Imaging hypoxia in gliomas

Abstract

Hypoxia plays a central role in tumour development, angiogenesis, growth and resistance to treatment. Owing to constant developments in medical imaging technology, significant advances have been made towards in vitro and in vivo imaging of hypoxia in a variety of tumours, including gliomas of the central nervous system. The aim of this article is to review the literature on imaging approaches currently available for measuring hypoxia in human gliomas and provide an insight into recent advances and future directions in this field. After a brief overview of hypoxia and its importance in gliomas, several methods of measuring hypoxia will be presented. These range from invasive monitoring by Eppendorf polarographic O₂ microelectrodes, positron electron tomography (PET) tracers based on 2-nitroimidazole compounds [¹⁸F-labelled fluoro-misonidazole (¹⁸F-MISO) or 1-(2-[(¹⁸)F]fluoro-1-[hydroxymethyl]ethoxy)methyl-2-nitroimidazole (FRP-170)], ⁶⁴Cu-ATSM Cu-diacetyl-bis(N4-methylthiosemicarbazone) (Cu-ATSM) or ^99mTc- and ⁶⁸Ga-labelled metronidazole (MN) agents to advanced MRI methods, such as blood oxygenation level dependent (BOLD) MRI, oxygen-enhanced MRI, diffusion-weighted MRI (DWI-MRI), dynamic contrast-enhanced MRI (DCE-MRI) and ¹H-magnetic resonance spectroscopy.

Mounting evidence over the last decade indicates that hypoxia plays a vital role in tumour development, angiogenesis, growth and resistance to treatment. Alterations in the malignant potential of tumours induced by hypoxia and changes in the tumour's gene expression lead to more aggressive survival patterns and result in resistance to radiation, photodynamic therapy and cytotoxic chemotherapy [1]. Hypoxia imaging may help select the patients who would be most likely to benefit from novel hypoxia-directed therapies and increase our understanding of the role tissue hypoxia plays in tumour biology.

Overview of hypoxia and its importance

In solid tumours, the vascular system fails to supply the rapidly growing tumoural mass with adequate amounts of oxygen, resulting in low oxygen tensions, nutrient deprivation and hypoxia. The major factors in the development of tumour cell hypoxia are structural and functional abnormalities in the tumoural microvasculature [2], increased diffusion distances between blood vessels, growing competition for oxygen between different regions of the expanding tumour cell mass and the reduced oxygen carrying capacity of blood due to disease- or treatment-related anaemia.

Three distinct types of tumour hypoxia can be identified [3]. (1) Acute (perfusion-related) hypoxia results from inadequate blood supply to and within tumours, a consequence of recognised structural and functional abnormalities of the tumour neovasculature. Acute hypoxia is often transient, caused by temporary occlusions and temporary rises in interstitial pressure and can affect vessels both in the vicinity and far from the vessel wall. (2) Chronic (diffusion-related) hypoxia is caused by the increase in diffusion distances of oxygen relative to the supplying blood vessel due to tumour expansion and affects cells at distances greater than 70–100 μm from the nearest capillary. This type of hypoxia also depends on where tumour cells lie in relation to the arterial or venous end of a capillary. (3) Anaemic hypoxia relates to reduced O₂-carrying capacity of the blood and may be tumour associated or treatment related.

Hypoxia measurements have been shown to correlate with the probability of metastatic spread [4], tumour recurrence [5], resistance to chemotherapy and radiation [6-9], invasion [10,11] and decreased patient survival [12,13]. A few studies suggested hypoxia-induced phenotypic changes, such as genomic instability, loss of apoptotic potential, alterations of gene expression, oncogene activation and induction of angiogenesis, are necessary conditions to malignant progression [14-20].

The critical level below which intratumoural partial oxygen pressure (pO₂) changes result in impaired cellular metabolism across tumoural cell types is still hotly debated. Establishing a pO₂ threshold is difficult because experimental findings in cell cultures may not be directly applied to in vivo environments and some of the variation in the published literature can be related to the tumour cell type and the demands of host tissues. The approximate values of critical pO₂ below which cellular functions progressively cease or anticancer treatments become less effective have been established as follows [21]:

• effectiveness of immunotherapy becomes impaired (30–35 mmHg);

• photodynamic therapy (15–35 mmHg);

• cell death on exposure to radiation (25–30 mmHg);

• binding of hypoxia immunohistochemical markers (10–20 mmHg);

• proteome changes (1–15 mmHg);

• genome changes (0.2–1 mmHg).

