3 at 2 h postinjection acquisition. The same group [104] had previously reported that ¹⁸F-FAZA has overall superior pharmacokinetics and that use of dynamic analysis offers further potential improvement [105-107].

So far ¹⁸F-FAZA has shown promising results in animal and patient studies [53,102,108-111] based on its selective accumulation in hypoxic tumors via a hypoxia-specific uptake mechanism [104]. In tumor-bearing mice with human SiHa cervix xenografts, for instance [112], intratumoral distribution of ¹⁸F-FAZA was strongly correlated with the regional density of the pimonidazole-positive cells (pimonidazole being a hypoxia marker).

The role of tumor hypoxia depicted by ¹⁸F-FAZA as a predictor of anticancer treatment response has been investigated in several preclinical models. The effect of hypoxia modulation with gefitinib, an epidermal growth factor receptor (EGFR) tyrosine kinase inhibitor, has been assessed with ¹⁸F-FAZA PET in human EGFR-expressing A431 squamous cell carcinoma xenografts [113]. Also the use of radiosensitizers, such as tirapazamine, has been investigated with ¹⁸F-FAZA PET in EMT6 tumor-bearing nude mice prior to treatment with concurrent chemoradiotherapy, RT alone, or chemotherapy alone [109]. In each case, hypoxia imaging proved efficient in predicting the beneficial effect of the treatment.

In a preclinical study, Mortensen et al. [114] investigated 92 female CDF1 mice with subcutaneous C3H mammary carcinomas prior to irradiation (55 Gy). The authors demonstrated a significant difference in local tumor control between “more hypoxic” and “less hypoxic” cases distinguished by either the median ¹⁸F-FAZA tumor-to-blood ratio or the fraction of oxygen partial pressure at the pO₂ Eppendorf electrode.

These data, taken together, have prompted the investigation of ¹⁸F-FAZA in clinical settings. One of the largest cohorts in which the tracer has been investigated in the clinical context is that reported by Postema et al. [115]. In a group of 50 patients with different types of solid tumor, i.e., head and neck squamous cell carcinoma (HNSCC), small cell lung cancer (SCLC), NSCLC, malignant lymphoma, and high-grade gliomas, the authors aimed first to evaluate the safety and general biodistribution of ¹⁸F-FAZA. They observed highly increased uptake of the tracer in all gliomas, with a tumor-to-background (T/B) ratio range of 1.9-15.6, and variable uptake in the remaining tumors, with a T/B closer to the average cut-off value of 1.6-2.0.

Recently, a group from Melbourne [56] investigated the role of ¹⁸F-FAZA in 17 patients with locoregionally advanced NSCLC before concurrent chemoradiation. Intralesional hypoxia was identified in 65% of patients (11/17), and in those investigated with ¹⁸F-FAZA PET after chemoradiation (60 Gy) (8/11), imageable hypoxia had resolved in the majority (6/8). Disease-free survival, however, did not differ significantly between patients with hypoxic and those with non-hypoxic tumors.

The first report on RT planning with ¹⁸F-FAZA dates back to 2007 and focused on dose-painting according to hypoxia image-guided RT in 18 patients with advanced HNSCC [116]. In this report, Grosu et al. outlined the gross tumor volume (GTV) on ¹⁸F-FAZA PET by applying a threshold of 50% with regard to background. This led to the inclusion of any PET-positive area with a T/M ratio ≥1.5. For primary localizations, GTV-FAZA presented with a single confluent hypoxic area in 61% of cases and with multiple diffused areas in 22%. In all cases, however, GTV-FAZA was inside the GTV outlined on CT. Although no comparison with ¹⁸F-FDG distribution was performed to determine the effective benefit of GTV-PET delineation, the conclusion was that dose-painting on hypoxic areas is potentially feasible.

The use of ¹⁸F-FAZA before RT also appears feasible in cervical cancer, as documented by Schuetz et al. [54], although the authors did not find a clear impact on survival in their limited cohort. Mortensen et al. [52] reported some more thorough results from the DAHANCA 24 trial on the role of ¹⁸F-FAZA PET in head and neck cancer before RT. The 40 patients investigated had undergone hypoxia PET before RT (66-76 Gy) and during treatment. In 25 cases (63%), PET showed a hypoxic volume with a tumor-to-muscle ratio (T/M) in the range of 1.1-2.9 (median 1.5). In this study, the prognostic significance of ¹⁸F-FAZA PET was confirmed (p=0.04): at a median follow-up of 19 months, disease-free survival was 93% for patients with non-hypoxic tumors and 60% for patients with hypoxic tumors.

