A Clinical PET Imaging Tracer ([¹⁸F]DASA-23) to Monitor Pyruvate Kinase M2–Induced Glycolytic Reprogramming in Glioblastoma

Enhanced glucose uptake is a hallmark of multiple cancers and has been developed as a clinical diagnostic tool through PET imaging of the radiolabeled glucose analogue, 2-[¹⁸F]fluoro-2-deoxy-D-glucose ([¹⁸F]FDG; refs. 1, 2). [¹⁸F]FDG is an indispensable tool in oncology for the staging and restaging of cancer, detection of recurrence, monitoring of treatment response, and facilitating timely modifications of therapeutic strategies (3, 4). Unfortunately, high background signal from glucose metabolism within the healthy brain limits the diagnostic utility of [¹⁸F]FDG in brain cancers (5–7). Radiolabeled amino acids represent another class of PET tracers used in the evaluation of brain cancers. They have shown improved utility for imaging brain tumors due to the increased uptake in tumor tissue, low uptake in healthy brain tissue, and include [¹¹C]-methionine, L-3,4-dihydroxy-6-[¹⁸F]-fluorophenyl-alanine ([¹⁸F]FDOPA), and [¹⁸F]-fluoroethyltyrosine ([¹⁸F]-FET; refs. 8–12). Their uptake is predominantly mediated by the L-type amino acid transporter (LAT) and generally reflects nutrient uptake to support increased biomass and proliferative energy demands (6, 13). Indeed, these agents provide important diagnostic information regarding delineation of tumor extent, diagnosis of treatment-related changes, and assessment of treatment response. But, given the key role of glycolysis in cancer cell metabolism and response to therapy (14), brain cancers would benefit from noninvasive imaging of glycolysis.

Pyruvate kinase (PK) catalyzes the conversion of phosphoenolpyruvate to pyruvate, simultaneously producing ATP (Fig. 1A; ref. 15). The expression of the pyruvate kinase M2 (PKM2) isoform is reported to contribute to the distinctive glycolysis in cancers, and replacement of PKM2 with its splice variant, PKM1, results in insufficient biosynthesis, thus inhibiting cancer growth (16). PKM1 is present in differentiated tissues including the brain, while PKM2 is found in most proliferating cells, including the majority of cancer cell lines and malignancies studied to date. PKM1 has high constitutive enzymatic activity, whereas PKM2 is allosterically regulated through the stabilization of a highly active tetramer relative to the inactive dimeric PKM2 (17, 18). The reduced activity of dimeric PKM2 results in a diminished production of pyruvate, enabling accumulation of upstream glycolytic intermediates and a metabolic shift towards an anabolic state (18). The dynamic equilibrium between the dimeric and tetrameric states of PKM2 enables proliferating cancer cells to regulate their needs for anabolic and catabolic metabolism and, to favor high levels of dimeric PKM2 (19). The Warburg effect is therefore partly mediated by PKM2 expression, with high expression of dimeric PKM2 in cancer cells contributing to anabolic glucose metabolism, promoting macromolecular biosynthesis, and benefiting cancer cell proliferation and growth (20). To date, there have been no reports of radiopharmaceuticals that can provide a noninvasive measure of PKM2 status in human subjects.

Glioblastoma (GBM) is the most common and lethal primary central nervous system cancer in adults (21). It is a WHO grade IV glioma representing approximately 15% of all primary intracranial malignancies (21). Despite aggressive surgical resection, radiotherapy, and chemotherapy, prognosis remains dismal with overall survival of newly diagnosed GBMs being only approximately 5% at 5 years after diagnosis (21). PKM2 expression in gliomas is well established; grade I–III gliomas exhibit modestly increased levels of PKM2 protein expression relative to normal brain tissue. GBM displays a further 4- to 5-fold increase in PKM2 protein expression (22). Given the absence of PKM2 in healthy brain and its role in glucose metabolism, PKM2 represents an attractive imaging target for investigating patients with GBM and other brain malignancies. Our group has previously reported the development of 1-((2,6-difluorophenyl)sulfonyl)-4-((4-(methoxy-[¹¹C])phenyl)sulfonyl)piperazine ([¹¹C]DASA-23) as a novel radiotracer that provided a direct measure of PKM2 in preclinical models of GBM (23). These studies demonstrated the in vitro and in vivo specificity of [¹¹C]DASA-23 for the detection of PKM2 (23). Owing to the favorable properties of fluorine-18 for clinical translation, we subsequently reported the radiosynthesis of 1-((2-fluoro-6-[¹⁸F]fluorophenyl)sulfonyl)-4-((4-methoxyphenyl)sulfonyl)piperazine ([¹⁸F]DASA-23), an isotopolog PKM2-specific radiotracer (Fig. 1A; refs. 24–26). In this study, we report the preclinical development and initial clinical evaluation of [¹⁸F]DASA-23 to image PKM2 status in patients with GBM.

We purchased LN18, U87, and U87 IDH1R132H cells from ATCC and maintained them in culture according to the manufacturer's protocol. We generated U87-GFP/luc by transduction with a lentiviral vector that expressed a fusion protein of GFP and firefly luciferase (luc) followed by puromycin selection (125 ng/mL). We received GBM206 cells as a gift from Professor Jann Sarkaria (Mayo Clinic). Mycoplasma testing was performed on a monthly basis. Cells were kept within 15 passages of the original frozen vial and authenticated using short tandem repeat profiling.