A large amount of clinical evidence suggests the hypoxia-mediated aggressive behaviour of cancer cells and their resistance to therapy is mediated by the hypoxia inducible factor-1 α (HIF-1α) through numerous molecular pathways required for the adaptation of tumour cells to hypoxia [22]. The emergence of new and more aggressive cell clones capable of overcoming nutrient deprivation and their hostile environments is facilitated by hypoxia-induced adaptations in the proteome and genome of neoplastic cells. Hypoxia initiates the selection of more aggressive cell types that, in turn, results in exacerbation of regional hypoxia, with development of further resistance to chemotherapy and radiotherapy. HIF-1α overexpression is described in a host of human cancers: prostate [23], squamous cell carcinoma [24], lung [25], breast [26,27], bladder [28] and pancreas [29]. Several studies have also demonstrated that increased HIF-1α activity is a predictor of a more aggressive tumour grade, tumour invasion, resistance to radiation therapy, metastatic potential, and is associated with a poorer prognosis [26,30-36]. Cell lines genetically altered to knock down HIF-1α show decreased cell growth both in vivo and in vitro [37-40].

Glioma cell lines in vitro overexpress HIF-1α in both normoxic and hypoxic conditions, so that most malignant glioma cell lines have increased vascular endothelial growth factor (VEGF) and HIF-1α expression at baseline [41-45]. Several other HIF-1α downstream-regulated proteins and their association with cancer is less known, although CA-IX and Glut-1 expression has been reported to correlate with poor response to adjuvant treatments [46-48]. Overexpression of CA-IX and Glut-1 has been established in higher-grade brain tumours [41,49-57] and malignant glioma cell lines [56,58], whereas CA-IX expression in malignant gliomas has been shown to predict radiological response and survival of patients treated with bevacizumab and irinotecan [59]. High-grade gliomas were also more likely to be immunohistochemically positive for HIF-1α, VEGF, Glut-1 and CA-IX than are low-grade tumours [41,49].

Several studies have described genetic instability and amplification, disruption of apoptotic pathways, abnormal oncogene expression and abnormal angiogenesis in high-grade gliomas [15,60-67]. There are several reasons to believe that hypoxia plays a role in high-grade glioma [glioblastoma multiforme (GBM)] development, angiogenesis and growth, the most obvious being the presence of intratumoural necrosis. Animal glioma model studies have shown that while small tumours (<1 mm in diameter) are intensely hypoxic and poorly perfused, large tumours (1–4 mm in diameter) are rich in vasculature and not significantly hypoxic, indicating that tumour necrosis is not simply due to inadequate vascular supply [17,68]. In spite of being highly vascular, the microcirculation in GBMs is inefficient and may contribute to relative hypoxia and necrosis within a given tumour [69-73]. Measurements of intratumoural hypoxia using direct and indirect methods and attempts at correlating this with tumour blood flow and necrosis have not provided an answer to this controversy [74-77]. Several studies have also found that tumour cells found in the areas surrounding the necrotic centres are implicated in hypoxia-regulated migration away from necrotic areas [78-80], suggesting tumour hypoxia results in increased GBM cell migration and possibly invasion [81-83].

Chemotherapy is clinically useful in the treatment of patients with GBM [84,85], and GBM chemoresistance is thought to be hypoxia related [86,87]. Inhibiting hypoxia-regulated survival mechanisms, by stopping the expression of HIF-1α, renders these malignant cells more sensitive to doxorubicin and etoposide [86]. Experiments on malignant glioma cell lines found these cells secrete proteins that suppress hypoxia-induced endothelial cell apoptosis and promote the angiogenic process [88].

Radiation therapy is also extensively used in the treatment of human GBMs [89,90]. Hypoxia-induced radioresistance is multifactorial with the presence of oxygen mediating DNA damage through the formation of oxygen free radicals, occurring after the interaction of ionising radiation with intracellular water [91]. Accumulating evidence shows hypoxia-mediated proteomic and genomic changes are likely to contribute towards tumoural radioresistance by increasing the levels of heat shock proteins (HSPs). HSPs are induced in response to environmental stresses like heat, cold and oxygen deprivation [17] or by increasing the number of tumour cells that can resist apoptosis by mutating a protein brake called p53. A large proportion (50–55%) of human cancers involve disruptions in the function of p53, and these could be, at least partially, hypoxia related. There is also strong evidence hypoxia makes GBM tumoural cells radioresistant and the intensity of hypoxia in GBM before radiotherapy is associated with decreased time to tumour progression or overall patient survival [92-94].

Tissue oxygenation status can be assessed in vivo, using both invasive and non-invasive methods, or in vitro using material from a biopsy. Non-imaging methods of assessing the presence of hypoxia in tissues include histological appearance, immunohistochemical staining for intrinsic markers of hypoxia (e.g. CA-IX and HIF-1α) and for the binding of externally administered nitroimidazoles [91,95-97]. Hypoxic cancer cells, both human and animal, can be labelled using nitroimidazole derivatives, such as EF5 and pimonidazole [98-100]. In a low O₂ concentration environment, nitroimidazole molecules are bioreduced by nitroreductases, bind to cellular macromolecules and are trapped intracellularly [101,102]. EF5- and pimonidazole-specific antibodies can be used to measure these molecules using various immunohistochemical methods [99,100,103,104]. Other methods of measuring intratumoural hypoxia are the use of polarographic O₂ microelectrodes and advanced PET and MRI. Measurement of tumour hypoxia is even more challenging in a clinical setting and there is no easy way to predict its presence or readily measure it. Imaging is one possible way of non-invasively selecting cancer patients who would benefit from treatments that take advantage of the presence of hypoxia. Since tumour hypoxia plays a key role in radioresistance, it has been suggested hypoxia mapping could be combined with radiotherapy techniques to improve target delineation and dose delivery. Imaging could also be used to document the extent to which re-oxygenation of tumours occurs during radiotherapy and the heterogeneity between and within tumours.

