|Home | About | Journals | Submit | Contact Us | Français|
Hypoxia has long been recognized to influence solid tumor response to therapy. Increasingly, hypoxia has also been implicated in tumor aggressiveness, including growth, development and metastatic potential. Thus, there is a fundamental, as well as a clinical interest, in assessing in situ tumor hypoxia. This review will examine diverse approaches focusing on the pre-clinical setting, particularly, in rodents. The strategies are inevitably a compromise in terms of sensitivity, precision, temporal and spatial resolution, as well as cost, feasibility, ease and robustness of implementation. We will review capabilities of multiple modalities and examine what makes them particularly suitable for investigating specific aspects of tumor pathophysiology. Current approaches range from nuclear imaging to magnetic resonance and optical, with varying degrees of invasiveness and ability to examine spatial heterogeneity, as well as dynamic response to interventions. Ideally, measurements would be non-invasive, exploiting endogenous reporters to reveal quantitatively local oxygen tension dynamics. A primary focus of this review is magnetic resonance imaging (MRI) based techniques, such as 19F MRI oximetry, which reveals not only hypoxia in vivo, but more significantly, spatial distribution of pO2 quantitatively, with a precision relevant to radiobiology. It should be noted that pre-clinical methods may have very different criteria for acceptance, as compared with potential investigations for prognostic radiology or predictive biomarkers suitable for use in patients.
Hypoxia and tissue oxygenation comprise a vast field of study relevant to diverse diseases. This review focuses on its relevance to solid in situ tumor development, progression, and primarily response to therapy. Studies range from molecular biology to therapeutic oncology with a common thread of imaging based on developments in reporter molecules, physics, engineering and data processing.
Ever since the classic studies of Gray et al. fifty years ago (1), it has been appreciated that hypoxia can influence the efficacy of radiotherapy on tumors. There have been many attempts to modulate tumor hypoxia in order to enhance radiotherapy, but translation to the clinic has shown marginal efficacy (2). The meta analysis by Overgaard et al. (3) of some 10,000 patients indicated a clinical benefit for manipulating tumor hypoxia. However, the overall conclusion was that there was a pressing need to identify those tumors (viz. patients), who would actually benefit.
In terms of hypoxia in human tumors, we perceive three needs: 1) to develop an effective method of detecting hypoxia non-invasively in tumors of patients prior to therapy; 2) to assess the ability to modulate tumor hypoxia to overcome resistance to therapy; 3) to exploit hypoxia using specific individualized patient therapies, which are effective on hypoxic cells. We believe we are at a historic juncture, where we not only have technologies for identifying hypoxia, but more importantly methods of tailoring therapy successfully to accommodate or exploit the killing of hypoxic cells.
One aspect of pre-clinical research is simply to develop methods satisfying the clinical requirements. However, for pre-clinical applications themselves, we perceive slightly modified needs, firstly: effective methods of assaying tumor hypoxia, and secondly: well-characterized animal models with tumors exhibiting well-defined differential levels of hypoxia and responses to interventions. Criteria for useful methods are also somewhat different, e.g., imaging methods may rely on reporter molecules, which are efficient and ethical for use in small animal studies, but do not have, and might not meet, requirements for patients (notably, IND). The requirements are not necessarily less stringent.
Early work on effects of hypoxia examined cells in vitro, where ambient oxygen concentrations are readily controlled. In vivo, hypoxia may be achieved by clamping the blood supply to a tumor (4), but other levels of oxygenation reflect the interplay of supply and consumption (5). Given the importance of hypoxia many methods have been developed for probing tumor oxygenation including hypoxia reporter agents, polarographic electrodes, fiber optic probes, NIR spectroscopy, and various NMR techniques, but most are either highly invasive, lack spatial resolution or lack the ability to dynamically and quantitatively sample the response to intervention (2, 6–8) (Table 1).
The concept of hypoxia may be debated, but presumably indicates oxygen insufficiency. Anoxia is presumed to be complete absence of molecular oxygen and this may be achieved in model systems and cell cultures. Indeed, cell culture incubators are available to exclude and eliminate oxygen, as required for the culture of facultative anaerobes. More commonly, cell cultures are saturated with a gaseous atmosphere of specific oxygen concentration and cells are adapted to growth under low pO2 saturation.
In the heart and brain, hypoxia may be attributed to ischemia, where the delivery of oxygen is insufficient to maintain normal function. Ischemia may be total, or partial, and it may manifest itself as chest pain, angina or even be silent for many years, while chronic damage is established. Hypoxia may occur due to local lack of vascular flow or insufficient hemoglobin saturation. An acute myocardial infarct will result in rapid oxygen depletion and loss of mechanical function. We have shown, as expected, using 19F NMR of perfluorocarbons sequestered in the myocardium that hypoxiation is much more rapid in a beating heart than an arrested heart, and indeed, the rate of loss of developed pressure is directly proportional to decrease in pO2 (9).
