Evaluation of the relationship between lSO2 MR measurement and hypoxia : impact of an antiangiogenic treatment on a gliosarcoma model

Introduction: Glioblastomas (GBM), the most angiogenic brain tumors, exhibit either hyper-vascularised and necrotic areas. Despite this highly vascular phenotype, most tumor cells are in hypoxia (1). Indeed, the new vessels are structurally abnormal. They are dilated and tortuous, leading to a chaotic blood flow. The existence of vessels does not guarantee a good oxygenation of tumor tissue. This hypoxia can select a sub-class of cancer cells which possess the ability to survive without oxidative metabolism and continue to proliferate despite these conditions. Hypoxia is associated with an aggressive phenotype that promotes tumor growth metastasis (2). Furthermore, hypoxia renders tumor chemo-and radio-resistant. Recently, different studies showed that local blood oxygen saturation (lSO2) could be estimated by MRI (3). Antiangiogenic therapies, which have demonstrated their ability to change tumor vasculature, should also alter tumoral oxygenation. In this study, we evaluate (i) the relationship between lSO2 estimated by MRI and tissue hypoxia estimated by immunohistology and (ii) the impact of an antiangiogenic (Sorafenib) treatment on the vasculature (blood volume fraction; BVf) and the oxygenation (lSO2) of a gliosarcoma model (9L). Material and methods: Fisher 344 rats were orthotopically injected at day 0 (D0) with 10 9L glioma cells (n=48). At D7, T2-weighted images were acquired to measure tumor size (4.7T, Bruker Avance III console). Rats were then randomized in 2 groups (n=24 per group) with similar tumor volume (4.6±2.5 mm, data not shown). Treatment started at D10 (D10(T0)). Untreated group received no treatment. Treated group received a daily oral administration of Sorafenib (100 mg.kg; Nexavar, Bayer Corporation) between the 1 and the 8 day after the start of the treatment (D10(T0) to D18(T8)). BVf and lSO2 were mapped 1 day before and 1, 3, 5 and 8 days after starting the treatment (D9(T-1), D11(T1), D13(T3), D15(T5) and D18(T8), respectively). BVf was mapped using a steady-state approach (a multiple gradient-echo/spin-echo MR sequence was acquired before and after intravenous injection of ferumoxtran-10 (Sinerem, 200 μmol Fe.kg, obtained from Guerbet). lSO2 was mapped using a recently proposed method and based on the quantitative BOLD approach(3). For each group, 4 rats were imaged at every time point. At each time, 4 additional animals per group were imaged and then sacrificed for ex-vivo studies. We therefore obtained, for each group and each time, MRI data from 8 animals and ex-vivo data from only 4 animals. One hour before sacrifice, rats were injected with pimonidazole (100 mg.kg; Hypoxyprobe-1, Chemicon). After sacrifice, immunostaining of pimonidazole was performed (10 μm thick slice). The percentage of necrotic/ hypoxic area within the tumor was calculated from the sum of pimonidazole stained areas and necrotic areas. The fraction of pixels within the tumor ROI and with lSO2 < 40 % was estimated on MRI data as a blood SO2 <40 % is known to induce tissue hypoxia (4). Then, the correlation between the percent of low lSO2 values (< 40 %) estimated by MRI and the percent of hypoxic/necrotic area estimated by immunohistology was computed. Results: Visually, in untreated tumors, we observed an increase in the number of pixels with low lSO2 values over time. Meanwhile, in the same group, an increase in pimonidazole stained areas was observed (Fig1a). In untreated tumor, a good correlation was found (R2 = 0.812) between the hypoxic-necrotic areas measured from pimonidazole staining and the fraction of tumor pixels with lSO2 < 40% (Fig1b). This correlation was not found in treated tumors. lSO2 in contralateral striatum did not change over time (68 ± 2 %; mean across all time points; Fig1c). Values of lSO2 in untreated tumors were not different from that in contralateral striatum (70 ± 2 %; mean across all time points). In contrast, there was a decrease between D9(T-1) and D15(T5) in lSO2 of tumors treated with Sorafenib (from 71 ± 5 to 54 ± 5 %, respectively; Fig1c). At the last imaging time, lSO2 in treated tumors was similar to that estimated in untreated tumors and contralateral striatum (61 ± 4, 69 ± 3 and 66 ± 8 % at D18(T8) respectively; Fig1c). Before treatment, BVf measured in tumors of both groups was higher than in contralateral striatum (5.9 ± 0.5, 5.5 ± 0.5 % versus 3.0 ± 0.5 % for untreated, treated tumor and contralateral striatum respectively; Fig1d). In untreated tumor and in contralateral striata, BVf did not vary over time (Fig1d). In treated tumors, however, there was a decrease in BVf between D9(T-1) and D15(T5) (from 5.5 ± 0.5 to 3.3 ± 0.7, respectively; Fig1d) and a trend to increase at the last imaging time (Fig1d). The fraction of lSO2 <40% was similar in both groups before treatment (≈ 3 %). During tumor growth, this fraction slowly increased in untreated tumors (from 3 ± 3 to 22 ± 10%, between D9(T-1) and D18(T8)). This increase was more pronounced in treated tumors (from 3 ± 3 to 39 ± 10%, between D9(T-1) and D15(T5)) before a decrease at the last imaging time (33 ± 9; at D18(T8)).