Targeting Catechol-O-Methyltransferase Induces Mitochondrial Dysfunction and Enhances the Efficacy of Radiotherapy in Glioma

Glioma is the dominant form of brain cancer, accounting for almost 80% of malignant cancers in the central nervous system (CNS; refs. 1, 2). There are three types of gliomas based on the cell of origin: astrocytoma, oligodendroglioma, and ependymoma (3). Glioblastoma multiforme (GBM) is the most common astrocytoma (∼80% total) with high-grade malignancy and aggressiveness (3, 4). Currently, treatment options for brain cancer remain very limited, with most patients undergoing surgical resection. Radiotherapy (RT) is often used to eliminate the remaining tumor cells. However, with RT, the median survival for GBM is about 12 months (5). The addition of chemotherapy, which consists mainly of temozolomide (TMZ), only increases median survival by around 2.5 months, with severe side effects (5–7). Therefore, discovering novel and effective therapeutic agents to enhance current GBM treatments is an unmet medical need.

We focused on the currently approved drugs to address this need because of the potential for quick clinical translation. Dopamine (DA) is an important neurotransmitter (8). In Parkinson’s disease (PD), the loss of neurons in the substantia nigra region of the midbrain leads to reduced DA synthesis and unintended movements (8). Past studies have suggested that genes involved in DA biosynthesis and signaling may also be involved in cancer development and treatment (9). Other studies have demonstrated an association between PD and brain cancers, although the connection remains controversial (10). Interestingly, genes involved in PD development have also been implicated in brain cancer. For example, about 10% patients with PD have genetic mutations, including SNCA, PARK2, and PARK6 that regulate mitochondrial functions. Mutations in these genes are also prevalent in different brain tumors (10, 11).

Because of the potential connection between PD and brain cancer, an interesting question is whether we can repurpose anti-PD drugs for brain cancer treatment. Such a possibility is attractive because of the excellent safety record of the anti-PD drugs demonstrated in several decades of clinical use (11). In addition, in support of the antitumor potential of anti-PD drugs, several anti-PD drugs have shown efficacy in non-brain cancer animal models. These drugs include benztropine, pimavanserin, and carbidopa (12–14).

One class of anti-PD drugs is catechol-O-methyltransferase (COMT) inhibitors. Their intended use is to slow down DA degradation (15). COMT is a key enzyme responsible for degrading DA in the brain (16). Entacapone, tolcapone, and opicapone are three FDA-approved COMT inhibitors (17). While entacapone and opicapone primarily work on peripheral neurons, tolcapone can efficiently cross the blood-brain barrier (BBB) to slow down DA metabolism in the brain (15). In this study, we investigated the involvement of COMT in GBM growth and evaluated the use of COMT inhibitors in combination with glioma radiotherapy.

We accessed human COMT expression in tumor tissues and normal tissues using The Cancer Genome Atlas (TCGA; RRID:SCR_003193) and Genotype-Tissue Expression (GTEx) portals (RRID: SCR_013042), respectively, and downloaded data via the Xena platform (RRID:SCR_018938; ref. 18). The expression levels were visualized and compared using GraphPad Prism version 9.5.1 (RRID:SCR_002798).

To identify enriched pathways of patients with GBM with high COMT expression, we obtained a cohort of data of patients with GBM from the TCGA PanCancer collection using an open-access online tool cBioPortal (http://www.cbioportal.org; RRID: SCR_014555; ref. 19). To determine coexpression, we ranked genes by Spearman correlation to COMT mRNA expression and selected the top 10% of genes showing a positive correlation with COMT for gene ontology (GO) analysis (RRID:SCR_002811; ref. 20). We used DAVID GO terms (https://david.ncifcrf.gov) to identify enriched biological pathways and plotted the top-ranked pathways with the “ggplot2” R package (RRID:SCR_014601). Scripts are available upon request.

We obtained CT2A cells from Dr. Peter Fecci’s laboratory at Duke University School of Medicine (RRID:SCR_011194). U87 (RRID:CVCL_0022) cells and HEK293T (RRID:CVCL_0063) from the Cell Culture Facility of Duke University School of Medicine. The immortalized C57BL/6J macrophage cell line was purchased from Kerafast (cat. #ENH167-FP, RRID: CVCL_C0MW). CT2A cells were cultured in high-glucose DMEM (Gibco; cat. #11995-065) supplemented with 10% heat-inactivated (HI) fetal bovine serum (FBS; HyClone Cytiva; cat. #SH30396.03) and 1× penicillin-streptomycin (Gibco; cat. #15140-122). U87 cells were cultured in Modified Eagle Medium (MEM; Gibco; cat. #11095-080) supplemented with 10% non-HI FBS, 1× MEM non-essential amino acids (Gibco; cat. #11140-050), 1 mmol/L sodium pyruvate (Gibco; cat. #11360-070), and 1× penicillin-streptomycin. C57BL/6J macrophage cell line was cultured in high-glucose DMEM supplemented with 10% HI-FBS and 2 mmol/L L-glutamine (Gibco; cat. #25030-081).

