https://orcid.org/0000-0002-1723-604X
Abstract
Reactive astrogliosis is a hallmark of various brain pathologies, including neurodegenerative diseases and glioblastomas. However, the specific intermediate metabolites contributing to reactive astrogliosis remain unknown. This study investigated how glioblastomas induce reactive astrogliosis in the neighboring microenvironment and explore 11C-acetate PET as an imaging technique for detecting reactive astrogliosis.
Through in vitro, mouse models, and human tissue experiments, we examined the association between elevated 11C-acetate uptake and reactive astrogliosis in gliomas. We explored acetate from glioblastoma cells, which triggers reactive astrogliosis in neighboring astrocytes by upregulating MAO-B and monocarboxylate transporter 1 (MCT1) expression. We evaluated the presence of cancer stem cells in the reactive astrogliosis region of glioblastomas and assessed the correlation between the volume of 11C-acetate uptake beyond MRI and prognosis.
Elevated 11C-acetate uptake is associated with reactive astrogliosis and astrocytic MCT1 in the periphery of glioblastomas in human tissues and mouse models. Glioblastoma cells exhibit increased acetate production as a result of glucose metabolism, with subsequent secretion of acetate. Acetate derived from glioblastoma cells induces reactive astrogliosis in neighboring astrocytes by increasing the expression of MAO-B and MCT1. We found cancer stem cells within the reactive astrogliosis at the tumor periphery. Consequently, a larger volume of 11C-acetate uptake beyond contrast-enhanced MRI was associated with a worse prognosis.
Our results highlight the role of acetate derived from glioblastoma cells in inducing reactive astrogliosis and underscore the potential value of 11C-acetate PET as an imaging technique for detecting reactive astrogliosis, offering important implications for the diagnosis and treatment of glioblastomas.
11C-acetate PET imaging visualizes reactive astrogliosis in the periphery of the glioma.
Acetate, derived from glioma cells, triggers reactive astrogliosis in neighboring astrocytes.
The volume of reactive astrogliosis of glioma is associated with patient prognosis.
Reactive astrogliosis plays a critical role in various brain pathologies, but the intermediate metabolites involved in the metabolic pathways that may cause reactive astrogliosis in glioblastomas remain unknown. Our findings reveal an association between elevated 11C-acetate uptake and reactive astrogliosis in the peritumoral region of glioblastomas. We demonstrate that acetate, derived from glioblastoma cells, triggers reactive astrogliosis in neighboring astrocytes by upregulating the expression of MAO-B and monocarboxylate transporter 1 (MCT1). The presence of cancer stem cells in the reactive astrogliosis region suggests a potential association with poor prognosis in patients with glioblastoma. Furthermore, a larger volume of 11C-acetate uptake beyond contrast-enhanced MRI correlates with worse clinical outcomes. These findings highlight the role of acetate derived from glioblastoma cells in inducing reactive astrogliosis and underscore the potential value of 11C-acetate PET as an imaging technique for detecting reactive astrogliosis, offering important implications for the diagnosis and treatment of gliomas.
Astrocytes are sensitive to environmental conditions and undergo dynamic changes in their molecular, functional, and morphological properties in response to various physical and chemical triggers in the CNS. These astrocytes are called “reactive astrocytes.”1 Reactive astrocytes, also observed in diseases such as Alzheimer’s disease (AD) and Parkinson’s disease (PD), stroke, and glioblastoma, play a significant role in CNS pathologies.2–6 In AD and PD, pathogenic molecules like amyloid beta, α-synuclein, cytokines, and hydrogen peroxide (H2O2) have been identified as key inducers of reactive astrogliosis.3,4 We have recently reported that reactive astrogliosis induced by these pathogenic molecules generates excessive GABA and H2O2 via the monoamine oxidase B (MAO-B)-dependent putrescine degradation pathway by activating the urea cycle, which causes neuronal dysfunction and neurodegeneration in AD.2,3,7 These studies have shown that pathogenic molecules and the aberrant activation of metabolic pathways can trigger reactive astrogliosis. However, the identities and characteristics of the intermediate metabolites involved in the metabolic pathways that may cause reactive astrogliosis in glioblastomas remain unknown.
Glioblastoma, the most aggressive type of glioma with a median overall survival (OS) of approximately 15 months, accounts for the majority of primary malignant brain tumors.8 One of the most conspicuous features of glioblastomas is the formation of a tumor microenvironment (TME).9 The TME comprises various cell types, including cancer stem cells and nontumorous cells, such as microglia/macrophages, astrocytes, neutrophils, dendritic cells, and endothelial cells.10 Among these cell types, astrocytes undergo reactive astrogliosis, characterized by the upregulation of glial fibrillary acidic protein (GFAP) within the TME.11 Reactive astrocytes contribute to the proliferation, invasion, and drug resistance of glioblastoma by secreting cytokines and chemokines.6,12–14 They enhance tumor cell proliferation by releasing growth factors, such as interleukin-6, transforming growth factor-β, and insulin-like growth factor 1. Additionally, reactive astrocytes influence tumor cell invasion by activating matrix metalloproteinase-2 (MMP2).14,15 Mutual interactions between reactive astrogliosis and metastatic tumor cells have also been observed in brain metastasis.15 Reactive astrocytes surround gliomas in both human biopsy tissues and murine models.6,12 Moreover, in a murine glioma resection and recurrence model, the presence of reactive astrocytes after resection promoted tumor proliferation and invasion, suggesting the potential presence of cancer stem cells within regions of reactive astrogliosis.16 Therefore, understanding the relationship between tumors and neighboring reactive astrocytes is crucial for combating glioblastomas.