However, in tumors, a definition of hypoxia may be far more complex, since the tissue is aberrant with no particular normal function, yet well documented ability to adapt to stress. Hypoxia has been shown to stimulate mutations, and hypoxic stress appears to stimulate angiogenesis and metastasis, particularly, as a result of intermittent bouts of hypoxic stress (10–12). Tumor cells may survive in quite hypoxic conditions, but for a solid tumor to grow beyond a few hundred microns in diameter, an angiogenic switch is required to generate new blood vessels for the delivery of nutrients and oxygen (13). In hypoxic conditions, tumor cells release angiogenic factors stimulating blood vessel sprouting and a new vascular network. Figure 1 shows typical chaotic vascular extent, and thus tumors are expected to be hypoxic. In terms of Figure 1d, it is noteworthy that externalization of phosphatidylserine (PS) to the luminal surface of the vascular endothelium is stimulated by stress including hypoxia (14). Consequently, phosphatidylserine is a marker of tumor vasculature and a potential target for anti-cancer drugs. The exposed PS is one of the most specific markers of tumor vasculature yet discovered and provides an excellent marker for imaging with radionuclide-bavituximab conjugates (15).
The expression of many proteins is influenced by hypoxia (16, 17) and hypoxia sensitive promoter regions have been developed to generate hypoxia responsive reporter genes. A specific promoter regulating a reporter gene, such as luciferase or fluorescent protein (e.g., eGFP), can selectively turn on expression under hypoxic conditions. Often, multiple promoter copies are exploited to avoid expression leakage and an oxygen degradation domain may be incorporated (18). This approach requires transfection of cells and reporter proteins may remain for several hours after amelioration of hypoxia, preventing acute dynamic studies. This approach is particularly suitable for observing onset of chronic hypoxia (19). In small animals, optical imaging has the advantages of being cheap, generally simple to implement on commercial instruments, and allows high throughput. Fluorescent proteins may be detected non-invasively, although depth of light penetration limits studies to small animals or surface tissues. Luciferase does require administration of luciferin substrate.
It has long been appreciated that hypoxic tumor cells are more resistant to radiotherapy (1). Indeed, a threefold increase in radio-resistance may occur when cells are irradiated under hypoxic conditions compared with pO2 > 15 torr for a given single radiation dose. However, recent modeling has indicated that the proportion of cells in the range 0 – 20 torr may be most significant in terms of surviving a course of fractionated radiotherapy (20). The clinical significance of hypoxia for radiotherapy was recently illustrated by modeling the radiation response of a heterogeneous cell population. Based on an oxygen enhancement ratio of 3.0, Fowler et al. (21) demonstrated that while a fractionated dose of 70–80 Gy would lead to reduction in cell survival from 1 to 10−12 for a well-oxygenated tumor, a similar reduction in survival would require a dose exceeding 200 Gy for a tumor containing 20% hypoxic cells. In the clinical setting, we expect hypoxia to be even more important for high dose fractionation protocols, which are gaining feasibility based on precise treatment plans and effective targeting. Timmerman et al. (22) recently reported results of The Radiation Therapy Oncology Group (RTOG) protocol RTOG 0236 Phase II trial utilizing stereotactic body radiation therapy (SBRT) with ablative prescription dose to treat early-stage medically inoperable non–small-cell lung cancer (NSCLC) patients. Using three doses of 20 Gy in less than two weeks, the estimated 2-year survival exceeded 90%, which is a remarkable improvement over traditional therapy. When few doses are applied, the hypoxic fraction is expected to have much greater impact on radioresistance, since there is no opportunity for the progressive reoxygenation and repopulation attributed to long-term fractionated irradiation and thus methods of assessing hypoxia are increasingly relevant.
When hypoxia is identified several options arise: i) “dose-painting”, booster regimens and IMRT have been proposed and are being tested (23–26), ii) administering adjuvant hypoxic cell selective cytotoxins: clinical trials with tirapazamine were halted, but new, potentially, more effective drugs are being developed, e.g., AQ4N and TH302 (27, 28); iii) a simpler intervention would be breathing hyperoxic gas to ameliorate the hypoxia. Indeed, many clinical trials have explored oxygen, carbogen or hyperbaric gas breathing, but with limited success, likely due to the inability to identify those patients with hypoxic tumors, who would benefit. By analogy, assaying HER2/neu status is vital to the effective use of Herceptin (Trastuzumab) to treat breast cancer patients. Thus, a crucial concern is identifying those patients, who may need the adjuvant intervention and benefit and there is an urgent need to develop methods of indentifying potentially resistant hypoxic tumors and implementing adjuvant interventions. We note the report from Nijmegen (29), which showed that patients with large head and neck hypoxic tumors (identified using pimonidazole in biopsies) could be brought into the realm of effective response by applying the ARCON protocol (Accelerated Radiation therapy with Carbogen and Nicotinamide).
