Tuberous sclerosis complex (TSC) is an autosomal dominant disorder caused by inactivating mutations of the TSC1 or TSC2 gene, characterized by neurocognitive impairment and benign tumors of the brain, skin, heart, and kidneys. Lymphangioleiomyomatosis (LAM) is a diffuse proliferation of α-smooth muscle actin–positive cells associated with cystic destruction of the lung. LAM occurs almost exclusively in women, as a TSC manifestation or a sporadic disorder (TSC1/TSC2 somatic mutations). Biomarkers of whole-body tumor burden/activity and response to rapalogs or other therapies remain needed in TSC/LAM.
These preclinical studies aimed to assess feasibility of [18F]fluorocholine (FCH) and [18F]fluoroacetate (FACE) as TSC/LAM metabolic imaging biomarkers.
We previously reported that TSC2-deficient cells enhance phosphatidylcholine synthesis via the Kennedy pathway. Here, we show that TSC2-deficient cells exhibit rapid uptake of [18F]FCH in vivo and can be visualized by PET imaging in preclinical models of TSC/LAM, including subcutaneous tumors and pulmonary nodules. Treatment with rapamycin (72 hours) suppressed [18F]FCH standardized uptake value (SUV) by >50% in tumors. Interestingly, [18F]FCH-PET imaging of TSC2-deficient xenografts in ovariectomized mice also showed a significant decrease in tumor SUV. Finally, we found rapamycin-insensitive uptake of FACE by TSC2-deficient cells in vitro and in vivo, reflecting its mitochondrial accumulation via inhibition of aconitase, a TCA cycle enzyme.
Preclinical models of TSC2 deficiency represent informative platforms to identify tracers of potential clinical interest. Our findings provide mechanistic evidence for testing the potential of [18F]FCH and [18F]FACE as metabolic imaging biomarkers for TSC and LAM proliferative lesions, and novel insights into the metabolic reprogramming of TSC tumors.
Tuberous sclerosis complex (TSC) is an autosomal dominant disease characterized by multi-organ hamartomas. Lymphangioleiomyomatosis (LAM) is a diffuse proliferation of TSC2-deficient cells associated with progressive cystic destruction of the lung. mTOR complex 1 (mTORC1) inhibition by rapamycin, the FDA-approved agent for TSC and LAM, exerts a cytostatic but not cytotoxic effect on tumor cell growth. Metabolic imaging biomarkers of disease activity are not currently available for TSC/LAM and would be critical to streamline the design of clinical trials and facilitate therapeutic decision-making, including personalization of rapamycin dosage. Our data provide the first in vivo evidence that TSC2-deficient cells have rapamycin-sensitive uptake of [18F]fluorocholine (FCH) and that TSC2-deficient tumors can be detected using [18F]fluoroacetate (FACE)-PET. Use of [18F]FACE or [18F]FCH as imaging biomarkers of mTOR-dependent aberrant metabolism may provide early, noninvasive measures of disease progression (pulmonary and extrapulmonary lesions) or therapeutic response in TSC, LAM, and potentially other mTOR-driven proliferative diseases.
Tuberous sclerosis complex (TSC) is an autosomal dominant disease characterized by neurologic manifestations and benign tumors of the brain, skin, heart, and kidneys (cysts and angiomyolipomas; ref. 1). Renal cell carcinoma occurs in 2% to 3% of the TSC population (1, 2). Up to 80% of women with TSC develop lymphangioleiomyomatosis (LAM; ref. 3). LAM is a diffuse proliferation of α-smooth muscle actin–positive "LAM cells" associated with cystic destruction of the lung. Severe LAM requires oxygen therapy and eventually lung transplantation. LAM can also occur as a sporadic disorder where LAM cells harbor inactivating somatic mutations of the TSC1 or TSC2 gene (1, 4–6). Extrapulmonary manifestations of sporadic LAM include abdominal tumors such as renal angiomyolipomas and lymphangiomyomas (5). TSC2 deficiency leads to activation of mammalian/mechanistic target of rapamycin complex 1 (mTORC1), a serine/threonine protein kinase complex that acts as a master regulator of cell-growth, proliferation, and metabolism (7–13).
Rapamycin is an FKBP12-dependent allosteric inhibitor of mTORC1 approved by the FDA for the treatment of LAM and TSC-associated renal angiomyolipomas. Clinical trials of TSC and LAM have shown that 12-month treatment with rapamycin induces response of renal angiomyolipomas and stabilization of pulmonary function (14–16). However, lung function decline and tumor growth resumes when treatment is discontinued. In addition, magnitude of clinical response to rapamycin varies across patients. For instance, in the EXIST-2 trial (double-blind, placebo-controlled, phase III trial), the response rate was 42%, with the primary efficacy endpoint being the proportion of patients with confirmed angiomyolipoma response of at least 50% reduction in the total volume of target angiomyolipomas relative to baseline (15).
Therapeutic regimens able to induce durable clinical response represent a critical unmet need in TSC and LAM, and several questions remain regarding personalized dosage and length of treatment for rapamycin and its analogs (rapalogs). This clinical evidence highlights the critical need for sensitive and specific biomarkers of response to therapy and/or disease progression, including biomarkers of tumor metabolic activity. Predictive biomarkers of disease progression and therapeutic response, including metabolic imaging biomarkers, would streamline the design, cost, and duration of TSC/LAM clinical trials to answer these outstanding clinical questions.