An ideal imaging test should:

1. distinguish normoxia, hypoxia, anoxia or necrosis;

2. distinguish between perfusion-related (acute) and diffusion-related (chronic) hypoxia;

3. reflect cellular, in preference to vascular/extracellular, pO₂;

4. be applicable to any tumour site with complete locoregional evaluation;

5. be simple to perform, non-toxic and allow repeated measurements; and

6. be sensitive at pO₂ levels relevant to tumour therapy [105].

The challenge for hypoxia imaging is to measure low levels of tissue pO₂ on a small spatial scale (70–100 μm), a much smaller dimension than can be achieved with current human imaging techniques.

Several methods have been proposed and evaluated for clinical imaging of hypoxia and only a few techniques have potential for in vivo assessment in humans, particularly for repeated, sequential measurements. These methods use either PET tracers [¹⁸F-fluoro-misonidazole (¹⁸F-MISO) and ^60/64Cu-ATSM PET being the most common ones used] or MRI techniques sensitive to variations in local oxygen changes [such as blood oxygenation level dependent MRI (BOLD-MRI) or dynamic contrast-enhanced MRI (DCE-MRI)].

Eppendorf O₂ microelectrodes

Measuring intratumoural hypoxia is a difficult task and the only method for direct and invasive measurements is to use Eppendorf polarographic O₂ microelectrodes (POEs) for spatial mapping of tumour hypoxia in experimental models. This technique has been used in various tumour types to demonstrate tumour hypoxia, but its use was limited in human brain tumour studies [76,104,106-110].

The studies that have shown positive relationships between Eppendorf oxygen probe measurements and treatment outcomes have used different endpoints when making measurements. Estimating the median pO₂ or a range of pO₂ values across a tumour requires many measurements, whereas stratifying patients into diagnostic categories based on median pO₂ requires fewer. One study reported that 20 independent measurements are necessary to classify tumours as oxic or hypoxic and to reduce the proportion of false classifications [111].

The use of hypoxic biomarkers in routine clinical practice has been slow due to the invasive nature of the electrodes and by the inability to specifically target hypoxic tumour regions of variable size and location during the course of treatment [112]. Concurrent HIF-1α expression and pimonidazole staining of U87 glioma cell lines has been demonstrated in vitro but not in tumour xenograft experiments [104,113]. HIF-1α, CA-IX and Glut-1 expression are inversely distributed relative to vascular perfusion and correlate with POE and pimonidazole staining in many tumour types. However, this relationship has not been proven in GBM [47,114-117].

The difficulty in using biomarkers as a clinical tool is the inability to determine which hypoxia-regulated protein or method provides the best "hypoxia marker" [114,118]. There are also questions surrounding which of these molecules and methods can measure "real-time" or "acute" hypoxia as opposed to cumulative cellular hypoxia [104]. The lack of direct correlation between POE measurements, biochemical data and imaging studies (discussed in the next section) has contributed to the confusion of the role of hypoxia in GBM.

In spite of their limitations, Eppendorf probe measurements are commonly considered a gold standard for hypoxia, but in reality, it is difficult to determine which technique is the best. One technique might be preferable in mapping absolute oxygen values in tumours, whereas another could be better for clinical trials testing hypoxia-dependent drug effects. Furthermore, all techniques available to measure hypoxia differ in sensitivity, accuracy and ability to measure oxygen availability. While pO₂ histography and electron paramagnetic resonance (EPR) measure oxygen availability directly, others provide information indirectly through reduced drug levels, haemoglobin saturation or perfusion. Direct measurements rely on measuring collision rates, which depend on oxygen concentration and diffusion, as well as on the type of probe used. Indirect measurements, although valuable, require a set of assumptions to relate the measurement to pO₂ or oxygen concentration. The Eppendorf electrode is primarily an extracellular measurement and does not distinguish between acute and chronic hypoxia or between hypoxic and necrotic tissue.

Positron emission tomography

Ideally, nuclear medicine techniques should allow the evaluation of both chronic and transient hypoxia and reflect cellular pO₂ without contamination by vascular signals. They should be simple, use non-toxic radiopharmaceuticals and be repeatable. They should also provide complete locoregional evaluation with a spatial resolution adequate for assessing the heterogeneity of local pO₂. For the patient's convenience, nuclear medicine techniques should permit imaging on the same day the radiopharmaceutical is administered and short half-life isotopes should be used to minimise radiation exposure.