Given the importance of hypoxia diverse classes of reporter molecules have been developed to reveal hypoxia, e.g., pimonidazole (29, 30), EF5 (31, 32), CCI-103F (33), Cu-ATSM (34), galactopyranoside IAZA (35). Many reporters are based on the structure of misonidazole, which had been explored extensively as a radiation sensitizer. Unacceptable toxicity (neuropathy) halted clinical applications, but it was realized that reporter molecules could be used at much lower concentrations, e.g., 15 μg as opposed to 10 g per patient (17) allowing effective application. Indeed, beyond pre-clinical applications EF5, FAZA, Cu-ATSM and F-MISO are being evaluated in ongoing clinical trials (http://clinicaltrials.gov/).
Following IV infusion, the agents diffuse into tissues, where they are reduced forming highly reactive intermediates. In the presence of oxygen, they are reoxidized and ultimately clear from the body. However, in the absence of molecular oxygen, further reductions occur yielding radical anions and reactive alkylating amine derivatives, which are trapped in the cell (17). In biopsies, EF5 and pimonidazole are both routinely used in conjunction with fluorescent immunohistochemical analysis providing microscopic indications of local hypoxia (e.g., Fig. 1c shows hypoxia in a rat breast tumor distant from the well perfused regions) (36). EF5, pimonidazole, and Cu-ATSM are currently being tested in clinical trials and correlations have been reported relating uptake to clinical outcome (29, 32, 37). Over the past 20 years, many variant reporter molecules have been proposed focusing on modifying lipophilicity, reduction potential and reactivity. MRI has been applied to several fluoronitroimidazole derivatives including EF5 and SR-4554 (38, 39).
Incorporation of radionuclides has facilitated non-invasive investigations using PET, SPECT or γ-scintigraphy, as reviewed extensively by others recently (8, 17, 40–42). The most widely used reporter has been fluoromisonidazole (F-MISO), which has been tested extensively in animals (Figure 2) and is being evaluated in clinical trials. F-18 has a relatively short half-life (110 mins), but an efficient synthesis was developed based on 18F-fluoride and a pure racemic mixture suitable for injection into patients can be produced in 35 minutes with a specific activity of 250 GBq/μmol (17). Thus, a typical patient dose of 260 MBq requires only 0.2 μg. Patients are normally imaged 1.5 hours after administration.
Early autographic studies based on tritiated F-MISO showed effective penetration into cultured cell spheroids with greatest concentration being at the interface of viable cells and the necrotic center. Under anoxic conditions, binding was found to be 25-fold greater than in normoxic conditions, where binding dropped to 40% at 4 mmHg (torr). In vivo, there is concern that distribution and uptake may be influenced by relative blood flow, which is likely to be particularly low in hypoxic regions. Correlative studies with the blood flow indicator 14C-iodoantipyrine showed no correlation between flow and F-MISO binding (17). Krohn et al. (2) state that a tumor to background ratio less than unity is an indication of normoxic tissues, while values greater than 1.2 infer hypoxia. However, slow washout is a concern and alternative fluorinated derivatives have been evaluated, particularly exploring differential lipophilicity based on degree of fluorination (Table 2) (17). FAZA showed much faster clearance from the blood of mice and while tumor uptake was less, the tumor to blood was significantly greater than for F-MISO (43).
Alternate agents use iodoazogalactosides or copper lactams and each is undergoing clinical trials (Table 2). I-124 allows PET and the longer half-life allows imaging after 24 to 48 hrs when background has cleared. Several studies have shown efficacy of this approach, but the signal to noise appears inferior to F-MISO (Fig. 2) (44). Iodine labeling has the added advantage of allowing several isotopes for different applications including PET (124I), SPECT (123, 125I), γ-scintigraphy (125I) and radioimmunotherapy (131I), although instability and dehalogenation can be problematical. Copper based agents are attractive, since they also allow multiple isotopes with longer half lives (60Cu t½ = 23 mins, 93% β+; 61Cu t½ = 3.3 hrs, 61% β+; 62Cu t½ = 9.7 mins; 64Cu t½ = 12.7 hrs, 17.4% β+ for PET; and 67Cu t½ = 61 hrs for therapy) (45). Cu-ATSM shows specificity for hypoxic tissues and has shown correlation with clinical outcome for lung and cervical cancer patients (34, 37). However, uptake and retention appears to be inconsistent in some cells and may be influenced by protein expression related to multi drug resistance (46, 47). Early PET studies had used 60Cu-PTSM, but recently Lewis et al. (48) showed that 64Cu-PTSM was equally applicable in patients with cervical cancer and indeed the image quality was reported to be better (Figure 3). However, differences in image quality may be attributed to some variation in data acquisition including 2D versus 3D to avoid contamination from daughter nuclides. Copper-64 is increasingly available from commercial sources to generate such labels, e.g., MDSNordion (Canada), ACOM(Italy), Trace Life Sciences (United States), IBA Molecular (United States and Europe), and IsoTrace (United States). The longer half-life promises a more convenient radiopharmaceutical. Beyond nuclides of iodine and copper, one can envision an arsenic nitroimidazole, where the isotopes 72, 74 and 77As provide diverse opportunities for PET and radiotherapy (15, 49).
