|Home | About | Journals | Submit | Contact Us | Français|
Oxygen-Enhanced Magnetic Resonance Imaging (OE-MRI) techniques were evaluated as potential non-invasive predictive biomarkers of radiation response. Semi quantitative blood-oxygen level dependent (BOLD) and tissue oxygen level dependent (TOLD) contrast, and quantitative responses of relaxation rates (ΔR1 and ΔR2*) to an oxygen breathing challenge during hypofractionated radiotherapy were applied. OE-MRI was performed on subcutaneous Dunning R3327-AT1 rat prostate tumors (n = 25) at 4.7 T prior to each irradiation (2Fx15 Gy) to the gross tumor volume. Response to radiation, while inhaling air or oxygen, was assessed by tumor growth delay measured up to four times the initial irradiated tumor volume (VQT). Radiation-induced hypoxia changes were confirmed using a double hypoxia marker assay. Inhaling oxygen during hypofractionated radiotherapy significantly improved radiation response. A correlation was observed between the difference in the 2nd and 1st ΔR1 (ΔΔR1) and VQT for air breathing rats. The TOLD response before the 2nd fraction showed a moderate correlation with VQT for oxygen breathing rats. The correlations indicate useful prognostic factors to predict tumor response to hypofractionation and could readily be applied for patient stratification and personalized radiotherapy treatment planning.
Hypoxia is increasingly recognized to play a fundamental role in aggressiveness and therapeutic resistance in many tumors including prostate [1–5]. Hypoxia has been associated with radioresistance in cells , pre-clinical animal studies [7–12] and human patients [3, 5, 13]. However, the meta-analysis of Overgaard et al.  indicated that interventions to overcome hypoxia provided only marginal benefit and it was concluded that the lack of efficacy was likely related to the inability to identify which patients would benefit. Consequently, there is a substantial effort to develop non-invasive measurements of the dynamics of tumor oxygenation, as potential biomarkers for patient stratification [2, 15, 16].
Robust evidence for hypoxia in human tumors has been established at multiple disease sites using the Eppendorf Histograph electrode system [1, 4, 5, 17–20]. This has also been applied extensively in pre-clinical studies, but is highly invasive, technically challenging and no longer commercially available. Analogous measurements of tumor pO2 and hypoxic fractions have been achieved by direct intra tumoral administration of reporter molecules for 19F [8, 9, 21] and 1H MRI , and ESR [11, 23, 24]. These have the distinct benefit of allowing dynamic response to interventions to be assessed non-invasively [8, 9, 11, 22–25]. To avoid violating tumor integrity, reporter molecules may also be delivered intravenously [10, 26], but such measurements invariably bias results towards better perfused and likely less hypoxic regions. The need for reporter molecules complicates potential translation to the clinic.
Hypoxia may be directly observed using nuclear medicine reporters, typically, 18F labeled nitroimidazoles [15, 16, 27], but the associated radioactivity makes them expensive and assessment of dynamic modulation of hypoxia is generally not practical. Analogous use of immunochemistry of nitroimidazole trapping has allowed pulse chase evaluation of hypoxia modulation, but requires biopsy .
Oxygen enhanced MRI has been suggested as a potential alternative approach, since it is entirely non-invasive and can be readily added to routine clinical MRI, which is increasingly applied to radiation planning and execution . The tissue water proton apparent transverse relaxation rate (R2*) is strongly influenced by the concentration of deoxyhemoglobin, which is paramagnetic [2, 28]. This provides blood oxygen level dependent (BOLD) contrast, which is the basis of fMRI used in studies of neuronal activation. R2* is influenced by conversion of deoxy- to oxyhemoglobin, but is also subject to alteration in flow, hematocrit, and vascular volume, as described by the so-called FLOOD effect . Meanwhile, the spin lattice relaxation rate (R1) is directly sensitive to the concentration of free oxygen molecules and hence pO2. This is the basis of tissue oxygen level dependent (TOLD) contrast . Several investigations have examined correlations between BOLD and TOLD based on semi quantitative changes in signal intensity or quantitative relaxation maps. Notably, studies in human tumor xenografts in mice , as well as syngeneic tumors in rats [33, 34] and rabbits . The two approaches have also been assessed in humans including volunteer patients [36–38].
