Evaluation of Glycolytic Response to Multiple Classes of Anti-glioblastoma Drugs by Noninvasive Measurement of Pyruvate Kinase M2 Using [¹⁸F]DASA-23

Abstract

Purpose:

Pyruvate kinase M2 (PKM2) catalyzes the final step in glycolysis, the key process of tumor metabolism. PKM2 is found in high levels in glioblastoma (GBM) cells with marginal expression within healthy brain tissue, rendering it a key biomarker of GBM metabolic reprogramming. Our group has reported the development of a novel radiotracer, 1-((2-fluoro-6-[¹⁸F]fluorophenyl)sulfonyl)-4-((4-methoxyphenyl)sulfonyl)piperazine ([¹⁸F]DASA-23), to non-invasively detect PKM2 levels with positron emission tomography (PET).

Procedure:

U87 human GBM cells were treated with the IC50 concentration of various agents used in the treatment of GBM, including alkylating agents (temozolomide, carmustine, lomustine, procarbazine), inhibitor of topoisomerase I (irinotecan), vascular endothelial and epidermal growth factor receptor inhibitors (cediranib and erlotinib, respectively) anti-metabolite (5-fluorouracil), microtubule inhibitor (vincristine), and metabolic agents (dichloroacetate and IDH1 inhibitor ivosidenib). Following drug exposure for three or 6 days (n = 6 replicates per condition), the radiotracer uptake of [¹⁸F]DASA-23 and 2-deoxy-2-[¹⁸F]fluoro-D-glucose ([¹⁸F]FDG) was assessed. Changes in PKM2 protein levels were determined via Western blot and correlated to radiotracer uptake.

Results:

Significant interactions were found between the treatment agent (n = 12 conditions total comprised 11 drugs and vehicle) and the duration of treatment (3- or 6-day exposure to each drug) on the cellular uptake of [¹⁸F]DASA-23 (p = 0.0001). The greatest change in the cellular uptake of [¹⁸F]DASA-23 was found after exposure to alkylating agents (p < 0. 0001) followed by irinotecan (p = 0. 0012), erlotinib (p = 0. 02), and 5-fluorouracil (p = 0. 005). Correlation of PKM2 protein levels and [¹⁸F]DASA-23 cellular uptake revealed a moderate correlation (r = 0.44, p = 0.15).

Conclusions:

These proof of principle studies emphasize the superiority of [¹⁸F]DASA-23 to [¹⁸F]FDG in detecting the glycolytic response of GBM to multiple classes of anti-neoplastic drugs in cell culture. A clinical trial evaluating the diagnostic utility of [¹⁸F]DASA-23 PET in GBM patients (NCT03539731) is ongoing.

Introduction

Glioblastoma (GBM) is the most common and lethal primary central nervous system (CNS) tumor [1–3]. Despite aggressive surgical resection, radiotherapy, and chemotherapy, prognosis remains dismal with overall survival of only ~ 5 % at 5 years [3, 4]. The recurrence of GBM is inevitable and its management often individualized based on pathological and clinical characteristics. Following failure of initial treatment, therapeutic options for recurrent GBM are diverse, with no standard identified [5]. Common agents administered in the treatment of newly diagnosed and recurrent GBM include temozolomide, nitrosoureas (carmustine and lomustine), anti-angiogenic agents (bevacizumab and cediranib), epidermal growth factor receptor (EGFR) inhibitors (erlotinib), topoisomerase I inhibitors (irinotecan), and others such as vincristine [5]. Metabolic therapies (dichloroacetate and the IDH1 inhibitor ivosidenib) have been studied in clinical trials of patients with recurrent GBM and have demonstrated safety and preliminary signs of efficacy [6, 7].

Great difficulties exist in the neuroimaging evaluation of patients undergoing treatment for malignant glioma [8]. Differentiation of treatment response and tumor progression is problematic and combines follow-up magnetic resonance imaging (MRI), with clinical status and corticosteroid-dependency assessments [9–12]. Response and progression-free endpoints depend on MRI and are complicated by lack of reliable interpretation of treatment-related radiography changes, including pseudoprogression secondary to radiation and chemotherapy and pseudoresponse due to antiangiogenic therapy [8–10, 13, 14]. Positron emission tomography (PET) is a sensitive, quantitative technology ideal for the molecular imaging of cancer biology [15, 16]. Early tumor metabolic changes can be detected with 2-deoxy-2-[¹⁸F]fluoro-D-glucose ([¹⁸F]FDG). This radiotracer has been extensively utilized in oncology for assessing response to therapy [17]; however, the role of [¹⁸F]FDG PET is limited in brain tumors due to the relatively high background signal from glucose metabolism in the normal brain [18].

