A Novel Radiotracer to Image Glycogen Metabolism in Tumors by Positron Emission Tomography

The majority of tumors exhibit substantially increased levels of glucose uptake and utilization in comparison with nonneoplastic tissues. These tumors use rapid glycolysis and lactate secretion even under normoxic conditions, a phenomenon first reported by Warburg in the 1920s (reviewed in ref. 1). This transition—the Warburg effect (2)—provides a selective advantage for tumor growth, progression, and protection from death (1, 3–5).

In addition to increased glycolytic flux, tumor cells originating from epithelial tissues also accumulate glycogen (6–8). Initially thought to be limited to specific histotypes such as clear cell cancers, recent data suggest that this is a more generalized phenomenon, with potential prognostic and therapeutic implications (9, 10). Glycogen is a branched glucose polymer, and the principal glucose store in mammalian cells. Glycogen is synthesized from the activated glucose donor, uridine diphosphoglucose (UDP), catalyzed by glycogen synthase, the key regulatory enzyme in glycogenesis. Glycogen synthase (GYS) has two isoforms: GYS1 (muscle) and GYS2 (liver; ref. 11), controlled both by allosteric effectors (G-6-P, ATP, and AMP) and by inhibitory phosphorylation of glycogen synthase at multiple sites (12).

Distinct from our perception of the cancer cell as a cell deficient in energy stores requiring glucose uptake and glycolysis to meet energy demands, emerging data suggest that oncogenic signaling and the quiescent state can induce glycogen storage in cancer cells (glycogenesis), and buffer bioenergetic stress. In ovarian cancer, glycogen stores are particularly high in chemoresistant clear cell adenocarcinomas, and the oncogenic small GTPase, Rab25, has recently been shown to increase glycogen stores, resulting in increased resistance to nutrient deprivation–induced bioenergetic stress, therapy-induced apoptosis, and poor prognosis (9). In addition, when cancer cells enter the stationary growth phase (quiescence), like budding yeast (13, 14), they synthesize glycogen (or trehalose), with the switch to growth being associated with diminished glycogenesis (6, 7, 15). Although the molecular mechanisms responsible for enhanced glycogenesis in mammalian cells are not well understood, it can be appreciated that noninvasive assessment of this metabolic switch could have important applications in cancer. Cytotoxic drugs used in cancer chemotherapy generally cause DNA damage, inhibit DNA replication, and/or cell division and are, therefore, active predominantly in proliferative cycling cells. Quiescent cells in tumors, such as dormant cells following chemotherapy and/or stem cells, therefore, exhibit relative chemoresistance.

It is against this background that we sought to develop a novel noninvasive measure of tissue glycogen for clinical use. Previous methods used to detect glycogen, such as the incorporation of tritiated glucose (16), immunocytochemistry (17), or glucose quantification following glycogen hydrolysis (18), are invasive techniques requiring biopsies. These methods also raise concerns about sampling a small region, and the ability to generalize to a heterogeneous tumor, that are well documented (19). More recently, 2-[N-(7-nitrobenz-2-oxa-1,3-diazol-4-yl) amino]-2-deoxy-d-glucose (2-NBDG), a fluorescent glucosamine derivative developed to measure cell viability and glucose transport, has been used to quantify glycogen storage in rat hepatocytes in culture (20). Fluorescent imaging of glycogen synthesis by 2-NBDG is ideal for in vitro measurements in intact cells, but poor tissue penetration of the emitted green fluorescence precludes its use in vivo. The substitution of fluoride in ¹⁸F-FDG for the C-2 hydroxyl moiety in glucose prevents metabolism past the second step of glycolysis catalyzed by glucose-6-phosphate isomerase; and ¹⁸F-FDG is not incorporated into glycogen. In contrast, we present here a noninvasive method for the in vivo detection of glycogen storage by positron emission tomography (PET) exploiting a novel glucosamine derivative, ¹⁸F-N-(methyl(2-fluoroethyl)-1H-[1,2,3]triazole-4-yl)glucosamine (¹⁸F-NFTG). ¹⁸F-NFTG retains the α-proton required for the initial flux into newly formed glycogen, analogous to 2-NBDG (Fig. 1A).