After obtaining informed consent under Stanford Institutional Review Board (IRB) protocol 12625, we acquired patient tissue used to generate the human glioblastoma cell line TP459 through Stanford IRB protocol 34363. We obtained a single-cell suspension of TP459 cells by mechanical dissociation using a scalpel, followed by enzymatic dissociation using DNAse I (250 U/mL, DPRFS) and Liberase TM (LIBTM-RO, Sigma Aldrich) on a shaker at 37°C. Following enzymatic digestion, a Percoll gradient was used to separate myelin. We depleted red blood cells using a density gradient (Histopaque 1119, Sigma Aldrich). The remaining cell suspension was then harvested, washed, and stained with anti-human CD45 antibody conjugated to phycoerythrin using a 1:20 ratio (5 μL in 100 μL staining solution; clone HI30, BD Biosciences) followed by anti-PE microbeads using a 1:5 ratio (20 μL in 100 μL staining solution; 130–048–801, Miltenyi Biotec). We collected CD45⁻ cells via magnetic separation and transferred to tissue culture flasks for in vitro culture at 37°C and 5% CO₂. We maintained TP459 cells in culture as a neurosphere suspension in serum-free medium.

We completed the radiosynthesis of [¹⁸F]DASA-23 as per previously reported methods (24).The radiochemical yield was 1.68% ± 0.92%, nondecay corrected at end of synthesis and the molar activity was 100.47 ± 29.58 GBq/μmol (n = 15).

We plated all cells (2 × 10⁵ per well) in 12-well plates the day before uptake studies. Cell uptake studies using U87, U87 IDH1R132H, LN18, GBM206, and TP459 cell lines were performed per previously reported methods (25, 26). TP459 neurospheres were processed into single-cell suspensions prior to plating following treatment with Accustase (A1110501, Thermo Fisher Scientific Inc.).

We obtained approval for all experimental procedures involving animals from the Stanford University Institutional Animal Care and Use Committee under protocol 12040. Orthotopic brain tumor models were completed according to previously described methods (23). We implanted 4 × 10⁵ U87 cells in the right hemisphere, 0.5 mm anterior and 2 mm lateral to the skull lambda, in the brains of 6- to 8-week-old nude mice held in place with a stereotaxic frame. Cells were suspended in 3 μL of Hank's balanced salt solution (HBSS) and were injected at a depth of 3 mm over 5 minutes using a Hamilton syringe, which was subsequently held in place for a further 2 minutes. We subsequently imaged tumor-bearing mice approximately 21 days after intracranial injection.

We completed MRI studies at the Stanford Small Animal Imaging facility using a Discovery MR901 General Electric 7T horizontal bore scanner (GE Healthcare) according to previously described methods (23). We completed small-animal PET imaging scans on a docked Siemens Inveon PET/CT scanner (matrix size 128 × 128 × 159; CT attenuation corrected; nonscatter corrected) after a bolus of intravenous injection of 8 ± 0.9 MBq [¹⁸F]DASA-23 in approximately 150 μL. We acquired dynamic scans in listmode format over 30 to 60 minutes. The acquired data were then arranged into 0.5-mm sonogram bins and 15–19 frames for image reconstruction (4 × 15 s; 4 × 60 s, and 11 × 300 s). The densities were averaged for all regions of interest at each time point to compute a time–activity curve. We normalized tumor and tissue time–activity curves to the injected amount of radioactivity and expressed values as %ID/g, assuming 1 g/mL. The normalized uptake of radiotracer at 30 minutes within the healthy contralateral region of the brain was used for comparison. We completed PET/MR registration in Inveon Research Workplace (IRW, Siemens) software and used the CT image for alignment of the skull. The CT intensity was then lowered to allow visualization of fused PET/MR images. We used IRW for visualization of radiotracer uptake within the tumor, to define the volumes of interest and to create the images. Tumor [¹⁸F]DASA-23 uptake was quantified using threshold-based three-dimensional (3D) volume of interest (VOI) in IRW. VOIs were defined on summed images for 10 to 30 minutes postinjection and transferred to the dynamic images to determine time–activity curves.

We completed autoradiography per previously described methods (27). For blocking studies, we administered [¹⁸F]DASA-23 one hour after intraperitoneal injection of TEPP-46 [50 mg/kg in 40% w/v (2-hydroxypropyl)-b-cyclodextrin in PBS; Cayman Chemical]. Thirty minutes after injection of radioactivity, we transcardially perfused mice with saline (15 mL) and the brain embedded in optimal cutting temperature and processed for autoradiography according to published methods (27).

We stained whole tissue sections with hematoxylin and eosin (H&E) and anti-PKM2 immunofluorescence according to published methods (26).

We orthotopically implanted nude mice (n = 4) with U87-GFP/luc glioma cells as described previously (23). After 21 days of tumor implantation, the brain was rapidly removed and placed into approximately 5 mL ice-cold HBSS containing HEPES (12.5 mL 1 mol/L HEPES per 487.5 mL HBSS). We homogenized the brains and passed the homogenate through a 40-μm cell strainer. The sample was centrifuged at 300 × g for 5 minutes at 4°C to remove excess buffer, then resuspended in a standard isotonic Percoll solution, and centrifuged at 800 × g for 20 minutes at 4°C to remove myelin. We then resuspended cells in 2% BSA in PBS and stained with antibodies for flow cytometry analysis. Cell surface marker antibodies were added at a concentration of 0.2 μg per 1 × 10⁶ cells in 100 μL volume. Antibodies included anti-CD45 (30F11, BioLegend), and anti-CD11b (M1/70, BioLegend) and anti-PKM2 (D78A4, Cell Signaling Technology). We washed all samples and then analyzed them using a FACSAria II (BD Biosciences). We analyzed the data using FlowJo. We first gated cells using forward scatter and side scatter. We then gated immune cells on CD11b and CD45, while tumor cells were gated on the basis of GFP signal.

We purchased temozolomide from SellekChem and determined the IC₅₀ value in TP549 cells according to reported methods (25).