There is a need for neuroimaging biomarkers that provide a rapid and accurate measurement of aberrant glycolysis in GBM. To address this, we have reported the development of 1-((2-fluoro-6-[¹⁸F]fluorophenyl)sulfonyl)-4-((4-methoxyphenyl)sulfonyl)piperazine ([¹⁸F]DASA-23) for the non-invasive measurement of pyruvate kinase M2 (PKM2) [19]. Pyruvate kinase (PK) catalyzes the final step in glycolysis, converting phosphoenolpyruvate (PEP) to pyruvate, yielding ATP and lactate. Reduced PK activity results in a diminished production of pyruvate, enabling accumulation of upstream glycolytic intermediates and shifting metabolism towards an anabolic state [20]. Cancer cells exploit this by utilizing the PKM2 isoform of PK, whose activity can be regulated between the less active PKM2 dimer and the highly active PKM2 tetramer [20, 21]. [¹⁸F]DASA-23 is a PKM2 activator, reported to bind a pocket at the dimeric PKM2 subunit interface and promote association of PKM2 subunits into stable tetramers [22]. However, the sub-pharmacological amounts of [¹⁸F]DASA-23 present in radiotracer cellular uptake studies unlikely have any effect on the enzymatic state of PKM2 [23]. Molecular imaging with [¹⁸F]DASA-23 therefore provides a non-invasive measurement of PKM2 dimer levels (Fig. 1). The relatively high lipophilicity of DASA-23 (cLogP = 3.4, ChemDraw 14.0) allows it to diffuse across the blood-brain barrier and cell membrane; this was previously confirmed in animal models when DASA-23 was radiolabeled with carbon-11 ([¹¹C]DASA-23) [24]. In the present work, we have radiolabeled DASA-23 with fluorine-18 ([¹⁸F]DASA-23) due to its longer half-life and improved properties for clinical application and widespread use. This was achieved without altering the native structure of DASA-23 (which includes fluorine) and thereby preserving its biological activity and properties, including blood-brain barrier penetration and low uptake in healthy brain tissue. In this study, we evaluated the ability of [¹⁸F]DASA-23 and [¹⁸F]FDG to detect GBM metabolic changes in response to seven classes of clinically relevant anti-neoplastic agents (n = 11 drugs total) in human U87 GBM cells in culture. The aim of this study was to validate [¹⁸F]DASA-23 as an imaging agent that provides elevated cellular uptake in untreated GBM cells relative to [¹⁸F]FDG and following treatment has a significantly reduced cellular uptake. Given the minimal levels of PKM2 in the healthy brain, this would enable high levels of [¹⁸F]DASA-23 radiotracer uptake within GBM tumors with low uptake in the healthy brain tissue. The elevated uptake and high signal-to-background of [¹⁸F]DASA-23 would therefore facilitate the ability to detect subtle changes in response to therapy compared to [¹⁸F]FDG.

Fig. 1.

Pyruvate kinase (PK) catalyzes the final step of glycolysis, resulting in net ATP synthesis through dephosphorylation of PEP. The PKM2 isoform predominates in tumor cells. Two quaternary PKM2 conformations exist as homo-dimeric or - tetrameric forms (shown as two and four blue spheres, respectively). Dimeric PKM2 has a reduced affinity for PEP in comparison to the tetramer, with tumor PKM2 mainly present in the dimeric form. Reduced conversion of PEP to pyruvate allows buildup of glycolytic precursors for the biosynthesis of macromolecules. Tetrameric PKM2 predominates in non-tumor cells. [¹⁸F]DASA-23 is a PKM2 activator, reported to bind a pocket at the dimeric PKM2 subunit interface that promotes association of PKM2 subunits into stable tetramers, due to the conformational change associated with tetramer formation, the binding pocket is not available on constituently active tetramers.

Materials and Methods

General

U87 human glioblastoma cells were obtained from ATCC (Manassas, VI) and maintained per the manufacturer’s instructions. Chemicals were purchased from Med Chem Express (Monmouth Junction, NJ) and Selleck Chem (Houston, TX).

Radiochemistry

[¹⁸F]DASA-23 was produced according to a reported procedure [19] and obtained in >95 % chemical and radiochemical purity and specific activity of 3029 ± 733 mCi/μmol. [¹⁸F]FDG was obtained from routine production at the Stanford University Cyclotron and Radiochemistry Facility.

