Monitoring Tumor Response to Carbogen Breathing by Oxygen-Sensitive Magnetic Resonance Parameters to Predict the Outcome of Radiation Therapy: A Preclinical Study

Abstract

In an effort to develop noninvasive invivo methods for mapping tumor oxygenation, magnetic resonance (MR)-derived parameters are being considered, including global R1, water R1, lipids R1, and R2*. R1 is sensitive to dissolved molecular oxygen, whereas R2* is sensitive to blood oxygenation, detecting changes in dHb. This work compares global R1, water R1, lipids R1, and R2* with pO2 assessed by electron paramagnetic resonance (EPR) oximetry, as potential markers of the outcome of radiation therapy (RT). R1, R2*, and EPR were performed on rhabdomyosarcoma and 9L-glioma tumor models, under air and carbogen breathing conditions (95% O2, 5% CO2). Because the models demonstrated different radiosensitivity properties toward carbogen, a growth delay (GD) assay was performed on the rhabdomyosarcoma model and a tumor control dose 50% (TCD50) was performed on the 9L-glioma model. Magnetic resonance imaging oxygen-sensitive parameters detected the positive changes in oxygenation induced by carbogen within tumors. No consistent correlation was seen throughout the study between MR parameters and pO2. Global and lipids R1 were found to be correlated to pO2 in the rhabdomyosarcoma model, whereas R2* was found to be inversely correlated to pO2 in the 9L-glioma model (P=.05 and .03). Carbogen increased the TCD50 of 9L-glioma but did not increase the GD of rhabdomyosarcoma. Only R2* was predictive (P<.05) for the curability of 9L-glioma at 40Gy, a dose that showed a difference in response to RT between carbogen and air-breathing groups. (18)F-FAZA positron emission tomography imaging has been shown to be a predictive marker under the same conditions. This work illustrates the sensitivity of oxygen-sensitive R1 and R2* parameters to changes in tumor oxygenation. However, R1 parameters showed limitations in terms of predicting the outcome of RT in the tumor models studied, whereas R2* was found to be correlated with the outcome in the responsive model.