To make stable GFP-CT2A cells and tdTomato-C57BL/6J macrophages, we generated pLenti-CMV-tdTomato-Neo plasmid in-house by modifying pLenti-CMV-GFP-Neo (Addgene #17447; RRID:Addgene_17447). We made lentivirus in HEK293T cells by cotransfecting the respective lentiviral vector with second-generation packaging plasmids psPAX2 (Addgene #12260; RRID: Addgene_12260) and pMD2.G (Addgene #12259; RRID:Addgene_12259). We then transduced CT2A cells with GFP-encoded lentivirus and transduced C57BL/6J macrophages with tdTomato-encoded lentivirus for 24 hours. We selected cells in a complete cell growth medium with 1 mg/mL neomycin (G418) for 7 days and sorted out stable GFP⁺ CT2A cells and tdTomato⁺ C57BL/6J macrophages.

All cells were periodically subjected to Mycoplasma tests using the Universal Mycoplasma Detection Kit from ATCC (cat. #30-1012K), and only Mycoplasma-free cells were used for our experiments.

We generated knockout cells with genes of interest using the CRISPR/Cas9 gene editing system. We chose single-stranded guided RNA (sgRNA) sequences from CHOPCHOP (https://chopchop.cbu.uib.no), an online CRISPR sgRNA design tool. The sequence of sgRNAs for each gene is listed in Supplementary Table S1. We used the lentiCRISPRv2 vector (Addgene #52961; RRID:Addgene_52961) to generate constructs encoding sgRNAs following the protocol published from the Zhang lab (21). Briefly, LentiCRISPRv2 was cut with BsmBI (NEB, cat. #R0580) and gel purified using the Gene JET Gel Extraction kit (Thermo Fisher Scientific; cat. #K0692). Oligos encoding sgRNA sequences were phosphorylated, annealed, and subsequently ligated into digested LentiCRISPRv2.

To generate sgRNA-encoding lentivirus, we used HEK293T cells to cotransfect lentiviral constructs encoding the target sgRNAs with second-generation packaging plasmids psPAX2 and pMD2.G following the instruction from the Trono Lab (https://www.epfl.ch/labs/tronolab/laboratory-of-virology-and-genetics/lentivectorstoolbox/). Lentivirus produced from LentiCRISPRv2 empty vector, psPAX2, and PMD2.G was used as control.

To make stable gene-specific knockout (KO) cells, we infected cells with sgRNA-encoding CRISPR/Cas9 lentivirus for 24 hours and selected CT2A and U87 cells in a complete cell growth medium with 1 μg/mL puromycin for 7 to 10 days. We then prepared cell lysates to test the expression of target genes by immunoblot. Those cell populations showed significant downregulation of the target gene were then used to select for clonal knockout populations. To generate knockout cell single clones, we seeded cells in 96-well plates after a 7-day puromycin selection and screened for pure knockout clones verified by immunoblot analysis.

To generate COMT knockout in U87 cells, we digested the px330-mCherry (Addgene #98750; RRID: Addgene_98750) and pSpCas9 (BB)-2A-GFP (PX458; Addgene #48138; RRID:Addgene_48138) with BbsI (NEB; cat. #R0539S), and then gel-purified, and ligated the vectors with annealed oligos encoding sgRNA_1 or sgRNA_2 targeting human COMT. We transfected U87 cells with sgRNA-encoding constructs using Lipofectamine 2000 (Thermo Fisher Scientific; cat. #11668019) and incubated them for 24 hours. U87 cells cotransfected with empty px330-cherry and pSpCas9 (BB)-2A-GFP (PX458) were used as control. We then cultured the cells for 6 days using a complete MEM medium [MEM base medium with 10% non-heat inactivated FBS, 1× non-essential amino acid (NEAA), 1 mmol/L sodium pyruvate, and 100 U/mL penicillin-streptomycin (Gibco by Life Technologies; cat. #15140122)] before sending them for FACS sorting (Duke Cancer Institute flow cytometry core facility). After obtaining GFP⁺ mCherry⁺ cells, we seeded them in 96-well plates and screened for COMT-KO single clones. Pure KO clones were verified by immunoblot analysis. Supplementary Table S1 shows the sgRNA primer sequences used for targeting various genes in this study.