IC₅₀ Determination in U87 Cells

The IC₅₀ of GBM drugs were determined using a 72-h incubation period in U87 cells [25]. U87 cells were plated in a 96-well plate (5 × 10³ cells per well, n = 3 per condition) and exposed to serial dilutions of appropriate drug stock (1–0.1 M in dimethyl sulfoxide (DMSO)) with a resulting dose range from 33 mM to 1 μM. DMSO vehicle was used for control wells. After 72 h, the Presto Blue® Viability Assay (Thermo Fisher Scientific Inc., Waltham, MA) was performed.

Drug Treatment and Radiotracer Uptake Studies

U87 cells were plated into 12-well plates (2 × 10⁴ per well for 6-day studies, 5 × 10⁴ per well for 3-day studies, n = 6 per condition) and the appropriate drug (n = 11) added to provide the IC₅₀ concentration in each well, the amount of DMSO in each well was less than 0.5 %. Cells were exposed to each drug for either 3 or 6 days. These treatment periods were selected to encompass 1 and ≥ 2 doubling cycles, respectively [25, 26]. After 3 or 6 days of exposure, the cellular uptake of [¹⁸F]DASA-23 and [¹⁸F]FDG was evaluated. For uptake studies, media was used for [¹⁸F]DASA-23 experiments and HBSS used for experiments involving [¹⁸F]FDG. Prior to addition of radioactivity, the media from each well was removed and cells washed with PBS (1 ml) in order to remove any dead cells from the drug treatment and ensure that changes in radiotracer uptake were reflective of metabolic changes within remaining living cells. A total of 5 μCi of the appropriate radiotracer in either 1 ml of pre-warmed media or HBSS was added per well. Cells were incubated in the presence of radiotracer for 30 min at 37 °C and 5 % CO₂. At 30 min post-addition of radiotracer, incubation was stopped by removing the radioactive solution and placing the plates on ice.

Wells were washed with ice-cold PBS (3 × 1 ml) and treated with radioimmunoprecipitation assay (RIPA) buffer (Thermo Fisher Scientific Inc.; 200 μl) supplemented with Halt Protease inhibitor cocktail (Thermo Fisher Scientific Inc.). A fraction of the corresponding cell lysate (150 μl) was used to evaluate the amount of decay-corrected radioactivity with a gamma counter (Cobra II Auto-Gamma Counter; Packard Biosciences Co.). Post-radioactivity decay, the cell lysate was evaluated for total protein concentration using a bicinchoninic acid (BCA) assay (Thermo Fisher Scientific Inc.) and then frozen at − 80 °C for subsequent Western blot analysis. Radiotracer uptake values were determined as the percentage uptake of total radiotracer added and normalized to total protein content of remaining living cells in each well.

Western Blotting

Cellular lysates from uptake studies were evaluated for PKM2 protein expression. PKM2 monoclonal rabbit antibody (1:3000 dilution, Cell Signaling Technology, Cambridge, MA) was used in a standard Western blotting protocol according to the manufacturer’s instructions. Rabbit anti-GAPDH antibody (1:5000 dilution, Cell Signaling Technology) was used as loading control. Anti-rabbit IgG, HRP-linked antibody (1:5000 dilution, Cell Signaling Technology) was used as a secondary antibody. Chemiluminescent signals from PKM2 were normalized to that of GAPDH to account for the amount of protein loaded for each sample.

Statistical Analysis

Results are reported as mean ± SD with n = 6 per drug/duration/radiotracer condition unless otherwise specified. For comparison between drug treatment and treatment duration on radiotracer uptake, two-way analysis of variance (ANOVA) was used with Bonferroni correction for multiple comparisons. For comparison between drug conditions and [¹⁸F]DASA-23 radiotracer uptake, one-way ANOVA was used with Bonferroni correction for multiple comparisons. A Student’s t test was used to compare radiotracer uptake values between [¹⁸F]DASA-23 and [¹⁸F]FDG. Association between normalized [¹⁸F]DASA-23 radiotracer uptake values and normalized PKM2 protein expression values was analyzed using Pearson’s correlation.

Results

Alkylating Agent Radiotracer Uptake Studies

Radiotracer uptake of [¹⁸F]FDG and [¹⁸F]DASA-23 was evaluated in cells that had been previously exposed for 3 or 6 days to vehicle (untreated) or the IC₅₀ of temozolomide, procarbazine, carmustine, or lomustine. Uptake of [¹⁸F]DASA-23 in U87 untreated cells was significantly greater than that of [¹⁸F]FDG (25.2 ± 4.0 vs 2.5 ± 0.3 % uptake/mg protein at 3 days, p < 0.0001, and 29.4 ± 2.2 vs 1.8 ± 0.3 % uptake/mg protein at 6 days, p < 0.0001) (Fig. 2a).