One of the open questions regarding hypoxia image-guided RT is the reproducibility of the PET data. Busk et al. [112] performed ¹⁸F-FAZA PET twice before initiation of fractionated RT in mice bearing human SiHa cervix tumor xenografts and again following treatment. They found that ¹⁸F-FAZA results were highly reproducible when based on injected dose, whereas normalization using an image-derived non-hypoxic reference tissue (i.e., muscle) yielded highly unreliable results. The authors underlined the stability of the intratumoral tracer distribution at baseline and its strong correlation with regional density of hypoxic cells. No evidence of general reoxygenation was observed during treatment, however, despite changes in overall tracer retention in individual mice. Consequently the question of reproducibility remains open when dealing with fractioned RT, especially with the intent of image-guided dose escalation.

Other nitroimidazole-like tracers

In view of the limitations of the above-mentioned compounds, other nitroimidazole-like tracers with high avidity for hypoxic tissue have been investigated and developed. One promising new radiopharmaceutical is ¹⁸F-fluoroery-thronitroimidazole (¹⁸F-FETNIM), which is more hydrophilic than ¹⁸F-FMISO and can be washed out more rapidly from well-oxygenated tissues, theoretically allowing a higher tumor-to-background ratio [117]. Pilot studies in patients with head and neck, esophageal, and lung cancer have demonstrated ¹⁸F-FETNIM PET to be feasible and useful in hypoxia imaging [60,61,118], with the potential to predict response to treatment [118] and overall patient outcome [60]. Superior overall benefit in relation to ¹⁸F-FMISO has not been demonstrated, however, and the T/B ratio for this tracer was not significantly superior to the ratios for other nitroimidazole-like tracers [119].

Similar results have been obtained with ¹⁸F-fluoroetanidazole (¹⁸F-FETA), which is a well-known nitroimidazole-like compound that has shown a better biodistribution than ¹⁸F-FMISO owing to its lower levels of liver and lung retention [67,68]. However, in spite of potential benefits, the diffusion of this tracer into tumor tissues appears limited [84].

Another group of hypoxia-avid radiopharmaceuticals, with a more stable but also more complex labeling chemistry, is represented by ¹⁸F-2-nitroimidazol-pentafluoropropyl acetamide (¹⁸F-EF5) and ¹⁸F-2 nitroimidazol-trifluoropropyl acetamide (¹⁸F-EF3) [66,120,121]. These tracers are slightly more lipophilic than the formerly described compounds and have been investigated in animal models as well as in clinical studies on head and neck cancer and cervical and brain tumors [62-64]. For these tracers, too, the optimal tumor-to-muscle cut-off value is low (T/M 1.5) and the potential advantage over ¹⁸F-FMISO is still negligible.

Valuable alternatives may be biochemically similar tracers labeled with other nuclides, including iodine-124 (¹²⁴I), e.g., ¹²⁴I-iodoazomy-cin galactopyranoside (¹²⁴I-IAZG) [69], and gallium-68 (⁶⁸Ga), e.g., ⁶⁸Ga-NOTA-nitroimidazole, ⁶⁸Ga-DOTA-nitroimidazole, and ⁶⁸Ga-SCN-NOTA-nitroimidazole [70,71] (Table 2). The ⁶⁸Ga-labeled tracers have the additional advantage of utilizing a nuclide produced by a generator (⁶⁸Ga/⁶⁸Ge) and are thus potentially applicable in PET centers without an onsite cyclotron. Up to now, however, these tracers have not proved superior to the principal nitroimidazole representative, ¹⁸F-FMISO, for hypoxia imaging. As a consequence, their application is still limited to preclinical studies [70,71,84].