IGROV-1 ovarian adenocarcinoma cells were a kind donation from the Institute of Cancer Research (Sutton, Surrey, UK; 2006). Generation of HEY Rab25, HEY pcDNA, and IOSE80ht pcDNA cells have been previously described (21) and were obtained from the MD Anderson Cancer Center (Houston, TX; 2010). All cells were maintained in RPMI-1640 medium, supplemented with 10% fetal calf serum, 2 mmol/L l-glutamine, 100 U/mL penicillin, and 100 μg/mL streptomycin (Invitrogen) and were maintained at 37°C in a humidified atmosphere containing 5% CO₂. About cell line authentication, cells were analyzed by short tandem repeat profiling (DNA Diagnostics Centre) using the Promega PowerPlex Fusion System just before publication.

The effect of HEY Rab25 cell density on 2-NBDG fluorescence and glycogen immunostaining was performed in 8-well Lab-Tek chamber slides (Thermo Fisher Scientific Inc.), seeded with 1.25 × 10⁴ cells/cm². Measurements were made at 1, 3, and 6 days after seeding. For all other density-dependent measurements, cells were seeded at 12.8 × 10⁴, 6.4 × 10⁴, 3.2 × 10⁴, and 1.6 × 10⁴ cells/cm², with analysis performed 24 hour after seeding.

Of note, 2-NBDG fluorescence microscopy was performed as previously described (20). Briefly, cells were incubated with 500 μmol/L 2-NBDG at 37°C for 3 hours in complete growth medium in the dark. Cells were washed three times with warm PBS, followed by fresh warm media. Fluorescence measurements and images were obtained with at ×400 magnification on a Zeiss Axiovert S100 inverted microscope (Carl Zeiss Ltd.). For washout experiments, cells were maintained at 37°C and 5% CO₂ within the microscope imaging chamber, with images taken every 20 minutes.

For fluorescence quantitation, HEY Rab25 cells were seeded in black 96-well tissue culture plates (Thermo Fisher Scientific Inc.), at cell densities described above. Approximately 24 hours after seeding, fresh media containing 500 μmol/L 2-NBDG was added to cells. Following 3 hours incubation, cells were washed three times with ice-cold PBS and lysed in 50 μL radioimmunoprecipitation assay (RIPA) buffer (Thermo Fisher Scientific Inc.). Fluorescence was subsequently quantified on a PHERAstar^plus fluorescence plate reader (BMG LABTECH GmbH; λEx/Em = 65/540 nm) and normalized for protein content using a bicinchoninic acid (BCA) 96-well plate assay (Thermo Fisher Scientific Inc.).

Total cellular glycogen was evaluated in cells, seeded in 10-cm plates for 24 hours. Cells were subsequently evaluated using a colorimetric Glycogen Assay Kit (Biovision), according to the manufacturer's instructions. Samples were transferred to 96-well plates, with colorimetric measurements read in a Tecan Sunrise plate reader at 570 nm. Total cellular glycogen was normalized to protein levels using a BCA assay.

Of note, ¹⁸F-NFTG synthesis and radiolabeling was performed according to previously described methodology (22).

Cells were plated into 6-well plates 24 hours before analysis. For noncell density–dependent measurements, cells were seeded at 2.4 × 10⁵ cells per well of a 6-well plate. A total of 1.0 × 10⁶ cells per well were seeded for GYS1 short-hairpin RNA (shRNA) experiments. On the day of the experiment, fresh growth medium, containing approximately 0.74MBq ¹⁸F-NFTG or ¹⁸F-FDG, was added to individual wells (1 mL). Cell uptake was measured following incubation at 37°C in a humidified atmosphere of 5% CO₂ for 3 hours. The plates were subsequently placed on ice, washed three times with ice-cold PBS and lysed in RIPA buffer (Thermo Fisher Scientific Inc.; 1 mL, 10 minutes). Cell lysates were transferred into counting tubes and decay-corrected radioactivity was determined on a gamma counter (Cobra II Auto-Gamma counter; Packard BioScience Co.). Aliquots were snap-frozen and used for protein determination following radioactive decay by a BCA assay. Data were expressed as percentage of total radioactivity incorporated into cells, normalized for total cellular protein.

For the assessment of direct labeling of pro- and macroglycogen with ¹⁸F-NFTG, 5.9 × 10⁶ HEY Rab25 cells were seeded in triplicate into 10-cm plates. Twenty-four hours after seeding, approximately 3.7MBq ¹⁸F-NFTG were added to wells and incubated for 3 hours at 37°C in a humidified atmosphere containing 5% CO₂. Cells were washed three times with ice-cold PBS and scraped into 1 mL of RIPA buffer, 10% trichloroacetic acid or 20% potassium hydroxide, 65% ethanol for analysis of whole-cell lysates, proglycogen, and macroglycogen, respectively. Cells were lysed following 10× passages in an Ultra-Turrax T-25 homogeniser (IKA Werke GmbH and Co. KG). Aliquots were taken for protein analysis, with glycogen isolation performed as in Huang and colleagues (23). Samples were transferred into counting tubes and decay-corrected radioactivity was determined on a gamma counter. Total glycogen was assessed in samples by the glycogen colorimetric assay and glycogen-associated radioactivity was expressed as a percentage of radioactivity in whole-cell lysates, normalized for glycogen content.