Fig. 2.

[¹⁸F]DASA-23 and [¹⁸F]FDG cellular uptake in untreated cells and cells treated with alkylating agents temozolomide, procarbazine, carmustine, and lomustine. a Cellular uptake of [¹⁸F]DASA-23 and [¹⁸F]FDG in U87-untreated cells, ****p < 0.0001. b Three-day treatment; untreated vs temozolomide, ***p = 0.0004; untreated vs carmustine, ****p < 0.0001; untreated vs lomustine, **p = 0.002; untreated vs procarbazine, ****p < 0.0001; n = 6 per condition, p values obtained from one-way ANOVA with Bonferroni correction for multiple comparisons. # = significant interaction on [¹⁸F]FDG uptake between the alkylating agents and duration of treatment, two-way ANOVA, 6-day data is shown in Fig. 2b. c Six-day treatment; untreated vs temozolomide, ****p < 0.0001; untreated vs carmustine, ****i < 0.0001; untreated vs lomustine, ****p < 0.0001; untreated vs procarbazine, ****p < 0.0001; n = 6 per condition, p values obtained from one-way ANOVA with Bonferroni correction for multiple comparisons. # = significant interaction on [¹⁸F]FDG cellular uptake between the alkylating agent and length of treatment duration, two-way ANOVA, 3-day data is shown in Fig. 2a. Combined 3- and 6-day data is shown for [¹⁸F]DASA-23 and [¹⁸F]FDG in Suppl. Fig. 4a and 5a, respectively. d Percentage reduction of [¹⁸F]DASA-23 and [¹⁸F]FDG radiotracer uptake relative to untreated cells in the presence of alkylating agents for 3-day treatment, carmustine **p = 0.007, procarbazine *p = 0.03. e Percentage reduction of [¹⁸F]DASA-23 and [¹⁸F]FDG radiotracer uptake relative to untreated cells in the presence of alkylating agents for 6-day treatment, temozolomide ****p < 0.0001, carmustine ****p < 0.0001, lomustine ****p < 0.0001, procarbazine **p = 0.0007.

A significant interaction was evident between the type of alkylating agent and treatment duration as measured by changes in [¹⁸F]FDG radiotracer uptake (two-way ANOVA, p = 0.0004, Suppl. Fig. S5a; see Electronic Supplementary Material (ESM)). Similarly, a significant interaction was evident between the alkylating agent and treatment duration as measured by changes in [¹⁸F]DASA-23 radiotracer uptake (two-way ANOVA, p < 0.0001, Suppl. Fig. S4a (ESM)). Further analysis into [¹⁸F]DASA-23 uptake at 3 days posttreatment with alkylating agents vs the untreated condition revealed significantly decreased uptake with treatment (one-way ANOVA, p < 0.0001, F = 17.1, Fig. 2b); the greatest reduction in radiotracer uptake relative to untreated cells was evident in the carmustine group (25.2 ± 4.0 vs 12.3 ± 1.7 % uptake/mg protein, p < 0.0001, 51.2 ± 6.8 % reduction). This was followed by procarbazine-treated cells (15.4 ± 2.9 % uptake/mg protein, p < 0.0001, 39.1 ± 11.4 % reduction), temozolomide-treated cells (18.0 ± 2.8 % uptake/mg protein, p = 0.0004, 28.7 ± 11.1 % reduction), and lomustine-treated cells (18.3 ± 2.2 % uptake/mg protein, p = 0.002, 27.1 ± 8.7 % reduction) (Fig. 2b, d).

After 6-day treatment, additional decreases were observed between [¹⁸F]DASA-23 uptake in treated cells compared to untreated cells (one-way ANOVA, p < 0.0001, F = 123.3, Fig. 2c). However, in contrast to the 3-day treatment findings, the greatest reduction in [¹⁸F]DASA-23 uptake compared to untreated cells was observed in the procarbazine group (29.4 ± 2.2 vs 7.3 ± 0.3 % uptake/mg protein, p < 0.0001, 75.0 ± 1.1 % reduction) and temozolomide (8.0 ± 1.5 % uptake/mg protein, p < 0.0001, 72.7 ± 5.0 % reduction), followed by carmustine whose uptake remained largely unchanged from 3-day treatment value (12.6 ± 1.2 % uptake/mg protein, p < 0.0001, 57.2 ± 4.2 % reduction) and lomustine (14.5 ± 2.7 % uptake/mg protein, p < 0.0001, 50.7 ± 9.2 % reduction) (Fig. 2c, e).