Several studies have examined correlations of BOLD with invasive oximetry in pre-clinical studies based on polarographic oxygen electrodes, fluorescent quenched fiber optic probes and 19F MRI [39–41]. Sometimes a strong direct correlation has been observed, while other studies indicated nonlinear correlative trends. Notably, a large BOLD response to a hyperoxic gas breathing challenge was associated with elimination of hypoxia in 13762NF rat breast tumors . A recent report indicated that syngeneic rat prostate tumors could be stratified in terms of radiation response based on TOLD MRI responses to an oxygen breathing challenge before a single dose of radiotherapy .
New hypofractionated treatment approaches are gaining popularity for several reasons: i) fewer treatment sessions are convenient to patients and physicians; ii) precise treatment plans may be developed for each irradiation; iii) recent clinical trials are showing enhanced outcome . However, it is thought that radiation response is more influenced by hypoxia, especially when large single- or multi-fraction dose regimens typical of stereotactic body radiotherapy (SBRT) are implemented, since tumor reoxygenation is minor compared to traditional conventional fractionated radiation therapy (CFRT) .
Noting the importance of hypoxia and desire for a robust non–invasive approach to assess tumor hypoxia and oxygen dynamics, prompted us to explore OE-MRI with respect to a hypofractionated radiation regimen.
Investigations were approved by the Institutional Animal Care and Use Committee. The experimental timeline for separate groups of tumors is shown in Table 1. Additional experimental details are provided in Supplementary Materials.
Dunning R3327-AT1 prostate tumors were surgically implanted subcutaneously in the flank of 25 adult male syngeneic Copenhagen rats . The AT1 is a well-characterized anaplastic prostate tumor often used for radiobiological studies [9, 21, 33, 44–46]. Tumors were used for OE-MRI around 19 days after implantation, when they reached a size in the range 0.7–2.1 cm3. Animals were divided into four groups: unirradiated “Control” (Group 1, n = 4), irradiated while inhaling “Air” (Group 2, n = 9), irradiated while inhaling “Oxygen” (Group 3, n = 9) and immunohistological correlates (Group 4, n = 3).
Anesthetized rats were provided with a warming pad to maintain body temperature, placed in a 4.7 T MR scanner and physiological parameters recorded using an MR-compatible monitoring and gating system. Baseline R1 measurements of the tissue water proton signal were obtained with a 2-D multi-slice spin-echo (SEMS) sequence, while the animals breathed air (1 dm3/min with 1.5–2.0% isoflurane) and at the end of the oxygen breathing challenge. Interleaved dynamic blood-oxygenation level dependent (BOLD or R2*) and tissue-oxygenation level dependent (TOLD or T1-weighted) measurements were performed for about 10 minutes for baseline air and during a hyperoxic oxygen breathing challenge (1 dm3/min O2 up to 10 minutes). BOLD acquisition used a 2-D multi-slice spoiled gradient-echo with multi-echo (MGEMS) sequence.
Tumors were irradiated about 24 hrs after OE-MRI experiments. Prior to, and during radiotherapy, the anesthetized animals inhaled either Air (Group 2, n = 9) or Oxygen (Group 3, n = 9) for at least 15 minutes. Unirradiated tumors (n = 4) provided controls. Radiation was applied to the gross tumor volume (GTV) with orthovoltage x-rays at 15 Gy using image-guided radiation therapy with a small animal x-ray irradiator. OE-MRI and irradiation were repeated one week later. Tumor growth was measured weekly until tumors reached 10% body weight or 90 days to assess the response to radiation. Tumor growth delay was determined by the time required for the tumors to reach two (volume doubling time, VDT) and four times (volume quadrupling time, VQT) the initial irradiated tumor volume using simple linear interpolation. Three additional tumors (Group 4) were examined to assess reoxygenation after the first fraction based on immunohistochemistry.