As previously described, we extracted total RNA using TRIzol Reagent (Ambion by life technologies; Cat #15596018) according to a protocol from StarrLab (https://sites.google.com/a/umn.edu/starrlab/protocols/rna/rna-isolation-using-trizol; ref. 22). Briefly, we rinsed the cells twice with ice-cold 1× PBS and added 1 mL TRIzol to lyse the cells. After scrapping the lysate off the Petri dish, we transferred the lysate to a 1.5 mL Eppendorf tube, incubated the lysate at room temperature for 5 minutes, and added 200 μL chloroform (Sigma; cat. #C2432) to the tube followed by vigorous vortex for 15 seconds. We then incubated the mixture at room temperature for 10 minutes, centrifuged it at 12,000 × g for 15 minutes at 4°C, and transferred the transparent top layer to a clean 1.5 mL tube. After the addition of 500 μL isopropanol, we further incubated the tube at room temperature for 10 minutes. The mixture was then centrifuged at 12,000 × g for 10 minutes at 4°C. A white pellet at the bottom should be visible. We removed the supernatant and washed the pellet with 75% ethanol and centrifuged the tube at 7,500 × g for 5 minutes at 4°C. The pellet was then allowed to air dry and dissolved in 85 μL RNase-free H₂O. Subsequently, we added 5 μL of TURBO DNase (Invitrogen by Thermo Fisher Scientific; cat. #AM2238) and 10 μL of 10× Turbo DNase buffer to degrade the remaining DNA at 37°C for 30 minutes. To stop the reaction, we added 200 μL chloroform again and repeated the RNA extraction procedures as described above. Finally, we dissolved total RNA in RNase-free H₂O and used it for quantitative PCR and bulk RNA sequencing analysis.

The RNA expression levels were quantified by qRT-PCR or qPCR as previously described (22). We used the TRIzol-extracted total RNA (described above) as the template for cDNA synthesis using random hexamer primers (Invitrogen by Thermo Fisher Scientific; cat. #SO142) and SuperScript II Reverse Transcriptase (Invitrogen Thermo Fisher Scientific; cat. #18064014) following the manufacturer’s instructions. Afterward, we performed qRT-PCR of the cDNA using qPCRBIO SyGreen Blue Mix Hi-ROX (Genesee Scientific; cat. #17-506C) and the Applied Biosystems ViiA seven Real-Time PCR System with 384-well Block (Thermo Fisher Scientific; cat. #4453536). We used the comparative Ct (ΔΔCt) method to normalize the expression of individual genes to β-actin and to compare the relative changes in gene expression among different groups.

To quantify cytosolic DNA levels, we obtained cytosolic extracts as described above. We then conducted qPCR analysis using the cytosolic extracts as templates. To quantify cytosolic mitochondria DNA (mtDNA) and nuclear DNA (nucDNA) levels for any individual gene, we set the ratio between the cytosolic DNA level and the whole cell lysate as in the control cells as 1 and used it to obtain the relative levels of cytosolic DNA in other cells.

Supplementary Table S2 lists the primers used for qRT-PCR and qPCR analysis of different target genes.

Anti-COMT (cat. #14754-1-AP, for immunofluorescent microscopy, RRID:AB_2083706), anti-RIG-1/DDX58 (cat. #20566-1-AP; RRID:AB_10700006), and anti-MAVS (cat. #14341-1-AP; RRID:AB_10548408) antibodies were purchased from Proteintech. Anti-MAVS (D5A9E, cat. #24930; RRID:AB_2798889), anti-MDA5 (D74E4, cat. #5321; RRID:AB_10694490), anti-STING (D2P2F; cat. #13647; RRID:AB_2732796), anti-TBK1/Nak (D1B4, cat. #3504, RRID:AB_2255663), anti-pTBK1/pNak (Ser172; D52C2, cat. #5483; RRID:AB_10693472), anti-IRF-3 (D83B9; cat. #4302; RRID:AB_1904036), anti-IRF7 (D2A1J, cat. #13014; RRID:AB_2737060), anti-HSP60 (D307; cat. #4870; RRID:AB_2295614), and anti-H3 (D1H2, cat. #4499; RRID:AB_10544537) antibodies were purchased from Cell Signaling Technology. Anti-actin (ACTN05-C4; cat. #MA5-11869; RRID:AB_11004139) and anti-COMT (cat. #PA5-76864; RRID:AB_2720591) antibodies were purchased from Thermo Fisher Scientific. Anti-J2 (cat. #10010200; RRID:AB_2651015) antibody was purchased from SCICONS. Anti-DNA (AC-30-10; cat. #CBL186; RRID:AB_11213573) and antiphospho-histone H2A.X (Ser139) antibody (clone JBW301, cat. #05-636; RRID:AB_309864) were purchased from Millipore Sigma. Alexa Fluor 488 goat anti-mouse IgG (cat. #A28175; RRID:AB_2536161) and Alexa Fluor 555 goat anti-rabbit IgG (cat. #A27039; RRID:AB_2536100) were purchased from Invitrogen by Thermo Fisher Scientific.