Acetate, a key metabolite in energy metabolism, has recently been reported as a bioenergetic substrate for human glioblastomas, suggesting that glioblastoma cells take up acetate.17 Additionally, acetate serves as an astrocyte-specific substrate and an alternative energy source to glucose in the normal brain.18–21 Astrocytes transport and utilize acetate primarily via monocarboxylate transporter 1 (MCT1).19,22,23 Our previous studies demonstrated that reactive astrocytes activate the urea cycle to detoxify toxic molecules in AD, raising the possibility that acetate can facilitate the urea cycle.7 Indeed, we have previously reported that 11C-acetate, a radiotracer version of acetate, can be used to visualize reactive astrogliosis in patients with AD.24 Currently, 11C-acetate is clinically used to detect low-grade tumors with acetate dependence in the body.25 Consistently, we also discovered that glioblastomas were associated with high 11C-acetate uptake in patients undergoing PET.26,27 However, whether reactive astrocytes or glioblastoma cells take up 11C-acetate is unclear. In this study, we aimed to investigate whether glioblastoma induces reactive astrogliosis and whether reactive astrocytes surrounding gliomas exhibit high 11C-acetate uptake. We also aimed to identify acetate-mediated reactive astrogliosis and evaluate the potential of 11C-acetate as an imaging probe for reactive astrogliosis in patients with glioblastoma. Finally, we assessed the prognostic implications of these findings.
Materials and Methods
Glioma Cell Culture
U87MG, U373MG, and T98G cancer cells were obtained from the Korean Cell Line Bank (KCLB) of Seoul National University. U87-IDH1-mt (U87mt) cells were obtained from the American Type Culture Collection (ATCC). To prepare conditioned media (CM), U87MG and U87-IDH1-mt (U87mt) cells were incubated for 1 day (approximately 80% confluency). Details are available in Supplementary Information.
Quantitative Real-Time Polymerase Chain Reaction
RNAs were isolated from mouse primary astrocytes (AST) and treated with U87MG-CM, U87mt-CM, and Dulbecco’s Modified Eagle Medium (DMEM) (control) using TRIzol reagent (Invitrogen). cDNA was synthesized from 1 μg of RNA by iScript™ cDNA Synthesis Kit (BioRad) and quantitatively amplified using SYBR Green Real-Time PCR Master Mix (BioRad). Details are available in Supplementary Information.
RNA Sequencing
Total RNAs were extracted from AST cells treated with U87MG-CM and DMEM using TRIzol reagent (Invitrogen) for RNA expression profiling. RNA sequencing and data analysis were performed using EBIOGEN, Inc.
Acetate Assay
To measure the amount of acetate in U87MG-CM, a colorimetric assay was performed using an acetate assay kit (Abcam), according to the manufacturer’s recommendations. DMEM was used as a negative control. Details are available in Supplementary Information.
13C-Acetate Measurement
U87MG cells (2 × 105) were seeded into 6-well plates. After 24 h of adaptation, the cells were treated with glucose-free DMEM supplemented with 10% FBS and 20 mM U–13C6–glucose at 37°C in a humidified incubator with 5% CO2. After 6 h and 24 h of treatment, the U87MG-CM were centrifuged for 10 min at 1000 rpm and the supernatant was collected. The media were mixed with 10% nuclear magnetic resonance (NMR) buffer and measured using an 800 MHz Bruker Avance spectrometer (Bruker BioSpin) equipped with a cryogenic triple resonance probe (College of Pharmacy, Seoul National University). J-scaled distortion-free 1H–13C Heteronuclear Single Quantum Coherence (HSQC) NMR spectra, which show a 6 fold increased 13C–13C coupling constant within multiplets, were assessed.28 Relative peak volumes were obtained by normalizing to the TSP peak volume and bicinchoninic acid assay (BCA) protein concentration. Details are available in Supplementary Information.
Primary Cells
Primary astrocytes were isolated from ICR mouse brain cortices on postnatal day 1, cultured in poly-d-lysine-coated flasks with MEM and supplements, and purified by removing nonastrocytic cells after 7 days. Details are available in Supplementary Information.
Animal Model
All animal experimental procedures were approved and performed following the protocols of the Institutional Animal Care and Use Committee (IACUC) of Yonsei University College of Medicine. The animals were maintained under maximum barrier specific-pathogen-free conditions with 12 h light–dark cycle at 20‒22°C temperature and 35‒55% humidity.
U87MG-Orthotopic Xenograft.—
Balb/c nude mice (female; 5-weeks-old; Orient Bio) were anesthetized with isoflurane. For implantation, cells were suspended in phosphate-buffered saline (PBS) at a concentration of 1 × 105 cells/μL, and each mouse was stereotactically injected with 2 μL of suspended cells. Details are available in Supplementary Information.
Patient-Derived Glioblastoma-TS Isolation and Orthotopic Xenograft.—
Four tumorsphere (TS)-forming glioblastoma cell lines, TS13-64, TS13-100, TS19-117, and TS19-156, were established from fresh glioblastoma tissue specimens. TS formation following previous methods and dissociated TS were implanted into the right frontal lobe of Balb/c nude mice using the stereotaxic instrument. Details are available in Supplementary Information.
KDS2010 Administration
One week after the stereotactic injection, tumor formation was confirmed using a T2-weighted MRI. The mice were randomly divided into 2 groups and allowed to drink water (control) or KDS2010 (10 mg/kg daily) containing water ad libitum. KDS2010, a recently developed selective, reversible MAO-B inhibitor, was synthesized as previously described.29 One week after implantation, the animals were monitored using T2-weighted MRI for the presence of xenografted tumors in the brain. Mice with MRI-identified brain tumors were used for microPET imaging studies.
Virus Injection
For astrocyte-specific gene-silencing of MCT1, we injected AAV-GFAP-Cre-mCherry (0.5 μL) and AAV-DJ vector containing pSico-MCT1-shRNA-GFP cassette (or pSico-scrambled-shRNA-GFP for control) (0.5 μL) into stereotaxic coordinates of AP +0.5; ML –2; DV –3.2 and –2.8 to bregma using stereotaxic apparatus with a rate of 0.15 mL/min. Details are available in Supplementary Information.
Animal microPET and MRI
All the mice (n = 5) were imaged using a microPET scanner (Inveon, Siemens Healthcare) and a 9.4 T preclinical MR (Bruker BioSpec 94/20 USR, Ettlingen; software: ParaVision 5.0). Details are available in Supplementary Information.
Autoradiography
Mouse orthotopic xenograft models (4‒5 mice per group) were i.v., treated with 14C-acetate (3 μCi, PerkinElmer). The brains were quickly isolated and frozen in an optimal cutting temperature compound (OCT, Leica). Frozen mice brains were sectioned at 20 μm thickness using a cryostat microtome and exposed to an imaging plate for 2 weeks. The plates were scanned using a phosphor imager (Typhoon FLA 7000; GE Healthcare). Details are available in Supplementary Information.