Currently, elevated levels of VEGF-D in the blood of patients with LAM represent the only biomarker with diagnostic role in LAM (17, 18) and potentially a role as a biomarker of therapeutic response to mTORC1 inhibitors (19). However, not all patients with LAM have clinically significant blood VEGF-D levels. A metabolic imaging biomarker and circulating VEGF-D could be used as complementary diagnostics to enhance clinical decision making.
[18F]fluorodeoxyglucose (FDG)-PET has shown utility to identify malignant neoplasms including malignant perivascular epitheliod cell tumors (PEComas), primary lung cancers, and lymphomas, in patients with TSC and/or LAM. However, TSC and LAM lesions (renal angiomyolipomas, pulmonary LAM, lymphangiomyomas) did not show abnormal FDG uptake, likely due to aberrant trafficking of glucose transporters (20–22).
We previously showed that TSC2-deficient cells have an aberrant complex lipid metabolism, including enhanced incorporation of choline into phosphatidylcholine, and increased levels of neutral lipids, which are in part dependent on the activity of sterol regulatory element binding protein (SREBP) downstream of mTORC1 (9). Building on this premise, here we tested the potential for [18F]fluorocholine (FCH) and [18F]fluoroacetate (FACE) PET to detect TSC2-deficient cells and to monitor response to rapamycin in vivo, in preclinical models of TSC and LAM. We show that TSC2-deficient tumors have rapid uptake of [18F]FCH and [18F]FACE. Interestingly, tumor incorporation of [18F]FCH but not [18F]FACE was significantly decreased by 72-hour treatment with rapamycin. Using precursors of the radiocompounds labeled with the stable isotope [19F], we traced the metabolic fate of FCH and FACE in TSC2-deficient cells by nuclear magnetic resonance (NMR) and [14C]acetate incorporation into lipids, validating the in vivo results. We confirmed that FCH is incorporated into lipids in a rapamycin-sensitive manner, whereas FACE was not found in lipids. Rapamycin-resistant accumulation of FACE in the cell impacted the mitochondrial function, leading to a decrease in the basal respiration and maximal respiratory capacity via inhibition of a tricarboxylic acid (TCA) cycle enzyme, aconitase. Of relevance, rapamycin did not suppress mitochondrial function in TSC2-deficient cells and induced an increase in mitochondrial acetate oxidation, suggesting that rapamycin-insensitive uptake of FACE by TSC tumors reflects a specific metabolic feature of TSC2 deficiency.
In summary, our findings provide mechanistic evidence for testing the potential of [18F]FCH and [18F]FACE as metabolic imaging biomarkers in TSC and LAM, and novel insights into the metabolic reprogramming of TSC tumors.
Materials and Methods
Cell lines and preclinical models of TSC
TSC2-deficient ELT3-V3 cells were derived from an Eker rat uterine leiomyoma (23, 24). 621-101 cells (gift from Dr. Elizabeth Henske, Brigham and Women's Hospital, Boston, MA)) were derived from a LAM patient renal angiomyolipoma (25) and carry the same somatic biallelic TSC2 gene-inactivating mutations as the patient's LAM cells (G1832A missense mutation of one allele, and loss of the other allele; ref. 4), which was confirmed in the cells used in these studies. All cell lines were grown in DMEM supplemented with 10% FBS, 100 units of penicillin, and 100 mg/mL of streptomycin. TSC2 deficiency, constitutive activation of mTORC1, and rapamycin activity were validated by immunoblotting of tuberin and phospho-S6 kinase.
Four in vivo models were generated: (i) subcutaneous tumors generated by injection of ELT3-V3 cells in intact mice; (ii) subcutaneous tumors generated by injection of ELT3-V3 cells into mice that were ovariectomized and implanted under the skin with either a 90-day release 17β-estradiol or placebo pellet; (iii) subcutaneous tumors generated by injection of LAM patient renal angiomyolipoma–derived 621–101 cells; and (iv) a pulmonary LAM model generated by injecting ELT3-V3 cells in the tail vein of mice.
For all models, female NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) mice (The Jackson Laboratory, JAX stock #005557) were used. For subcutaneous tumors, 2.5 × 106 cells were resuspended in 100 μL of PBS and injected with Matrigel (1:1) in the flank of mice (all except one mouse were injected in a single flank). For the pulmonary model, 1 × 106cells resuspended in 100 μL of PBS were injected.
Mice with subcutaneous tumors were imaged 6 to 8 weeks (ELT3-V3, both intact and ovariectomized) or 12 to 14 weeks (621–101) after cell injection. Mice with lung lesions were imaged 25 to 90 days after cell injection. The animal studies were conducted under protocols approved by the Institutional Animal Care and Use Committee (IACUC) at Massachusetts General Hospital or Brigham and Women's Hospital (Boston, MA).