Lentiviral plasmid containing a shRNA targeting human GYS1 mRNA was cloned in pLKO1 lentiviral vector (Sigma) using the following sequence: 5′-CCGGCAAGGGCTGCAAGGTGTATTTCTCGAGAAATACACCTTGCAGCCCTTGTTTTTG-3′ (TRCN0000045696). Recombinant lentiviruses were generated using a three-plasmid system in 293T cells as previously described (24). Viral titers were determined on 293T cells and target cells (HEY Rab25) transduced twice at a multiplicity of infection 10. Transduced cells were continuously selected in 1 μg/mL puromycin (Sigma). Knockdown of GYS1 activity was measured by Western blotting.

DNA-based flow cytometric analysis of cell-cycle progression was performed using propidium iodide in fixed, permeabilized cells according to the standard procedures (25). Cells were analyzed in an FACS Canto Cytometer (BD Biosciences), with 10,000 cells counted per treatment.

Rabbit antibody to Rab25, p27 Kip1, glycogen synthase, cdk-4, pRb, phospho-pRB (Ser807/811), and mouse anti-cyclin D1 and cdk-4 (1:1,000 dilution; Cell Signaling Technology) were used in a standard Western blotting protocol. A rabbit antibody to actin (Sigma-Aldrich Co.; 1:5,000) was used as a loading control.

To measure metabolic rates, 7.5 × 10⁵ cells were plated into 6-cm dishes and cultured until 90% confluent. At time 0, the cells were rinsed in PBS, fed with 2 mL of the test medium, and cultured. Endpoint experiments proceeded for 6 hours, then the medium was collected and analyzed for metabolite abundance. Of note, 0.6 mL tissue culture medium was processed by an automated chemical analyzer to determine glucose, lactate, glutamine, and glutamate concentrations (BioProfile Basic4; NOVA Biomedical). Differences in metabolite concentration between time-0 and end medium were used to calculate rates of nutrient utilization and metabolite excretion, normalized to the cellular protein content of the dish (26).

All animal experiments were performed by licensed investigators in accordance with the United Kingdom Home Office Guidance on the Operation of the Animal (Scientific Procedures) Act 1986 and within published guidelines for the welfare and use of animals in cancer research (27). Female BALB/c nude mice (ages 6–8 weeks; Charles River Laboratories) were used. HEY Rab25 or IGROV-1 tumor cells (2 × 10⁶) were injected subcutaneously on the back of mice and animals were used when the xenografts reached approximately 100 mm³. Tumor dimensions were measured continuously using a caliper and tumor volumes were calculated by the equation: |${\rm volume}\, = \,(\pi/6) \times a \times b \times c$|⁠, where a, b, and c represent three orthogonal axes of the tumor.

Dynamic ¹⁸F-NFTG imaging scans were carried out on a dedicated small animal PET scanner (Siemens Inveon PET module; Siemens Medical Solutions USA, Inc.) following a bolus intravenous injection of approximately 3.7MBq of the radiotracer into tumor-bearing mice (28). Dynamic scans were acquired in list mode format over 60 minutes. The acquired data were then sorted into 0.5-mm sinogram bins and 19 time frames for image reconstruction (4 × 15, 4 × 60, and 11 × 300 s), which was done by filtered back projection. The Siemens Inveon Research Workplace software was used for visualization of radiotracer uptake in the tumor; 30 to 60 minutes cumulative images of the dynamic data were used to define three-dimensional (3D) regions of interest (ROI). Radiotracer blood retention was estimated as in ref. 28. Care was taken to minimize ROI overlap with the myocardium. The count densities were averaged for all ROIs at each time point to obtain a time versus radioactivity curve (TAC). Tumor TACs were normalized to injected dose, and expressed as percentage injected dose per mL tissue. The area under the TAC, calculated as the integral of %ID/mL from 0 to 60 minutes, and the normalized uptake of radiotracer at 60 minutes (%ID/mL₆₀) were also used for comparisons. For image visualization, iterative reconstruction was performed [three dimensional ordered subset expectation maximization (3D-OSEM)].