Animal Tissue Immunostaining
The mice were deeply anesthetized with isoflurane and perfused. The excised brains were postfixed overnight in 4% PFA and immersed in 30% sucrose for 48 h for cryoprotection. After fixation, the brains were mounted into an OCT embedding solution (Leica) and frozen at −20° to −80°C. Tissues were cut in a cryostat by 30-μm coronal sections. Sections were processed with 3 additional washes in PBS, incubated for 1 h in a blocking solution, and immunostained with a mixture of primary antibodies in a blocking solution at 4°C on a shaker overnight. Details are available in Supplementary Information.
In vitro 14C-Acetate Uptake Assay
In all experiments related to in vitro 14C-acetate uptake, 14C-acetate (PerkinElmer) 0.5 μCi/0.5 mL in the external solution was added into each well. Details are available in Supplementary Information.
siRNA Transfection
Primary astrocytes were transfected with siRNA (AccuTarget™ Genome-wide Predesigned siRNA, Bioneer) with OPTIMEM medium (Gibco) following the manufacturer’s instructions. Details are available in Supplementary Information.
Western Blotting
The cells were subsequently lysed with radioimmunoprecipitation assay lysis buffer (Thermo Fisher Scientific) containing protease inhibitor (Roche), and 10 μg of proteins were separated by electrophoresis in a 10% sodium dodecyl sulfate-polyacrylamide gel and transferred onto a polyvinylidene fluoride membrane (Millipore). Details are available in Supplementary Information.
Image Analysis of Immunostaining
Immunostained sections were imaged using a Zeiss LSM710 confocal microscope. Confocal images were analyzed using ImageJ software (NIH). Details are available in Supplementary Information.
Statistical Analysis for In Vitro and In Vivo Animal Studies and Clinical Data
Prism 8 software (GraphPad Software, Inc.) and IBM SPSS Statistics for Windows (version 25.0; IBM Corp.) were used for statistical analysis. The data are presented as mean ± SEM for parametric analysis and as median with quartiles for nonparametric analysis. The significance level is represented by asterisks (*P < .05, **P < .01, ***P < .001, ****P < .0001; ns, not significant). Detailed descriptions of the statistics are presented in the Supplementary Information.
Human Studies
From January 2016 to October 2017, 211 patients with glioma were surgically treated at our institution. Among them, 102 patients underwent 11C-acetate PET/CT as part of a nonrandomized prospective clinical trial at Severance Hospital with at least 5 years of follow-up (Supplementary Table 1). This study was approved by the IRB of Severance Hospital (4-2011-0697), and all patients provided written informed consent. The study was registered with the Clinical Research Information Service of the Korea Centers for Disease Control and Prevention (KCT0004468). Details are available in Supplementary Information.
Tissue Preparation for Immunostaining of Targeted Biopsied Glioma Tissues
For immunostaining of human glioma tissues from the targeted biopsy, tissues were fixed with 4% PFA for 1 day, made into paraffin blocks, and cut into 2-μm sections. After antigen retrieval, the slides were blocked with a blocking solution and incubated with primary antibodies. After washing in PBS 3 times, sections were incubated with corresponding fluorescent secondary antibodies (Invitrogen) for 2 h at room temperature. Following several washes with PBS and counterstaining with Hoechst 34580 (Molecular Probes, 1:1000), the tissue slices were mounted in a fluorescent mounting medium (Dako). Several fluorescent images were obtained with an LSM710 confocal microscope (Carl Zeiss) and 30 μm Z-stack images in 5-μm steps were processed for further analysis using ZEN software (Carl Zeiss). Antibodies are in Supplementary Information.
World Health Organization 2021 Classification
The fifth edition of the 2021 WHO CNS tumor classification was published with extensive knowledge of genetic and molecular characteristics.30 Our patients underwent surgery from 2016 to 2017. Those with relevant genetic information were classified accordingly and those without relevant genetic information were left as not otherwise specified (NOS). Details are available in Supplementary Information.
PET Imaging and Analysis
11C–acetate PET/CT was performed using a PET/CT scanner (Discovery 710, General Electric Medical Systems). All PET/CT and MRI images were reviewed and analyzed by a double-board-certified radiologist, a nuclear medicine physician, and a nuclear medicine physician using a fusion module in the imaging software (MIM-6.5; MIM Software Inc.), and decisions were reached by consensus. The standardized uptake value ratio (SUVR) of the tumor to the choroid plexus was determined as the SUVmax of the tumor to the SUVmean of the choroid plexus.31 Tumor volume on 11C-acetate PET/CT was calculated using an iso-contour volume of interest (VOI) determined using a threshold (SUVmean of the contralateral choroid plexus + 2 standard deviation).26 We defined the 11C-acetate uptake area beyond the contrast-enhanced MRI as VolumeACE-MRI. Details are available in Supplementary Information.
MRI Imaging and Volume Measurement
Anatomical datasets were obtained using a 3T whole-body MR scanner (Magnetom Trio, Siemens Medical Solutions) with T2-weighted fluid-attenuated inversion recovery (FLAIR) and contrast-enhanced 3-dimensional (3D) T1-weighted sequences. For tumor volume measurement, tumor margins were manually delineated on contrast-enhanced axial images of the tumor and all regions of interest were integrated to measure tumor volume.32 Details are available in Supplementary Information.
Results
Regions of 11C-Acetate Uptake at the Tumor Boundary in Patients with Glioblastomas
To test the possibility that reactive astrocytes surrounding gliomas take up high levels of 11C-acetate, we compared 11C-acetate uptake between gliomas. Upon visual inspection, IDH1-wildtype diffuse gliomas showed significantly higher 11C-acetate uptake, mostly at the tumor periphery (Figure 1A; upper row), whereas IDH1-mutant (IDH1-mt) diffuse gliomas demonstrated minimal or no significant 11C-acetate uptake (Figure 1A; lower row).