Ovariectomy was performed at The Jackson Laboratory using aseptic and atraumatic surgical technique approved by the IACUC. The Jackson Laboratory is an Association for Assessment and Accreditation of Laboratory Animal Care International–accredited facility. Briefly, the mouse was placed in right ventral-lateral recumbency and a 4 to 7 mm skin incision was made parallel and ventral to the spine midway between the last rib and the iliac crest. An incision was made in the underlying left abdominal wall. Forceps were used to grasp and exteriorize the left ovarian fat pad. The ovary was excised using a crush and tear technique and the uterine horn was replaced in the abdominal cavity. The abdominal wall was sutured with absorbable suture and the skin incision was closed with a wound clip. Bupivacaine was applied to the edges of the skin incision prior to closure. The same procedure was used for removal of the right ovary. Warmed sterile saline was administered subcutaneously to aid recovery. Postoperatively, mice were examined daily by surgery technicians and additional analgesia was given as needed. Mice were recovered from the surgical procedure for 5 to 7 days and the skin closure material was removed prior to shipment.
In vivo drug treatments
Rapamycin (LC Laboratories) was dissolved in absolute ethanol (20 mg/mL), diluted in sterile vehicle (0.25% Tween 80/0.25% PEG 200 in distilled water), and administered intraperitoneally (3 mg/Kg/day; 100 μL maximum volume) using a 27 G needle. Ninety-day release placebo and 17β-estradiol pellets were purchased from Innovative Research of America and implanted under the skin of NSG ovariectomized mice.
Synthesis of [18F]FCH and [18F]FACE
[18F]FCH and [18F]FACE were prepared following reported methods (26–28) with minor modifications. A detailed description of radiosynthetic methods for [18F]FCH and [18F]FACE is provided in Supplementary Methods.
Mice were imaged on a microPET/CT scanner (Sedecal Argus). A flowchart of the PET/CT procedures and data analysis is included in Fig. 1.
Of 37 intact mice bearing subcutaneous ELT3 xenografts (model 1), 12 mice underwent a single [18F]FCH-PET scan, 5 mice underwent [18F]FCH-PET imaging before and after 72-hour treatment with rapamycin (4 doses, 1 dose/per day, the last dose 4–5 hours prior to imaging), 6 mice underwent [18F]FCH-PET imaging before and after 24-hour treatment with rapamycin (1 dose), 4 mice (n = 5 tumors; one mouse was injected with cells in both flanks) underwent [18F]FCH-PET imaging before and after 72-hour administration of vehicle (control), 5 mice underwent [18F]FACE-PET imaging before and after 72-hour treatment with rapamycin, and 5 mice underwent [18F]FACE-PET imaging before and after 72-hour administration of vehicle (control). Five mice of the FCH group were imaged with [18F]FCH-PET and [18F]FACE-PET on consecutive days at baseline (n = 5) and posttreatment (n = 3) to test specificity of the uptake; therefore, they underwent [18F]FACE-PET after 48-hour treatment and [18F]FCH-PET after 72-hour treatment with rapamycin (n = 2) or vehicle (n = 1; 2 tumors). Of 7 ovariectomized mice bearing subcutaneous ELT3 xenografts (model 2), 4 were implanted with placebo pellet and 3 with 17β-estradiol pellet. All of these mice underwent [18F]FCH-PET. Three mice underwent [18F]FCH-PET 25–90 days after intravenous injection of ELT3 cells (model 3) and their lungs were compared with those of 3 healthy mice. Four mice bearing 621–101 xenografts (model 4) underwent [18F]FCH-PET (n = 1) or [18F]FACE-PET (n = 3) to validate rapid uptake of the tracers in human TSC2-deficient cells (Supplementary Table S1).
For each scan, the mouse was anesthetized with isoflurane, weighed, and cannulated in the tail vein (30-gauge needle). A dynamic PET scan was acquired, with bolus tracer injection (∼250 μCi; flushed with 0.05 mL saline) through the catheter 30 seconds after starting acquisition and total duration of 40 minutes for [18F]FCH and 90 minutes for [18F]FACE. A CT scan (40 kV, 300 μA) was acquired to provide anatomic information and for attenuation correction purposes. PET data were reconstructed using a two-dimensional ordered subset expectation maximization (2D-OSEM) algorithm with 16 iterations and 4 subsets, into 33 frames (1 × 30, 9 × 10, 6 × 30, 8 × 60, 6 × 120, 3 × 300 s, for PET start 30 seconds before tracer injection; 8 × 15, 8 × 30, 8 × 60, 8 × 120, 3 × 300 s for PET start 30 seconds after tracer injection), with corrections for dead time, random coincidences, detector inhomogeneity, and radioactive decay. Final images had a spatial resolution of approximately 1.5 mm (29).
PET data analysis.
Reconstructed data were converted to NifTi format (MatLab, Nifti toolbox 20140122-2). PET data were converted to standardized uptake value (SUV) and average images at 30 to 40 minutes postinjection (p.i.) were derived. CT scans were rigidly aligned to the averaged PET image using Vinci (30) and resliced to match PET geometry. Volumes of interest (VOI) were manually defined onto the resliced CT image using ITK-SNAP (31). Zero to 20 seconds p.i. PET images were used to define blood VOI in the descending aorta and to exclude large blood vessels from tissue VOI. For technical reasons, a few mice received the radiotracer dose directly into the tail vein (without cannulation) at the beginning of the scan; therefore, for these scans, 0 to 20 seconds p.i. PET images were not available.
Whole-lung VOIs were defined semiautomatically using MatLab. To minimize overestimation due to partial volume effects from surrounding tissues (e.g., the liver), a VOI was defined within the center of the right lung.
