INTRODUCTION

The cerebral metabolic rate of oxygen consumption (CMRO₂) is an important metric of brain metabolism. Accumulation of amyloid-beta in the brain, one of the hallmarks of Alzheimer’s disease, has been associated with a defect in mitochondrial function.¹ Relevant non-invasive tools to measure CMRO₂ with high spatial resolution are required to assess mechanisms leading to AD-related mitochondrial dysfunction and propose new therapies. Positron Emission Tomography of inhaled ¹⁵O₂ is the gold standard for CMRO₂ measurements, but suffers from low spatial resolution, which is limiting for mice studies. Moreover, the half-life of ¹⁵O is very short (120s) and direct access to a cyclotron is required. As an alternative, the MR visible stable isotope ¹⁷O can be used in a similar manner: ¹⁷O-labeled oxygen gas is administered and metabolically produced H₂¹⁷O MR signal can be detected.² In this study, we used 3D zero echo time (ZTE-)MRI of H₂¹⁷O in the APP/PS1_dE9 mouse model of amyloidosis and showed a reduction of CMRO₂.

METHODS

Subjects. APP/PS1_dE9 mice (n=4, 24.4±1.4g, age=13±1 months) and wild-type littermates (CTR, n=4, 24.8±2.1g, age=11.6±0.3 months) were used.
Anesthesia. Animals were first anesthetized with isoflurane mixed in medical air (1-2%) for insertion of a tail vein catheter. Mice were placed prone on a water-heated bed, equipped for temperature and breathing rate monitoring. Mice received an initial I.V. bolus of medetomidine (0.1 mg/kg domitor®, Vetoquinol) followed by infusion (0.2mg/kg/h) for maintenance and isoflurane was progressively decreased to 0% within 10min. CMRO₂ measurements were performed 45-60min later, when physiological parameters were stable (~120-140 breath/min, ~36.5°C).
MR experiments. Acquisitions were performed at 11.7T (Bruker) using Paravision 6.0.1. A quadrature volume ¹H-coil was used for shimming and acquisition of RARE images (TR/TE=3000/30ms; RARE factor 8; FOV: 20x20mm²; matrix size: 192x192; slice thickness: 0.5mm). Because ¹⁷O relaxation times are very short (T1/T2~6/1.5ms), choosing an optimal short TR/TE acquisition strategy is critical. We compared the performance of an optimized 3D-MRSI sequence (TE=0.3ms) with that of 3D-ZTE with identical TR (1.8ms) and spatial resolution (1.5x1.5x1.5mm³) on an unlabeled-water phantom using a home-built linear surface ¹⁷O-coil (10mm). The signal-to-noise ratio (SNR) was more than 2-fold-higher in ZTE than in MRSI for similar acquisition times (SNR_ZTE= 45 vs. SNR_CSI=20; Fig. 1). Subsequent in vivo ¹⁷O studies were conducted using 3D-ZTE (5-μs broad pulse, TR:1.3ms, BW:18 kHz, FOV:48x48x48mm³, matrix size:32x32x32, NA: 4). These parameters yielded a nominal spatial resolution of 1.5x1.5x1.5mm³ and a time resolution of 17s. 3D-ZTE were continuously acquired before (10min), during (200s), and after (15min) inhalation of 70%-enriched ¹⁷O₂ (Nukem Isotopes). To that effect, the breathing circuit was transiently switched to a home-built gas delivery system (Fig. 2) that automatically delivered 160mL ¹⁷O-enriched O₂ over 200s.
Data analysis. After 3D regridding in Paravision, data were exported in Matlab and Hamming filter and Fourier transform were applied. This yielded an effective spatial resolution of 3x3x3mm³. Time-courses were normalized to baseline assuming [H₂¹⁷O]=20.35μmol/g, and CMRO₂ maps were derived by fitting signal changes during inhalation to a linear model, as previously described.³ For group comparison, 8 voxels (effective volume: 100μL) were conservatively chosen within the brain of each subject (Fig. 3a) for signal averaging. Whole brain and ventricles volumes were determined by automated segmentation using an in-house developed python library (https://sammba-mri.github.io/). Mann-Whitney’s U tests were used for group comparison. Data are presented as mean±standard deviation.

RESULTS

H₂¹⁷O signal started to increase immediately upon inhalation of labeled gas (Fig.3b), showing incorporation of ¹⁷O₂ into H₂¹⁷O via mitochondrial metabolism. CMRO₂ values were significantly lower in APP/PS1_dE9 (1.52±0.07 μmol/g/min) than in CTR (1.27±0.08 μmol/g/min) (Fig.3c, **p=0.0028). After switching back to ¹⁶O₂, H₂¹⁷O signal decayed and reached a new steady state, higher than baseline signal. Whole brain and ventricles volumes were not significantly different between groups (4.5±0.2 μL in APP/PS1_dE9 and 4.3±0.1μL in CTR).

DISCUSSION

We found that ZTE was advantageous over CSI to detect rapidly decaying ¹⁷O signal, allowing higher SNR. While it was possible to determine CMRO₂ on a single voxel basis (Fig.4), such resolution still yielded significant partial volume between the different structures at the mouse brain scale, and we chose to average the signal from 8 voxels for robust individual measurements. CMRO₂ values obtained in CTR (1.52±0.07 μmol/g/min) were lower than previously reported (2-3 μmol/g/min).^4,5 This difference may simply reflect a different physiological state in our work, possibly due to the older age of the animal, lower body temperature, blood oxygenation⁶ and/or, importantly, the effect of anesthetic on cerebral blood flow⁷ (CBF) and glucose oxidation rates.⁸ Previous studies used isoflurane, known to induce vasodilation and increase CBF, while medetomidine induces lower resting CBF,⁷ possibly affecting CMRO₂. Oxygen metabolism might also vary in hyperoxic conditions, such as the 100% O₂ inhalation used in our study. Partial volume could constitute a methodological bias, as we cannot exclude that ventricles and other non-brain tissues contribute to the measured ¹⁷O signal, which would lead to an underestimation of CMRO₂. However, partial volumes were also present in former works in rodents, and no significant difference was detected between ventricles sizes in APP/PS1_dE9 and CTR mice in this study. Therefore, the lower CMRO₂ in APP/PS1_dE9 compared to CTR in our conditions likely reflects mitochondrial dysfunction.

Acknowledgements

This work was partly funded by NeurATRIS: A Translational Research Infrastructure for Biotherapies in Neurosciences (“Investissements d'Avenir”, ANR-11-INBS-0011)