18F-misonidazole shows considerable promise in clinical trials, though the rapid decay of F-18 is inconvenient. There have been attempts to correlate the more standard fluorodeoxyglucose (FDG) uptake with hypoxia, but correlations are often poor (50).
Nuclear medicine approaches are costly and less suitable for repeat investigations. Generally, only a single time point is investigated and the measurement is thought to reflect chronic hypoxia, rather than acute hypoxia. For PET, each reporter provides identical signal and thus repeat doses or tandem reporters cannot be readily differentiated except perhaps based on differential decay rates of separate isotopes. Dynamic variations in hypoxia have been assessed, based on biopsy specimens, by applying pairs of hypoxia reporters in a pulse chase fashion with respect to an intervention, as shown by Ljungkvist et al. (33). NMR spectroscopy of 19F-misonidazoles potentially allows greater versatility, since multiple different reporters could be detected simultaneously based on separate resolved chemical shifts, but it must be noted that NMR requires much higher concentration of reporter molecule (9). An alternative approach is to exploit an MRI reporter, which has physical characteristics influenced by oxygen and detectable by relaxometry as described in the next sections (6).
While hypoxia may have a somewhat arbitrary definition, oximetry provides quantitative assessment of pO2. In turn pO2 can define hypoxic thresholds such as fraction of tumor less than 10 torr (HF10) or HF5 < 5 torr. It should be noted that various scales are used to report pO2 [760 torr = 1 atm = 101 kPa], and that pO2 is related to the concentration of dissolved oxygen [O2]. It is important to consider the location of the pO2 measurement, since there are steep gradients from vasculature to interstitium, cytosol and mitochondrion as emphasized by Swartz (51). Ultimately, a parameter must correlate with therapeutic outcome, be reproducible and simple to implement and interpret.
To date, the most extensive evidence for hypoxia in experimental solid tumors is provided by polarographic needle oxygen electrodes. Robust fine needle polarographic electrodes opened the possibility of measuring pO2 in tumors in situ in vivo to define local pO2 under baseline conditions or with respect to interventions. Cater and Silver (52) showed the ability to monitor pO2 at individual locations in patients’ tumors with respect to breathing oxygen. The Eppendorf Histograph has provided the most convincing evidence for hypoxia in human tumors (53–55) and has been considered a “gold standard”, but it is highly invasive. Following extensive studies in animals, the Histograph was used in the clinical setting and has unequivocally revealed hypoxia in many tumor types, e.g., head and neck, cervix, breast, and prostate and even intraoperative deflated lung (2). Moreover, pO2 distributions have been found to have prognostic value. Disease free survival is significantly worse for patients with hypoxic tumors and applications to cervical cancer have shown correlation between tumor hypoxia and response to irradiation (56). In addition hypoxic tumors have been shown to have poorer response with surgical resection and this is considered to reflect a more aggressive hypoxic phenotype (10, 56, 57). By stepping the electrode through a tumor pO2 distributions are determined providing a quasi map, though normally presented just in terms of a histogram. Chapman et al. (58) determined that as many as 100 individual locations were required to reproducibly/reliably represent tumor oxygenation. The Eppendorf reveals a pO2 distribution, which is invariably skewed towards hypoxia. The distribution reported by the histogram reveals heterogeneity, but requires effective data reduction to be useful as a prognostic or predictive biomarker. Variously, mean, median or hypoxic fraction have been reported to correlate with therapeutic response. Hypoxic fraction may be described as HF10, HF5 or HF2.5 and may be quoted as a ratio to normal adjoining tissue. The Eppendorf Histograph is not well suited for dynamic measurement of pO2 with respect to intervention, though repeated measurement in patients with cervical cancer showed that oxygen or carbogen increased tumor oxygenation (59, 60). The Histograph requires tissue access and is impractical in many tumors.
Fiber optic probes can also measure pO2 directly and are typically finer and do not consume oxygen during measurement. Typically, two or four locations are sampled simultaneously and these optical probes facilitate observation of dynamic changes in pO2 in response to interventions (61, 62). Fiber optic probes are generally most sensitive to low pO2 providing little sensitivity above 100 torr.
ESR has been exploited effectively, using both vascular probes and injected implanted particulates (63–65). Most importantly, ESR of implanted carbon chars revealed reoxygenation, in tumors and a correlation with response to irradiation based on measurements at a single location (66). Gallez, et al., investigated many drugs and interventions using ESR and found that several drugs will enhance tumor oxygenation (67).