Regions of 11C-acetate uptake at the tumor boundary in patients with glioblastomas. (A) Representative images of 11C-acetate PET and MRI in patients with glioblastoma and IDH1-mt astrocytoma. (B) SUVR of 11C-acetate uptake correlated with the WHO 2021 classification. (C) The median tumor volume in 11C-acetate PET images was significantly larger than that seen on MRI in glioblastoma (n = 38) (20.80 (IQR 8.99–59.39) versus 7.72 (IQR 1.80–26.14)). (D–E) Western blot for MCT1 expression and 14C-acetate uptake in mouse primary astrocytes and various human glioma cell lines (U87MG, U373MG, T98G, and U87mt). (F–G) Immunofluorescence images of GFAP, Ki-67, MAO-B, and MCT1 in human glioma. C, Cortical; S, Star-shape astrocyte; L, Linear-shape astrocyte; T, Tumor. (H–I) Immunofluorescence images showing GFAP, Ki-67, MAO-B, and MCT1 in human glioma. (J) Immunofluorescence images of CD133, MAO-B in human glioma. (K) Pearson’s correlation coefficients for MAO-B and CD133 colocalization assays. (L) The proportion of MAO-B+ or CD133+ in Ki-67+ cells in IDH1-wildtype glioma (n = 5). (M) Representative image of Sholl analysis of an astrocyte according to S, L, or T region in human glioma. (N–Q) Quantification of GFAP area, the number of Ki-67+ cells, MAO-B, and MCT1 immunoreactivity of GFAP + astrocytes in glioblastoma versus IDH1-mt gliomas (n = 3). Scale bars, 50 μm (F), (G); 20 μm (H and I). Data are presented as mean ± SEM. *P < .05, **P < .01, ***P < .001, ****P < .0001 by 1-way ANOVA with Tukey’s test (C), unpaired 2-tailed t-test (O and Q), Kruskal–Wallis test (B), (N) left, (P) left, or Mann–Whitney U-test (N) right, (P) right.
Semi-quantitatively, the median SUVratio (SUVR, the SUVmax of tumors to the SUVmean of choroid plexus) of 11C-acetate PET in glioblastomas (2.40, interquartile range (IQR) 2.05‒2.62, n = 38) was significantly elevated than that in oligodendrogliomas (1.19, IQR 0.99‒1.39, n = 25), IDH1-mt (1.01, IQR 0.92‒1.40, n = 30), and IDH1-wildtype astrocytomas (1.47, IQR 1.31‒2.00, n = 9). The 11C-acetate SUVR of IDH1-wildtype was significantly higher than that of IDH1-mt astrocytomas and oligodendrogliomas, but no difference between the SUVR of IDH1-mt astrocytomas and oligodendrogliomas was observed (Figure 1B).
The volume calculated using 11C-acetate PET was larger than that calculated using contrast-enhanced MRI (Figure 1C). This suggests that 11C-acetate PET provides clearer and larger boundaries than that of contrast-enhanced MRI. This also indicates that the regions of 11C-acetate uptake beyond the contrast-enhanced area on MRI might indicate the possibility of 11C-acetate uptake by cells other than glioblastoma.
To ascertain which cells within the glioblastoma take up acetate, we examined 14C-acetate uptake by mouse primary astrocytes (AST) and several human glioma cell lines, including U87MG, U373MG, T98G, and U87mt. We found that AST showed higher 14C-acetate uptake than in human glioma cell lines, including U87MG, U373MG, T98G, and U87mt (Figure 1D). In addition, because MCT1 is an acetate transporter,19,33 we found that AST showed higher expression of MCT1 than that of the aforementioned human glioma cell lines (Figure 1E). These results suggest that the cellular origin of acetate uptake is most likely from astrocytes rather than tumor cells.
To examine the presence and distribution of reactive astrocytes in human glioma tissues, we investigated regions with high 11C-acetate uptake. We performed double immunohistochemistry with targeted glioma biopsy tissues using antibodies against GFAP, an astrocyte marker, and Ki-67, a marker of proliferating cells (Figure 1F–I). We identified 4 distinct regions based on the presence of tumors and reactive astrocyte morphology. To distinguish between reactive astrocytes and cancer stem cells in the peritumoral regions, we utilized MAO-B as a reactive astrocyte marker1–3,7 and CD133 as a cancer stem cell marker.34,35 These 2 markers were nonoverlappingly expressed in the peritumoral regions (Figure 1J–L). We examined GFAP and MAO-B positive reactive astrocytes and categorized reactive astrocyte morphology based on GFAP expression (Figure 1M): a T tumor region with shriveled astrocytes, an L region with line-shaped astrocytes, an S region with star-shaped astrocytes, and C regions displayed low expression of GFAP (Figure 1F and H). Morphology was analyzed by Sholl staining of the GFAP-positive astrocytes (Figure 1M). Tangential line-shaped astrocytes delineated the boundary between the T and S regions to form a glial scar-like barrier around the tumor (Figure 1F and H). In contrast, we did not find the L region in IDH1-mt gliomas (Figure 1G and I). Astrocyte reactivity at the margins of gliomas was more severe in the S and L regions of glioblastomas than in the S region or in IDH1-mt gliomas, as demonstrated by the higher expression of GFAP and MAO-B (Figure 1H, I, K, and L). MCT1 was highly expressed in the periphery of glioblastomas (S and L regions), whereas in the S region of IDH1-mt gliomas, MCT1 expression was lower compared to glioblastomas (Figure 1H, I, and M), which was consistent with the 11C-acetate uptake results (Figure 1A–C). These results indicate that reactive astrogliosis is present in regions with high 11C-acetate uptake, particularly at the tumor margins of glioblastomas (S and L regions). Moreover, we observed a significant number of Ki-67-positive cells not only in the T region but also in the L and S regions of glioblastomas (Figure 1F, H, and Q), whereas they were very few in IDH1-mt gliomas (Figure 1G, I, and Q).
Reactive Astrogliosis and 11C-Acetate Uptake in Patient-Derived Xenograft Models
To investigate whether reactive astrogliosis with increased astrocytic MCT1 and MAO-B expression could be recapitulated in a mouse orthotopic xenograft model, we developed patient-derived xenografts using tumorspheres from patients with glioblastoma (glioblastoma-TS model; Figure 2A), as previously described.36,37 We then performed immunostaining with antibodies against GFAP, MAO-B, and MCT1 (Figure 2A–C) and detected distinct T, L, and S regions with GFAP-positive reactive astrocytes at the tumor margin (Figure 2A). These reactive astrocytes showed strong expression of GFAP, MAO-B, and MCT1 in the peritumoral L and S regions compared to that in the contralateral regions (Figure 2B–E). These results indicated that reactive astrogliosis with increased MCT1 levels exists around tumors in patient-derived glioblastoma-TS models, consistent with the findings presented in Figure 1D and E.
