Glutamine-based PET imaging facilitates enhanced metabolic evaluation of gliomas in vivo

¹⁸F-FGln shows high uptake in gliomas but negligible uptake in the normal brain

Because glutamine is an important source of replenishment of TCA cycle metabolites (termed anaplerosis) in many cancers, we first evaluated the contribution of glutamine to TCA cycle anaplerosis in glioma cell lines bearing various oncogenes: U87-MG (PTEN −/−), TS543 (PDGFRA, PTEN −/−), and TS603 (IDH1m). Glutamine was the main TCA cycle anaplerotic substrate in all cell lines tested (Figure S1), underscoring the importance of glutamine as a key nutrient in gliomas and the rationale for developing glutamine-based PET imaging of gliomas. To determine if fluorinated glutamine is metabolized to glutamate, we compared ¹⁹F-FGln (identical to ¹⁸F-FGln, except that ¹⁸F is replaced with the more stable fluorine isotope ¹⁹F to enable GC-mass spectroscopy analyses) with [U-¹³C]-glutamine metabolism in the above cell lines. Whereas [U-¹³C]-glutamine was metabolized to [U-¹³C]-glutamate, no ¹⁹F-glutamate was detected (in comparison to standards) suggesting that ¹⁹F-FGln is not converted to ¹⁹F-glutamate (Figure S2). These in vitro data with ¹⁹F-FGln are similar to ¹⁸F-FGln data in vivo, where a only minor fraction of ¹⁸F-FGln (~9% in animal tumor xenografts (8)) is converted to ¹⁸F-glutamate, suggesting that ¹⁸F-FGln PET imaging is mainly a measure of glutamine uptake.

To assess glutamine uptake in vivo in gliomas we used PET imaging with the glutamine analogue ¹⁸F-FGln (Figure 1A). Biodistribution studies with ¹⁸F-FGln in normal mice (Figure S3) showed that ¹⁸F-FGln crossed the blood-brain barrier (BBB) but showed minimal uptake in the normal brain compared to other organs such as the pancreas and the gut (Figures 1 and S3). We thus postulated that high glutamine utilization in gliomas (Figure S1) coupled with minimal uptake in the normal brain (Figures 1 and S3) would result in high tumor-to-background ratios. To test this, we compared ¹⁸F-FGln and ¹⁸F-FDG uptake in the normal mouse brain with glioma xenografts. Xenograft models were created by subcutaneously injecting TS543 (PDGFRA, PTEN −/−), TS603 (IDH1m), TS598 (EGFR), and U87-MG (PTEN −/−) glioma cells into SCID mice. Xenografts histopathologically resembled human gliomas (9) (Figure S4A–D). Fasting or perfusion did not have any effect on ¹⁸F-FGln uptake in glioma subcutaneous xenografts (Figure S5). Further, ¹⁸F-FGln uptake was significantly (p=0.0088) lowered in xenografts on co-injection of excess cold ¹⁹F-FGln (Figure S3). All glioma subcutaneous xenografts showed significantly (p<0.0001) higher uptake of ¹⁸F-FGln compared to the normal brain (Figure 1 and Figure S6, whole body images). ¹⁸F-FDG showed higher or equivalent uptake in the normal brain compared to all tested glioma subcutaneous xenografts (Figure 1 and Figure S6, whole body images).

In IDH1-mutant glioma cells, we found that [U-¹³C]-glutamine was metabolized to generate the oncometabolite 2-hydroxyglutarate. This was reversed by either shRNA knockdown of total IDH1 (mutant and wild type) or an IDH1m-specific inhibitor (IDH35) (11) in vitro (Figures S1 I and S7). Glioma xenografts bearing shRNA against total IDH1 did not show significant alterations in ¹⁸F-FGln uptake compared to controls (Figure S7). This suggests that ¹⁸F-FGln uptake in vivo in TS603 IDH1m cells is not altered when the activity of mutant IDH1 generating 2-HG is inhibited (Figure S7).