Human PET imaging of 2-nitroimidazole compounds, such as ¹⁸F-MISO, has been used to identify hypoxic tumours (including gliomas) [106,119-124]. ¹⁸F-MISO is probably the most widely used PET imaging agent for hypoxia. ¹⁸F-MISO can be generated by using a commercially available precursor molecule and is adaptable to existing PET technology for automated radiosynthesis. Clinically, PET ¹⁸F-MISO is a well-tolerated procedure by patients and takes around 20–30 min, starting from 75 to 150 min after injection and making it similar to the bone scan with which most cancer patients are familiar. Radiation dose is low, typically 250 MBq, and no arterial sampling or metabolite analysis is required. The initial ¹⁸F-MISO uptake, within 5 min after injection, correlates tightly with blood flow [125]. After an hour, the F-MISO–plasma ratio is unity for most normoxic tissues, consistent with the partition coefficient of ¹⁸F-MISO. Similar ratios are observed in experiments in mice, rats, guinea pigs, dogs and humans. A tumour-blood ratio of 1.2 has been demonstrated as a reasonable cut-off between normoxia and hypoxia. Hypoxia can be detected in the presence of normal or reduced blood flow, and the absence of hypoxia can be detected with flow that is only one-third normal. ¹⁸F-MISO images can be acquired 1 h after tracer injection, but best contrast will be achieved 90–150 min post injection.

Quantitatively, the degree of hypoxia can be defined by hypoxic fraction or volume, or can be expressed by its severity, defined as the region with the lowest oxygen concentration and its relative level. Calculating the hypoxic tumour fraction (percentage of pixels with values greater than the tumour–blood cut-off of 1.2) requires accurate delimitation of the tumoural volume. PET ¹⁸F-MISO imaging studies cannot provide values of regional oxygen concentration, but the maximum tumour–blood ratio is a convenient surrogate for the worst level of hypoxia in the image. It remains to be established whether a larger hypoxic volume, a more severe hypoxic area or heterogeneity of the hypoxic region is the strongest predictor of response to cancer treatment [91].

PET imaging with ¹⁸F-MISO has several advantages. Early ¹⁸F-MISO distribution reflects blood flow, while later distribution is purely a measure of tissue–plasma partitioning. ¹⁸F-MISO diffuses freely across the blood–brain barrier, is reduced at any hypoxic site where it is not excreted or highly metabolised and its metabolites have a rapid plasma clearance, so normalisation for delivery is not required. ¹⁸F-MISO has no protein binding and imaging is possible within 90–120 min of ¹⁸F-MISO administration. Further advantages are that ¹⁸F-MISO has been validated in several animal models and human disease conditions, and its signal is independent of other factors associated with hypoxia, such as regional glucose concentration, glutathione levels, other transporters and pH [91].

Limitations of ¹⁸F-MISO PET include the low signal-to-noise ratio (SNR) of raw ¹⁸F-MISO PET images, large range of fractional hypoxic volume values across tumour types (making standard correlation of hypoxic volume with that of the tumour difficult) and serial measurements can only be performed more than a day apart, limiting the ability of ¹⁸F-MISO PET to monitor temporal heterogeneity of tumour function. The low SNR can be improved by taking a venous blood sample during the mid-course of the imaging procedure and used to calculate a tumour–blood (T/B) ratio image and then electronically subtracting the normoxic uptake (T/B<1) to increase image contrast.

A recent study in GBM patients imaged pre-operatively using ¹⁸F-MISO showed positive correlations between relative hypoxia (defined as the ratio of the hypoxic volume to the MRI T₂ defined tumour volume) and the net rate of cell proliferation, as well as between the biological aggressiveness ratio (defined as the ratio between the net rate of cell proliferation and the net rate of invasion) and relative hypoxia, scaled to the blood activity of the tracer [126]. ¹⁸F-MISO and ¹⁵O-H₂O PET have also been used in brain tumours to measure tumour hypoxia and perfusion [119]. Increased ¹⁸F-MISO brain tumour retention on a delayed scanning time was found predominantly in GBM (7 out of 7) and this was associated with an increased ¹⁸F-MISO tumour distribution volume, which was also used as a quantitative criterion for hypoxia. ¹⁸F-MISO accumulated in both hypo- and hyperperfused tumour regions, suggesting hypoxia in glioblastoma may develop irrespective of the magnitude of perfusion (Figure 1).

Figure 1.

(a–c) ¹⁵O-H₂O positron emission tomography (PET) perfusion images in three patients with glioblastoma multiforme. (d–f) Corresponding late ¹⁸F-labelled fluoro-misonidazole (¹⁸F-MISO) PET images show tumour hypoxia in low perfusion, in intermediate perfusion with an inverse pattern compared with hypoxia and in high perfusion. PET images are normalised to their own maximum. Reproduced with permission from [119].