A double hypoxia marker approach [28, 47] was used to verify tumor reoxygenation. Immediately after OE-MRI, three tumor bearing rats, while breathing oxygen, were injected intravenously with pimonidazole as a baseline tumor hypoxia marker. About 24 hours later two of the tumors were irradiated with 15 Gy, while the animal was breathing oxygen. The third tumor served as a control. Three days later a second tumor hypoxia marker, CCI-103F, was injected intraperitoneally, while the rats were breathing oxygen. Two hours later, the rats were sacrificed and tumor tissue harvested.
Using in-house algorithms developed in Matlab, voxel-by-voxel %ΔSI in BOLD and TOLD responses with respect to inhaling oxygen were calculated from the whole tumor region-of-interest. BOLD images were selected at a single echo time (TE=20 ms) for analysis. Voxel-by-voxel R2* maps were generated from BOLD images by fitting the multi-echo data to the echo time (TE) in a nonlinear least squares equation and quantitative ΔR2* values were calculated. Likewise, R1 with respect to the repetition times (TR). A log-rank test Kaplan-Meier analysis was used to compare tumor growth for Air, Oxygen and Control groups.
Tumors showed considerable heterogeneity in terms of baseline R2* and R1, as well as responses to oxygen challenge (semi quantitative BOLD and TOLD; quantitative ΔR2* and ΔR1; Fig. 1). Mean R1 for individual tumors ranged from 0.3367 to 0.7122 s−1 with a population mean 0.499±0.0259 s−1. Baseline ΔR1 ranged from −0.006 to 0.12 s−1 with a mean 0.041±0.008 s−1 for the 18 tumors (Groups 2 and 3; Table 1). Mean R2* ranged from 27.4 to 69.9 s−1 with a mean 51.7±3.1 s−1 and ΔR2* ranged from −3.5 to 9.1 s−1 (mean 0.55±0.75 s−1). There were no significant differences between Groups before irradiation, but there was a significant difference in ΔR2* between Air and Oxygen breathing groups (P<0.03) one week after 15 Gy (before the 2nd irradiation). The semi quantitative parameters showed a significant correlation between mean TOLD and BOLD responses for individual tumors before the 1st irradiation (R>0.6, P<0.005) and before the 2nd irradiation (R>0.6, P=0.0051).
Non-irradiated tumors showed a typical VDT = 6.5±0.3 days and VQT = 13±0.4 days with exponential growth up to time of sacrifice at 10% body weight (Fig. 2). All tumors responded to 2×15 Gy with a growth delay regardless of inhaling air or oxygen (Fig. 2A). Tumors growing on animals breathing O2 tended to express a greater growth delay than tumors in the Air Group (Fig. 2), which was significant for VDT (P<0.001) and VQT (P<0.04) (Table 1, Fig. 2B) and confirmed by the log rank test (Fig. 2C).
Potential correlates of VDT and VQT with OE-MRI were examined (Table 1 and Supplementary Table S1). Moderate correlations were seen for pre irradiation ΔR1 for both Air and O2-breathing groups in addition to ΔR2* for Air and O2-breathing groups combined (data not shown). The strongest correlation was observed in Group 2 (Air) for change in ΔR1 response to O2 challenge between the first and second measurements (ΔΔR1). Those tumors showing the greatest increase in ΔR1 and had the longest VQT (R>0.9, P<0.002, Fig. 3A). For Group 3, but not for Group 2, there was a modest correlation between TOLD before the 2nd irradiation and VQT (R>0.6, P<0.04, Fig. 3B). Mean BOLD and TOLD responses did not change significantly between fractions (Fig. 4, Table 1), but the fraction of voxels within the tumors showing a TOLD response to O2 challenge increased significantly (Fig. 4). The BOLD response appeared largely unchanged. These data are consistent with reoxygenation revealed by pulse chase immunohistochemistry (Fig. 5). Extensive hypoxia was observed pre irradiation, which appeared consistent 3 days later in the absence of radiation (HFpimo =28%, HFCCI-103F =27%; Fig. 5). Irradiated tumors showed significant decrease in hypoxic fraction after 72 hours (HFpimo =9.7%, HFCCI-103F =0.3%).