The VOIs were applied to the dynamic PET scan to derive averaged SUV time–activity curves (TAC). In addition, for each scan, the SUVpeak was derived from a spherical VOI of 1.9 mm diameter centered on the maximum SUV voxel within the tissue VOI of the 30- to 40-minute PET image that was smoothed using a three-dimensional Gaussian filter with σ = 1.5 mm (SUVpeak; excluding tissue VOI edges to account for partial volume effects from surrounding tissues).
For quality-assurance purposes, SUV changes were confirmed using tumor VOI thresholded at 50% SUVpeak and for each scan the whole field-of-view (FOV)-TAC was derived. Figures were generated using MatLab. To facilitate comparison, PET windowing levels for each tracer were set to 25% to 75% of the maximum baseline tumor SUV for the study population displayed.
ELT3 cells were treated with rapamycin (20 nmol/L) for 24 hours and labeled with [19F]FCH (200 μmol/L; obtained from Advanced Biochemical Compounds) for 6 hours in serum-free media before lipid extraction (methanol:chloroform:water, 2:1:0.5 v/v/v). 19F NMR spectra analysis of FCH-labeled cell lipid extracts was carried out on a Bruker Avance-500 instrument equipped with a BBFO probe. Fluorine-19 on this instrument is at 470.51 MHz, and chemical shifts are referenced to CFCl3 at 0.00 ppm. One-dimensional spectra were obtained using a standard one-pulse sequence with a 1.3-second overall recycle delay and proton-decoupling; typically, a total of 4096–8192 scans were collected. Line broadening of 2 Hz was applied to a final 64k points. Quantitation was performed by adding a solution of 4.94 mmol/L FCH in deuterated methanol as an internal reference standard to a final concentration of 82 μmol/L. This internal standard gives a peak at −218.81 ppm, and the concentration of fluorine-containing metabolite was measured by integrating the metabolite peak relative to the internal standard. Concentration of 19F compounds in lipid extracts was normalized to total protein content.
De novo lipid synthesis
ELT3 cells were labeled with [1-14C]acetic acid or [1-14C]FACE (0.5 μCi/mL) obtained from American Radiolabeled Chemicals for 6 hours, then washed three times and collected for lipid extraction using methanol/chloroform/H2O (2:1:0.5 v/v/v). Radioactivity was counted on a Packard Tri-Carb Liquid Scintillation Analyzer.
[1-14C]FACE cellular incorporation
ELT3 cells were seeded in 12-well plates (50,000 cells/well) in DMEM supplemented with 10% FBS, treated with rapamycin (20 nmol/L) or vehicle (DMSO) for 24 hours, labeled with [1-14C]FACE (1 μCi/mL) for 6 hours, carefully washed three times with PBS, and collected for radioactivity count on a Packard Tri-Carb Liquid Scintillation Analyzer.
Cells were seeded in 12-well plates, treated with rapamycin (20 nmol/L) for 24 hours, and incubated for 3 hours at 37°C with 1 μCi/mL of [1-14C]acetate (PerkinElmer Inc.). Perchloric acid (3 mol/L) was added to the culture media and the dishes were sealed with a phenylethylamine (Sigma-Aldrich)-saturated Whatman filter paper to capture 14C-CO2, as previously done (32). Following 3-hour incubation at room temperature on a gentle shaker, the filter paper was removed, placed into Ultima Gold F Scintillation Fluid (PerkinElmer Inc.), and radioactivity was counted on a Packard Tri-Carb Liquid Scintillation Analyzer.
Oxygen consumption rate
Cells were treated with rapamycin (20 nmol/L), [19F]FACE (40 mmol/L; obtained from Advanced Biochemical Compounds), combination of the two, or vehicle (DMSO) for 20 hours. Following treatment, oxygen consumption rate (OCR) was measured under basal conditions and in the presence of oligomycin (1 μmol/L), trifluoromethoxy carbonylcyanide phenylhydrazone (FCCP; 1.5 μmol/L), or the combination of antimycin A (0.5 μmol/L) and rotenone (0.5 μmol/L) using the Seahorse Bioscience XF24 analyzer, as we have described previously (32). Results were normalized to crystal violet staining (optical density, O.D.).
Metabolites were extracted from cultured cells on dry ice using 80% aqueous methanol precooled at −80°C. An internal standard, [D8]-valine (Cambridge Isotope Laboratories), was added during metabolite extraction. Insoluble material was removed by centrifugation. The supernatant was evaporated to dryness by SpeedVac at 42°C. The pellet was resuspended in LC/MS water and metabolites were analyzed by LC/MS.