This clinical trial (NCT03539731) was approved by the local IRB (IRB-44597) and the local Cancer Center's scientific review committee (BRN0038), and is compliant with federal, state, and local regulation on medical research, and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. We obtained written informed consent from all individual participants. Inclusion criteria included age ≥ 18 years; adequate organ function as assessed by blood analysis up to 14 days prior to the [¹⁸F]DASA-23 PET/MRI scan and defined by absolute neutrophil count (ANC) ≥ 1.5 × 10⁹/L without myeloid growth factor support; hemoglobin (Hgb) ≥ 9 g/dL; platelet count ≥ 100 × 10⁹/L; bilirubin ≤ 1.5 × upper limit of normal (ULN) except for documented history of Gilbert disease; alanine aminotransferase (ALT) and aspartate aminotransferase (AST) each ≤ 2.5 × ULN; alkaline phosphatase (AP) ≤ 3 × ULN; and in women of childbearing potential, a negative serum pregnancy test. Exclusion criteria included pregnant and nursing participants, implanted devices contraindicated for MRI, prespecified comorbid diseases, incurrent illness, prior or current malignancy, known kidney disease, and history of allergic reactions to skin adhesives (e.g., tapes) used in medical care. We requested no specific patient preparation (i.e., fasting) prior to the [¹⁸F]DASA-23 PET/MRI scan. We monitored vital signs, including heart rate, pulse oximetry, and body temperature, pre- and post-administration of [¹⁸F]DASA-23. To evaluate toxicity of the injected dose of [¹⁸F]DASA-23, we obtained blood specimens (complete blood count with differential and complete metabolic panel) from each participant and analyzed before and within 7 days postinjection of [¹⁸F]DASA-23. Adverse events were recorded on the day of imaging as well as during follow-up up to 7 days.

Inclusion criteria included age ≥ 18 years, adequate organ function as assessed by blood analysis within 14 days prior to the [¹⁸F]DASA-23 PET/MRI scan, a negative pregnancy test in women of childbearing potential, and any of the following: (i) radiographic or pathologic evidence of newly diagnosed intracranial tumor that was presurgical resection, (ii) radiographic or pathologic evidence of progressive/recurrent intracranial tumor, (iii) question of pseudoprogression versus true progression on the most recent standard-of-care brain MRI, or (iv) evidence on the most recent standard-of-care brain MRI scan of intracranial metastasis/metastases in a patient with known extracranial primary cancer. Exclusion criteria included pregnant and nursing participants, implanted devices contraindicated for MRI, prespecified comorbid diseases, incurrent illness, known kidney disease, history of allergic reactions to skin adhesives, and other chemotherapy (besides what was being used to treat the intracranial tumor).

Whole-body tracer distribution was assessed in a cohort of healthy volunteers (Group I, NCT03539731) using serial PET/MRI scans on a time-of-flight simultaneous PET/MRI scanner (SIGNA PET/MRI; GE Healthcare) as described previously (28). Patients with brain tumor (groups II and III, NCT03539731) and an additional cohort of healthy volunteers (group IV, NCT03539731) underwent 60-minute brain acquisitions; a dynamic emission acquisition sequence in 3D mode over 60 minutes was started immediately prior to intravenous administration of [¹⁸F]DASA-23. Participants received 294 ± 12 MBq of [¹⁸F]DASA-23. We reconstructed the PET data using a fully 3D iterative ordered subsets expectation maximization algorithm (28 subsets, 2 iterations) and corrected for attenuation, scatter, dead time, and decay. We applied MR-based ZTE attenuation correction and reconstructed the image into the following frames: 8 × 15 s; 16 × 30 s, 20 × 60 s, 10 × 200 s, and 2 × 300 s. We completed a standardized brain tumor MRI protocol (29). As per standard of care, 330 MBq of [¹⁸F]FDG was administered to intracranial patient 1 (IC-1). Approximately 45 minutes after intravenous administration of [¹⁸F]FDG, a plain CT was obtained from the vertex to the skull base in 3D model for use in attenuation correction and anatomic localization of radiotracer activity. Emission scans were completed over the same anatomic regions. [¹⁸F]FDG PET images were retrospectively accessed, reconstructed, and reviewed in the axial, coronal, and sagittal planes under IRB protocol 42357.

The whole-body dosimetry of [¹⁸F]DASA-23 was assessed as described previously (28). The brain distribution and kinetics of [¹⁸F]DASA-23 were assessed using a brain region atlas approach implemented in the PMOD PNEURO tool. The anatomic MRI was segmented and matched to a template before transferring the atlas regions to the PET data. Regional time–activity curves were extracted and adjusted for injected dose and body weight by calculating standard uptake value (SUV).

Tumor [¹⁸F]DASA-23 uptake was quantified using threshold-based 3D VOIs in the PMOD PFUS tool. VOIs were defined on summed images for 30 to 60 minutes postinjection and transferred to the dynamic image series to extract time–activity curves. Tumor VOIs were defined using a 70% peak-voxel activity isocontour. The normal white matter reference brain region for tumor-to-brain calculation was the contralateral centrum semiovale. The SUV in gray matter was measured using manually drawn VOIs along the cortical band of the contralateral frontal and parietal cortices. Ratios of tumor uptake were defined as tumor-to-brain ratio (TBR), tumor-to-brain maximum ratio (TBR_max), and tumor-to-gray matter ratio. SUV histograms for the tumors in IC-1 to IC-4 were calculated by extracting the voxel values for voxels within the 70% peak-voxel activity isocontour VOI using the Pixeldump functionality in PMOD.

We obtained informed consent under IRB protocol 12625, and then obtained patient tissue used to evaluate PKM2 expression by Western blot analysis through Stanford IRB protocol 34363. We homogenized tissues on ice in RIPA buffer (89900, Thermo Fisher Scientific Inc.) containing 1× Halt protease inhibitors (78429, Thermo Fisher Scientific Inc.). We prepared the U87 cell lysate by lysing U87 cells in RIPA buffer containing protease inhibitors for 10 minutes on ice. We determined the protein concentration of these samples using the BCA assay and evaluated the cellular lysates for PKM2 protein expression using a standard Western blotting protocol (25).