¹⁸F-FGln enables distinct delineation of gliomas in vivo

We next evaluated well-characterized genetically engineered glioma mouse models using the RCAS/tv-a system driven by PDGF with a PTEN-null background (RCAS-PDGF, PTEN −/−), which develop GBM histologically identical to human GBM (Figure S4E and G) (12). PET imaging in these animals showed marked ¹⁸F-FGln uptake with distinct tumor delineation (compared to the surrounding brain) in a region directly corresponding to the tumor, as detected by MRI and confirmed by autoradiography and histopathology (Figure 2A–F). Tumor-to-background ratios with ¹⁸F-FGln ranged from 4:1 to 6:1 compared to ~1:1 tumor-to-background ratio with ¹⁸F-FDG (Figure 2F). Similar findings were observed in RCAS-PDGF animals on a PTEN wild type background (Figure S8A–D) and in an orthotopic IDH1m glioma model (Figure 2G–L and S8 E–G).

¹⁸F-FGln uptake is not observed in neuroinflammation or BBB disruption

To evaluate if inflammatory cells contributed to ¹⁸F-FGln uptake, we created mouse models of neuroinflammation (Figure 3A). Lipopolysaccharide (LPS) injection into the brain resulted in global activation of IBA1 immunoreactive microglia/macrophages in the injected hemisphere well beyond the injection site (Figures 3B and S9A–D). We also used IFN-γ which polarizes macrophages/microglia towards classic activation, and IL-4, which mediates an alternative activation phenotype similar to tumor-associated macrophages (13). Robust neuroinflammation was confirmed histopathologically (Figure 3B and Figure S9A–D). Mice were imaged at the peak of neuroinflammation with ¹⁸F-FGln and ¹⁸F-FDG (Figures 3C–E and S9E–G). ¹⁸F-FDG PET is generally a poor measure of neuroinflammation in mouse brains in vivo because of the high background uptake in the brain (14). Consistent with this, no lesion was discernible with ¹⁸F-FDG PET (Figure 3D and S9F). Conversely, ¹⁸F-FGln did not show any uptake at the site of the lesion in neuroinflammatory models, as opposed to the high uptake seen in gliomas (Figure 3C and Figure S9E).

To address the concern of BBB permeability as a factor in causing increased ¹⁸F-FGln delivery to the brain, we used an adenosine receptor agonist that is an established means of increasing BBB permeability by increasing spaces between endothelial cells in brain capillaries (15, 16). NECA [1-(6-amino-9H-purin-9-yl)-1-deoxy-N-ethyl-β-D-ribofuranuronamide], a combined A1 and A2 adenosine receptor agonist, significantly (p=0.034) increased BBB permeability in mice as determined by measuring extravasation of fluorescent-labeled dextran into the brain (Figure 3F and G). We did not see increased ¹⁸F-FGln delivery to the brain in NECA-treated animals (Fig 3H–K), suggesting that increased BBB permeability by itself does not artificially increase ¹⁸F-FGln delivery to the brain. This demonstrates that neuroinflammation or a breach in the BBB using NECA do not significantly contribute to ¹⁸F-FGln uptake.

¹⁸F-FGln uptake in gliomas is reduced after chemo/radiation therapy

To evaluate if ¹⁸F-FGln can monitor treatment efficacy and differentiate tumor from post-treatment changes, we studied a well-characterized RCAS-PDGF-PTEN null mouse model of radiation/chemotherapy (17) (Figures 4A and S10A–C). The same mouse was imaged before therapy and after treatment with MRI, ¹⁸F-FGln, and ¹⁸F-FDG (Figure 4B–F). As can be seen in post-treatment changes in human patients, the T2-weighted MRI showed no appreciable differences when comparing before-treatment scans with after-treatment scans (18). In contrast, ¹⁸F-FGln showed a significant reduction (p=0.0078) in tumor uptake after treatment, which was confirmed by autoradiography (Figure 4C–F).