For in vivo experiments, rat IgG2 isotype (clone 2A3, cat. #BE0089; RRID:AB_1107769), rat anti-mouse aPD-1 (clone RMP1-14, cat. #BE0146; RRID:AB_10949053), mouse IgG1 isotype (clone MOPC-21; cat #EB0083; RRID:AB_1107784), mouse αIFNAR1 (clone MAR1-5A3; cat. #BE0241; RRID:AB_2687723), αCD4 (Clone GK1.5; cat. #BE0003-1; RRID:AB_1107636), αCD8b (Clone 53-5.8; cat. #BE0223; RRID:AB_2687706), and αNK1.1 (Clone PK136; cat. #BE0036; RRID:AB_1107737) antibodies were purchased from BioXCell.

For immunohistochemical staining, anti-IBA1 (cat. #019-19741; RRID:AB_839504) was purchased from FUJIFILM Wako Pure Chemical Corporation. Anti-CD68 (Clone FA-11, cat. #137002; RRID:AB_204400) was purchased from BioLegend.

Entacapone (cat. #HY-14280) and tolcapone (cat. #HY-17406) were purchased from MedChemExpress. Actinomycin D (Act-D; cat. #A1410), a transcriptional inhibitor to deplete mitochondrial RNA (mtRNA), and ROS inhibitor N-acetyl-L-cysteine (NAC; cat. #A7250) were purchased from Millipore Sigma. MitoTracker Red CMXRos (cat. #M7512) was purchased from Invitrogen.

We prepared cellular lysates by washing cells twice with ice-cold 1× PBS and adding in 1× RIPA buffer (Sigma; cat. #R0278) supplemented with 1× protease inhibitors (Sigma; cat. #P8340) to detach cells from the culture dishes using a cell scraper. We then collected the lysates in a 1.5 mL tube, incubated them on ice for 10 minutes, and centrifuged them at 15,000 rpm for 15 minutes at 4°C. We transferred the supernatants to new 1.5 mL tubes, measured protein concentrations in the lysates, and adjusted them to the same levels. We boiled the lysates for 5 minutes and loaded equal amounts of them into SDS PAGE gels. After electrophoresis, we used the wet transfer chamber system to transfer the proteins from gels to PVDF membranes, followed by incubating the membranes with blocking buffer (5% milk in 1× TBS with 1% Tween 20) for 1 hour at room temperature. We then added primary antibodies to the 5% BSA in 1× TBST with 0.2% NaN3 and incubated the PVDF membranes overnight at 4°C. We washed the membranes three times on a shaker with blocking buffer, with 10 minutes per wash, then incubated the membranes with HRP-conjugated secondary antibodies at room temperature for 1 hour. After washing the membranes three times with 1× washing buffer (1× TBST; 1× TBS with 1% Tween 20), we incubated the membranes with the SuperSignal West Pico PLUS Chemiluminescent Substrate (Thermo Fisher Scientific; cat. #34580) and detected target signals with the Odyssey Fc imaging system (LI-COR Biosciences; RRID:SCR_023227).

Duke University Institutional Animal Use and Care Committee–approved all mouse experiments in this study. We purchased six-week-old C57BL/6J (Strain #000664; RRID:IMSR_JAX:000664) female mice from the Jackson Laboratory for all in vivo tumor growth experiments. For experiments using subcutaneous mouse models, we injected about 5 × 10⁵ CT2A cells (resuspended in 50 μL 1× PBS) to the right flanks of syngeneic mice. Some mice receiving radiotherapy were exposed to 8 Gray (Gy) radiation at the right flank using the CIX3 irradiator from Xstrahl. In other mice, fractionated radiation was given as five consecutive daily 2 Gy treatments. We measured the length and width of the tumors every 2 to 3 days using a digital caliper and calculated the tumor sizes using the formula: (length) × (width)²/2. The endpoint for each mouse was when its tumor volume reached 2,000 mm³.