Materials may be monitored for weeks to months with no apparent deleterious side effects in normal tissue. However, ESR is not readily available for pre-clinical applications and it is essentially impossible in patients due to lack of instrumentation.
Given the importance of tumor oxygenation, many diverse methods have been developed and multiple reviews compare and contrast efficiency of imaging approaches. As such, the reader is directed to these articles for historical perspective. In 2004, the NCI sponsored a workshop on hypoxia imaging technologies as summarized in a comprehensive review by Arbeit, et al. (2) featuring contributions from many of the leaders in the field. Imaging approaches, together with needs for the measurements, are extensively described. Other recent reviews have focused on nuclear imaging approaches (8, 17), optical methods (68), ESR (51, 69) and MRI (6, 70). Others have focused on the molecular biology and relevance to therapy and clinical trials (71, 72). In Table 1, we summarize virtues and capabilities of popular methods. In the present review, we will focus on the latest innovations and our own experience with MRI.
In choosing a method, it is crucial to match the capabilities to the problem being addressed. Thus, predicting the efficacy of a hypoxic cell selective cytotoxin may well be ideal using a nitroimidazole reporter, which could exploit optical imaging, MRI or PET (73).
Most MRI oximetry exploits perfluorocarbons (PFCs), where the spin lattice relaxation rate (R1=1/T1) is exquisitely sensitive to pO2 and obeys a linear relationship (R1= a + b pO2) at any given magnetic field and temperature. Historically, PFC emulsion blood substitutes were administered systemically (74). Following iv administration, they exhibit prolonged vascular circulation and can provide measurements of vascular pO2 for several hours, though measurements are potentially susceptible to flow artifacts. Ultimately, PFCs become sequestered in tissue, primarily in the reticuloendothelial system (RES, liver, spleen and bone marrow) allowing pO2 measurements (74). Small amounts are found in myocardium and tumor and have been exploited for oximetry. The early PFC emulsions (FluosolTM; OxypherolTM (perfluorotributylamine); OxygentTM (perfluorooctylbromide); TheroxTM (F44E)) had multiple resonances and the requirement for chemical shift selective imaging often caused signal loss. However, each signal showed unique sensitivity to pO2 and temperature and this phenomenon could be exploited to determine both parameters simultaneously (75). Systemic administration had the great advantage of being non-invasive, but most of the small fraction of PFC, which became sequestered in the tumor was found in well perfused tumor regions biasing measurements (76).
Symmetrical PFCs may have a single 19F resonance offering better signal to noise ratio. Fifteen-crown-five-ether has been used as an emulsion to probe liver and tumor pO2 (77). Hexafluorobenzene (HFB) provides an alternative, which is more readily commercially available, more sensitive to changes in pO2 and less sensitive to changes in temperature (6). Hexafluorobenzene has an extremely high vapor pressure bringing potential benefits or problems. It does not form stable biocompatible emulsions and requires direct injection into tissue. It also clears rapidly so that pO2 measurements are generally only possible on the day of administration. Direct injection into a tissue has the distinct advantage that any region of interest may be interrogated immediately without needing to wait for vascular clearance. Moreover, there are no concerns about potential toxicity related to long-term retention. Hexafluorobenzene exhibits an extremely broad range of T1 from 0.7 s to 13 s at 4.7 Tesla, so that imaging may be slow. Since signal to noise ratio (SNR) influences the precision of pO2 measurements, the benefits of HFB may be less than anticipated (78). The potential long acquisition times required to assess long T1s were overcome by applying echo planar imaging (EPI). We routinely achieve pO2 maps with 50 to 150 individual locations simultaneously in 6½ minutes in rat tumors with a precision of 1 to 3 torr in hypoxic regions following administration of 50 μl HFB. We have used the so-called FREDOM approach (Fluorocarbon Relaxometry using Echo planar imaging for Dynamic Oxygen Mapping) to examine multiple tumor types implanted in rats and mice including breast, lung and prostate. Representative images and data are shown in Figures 4–7 and Table 3. As expected, median pO2 and hypoxic faction observed in rat tumors were inversely related (Figure 5).
We have applied this approach to multiple studies and have demonstrated measurements commensurate with more traditional techniques such as polarographic electrodes (including the Eppendorf Histograph) (79, 80), optical fiber probes (62), immunohistochemistry (81), and near-infrared spectroscopy (82), in terms of baseline pO2 and response to interventions.
The greatest strength of 19F MRI is the ability to make sequential measurements over a period of hours with respect to interventions. Maps of pO2 allow individual tumor regions to be monitored or cluster analysis reveals differential behavior for regions showing shared baseline characteristics. In some cases, the baseline measurement is indicative of response to intervention such as breathing hyperoxic gas (Figure 5b). For most tumors, we find that initially well oxygenated tumor regions show a particularly large response to hyperoxic gas breathing (83). In virtually every mouse and rat tumor system examined to date, we found progressive hypoxiation with increased size (Table 3 + Figure 6) (62, 81, 84). Small tumors are generally well oxygenated with small hypoxic fractions, but somewhere between 1 and 3 cm3 a catastrophic decline occurs in oxygen tension (62, 81, 84).