¹⁸F-MISO PET was also evaluated in advanced head and neck cancer during hypoxia-targeting therapy. All sites of corresponding fluorodeoxyglucose (FDG) and ¹⁸F-MISO abnormality at baseline showed marked qualitative reduction of uptake within 4 weeks of commencing therapy, consistent with effective hypoxia-targeted therapy. The high prevalence of hypoxia demonstrated on ¹⁸F-MISO PET imaging was consistent with the advanced disease stage of these patients and predicted adverse prognosis.[127] Another study [128] showed ¹⁸F-MISO uptake was present in all high-grade gliomas but not in low-grade gliomas (LGGs). In the same study, a significant relationship was found between FDG or ¹⁸F-MISO uptake and expression of vascular endothelial growth factor (VEGF)-R1 and Ki-67 expression. The authors concluded that ¹⁸F-MISO PET provides a non-invasive assessment of hypoxia in glioma and is prognostic for treatment outcomes in the majority of patients. There was a correlation between ¹⁸F-MISO uptake and tumour grade, and all high-grade lesions showed uptake that was frequently heterogeneous. There was only partial overlap between regions of ¹⁸F-MISO uptake and FDG uptake (Figure 2).

Figure 2.

High-grade glioma (grade IV). (a) ¹⁸F-labelled fluoro-misonidazole (¹⁸F-MISO) positron emission tomography (PET). (b) ¹⁸F-fluorodeoxyglucose (FDG) PET. (c) MRI. ¹⁸F-FDG and ¹⁸F-MISO uptake is evident in the left posterior parietal glioma, with uptake absent in the central necrotic area, although different patterns of maximal glucose metabolic rate and hypoxia are evident. Reproduced with permission from [128].

Although promising, the above-mentioned studies need to be extended to larger patient trials with serial imaging and follow-up to establish the full clinical potential of ¹⁸F-MISO PET in gliomas. Possible applications could include pre-surgical planning to identify the regional distribution of hypoxia and accurate targeting of higher doses of radiotherapy to hypoxic areas of the tumour.

¹⁸F-MISO autoradiography correlates with pimonidazole (cellular hypoxia marker) studied by immunohistochemistry in human xenograft tumours [129]. In GBMs imaged by ¹⁸F-MISO PET prior to radiotherapy, the volume and intensity of hypoxia showed a strong association with poorer time to tumour progression and overall survival. New treatment strategies to target hypoxia more aggressively in GBM could benefit from ¹⁸F-MISO PET imaging, which could also be applied to assess treatment outcomes [94]. Because ¹⁸F-MISO is not commonly available at all institutions and has a relatively slow clearance from normoxic tissues, there has been an intense search for newer radiotracers for hypoxia imaging.

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

⁶⁴Cu-ATSM Cu-diacetyl-bis(N4-methylthiosemicarbazone) (Cu-ATSM) is a promising new PET agent for delineating the extent of hypoxia within tumours. Cu-ATSM is retained in hypoxic cells at a higher level than ¹⁸F-MISO [130]. Cu-ATSM enters and exits normoxic cells metabolically unchanged, but in hypoxic cells undergoes alteration and is retained. It is not retained in necrotic tissues. The mechanism of retention of Cu-ATSM in hypoxic tissues is attributed to the low oxygen tensions and the subsequent altered redox environment of hypoxic tumours [increased nicotinamide adenine dinucleotide dehydrogenase (NADH) levels]. Cu-ATSM images hypoxia both in acute ischaemic syndromes where flow is limited and in instances when ischaemia is induced without severe restrictions in flow. With the short-lived copper-PET radionuclides, imaging can be repeated at 120 min intervals, enabling investigators to perform reperfusion studies [91].

Cu-ATSM was validated as a hypoxia PET marker by comparing autoradiographic distributions of Cu-ATSM with a well-established hypoxia marker drug, EF5, in animal R3230 mammary adenocarcinomas, fibrosarcomas and 9L gliomas. It was shown that Cu-ATSM is a valid PET hypoxia marker in some tumour types, but not for all; this tumour type-dependent hypoxia selectivity of Cu-ATSM challenges the use of Cu-ATSM as a universal PET hypoxia marker [131]. Cu-ATSM has a higher uptake in hypoxic tissues than the nitroimidazoles, allowing for high-quality images as soon as 20 min after injection [130,132,133]. ATSM is also flexible in that it can be labelled with any of several positron-emitting copper radioisotopes. Cu-ATSM has been used in a small number of human clinical trials, but no applications in human gliomas have yet been reported [134-136].

Other radionuclide agents

Recently, several other PET agents have been assessed for tumour hypoxia imaging: ^99mTc- and ⁶⁸Ga-labelled metronidazole (MN), ^99mTc-labelled iminodiacetic acid (IDA) derivative of 2-methyl-5-nitroimidazole and 1-(2-[¹⁸F] fluoro-1-[hydroxymethyl]ethoxy)methyl-2-nitroimidazole, a ¹⁸F-labelled 2-nitroimidazole analogue [137-139]. The last of these agents, 1-(2-[¹⁸F]fluoro-1-[hydroxymethyl]ethoxy)methyl-2-nitroimidazole (FRP-170), was used to visualise hypoxic tissues in eight patients with glioma (three GBM, two oligodendroglioma and one each with diffuse astrocytoma, anaplastic ganglioglioma, and recurrent anaplastic astrocytoma). The FRP-170 PET images showed marked uptake with upregulation of HIF-1α in the three GBM, and moderate uptake in the recurrent anaplastic astrocytoma and one oligodendroglioma, but no uptake in the other tumours. The FRP-170 PET images showed positive correlation with HIF-1α immunoreactivity and some correlation with FDG PET and MRI enhancement (Figure 3), but no correlation with [¹¹C]-methionine PET. Imaging with FRP-170 PET also seemed to be more sensitive for detecting hypoxia than identifying the lactate peak on proton MR spectroscopy [139].