To assess the mechanism by which chemo/radiation therapy results in a decrease in ¹⁸F-FGln uptake, we compared before-treatment and after-treatment glioma brain tissues. Tumor burden was dramatically reduced after treatment, as demonstrated by a reduction in the Olig2-positive tumor cells (p=0.031) (Figure S10A–B). Brain tissues after treatment showed increased reactive gliosis in the form of GFAP-positive astrocytosis and abundant IBA1-positive macrophages/microglia (Figure S11A–B). No significant changes were noted in CD3-positive lymphocyte infiltration or in blood vessels as delineated by CD31 staining for endothelial cells and collagen staining for vessel walls (Figure S11C–E). Moreover, expression of the glutamine transporter SLC1A5 in residual tumor cells after treatment was not significantly different from that seen in tumor cells before treatment (Figure S11F). These data suggest that reduction in ¹⁸F-FGln uptake after chemo/radiation therapy is due a decrease in tumor burden. Further, these data imply that ¹⁸F-FGln is taken up specifically in tumor cells, can monitor treatment responses, and potentially differentiate post-treatment changes from tumor in glioma animal models.

¹⁸F-FGln shows high tumor to brain uptake ratio in human gliomas with progression

To determine if our findings could be translated to a spectrum of human gliomas apart from the animal models studied, we evaluated the expression of the glutamine transporter SLC1A5, which in part mediates ¹⁸F-FGln uptake (8, 19), as demonstrated by siRNA knockdown and pharmacologic inhibition (Figure S12A–B). Immunohistochemistry in normal brain tissue samples (n=8, Figure S12C) showed minimal expression, mirroring low background ¹⁸F-FGln uptake. In contrast, SLC1A5 demonstrated higher expression in various gliomas (n=64), consistent with high ¹⁸F-FGln-glioma uptake (Figure S12 C–D). Further, we queried the GBM cancer genome atlas for SLC1A5 mRNA expression, which was noted in all GBM subtypes (Figure S12E). These data suggest that ¹⁸F-FGln uptake is not restricted to the animal model glioma genotypes tested and may be of broader utility in evaluating uptake of glutamine as a key nutrient in gliomas.

We translated these findings in vivo to human glioma subjects as an initial investigation of ¹⁸F-FGln PET imaging in human cancer patients (Phase I study, Figures 5A–L, S13, S14 A–C and S15 A–C). As part of this study, we compared ¹⁸F-FGln uptake in 3 glioma patients with clinical progression of disease and 3 patients with stable disease (Tables S1 and S2). Normal brain parenchyma showed minimal ¹⁸F-FGln uptake, and ¹⁸F-FGln-avidity was noted in all tumors that showed progression within the 3 patients (tumor:brain ratio range: 3.7–4.8, Figures 5L and S14 and Tables S2–S3). In contrast, clinically stable tumors showed minimal or no ¹⁸F-FGln-avidity on PET (Figure S15 and Table S2). Normal brain tissues in these same patients demonstrated high ¹⁸F-FDG avidity, with normal brain ¹⁸F-FDG concentrations (SUV) equivalent to or greater than tumor SUV values (tumor:brain ratio range: 0.9–1.0, Figure 5L and Table S2). For example, in patient #5, ¹⁸F-FDG could distinguish the posterior portion of the tumor (Figure 5C, 3 red arrows) from the surrounding brain, but not the anterior part (2 red arrows, Figure 5C). In contrast, both regions of the tumor showed high uptake with ¹⁸F-FGln (Figure 5B & E). This is an important issue considering the infiltrative nature of gliomas. Further, this patient’s tumor demonstrated mild contrast enhancement on gadolinium-enhanced MRI (Figure 5A), but high ¹⁸F-FGln avidity (Figures 5B and E) and retention of ¹⁸F-FGln compared to its rapid clearance in the blood (Figure 5F). Whole body PET images, plasma and blood time activity curves, biodistribution, tracer metabolite analysis, and dosimetry from these glioma patients are provided in the supplementary information (Figures S14–S16 and Table S3–5) These findings in human subjects recapitulate our ¹⁸F-FGln PET and ¹⁸F-FDG PET findings in mouse models and demonstrate that clinical ¹⁸F-FGln PET can evaluate high-grade glioma in vivo and may be potentially useful in identifying tumors undergoing transformation.