It is well documented that hypoxia leads to radiation resistance. During a course of conventional fractionated radiotherapy, progressive re-oxygenation is reported to occur. However, for a single dose or limited sequential doses, hypoxic fraction may be far more important. We have shown that irradiation (single dose 30 Gy) of large hypoxic Dunning prostate R3327-HI tumors caused limited tumor growth delay (Figure 6). However, this tumor type shows a pronounced oxygenation accompanying hyperoxic gas breathing, essentially eliminating the hypoxic fraction (HF10 < 20%; Table 3). With oxygen breathing there was a significant tumor growth delay accompanying the same single dose of irradiation (85). By contrast, in small HI tumors, where the HF10 <20 torr for air breathing, there was a significant tumor growth delay accompanying irradiation and no additional benefit was seen for breathing oxygen (Figure 6). We note that large AT1 tumors, which exhibit a high hypoxic fraction that resists modulation also show minimal benefit from breathing oxygen (86). Thus, measurements of baseline pO2 may be indicative of radiation response and assessment of the ability to manipulate tumor oxygenation using an intervention may be relevant to predicting such adjuvant interventions for enhancing radiation. Noting that some tumors resist hyperoxic gas breathing, it is tempting to apply hyperbaric oxygen and indeed using fiber optic probes, we showed that hypoxic regions of tumors, which resisted modulation by 100% normobaric oxygen, did respond to hyperbaric oxygen (87). We have previously shown that vasoactive drugs, such as hydralazine caused reduced tumor oxygenation (88). As expected, the vascular disrupting agent Combretastatin A4P caused rapid hypoxiation (89) and preliminary data show that this strongly influences the efficacy of combined radiation therapy. Meanwhile, the vascular targeting agent, bavituximab, has a minimal acute effect on pO2 over a period of two hours (Figure 7).
For small animal studies, quantitative 19F MRI oximetry of hexafluorobenzene (HFB) is very effective. There is no background signal and single resonance provides very high sensitivity to pO2. However, we have been exploring proton MRI alternatives with view to greater translatability to the clinical setting. Noting the requisite properties of the 19F perfluorocarbon reporters (hydrophobic, liquid, single resonance, unreactive, high gas solubility), we sought analogous proton agents and identified hexamethyldisiloxane (HMDSO) (90). Detection does require fat and water suppression, but this is routinely achieved for imaging metabolites such as N-acetyl aspartate, lactate, choline and citrate in brain or prostate of patients. Thus, proton magnetic resonance spectroscopy (90) or magnetic resonance imaging (91) may be used analogously to FREDOM (6) and we have initiated investigations using HMDSO (e.g., PISTOL: Proton Imaging of Siloxanes to map Tissue Oxygenation Levels). We show an example of the detection of hypoxiation following administration of the vascular disrupting agent Combretastatin (CA4P ) to a prostate tumor bearing rat in Figure 8 and this closely matches previous 19F MRI results (89).
Many reporter molecules are routinely exploited with diverse imaging modalities to measure tumor hypoxia or pO2. Ideally, oxygenation could be related to endogenous characteristics. Since many biochemical pathways are under oxygen regulation, they can provide an elegant window on hypoxia, e.g., induction of HIF-1 and Glut-1 together with secondary responses, such as increased production of VEGF, NIP3 and tumor associated macrophage activity. However, assessment of expression requires biopsy. Additionally, the optical absorption spectra of oxy- and deoxyhemoglobin are quite different allowing interrogation of vascular oxygenation by spectroscopy and imaging though spatial resolution is generally quite poor (68, 82). Deoxyhemoglobin is also paramagnetic and forms the basis of BOLD MRI.
BOLD contrast depends on deoxyhemoglobin, which is paramagnetic. Deoxygenated red blood cells generate local magnetic field gradients causing signal dephasing, which appears as T2* relaxation and can be detected as signal loss in T2*-weighted images. The magnitude of the effect depends on hematocrit, vascular volume, and concentration of deoxyhemoglobin. Images may further be influenced by flow effects leading to the concept of FLOOD (FLow and Oxygen Level Dependant) (92). The BOLD effect is widely exploited in studies of brain activation, where stimuli generate regional changes in blood flow, altering deoxyhemoglobin concentration and generating contrast, which is the foundation of fMRI (93). In blood, a direct relationship is found between the T2* and pO2 based on the sigmoidal O2 binding curve of hemoglobin. For large blood vessels T2* may be used to measure vascular oxygenation, and hence pO2, directly (94) and a recent development in the brain is sometimes referred to as qBOLD (95).