Figure 3.

(a–c) A 68-year-old man with a glioblastoma multiforme (GBM). Axial T₁ weighted gadolinium-enhanced MR image (a) showing ring enhancement in the left insula, FRP-170 positron emission tomography (PET) image (b) showing marked uptake, and photomicrograph with hypoxia inducible factor-1α (HIF-1α) antibody (c) indicating strong immunoreactivity. Axial T₂ weighted MR image (d) showing a diffuse infiltrative lesion in the left frontal lobe, FRP-170 PET image (e) showing moderate uptake within the lesion, and photomicrograph with HIF-1α antibody (f) showing moderate immunoreactivity. Reproduced with permission from [139].

All these agents are in their initial phases for research applications and none of them have been used in routine clinical imaging.

Magnetic resonance imaging

Advanced MRI techniques can be used to measure tumour perfusion and indirectly provide insights into hypoxia and angiogenesis in malignant tissues [37,140-147]. Several MRI techniques are available, including DCE-MRI, which allows measurements of microvascular properties, magnetic resonance spectroscopy (MRS), which measures metabolite ratios, MR diffusion imaging and blood oxygen level-dependent MRI [148]. DCE-MRI uses paramagnetic contrast agents injected into the bloodstream and pharmacokinetic models of tumour contrast uptake over time to measure tumour perfusion and vascularity [149-155]. Several studies have shown DCE-MRI can be used to determine intratumoural blood flow and predict vascular permeability [143,147,149,153]. Recently, advances in MR technology have enabled complex acquisitions of multiparametric data in gliomas, allowing complementary information to be obtained from the above-mentioned MRI techniques [148].

Blood oxygenation level dependent MRI

BOLD-MRI, also known as intrinsic susceptibility-weighted MRI, is a non-invasive MRI technique to indirectly measure oxygenation changes in tissues. In BOLD-MRI tissue contrast is affected by the intrinsic tissue properties (spin–lattice and spin–spin relaxations) and also by blood flow and paramagnetic deoxyhaemoglobin within red blood cells (in contrast to oxyhaemoglobin, which is not paramagnetic).

Deoxyhaemoglobin increases the MR transverse relaxation rate (R₂*), the inverse of the transverse relaxation time (T₂*), of water in blood and surrounding tissues. Thus, BOLD-MRI is sensitive to changes in pO₂ within vessels and in tissues adjacent to perfused vessels [156]. Susceptibility weighted images are also affected by iron content (e.g. myoglobin found in muscle), blood flow, carbon dioxide tension, haematocrit, pH and the presence of fibrosis or ligamentous structures (e.g. in benign prostatic hyperplasia and the suspensory ligaments of the breasts). To distinguish between the effects of flow from deoxyhaemoglobin and static tissue components, the T₂* relaxation rate (R₂*=1/T₂*) has to be measured. This can be done by using a multi-echo GRE sequence, sensitive to variations in the properties of the local magnetic field. Decoupling of flow from static effects on R₂* images occurs because the flow component can be thought of as affecting individual T₂* images of a multigradient echo sequence. Although synthetic R₂* images are free of blood flow contribution, reflecting mainly deoxyhaemoglobin content and static tissue components, improved blood flow and vascular function will also increase tissue oxygenation, which can be seen by changes in R₂* images.

BOLD-MRI images are more likely to reflect acute (perfusion-related) tissue hypoxia, due to transient occlusions of small blood vessels, while chronic hypoxia is less likely to be reflected by BOLD-MRI because of a greater distance between the red blood cells in the vessels and the area of hypoxic tissue. BOLD-MRI can only provide accurate information on tissue oxygenation if red blood cells are delivered to the tissue of interest. Human and xenograft studies have shown tumour perfusion varies widely and capillary blood delivery is not simply related to the absence/presence of vessels [157]. This observation could explain, in part, why no direct correlations between baseline R₂* and tissue pO₂ have been observed. Tissue oxygenation status can be accurately interpreted from R₂* images only if the distribution of blood volume in the tissue is known or experimentally determined. Thus, if a tissue is perfused and has a homogeneous blood volume fraction but has a high baseline R₂* in one area compared with another area in the same tissue, it can be inferred the high R₂* region is relatively more hypoxic; this hypothesis is supported by recent pre-clinical and clinical data [158].