Reactive astrogliosis and 11C-acetate uptake in patient-derived xenograft models. (A–B) Immunofluorescence images showing GFAP, MAO-B, and MCT1 in glioblastoma-TS models. (C) Immunostaining images showing MCT1 in glioblastoma-TS models. (D–E) Quantification of MAO-B immunoreactivity of GFAP + astrocyte and MCT1 positive area on the peritumoral (Peri-T.) and contralateral regions (Ctr.) in glioblastoma-TS models (n = 3). Scale bars, 500 μm (A) left; 20 μm (A) middle; 10 μm (A) right, (B), and (C). Data are presented as mean ± SEM. ***P < .001, ****P < .0001 by unpaired 2-tailed t-test (D and E).
Molecular Factors of Glioblastoma that Induce Reactive Astrogliosis
To investigate the relationship between glioblastoma and neighboring reactive astrocytes, we performed RNA sequencing of AST cells treated with U87MG-controlled media (U87MG-CM) (the U87MG cell line was used to simulate glioblastoma) and compared it to that of AST cells treated with DMEM, which served as a negative control. AST treated with U87MG-CM exhibited increased expression of mRNAs associated with reactive astrogliosis, including GFAP, STAT3, SOX9, MAO-B, and MCT1 (SLC16A1), compared with AST treated with DMEM (Figure 3 and B).
Molecular factors of glioblastoma that induce reactive astrogliosis. (A) Heatmap of Z-scores calculated from count per million values of GFAP, MAO-B, STAT3, SOX9, SLC16A1, SLC16A3, CD44, ALDH1A1, ACSS1, and ACSS2 related to astrocytic reactivity or acetate metabolism in DMEM or U87MG-CM-treated primary mouse AST. (B) Z-scores and raw data of GFAP, STAT3, SOX9, MAO-B, and SLC16A1. (C) Schematic diagram illustrating preparation of U87MG or U87-IDH1-CM. (D) Western blot for GFAP, MAO-B, and MCT1 expression. (E) MCT1 mRNA expression. (F) 14C-Acetate uptake in U87MG and U87mt-CM-treated AST. (G) MCT1 expression in shMCT1-transfected AST. (H) MCT1 expression (mRNA) in shMCT1-transfected AST. (I) 14C-Acetated uptake after shMCT1-transfection in AST. (J) The concentration of acetate in U87MG-CM. (K) J-scaling HSQC spectra of the methyl group on 13C2-acetate on the CM of U87MG treated with 20 mmol/L U-13C6-glucose. (L) Peak volume for the doublet of the methyl group from the spectra normalized to trimethylsilyl propionate (TSP) and protein. (M) Western blot for GFAP, MAO-B, and MCT1 expression in acetate-treated AST. Data are presented as mean ± SEM. *P < .05, **P < .01, ***P < .001, ****P < .0001 by 1-way ANOVA with Tukey’s test (E, F, and J) or unpaired 2-tailed t-test (H, I, and L).
To date, we have consistently demonstrated elevated levels of acetate uptake and reactive astrogliosis in glioblastoma patients and animal models but not in patients and animal models with IDH1-mt glioma. This raises an interesting possibility regarding the potential differences in tumor-derived molecular factors that induce reactive astrogliosis between glioblastomas and IDH1-mt gliomas. To explore this hypothesis, we exposed astrocytes to CM derived from 2 types of human glioma cells (Figure 3C). Treatment with U87MG-CM significantly increased MAO-B and MCT1 expression, whereas treatment with U87mt-CM did not (Figure 3D and E). Additionally, we measured 14C-acetate uptake and found that astrocytes treated with U87MG-CM showed higher uptake than those treated with DMEM or U87mt-CM (Figure 3F). Gene silencing of MCT1 in AST (Figure 3G and H, and Supplementary Figure 1A and B) significantly prevented 14C-acetate uptake (Figure 3I, and Supplementary Figure 1C), indicating that MCT1 is necessary for acetate uptake. These results indicate that tumor-derived molecular factors induce reactive astrogliosis with elevated acetate uptake and expression of MAO-B and MCT1.
We previously demonstrated that acetate facilitates reactive astrogliosis in an AD model.24 This raises an interesting possibility that acetate, as one of the tumor-derived metabolites, causes reactive astrogliosis. To test this possibility, we measured the amount of acetate in U87MG-CM using an acetate assay kit. We observed a time-dependent increase in acetate secretion by U87MG cells (Figure 3J), which is consistent with the results of a previous report on other cancers.38
Furthermore, to investigate the origin of acetate in U87MG cells, we treated cells with U–13C6–glucose and examined the methyl position of acetate using the newly developed J-scaled distortion-free HSQC spectra in U87MG-CM cells (Figure 3K and L). Notably, a doublet was observed for the methyl group of acetate (J = 314 Hz), indicating that the acetate was derived from U–13C6–glucose. Moreover, the levels of acetate originating from glucose showed an approximately 2.5 fold increase at 24 h compared to that at 6 h. These findings strongly suggest that glioma cells exhibit increased acetate production as a result of glucose metabolism, with subsequent secretion of acetate.
To investigate whether acetate alone induces reactive astrogliosis, we treated cells with various concentrations of acetate for 1 day and 0.5 mM acetate daily for 2 days. We found that acetate increased the expression of MAO-B and MCT1 (Figure 3M and Supplementary Figure 1D). These results suggest that excess acetate from glioblastomas can induce reactive astrogliosis in the neighboring astrocytes.
Inhibition of MAO-B and MCT1 Decreases Reactive Astrogliosis and 11C-Acetate Uptake
To investigate whether reactive astrogliosis is necessary for elevated 11C-acetate uptake in glioma, we used the selective and reversible MAO-B inhibitor, KDS2010, which blocked reactive astrogliosis in animal models of AD, PD, and subcortical stroke in our previous reports,2,4,5,31 and performed 11C-acetate microPET imaging and 14C-acetate autoradiography in orthotopic mouse tumor models. The human glioma cell line, U87MG, was used to mimic glioblastoma, respectively. Similar experiments were also conducted on patient-derived xenografts using tumorspheres from patients with glioblastoma (PDX model).