Under IRB protocol 42357, we obtained formalin-fixed paraffin-embedded brain tumor tissue. The specimens were reviewed with a neuropathologist to identify regions of neoplasm and these regions were sectioned at 5 μm. Adjacent sections were stained with standard H&E and, following antigen retrieval at pH 6.0, by IHC for PKM2 (1:600, D78A4 XP rabbit mAb, Cell Signaling Technology, catalog no. 4053). We scanned the slides on a digital pathology slide scanner (Aperio, Leica Biosystems) and images were viewed using Aperio ImageScope version 12.4.0.5043 (Leica Biosystems).

Data analysis and visualization were performed using Prism 7.0 (GraphPad Software). All data are presented as the average value ± SD of at least three independent measurements. Statistical analysis for cell and animal studies were performed by Student t test and one-way ANOVA using Prism as indicated in the figure legends. Multiple comparisons were assessed using the Tukey multiple comparisons test. Significance was assigned for P values <0.05.

We radiolabeled DASA-23 with ¹⁸F in the native position of fluorine on the DASA-23 molecule, as reported previously (Supplementary Fig. S1; ref. 24), and obtained [¹⁸F]DASA-23 with a radiochemical yield of 1.68% ± 0.92%, molar activity of 100.47 ± 29.58 GBq/μmol, and chemical and radiochemical purity >95% (n = 15). We performed initial cellular uptake studies in human GBM cell lines possessing diverse molecular characteristics, namely isocitrate dehydrogenase 1 (IDH1) mutation and methylguanine-DNA methyltransferase (MGMT) promoter methylation status. Rapid and extensive cellular uptake was observed in all GBM lines evaluated after addition of [¹⁸F]DASA-23. At 30 minutes after addition of radioactivity, the highest uptake values were evident in U87 (IDH1 wild-type and MGMT promoter methylated) and LN-18 (IDH1 wild-type and MGMT promoter unmethylated) cells, reaching 25.3% ± 7.2% and 23.9% ± 1.4% uptake/mg protein, respectively (Fig. 1B). This was followed by U87 IDH1 R132H (MGMT promoter methylated) and patient-derived GBM206 (IDH1 wild-type, and MGMT promoter methylated) cells that reached [¹⁸F]DASA-23 uptake values of 15.9% ± 3.6% uptake/mg protein and 15.5% ± 3.0% uptake/mg protein, respectively. [¹⁸F]DASA-23 uptake continued to increase over the 60-minute incubation period in all cell lines evaluated. Removal of exogenous radioactivity resulted in the efflux of cell-associated activity, with approximately 42% retention of the initial intracellular radioactivity remaining 30 minutes after removal of exogenous [¹⁸F]DASA-23 for all cell lines (Fig. 1B).

In view of the favorable cancer cell uptake of [¹⁸F]DASA-23, we assessed the potential use of this tracer for in vivo imaging of PKM2 status. We first assessed the brain penetrance and biodistribution of [¹⁸F]DASA-23 using small-animal PET/CT imaging of healthy nude mice. Dynamic imaging following intravenous administration of approximately 8 MBq [¹⁸F]DASA-23 revealed it passively crossed the intact blood–brain barrier (BBB), with peak uptake levels of 8.6% ± 0.6%ID/g (n = 4) occurring at approximately 1 minute after tracer administration (Fig. 1D). Subsequently, [¹⁸F]DASA-23 radioactivity within the healthy brain rapidly cleared over the 60-minute acquisition period, with low levels remaining at 30 and 60 minutes after tracer administration (0.8% ± 0.1%ID/g and 0.7% ± 0.1%ID/g, respectively; Fig. 1D). The clearance of [¹⁸F]DASA-23 was predominantly found to be through renal and hepatobiliary pathways (Fig. 1E).

Owing to the low levels of tracer accumulation within healthy brain, we were motivated to explore the ability of [¹⁸F]DASA-23 to measure PKM2 status in mice orthotopically implanted with human U87 GBMs genetically modified to express GFP and firefly luciferase (U87-GFP/luc). We used T2-weighted MRI to confirm the presence of the intracranial tumor in each mouse (Fig. 2A). Small-animal PET imaging following intravenous administration of [¹⁸F]DASA-23 (∼8 MBq) allowed clear visualization of aberrantly expressed PKM2 within the brain tumors (Fig. 2A). The kinetics of [¹⁸F]DASA-23 within the U87-GFP/luc tumors consisted of rapid tracer delivery followed by slow washout (Fig. 2B). At 30 minutes post radiotracer administration, the levels of radioactivity within the U87-GFP/luc GBMs were 1.9% ± 0.4%ID/g, which were significantly higher than those of the contralateral normal brain (0.6% ± 0.1%ID/g, P < 0.0001, n = 8; Fig. 2C), establishing a TBR of 3.6 ± 0.5. We compared the ability of [¹⁸F]DASA-23 to delineate orthotopic U87-GFP/luc GBMs compared with an established amino acid PET agent, 3,4-dihydroxy-6-[¹⁸F]-fluoro-l-phenylalanine ([¹⁸F]FDOPA) currently explored in the clinical management of patients with GBM (30). Similar to [¹⁸F]DASA-23, [¹⁸F]FDOPA delineated the orthotopically implanted U87-GFP/luc GBMs (Supplementary Fig. S2) with a TBR of 2.5 ± 0.3 (n = 4) at 30 minutes after tracer administration.

After [¹⁸F]DASA-23 PET imaging, we excised brains of tumor-bearing mice for analysis using ex vivo autoradiography and histopathology. Similar to the PET images, ex vivo autoradiography revealed the radioactive signal within the U87-GFP/luc GBMs to be clearly defined from the surrounding brain tissues (Fig. 2D). H&E staining of adjacent sections demonstrated the presence of highly cellular tumor tissue with good localization to [¹⁸F]DASA-23 autoradiography (Fig. 2E). Immunofluorescence staining of PKM2 within adjacent tumor sections revealed excellent correlation with the GFP-positive U87 cells and cytosolic PKM2 expression (Supplementary Fig. S3A and S3B). PET imaging and autoradiography completed under blocking conditions with structurally distinct PKM2 activator TEPP-46 (50 mg/kg) attenuated the [¹⁸F]DASA-23 signal within the U87 GBMs at 30 minutes postinjection, highlighting the in vivo specificity of [¹⁸F]DASA-23 for PKM2 (Supplementary Fig. S3C–S3E). We additionally processed the brains of U87-GFP/luc tumor-bearing mice (n = 4) for analysis of PKM2 levels in multiple cell populations within the tumor microenvironment using flow cytometry. This revealed PKM2 expression was almost exclusively found in U87 GFP–positive cancer cells with minimal PKM2 found in infiltrating myeloid, microglial, or lymphoid cell populations (Fig. 2F). Collectively, these data highlight the potential for clinical translation of [¹⁸F]DASA-23 to evaluate PKM2 status in patients harboring GBMs.