Metabolic Therapy Radiotracer Uptake Studies

We next evaluated [¹⁸F]DASA-23 and [¹⁸F]FDG radiotracer uptake in cells treated with the metabolic agents dichloroacetate and ivosidenib after 3 or 6 days of drug treatment (Fig. 3). There was a significant interaction between metabolic drug (dichloroacetate or ivosidenib) and treatment duration (3 or 6 days) on the uptake of [¹⁸F]FDG (two-way ANOVA, p = 0.04, Suppl. Fig. S5b (ESM)) and [¹⁸F]DASA-23 (two-way ANOVA, p < 0.0001, Suppl. Fig. S4b (ESM)). Further analysis of [¹⁸F]DASA-23 uptake in the 3-day treatment group revealed a significant reduction in the mean radiotracer uptake in treated vs. untreated cells (one-way ANOVA, p < 0.0001, F = 13.24, Fig. 3a). Greater decreases in [¹⁸F]DASA-23 uptake were observed between untreated cells (25.2 ± 4.0 % uptake/mg protein) and ivosidenib-treated cells (15.5 ± 2.1 % uptake/mg protein, p = 0.0004, 38.7 ± 8.4 % reduction), compared to dichloroacetate-treated cells (17.8 ± 4.0 % uptake/mg protein, p = 0.004, 33.8 ± 13.0 % reduction) (Fig. 3a, c).

Fig. 3.

[¹⁸F]DASA-23 and [¹⁸F]FDG uptake in untreated cells and cells treated with metabolic agents dichloroacetate and ivosidenib. a Three-day drug treatment; untreated vs dichloroacetate, **p = 0.004; untreated vs ivosidenib, ***p = 0.0004; n = 6 per condition, p values obtained from one-way ANOVA with Bonferroni correction for multiple comparisons. # = significant interaction in [¹⁸F]FDG uptake between the metabolic agents and duration of treatment, two-way ANOVA, 6-day data is shown in Fig. 3b. b Six-day treatment; untreated vs dichloroacetate, ****p < 0.0001; untreated vs ivosidenib, ***p = 0.0005; n = 6 per condition, p values obtained from one-way ANOVA with Bonferroni correction for multiple comparisons. # = significant interaction on [¹⁸F]FDG cellular uptake between the alkylating agent and length of treatment duration, two-way ANOVA, 3-day data is shown in Fig. 3a. Combined 3- and 6-day data is shown for [¹⁸F]DASA-23 and [¹⁸F]FDG in Suppl. Fig. 4b and 5b, respectively. C percentage reduction of [¹⁸F]DASA-23 and [¹⁸F]FDG radiotracer uptake relative to untreated cells in the presence of metabolic agents is shown for 3-day treatment, ivesodinib ***p = 0.0001. d Percentage reduction of [¹⁸F]DASA-23 and [¹⁸F]FDG radiotracer uptake relative to untreated cells in the presence of metabolic agents is shown for 6-day treatment, dichloroacetate **p = 0.001.

Following 6-day exposure to metabolic drugs, additional reductions were detected in [¹⁸F]DASA-23 uptake in treated vs. untreated cells (one-way ANOVA, p < 0.0001, F = 114.2, Fig. 3b). However, the greatest differences at this time point were observed between untreated cells (29.4 ± 2.2 % uptake/mg protein) and dichloroacetate-treated cells (11.7 ± 1.8 % uptake/mg protein, p < 0.0001, 59.6 ± 6.4 % reduction), followed by ivosidenib-treated cells (23.5 ± 1.7 % uptake/ mg protein, p = 0.0005, 19.5 ± 5.7 % reduction) (Fig. 3b, d).

Radiotracer Uptake Studies of Remaining Five Drug Classes (One Drug/Class)

The remaining five therapeutic agents studied were in separate drug classes: cediranib (inhibitor of vascular endothelial growth factor [VEGF] receptor), irinotecan (inhibitor of topoisomerase I), erlotinib (inhibitor of epidermal growth factor receptor [EGFR]), vincristine (tubulin inhibitor), and 5-fluorouracil (5-FU, thymidylate synthase inhibitor).