We placed a region of interest (ROI) over the peritumoral and mirror-image contralateral regions to measure the SUV ratio (SUVr) on 11C-acetate microPET images. After the oral administration of KDS2010, we observed a significant decrease in 11C-acetate SUVr on microPET images (Figure 4A and B, and Supplementary Figure 2A and E) decreased 14C-acetate uptake as visualized by autoradiography in the peritumoral regions (Supplementary Figure 2A). In addition, GFAP, S100β, MAO-B, and MCT1 levels in reactive astrocytes were significantly reduced following KDS2010 treatment (Figure 4C–I, Supplementary Figure 2B–J). These results indicated that reactive astrogliosis is necessary for elevated MCT1 expression and 11C-acetate uptake. The MRI-measured tumor volumes were statistically significantly larger in the KDS2010-treated PDX mice than in nontreated mice on day 18 (Supplementary Figure 4).
Inhibition of MAO-B and MCT1 decreases reactive astrogliosis and 11C-acetate uptake in the PDX model. (A) Representative images of MRI, 11C-acetate PET, PET, and MRI fusion with or without KDS2010 administration in the PDX mouse model. (B) SUVr of 11C-acetate in the peritumoral and contralateral regions (n = 4). (C) Immunofluorescence images of GFAP expression in mouse models. (D and G) Immunofluorescence images of S100β, MCT1, GFAP, and MAO-B expression. (E, F, H, and I) Quantification of GFAP, S100β, MCT1, and MAO-B expression ± KDS2010 administration in PDX mouse models (n = 4). (J) Representative images of MRI, 11C-acetate PET, PET, and MRI fusion with or without astrocyte-specific MCT1 gene-silencing around PDX glioblastoma cells. (K) SUVr of 11C-acetate in the peritumoral and contralateral regions (n = 4). (L) Immunofluorescence images showing GFAP expression in mouse models. (M and P) Immunofluorescence images showing S100β, MCT1, GFAP, and MAO-B expression. (N, O, Q, and R) Quantification of GFAP, S100β, MCT1, and MAO-B expression ± astrocyte-specific MCT1 gene-silencing of PDX glioblastoma cells (n = 4). Scale bars, 500 μm (C), (L); 10 μm (D), (G), (M), (P). Data are presented as mean ± SEM. *P < .05, **P < .01, ***P < .001, ****P < .0001 by unpaired 2-tailed t-test (B) or Mann–Whitney U-test (E–I, N–R).
To investigate whether astrocytic MCT1 is necessary for 11C-acetate uptake, we performed Cre-loxp-dependent astrocyte-specific gene silencing of MCT1 using AAV-GFAP-cre-mCherry and AAV-pSico-rMCT1sh-GFP viruses around the tumor in the PDX model and U87MG model. We found that MCT1 gene silencing significantly decreased 11C-acetate uptake on microPET images (Figure 4J and K, and Supplementary Figure 2K and O) and 14C-acetate uptake as visualized by autoradiography in the peritumoral regions (Supplementary Figure 2K). In addition, the levels of GFAP, S100β, and MAO-B were significantly reduced by astrocytic MCT1 (Figure 4L–R, Supplementary Figure 2L–T), indicating that astrocytic MCT1 is necessary for elevated 11C-acetate uptake and expression of reactive astrogliosis-related proteins. Overall, these results suggested that elevated 11C-acetate uptake requires both reactive astrogliosis and astrocytic MCT1 at the glioblastoma boundary.
Relationship Between Volume of Reactive Astrogliosis and Prognosis in Patients with Glioblastoma and Cancer Stem Cells
We identified Ki-67-positive cells within the regions of reactive astrogliosis in glioblastoma tissues, whereas these cells were absent in astrocytoma and IDH1-mt tissues (Figure 1F–I, and Q). Subsequently, we performed additional analyses to investigate whether these Ki-67-positive cells within regions of reactive astrogliosis were associated with a poor prognosis in glioblastoma. Ki-67-positive cells were observed within the regions of reactive astrogliosis in human glioblastoma tissue. Consequently, we conducted immunostaining for CD133, a potential biomarker of cancer stem cells,34,35 to determine whether Ki-67 positive cells represented proliferating reactive astrocytes or cancer stem cells. Interestingly, we found that Ki-67-positive cells largely colocalized with CD133 (Figure 5A and B), suggesting the existence of cancer stem cells double-positive for Ki-67 and CD133 within regions of reactive astrogliosis at the glioblastoma tumor margin (S and L regions). Moreover, we found that GFAP-positive cells are not colocalized with CD133-positive cells in the patients with glioblastoma and PDX models (Figure 5B, and Supplementary Figure 3). Collectively, these results indicate the potential presence of cancer stem cells within regions of reactive astrogliosis.
Relationship between volume of reactive astrogliosis and prognosis in patients with glioblastoma and cancer stem cells. (A) Immunofluorescence images showing GFAP, Ki-67, and CD133 in the peritumoral region of glioblastomas. (B) The proportion of GFAP+ and/or CD133+ in Ki-67+ cells in glioblastomas (n = 3). (C) The 11C-acetate uptake area beyond the contrast-enhanced MRI was defined as VolumeACE–MRI. (D) Kaplan–Meier survival analysis showed significant differences in PFS between the 2 groups with a VolumeACE–MRI cutoff value of >5.06 cc (P = .0134). (E) Kaplan–Meier Survival analysis also showed significant differences in overall survival (OS) between the 2 groups with a VolumeACE–MRI cutoff value of >5.06 cc (P = .0206). Scale bars, 10 μm (A). Data are presented as mean ± SEM. ****P < .0001 by 1-way ANOVA with Tukey’s test (B).
To investigate the correlation between the prognosis of patients with glioblastoma and the volume of 11C-acetate uptake beyond contrast-enhanced MRI, which represents the area of reactive astrogliosis where cancer stem cells could potentially be present, we defined the 11C-acetate uptake area beyond contrast-enhanced MRI as VolumeACE-MRI (Figure 5C) and analyzed its correlation with progression-free survival (PFS) and OS. The optimal cutoff value for VolumeACE-MRI was >5.06 cc. The characteristics of the 2 VolumeACE-MRI groups, including age, sex, the extent of surgery, KPS, O6-methylguanine-DNA methyltransferase (MGMT), Ki-67, tumor size, progression, and death, are summarized in Table 1. There were no statistically significant prognostic factors between the 2 groups except for progression and death (Table 1). For univariate analysis, independent variables, such as patient age, sex, the extent of surgery, KPS, MGMT, Ki-67, tumor size, and VolumeACE-MRI were included. For PFS, the extent of surgery and VolumeACE-MRI were significant in the univariate and multivariate analysis (Supplementary Table 2). For OS, only VolumeACE-MRI was significant in the univariate and multivariate analysis (Supplementary Table 3).