We intravenously administered [¹⁸F]DASA-23 to 4 healthy male and 2 healthy female adult volunteers to study radiation dosimetry and brain kinetics using simultaneous PET/MRI (Supplementary Table S1). We noted no adverse events at the time of imaging or after 7 days; the effective dose of [¹⁸F]DASA-23 was 23.5 ± 5.8 μSv/MBq (28). The pharmacokinetics and biodistribution of [¹⁸F]DASA-23 in a healthy volunteer (at 0.5, 1, 2, and 3 hours after administration) are shown in Fig. 3A. The dose-limiting organ was determined to be the gallbladder, which received an absorbed dose of 0.61 ± 0.52 mSv/MBq (28). We evaluated brain uptake patterns of [¹⁸F]DASA-23 over the course of a 60-minute brain acquisition beginning with intravenous administration of [¹⁸F]DASA-23 (Fig. 3B). The tracer exhibited high initial localization (peak SUV ∼ 5) to most brain structures, including cerebral cortex (Fig. 3C) and the posterior fossa (Fig. 3D), followed by rapid washout over the 60-minute acquisition period. This was in contrast to white matter, which was characterized by lower delivery of the tracer (peak SUV ∼ 2) and slower washout (Fig. 3E).

We intravenously administered approximately 300 MBq [¹⁸F]DASA-23 to 2 female and 2 male patients diagnosed with malignant gliomas, and completed a 60-minute dynamic brain acquisition using simultaneous PET/MRI (Supplementary Tables S2–S6). There were no adverse clinical reactions to [¹⁸F]DASA-23 injections. Representative decay-corrected time–activity curves for the tumor, centrum semiovale (white matter), and the contralateral frontal and parietal cortices [gray matter (GM)] were averaged for patients undergoing [¹⁸F]DASA-23 PET imaging (Supplementary Fig. S4). We used this information to select the timeframe for obtaining a suitable ratio of tumor uptake relative to normal brain tissue. The optimal timeframe was determined to be 30 to 60 minutes after radiotracer administration, with the criterion that any tracer accumulation above background was considered abnormal. We used white matter at the level of the contralateral centrum semiovale as the reference region for TBR calculations. The corresponding SUV, SUV_max, TBR, TBR_max, and tumor-to-gray matter (T/GM and T/GM_max) values are summarized in Table 1.

The first patient (IC-1) was diagnosed with recurrent GBM, having IDH wild-type and methylated MGMT promoter status. Contrast-enhanced MRI revealed a large enhancing tumor within the left frontal lobe (Fig. 4A). Dosages of 337 MBq [¹⁸F]FDG and 300 MBq [¹⁸F]DASA-23 were administered to the patient on separate days. [¹⁸F]FDG PET/CT revealed subtle although asymmetric [¹⁸F]FDG uptake corresponding to the region of MRI enhancement (Fig. 4B; Table 1). [¹⁸F]DASA-23 clearly delineated this lesion within the left frontal lobe with a TBR_max of 2.6 (Fig. 4C; Table 1). The patient then began treatment with bevacizumab (7.5 mg/kg) and temozolomide chemotherapy (150 mg/m²), and we then obtained a second [¹⁸F]DASA-23 PET/MRI 6 days later to detect early metabolic treatment response (Fig. 4D and E). We compared that result with the outcome noted at 3 months (Fig. 4F). In the posttherapy [¹⁸F]DASA-23 PET, we observed an increased TBR_max of 2.9 (Fig. 4D and G; Table 1). We observed a discordance between [¹⁸F]DASA-23 PET and MRI, the latter of which demonstrated a reduction in contrast-enhancing tumor volume after administration of bevacizumab and temozolomide. We also noted that the highest levels of [¹⁸F]DASA-23 tracer uptake in both pre- and post-therapy scans were observed in the posterior medial aspect of the left frontal mass, which did not exhibit substantial contrast enhancement on both the pre- and post-therapy MRI scans (Supplementary Fig. S5). A standard-of-care MRI 3 months after [¹⁸F]DASA-23 PET revealed progressive disease within the areas of [¹⁸F]DASA-23 uptake at the medial and lateral posterior portions of the GBM (Fig. 4F), and the patient subsequently underwent tumor re-resection. H&E staining and IHC of the resected tissue confirmed the expression of PKM2 in the cancer cells shown in the adjacent tissue sections (Fig. 4H and I). Low levels of PKM1 were detected within the cancer cells (Supplementary Fig. S5). Western blot analysis additionally confirmed the presence of PKM2 within the resected tissue compared with a nonmatched normal brain sample and U87 cell lysate (Supplementary Fig. S6). We developed a primary patient-derived cell line (TP459) from IC-1′s tumor tissue, maintained it in culture as neurospheres, and subsequently confirmed the cellular uptake of [¹⁸F]DASA-23 in the ex vivo–cultured cells (Fig. 4J). We also subjected TP459 cells to the IC₅₀ dosage of temozolomide for 72 hours (Supplementary Fig. S7A), and then assessed uptake of [¹⁸F]DASA-23 (Fig. 4J). There was a nonsignificant difference between [¹⁸F]DASA-23 uptake in untreated TP459 cells and temozolomide-treated TP459 cells (29.5% ± 2.3% uptake/mg protein vs. 26.1% ± 1.9% uptake/mg protein, P = 0.12) from this patient who was treated with temozolomide and did not respond. We confirmed via Western blot analysis that there was a nonsignificant difference in PKM2 protein expression between untreated and temozolomide-treated TP459 cells (Supplementary Fig. S7B and S7C).