Data were expressed as mean ± SD and statistical significance determined using the Student two-tailed t test. The ANOVA was used for multiple comparisons (Prism v5.0 software for windows; GraphPad Software). Differences between groups were considered significant if P ≤ 0.05.

Addition of 2-NBDG to isogenic HEY Rab25 and HEY pcDNA cells resulted in rapid cellular uptake and retention. The overexpression of Rab25, in HEY Rab25 cells was confirmed by Western blotting (Fig. 1B). Washout of diffuse 2-NBDG–derived background fluorescence revealed punctate granular cytosolic fluorescence in both HEY Rab25 and HEY pcDNA cells. The number and intensity of punctate 2-NBDG fluorescent foci were qualitatively higher in Rab25-overexpressing HEY cells compared with empty vector HEY pcDNA cells, and fluorescence remained up to 4 hours after removal of 2-NBDG (Fig. 1C). In nontumorigenic immortalized human ovarian epithelial IOSE80ht cells, no punctate staining was observed in IOSE80ht after washout, with intracellular fluorescence lost by 2 hours (Fig. 1C).

We synthesized ¹⁸F-NFTG by a copper-catalyzed 1,3-dipolar cycloaddition reaction (22). For in vivo PET studies injection-ready ¹⁸F-NFTG was obtained from ¹⁸F-fluoride with a radiochemical yield of 15% ± 5% (nondecay-corrected; ref. 22). The radiochemical purity was >98% from n = 12 individual syntheses.

Temporal uptake of ¹⁸F-NFTG was assessed in HEY Rab25 and HEY pcDNA cells under the same conditions as 2-NBDG. Significantly higher ¹⁸F-NFTG uptake was detected in HEY Rab25 (2.80% ± 0.26% radioactivity/mg protein) compared with HEY pcDNA cells at 3 hours (2.18% ± 0.11% radioactivity/mg protein; P = 0.019; Fig. 1D). A 30% higher ¹⁸F-NFTG uptake was observed in HEY Rab25 in comparison with HEY pcDNA cells at 4 hours after radiotracer addition (3.17% ± 0.25% and 2.45% ± 0.17% radioactivity/mg protein, respectively; P = 0.014). In HEY Rab25 cells, total cellular glycogen was 51% higher than pcDNA control cells (Fig. 1E), with 2.95 ± 0.53 μg/mg protein and 1.95 ± 0.15 μg/mg protein measured, respectively, consistent with differences in glycogen stores measured previously (9).

Direct labeling of glycogen with ¹⁸F-NFTG was shown following glycogen extraction, 3 hours after HEY Rab25 cells were exposed to ¹⁸F-NFTG. In total, 69.2% ± 2.3% intracellular radioactivity was associated with proglycogen, whereas macroglycogen labeling was below the limit of detection (Fig. 1F).

Although the metabolic fate of glucose and glucosamine has been well documented, the effect of their derivatization on metabolic flux is not known. The potential metabolic pathways in which ¹⁸F-NFTG is substrate are illustrated in Fig. 2A, compared with the known fate of ¹⁸F-FDG. To explore the dependence of 2-NBDG and ¹⁸F-NFTG cell retention on glycogenesis, HEY Rab25 cells were transfected with GYS1 shRNA or control lentiviral vector. Greater than 90% knockdown of GYS1 was achieved in the GYS1 shRNA cells in comparison with controls (Fig. 2B). Three further GYS1 shRNA sequences produced equivalent target knockdown (Supplementary Fig. S1). GYS1 depletion resulted in a corresponding 5.4-fold reduction in total extractable glycogen (Fig. 2C), from 10.4 ± 0.8 μg/mg protein in empty vector control cells to 1.9 ± 0.1 μg/mg protein in GYS1 shRNA cells (P = 0.00004; n = 3). To understand the effect of GYS1 knockdown and the resulting glycogen depletion on metabolic pathway flux, cells were grown to confluency in high glucose-containing media (11.11 mmol/L), with media subsequently analyzed for glucose and glutamine consumption and the secretion of catabolic by-products (lactate and glutamate). Under these nutrient-rich conditions there was no significant difference between empty vector control and GYS1 knockdown cells in either glucose or glutamine utilization (Fig. 2D).

