Application of a modified quantitative BOLD approach to monitor local Blood Oxygen Saturation in two glioma models and a stroke model in rat

Introduction Brain oxygenation level is physiological information of interest in numerous cerebral pathologies. However, few MRI techniques for quantifying oxygen-level parameters (such as partial pressure of oxygen, oxygen extraction fraction...) are available. Recently, He and Yablonskiy proposed an in vivo MR approach – quantitative BOLD – to obtain local blood oxygen saturation (lSO2) maps [1] [2]. In this study, we evaluate a modified version of the qBOLD technique in two cerebral diseases: stroke and tumor. Theory The gradient echo MR signal decay can be described by: S(t)= Cte. F(t).exp(-t.R2).exp(-t.R2’) (water diffusion neglected) [Eq. 1] Where Cte is a proportionality constant, F(t) represents the contribution to signal attenuation caused by macroscopic field inhomogeneities [1], R2= 1/T2 and R2’= 1/T2’= 4/3.π.Δχ0.Hct.(1-lSO2).B0.γ.BVf. Δχ0 represents the change in magnetic susceptibility between oxy and deoxy-haemoglobin (0.264 ppm), Hct for hematocrit (%), and γ for magnetogyric ratio. Material and methods Animals (n=23) were anaesthetized using isoflurane (2%). The tail vein was equipped with a catheter. Four groups of animal were studied: • Control group: male Wistar rats were used as control (n=12). • C6 group: 10 C6 glioma cells were orthotopically implanted in the striatum of Wistar rats (n=4). Imaging was performed 18 days after tumor implantation. • U87-MG group: 10 U87-MG glioma cells were orthotopically implanted in the striatum of male nude rats (n=4). Imaging was performed 34 days after tumor implantation. • Stroke group: Transient (90min) focal brain ischemia was induced by occlusion of the right Middle Cerebral Artery using the intraluminal suture model [3] in male Sprague-Dawley rats (n=3). MRI was performed 2 days after ischemia. MR imaging was performed at 4.7T on a Bruker Avance 3 console using volume/surface cross coil configuration. All data were acquired with the same geometry (7 contiguous, 1mm-thick slices, FOV=30x30mm; matrix=64x64 or 128x128), except for B0 mapping (3D GE sequence, FOV=30x30mm, matrix=128x128x40, TR=100ms TEs=4 and 12ms). Acquisition protocol was: brain shimming, B0 mapping, T2 mapping (TR=1500ms, 20 spin-echoes, ΔTE=12ms), T2* mapping (TR=1500ms, 30 gradient echoes, ΔTE=2.5ms), BVf/VSI mapping [4] (multiple gradient-echoes spin-echo sequence, before and 3min after injection of 200μmol/kg of iron oxide particles (USPIO: Combidex®/Sinerem®, Amag Pharmaceuticals/Guerbet): TR=6000ms; ΔTEGE=3ms; TESE=60ms). The entire MRI protocol lasted 1h15 per animal. Processing was performed with the Matlab environment and using home-made software. B0 map was obtained by unwrapping the phase maps of the 3D GE sequence [5]. This B0 map was used to compute F(t) in [1]. T2 was computed using a non-linear fit algorithm and a two-parameter exponential decay. BVf and VSI were obtained with the formula given in [4] using 700μm2/s for the apparent diffusion coefficient and 0.28ppm for the increase in intravascular magnetic susceptibility due to the injection of USPIO [4]. To compute lSO2 maps, Eq [1] was fitted to the MR gradient-echo data. Since maps of BVf, R2, and F(t) are available, the fitted parameters were Cte and lSO2. Data, averaged across rats in each group, are presented for 2 regions of interest (healthy striatum and tumor or stroke region). Student t-tests (after assessment of variance homogeneity) were used to assess differences (**:p<0.01, ***:p<0.001). Results