We also conducted brain tumor growth experiments using an orthotopic mouse model via implantation into the right caudate nucleus (2–2.5 mm to the right of bregma at a depth of 3.5 mm). We measured the baseline weight of animals before intracranial injection with 1 × 10⁵ CT2A cells in 5 μL 1× PBS with 2% methylcellulose. For radiotherapy, the mice with VC and comt-KO brain tumors were treated with a dose of 6 Gy radiation at the head region, while those with vehicle and drug treatments were given 8 Gy radiation. Body weight loss is an objective and quantitative measurement for evaluating the endpoint of the experiment, which is set as more than 15% body weight loss from the highest record of each mouse. We measured the weights of mice every 1 to 2 days and monitored the post-surgery status of mice carefully for signs of pain, symptoms of distress, and tumor formation (including hunched posture, arched back, circling, walking on their tiptoes, balance issues, eyes that are not opened, hyperactivity, seizures, head doming). The animals with the onset of behavioral or physical impairments were euthanized by CO₂ and decapitation, followed by tissue harvest for analysis.

In experiments involving αIFNAR1 antibody treatments, we pretreated mice with 200 μg mouse IgG1 isotype control (clone MOPC-21; BioXCell; cat. #EB0083; RRID:AB_1107784) or 200 μg mouse αIFNAR1 antibody (clone MAR1-5A3; BioXCell; cat. #BE0241; RRID:AB_2687723) in 150 μL 1× PBS per mouse by intraperitoneal injection 1 day prior to subcutaneous tumor implantation. Control or αIFNAR1 antibodies were then given to tumor bearing mice on day 2, 5, 8, and 11 post tumor engraftment procedures.

To deplete CD4⁺ T cells, CD8⁺ T cells, and NK cells in vivo, mice were administered intraperitoneally with 100 μg αCD4 antibody (Clone GK1.5; BioXCell; cat. #BE0003-1; RRID:AB_1107636), 100 μg αCD8b antibody (Clone 53–5.8; BioXCell; cat. #BE0223; RRID:AB_2687706), and 100 μg αNK1.1 antibody (Clone PK136; BioXCell; cat. #BE0036; RRID:AB_1107737), respectively, 1 day before tumor implantation. After subcutaneous injection of tumor cells to the right flank of mice, three more doses of antibody treatments were given by intraperitoneal injection on day 2, 5, and 8, respectively. We also administered an equal amount of IgG isotype antibodies to control mice.

In experiments with drug treatments, we intraperitoneally administrated 30 mg/kg entacapone daily to CT2A-tumor bearing mice from day 3 to the endpoint for subcutaneous mouse models. For orthotopic CT2A brain tumor-bearing mice, we treated mice with 30 mg/kg tolcapone twice a day by intraperitoneal injection from day 3 to day 35. Animal conditions (including changes of behaviors, activities, and body weights) were carefully monitored during and after treatments.

To measure the cell proliferation and sensitivity to radiation in control and COMT-deficient CT2A cells, 2 × 10³ irradiated or non-irradiated cells were seeded in 12-well plates and incubated at 37°C with 5% CO2 for 7 to 9 days. In addition, 10 mmol/L NAC and 10 μg/mL anti-IFNAR1 antibodies were used to study the impact of ROS and type I IFN signaling on cell growth. Cell numbers were counted every 24 hours using the T20 automated cell counter (Bio-Rad; cat. #1450102). Cell growth curves were demonstrated with log scales.

We isolated cellular mitochondrion using a mitochondria isolation kit for cultured cells (Abcam; cat. #110170) according to the manufacturer’s instructions. Approximately 2 × 10⁶ cells were seeded 1 day prior to use. On the day of the experiment, we washed cells with 1× PBS twice and detached cells from the plate using a scrapper. Cells were then frozen and thawed to weaken the cellular membrane and resuspended in 400 μL reagent A, and homogenized by pipetting up and down harshly for 30 to 40 times. After incubation on ice for 10 minutes, cells were centrifuged at 1,000 × g at 4°C for 10 minutes. The supernatant was collected in a new tube. The pellet was resuspended in 400 μL reagent B, homogenized, incubated, and centrifuged as described for reagent A. Supernatant was collected and mixed with the supernatant from reagent A. The mixture was then centrifuged at 12,000 g for 10 minutes to get the mitochondria pellet. We resuspended the pellet in reagent C supplemented with protease inhibitor (Sigma; cat. #P8340) for further analysis.

We used the Complex I enzyme activity microplate assay kit (Abcam; cat. #ab109721) to determine the activities of mitochondrial oxidative phosphorylation (OXPHOS) complex I, which is quantified based on the oxidation of NADH to NAD⁺ and the corresponding reduction of a dye that increases the absorbance at OD = 450 nm.