Differences in Characteristics between 2 VolumeACE–MRI (the 11C-Acetate Uptake Area Beyond Contrast-Enhanced MRI) Groups
| Variables | VolumeACE-MRI ≤ 5.06 (n = 12) | VolumeACE-MRI > 5.06 (n = 26) | P-Value | |
|---|---|---|---|---|
| Age (y) | Median (range) | 49 (30–63) | 55 (25–78) | .1125 |
| Sex | Male | 6 | 18 | .2597 |
| Female | 6 | 8 | ||
| Extent of surgery | Total | 9 | 15 | .4722 |
| Subtotal | 3 | 11 | ||
| KPS | Median (range) | 90 (60–100) | 80 (70–100) | .6414 |
| MGMTa | Methylated | 6 | 7 | .1690 |
| Unmethylated | 6 | 19 | ||
| Ki-67 (%) | Median (range) | 9 (1–70) | 20 (3–50) | .1512 |
| Tumor size (cm) | Median (range) | 3.9 (2.1–8.0) | 5.0 (1.5–8.0) | .2262 |
| Progression | No progression | 4 | 0 | .0067 |
| Progression | 8 | 26 | ||
| Death | Survival | 6 | 0 | .0019 |
| Death | 6 | 26 | ||
| Variables | VolumeACE-MRI ≤ 5.06 (n = 12) | VolumeACE-MRI > 5.06 (n = 26) | P-Value | |
|---|---|---|---|---|
| Age (y) | Median (range) | 49 (30–63) | 55 (25–78) | .1125 |
| Sex | Male | 6 | 18 | .2597 |
| Female | 6 | 8 | ||
| Extent of surgery | Total | 9 | 15 | .4722 |
| Subtotal | 3 | 11 | ||
| KPS | Median (range) | 90 (60–100) | 80 (70–100) | .6414 |
| MGMTa | Methylated | 6 | 7 | .1690 |
| Unmethylated | 6 | 19 | ||
| Ki-67 (%) | Median (range) | 9 (1–70) | 20 (3–50) | .1512 |
| Tumor size (cm) | Median (range) | 3.9 (2.1–8.0) | 5.0 (1.5–8.0) | .2262 |
| Progression | No progression | 4 | 0 | .0067 |
| Progression | 8 | 26 | ||
| Death | Survival | 6 | 0 | .0019 |
| Death | 6 | 26 | ||
aMGMT, O6-methylguanine-DNA methyltransferase.
Differences in Characteristics between 2 VolumeACE–MRI (the 11C-Acetate Uptake Area Beyond Contrast-Enhanced MRI) Groups
| Variables | VolumeACE-MRI ≤ 5.06 (n = 12) | VolumeACE-MRI > 5.06 (n = 26) | P-Value | |
|---|---|---|---|---|
| Age (y) | Median (range) | 49 (30–63) | 55 (25–78) | .1125 |
| Sex | Male | 6 | 18 | .2597 |
| Female | 6 | 8 | ||
| Extent of surgery | Total | 9 | 15 | .4722 |
| Subtotal | 3 | 11 | ||
| KPS | Median (range) | 90 (60–100) | 80 (70–100) | .6414 |
| MGMTa | Methylated | 6 | 7 | .1690 |
| Unmethylated | 6 | 19 | ||
| Ki-67 (%) | Median (range) | 9 (1–70) | 20 (3–50) | .1512 |
| Tumor size (cm) | Median (range) | 3.9 (2.1–8.0) | 5.0 (1.5–8.0) | .2262 |
| Progression | No progression | 4 | 0 | .0067 |
| Progression | 8 | 26 | ||
| Death | Survival | 6 | 0 | .0019 |
| Death | 6 | 26 | ||
| Variables | VolumeACE-MRI ≤ 5.06 (n = 12) | VolumeACE-MRI > 5.06 (n = 26) | P-Value | |
|---|---|---|---|---|
| Age (y) | Median (range) | 49 (30–63) | 55 (25–78) | .1125 |
| Sex | Male | 6 | 18 | .2597 |
| Female | 6 | 8 | ||
| Extent of surgery | Total | 9 | 15 | .4722 |
| Subtotal | 3 | 11 | ||
| KPS | Median (range) | 90 (60–100) | 80 (70–100) | .6414 |
| MGMTa | Methylated | 6 | 7 | .1690 |
| Unmethylated | 6 | 19 | ||
| Ki-67 (%) | Median (range) | 9 (1–70) | 20 (3–50) | .1512 |
| Tumor size (cm) | Median (range) | 3.9 (2.1–8.0) | 5.0 (1.5–8.0) | .2262 |
| Progression | No progression | 4 | 0 | .0067 |
| Progression | 8 | 26 | ||
| Death | Survival | 6 | 0 | .0019 |
| Death | 6 | 26 | ||
aMGMT, O6-methylguanine-DNA methyltransferase.
Kaplan–Meier survival analysis demonstrated significant differences in PFS between the 2 groups, with a VolumeACE-MRI cutoff value of >5.06 cc (P = .0134, as shown in Figure 5D). Likewise, Kaplan–Meier survival analysis exhibited notable differences in OS between the 2 groups, using a VolumeACE-MRI cutoff value of >5.06 cc (P = .0206, as shown in Figure 5E). These findings suggest that a larger volume of 11C-acetate uptake correlates with poorer prognostic outcomes, underscoring the significant impact of the presence of cancer stem cells in the reactive astrogliosis region on prognosis.