Interpretation of BOLD in tumors is more complex due to the bed of tortuous small capillaries, such that voxels include both blood vessels and surrounding tissue. Several groups have examined the effect and its relationship to tumor oxygenation, notably Howe, Robinson and Griffiths at St George’s, London (92, 96, 97). A direct correlation between pO2 and the BOLD signal changes has been found in several studies. Al Hallaq et al. (98) reported a strong linear correlation between change in pO2 measured by electrode and change in the signal line widths (essentially T2*). Elas et al. and Dunn et al. compared EPR (electron paramagnetic resonance) estimates of oxygenation and BOLD response and each found consistent data (65, 99). We recently showed a strong correlation between changes in T2*-weighted signal and pO2 based on 19F MR oximetry (100). Perhaps most significantly, we found that a large BOLD response (>2% ΔSI) to hyperoxic gas challenge (carbogen) corresponded with essential elimination of hypoxic fraction in rat breast tumors. Meanwhile, a small BOLD response generally indicated larger residual hypoxic fraction resistant to manipulation. Baudelet and Gallez reviewed BOLD investigations of tumors thoroughly (101) and also reported correlation of BOLD signal changes versus fiber optic oxygen tension measurements and oxygen challenge (102). The results indicated that there was always a positive correlation, but that a 10% change in relative signal intensity in the BOLD experiment could correspond to an increase in pO2 < 25 torr or > 100 torr. We contend that either change would be radiobiologically pertinent, because once pO2 exceeds 10 torr there is relatively little further oxygen enhancement achieved by increased levels of oxygen.
Currently, Padhani et al. are pursuing the use of BOLD to investigate tumors at various sites in patients with an emphasis on prostate cancer (103–105). Their primary method differs somewhat from our approach in that they assess baseline R2* and relate it to inherent hypoxia rather than evaluating hypoxia based on the response to an oxygen breathing challenge. Rodrigues et al., showed that R2* itself was indicative of hypoxia and correlated with response to radiation in tumors (106).
We note the BOLD response reflects vascular oxygenation, which may be disconnected from tissue pO2. While increased oxygen delivery is expected to elevate local pO2, the additional oxygen could be consumed, as noted by Gullino and Vaupel in studies of perfused tumors (107, 108). Thus, we believe it is imperative to also include an assessment of tissue oxygenation (pO2).
Molecular oxygen (O2) is paramagnetic and hence [O2] influences tissue water spin-lattice relaxation (R1). Recently, Matsumoto et aI. (109) demonstrated T1 sensitive signal response in tumors to breathing hyperbaric oxygen, which reflects changes in tissue pO2. Others have explored normal tissue response to hyperoxic gas breathing, notably Edelman et al. (110) and recently O’Connor et al. (111, 112), Jones et al. (113) and Tademura et al. (114). However, except in rare circumstances, such as vitreous humor or CSF, it is probably not a suitable measure of absolute pO2, since many factors can alter R1 (115, 116). We believe it can provide suitable evaluation of acute changes in pO2.
We believe that combining BOLD and TOLD response to hyperoxic gas intervention provides particularly robust insight into tumor hypoxia and potential modulation. This is the basis of the simple test DOCENT (Dynamic Oxygen Challenge Evaluated by NMR T1 and T2*), as shown in Figure 9. To demonstrate the method, we chose a small HI tumor, which is known to be particularly well oxygenated and exhibit a large response to breathing hyperoxic gas (Table 3). We note a stable baseline for BOLD and a rapid response in signal intensity approaching 40% ΔSI within one minute of breathing carbogen (95% O2, 5% CO2) (Figure 9). Distinct heterogeneity was observed in any given slice, but quite similar data were seen by MRI in three adjacent tissue slices. The TOLD response was smaller and more sluggish, as expected, since O2 must diffuse from the vasculature into the tumor tissue. Importantly, the combined methods robustly demonstrate delivery and accumulation of oxygen. It remains to be seen whether quantitative R1 and R2* measurements are required or simple R1- and R2*-weighted signal intensity changes will be useful. We are currently evaluating whether such measurements are predictive for radiation response in rat tumors. We are also implementing these methods in patients.