Non-nitroimidazole compounds

Cu-diacetyl-bis(N4-methylthiosemicarbazone) (Cu-ATSM)

Radioactive copper (^60,61,62,64Cu) labeled with diacetyl-bis(N4-methylthiosemicarbazone) (Cu-ATSM) is a very promising PET radiopharmaceutical for hypoxia imaging. First investigated for this purpose in 1997 [122], the compound appeared immediately suitable for detection of hypoxia in living tissue. A series of copper radioisotopes is now available for labeling ATSM, each with its specific half-life (T _1/2), decay scheme, and production facilities (Table 3). The mechanism of uptake is still not fully understood, but as Fujibayashi et al. suggested [122], retention of the tracer in tumor cells is principally dependent on cytosolic/microsomal bioreduction [123]. In fact Cu-ATSM is a neutral lipophilic molecule, which is highly membrane permeable and can passively diffuse within the intracellular environment (Figure 1). Once inside the cell, the bivalent copper compound, Cu(II)-ATSM, undergoes reduction by thiols and is converted into Cu(I)-ATSM complex [124]. In hypoxic conditions this complex, less stable than the bivalent form, is progressively dissociated into H₂-ATSM and free Cu(I), which is rapidly entrapped in intracellular proteins [122,125]. The entrapment is reported to reflect the level of tissue oxygenation in many tumor types [34,72-74,76,78] and when directly compared to the principal 2-nitroimidazole family representative (¹⁸F-FMISO), Cu-ATSM uptake is significantly higher in target tissue than in non-hypoxic areas and occurs at an earlier time (10-15 min versus 2-4 h) [88,126]. These data have also been confirmed in relation to other nitroimidazole compounds, i.e., ¹⁸F-FAZA and ¹⁸F-HX4 [99], in nude mice bearing human xenografts.

Table 3.

Characteristics of copper nuclides utilized in PET imaging and comparison with otherpositron emitters [59,137-140]

Nuclides	T 1/2	Production	β+ emission (Emean)	Other emissions	Range of β+ in tissue	Use
Copper-60 (60Cu)	23.7 min	Cyclotron	93% (0.970 MeV)	γ emission	4.4 mm	Diagnostic
				1332 keV 88%
				1791 keV 45.4%
Copper-61 (61Cu)	3.33 h	Cyclotron	61% (0.500 MeV)	γ emission	2.6 mm	Diagnostic
				282 keV 12.20%
				656 keV 10.77%
Copper-62 (62Cu)	9.67 min	Generator/cyclotron	97.83% (1.319 MeV)	γ emission	6.6 mm	Diagnostic
				1172 0.74%
Copper-64 (64Cu)	12.7 h	Cyclotron	17.6% (0.278 MeV)	γ emission	1.4 mm	Diagnostic/therapeutic
				1345 keV 0.47%
				β- emission
				0.190 MeV 38.7%
Fluoride-18 (18F)	109.7 min	Cyclotron	96.7% (0.249 MeV)	β- emission (0.52 keV)	0.6 mm	Diagnostic
Iodine-124 (124I)	4.17 days	Cyclotron	22.7% (0.820 MeV)	γ emission	3 mm	Diagnostic
				0.602 keV 62.9%
				1690.9 keV 11.15%
Gallium-68 (68Ga)	67.71 min	Generator	88.91% (0.829 MeV)	γ emission	2.9 mm	Diagnostic
				1077 keV 3.2%

From recent investigations, it appears plausible that a significant role in the Cu-ATSM entrapment is played by copper itself. To test this assumption, Hueting et al. [127] analyzed the in vitro and in vivo distribution of ⁶⁴Cu-ATSM and ⁶⁴Cu-acetate in the same animal models (EMT6 and CaNT). They showed a similar tissue distribution of radio-copper for both tracers and suggested that copper metabolism can play a role in the mechanism of selectivity of Cu-ATSM in hypoxia. More thorough investigations and consolidated evidence in the clinical context are required to confirm these data.