Intracranial patient 2 (IC-2) was diagnosed with GBM, having IDH1 wild-type and unmethylated MGMT promoter status. Contrast-enhanced MRI revealed a heterogeneous, enhancing GBM centered in the left body of the corpus callosum, with involvement of the left frontal lobe and corona radiata, with contralateral extension shown in the coronal image in Fig. 5A. This patient was receiving bevacizumab therapy at the time of [¹⁸F]DASA-23 PET/MRI, an infusion of 7.5 mg/kg was completed 14 days prior to the [¹⁸F]DASA-23 PET/MRI. [¹⁸F]DASA-23 targeted and outlined this tumor (Fig. 5B), with homogeneous uptake of the tracer observed within the lesion compared with the heterogeneous MRI contrast enhancement (Fig. 5C). [¹⁸F]DASA-23 PET revealed high levels of tracer avidity and metabolic activity within this tumor, with TBR_max of 2.8 (Table 1; Supplementary Fig. S8). H&E staining and IHC analysis of the biopsied tissue confirmed that the GBM cells expressed PKM2 (Fig. 5D–G). Low levels of PKM1 were found within the lesion (Supplementary Fig. S8).

We also administered [¹⁸F]DASA-23 to one patient diagnosed with WHO grade IV diffuse midline glioma, H3 K27M–mutant (IC-3). At the time of enrollment, this patient had recently completed standard-of-care chemoradiation, and MRI revealed a peripherally enhancing, centrally necrotic mass in the right thalamus thought to represent a mixture of treatment-associated effects and residual tumor (Fig. 6A). [¹⁸F]DASA-23 PET revealed two focal areas of radiotracer uptake (Fig. 6B and C), with TBR_max of 2.9 (Table 1). SUV distribution within the lesion and the SUV histogram are shown in Supplementary Fig. S9. H&E staining (Fig. 6D) and PKM2 IHC (Fig. 6E) confirmed the expression of PKM2 within the malignant biopsied tissue (also shown in Supplementary Fig. S9). Finally, we explored the ability of [¹⁸F]DASA-23 to delineate a lower grade lesion, and administered [¹⁸F]DASA-23 to one patient with WHO grade III recurrent anaplastic astrocytoma (IC-4). IC-4 presented with a fluid-filled cyst and surrounding patchy contrast enhancement within the left frontal lobe (Fig. 6F). The uptake of [¹⁸F]DASA-23 was evident in this lesion, with a TBR_max of 2.4 (Table 1). [¹⁸F]DASA-23 uptake was heterogeneous within the lesion and the SUV distribution of the lesion is shown in Supplementary Fig. S10. H&E staining (Fig. 6I) and PKM2 IHC (Fig. 6J) confirmed the presence of PKM2 within the resected tumor tissue (Supplementary Fig. S10); the expression of PKM1 was also detected within this lesion.

PKM2-specific imaging agents are sought after owing to their potential impact on clinical decision making and therapeutic monitoring in the setting of brain malignancies. PKM2 is overexpressed in many cancer types, including GBM, and is found in varying amounts in most body tissues with the exception of adult muscle, liver, and brain. Given the absence of PKM2 in healthy brain, we set out to develop a clinically translatable radiotracer for noninvasive assessment of PKM2 status in GBM. Here, we developed and translated [¹⁸F]DASA-23, to measure aberrantly expressed PKM2 in GBM cells in culture, animal models, and patients. This is the first study to evaluate a PKM2-specific PET tracer in humans.

In cell culture, [¹⁸F]DASA-23 showed rapid and extensive uptake in all human GBM cell lines evaluated, including commercial cell lines (U87 and LN18) and patient-derived lines (GBM206 and TP459), with no variation in relation to IDH mutation and MGMT promoter methylation status (22). We have previously demonstrated the in vitro specificity of [¹⁸F]DASA-23 through competition studies with TEPP-46 and by modulation of PKM2 expression with PKM2-specific siRNA (24, 31). Using mouse models, we highlight the ability of [¹⁸F]DASA-23 to target U87-GFP/luc GBMs compared with [¹⁸F]FDOPA, an established amino acid PET tracer explored in the clinical management of patients with GBM (30, 32). We emphasize that the goal of this study was not to develop a comparatively better tracer for brain cancer diagnosis, but rather to develop a novel tracer that can provide new and actionable information about the glycolytic state of brain cancer, while maintaining a TBR similar to established tracers. In this article, we confirmed the in vivo specificity of [¹⁸F]DASA-23 with PET imaging and autoradiography under blocking conditions with TEPP-46. These results build on our previous validation of [¹⁸F]DASA-23, which demonstrated that mice stereotactically injected with AAV9-PKM2 showed a position correlation between PKM2 mRNA levels and [¹⁸F]DASA-23 PET uptake (33). We further confirmed specificity with the isotopolog, [¹¹C]DASA-23, by completing PET imaging studies in mice bearing orthotopic GBM under baseline and blocking conditions (23).