Here, Q1 is Red⁺Green⁻ Quadrant; Q2 is Red⁺Green⁺ Quadrant; and Q3 is Red⁻Green⁺ Quadrant.

We tested ROS levels of cells using DCFDA/H2DCFDA cellular ROS assay kit (Abcam; cat. #Ab113851) by flow cytometry. 2′-7′-Dichlorodihydrofluorescein diacetate (DCFDA) is a cell-permeable fluorogenic probe that can directly measure the redox status within cells. The addition of DCFDA to cells allows them to diffuse in cells. A non-fluorescent product is produced from DCFDA by interacting with cellular esterase, which is then oxidized by ROS into a highly fluorescent compound, 2′, 7′ –dichlorofluorescein, that can be detected with excitation/emission at 485 nm/535 nm (24). Here, we treated cells with 20 µmol/L DCFDA in 1× ROS buffer and incubated them at 37°C in dark for 30 minutes before proceeding to flow cytometry using FITC. Tert-butyl hydroperoxide is commonly used as an oxidant, therefore working as a positive control in this assay with the final concentration of 100 µmol/L.

For in vitro COMT inhibitor treatments, we treated cells with 50 µmol/L entacapone or 50 µmol/L and/or 100 µmol/L tolcapone for 24 hours prior to sample collection and experiments. For cell colony formation assays, CT2A cells were incubated in the complete DMEM medium supplemented with 50 µmol/L entacapone, which was changed every 3 days.

To inhibit mitochondrial RNA transcription, we treated cells with 0.5 μg/mL Act-D and incubated for 1 hour as described previously (25).

To attenuate ROS production, we pretreated cells with 10 mmol/L antioxidant N-acetylcysteine (NAC) for 16 hours before experiments.

To inhibit IFNα/β pathway in vitro, we pretreated CT2A cells with 10 μg/mL anti-IFNAR1 (αIFNAR1).

The subcellular fractionation from cells for cytoplasmic mitochondrial DNA was conducted using a protocol as described in previous publications (22, 26). Briefly, we seeded cells in 100 mm Petri dishes 2 days before the experiments so that the confluence could reach 70% on the day of experiments. We prepared two equal parts of 4 × 10⁶ cells from each plate. We then suspended one aliquot in 500 μL of 50 mmol/L NaOH, boiled the lysates for 30 minutes and used it to serve as total mtDNA control from whole cell lysate (WCL). Next, we resuspended the other aliquot in 500 μL of solution with 25 μg/mL of digitonin (Millipore Sigma; SKU #D141-100MG) in a buffer with 150 mmol/L of NaCl, 50 mmol/L HEPES at pH 7.4, and homogenized the cells by vigorous pipetting, and followed by a 10-minute incubation period on ice to allow selective permeabilization of the plasma membrane. We then centrifuged the homogenates at 980 × g three times for 3 minutes each at 4°C to pellet the intact cells. After the centrifugation, we rinsed the pellet with 1× PBS and used it as the pellet (Pel) fraction for immunoblot. Next, the supernatant was transferred to a new tube and centrifuged at 17,000 × g for 10 minutes at 4°C to spin down the remaining cellular debris. The supernatant from this centrifugation was transferred to a new tube and used as the cytosolic (Cyto) fraction free of nuclear, mitochondrial, and endoplasmic reticulum contamination. We then boiled the WCL, Pel, and Cyto lysates at 95°C for 5 minutes and loaded the lysates for western blot analysis to ensure the cytosolic preparation was free of contamination. Finally, we purified and concentrated both WCL_DNA and Cyto_DNA using DNA Clean & Concentrator-5 Kit (ZYMO RESEARCH; cat. #D4004) and used the DNA for quantitative PCR (qPCR) tests.

To extract cytoplasmic mitochondrial RNA, around 4 × 10⁶ Cells were seeded in 100 mm dish 2 days prior to experiments. On the day of the experiment, cells were washed with ice-cold 1× PBS twice and detached from plates using a scrapper. We saved some cells for WCL. Cells were spun down and resuspended in 300 μL 25 μg/mL digitonin buffer with the addition of RNase inhibitor (Promega; cat. #N2111; 40 U/μL; final concentration = 1 U/μL), then homogenized by vicious pipetting (20–30 times) following 10-minute incubation on ice to allow selective plasma permeabilization. The homogenates were then centrifuged and proceeded to cytoplasmic fractionation as described above. The ultimate supernatant was collected to a new tube and served as the cytosolic (Cyto) fraction that was free of nuclear, mitochondrial, and endoplasmic reticulum contamination. WCL, Pel, and Cyto lysates were boiled at 95°C for 5 minutes for immunoblot validation of pure cytosolic preparation. The cytosolic fraction was then treated with TURBO DNase (Invitrogen by Thermo Fisher Scientific; cat. #AM2238) and incubated in a 37°C water bath for 0.5 hours to remove genomic DNA. Cytoplasmic RNA was then isolated following the procedures as described in section total RNA extraction and dissolved in 15 μL RNase-free H₂O.