LC/MS analysis was performed on a Vanquish ultra-high performance liquid chromatography system coupled to a Q Exactive mass spectrometer (Thermo Fisher) that was equipped with an Ion Max source and HESI II probe. External mass calibration was performed every 7 days. Metabolites were separated using a ZIC-pHILIC stationary phase (150 mm × 2.1 mm × 3.5 mm; Merck) with guard column. Mobile phase A was 20 mmol/L ammonium carbonate and 0.1% ammonium hydroxide. Mobile phase B was acetonitrile. The injection volume was 2.5 μL, the mobile phase flow rate was 150 μL/minute, the column compartment temperature was set at 25°C, and the autosampler compartment was set at 4°C. The mobile phase gradient (%B) was 0 minutes, 80%; 20 minutes, 20%; 20.5 minutes, 80%; 28 minutes, 80%. The column effluent was introduced to the mass spectrometer with the following ionization source settings: sheath gas 40, auxillary gas 15, sweep gas 1, spray voltage ± 3.0 kV, capillary temperature 275°C, S-lens RF level 40, probe temperature 350°C. The mass spectrometer was operated in polarity switching full scan mode from 70 to 1,000 m/z. Resolution was set to 70,000 and the AGC target was 1 × 106 ions. Data were acquired and analyzed using TraceFinder software (Thermo Fisher Scientific) with peak identifications based on an inhouse library of authentic metabolite standards previously analyzed utilizing this method. Metabolite data are expressed as the ratio of peak areas of the target metabolite to the internal standard normalized to total cell proteins determined from a parallel experiment.
Histology and IHC
Dissected tumors were fixed in formalin for 24 hours and stored in 70% ethanol until processing. Hematoxylin and eosin (H&E) and immunohistochemistry (IHC) staining were performed on 5-μm sections of formalin-fixed and paraffin-embedded samples. IHC was performed by HistoWiz Inc. using a fully automated workflow and an automated autostainer (Bond Rx, Leica Biosystems). Tissue sections were deparaffinized, rehydrated, and subjected to heat-mediated antigen retrieval. The following primary antibodies were used: rabbit polyclonal anti-Ki67 (catalog # ab15580, Abcam) at 1:6,000 dilution and rabbit monoclonal anti-phosphoS6 ribosomal protein (cat# 4858, Cell Signaling Technology) at 1:200 dilution. Bond Polymer Refine Detection (Leica Biosystems) was applied according to the manufacturer's protocol. Sections were then counterstained with hematoxylin, dehydrated, and film coverslipped using a TissueTek-Prisma and Coverslipper (Sakura). Whole slide scanning (40×) was performed on an Aperio AT2 (Leica Biosystems). Whole slide images were quantified using automated image analysis software Halo from Indica labs and cytonuclear module.
Total proteins were extracted through 15-minute incubation with Nonidet P-40 lysis buffer and resolved on Bolt Bis-Tris Plus polyacrylamide gels (Life Technologies). Phospho-S6 kinase (catalog # 9234S), total S6-kinase (catalog # 2708S), and Tuberin (catalog # 4308S) antibodies were obtained from Cell Signaling Technology. β-Actin (catalog # A5316) was obtained from Sigma-Aldrich. Primary antibodies were diluted 1:1,000 in 5% BSA or 5% nonfat milk in TBS with 0.5% Tween 20, following the manufacturer's instructions, and probed on nitrocellulose membrane overnight at 4°C.
Statistical calculations were performed using GraphPad Prism 5. Data are reported as median values and range, unless otherwise noted. Statistical significance was defined as P < 0.05.
Rapamycin-sensitive uptake of [18F]FCH in TSC2-deficient tumor xenografts
To test whether enhanced phospholipid incorporation of choline in TSC2-deficient cells (9) may allow visualization and measurement of metabolic activity of TSC tumors, including LAM, we conducted a trial of dynamic PET using [18F]FCH in a preclinical model of TSC generated by subcutaneous injection of ELT3 cells. TSC2-deficient tumors showed rapid uptake of [18F]FCH (n = 27 mice, n = 28 tumors; typical images shown in Fig. 2A and B and Supplementary Fig. S1A and S1B). The high [18F]FCH uptake in the abdomen and neck reflects physiologic uptake by the liver, which occupies most of the abdominal cavity in the mouse, and by the salivary glands, which are very prominent in the mouse neck.
Next, we asked whether [18F]FCH uptake could represent a suitable imaging biomarker for assessment of rapamycin effectiveness in TSC and LAM. A group of mice that underwent [18F]FCH-PET imaging at baseline, was also scanned after 72-hour treatment with rapamycin (Fig. 2C and D; Supplementary Fig. S1C and S1D) or vehicle (Supplementary Fig. S2). Pre and posttreatment TACs were derived for the tumor, blood, and muscle (Fig. 2E; Supplementary Figs. S1E and S2I and S2J) in all mice, revealing a significant change from baseline in tumor SUV after treatment with rapamycin compared with vehicle (P < 0.01; Fig. 3A). Specifically, tumor [18F]FCH average SUV decreased by 49.9% (−75 to −31%) after 72-hour treatment with rapamycin (n = 5), whereas an increase by 12.8% (−10% to 29.7%) was observed in vehicle-treated mice (n = 4; 5 tumors). As expected, no SUV change from baseline was found for muscle and blood. A separate group of mice (n = 6) was imaged 24 hours after a single dose of rapamycin, showing heterogeneous response with a trend to decrease in tumor SUV versus baseline (−10%; −30% to 20.8%). Three of six tumors showed higher than 10% decrease in SUV (12%–30%; Fig. 3A).
All tumors in this study were confirmed to be viable by a pathologist using H&E staining (Fig. 3B). mTORC1 activity was suppressed by rapamycin at 24 hours and fully abrogated at 72 hours, as shown by IHC staining for phospho-S6, which is a substrate for the mTORC1 direct target S6 kinase (Fig. 3B). Tumor SUV did not correlate with proliferation index (Ki-67%; Fig. 3C) or tumor size (Fig. 3D).