Discussion
In this study, we investigated abnormal tumor–astrocyte interactions in glioblastoma, focusing on excessive acetate uptake in the TME. Previous studies have suggested glioma cells utilize acetate as a substrate for lipid metabolism.17,39 However, our findings indicated a more prominent acetate uptake by reactive astrocytes compared to glioma cells. We found that the aggressive proliferation of glioblastoma cells, driven by the monopolization of glucose uptake, led to the depletion of glucose in the surrounding astrocytes. Consequently, reactive astrocytes exhibit metabolic plasticity and utilize alternative metabolites such as acetate in the TME. Through a combination of in vitro experiments, mouse models, and human tissue analyses, we demonstrated high expression of MCT1 and increased acetate uptake by reactive astrocytes. These findings highlight the crucial role of metabolic plasticity in abnormal tumor–astrocyte interactions in glioblastoma, which share similarities with the metabolic pathways observed in neurodegenerative diseases and stroke recovery.4,5,24
Here, we introduce a novel method for visualizing reactive astrogliosis using 11C-acetate PET imaging in patients with gliomas. This imaging technique has several advantages over conventional MRI, which primarily focuses on the tumor itself.40–42 First, 11C-acetate PET accurately reflects the pathological grade of gliomas, as we previously demonstrated a positive correlation between the grade of glioma and the level of 11C-acetate uptake.27 Secondly, 11C-acetate PET imaging provides a clearer demarcation of the TME, particularly in areas of reactive astrogliosis that cannot be visualized using MRI.43 The presence of cancer stem cells at the leading edge of a tumor, where reactive astrogliosis occurs, plays a critical role in tumor expansion and invasion of the surrounding brain tissue.44 Our findings revealed a negative correlation between the volume of 11C-acetate uptake beyond contrast-enhanced MRI and prognosis (Figure 5D and E), emphasizing the importance of surgical resection of the reactive astrogliosis region to eliminate potential cancer stem cells characterized by the coexpression of Ki-67 and CD133. Therefore, 11C-acetate PET imaging represents a significant step toward visualizing the TME beyond the tumor itself.
The extent of surgical resection is a crucial prognostic factor in patients with glioblastoma, as residual cancer stem cells at the resection margin contribute to tumor recurrence.45 Previous studies have shown that patients who undergo supratotal tumor resection extending beyond the contrast-enhanced MRI region have better clinical outcomes.46–48 However, determining the supratotal resection margin based on MRI alone can be challenging because of the presence of scattered cancer stem cells, which may not be fully captured using MRI. Thus, techniques that enable clear visualization of the TME beyond the tumor are required to improve patient survival. Our study suggests that the prospective utilization of 11C-acetate PET imaging to guide tumor resection and assess prognosis provides the basis for promising future research.
Reactive astrocytes are involved in glial scar formation in glioblastoma as well. This involvement is characterized by a transformation into an elongated morphology,49 which is also observed in the line-shaped astrocytes within the L region of our study (Figure 1M). Research on astrocytic scarring in glioblastoma found that a decrease in reactive astrogliosis led to a reduced glial scar barrier and an increase in tumor volume.50 Our research supports these findings, showing a rise in tumor volume under similar conditions (Supplementary Figure 4). Furthermore, the diminished astrocytic barrier was associated with improved drug delivery into the tumor.50 These insights collectively underscore the potential of reactive astrogliosis to serve as a modifiable barrier, which could be targeted to restrict tumor expansion and improve drug delivery in glioblastoma.
We acknowledge the limitations of this study. Our human study was conducted at a single center, and future multicenter analyses are necessary to ensure the generalizability of our results. Additionally, the survival analysis was based on a relatively small sample size of only 38 patients with glioblastoma, calling for larger studies to validate our findings.
Conclusions
Our study revealed a connection between increased acetate uptake and reactive astrogliosis in glioblastomas. We proposed cellular mechanisms and developed imaging techniques to elucidate this relationship. Glioma cells producing excessive acetate trigger reactive astrogliosis in astrocytes adjacent to MCT1. Furthermore, 11C-acetate PET imaging is a promising tool for detecting reactive astrogliosis, which is the same approach that has been successfully applied in patients with AD and multiple sclerosis,24,51,52 with a larger volume of acetate uptake seen than in contrast-enhanced MRI, indicating poorer clinical outcomes. The concepts and techniques developed in this study have the potential to significantly enhance our understanding and management of glioblastoma.
Conflict of interest statement
None declared.
Funding
This work was supported by IBS-R001-D2 from the Institute for Basic Science funded by the Ministry of Science and ICT to C.J.L.; NRF-2018M3C7A1056898, NRF-2020R1A2B5B01098109, and RS-2022-00144475 from the National Research Foundation (NRF) of Korea to M.Y.; and NRF-2021R1C1C2011016 from the NRF to H.Y.K.
Acknowledgements
A schematic diagram (Figure 5C) was generated using Biorender.
Author Contributions
Conceptualization: M.Y., C.J.L., D.K., H.Y.K., J.C., J.H.C., S.K.; Methodology: D.K., H.Y.K., J.C., Y.M.P., M.Y., C.J.L., J.H.C., SK; Investigation: D.K., H.Y.K., J.C., Y.M.P., S.Y.K., J.K., M.L., K.H., Y.H.J., S.Y.L., S.P., H.L., J.P., G.H.L., H.K., S.K., U.P., H.R., J.S. and S.K.; Visualization: S.L., J.C., H.Y.K., D.K., Y.M.P.; Resources: J.H.C., K.S., M.Y., S.J.P., K.D.P., M.H.N., S.H.K.; Writing—original draft: D.K., H.Y.K., J.C., Y.M.P., Y.P.; Writing—review & editing: D.K., H.Y.K., J.C., Y.M.P., Y.P., C.J.L. and M.Y.; Writing—revised draft: D.K., H.Y.K., J.C., Y.M.P.; Writing—review & editing of revised draft: D.K., H.Y.K., J.C., Y.M.P., C.J.L. and M.Y.; Supervision by S.K., J.H.C., C.J.L., and M.Y.
Ethics Approval and Patient Consent Statement
This study was approved by the Institutional Review Board of Severance Hospital (4-2011-0697) and all patients provided written informed consent.
Clinical Trial Registration
This study was retrospectively registered under KCT0004468 on November 18, 2019.
Data Availability
The datasets generated and/or analyzed during the current study are available from the corresponding author upon request.
References
Author notes
Dongwoo Kim, Hae Young Ko, Jee-In Chung and Yongmin Mason Park contributed equally to this work.
