Introduction

Metabolic reprogramming, also known as the Warburg Effect in which lactate production via glycolysis (GLY) is favored over oxidative phosphorylation (OXPHOS), is a hallmark of cancer¹. This effect is particularly pronounced in glioma and other brain tumors and is the target of multiple oncologic therapies². However, limitations on measuring this abnormal tumor metabolism remain a crucial obstacle to clinical translation.

The recent introduction of deuterium metabolic imaging (DMI) offers a promising approach for in vivo imaging of the Warburg effect. de Feyter et al. showed DMI can map the balance of GLY and mitochondrial metabolism in patients with glioma at 4T 60-75 min following oral consumption of [6,6’-²H₂]Glc ³. The ratio of ²H-Lac/Glx (glutamate+glutamine) provides a direct measure of the Warburg effect. In a preclinical animal model study, Kreis, et al. further showed that dynamic DMI acquisitions could also be used to acquire quantitative and spatially resolved measurements of glycolytic flux in tumors, which can be used to assess treatment response⁴. In this preliminary study, we demonstrate high-quality whole-brain metabolic maps of deuterated glucose metabolic products in both healthy adults and patients with Central Nervous System (CNS) lesions, and explore the extent to which DMI can quantify the metabolic heterogeneity across these patients.

Methods

Two normal control volunteers and three patients with varying CNS lesions were scanned. Two of these patients had diseases which were considered well-managed at the time of imaging – one with an anaplastic oligodendroglioma, Grade III but without evidence of active disease (Patient A), and the second with a glioblastoma stable on an anti-VEGF drug for several months (Patient B). The third patient had a cerebellar pilocytic astrocytoma considered to be anaplastic and behaving relatively aggressively (Patient C).

All DMI scans were acquired on a GE 3T PET/MR scanner using a modified gradient filter and ²H volume RF coil approximately 45 minutes following the oral ingestion of 60g of deuterated glucose ([6,6’-²H₂]Glc) as in [5]. We used spherical k-space encoded chemical shift imaging (CSI) and a 90° non-selective RF excitation pulse, a 10x10x10 cm spatial array, 2.4 cm isotropic resolution, a repetition time (TR) of 350 ms, and 13 averages yielding a total scan time of 39 minutes. Additional proton scans (T1, T2 FLAIR, and ASL) were acquired using a multi-channel proton coil to better visualize anatomy and pathology. A lower-quality reference proton scan was also taken with the same reference position to the DMI scan with the scanner’s built-in body coil to co-register the anatomical proton scans to the DMI dataset.

DMI datasets were processed using DMIWizard, an open-source Matlab-based toolbox [6]. Raw k-space data were zero-padded, phase-corrected, Fourier Transformed and spectrally apodised by 5 Hz. The subsequent spectra were quantified through least-squares fitting to generate metabolic maps of deuterated water (HDO), glucose (²H-Glc), glutamine/glutamate (²H-Glx) and lactate (²H-Lac). The ratio of Lac/(Glx+Lac) was used to quantify the fraction of glucose metabolism that undergoes GLY (as opposed to OXPHOS), a possible measure of the Warburg effect. The additional proton scans acquired with a conventional multi-channel head coil were aligned to the body-coil proton scan (and therefore the DMI dataset) acquired with the ²H coil in place via rigid registration using the SimpleITK toolkit [7].

Results

Representative metabolic maps, corresponding anatomical scans, and example spectra of a healthy control, a patient with a well-controlled CNS lesion (Patient A), and a patient with a more aggressive CNS lesion (Patient C) are shown in Figures 1, 2 and 3 respectively. Figure 4 compares the Lac/(Glx+Lac) fraction across all five subjects. 3D whole-brain coverage is achieved in these DMI scans, including areas typically plagued by inhomogeneities in proton MRS such as near the brainstem. Good anatomical correspondence between the cerebrospinal fluid and lowered ²H-Glx is notable in the healthy control (Figure 1). Patient A shows a decrease in ²H-Glx in the necrotic region of the lesion without a corresponding increase in ²H-Lac (Figure 2). Patient C shows decreased Glx as well in the necrotic region of the lesion, with a moderate increase in Lac as seen most clearly on the Lac/(Glx+Lac) map (Figure 3). All patients show an elevated “Lac” signal near the skull due to partial voluming from the natural abundance ²H-lipid peaks overlapping the lactate resonance.

Discussion

Following the oral ingestion of deuterated glucose, we used a modified gradient filter, home-built ²H volume RF coil, and spherical k-space sampling in a 3D CSI acquisition to obtain high quality whole-brain metabolic images of ²H-labeled water and glucose metabolic products. We explored the ability to quantify local glucose metabolism in correspondence to anatomical landmarks across patients with varying CNS lesions, and see evidence of the Warburg effect in one patient with the most active disease. Quantification is limited by SNR and spatial resolution, which in future work will be improved through optimizing pulse sequence parameters and machine-learning based multi-modal image denoising.

Acknowledgements

NIH R01 CA245097, Berg LLC