We successfully translated [¹⁸F]DASA-23 for clinical applications under investigational new drug status. The estimated effective dose of [¹⁸F]DASA-23 (23.5 ± 5.8 μSv/MBq) is comparable to other fluorine-18 radiotracers evaluated in patients with GBM including [¹⁸F]FPPRGD2 (39.6 ± 18.1 μSv/MBq; ref. 34) and [¹⁸F]-fluoro-ethyl-tyrosine (16.5 μSv/MBq; ref. 35). An injected dose of 300 MBq exposes the patient to 7.1 mSv, which is below the whole-body dose limit of 30 mSv (single dose) specified by the FDA for research subjects. Radiometabolite analysis in healthy volunteers demonstrated that [¹⁸F]DASA-23 remained intact in human plasma up to 5 minutes post intravenous administration and then underwent some degradation at the 10- and 30-minute sampling time points (28). Although the identity of the radiometabolite found in the 10- and 30-minute samples is unknown, radio-high-performance liquid chromatography analysis signifies that this is likely a smaller, polar fragment and implies it will not be able to passively cross the BBB. We noted higher background [¹⁸F]DASA-23 signal within the cerebral white matter relative to gray matter, observed in both healthy volunteers and patients. There is no evidence of expression of PKM2 within myelinated axons (36–38), suggesting that the uptake of [¹⁸F]DASA-23 is likely nonspecific and potentially may be related to the lipophilicity and highly planar structure of the DASA-23 molecule.

We note that the uptake profile of [¹⁸F]DASA-23 by the GBM in patient IC-1 is different than the uptake profile of [¹⁸F]FDG by the same tumor (Fig. 4). The high background uptake of [¹⁸F]FDG within the cerebral cortex makes it difficult to accurately visualize the boundaries of the tumor. In contrast, [¹⁸F]DASA-23 delineates this lesion with high signal to background, and the uptake pattern of [¹⁸F]DASA-23 is also different to that of [¹⁸F]FDG. We observe the highest levels of [¹⁸F]DASA-23 uptake at the posterior medial aspect of the tumor that additionally did not show substantial MRI contrast enhancement, highlighting the ability of [¹⁸F]DASA-23 to cross the BBB and to provide information regarding the glycolytic status of the tumor independent of BBB integrity. We additionally demonstrate the lack of a metabolic response in IC-1 using [¹⁸F]DASA-23 PET within one week of initiating temozolomide and bevacizumab therapy. This was in comparison to contrast-enhanced MRI, which revealed a reduction in tumor volume after bevacizumab therapy, a phenomenon referred to as pseudoresponse (39, 40). The pattern of enhancement has been used as a surrogate for tumor grade and viability; however, processes that impact the permeability of the BBB will modify the degree and extent of enhancement, irrespective of the size and activity of the tumor (39, 41, 42). Bevacizumab is known to stabilize tumor vasculature and decrease BBB permeability, resulting in a reduction of tumor enhancement and pseudoresponse effect. Tumor progression was confirmed on a standard-of-care contrast-enhanced MRI 3 months later, highlighting the potential of PET imaging with [¹⁸F]DASA-23 to identify a metabolic nonresponder early on.

In patient IC-2, we observe the clear delineation of the GBM in the corpus callosum with contralateral extension on [¹⁸F]DASA-23 PET. This tumor definition was more conspicuous compared with the heterogeneous T1-weighted MRI-based contrast enhancement. IC-2 was receiving bevacizumab at the time of [¹⁸F]DASA-23 PET/MRI, which may account for the heterogeneous enhancement observed in IC-2. For all other patients evaluated in this pilot study, the uptake of [¹⁸F]DASA-23 in the intracranial tumors was evident and provided functional information regarding PKM2 status within these lesions. In patient IC-3, we observe the ability of [¹⁸F]DASA-23 PET to visualize aberrant PKM2 expression and metabolic activity that can potentially aid in distinguishing viable tumor as opposed to treatment effects. Finally, we highlight the ability of [¹⁸F]DASA-23 to visualize aberrant PKM2 in anaplastic astrocytoma as seen in IC-4. Although the SUV in IC-4 was slightly higher than IC-3, we note that various factors can impact radiotracer uptake such as perfusion and necrosis, and this will be further investigated in future studies.

We developed this imaging strategy as a critical component of ongoing efforts to establish successful therapies for GBM. One important aspect of such efforts is to identify a reliable molecular imaging method to assess response to therapy. We previously reported the development of [¹⁸F]FPPRG2 for evaluation of the integrin α_vβ₃ in GBM patients before and after bevacizumab therapy (43). Here, we describe the development of [¹⁸F]DASA-23 to assess aberrant glycolysis in GBM. The effect of PKM2 activity in glycolysis and the proliferation of cancer cells has been shown to balance the production of biomolecular building blocks and the generation of pyruvate and ATP (17). Given the important role of aberrant metabolism in GBM growth and infiltration (44, 45), it is possible that most treatments for GBM would likely benefit from a general metabolic biomarker for treatment monitoring. The novelty of our ability to track PKM2 activity, which is a key regulator of GBM metabolism, lies in allowing us to potentially monitor different classes of therapies for metabolic response. We previously reported the ability of [¹⁸F]DASA-23 to detect glycolytic changes in GBM cells in culture in response to multiple classes of antineoplastic agents (25). As an example of clinical translation of this concept, we highlight in this study the ability of [¹⁸F]DASA-23 to identify IC-1 as a metabolic nonresponder within one week of temozolomide and bevacizumab therapy. We additionally confirm the ability of [¹⁸F]DASA-23 to detect metabolic changes within IC-1 patient-derived recurrent GBM TP459 neurospheres in culture. That the [¹⁸F]DASA-23 uptake in these neurospheres was comparable between control and temozolomide IC₅₀ conditions, may not be surprising given that the patient's GBM progressed on a temozolomide-containing treatment regimen. Indeed, Park and colleagues had previously reported that temozolomide-treated GBMs have reduced PKM2 expression (46).

We acknowledge some limitations, including the small number of patients evaluated to date in this pilot study, and the participation of patients at different stages of disease progression. We note the lack of certain parameters relating to DASA-23, including K_D (equilibrium dissociation constant) and B_max (total number of receptors in a given tissue), which are currently under investigation. However, it should be noted that the primary aim of this study was the development and translation of a PKM2-specific radiopharmaceutical for evaluation of PKM2 status in GBMs. Future studies will evaluate the diagnostic accuracy of [¹⁸F]DASA-23 in larger patient cohorts that include newly diagnosed and treatment-naïve patients.