Study Design

The objective of this study was to address the clinically unmet need of assessing metabolic nutrient uptake in gliomas in vivo. This goal was addressed by (1) evaluating glutamine uptake in gliomas by PET imaging in clinically relevant glioma animal models, (2) assessing if glutamine PET-based imaging could monitor therapeutic response in glioma animal models, and (3) translating these findings to human subjects.

All animal experiments were performed after approval from the MSKCC Institutional Animal Care and Use Committee and were conducted as per NIH guidelines for animal welfare. The utility of glutamine-based PET imaging to monitor therapeutic response was assessed using standard chemo/radiation therapy protocols in glioma animal models. Animals were randomly assigned to various test groups in a blinded manner. Individuals handling the animals and conducting animal surgeries and therapies were blinded to the experimental design. Evaluation of ¹⁸F-FGln PET uptake measurements in animals was performed in a non-blinded fashion. Because ¹⁸F-FGln microPET imaging is a new imaging technology, it is difficult to estimate sample size with adequate power. An n = 3 – 9 was selected for these well-controlled models with a low (<10%) error in consecutive studies. Cell culture studies were conducted using three independent experiments (27). No samples or animals were excluded from data analyses.

Human glioma subjects were studied as part of an in-human microdose, open label, Phase I trial of ¹⁸F-FGln PET that was approved by the institutional review board and conducted under the auspices of an FDA-approved investigational new drug application in compliance with the STAndards for the Reporting of Diagnostic accuracy studies (STARD) guidelines for diagnostic studies (trial registered at www.clinicaltrial.gov; NCT01697930). Informed consent was obtained before imaging.

Animal models of gliomas

Subcutaneous glioma xenografts were created by inoculating SCID mice (4–6 weeks of age, male and female, NOD-SCID, Taconic Farms Inc.) with 1×10⁶ U87-MG, TS603, TS543, or TS598 cells resuspended in 200 μL of a 1:1 volume of cell culture medium and matrigel (BD Biosciences). Animals were imaged when xenograft tumor volumes approached 200 mm³. Genetically modified glioma models were generated using the RCAS/tva system. 4–6 week-old, male or female C57BL6 Nestin-tv-a/Ink4a-Arf−/− mice with or without a PTEN fl/fl background were anesthetized with ketamine (0.1 mg/g) and xylazine (0.02 mg/g) and injected using a stereotactic fixation device (Stoelting) (17). One μL of RCAS-PDGF or a 1:1 mixture of 4×10⁴ RCAS-PDGF and RCAS-Cre transfected DF1 cells were delivered using a 30-gauge needle attached to a Hamilton syringe. Cells were injected into the right frontal cortex (stereotactic coordinates: bregma +1.7 mm (anterior), lateral −0.5 mm (right), and at a depth of 2.5 mm). For orthotopic injection of glioma cells, the same procedure was followed, except that TS603 IDH1m (1×10⁶ cells) was implanted into the right frontal cortex (same coordinates as above). Individuals performing implantation and stereotactic injections were blinded to the experimental design.