We collected the whole brains of control and Comt-KO tumor-bearing mice at endpoints and fixed them in 10% formalin for 48 hours with subsequent transfer to 70% ethanol for long-term storage. The samples were sent to the Research Immunohistology Lab at Duke Pathology Department for paraffin processing, embedding, sectioning, and hematoxylin and eosin (H&E) staining.

Paraffin-embedded sections were deparaffinized sequentially as follows: 100% xylene, 10 minutes for three times; xylene and ethanol (1:1), 10 minutes; 100% ethanol, 5 minutes for twice; 95% ethanol, 5 minutes; 70% ethanol, 5 minutes; 50% ethanol, 5 minutes. We then placed slides in running cold tap water to rinse off ethanol and subjected the tissues to antigen retrieval by immersing slides in boiling sodium citrate buffer (10 mmol/L sodium citrate, 0.05% Tween 20, pH 6.0) for 20 minutes. We then rinsed the slides in running tap water for 10 minutes and blocked them with 10% goat serum at room temperature for 1 hour before immunohistochemical staining. We diluted rabbit anti-IBA1 (1:500, Wako, cat. #019-19741; RRID:AB_839504) and rat anti-CD68 (1:500, Clone FA-11, BioLegend, cat. #137002; RRID:AB_2044004) in TBS with 1% BSA and incubated slide with diluted antibodies overnight at 4°C. After washing slides with 1× TBS three times, with each for 5 minutes, we added Alexa 555 anti-rabbit (1:500) and Alexa 488 anti-rat (1:500) to the slides and incubated them at room temperature for 1 hour in the dark. The slides were washed with 1× TBS again three times and stained by VECTASHIELDantifade mounting medium with DAPI (Vector Laboratories, cat. #H-1200-10). Fluorescent images were acquired using the 20× objective of the Zeiss fluorescence microscope (Axio Observer Z1; RRID:SCR_021351).

Samples for imaging were prepared as previously described (22). Briefly, we seeded the cells in 35 mm glass-bottomed poly-D-lysine-coated dishes (MatTek Life Sciences; cat. #P35G-1.5-10-C) 2 days before experiments. We used 4% praformaldehyde (PFA) to fix cells at room temperature for 15 minutes and permeabilized the cells using 0.5% Triton X-100 in 1× PBS at room temperature for 10 minutes. We then washed the cells three times with PBS and blocked them with 5% bovine serum albumin (BSA; Sigma; cat. #A3983) at room temperature for 1 hour. Next, we added primary antibodies and incubated the cells at 4°C overnight, followed by adding fluorophore-conjugated secondary antibodies after washing with 1× PBS 3 times. Next, we incubated the cells at room temperature for 1 hour in the dark and washed them three times. Finally, we added VECTASHIELD Antifade Mounting Medium with DAPI (VECTOR LABORATORIES; cat. #H-1200-10) to the glass bottom of the dish before analysis. We took fluorescence images using the Leica TCS SP5 laser scanning confocal microscope (RRID:SCR_020233) in the Light Microscopy Core Facility of Duke University School of Medicine.

We seeded 1 × 10⁵ tdTomato-C57BL/6J macrophages in 35 mm glass-bottomed poly-D-lysine-coated dishes and stimulated cells with 20 ng/mL IFNγ and 1 μg/mL LPS for 24 hours. We then added GFP-labeled CT2A tumor cells to the dish, generating a 5:1 (tumor: macrophage) cell mixture. After 2 hours of coculture at 37°C, unattached cells were washed off using a complete medium. We fixed cells with 4% PFA at room temperature for 15 minutes and washed the dishes three times with 1× PBS. Fluorescent images were taken using 20× objective (For quantification) and 40× objective (For representing demonstration) of Leica TCS SP5 laser scanning confocal microscope in the Light Microscopy Core Facility of Duke University School of Medicine.