There was a significant interaction between the remaining five drugs, duration of treatment (3 or 6 days) on the uptake of [¹⁸F]FDG (two-way ANOVA, p = 0.0001, Suppl. Fig. S5c (ESM)) and [¹⁸F]DASA-23 (two-way ANOVA, p < 0.0001, Suppl. Fig. S4c (ESM)). Further analysis of [¹⁸F]DASA-23 uptake in the 3-day treatment group revealed a significant reduction in the mean radiotracer uptake in treated vs. untreated cells (one-way ANOVA, p < 0.0001, F = 25.49, Fig. 4a). The greatest reduction in [¹⁸F]DASA-23 uptake between untreated cells (25.2 ± 4.0 % uptake/mg protein) was in the vincristine group (11.0 ± 1.0 % uptake/mg protein, p < 0.0001, 56.4 ± 3.9 % reduction), followed by 5-fluorouracil (14.3 ± 1.9 % uptake/mg protein, p < 0.0001, 43.4 ± 7.5 % reduction), cediranib (15.7 ± 1.8 % uptake/mg protein, p < 0.0001, 37.8 ± 7.1 % reduction), irinotecan (19.1 ± 1.7 % uptake/mg protein, p = 0.0005, 24.4 ± 6.6 % reduction), and erlotinib (20.1 ± 2.4 % uptake/mg protein, p = 0.01, 20.2 ± 9.3 % reduction) treatments.

Fig. 4.

[¹⁸F]DASA-23 and [¹⁸F]FDG cellular uptake in untreated cells and cells treated with the remaining five classes of drugs: cediranib, irinotecan, erlotinib, vincristine, and 5-fluorouracil. a Three-day treatment; untreated vs cediranib, ****p < 0.0001; untreated vs irinotecan, ***p = 0.0005; untreated vs erlotinib, *p = 0.011; untreated vs vincristine, ****p < 0.0001; untreated vs 5-fluorouracil, ****p < 0.0001; n = 6 per condition, p values obtained from one-way ANOVA with Bonferroni correction for multiple comparisons. # = significant interaction in [¹⁸F]FDG cellular uptake between the five drug classes and duration of treatment, two-way ANOVA, 6-day data is shown in Fig. 4b. b Six-day treatment; [¹⁸F]DASA-23 uptake: untreated vs cediranib, ****p < 0.0001; untreated vs irinotecan, ****p < 0.0001; untreated vs erlotinib, ****p < 0.0001; untreated vs vincristine, ****p < 0.0001; untreated vs 5-fluorouracil, ****p < 0.0001; n = 6 per condition, p values obtained from one-way ANOVA with Bonferroni correction for multiple comparisons. # = significant interaction on [¹⁸F]FDG cellular uptake between the alkylating agent and length of treatment duration, two-way ANOVA, 3-day data is shown in Fig. 5a. Combined 3- and 6-day data is shown for [¹⁸F]DASA-23 and [¹⁸F]FDG in Suppl. Fig. 4c and 5c, respectively. c Percentage reduction of [¹⁸F]DASA-23 and [¹⁸F]FDG radiotracer uptake relative to untreated cells in the presence of remaining agents is shown for 3-day treatment, cediranib **p = 0.007, erlotinib **p = 0.002, 5-fluorouracil ***p = 0.002. d Percentage reduction of [¹⁸F]DASA-23 and [¹⁸F]FDG radiotracer uptake relative to untreated cells in the presence of remaining agents is shown for 6-day treatment, cediranib ***p = 0.0003, irinotecan ****p < 0.0001, erlotinib ****p < 0.0001, 5-fluorouracil ****p < 0.0001.

In the case of 6-day treatment studies of the remaining five drugs, a significant reduction in mean [¹⁸F]DASA-23 uptake in treated vs. untreated cells was also identified (one-way ANOVA, p < 0.0001, F = 69.16, Fig. 4b). Similar to the 3-day treatment studies of the same drugs, the greatest reduction in [¹⁸F]DASA-23 cellular uptake was observed between untreated cells (29.4 ± 2.2 % uptake/mg protein) and treated cells in the vincristine group (6.3 ± 0.9 % uptake/mg protein, p < 0.0001, 78.5 ± 3.0 % reduction), followed by 5-fluorouracil (7.5 ± 1.3 % uptake/mg protein, p < 0.0001, 74.5 ± 4.5 % reduction) treated cells. However, this was followed by cells treated with irinotecan (11.6 ± 1.6 % uptake/mg protein, p < 0.0001, 60.6 ± 5.3 % reduction), erlotinib (13.3 ± 1.5 % uptake/mg protein, p < 0.0001, 54.4 ± 5.0 % reduction), and finally cediranib (17.0 ± 5.0 % uptake/mg protein, p < 0.0001, 42.1 ± 16.7 % reduction) (Fig. 4b, d).