2-NBDG fluorescence microscopy qualitatively revealed a substantial decrease in staining in GYS1 knockdown cells versus empty vector controls (Fig. 2E), consistent with glycogen levels in the two cells, and despite minimal differences in glycolytic flux between the two lines. In contrast, differences in ¹⁸F-FDG cellular retention were unremarkable between GYS1 shRNA and empty vector HEY Rab25 cells, in keeping with the minimal differences in initial glycolytic flux between the two cell lines. ¹⁸F-FDG uptake was reduced by just 9% in GYS1 shRNA cells in comparison with control cells, with 10.2% ± 0.2% and 11.2% ± 0.6% radioactivity/mg protein measured in cells, respectively (P = 0.045; Fig. 2F). Conversely, knockdown of GYS1 expression resulted in a corresponding 7.3-fold reduction in ¹⁸F-NFTG uptake when compared with empty vector control cells (7.69% ± 0.68% and 1.05% ± 0.05% radioactivity/mg protein, respectively; P = 0.00007; Fig. 2G), supporting the utility of substituted glucosamine derivatives, 2-NBDG and ¹⁸F-NFTG, to report glycogen synthase–mediated glycogenesis.

High 2-NBDG punctate fluorescence was observed in HEY Rab25 cells when confluency was achieved in culture, 6 days after seeding (Supplementary Fig. S2). Similarly, antiglycogen immunofluorescence staining of fixed cells with an antiglycogen IgM (20, 29) also demonstrated granular cytosolic fluorescence, albeit with some nonspecific background staining (Supplementary Fig. S2). Overlay of 2-NBDG fluorescence and glycogen immunofluorescence images was not possible as 2-NBDG fluorescence was lost during fixation and permeabilization of cells. We noted in washout experiments a reduction in punctate 2-NBDG staining 40 minutes before cell division in comparison with nondividing cells; this effect also corresponded to an increase in diffuse background 2-NBDG fluorescence (Supplementary Fig. S3).

The role of growth arrest on glycogen accumulation was further assessed by varying cell seeding density (confluency) from 1.6 × 10⁴ to 12.8 × 10⁴ cells/cm², 24 hours before analysis. In the asynchronous cultures, increased seeding density resulted in a proportional increase in G₁–G₀ growth arrest, characterized by flow cytometric measurements of DNA content (Fig. 3A) and a decrease in phospho-pRb levels (Fig. 3B). Furthermore, the quiescence marker, p27/Kip1, and β-galactosidase staining for senescence, were both increased at the highest cell seeding density of 12.8 × 10⁴ cells/cm² (Fig. 3B and Supplementary Fig. S4, respectively), whereas total cyclin D1 levels remained unchanged (Fig. 3B). A linear relationship between cell seeding density and total extractable glycogen was observed in HEY Rab25 cells, increasing from 2.95 ± 0.53 μg/mg protein with the lowest seeding density (1.6 × 10⁴ cells/cm²) to 6.29 ± 0.75 μg/mg protein at the highest (12.8 × 10⁴ cells/cm²), a 2.1-fold increase (P = 0.0019; n = 3; Fig. 3C). A density-dependent increase in 2-NBDG fluorescence was also observed, rising 2.8-fold, from 229 ± 63 relative fluorescence units (RFU)/μg protein to 629 ± 188/μg protein for cells seeded at 1.6 × 10⁴ and 12.8 × 10⁴ cells/cm², respectively (Fig. 3D).

Having confirmed an association between confluency (quiescence)-induced glycogen storage and 2-NBDG fluorescence, in vitro uptake of ¹⁸F-NFTG was performed in HEY Rab25 cells. An increase in cell density increased tracer uptake and retention. Cells seeded at 1.6 × 10⁴ cells/cm² retained 3.51% ± 0.18% radioactivity/mg protein, increasing to 6.63% ± 0.25% radioactivity/mg protein at a density of 6.4 × 10⁴ cells/cm², a 1.9-fold increase (P = 0.0006; n = 3). The increase in ¹⁸F-NFTG uptake started to plateau at the highest density of 12.8 × 10⁴ cells/cm² − 7.69% ± 0.68% radioactivity/mg protein (Fig. 3E). Similar experiments were conducted with ¹⁸F-FDG, to assess corresponding levels of glucose uptake. Although overall cellular uptake of ¹⁸F-FDG was higher in HEY Rab25 cells in comparison with ¹⁸F-NFTG (12.08% ± 0.58% versus 7.69% ± 0.68% radioactivity/mg protein, respectively, at 12.8 × 10⁴ cells/cm²; Fig. 3F), seeding density did not alter ¹⁸F-FDG uptake. When compared, there was good correlation between both 2-NBDG fluorescence and ¹⁸F-NFTG uptake with cellular glycogen (R² = 0.9336 and R² = 0.8998, respectively), with no correlation between ¹⁸F-FDG and glycogen (R² = 0.0016; Fig. 3G).