Both, the tumor oxygenation status as well as its treatment-related variation have profound clinical implications, which represent the driving force to further exploit fast and sensitive imaging techniques. OE-MRI is such a non-invasive technology which provides parameters sensitive to changes in tissue oxygenation. Although the idea of altering radiation response based on the simple procedure of breathing hyperoxic gas [8–10, 23, 33, 48] was disappointing when translated to patients, there is a general consensus that lack of success was mainly due to inability to identify those patients who would benefit. This study serves to further assess the potential prognostic utility of OE-MRI . Hypofractionation of Dunning prostate R3327-AT1 tumors was selected as a model system due to evidence of a prompt oxygen response following the application of large radiation doses . We chose one week between two radiation doses to match ongoing clinical trials of hypofractionated SBRT in lung cancer (3×16 Gy over 1½ to 2 weeks )
As expected, a significant tumor growth delay was observed and rats breathing oxygen during irradiation showed a greater response (VDT and VQT, Table 1, Fig. 2). However, each cohort exhibited a range of growth delays with overlap between the Air and Oxygen Groups. It was previously reported that tumor growth delay in response to a single dose of 30 Gy, while breathing oxygen was related to pre irradiation TOLD response to an oxygen breathing challenge . A significantly greater VQT was observed for those tumors with large TOLD. Here, we observed a strong correlation between TOLD prior to the 2nd irradiation and VQT for Group 3 (but not Group 2, Fig. 3B). The strongest correlation was observed between VQT and the change in TOLD response (ΔΔR1: baseline vs. 1 week later) for the Air Group (Fig. 3A). Those tumors with increased response to oxygen breathing (positive ΔΔR1) showed a greater tumor growth delay than those showing a reduced response (P<0.005). The observed R1 and R2* (Fig. 1) are not dissimilar from previous reports for this tumor type at 4.7 T [33, 34]. Present results (ΔR1 = −0.006 to 0.121 s−1) are also comparable to reports of orthotopic gliomas implanted in mice , though overall population mean response was somewhat higher (0.04 vs. 0.01 s−1 with oxygen or carbogen) than reported for the orthotopic gliomas or squamous cell carcinomas with respect to hyperbaric oxygen .
Most parameters showed little change after irradiation, but ΔR2* response to breathing oxygen was significantly different for Group 3 versus Group 2, one week after 15 Gy irradiation (Table 1). The more negative ΔR2* is consistent with greater conversion of deoxy- to oxyhemoglobin in response to O2-breathing challenge and hence improved tumor oxygenation. By contrast Lin et al.  examined BOLD response of TRAMP-C1 tumors 6 days after 15 Gy and found significantly smaller response compared to controls implying hypoxiation. It was previously reported that only a sub-group of AT1 tumors benefited from oxygen breathing during a single dose of radiation (30 Gy) and these were characterized by larger ΔR1 . Here, for the split dose (2×15 Gy), O2-breathing enhanced radiation response, but in terms of OE-MRI parameters, there was a general trend rather than a stratifiable difference in response. However, for the Air–breathing group those with improved oxygenation (positive ΔΔR1), did as well as those with oxygen breathing, whereas those with negative ΔΔR1 did significantly less well. Noting the reported relaxivity of oxygen in tissue at 4.7 T (r1 = 9 *10–4 Torr−1.s−1 ) the observed changes (ΔR1 = −0.006 to 0.121) would correspond with ΔpO2 = −6 to +134 Torr in response to oxygen breathing challenge and ΔΔR1 ± 0.05 s−1 implies ΔΔpO2 = 55 Torr suggesting that the observed changes in oxygenation would cause distinct differences in radiation response.