Changes in [¹⁸F]DASA-23 Cellular Uptake Positively Correlates with Changes in PKM2 Protein Levels

We next determined the effect of a 6-day treatment with each of the 11 drugs on PKM2 protein levels by Western blot analysis and evaluated whether PKM2 protein levels correlated with the [¹⁸F]DASA-23 cellular uptake (Fig. 5a). PKM2 protein levels were reduced in cells treated with vincristine, erlotinib, and procarbazine (PKM2/GAPDH ratios of 0.07 ± 0.09, 0.12 ± 0.09, and 0.1 ± 0.05, respectively). Alkylating agents temozolomide, carmustine, and lomustine exhibited comparable PKM2/GAPDH ratios of 0.46 ± 0.12, 0.56 ± 0.1, and 0.67 ± 0.09, respectively. Finally, we normalized the mean PKM2/GAPDH ratios and mean [¹⁸F]DASA-23 cellular uptake values from each drug group (n = 11) to the corresponding mean value from the untreated group (Fig. 5b). There was a moderate correlation between normalized PKM2 protein expression (PKM2/GAPDH values in drug-treated cells normalized to untreated cells) and normalized [¹⁸F]DASA-23 uptake ([¹⁸F]DASA-23 uptake values normalized to untreated cells), r = 0.44, p = 0.15.

Fig. 5.

Correlation between [¹⁸F]DASA-23 cellular uptake and PKM2 protein expression. a Six-day cellular uptake values from each drug condition (n = 11 total) and untreated cells are shown on the y-axis and PKM2/GAPDH ratios obtained from Western blot studies are shown on the x-axis. Values shown are mean ± SEM in both y- and x-directions. b [¹⁸F]DASA-23 cellular uptake values for each drug exposure were normalized to those from untreated cells and plotted on the y-axis as normalized [¹⁸F]DASA-23 cellular uptake. PKM2/GAPDH ratios for each drug exposure were also normalized to untreated cells and plotted along the x-axis as normalized PKM2/GAPDH. Values shown are mean ± SEM in both y- and x-directions. Pearson’s correlation, r = 0.44, p = 0.15. Dotted lines represent 95 % confidence interval.

Discussion

The molecular imaging of the Warburg effect, one of the principal causes for the elevated glucose metabolism in tumors, with [¹⁸F]FDG-PET forms the basis of imaging-based diagnosis and response assessment in several cancers [17]. However, [¹⁸F]FDG-PET has limited use for molecular imaging of brain tumors due to the relatively high background signal from glucose metabolism in the brain [18]. We demonstrated that [¹⁸F]DASA-23 has a greater than 10-fold higher cellular uptake in untreated cells compared to [¹⁸F]FDG. A high differential uptake between [¹⁸F]DASA-23 and [¹⁸F]FDG was sustained in the presence of each of the 11 drugs tested that cover 7 classes of antineoplastic agents currently used in patients with GBM. This suggests that PET imaging of PKM2 expression with [¹⁸F]DASA-23 has the potential to play an important role in the noninvasive detection and measurement of malignancies where [¹⁸F]FDG has limitations, specifically GBM.

[¹⁸F]DASA-23 can bind dimeric PKM2 and promote association into active tetramers; however, the binding pocket for [¹⁸F]DASA-23 is reportedly unavailable on constituently active tetramers due to conformational changes associated with tetramer formation [22]. Reduced [¹⁸F]DASA-23 binding in response to therapeutic agents can be associated with multiple factors including (1) reduced PKM2 expression, (2) activation of PKM2 dimers to tetramers, and (3) modification of PKM2 through post-translational modifications [22, 27–29]. This was a proof-of-principle study, and no efforts were made to determine the mechanism by which the therapeutic agents studied affect PKM2 levels or enzymatic state. The moderate correlation between [¹⁸F]DASA-23 cellular uptake and PKM2 protein levels likely suggests that changes evident in [¹⁸F]DASA-23 cellular uptake are the result of additional factors beyond reduced PKM2 protein levels. Although [¹⁸F]DASA-23 cannot distinguish between these different states of PKM2, the clinical implication of this strategy is to detect a global cellular response to any variety of treatments and to correlate such a change with patient outcomes [29–31].