Finally, rapid uptake of [18F]FCH within 5 to 10 minutes was also observed in a tumor model generated using LAM patient renal angiomyolipoma–derived TSC2-deficient 621-101 cells (Supplementary Fig. S3A and S3B).
Ovariectomy induces a decrease in [18F]FCH uptake in TSC2-deficient tumor xenografts
LAM occurs almost exclusively in women of childbearing age, and a role for estrogen signaling in promoting aggressiveness in preclinical models of TSC2 deficiency has been proposed (25, 33, 34). We asked whether estrogen signaling plays a role in [18F]FCH uptake in TSC2-deficient cells. Indeed, ovariectomized mice showed significantly reduced tumor average SUV (0.36; 0.23–0.39, n = 4) compared with nonovariectomized mice (0.67; 0.40–1, n = 28 tumors, including pretreatment tumor SUV from the rapamycin group; P < 0.01), while no significant difference was found between intact mice and ovariectomized mice with 17β-estradiol supplementation (0.54; 0.42–0.96, n = 3; Fig. 3E).
[18F]FCH-PET imaging of a preclinical model of pulmonary LAM
To test feasibility of [18F]FCH-PET to image TSC2-deficient cells infiltrating the pulmonary parenchyma, we used a novel model of pulmonary LAM by injection of ELT3 cells into the mouse tail vein. Mice were imaged 25 to 90 days after cell injection. The diseased lung VOI showed approximately 2-fold higher SUV (1.3; 1.27–1.74, n = 3) than healthy control lungs (0.74; 0.7–0.77, n = 3; P < 0.05; Fig. 4A–E). The presence of TSC2-deficient nodules showing hyperactivation of mTORC1 signaling (by P-S6 IHC staining) within the diseased lungs was confirmed by histopathology (Fig. 4F).
[18F]FACE uptake in TSC2-deficient tumor xenografts is not rapamycin-sensitive
Next, we tested the potential for [18F]FACE to trace lipogenesis in TSC2-deficient cells as an acetate analogue. Mice were scanned for 90 minutes as suggested previously for [11C]acetate (35). TSC2-deficient xenografts generated by subcutaneous injection of ELT3 cells showed rapid uptake of [18F]FACE (n = 10 tumors; tumor SUV median = 1.33, range = 1.15 to 1.72 at 30–40 minutes, Fig. 5A and B, and tumor SUV median = 1.19, range = 0.98 to 1.44 at 80–90 minutes, Fig. 5E and F).
TACs were derived for blood (Fig. 5I), tumor, and muscle (Fig. 5J) before and after rapamycin (Fig. 5C, D, G, and H) or vehicle treatment (72 hours; Supplementary Fig. S4A–S4J), as it was done for [18F]FCH, revealing no significant decrease from baseline in tumor SUV after treatment with rapamycin at either time point compared with vehicle (Fig. 5K).
[18F]FACE images (Supplementary Fig. S5) were also obtained for 3 mice (4 tumors) that were imaged with [18F]FCH-PET on consecutive days (Supplementary Figs. S1 and S2A–S2D, S2I) to validate specificity of tumor uptake. These tumors showed a dramatic decrease in [18F]FCH but not [18F]FACE uptake.
Validation of rapid uptake was obtained in the LAM patient renal angiomyolipoma–derived model (621–101 cells), which showed rapid uptake of [18F]FACE within 5 to 10 minutes (Supplementary Fig. S3C and S3D; tumor SUV median = 1.04, range = 1.01 to 1.13, n = 3).
Incorporation of FCH but not FACE into lipids in TSC2-deficient cells
We have previously shown that incorporation of choline into phosphatidylcholine is enhanced in TSC2-deficient mouse embryonic fibroblasts (MEF) relative to TSC2-expressing MEFs (9). To determine whether the differential effect of rapamycin on in vivo uptake of FCH or FACE depends upon the metabolic fate of these nutrients, we traced [19F]FCH and [1-14C]FACE in ELT3 cells.
Lipid extracts from ELT3 cells treated with rapamycin (20 nmol/L) or vehicle (DMSO) for 24 hours and labeled with [19F]FCH for 6 hours were analyzed by NMR. A single 19F peak was detected in both treatment conditions and absolutely quantified using [19F]FCH internal standard that was spiked into the lipid extracts before spectra acquisition. As expected, [19F]FCH incorporation was significantly suppressed by rapamycin (∼20%, P < 0.01; Fig. 6A). Suppression of mTORC1 signaling and TSC2 deficiency in ELT3 cells was validated by immunoblotting for phospho-S6 kinase (P-S6K) and TSC2 (positive control: TSC2+/+ MEFs; Fig. 6B).
When we similarly labeled rapamycin or vehicle-treated ELT3 cells with [1-14C]FACE, counts per minute from cell lipid extracts were equal to background, whereas [1-14C]acetate, used as positive control, was highly incorporated into lipids by ELT3 cells (Fig. 6C).