To confirm cell-cycle–dependent changes in glycogen labeling as a prelude to in vivo studies, 2-NBDG confocal microscopy was performed on HEY Rab25 multicellular tumor spheroids, grown in 3D culture. Tumor cells in the core of spheroids are usually in G₀, with cycling cells closer to the spheroid surface (30). Approximately 20-μm Z-stacks of the spheroid (Supplementary Fig. S5) revealed heterogeneous uptake, with maximum fluorescence visualized in the spheroid core, and a further region of high uptake at the rim of the spheroid.

Of note, ¹⁸F-NFTG PET imaging was performed in size-matched HEY Rab25 xenograft-bearing nude mice and compared with ¹⁸F-FDG–PET (Fig. 4). ¹⁸F-NFTG radiotracer distribution was characterized by rapid liver uptake, followed by clearance through the kidney and bladder. Clearance through the intestines was also observed (Fig. 4A). Although ¹⁸F-NFTG did not accumulate in the brain, heart-specific radioactivity was still present at 60 minutes, accompanied by good tumor uptake and retention (Fig. 4B). In comparison, the typical ¹⁸F-FDG distribution of high brain, heart, bladder, and tumor radioactivity was observed (Fig. 4C and D).

The kinetic profiles of tumor retention also varied considerably between the two tracers, as shown by the TACs (Fig. 4E). Rapid tumor uptake of ¹⁸F-NFTG, peaking at 5 minutes, preceded a slow washout of radioactivity over the remaining 55 minutes. This was contrasted with ¹⁸F-FDG, with a maximal tumor radioactivity achieved 15 minutes after injection, with subsequent stabilization of tumor-associated radioactivity (Fig. 4E). Radio high-performance liquid chromatography analysis of plasma ¹⁸F-NFTG radioactivity revealed good tracer stability, with 100% of the parent compound still present 15 minutes after injection (n = 3 mice; Supplementary Fig. S6). Radiotracer uptake in the tumor was 2.9-fold higher with ¹⁸F-FDG than ¹⁸F-NFTG 60 minutes after injection (4.10 ± 0.38 and 1.40 ± 0.29 %ID/mL, respectively; P = 0.009).

¹⁸F-NFTG pharmacokinetics in other major organs varied substantially compared with ¹⁸F-FDG. High heart-associated radioactivity, typical of ¹⁸F-FDG, was not present with ¹⁸F-NFTG, with 42.5 ± 5.9 %ID/mL measured with ¹⁸F-FDG in comparison with just 4.2 ± 0.3 %ID/mL with ¹⁸F-NFTG at 60 minutes after injection (Fig. 4F). Liver ¹⁸F-NFTG radioactivity peaked at 5 minutes, followed by slow clearance, whereas rapid washout from the liver was observed with ¹⁸F-FDG, with maximal radioactivity measured <1 minute after injection (Fig. 4G). A similar profile of rapid extraction from blood and urinary elimination through the kidneys was evident with both radiotracers (Supplementary Fig. S7 and Fig. 4H, respectively).

A common confound of many oncologic PET radiotracers, including ¹⁸F-FDG, is their inability to differentiate between neoplastic and inflammatory tissue, leading to reduced specificity (31). Here, we assessed uptake of both ¹⁸F-FDG and ¹⁸F-NFTG in an aseptic inflammation (turpentine) model (Supplementary Fig. S8). Turpentine-induced inflammation, confirmed in histologic sections stained with hematoxylin and eosin, resulted in a 2.7-fold increase (P = 0.003) in ¹⁸F-FDG uptake in treated verses untreated muscle, clearly visible in the 50 to 60 minute static PET image. Conversely, no significant change in turpentine-treated muscle ¹⁸F-NFTG uptake was detectable by either PET or ex vivo biodistribution when compared with the untreated control contralateral muscle (P = 0.10).

A panel of eight human tumor cell lines were assessed for glycogen content (Supplementary Fig. S9). High-total cellular glycogen (44.2 ± 14.8 μg/mg protein) was measured in stationary-phase IGROV-1 ovarian carcinoma cells, 4.8-fold higher than in HEY Rab25 cells (9.30 ± 1.96 μg/mg protein; P = 0.01; n = 3; Fig. 5A) and, thus, were selected for further evaluation. This difference in extractable glycogen correlated well with in vitro uptake of ¹⁸F-NFTG, which was 12.26% ± 1.97% and 2.79% ± 0.26% radioactivity/mg protein in IGROV-1 and HEY Rab25 cells, respectively (4.4-fold increase; P = 0.001; n = 3; Fig. 5B). Similarly, ¹⁸F-NFTG uptake in implanted IGROV-1 xenografts in vivo was 1.6-fold higher than in HEY Rab25 tumors (area under the curve of 122.7 ± 12.3 and 73.7 ± 2.42 %ID/mL/min, respectively; P = 0.028), consistent with the higher glycogen levels measured in vitro. In contrast with HEY Rab25 xenografts, tumor-associated radioactivity in the higher glycogen content IGROV-1 xenograft was stable between 15 and 60 minutes (Fig. 5C and D).