In this study, we have demonstrated significant reductions in [¹⁸F]DASA-23 cellular uptake in response to seven classes of antineoplastic agents used in the treatment of newly diagnosed and recurrent GBM. The most significant decreases in [¹⁸F]DASA-23 uptake were evident in response to treatment with alkylating agents (p < 0.0001), irinotecan (p = 0.0012), erlotinib (p = 0.02), and 5-fluorouracil (p = 0.005) (Suppl. Fig. S6 (ESM)) suggesting the potential to non-invasively monitor GBM therapeutic response with [¹⁸F]DASA-23 to these agents in living subjects. Although we used an isocitrate dehydrogenase (IDH) wild-type U87 cell line, the IDH1 inhibitor, ivosidenib, resulted in a ~ 38 % and ~ 20 % reduction in [¹⁸F]DASA-23 uptake at 3 and 6 days of treatment, respectively, compared to untreated cells. The smaller reduction in [¹⁸F]DASA-23 uptake in ivosidenib-treated cells at 6 days vs 3 days of treatment suggests a selection pressure favoring IDH-wild type cells after 3 days of incubation. This was in contrast to treatment with DCA, where [¹⁸F]DASA-23 cellular uptake continued to decrease over the 3- and 6-day treatment and is likely the result of continued inhibition of the mitochondrial enzyme pyruvate dehydrogenase kinase which subsequently promotes mitochondrial oxidation and decreased lactic acid formation. The cellular uptake of [¹⁸F]DASA-23 was higher than that of [¹⁸F]FDG in all conditions studied and may result from competition of glucose within HBSS for [¹⁸F]FDG uptake. HBSS containing 5 mM glucose was selected in order to mimic physiological glucose concentrations [32–34] in patients who have fasted prior to their [¹⁸F]FDG PET scan. These results further highlight the potential utility of [¹⁸F]DASA-23 for PET imaging of brain tumors compared to [¹⁸F]FDG.

There exist several PET radiotracers used to evaluate various aspects of brain tumors including amino acid and fatty acid metabolism, cell proliferation, and hypoxia [35, 36]. Protein synthesis rates have been measured in GBM utilizing [¹¹C]tyrosine and [¹¹C]methionine [37, 38]. Radiolabeled unnatural amino acids have been used to visualize amino acid transporter activity including L-3,4-dihydroxy-6-¹⁸F-fluorophenyl-alanine ([¹⁸F]FDOPA) and [¹⁸F]fluoroethyltyrosine ([¹⁸F]FET). The uptake of these radiotracers is primarily mediated by L-type amino acid transporter (LAT) mechanisms and principally reflects nutrient uptake to support increased biomass and proliferative energy demands [39]. While these agents have utility in the diagnosis and grading of brain tumors and have been explored in treatment monitoring and prognosis [36, 40–42], the field of neuro-oncology would benefit from an imaging biomarker that provides a timely and accurate measurement of alterations in the glioblastoma’s aberrant glycolysis in response multiple antineoplastic therapies. Aberrant glycolysis is a fundamental feature of cancer cells [43], making the glycolytic pathway an appealing target for imaging brain tumors undergoing treatment [35]. Future studies will investigate the utility of [¹⁸F]DASA-23 to detect response to therapy compared to measures of amino acid metabolism (e.g., [¹⁸F]FDOPA or [¹⁸F]FET).

While these results are promising and suggest potential for early response assessment to multiple therapeutic agents using [¹⁸F]DASA-23, there are several limitations. These include the use of a single cell line and evaluation of one concentration of each drug studied (IC₅₀ from 3-day incubation). Future studies will involve the evaluation of glycolytic response in multiple GBM cell lines including patient-derived cell lines with varying molecular characteristics (i.e., O6-methylguanine-DNA methyltransferase [MGMT] promoter methylation status and IDH mutation status) and will evaluate multiple concentrations of therapeutic agents.

Conclusions

In summary, decreases in [¹⁸F]DASA-23 cellular uptake are evident in U87 cells treated with multiple classes of therapeutic agents. [¹⁸F]DASA-23 uptake in GBM cells was significantly higher compared to [¹⁸F]FDG, regardless of the presence of any of seven classes of antineoplastic agents. This suggests a new approach for the early detection of metabolic response to therapy in GBM. Future studies will expand upon this work and explore multiple cell lines as well as pre-clinical models to better elucidate the mechanism of the changes. In parallel, a clinical trial is underway to investigate the safety and diagnostic utility of [¹⁸F]DASA-23 PET in GBM patients (NCT03539731).