The first human use of Cu-ATSM dates back to 2000, when Takahashi et al. [128] studied its application in four normal subjects and six patients with lung cancer. The tracer, in this case ⁶²Cu-ATSM, accumulated within a few minutes in all patients with cancer, giving a tumor-to-background ratio of 3.0, whereas it rapidly cleared from the blood of all normal subjects. Similar findings have been documented for the other copper radioisotopes labeled with ATSM. In their feasibility study, Dehdashti et al. [34] analyzed the role of ⁶⁰Cu-ATSM in patients with NSCLC and correlated imaging findings with follow-up (n=19) and response to therapy (n=14). As expected, the tracer had a variable distribution in tumor masses, depending on hypoxia level, and the authors were able to define a tumor-to-muscle (T/M) ratio of 3.0 as effective in distinguishing treatment responders from non-responders.

The same group [74] evaluated the prognostic significance of ⁶⁰Cu-ATSM in 14 patients with cervical cancer before RT and chemotherapy. This time the selected T/M ratio was 3.5, which could optimally distinguish patients experiencing recurrence from those free of disease at last follow-up. Similar results were obtained in a more recent study in 38 patients with cervical carcinoma [128]. In this case, ⁶⁰Cu-ATSM performed before treatment gave relevant information on tumor oxygenation and was predictive of patient outcome.

The group from Yokohama City University [72] investigated use of ⁶²Cu-ATSM in 17 patients with locally advanced head and neck cancer (stage III and IV) prior to chemotherapy or RT. In 15 cases the authors assessed the relationship between clinical outcome and ⁶²Cu-ATSM uptake. The SUV_max in their analysis differed significantly (p <0.05) in patients free of disease at 2 years postirradiation follow-up versus those with residual/recurrent disease. More specifically, all cured patients had a SUVmax <5.0 and all patients (n=10) with persistent disease had a SUV_max >5.0 (Figures 2 & 3).

Figure 2.

Example of a patient with localized head and neck squamous cell carcinoma (HNSCC) who was investigated with ¹⁸F-FDG PET before (A) and after the end of RT (B). At staging the patient had undergone ⁶⁴Cu-ATSM PET/CT (C-E) documenting some mild uptake at the level of the primary tumor in the left tonsil (SUV_max 1.85). As is visible in (B) the patient achieved a complete response after treatment.

Figure 3.

Example of a patient with advanced HNSCC who was investigated with ¹⁸F-FDG PET at staging (A) and after the end of combined chemoradiotherapy (B). Before treatment the patient underwent ⁶⁴Cu-ATSM PET/CT (C-E), documenting intense tracer uptake (SUV_max 17.86) both in the primary tumor, involving the right tonsil, and in numerous bilateral cervical nodes. Despite the high-dose therapeutic regimen utilized, the patient presented some residual disease at end-of-treatment evaluation (B), as confirmed during follow-up.

Recently ⁶²Cu-ATSM was investigated in 22 patients with gliomas [78] with the intent of differentiating tumor grade according to uptake and correlating findings with contrast-enhanced regions on MRI and HIF-1α expression at immunohistochemistry. Using a tumor-to-background ratio threshold of 1.8 at 30-40 min post injection, ⁶²Cu-ATSM uptake was found to be predictive of HIF-1α expression, with 92.3% sensitivity and 88.9% specificity. Moreover, it correlated significantly with the presence of a necrotic component (p=0.002) and defined regional uptake in 61.9% (13/21) of tumors within the contrast-enhanced region on MRI.

However, other preclinical data suggest that Cu-ATSM may not be suitable for hypoxia detection in all types of tumor. In a fibrosarcoma animal model (FSA) and in prostate cancer cell lines (PC-3, 22Rv1, LNCaP, LAPC-4, and R3327-AT) [130-132], the tracer showed limited selectivity for hypoxia, suggesting the need for specific tumor-type studies with Cu-ATSM and also for more clinical evidence in these types of solid tumor. This latter aspect is crucial because animal models do not completely match human cancer and tend to give discordant findings based on the cell line utilized. For example, the rat model of fibrosarcoma investigated by Jalilian et al. in 2009 [133] with ⁶¹Cu-ATSM yielded completely different results from the findings of Yuan et al [130] using the FSA model.