For more than 30 years, aberrant glycogen metabolism has been known to be associated with the metabolic reprogramming that occurs in many cancers (7). Although initially thought to be confined to specific histotypes, recent data have shown that this is a general phenomenon. Little is known about the extent of glycogen storage, the mechanisms governing this phenotypic adaptation and role in resistance and poor prognosis. In this study, we unequivocally demonstrate that measurements of ¹⁸F-NFTG uptake and retention can be used to detect de novo glycogenesis in tumor cells growing in culture or in tumor xenografts in vivo, and that the ¹⁸F-NFTG readout of glycogenesis could be used to assess the proliferative state of cancer cells.

Recently, the phosphoinositide 3-kinase (PI3K) signaling pathway has been suggested as a key regulator of glycogenesis, linked to increased expression of the small GTPase, Rab25 (9). In Rab25-overexpressing cells, AKT activity was increased, leading to an augmentation of GSK3 phosphorylation and subsequent inactivation. In these cells GSK3 was unable to phosphorylate and therefore inhibit glycogen synthase activity, resulting in increased glycogen accumulation. Furthermore, it was shown that inhibition of PI3K abrogated Rab25-mediated glycogen storage. Of note, 2-NBDG and ¹⁸F-NFTG were able to discriminate between small differences in glycogen synthesis in HEY cells induced by Rab25 transfection, and suggest potential utility as a measure of oncogenic signaling–induced changes in glycogen. A measure of oncogene-driven glycogen status may be of prognostic significance, given that Rab25 seems to confer a functional survival advantage to cancer cells; allowing cellular survival under acute nutrient stress conditions (9).

In cell experiments, specificity of glycogen labeling with 2-NBDG was demonstrated in comparison with antiglycogen immunofluorescence and through extraction of ¹⁸F-NFTG–labeled proglycogen. Glycogen labeling by these substituted glucosamine derivatives was proposed via the putative signaling pathway illustrated in Fig. 2A, controlled by the key regulator of glycogenesis, glycogen synthase. To further elucidate the specificity of these tracers to report glycogenesis, specific knockdown of muscle GYS1 by shRNA was used. GYS1 knockdown depleted intracellular glycogen stores, reduced ¹⁸F-NFTG uptake more than 7-fold when compared with empty vector control cells and similarly reduced 2-NBDG retention. Furthermore, GYS1 silencing had no effect on glycolytic flux, ruling out nonspecific metabolic effects of GYS1 silencing or off-target toxicities of the shRNA. As expected, ¹⁸F-FDG was unable to annotate the glycogen status of these cells. It is important to note that although complementary, measurements of the total intracellular glycogen pool (measured biochemically) and rate of glycogen synthesis (measured by ¹⁸F-NFTG) differ: the total glycogen pool is a sum of both glycogenesis and glycogenolysis, whereas ¹⁸F-NFTG has only been shown to measure glycogen synthase–mediated glycogenesis. Future studies are required to elucidate whether ¹⁸F-NFTG–associated glycogen is substrate for glycogen breakdown by glycogen phosphorylase.

We examined whether modulation of glycogen storage during the nonproliferative state of cancer cells could be detected by 2-NBDG and ¹⁸F-NFTG. Induction of G₀–G₁ growth arrest in vitro by increased cell density resulted in an increase in glycogen storage, proportional to the initial cell seeding concentration. In cells, G₀–G₁ arrest led to increased 2-NBDG and ¹⁸F-NFTG uptake, whereas ¹⁸F-FDG retention remained unchanged. The mechanism behind this increased glycogen storage remains to be fully determined. We speculate that the high ¹⁸F-FDG/glucose uptake attributed to the Warburg effect aims to provide the building blocks for cellular proliferation through, for example, the pentose phosphate shunt, whereas in quiescent tumor cells the excess glucose is instead diverted to glycogen. This could then be potentially mobilized when the cell reenters the cell cycle for both energy and as a carbon source for the synthesis of proteins, nucleotides, and other building blocks, analogous to nutrient stress–induced quiescence, and synthesis of glycogen and trehalose by budding yeast (13, 14).

