Measuring Tumor Metabolism in Pediatric Diffuse Intrinsic Pontine Glioma Using Hyperpolarized Carbon-13 MR Metabolic Imaging

Objective The purpose of this study was to demonstrate the feasibility of using hyperpolarized carbon-13 (13C) metabolic imaging with [1-13C]-labeled pyruvate for evaluating real-time in vivo metabolism of orthotopic diffuse intrinsic pontine glioma (DIPG) xenografts. Materials and Methods 3D 13C magnetic resonance spectroscopic imaging (MRSI) data were acquired on a 3T scanner from 8 rats that had been implanted with human-derived DIPG cells in the brainstem and 5 healthy controls, following injection of 2.5 mL (100 mM) hyperpolarized [1-13C]-pyruvate. Results Anatomical images from DIPG-bearing rats characteristically exhibited T2-hyperintensity throughout the cerebellum and pons that was not accompanied by contrast enhancement. Evaluation of real-time in vivo13C spectroscopic data revealed ratios of lactate-to-pyruvate (p < 0.002), lactate-to-total carbon (p < 0.002), and normalized lactate (p < 0.002) that were significantly higher in T2 lesions harboring tumor relative to corresponding values of healthy normal brain. Elevated levels of lactate in lesions demonstrated a distinct metabolic profile that was associated with infiltrative, viable tumor recapitulating the histopathology of pediatric DIPG. Conclusions Results from this study characterized pyruvate and lactate metabolism in orthotopic DIPG xenografts and suggest that hyperpolarized 13C MRSI may serve as a noninvasive imaging technique for in vivo monitoring of biochemical processes in patients with DIPG.


Introduction
Diffuse intrinsic pontine glioma (DIPG) comprises a heterogeneous class of childhood brainstem cancers that defy molecular stratification and surgical intervention because of their sensitive location. After forming in the pons, this relatively inaccessible disease often undergoes rapid growth that is characterized by diffuse infiltration across the midline through healthy tissue. Despite decades of clinical trials investigating the efficacy of novel treatment regimens, DIPG remains the leading cause of death among pediatric patients with central nervous system cancers and carries an average survival of only 9 months [1].
Magnetic resonance imaging (MRI) serves as the standard modality for diagnosing DIPG and monitoring disease status in response to treatment [2]. While conventional MRI provides information regarding anatomical changes, its prognostic value and ability to assess physiologic or functional alterations associated with therapeutic efficacy are highly limited [3]. Attempts at using positron emission tomography to evaluate DIPG metabolism have also presented challenges owing to the risks of ionizing radiation exposure [4,5]. Given the heterogeneous presentation of the disease on imaging, which lacks features for stratifying aggressiveness [6], the development of noninvasive metabolic imaging methods may enhance the evaluation of molecular characteristics as well as response to therapy.
Representing a novel advance in solid state physics, dissolution dynamic nuclear polarization (d-DNP) enables the acquisition of carbon-13 ( 13 C) magnetic resonance data with an appreciable gain in sensitivity over conventional methods [7]. A phase I first-in-human study using hyperpolarized 13 C magnetic resonance spectroscopic imaging (MRSI) has already demonstrated the safety and feasibility of such technology for probing real-time metabolism in prostate cancer patients [8]. e purpose of this study was to explore the feasibility of using hyperpolarized 13 C metabolic imaging with [1-13 C]-pyruvate for evaluating real-time in vivo metabolism of orthotopic DIPG xenografts.