Assessing tumor oxygenation using endogenous contrast based on tissue blood and water is appealing, but there is a potential caveat particularly in terms of BOLD responses. These are predicated on conversion of deoxy- to oxyhemoglobin. Historically, there are reports that hematocrit may decrease in patients during a course of radiation therapy. This is expected to be less of an issue with modern conformal targeting capabilities. Leonard et al. recently reported no significant change in the hematocrit of patients undergoing definitive IMRT for prostate cancer , though a significant decline, albeit only 1%, was seen in patients receiving hormone ablation therapy. Lower hematocrit would imply smaller BOLD response, but also less capability of delivering oxygen, and thus any reduced BOLD response would actually reflect weaker modulation. Noting that TOLD is primarily sensitive to pO2 rather than hemoglobin status, again emphasizes why TOLD mat be expected to be a more robust marker of tumor hypoxia and modulation. To further investigate improved oxygenation for Group 3, we applied the pulse chase approach pioneered by van der Kogel et al. . The distribution of the bioreductive hypoxia markers pimonidazole and CCI-103F administered before IR and 3 days after irradiation in the control non-irradiated tumor closely matched, indicating consistent hypoxia. In contrast, irradiated tumors showed significantly decreased marker distribution indicating reduced hypoxia (Fig. 5).
Previous results regarding post irradiation reoxygenation appear contradictory. A significant increase in tumor oxygenation was observed by 19F MRI in Dunning prostate AT1 tumors up to 10 hours after 20 Gy irradiation  and by EPR in RIF-1 tumors, 72–120 hours after 20 Gy . Meanwhile, a recent study in FSaII tumors demonstrated secondary cell death, based on clonogenic survival assays 2–5 days, after 20 Gy, attributed to by vascular damage, which was accompanied by a significant increase in hypoxic markers . As levels of hypoxia and response are expected to vary with radiation dose, tumor type, host species and method of anesthesia the need for further investigations is obvious.
OE-MRI is particularly attractive since it is readily translatable to human applications. Several preliminary studies have already shown the use of hyperoxic gas challenge to stimulate BOLD and/or TOLD MRI signal responses in tumors at distinct disease sites (breast [56, 57], cervix [49, 58, 59], brain [38, 60], prostate , and liver ). Breathing oxygen is particularly appropriate, since this is standard intervention in emergency medicine and thus it is quite straightforward to gain IRB approval. Carbogen (95%O2, 5%CO2) has been favored by some investigators, but it is reported to cause some respiratory distress and is les swell tolerated. In terms of adding OE-MRI to a routine radiological examination, it would need to be applied before any Gd-contrast agents are applied. As seen in Fig 4, the response to oxygen breathing is quite rapid and therefore do not add an undue burden to the length of radiological exam. Overall the non-invasive nature of the measurements, lack of radioactivity and ease of conducting an oxygen gas challenge make OE-MRI particularly attractive.
In summary, the results further indicate the feasibility of OE-MRI and again suggest that R1 is more relevant to stratifying tumors than R2*. Given the non-invasive nature of the measurements and relatively rapid data acquisition, the approach could be rapidly incorporated in future clinical investigations, and should ultimately indicate which patients are likely to benefit from oxygen breathing during irradiation.
Supported in part by funds from the National Cancer Institute (5R01 CA139043). MRI experiments were performed in the AIRC, supported by National Institute of Biomedical Imaging and Bioengineering Resource grant (EB015908) and subsidized by NCI Cancer Center Support Grant (1P30 CA142543) and irradiation facilitated by Shared Instrumentation Grant (S10 RR028011). Histology was facilitated by the Imaging Core of the Cancer Center under the guidance of Dr. Kate Phelps. We thank Joshua Gunpat and Rebecca Denney for tumor implantation.
*Presented in part at 23rd Annual meeting of ISMRM, Toronto, Canada, June 2015, 57th Annual meeting AAPM, Anaheim CA, July 2015 and 61st Annual Meeting Radiation Research Society, Weston, FL, September, 2015.
The authors have no conflicts of interest to disclose.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.