The use of BOLD-MRI to evaluate tissue hypoxia is based on the assumption that haemoglobin oxygenation is proportional to blood arterial pO₂ which, in turn, is in equilibrium with the tissue oxygenation. Although changes in R₂* in response to vasomodulation with carbogen (95% CO₂:5% O₂) inhalation are temporally correlated with changes in tissue pO₂, only 50–60% of human tumours show changes in R₂* after carbogen inhalation [159,160]. These limited and heterogeneous responses can partially be attributed to different perfusion patterns within tumours and even when vessels are present, red blood cell transport along intratumoural vessels may not be effective [157]. Hypoxic tumours with high blood volume (due to high microvessel density coupled with large vessels) will not only have raised baseline R₂* values but are more likely to respond to carbogen, with subsequent large changes in R₂* and positive radiosensitisation. Hypoxic tumours with low blood volume (due to lower microvessel density or small vessels) will have lower baseline R₂* values and are less likely to respond to carbogen, with negligible changes in R₂* and radiosensitisation with carbogen [161]. In these cases, BOLD response to carbogen is also dependent on the maturity of the underlying vasculature, with mature blood vessels being able to respond more actively to vasoconstrictory and vasodilatory stimuli [162].

The advantages of the BOLD-MRI technique are no need for externally administered contrast media or radioactive isotopes, easily repeatable and near real-time visualisation of time-dependent changes. BOLD-MRI is sensitive to the presence of changing vascular oxygen tension in the tissue and flow dependence of the MRI signal can also be decoupled. Major limitations of BOLD-MRI include the technique doesn't measure tissue pO₂ directly (either in blood or tissues because of a non-linear relationship of R₂* and tissue pO₂), the images obtained have low SNR and clinical studies with carbogen vasomodulation are technically challenging (approximately 25–35% of patient examinations fail due to respiratory distress caused by an increased respiratory drive induced by carbogen) [159,160]. BOLD-MRI appears most sensitive to oxygen levels adjacent to perfused vessels (perfusion related or acute hypoxia) and BOLD-MRI sensitivity to more distant diffusion related (chronic hypoxia) is unknown.

Several studies explored the relationship between the BOLD-MRI signal and tumour oxygenation status in animal models and patients with gliomas. A recent study in 9L and CNS-1 intracranial rat tumour models, before and during carbogen breathing, indicated glial sarcomas may be radiobiologically hypoxic, as evaluated by concurrent EPR oxymetry and BOLD-MRI measurements, and both techniques could reliably detect increases in brain tumour oxygenation after carbogen inhalation [163]. Hypoxic behaviour of cerebral gliomas has also been identified using BOLD-MRI in a small number of patients (six low grade, one high grade) during breath-holding [164]. The authors noted an absence of significant increase in the BOLD signal in both low-grade and high-grade gliomas, explained by either overwhelming hypoxia within the tumour, inadequacy or absence of hypercapnia-induced vasodilatation of tumour vessels, or both. Breath-hold regulated decreases in BOLD signals occurred only in the high-grade glioma, which is most likely due to the hypercapnia-induced steal effect that redistributes blood flow from tumour regions with unresponsive neovasculature to surrounding normal tissue. Another study looked at tumour heterogeneity and carbogen response in four patients with GBM and one patient with grade II astrocytoma [165]. The glioblastomas showed strong but heterogeneous signal changes between carbogen and air breathing, especially in the peritumoural areas corresponding to high-signal regions on T₂ weighted images, while the astrocytoma displayed a signal decrease during carbogen breathing in the peritumoural regions. The authors concluded that BOLD-MRI provides high-resolution images of cerebral anatomy and venous vascularisation and the technique, combined with hypercapnia challenge, allows for regional assessment of brain tumours.

Other studies have shown malignancies in the brain (meningiomas, gliomas or metastasis) can be distinguished from normal brain tissues using BOLD-MRI and in older patients with grade IV gliomas, BOLD signal intensity is equivalent to that measured in younger patients with grade IV gliomas, whereas in patients with grade II and III gliomas, BOLD-MRI signal change correlated only with increasing age [166,167].

Oxygen-enhanced MRI

Since molecular oxygen is paramagnetic, elevated concentrations of oxygen dissolved in blood, plasma and tissue fluid increase the longitudinal relaxation rate (R1=1/T₁) via dipolar interactions. This effect has been quantified on T₁ weighted imaging in studies of normal tissue [168-173], emphysema [174] and tumours [175]. In contrast, T₂* weighted sequences are sensitive to changes in oxygenation status by an entirely different mechanism. Here, paramagnetic deoxygenated haemoglobin, compartmentalised in red blood cells, creates local variation in magnetic susceptibility, increasing the effective transverse relaxation rate (R₂*=1/T₂*) of vascular and perivascular water. The effects of breathing carbogen (95% oxygen and 5% carbon dioxide) and other hyperoxic hypercarbic gases on R₂* have been evaluated in normal tissue [159,171,176] and tumours [159,172].