Materials and Methods
Eight six-week-old male athymic rats (rnu/rnu, homozygous, and median weight � 290 g) purchased from Harlan (Indianapolis, IN) were implanted with patient-derived human DIPG cells (SF8628) in the brainstem to create an orthotopic DIPG model, while 5 healthy rats served as experimental controls. e details of the cell culture and intracranial implantation procedures have been described elsewhere [9]. Study procedures were approved by the Institutional Animal Care and Use Committee.
All animals were scanned on a 3Tclinical MRI system (GE Healthcare, Waukesha, WI, USA) equipped with a customdesigned 1 H/ 13 C rat coil on approximately the 58th day after tumor implantation. e body temperature was maintained using a heated pad positioned inside the RF coil. Anesthesia was maintained with a constant delivery of 1.5% isoflurane. For the polarization of pyruvate, a 35 μL of [1-13 C]-pyruvate mixed with 15 mM OX063 trityl radical (GE Healthcare, Oslo, Norway), and 1.5 mM gadolinium (Gd)-DOTA was polarized using a HyperSense ® (Oxford Instruments, Abingdon, UK) [7,10]. After 60 minutes of microwave irradiation, the mixture was dissolved in a saline solution with 5.96 g/L Tris (40 mM), 4.00 g/L NaOH (100 mM), and 0.1 mg/L Na 2 ethylenediaminetetraacetic acid. e final solution had a concentration of 100 mM pyruvate, and pH∼7.5. 2.7 mL of the dissolved pyruvate solution was injected into the tail vein of the rats over 10 s duration. e following 1 H and 13 C data were acquired in sequence for each scan: (1) axial T 2 -weighted images using a fast spinecho sequence (echo time/repetition time � 60/4000 ms, 8 cm field of view, 256 × 256 matrix, and 2 mm slice thickness), (2) compressed-sensing 13 C 3D MRSI data (echo time/repetition time � 140/215 ms, phase encoding in x and y axes, flyback echo-planar readout in z-axis, 20 × 16 × 16 matrix, and 2 × 2 × 5.4 mm spatial resolution) [11] acquired at 20 s from the start of the pyruvate injection, and (3) axial T 1 -weighted images using a spin-echo sequence (echo time/repetition time � 10/700 ms, 8 cm field of view, 320 × 192 matrix, and 1.2 mm slice thickness) after the injection of 0.2 mmol/kg Gd-DTPA. e methods for processing 13 C MRSI data have been described previously [11]. For quantification of 13 C metabolites, the ratio of lactate-to-pyruvate and lactate-to-total carbon (tC: sum of lactate, pyruvate-hydrate, alanine, and pyruvate) was calculated. In addition, the lactate and pyruvate signals in the brainstem were normalized with respect to the relative signals of the normal brain in the supratentorial region (Figure 1(a)). ROIs were manually contoured on T 2 -weighted images for the T 2 -hyperintense lesion, and the percentage of T 2 lesion volume was calculated for each voxel. Comparison of mean 13 C parameters from the T 2 -hyperintense lesion (voxels with >75% T 2 lesion) and similar region of the infratentorial brain of healthy control animals was performed using the Mann-Whitney rank-sum test. In order to evaluate the spatial variation of 13 C metabolites, the T 2 -hyperintense lesion was also compared with the contralateral brain (voxels with nonhyperintense tissue on the opposite side of the T 2 lesion).
Immediately following the scan, tumor-bearing animals were sacrificed and harvested for their brains, which were fixed in phosphate-buffered 4% formalin. Samples were then dehydrated by graded ethanol and embedded in Paraplast Plus wax (McCormick Scientific). 5 µm sections were examined following haematoxylin and eosin (H&E) staining.

Results and Discussion
Representative anatomical data from a rat injected with DIPG cells are shown in Figure 1, with panel (a) providing an overview of an orthotopic lesion around the brainstem on a sagittal T 2 -weighted image. e corresponding axial T 2weighted image exhibited hyperintensity throughout the cerebellum and pons (Figure 1(b)), while no contrast enhancement was visible from the post-Gd T 1 -weighted image acquired at the same location (Figure 1(c)). e axial T 2 -weighted image in Figure 2(a) depicts a 13 C MRSI grid superimposed over the brainstem. e corresponding hyperpolarized 13 C spectra from the same animal (Figure 2(b)) demonstrated the spatial distribution of high 13 C-labeled lactate and pyruvate signals over the entire brainstem region. Illustrative of the full cohort, these hyperpolarized 13 C MRSI data demonstrated levels of lactate in the T 2 lesions (pink voxels ; Figures 2(a) and 2(b)) that were elevated relative to the contralateral normal brain (blue voxels; Figures 2(a) and 2(b)).
Metabolite parameters derived from hyperpolarized data are compared between DIPG xenografts and healthy control brains in Table 1.
e T 2 -hyperintense tumors exhibited highly elevated metabolism compared to both healthy controls and the contralateral hemisphere, which may contain infiltrating tumor. e ratios of lactate-topyruvate, lactate-to-total carbon, and normalized lactate in T 2 lesions (0.70 ± 0.24, 0.36 ± 0.08, and 2.9 ± 1.1, resp.) were significantly higher than the corresponding values in the healthy normal brain (0.20 ± 0.06, 0.14 ± 0.03, and 1.1 ± 0.25, resp.). e normalized lactate map in Figure 2(c) shows the differential production of lactate between the DIPG xenograft and tissue in the contralateral hemisphere. From the corresponding H&E-stained slice, there was observed infiltrative, viable tumor that recapitulated the histopathology of pediatric DIPG (Figure 2(d)). In contrast, the normalized pyruvate was found to be similar across both regions and comparable to the healthy brain (Table 1).