With regard to RT planning, the principal advantage of Cu-ATSM is its high tumor-to-background ratio (T/B >3.0). As reported by Dalah et al. [134] in their simulation of tissue activity curves for ⁶⁴Cu-ATSM and ¹⁸F-FMISO for sub-target volume delineation, a good tumor-to-background ratio allows high sensitivity and specificity targeting of positive lesions on PET. This was also shown in the feasibility study reported by Chao et al [135] in head and neck tumors, where the use of IMRT based on Cu-ATSM led to a higher dose (80 Gy) in hypoxic areas and spared more than half of the parotid glands to less than 30 Gy [116]. However, one weakness of ⁶⁴Cu-ATSM needs to be underlined: the total body irradiation at diagnostic administered activities (500-800 MBq) is twice as high as the dose calculated for ¹⁸F-FMISO [59,136].

Nevertheless, among the different copper nuclides, ⁶⁴Cu represents the best compromise based on T _1/2, intrinsic image resolution and production yield (Table 3). In a direct comparison of ⁶⁰Cu-ATSM and ⁶⁴Cu-ATSM in ten patients with cervical carcinoma [75], for instance, ⁶⁴Cu-ATSM proved as safe as ⁶⁰Cu-ATSM while also offering the advantage of better image quality. Another advantage of the use of ⁶⁴Cu-ATSM is the theranostic potential of the nuclide, which emits medium-energy β- particles, along with positrons, and produces high linear energy transfer (LET) Auger electrons [136,141]. At adequate doses and thanks to the short path length of the emissions, the nuclide can produce a toxic effect on targeted cells with minimal effects on neighboring tissue, as already reported in some preclinical studies [141,142]. This therapeutic effect could also be seen by Yoshii et al. [143] in a mouse colon carcinoma (Colon-26) model, where ⁶⁴Cu-ATSM administration at 37 MBq twice a week reduced tumor volume as well as the percentage of CD133+ cells and the metastatic ability of Colon-26 tumors. However, this potential application needs more clinical evidence, so that for the time being the major use of ⁶⁴Cu-ATSM in the diagnostic field is for hypoxia assessment.

¹²⁴I-cG250 and ⁸⁹Zr-cG250-F(ab′)

Carbonic anhydrase IX (or CAIX) is a transmembrane enzyme involved in the cellular regulation of pH homeostasis and represents one of the downstream targets of HIF-1α [81]. Its role is to hydrolyze the carbon dioxide (CO₂) into carbonic acid (H₂CO₃) and stabilize intracellular pH [81,144]. With the exception of renal cell carcinoma, where the CAIX expression is not related to hypoxia, in tumors this enzyme is up-regulated as a result of reduced levels of oxygenation, namely <20 mmHg, and can therefore be targeted for hypoxia imaging [79,81,145].

The first compound developed for the identification of CAIX, although at the time the enzyme was not known, was the antibody Grawitz250 (G250) [146]. Later the chimeric version of the antibody was labeled with ¹²⁴I-iodine as a tracer for PET imaging (¹²⁴I-cG250) [79,147], and more recently selected antibody fragments have been labeled with ⁸⁹Zr-zirconium (⁸⁹Zr-cG250-F(ab′) for the same purpose [81]. This category of tracers has the potential to detect hypoxia in tumors, other than renal cell carcinoma, owing to the good correlation reported between tracer uptake and CAIX expression, although the evidence is still too limited and is reliant only on preclinical studies.

Conclusions

The clinical relevance of hypoxia in patients with cancer means that it has the potential to become a useful prognostic biomarker. Furthermore, the possibility of identifying hypoxia in vivo without any invasive intervention may be of great value in improving treatment. Among the numerous PET tracers investigated, a broad range of radiopharmaceuticals have been found to specifically identify hypoxia expression in tumors. It is nevertheless not straightforward to determine the most useful tracer for this purpose because many factors influence the choice. Evidence-based data favor the use of ¹⁸F-FMISO, but the issue of suboptimal imaging persists. On the other hand, if importance is placed on high PET image quality, ⁶⁴Cu-ATSM would be selected; in this case, however, evidence is more limited and the mechanism of uptake in hypoxic tissue is still not completely clear. Alternatively, “new” tracers labeled with cyclotron-independent nuclides hold appeal despite the apparent lack of superiority compared to ¹⁸F-FMISO. Nonetheless, if a “winner” has to be chosen in this “competition”, we would select the tracer that demonstrates better image quality.

Disclosure of conflict of interest

None to declare.