Like acetate, FACE enters mitochondria and is converted into fluorocitrate. Fluorocitrate is an irreversible inhibitor of aconitase, an enzyme that catalyzes the conversion of citrate to isocitrate in the Krebs cycle. Therefore, FACE is expected to accumulate into mitochondria and could be used as a surrogate biomarker of mitochondrial activity. We asked whether rapamycin-insensitive uptake of [18F]FACE in vivo could reflect this differential metabolic fate of FACE compared with acetate. ELT3 cells were incubated with [1-14C]FACE for 6 hours, in the presence or absence of rapamycin (20 nmol/L, 24 hours), and then collected to count radioactivity (tracer incorporation). Consistent with the in vivo data, no significant difference in intracellular [1-14C]FACE was found between vehicle and rapamycin-treated cells (Fig. 6D), suggesting that mitochondrial oxidative metabolism is sustained even in the presence of rapamycin. We tested this hypothesis using two independent approaches. First, we used Seahorse analyzer to measure mitochondrial OCR in TSC2-deficient cells treated with rapamycin, FACE, or combination of the two. Indeed, rapamycin did not suppress OCR in TSC2-deficient cells, whereas FACE significantly suppressed basal respiration and maximal respiratory capacity, in the presence and absence of rapamycin (Fig. 6E). Similarly, ATP production, spare respiratory capacity, and proton leak were suppressed by FACE in the presence and absence of rapamycin (Supplementary Fig. S6A–S6C). Second, we measured levels of citrate, the aconitase substrate, using LC/MS, under the same experimental conditions. FACE induced significant accumulation of citrate (Fig. 6F), as expected.
We then assayed mitochondrial acetate oxidation in TSC2-deficient cells upon treatment with rapamycin, revealing an increase in this bioenergetic process by mTORC1 inhibition (Supplementary Fig. S6D). Interestingly, we have previously shown that rapamycin leads to a decrease in glucose oxidation in TSC2-deficient cells (32), suggesting a switch in preferential mitochondrial oxidation substrate by these cells upon rapamycin-induced cytostasis.
Taken together, these data substantiate the hypothesis that rapamycin-insensitive uptake of fluoroacetate by TSC tumors reflects a specific metabolic feature of TSC2 deficiency.
This preclinical study was undertaken to assess feasibility of [18F]FCH and [18F]FACE as TSC and LAM imaging agents. Both tracers have been shown to be safe in humans and to accumulate in cancer cells allowing PET imaging of prostate cancer metastases in patients (36–38).
We previously found that TSC2-deficient cells have enhanced incorporation of choline into phosphatidylcholine (Kennedy pathway) compared with TSC2-expressing cells (9). Here, we show that TSC2-deficient cells exhibit rapid uptake of [18F]FCH in vivo and can be visualized by PET imaging in preclinical models of TSC and LAM, including subcutaneous xenografts and pulmonary nodules. Importantly, 72-hour treatment with rapamycin suppressed [18F]FCH uptake by approximately 50% in subcutaneous tumors, and [18F]FCH uptake was independent of tumor volume and proliferation rate. We aimed to obtain close to full abrogation of mTORC1 signal in the TSC2-deficient tumors (assessed by IHC for P-S6 as a surrogate of mTORC1 signaling) to evaluate efficiency of rapamycin treatment in regulating choline uptake signal. These results suggest a role for rapamycin-sensitive signaling pathways in the regulation of choline uptake and metabolism in TSC2-deficient cells. Similar results would be expected in other models of TSC loss-associated mTORC1 hyperactivation, such as TSC1 deficiency.
Pulmonary uptake of [18F]FCH was also higher in the mice that developed small, diffuse nodules in the lungs after intravenous injection of TSC2-deficient cells compared with healthy control mice, establishing proof of concept for using [18F]FCH to detect metabolic activity in patients with pulmonary LAM.
Interestingly, [18F]FCH-PET imaging of TSC2-deficient tumors generated in ovariectomized mice showed a significant decrease in tumor SUV. LAM aggressiveness could exacerbate with pregnancy (39), and a role for estrogen signaling in TSC2-deficient cell migration and invasion has been reported (25, 33, 34). An aromatase inhibitor has been tested in a clinical trial of LAM (ref. 40; NCT01353209). Our results suggest that menopause, cyclic hormonal changes that regulate the menstrual cycle, or hormonal therapies could affect [18F]FCH uptake by TSC2-deficient cells, with potential impact on the detection of pulmonary and extra-pulmonary LAM lesions.
Besides being incorporated into cellular lipids, [18F]FCH can be oxidized to [18F]fluorobetaine in certain tissues (primarily liver and kidneys), lowering the tracer parent fraction in the blood of patients to about 10% within 5 to 10 minutes (37, 41) and potentially leading to variability in measured SUV (42, 43). Importantly, in our study, we found consistent decrease in [18F]FCH tumor uptake after treatment with rapamycin, but not after vehicle treatment. Moreover, muscle and blood SUV did not show significant changes between pre and posttreatment scans. These results corroborate the specificity of our findings. However, serial measurements of concentration of [18F]FCH and its metabolites, including [18F]fluorobetaine, in arterial blood would be recommended in first in-human evaluation of [18F]FCH as a biomarker for rapamycin activity. In addition, using [18F]FCH analogues (i.e., deuterated [18F]FCH) could improve image quality (44).