Within the past year, several innovations have been presented promising new insight into tumor hypoxia. There has been notable progress in quantitative oximetry based on 19F MRI with new reports of applications. Gallez et al. implemented a faster imaging approach for 19F MRI of hexafluorobenzene, Look-Locker type SNAP-IR measurements giving 90 s time resolution (117, 118). Spatial resolution was similar to that previously reported for FREDOM (requiring 6.5 mins), though a larger volume of hexafluorobenzene reporter molecule was used (90 vs. 50 μl) and it remains to be seen whether the precision of measurements will be equivalent, since SNR is known to strongly influence the quality of relaxation rate measurements (78). We do ourselves sometimes use larger volumes of HFB in tumors, e.g. Figure 4. In any case, the faster temporal resolution opens new opportunities to explore rapid dynamic pO2 fluctuations and transients: a new window on acute hypoxic episodes. Notably, both the groups in Louvain and UCSF are using 19F MR oximetry of hexafluorobenzene to explore oxygen dynamics in response to pharmaceutical interventions (119–121). The use of HFB as a 19F pO2 reporter molecule is gaining popularity with new reports of applications. Diepart et al. have further used this technique to measure oxygen consumption in tumors (120). Using the FREDOM technique, Liu et al. (122) compared pO2 changes in multiple organs during isovolemic anemic hemodilution using hemoglobin based oxygen carriers under normoxic and hyperoxic conditions in a rat model.
While we favor direct intra tumoral injection of HFB for immediate interrogation of regions of interest, we note a novel approach presented by Ahrens et al. (123). Several groups have pre-labeled cells in culture with PFC to allow cell tracking after implantation in vivo (124) and this has now been used to measure changes in intracellular pO2 following implantation of cells.
Interrogation of hypoxia by NMR has received new impetus with the presentation of a trifluoronitroimidazole, which provided sufficient signal to noise to allow chemical shift imaging, as opposed to the more traditional signal limited spectroscopy (125). A novel approach uses a gadolinium liganded nitroimidazole (GdDO3NI), which showed increased MRI contrast in hypoxic central regions of rat prostate AT1 tumors compared to peripheral, well-perfused regions. Increased uptake of GdDO3NI in the tumor center was further confirmed by flame ionization spectroscopy post mortem (126).
Well characterized animal models exhibiting differential hypoxia are important for evaluating new methods of detecting hypoxia and evaluating novel therapies. Based on 19F MRI, we have extensively characterized a series of Dunning prostate R3327 rat tumors, which exhibit a wide range of oxygen characteristics, as shown in Table 3. The Dunning prostate tumors have the great advantage of being syngeneic in rats allowing investigations in immunocompetent animals. Moreover, rats show much more stable physiology than mice under anesthesia and greater tumor burden is feasible. Indeed, knowledge of the hypoxia status was crucial in choosing a model to evaluate GdDO3NI
We find three distinct characteristics for tumor oxygenation and dynamics:
Increasingly, there is evidence that hypoxia influences angiogenesis, tumor invasion and metastasis. Moreover, repeated bouts of intermittent hypoxic stress may be important in stimulating tumor progression. Thus, the ability to assess tumor oxygenation non-invasively, and repeatedly, with respect to acute or chronic interventions becomes increasingly important.
PET approaches are costly for small animal research, but they have already made the transition to clinical trials. They do require reporter molecules and suffer from the complications of radioactivity, notably substrate decay and waste disposal.
MRI approaches are very attractive for pre-clinical investigations of hypoxia and oxygen dynamics. They offer high spatial resolution, arbitrary tissue depth penetration and avoids complications of radioactivity. However, data acquisition is not trivial, throughput is quite modest and instrumentation is expensive. Uniquely, MRI can provide quantitative maps of pO2 and reveal dynamic changes in response to interventions based on inert reporter molecules. 19F MR oximetry, e.g., FREDOM, based on hexafluorobenzene is perfectly suited to measurements in pre-clinical animal studies providing quantitative pO2 measurements with useful spatial and temporal resolution and precision relevant to radiotherapy (6). Given the well documented lack of toxicity clinical application ultimately be feasible, but currently 19F MRI remains esoteric on clinical MR scanners. PISTOL offers an analogous proton MRI approach (91). Repeat measurements of pO2 are non-invasive, but both PISTOL and FREDOM approaches do require a reporter molecule.
Meanwhile, BOLD and TOLD MRI approaches offer entirely non-invasive assessments of tissue oxygenation dynamics and are clearly preferable for clinical applications. While they do not provide quantitative measurements of pO2 there is increasing evidence that changes in R1 and R2* relate to changes in tissue and vascular oxygenation. Together the offer inter-dependant insight into tumor oxygenation and DOCENT (Dynamic Oxygen Challenge Evaluated by NMR T1 and T2*) promises to become a test for human tumor hypoxia and response to intervention based on the simple act of breathing oxygen during MRI.
This work was supported in part by the DOD Congressionally Directed Medical Research Programs DAMD#17-00-1-0437, W81XWH-08-1-0583, W81XWH-06-1-0475, W81XWH-06-1-0149, as well as NCI R01 CA139043-01A1 and 1R21 CA132096-01A1 in conjunction with Cancer Imaging Program SAIRP U24 CA126608 and NIH BRTP Facility P41-RR02584. We are grateful to Michael Long and Drs. Alex Hermanne, Xiankai Sun, Marc Jennewein and Frank Rösch for facilitating the radioarsenic study presented in Fig. 1d.