An oxygen-induced change in R1 is thought to reflect chiefly the amount of supraphysiological levels of dissolved oxygen in plasma and tissue fluid. Changes in arterial blood volume may modulate this effect, especially when carbogen is inhaled [172,177]. A marked reduction in R₂* has been reported in several studies of diseased vasculature on carbogen inhalation, which may be dependent on vessel calibre [157], blood flow and blood volume effects [156,177]. In distinction, breathing 100% oxygen has been reported to have a negligible effect on R₂* in normal tissues outside the brain [171]. In normal tissue, significant reductions in R1 were demonstrated following both oxygen and carbogen inhalation in the spleen, liver, skeletal muscle and renal cortex. No significant changes in R₂* occur with oxygen alone. In patients with advanced cancer in the abdomen and pelvis an oxygen inspiration challenge showed significant increases in R1 in the majority of cases [178]. Comparison with regional perfusion data showed congruence in most areas of the tumour, but some areas of mismatch with high levels of perfusion were not associated with significant changes in oxygen concentration. These observations support the use of oxygen-enhanced imaging as a biomarker for tumour oxygenation, although the relationship between the signal changes resulting from variations in dissolved oxygen pressure and true tumour hypoxaemia remain to be elucidated. Oxygen-enhanced imaging studies have not been applied in human brain tumours to date; however, studies in orthotopic tumours in the mouse brain showed close agreement between R₂* and R1 changes in response to oxygen inhalation [179].

Diffusion-weighted MRI

A study using diffusion MRI, an MRI technique that measures the mobility of water within tissues at a cellular level, in patients with high-grade gliomas, concluded that functional diffusion map analysis can provide a meaningful early assessment of treatment response in patients with high-grade gliomas [180]. Recently, bevacizumab-induced diffusion-restricted lesions have been reported in 18 patients with recurrent malignant gliomas before and after exposure to this drug [181]. These lesions were detectable as early as 4 weeks after initiation of therapy and were maintained for up to 80 weeks. Within the tumour bed, bevacizumab-induced diffusion-restricted lesions in the presence of reduced rCBF and rCBV. Although the cause of these alterations is unclear, it may involve atypical necrosis and chronic hypoxia. However, the relationship between hypoxia in gliomas and diffusion-weighted imaging measurements has not been explored in detail.

Dynamic contrast-enhanced MRI

DCE-MRI, using contrast agents of low molecular weight, may be used to estimate tumour parameters such as local perfusion and extracellular–extravascular volume [182]. Contrast agents generally contain chelates of gadolinium (Gd), which is an effective inducer of magnetic T₁ relaxation and thus induces an increase in signal intensity under T₁ weighted conditions.

Tumour vessels are highly disorganised, tortuous and dilated, uneven diameters, with biochemical and empirical concordance between vessel permeability and hypoxia [183,184]. Imbalances of angiogenic mediators such as VEGF and angiopoietins may contribute to this extensive branching and shunting. The excessive leakiness or hyperpermeability of tumour vessels is in part attributed to cytokines and angiogenic factors, which dynamically alter the structure of the microvessel walls [185].

It has been suggested DCE-MRI may be used as a surrogate for invasive pO₂ measurements [186]. Functional parameters derived from DCE-MRI have been shown to be related to pO₂ levels in murine tumours and correlate with microvessel density and tumour oxygenation in human cervix carcinomas [187-190]. Also, quantities derived from DCE-MRI have displayed a predictive value when evaluating the treatment response following radiotherapy of cervix cancers [191-193]. Furthermore, functional DCE-MRI images have been integrated in radiotherapy treatment planning in combination with morphological CT images [194].

¹H-magnetic resonance spectroscopy

¹H-magnetic resonance spectroscopy can provide useful metabolic information for brain tumour diagnosis and glioma grading [195,196]. Two metabolites, lipids and lactates, both of which are present on short- and long-echo time (TE) sequences, have been studied as hypoxia markers. Lipids, which are related to necrosis, also represent an early response to cellular hypoxic stress [197-199]. The relationship between lipids and glioma grade is well established [196,200]. The significance of lactates, which are the end product of anaerobic metabolism, in grading is more controversial [195,200-203]. Either lipid or lactate signals have been identified in hypoxic areas and have been correlated with the rCBV in malignant gliomas [204]. Interestingly, in short-TE acquisitions, which are more able to detect mobile lipid droplets, some LGGs demonstrated small lipid peaks [195,200,205,206]. To date, the predictive value of such peaks in LGGs for early angiogenesis switching and poor clinical outcome remains to be established, mainly by multivoxel sequences that can demonstrate glioma spatial heterogeneity.

Conclusions

Hypoxia probably plays a role in the development, angiogenesis and growth of malignant brain tumours; its impact in benign tumours is less clear. More research, especially regarding methods of measuring tumour hypoxia directly or—even more attractively—indirectly through imaging modalities, is necessary. Hypoxia is a logical therapeutic target for selective GBM because there are several ingenious methods to exploit tumour hypoxia available, but much more work will be required. An understanding of the role of hypoxia in tumour development and growth is important for physicians involved in the care of patients with brain tumours.