To provide further evidence for uptake of the substituted glucosamines in the quiescent state, we used a more sophisticated model of heterogeneous tumor cell growth, multicellular tumor spheroids (reviewed in ref. 30). In this model, we revealed elevated 2-NBDG uptake in the hypoxic spheroidal core, in which cells are relatively quiescent (32) and express elevated GYS1 (10). The high uptake of 2-NBDG in the rim of the spheroid may reflect improved nutrient, and therefore 2-NBDG, access in this region. Although our focus has been on glycogenesis, it should be noted that other mechanisms could be responsible for the high glycogen levels in quiescent cells. For instance, it was shown recently that inhibition of glycogen phosphorylase activity and glycogenolysis in cancer leads to a p53-dependent induction of senescence, providing a mechanistic link between cellular glycogen and senescence (10). The ability to noninvasively measure tumor proliferative state, be it cellular quiescence or senescence may be of great importance.

Given the ability of ¹⁸F-NFTG–PET to measure cellular glycogen levels as a metabolic biomarker of oncogene activation (e.g., PI3K/AKT pathway by Rab25), and G₁–G₀ quiescence, we envisage that ¹⁸F-NFTG–PET could potentially be used to complement ¹⁸F-FDG–PET imaging of tumor glycolysis. Dynamic imaging revealed distinct patterns of ¹⁸F-NFTG tissue retention in comparison with ¹⁸F-FDG, characterized by high uptake in the kidneys and in glycogen-storing liver. As the liver is a frequent site of metastasis, this could limit the utility ¹⁸F-NFTG in these patients. However, some tumors are not ¹⁸F-FDG avid due to low metabolic rates. These are frequently slow growing tumors. On the basis of the relationship between cell cycle and glycogen levels, ¹⁸F-NFTG could complement ¹⁸F-FDG in imaging of such tumors. In oncology ¹⁸F-FDG–PET is also used as a marker of treatment response. ¹⁸F-FDG can overestimate the efficacy of cytotoxic chemotherapy due to a stunning effect, resulting from a reduction in metabolism due to cellular “stress,” or quiescence of residual cells. The potential ability of ¹⁸F-NFTG to specifically image these quiescent cells may be an advantage. Furthermore, ¹⁸F-NFTG provides improved specificity over ¹⁸F-FDG for tumor imaging under conditions of inflammation. We envisage ¹⁸F-NFTG to have great applicability to image tumors of the upper thoracic, for example, those of the breast, in which background is minimal. Utility in ovarian cancer, as depicted in this study, is also envisaged.

We believe that this technique can be successfully transferred to the clinic, with potential application in both malignant and nonmalignant diseases; in malignant disease, the tracer may have a role in the assessment of activation of oncogenic pathways related to glycogenesis, their modulation by drugs in development, and the detection of posttreatment quiescent residual tumor burden.

No potential conflicts of interest were disclosed.

Conception and design: T.H. Witney, L. Carroll, Q.-D. Nguyen, R. Agarwal, E.O. Aboagye

Development of methodology: T.H. Witney, L. Carroll, A. Chandrashekran, Q.-D. Nguyen, E.O. Aboagye

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T.H. Witney, L. Carroll, I.S. Alam, Q.-D. Nguyen, R. Sala, R. Harris, R. Agarwal, E.O. Aboagye

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T.H. Witney, R. Harris, R.J. DeBerardinis, R. Agarwal, E.O. Aboagye

Writing, review, and/or revision of the manuscript: T.H. Witney, L. Carroll, I.S. Alam, R. Sala, R.J. DeBerardinis, R. Agarwal, E.O. Aboagye

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T.H. Witney, I.S. Alam, E.O. Aboagye

Study supervision: E.O. Aboagye

The authors thank W. Gsell and J. Tremoleda for their assistance with the PET imaging studies, and K. Cheng and G. Mills for the HEY Rab25/pcDNA and IOSE80ht cell lines.

This work was funded by Cancer Research UK, Engineering and Physical Sciences Research Council grant (in association with the Medical Research Council and Department of Health, England) grant C2536/A10337. E.O. Aboagye laboratory receives core funding from the UK Medical Research Council (MC_A652_5PY80). R. Agarwal is funded by a Cancer Research UK Clinician Scientist Fellowship. R.J. DeBerardinis is supported by the NIH (R01 CA157996-02) and Robert A. Welch Foundation (I-1733).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.