Small animal PET imaging

Animals were imaged with ¹⁸F-FDG and ¹⁸F-FGln (24 hours apart to allow complete decay of radiotracer) by injecting 200 μCi of the radiotracer into the lateral tail vein. PET imaging was performed using a dedicated small-animal microPET scanner (Concorde Microsystems) under 2% isoflurane anesthesia, with the tumors centered in the field of view. Dynamic imaging was performed by obtaining sixty-minute acquisitions with an energy window of 350–750 keV with a coincidence-timing window of 6 ns. For static imaging, acquisitions were collected 0.5 h, 1 h and 2 h after injection. In each animal, the MRI and PET images were directly compared. Region-of-interest (ROI) analysis of the acquired images was performed using ASIPro software (Siemens) in a non-blinded manner, and the observed maximum pixel value was represented as percent-injected dose/cubic centimeter (%ID/cc). Tumor to brain ratios were determined by normalizing tumor uptake to surrounding brain uptake.

¹⁸F-FGln synthesis

Synthesis of the precursor for ¹⁸F-(2S,4R)4F-Gln (¹⁸F-FGln) was performed as described(8, 28). Briefly, the radiolabeling procedure was similar except that the radiolabeling was performed at 90°C, and the final compound (in saline) was passed through a C¹⁸ cartridge and AG11A8 resin to formulate the final solution for injection. The purity of ¹⁸F-FGln was analyzed with a chiral column (Chirex 3126 (d)-penicillamine, 1 mM CuSO₄ solution, 1 mL/min). For human subjects, ¹⁸F-FGln was manufactured at the MSKCC Radiochemistry and Molecular Probe Core Facility pursuant to an Investigational New Drug Application (IND) acknowledged by the United States Food and Drug Administration (US FDA). Every batch of final drug product was tested to assure conformance to the drug product acceptance specifications, which included radiochemical purity (≥80%) and identity, residual solvent content, endotoxin content, radionuclidic identity, pH, and appearance. 125 MBq was adequate for imaging the brain region-of-interest and provided satisfactory PET image quality and count statistics.

¹⁸F-FGln PET-CT protocol

Microdose ¹⁸F-FGln tracer was administered to patients by single peripheral intravenous injection of 125 MBq as a slow bolus (1 minute duration) in a volume of 5–10 mL. Imaging of ¹⁸F-FGln biodistribution was obtained at approximately 30, 90, and 160 min after injection, using a Discovery DSTE PET-CT scanner (GE Healthcare) with the patient supine upon the scanner bed. After a scout X-ray, CT data were acquired with 140 kvp; 70 mA; pitch of 1.75:1; reconstructed slice thickness of 3.75 mm; 0.8 s per rotation. CT data were reconstructed in a 512 × 512 matrix using a filtered back-projection algorithm. PET emission scans were acquired starting at the proximal thigh region and ending at the head region; emission scanning specifically of the brain tumor field of view shown was acquired as a single bed position of 35 minutes duration in 3–D mode. PET images were reconstructed using an ordered subset expectation maximization iterative algorithm. PET/CT scans obtained at the 60 and 150 minutes post-injection time-points also spanned thighs to skull using the same X-ray/CT and PET emission data acquisitions and reconstructions. PET emission data were reconstructed using an ordered subset expectation maximization iterative algorithm. Emission data were corrected for random detector inhomogeneity, scatter, attenuation, deadtime, and decay. PET, CT, and fusion PET-CT images shown were generated for display using an integrated GE PACS AW Suite workstation (GE Healthcare). PET data was analyzed using HERMES workstation (Hermes Medical Solutions).

Statistical analysis

Statistical analyses were performed in consultation with the Sloan Kettering biostatistics facility. Data are represented as the means ± standard errors (s.e.m.) unless specified otherwise. Graphs were plotted and statistical analyses were performed using Prism software (version 6, Graphpad). All statistical tests were two-sided. Unpaired, two-tailed, Student’s t test or one-way ANOVA followed by post hoc Bonferroni, Dunnett’s, or Tukey’s multiple comparisons analysis were used to analyze data. Wilcoxon matched-pairs signed rank t test was used when the same parameter was assessed in the same animal in any given experiment (chemo/radiation therapy when the same animal was imaged with a given radioligand before and after treatment). Data were considered significant if p values were below 0.05 (95% confidence intervals).