Contrast Media & Molecular Imaging
In order to assess the ability to observe longitudinal changes in metabolism, hyperpolarized 13 C MRSI data were obtained from an additional single animal imaged over a period of 7 days in the course of tumor development (Figure 3). e longitudinal change in normalized lactate signal and T 2hyperintensity are shown in Figure 3. Normalized lactate from the 13 C spectral data imaged at 42, 46, and 48 days from implantation were 1.2 ± 0.2, 2.5 ± 0.1, and 3.5 ±1.0, respectively. ese data show a severalfold increase in the metabolic abnormality associated with the evolution of the anatomic lesion.
is study has demonstrated the feasibility of using hyperpolarized 13 C metabolic imaging to assess in vivo metabolism in orthotopic brainstem xenografts that contain patient-derived primary DIPG cells. By using hyperpolarized [1-13 C]-pyruvate in conjunction with rapid 3D MRSI acquisition techniques, it was shown that nonenhancing brainstem glioma can be evaluated on the basis of real-time molecular data, as an initial step towards noninvasive disease characterization. To the best of our knowledge, this is the first study to apply hyperpolarized 13 C techniques in brainstem tumor as well as in nonenhancing brain tumor.
An important feature of the orthotopic murine tumor model adopted here was its ability to recapitulate aspects of disease observed in patients. With regard to imaging, the longitudinal data revealed a similar pattern of disease progression, wherein cells implanted in the pons spread from a localized lesion to the cerebellum after a period of rapid growth, and without visible enhancement [12,13]. Analysis of the resected brain by histopathology confirmed viable DIPG in the pons, along with the associated cerebellar infiltration, which supported findings from metabolic imaging indicating temporal changes.
Because DIPG is radiographically characterized by poorly perfused and faintly enhancing heterogeneous lesions, its diagnostic assessment remains challenging. In this context, the relative elevation of lactate in nonenhancing lesions compared to healthy control tissue was a defining feature that may hold diagnostic value for patients as an in vivo marker of disease. As these tumors frequently display high levels of LDHA [14,15] that preferentially convert pyruvate to lactate, imaging of hyperpolarized [1-13 C]-pyruvate might offer a targeted means of monitoring tumor growth and disease status. e nuclear polarization techniques implemented here provided sufficient signal enhancement to detect real-time pyruvate-to-lactate conversion in the brainstem with high sensitivity, as well as distinguish longitudinal variation in metabolism from growing tumor over relatively brief intervals. Based on the quality and spatial resolution of the 13 C spectra achieved via hardware and sequence performance, it was possible to evaluate metabolic differences between T 2 -hyperintense lesions and contralateral brain tissue.
Although our focus was on demonstrating feasibility, we believe that a promising application of this technique may be monitoring response to treatment in patients with DIPG, Significant difference between T 2 -hyperintense lesion and contralateral brain (p < 0.001). b Significant difference between T 2 -hyperintense lesion and healthy rat brain (p < 0.002). given the inadequacy of conventional MR imaging. By administering [1-13 C]-pyruvate as a hyperpolarized substrate with measurable conversion to [1-13 C]-lactate, studies have already managed to provide evidence for both localization of malignant tissue and treatment-induced reduction of metabolic activity arising from growth arrest or apoptosis [16,17]. A recent study has demonstrated the first application of hyperpolarized 13 C MR metabolic imaging in patients with supratentorial glioma and presented the safety and feasibility of using hyperpolarized [1-13 C]-pyruvate to evaluate in vivo brain metabolism [18]. Perhaps the greatest technical challenge to translating hyperpolarized imaging to the clinic for diffusing intrinsic pontine glioma is ensuring adequate SNR in the brainstem, where the surrounding tissue is less perfused and far removed from coil elements. While several single-and multivoxel 1 H spectroscopy studies have indicated that 1 H magnetic resonance spectroscopy (MRS) may be useful for assessing disease progression and monitoring response to treatment [19][20][21][22], the acquisition of proton spectra in the infratentorial region is generally challenging due to susceptibility effects around the brainstem region and confounded by overlapping lipid peaks that reflect contamination from the surrounding skull. e susceptibility effects encountered in 13 C MR are considerably smaller relative to 1 H MR by virtue of the 13 C gyromagnetic ratio, which is one-fourth that of 1 H. e proposed method of assessing real-time metabolism using hyperpolarized 13 C MRSI, combined with anatomical MRI and 1 H MRS, may provide complementary information that is of value in assessing disease status and response to treatment in DIPG.
Interestingly, the ratio of lactate-to-pyruvate in the nonenhancing T 2 lesion from this study (0.70 ± 0.24) was significantly smaller than that of enhancing tumor from supratentorial orthotopic glioblastoma xenografts in a previous study (1.0 ± 0.36) (p < 0.02, unpaired t-test), while it remained similar between contralateral brainstem tissue (0.28 ± 0.11) and contralateral supratentorial brain tissue (0.29 ± 0.17) [11]. Future studies will attempt to elucidate the molecular and pathologic mechanisms that produce different characteristics in pyruvate metabolism depending on the type of glioma.

Conclusions
e results from this study characterized pyruvate and lactate metabolism in orthotopic DIPG xenografts and suggest that hyperpolarized 13 C pyruvate MRSI is a promising noninvasive imaging tool for the in vivo monitoring of biochemical processes in DIPG.

Data Availability
e data used to support the findings of this study are available from the corresponding author upon request.

Disclosure
An earlier version of this work was presented as a conference abstract at the Joint Annual Meeting ISMRM-ESMRMB on 10-16 May 2014 in Milan, Italy.

Conflicts of Interest
ere are no conflicts of interest to report in this study.