In conclusion, we successfully developed and clinically translated [¹⁸F]DASA-23 as a PKM2-specific radiotracer for noninvasive visualization of PKM2 status in patients with brain cancer. [¹⁸F]DASA-23 can be prepared with high radiochemical yield and molar activity suitable for clinical imaging, and is well tolerated by human subjects as determined in these initial studies. Of note, we observed significant binding of [¹⁸F]DASA-23 in brain tumors, with low background in healthy brain tissue. We additionally highlight the potential ability of [¹⁸F]DASA-23 to identify early metabolic nonresponse to conventional chemotherapy and bevacizumab. Taken together, our results represent the first example of PKM2 visualization in human subjects. A comprehensive evaluation of [¹⁸F]DASA-23 in larger patient populations is anticipated.

C. Beinat reports grants from GE Healthcare and from Ben and Catherine Ivy Foundation during the conduct of the study, as well as grants from Stanford SPARK and NIH outside the submitted work; in addition, C. Beinat has a patent for US-2018–0043040-A1 issued and a patent for US-2021–0199640-A1 issued. C.B. Patel reports grants, personal fees, and non-financial support from Novocure, Ltd., as well as grants from Aveta Biomics, Inc. outside the submitted work; in addition, C.B. Patel has a patent for Using Alternating Electric Fields to Increase Cell Membrane Permeability pending to Novocure, Ltd. and a patent for Methods of Normalizing Aberrant Glycolytic Metabolism in Cancer Cells pending to Novocure, Ltd. I.S. Alam reports a patent for Imaging Tumor Glycolysis by Non-invasive Measurement of Pyruvate Kinase M2 (US-2018–0043040-A1) issued. K. Halbert reports grants from GE Healthcare during the conduct of the study. I. Weissman was a founder and director at SyStemix, Stem Cells, Inc., Cellerant, and Forty Seven Inc.; in addition, I. Weissman is founder and director of companies not yet established: Bitterroot Bio and Pheast, Inc. M. Khalighi reports grants from GE Healthcare during the conduct of the study. A. Iagaru reports grants from GE Healthcare and Advanced Accelerator Applications, as well as non-financial support from Progenics Pharmaceuticals outside the submitted work. R. Thomas reports other support from Oncoceutics and Bristol Myers Squibb, as well as grants from Health Resources and Services Administration, NIH outside the submitted work. S.S. Gambhir reports grants from GE Healthcare and from Ben and Catherine Ivy Foundation during the conduct of the study; in addition, S.S. Gambhir has a patent for US-2018–0043040-A1 issued and a patent for US-2021–0199640-A1 issued. No disclosures were reported by the other authors.

C. Beinat: Conceptualization, data curation, formal analysis, investigation, methodology, writing–original draft, writing–review and editing. C.B. Patel: Conceptualization, data curation, formal analysis, investigation, methodology, writing–original draft, writing–review and editing. T. Haywood: Data curation, investigation, methodology, writing–review and editing. S. Murty: Data curation, writing–review and editing. L. Naya: Data curation, validation, project administration, writing–review and editing. J.B. Castillo: Data curation, methodology, project administration, writing–review and editing. S.T. Reyes: Data curation, writing–review and editing. M. Phillips: Data curation, writing–review and editing. P. Buccino: Data curation, writing–review and editing. B. Shen: Data curation, supervision, methodology, project administration, writing–review and editing. J.H. Park: Data curation, methodology, writing–review and editing. M.E.I. Koran: Formal analysis, visualization, writing–review and editing. I.S. Alam: Data curation, methodology, writing–review and editing. M.L. James: Data curation, methodology, writing–review and editing. D. Holley: Investigation, methodology, writing–review and editing. K. Halbert: Investigation, methodology, writing–review and editing. H. Gandhi: Investigation, methodology, writing–review and editing. J.Q. He: Methodology, writing–review and editing. M. Granucci: Data curation, project administration, writing–review and editing. E. Johnson: Data curation, project administration, writing–review and editing. D.D. Liu: Methodology, writing–review and editing. N. Uchida: Methodology, writing–review and editing. R. Sinha: Methodology, writing–review and editing. P. Chu: Data curation, methodology, writing–review and editing. D.E. Born: Formal analysis, investigation, writing–review and editing. G.I. Warnock: Formal analysis, writing–review and editing. I. Weissman: Supervision, methodology, writing–review and editing. M. Hayden-Gephart: Data curation, investigation, writing–review and editing. M. Khalighi: Formal analysis, supervision, writing–review and editing. T.F. Massoud: Formal analysis, writing–original draft, writing–review and editing. A. Iagaru: Formal analysis, supervision, investigation, writing–review and editing. G. Davidzon: Formal analysis, supervision, investigation, writing–review and editing. R. Thomas: Formal analysis, supervision, investigation, writing–review and editing. S. Nagpal: Formal analysis, supervision, investigation, writing–review and editing. L.D. Recht: Supervision, investigation, writing–review and editing. S.S. Gambhir: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, investigation, writing–original draft, writing–review and editing.

We thank the Cyclotron and Radiochemistry Facility at Stanford University for the ¹⁸F production, in particular Dr. Fred Chin and Carmen Azevedo. We thank the Small Animal Imaging facility at Stanford, in particular Drs. Timothy Doyle, Frezghi Habte, and Laura Pisani. We thank Dr. Yuanyang Xie for helpful discussions about this study. This work was supported by the Ben and Catherine Ivy Foundation (to S.S. Gambhir), GE Healthcare (to S.S. Gambhir), and R01 CA216054–01 (to M. Hayden-Gephart). C. Beinat is grateful for support from Stanford School of Medicine Translational Research and Applied Medicine Fellowship and C.B. Patel acknowledges receipt of the Stanford Cancer Institute Fellowship Award for Cancer Research and NIH NINDS Research Education Grant (5R25NS065741–07). J.Q. He acknowledges support from the NIH NCI (F30CA228215). We dedicate this manuscript to the loving memory of our mentor, Dr. Sanjiv Sam Gambhir.

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.