The relatively high blood volume fraction and the presence of air should be considered in lung PET quantitative analysis to enable direct estimation of the metabolic rate of nutrients in lung tissue (45, 46). Correction for tissue fraction effects in lung PET remains to be addressed in patients, where full pharmacokinetic modeling can be used to estimate the blood volume fraction and subtract the signal originating from the blood (45, 47). For evaluation of the performance of [18F]FCH as an imaging biomarker for pulmonary LAM in patients, it is therefore recommended to perform a pharmacokinetic modeling study that includes arterial blood sampling or image-derived input functions obtained from volumes of interest in blood-pool structures distant from tissues with high [18F]FCH uptake (42).
A metabolic imaging biomarker would be powerful for patients with LAM who do not have clinically significant blood VEGF-D levels or could be used together with circulating VEGF-D as complementary diagnostics to enhance clinical decision making.
Acetate is a critical nutrient in TSC2-deficient cells, contributing to both bioenergetic (mitochondrial oxidative metabolism) and biosynthetic (de novo fatty acid and cholesterol synthesis) processes. The short half-life (20.4 minutes) of 11C limits the general availability of [11C]acetate compared with [18F]FACE ([18F] half-life is 110 minutes), which makes it less suitable in a routine clinical setting. Therefore, we tested the potential for [18F]FACE as an analogue of acetate and an imaging tracer in TSC and LAM. [14C]FACE was not incorporated into cellular lipids, and its uptake by TSC2-deficient cells, which reflects mitochondrial accumulation, was unchanged after 24-hour treatment with rapamycin, suggesting that inhibition of mTORC1 by rapamycin does not suppress mitochondrial metabolism. Consistent with these results, [18F]FACE tumor SUV did not decrease after 72-hour rapamycin treatment. However, [18F]FACE tumor SUVs were higher than [18F]FCH tumor SUVs, revealing a potential role for [18F]FACE in measuring whole-body tumor burden as well as metabolic activity of TSC2-deficient lesions in TSC and LAM.
Studies conducted in healthy volunteers and/or patients with cancer have shown safety of diagnostic doses of [18F]FACE. These include a study from Takemoto and colleagues (48) reporting an imaging trial of 24 healthy volunteers (age 48.2 ± 12.9 years old; 15 male and 9 female) and 10 patients with liver tumor (age 72.1 ± 7.0 years old; 6 male and 4 female).
In addition, to facilitate the clinical translation of [18F]FACE-PET, the pharmacokinetics, biodistribution, radiation dosimetry, and pharmacologic safety of diagnostic doses of [18F]FACE were determined in nonhuman primates (49). In this study, electrocardiograms and hematology analyses were conducted to evaluate the acute and delayed toxicity of diagnostic dosages of [18F]FACE. The time-integrated activity coefficients for individual organs and the whole body after administration of [18F]FACE were obtained using quantitative analyses of dynamic and static PET images and were extrapolated to humans. The conclusion of the study was that diagnostic dosages of [18F]FACE could be safely administered to humans for PET imaging.
In summary, our data validate the hypotheses that TSC2 loss–associated aberrant cellular metabolism may provide opportunities for metabolic imaging in TSC and LAM, and that preclinical models of TSC2-deficiency represent informative platforms to identify tracers of potential clinical interest. Finally, results from our preclinical trial support a potential role for [18F]FCH and [18F]FACE as metabolic imaging agents for TSC and LAM proliferative lesions, and for [18F]FCH-PET as a noninvasive means to monitor response to rapamycin in these tumors.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: E.E. Verwer, M.F. Kijewski, G.E. Fakhri, M.D. Normandin, C. Priolo
Development of methodology: E.E. Verwer, T.M. Shoup, W. Massefski, Y. Cui, W.M. Oldham, G.E. Fakhri, M.D. Normandin, C. Priolo
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E.E. Verwer, T.R. Kavanagh, W.J. Mischler, Y. Feng, K. Takahashi, S. Wang, N.J. Guehl, W. Massefski, S. El-Chemaly, W.M. Oldham, M.F. Kijewski, C. Priolo
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E.E. Verwer, T.R. Kavanagh, W.J. Mischler, N.J. Guehl, W. Massefski, P.M. Sadow, W.M. Oldham, M.D. Normandin, C. Priolo
Writing, review, and/or revision of the manuscript: E.E. Verwer, T.R. Kavanagh, W.J. Mischler, Y. Feng, K. Takahashi, R. Neelamegam, C. Ran, Y. Cui, W. Massefski, S. El-Chemaly, P.M. Sadow, M.D. Normandin, C. Priolo
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E.E. Verwer, T.R. Kavanagh, W.J. Mischler, Y. Feng, K. Takahashi, J. Yang, N.J. Guehl, G.E. Fakhri, C. Priolo
Study supervision: E.E. Verwer, C. Priolo
Other (writing software tools for image analysis): E.E. Verwer
Other (synthesis of 18F-radiotracer for the preclinical studies): R. Neelamegam
We are thankful to the Harvard Catalyst Biostatistical Consulting Program for assisting with the statistical analyses. This work was supported through funds from the Department of Defense (W81XWH-13-1-0262 and W81XWH-16-1-0165), NIH R01HL130336, the Tuberous Sclerosis Alliance, and the LAM Foundation Biomarker Innovation Grant to C. Priolo. The effort of E.E. Verwer was partly supported by the Cassen Postdoctoral Mentoring Award from the Education and Research Foundation at the Society of Nuclear Medicine and Molecular Imaging to M.D. Normandin. W.M. Oldham was supported by NIH 1K08HL128802-01A1.
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.