We plated 1 × 10⁵ tdTomato-C57BL/6J macrophages in 35 mm glass-bottomed poly-D-lysine-coated dishes and stimulated cells with 20 ng/mL IFNγ and 1 μg/mL LPS for 24 hours. We then added 5 × 10⁵ GFP-VC and GFP-Comt-KO CT2A cells to the respective dishes and cocultured them with macrophages at 37°C for 2 hours, then removed unattached cells using the complete medium. Finally, we anchored the dishes to the cassette within the 37°C incubator (supplied with 5% CO₂) connected to the Olympus VivaView FL incubator microscope in the Light Microscopy Core Facility of Duke University School of Medicine. Live-cell imaging was recorded every 5 seconds for 6 minutes.

We performed genome-wide transcriptome analysis of VC and COMT-KO U87 cells using a protocol as previously described (22). First, we prepared total RNAs from the cells using TRIzol Reagent as described above. We then submitted our RNA samples to the Duke University Sequencing and Genomic Technologies Core Facility (RRID:SCR_017748) for sequencing, which QCed the samples and prepared cDNA libraries for analysis using Illumina NovaSeq 6000 (RRID:SCR_016387). We processed the RNA-seq data using the Trim Galore toolkit (RRID:SCR_011847), which employs Cutadapt to trim low-quality bases and Illumina sequencing adapters from the 3′ end of the reads (27). Only reads that were 20 nt or longer after trimming were kept for further analysis. Reads were mapped to the GRCh38.p13 of the human genome and transcriptome using the STAR RNA-seq alignment tool (RRID:SCR_004463; ref. 28). Reads were kept for subsequent analysis if they mapped to a single genomic location using the SAMtools (RRID:SCR_002105; ref. 29). Gene counts were compiled using the HTSeq tool (RRID:SCR_005514; ref. 30). Only genes that had at least 10 reads in any given library were used in subsequent analysis. Normalization and differential expression were carried out using the DESeq2 Bioconductor package with the R statistical programming environment (RRID:SCR_001905). Software for gene set enrichment analysis (GSEA; version 4.1.0; RRID:SCR_003199) was used to identify differentially regulated pathways. The source RNA-seq data will be deposited in the NCBI’s Gene Expression Omnibus (GEO) database (RRID:SCR_006472; RRID:SCR_00501).

Brain tumors and their surrounding tissues were dissected from the orthotopic mouse models and collected from two independent experiments. Control tumors were collected at 12 days post injections (n = 7), and Comt-KO tumors (n = 3 for each time point) were collected at 12 and 25 days post injections, respectively. Tissues were digested with 2 mg/mL collagenase D and 30 μg/mL DNase I for 45 minutes at 37°C with agitation. Single-cell suspensions containing brain tumor cells were passed through 70 μm filters and stained with LIVE/DEAD aqua (Invitrogen; Cat #L34957) for viability, APC anti-mouse CD45 (Clone 30F11; BioLegend; cat. #103112; RRID:AB_312977), and BD Pharmingen PE rat anti-mouse H-2 class I (Clone M1/42; BD Biosciences; cat. #566776; RRID:AB_2869859) and processed with single-cell multiplexing kit (BD). Viable CD45⁺ single cells were sorted using BD Biosciences FACSAria III Cell Sorter (RRID:SCR_016695) and captured with BD Rhapsody single-cell analysis system according to the manufacturer’s protocols. The cDNA libraries were prepared using BD Rhapsody whole transcriptome analysis amplification kit. Libraries of vector control and Comt-KO were pooled and sequenced more than 100,000 unique reads per cell using Illumina NovaSeq6000 with the read length of 100 base-pairs and paired-end at Duke University Sequencing and Genomic Technologies Core Facility (RRID: SCR_017748).

Raw sequencing data were processed using BD Rhapsody Seven Bridges pipelines. Briefly, FASTQ files were demultiplexed and aligned to the mouse genome reference mm10. Feature barcode processing and unique molecular identifier (UMI) counting were then performed according to the standard workflow. Single-cell data were further analyzed using Seurat 4 (RRID:SCR_007322; ref. 31) with the following exclusion criteria: (i) cells that had fewer than 200 UMI counts or genes that were expressed by fewer than three cells were removed from analysis; (32) cells that had more than 10,000 UMI counts or greater than 25% of mitochondrial genes were also excluded. After filtering, 3,000 variable features were identified. Top 30 PCs were used to generate Uniform Manifold Approximation and Projection (UMAP) clustering. Myeloid cells were extracted for subclustering analysis with Seurat 4, pseudotime analysis with Monocle 3 (RRID:SCR_018685) and gene module analysis (33). Gene ontology database (RRID:SCR_002811) of biological process was used for pathway enrichment analysis with top 200 marker genes ranked by average fold change of tumor-associated macrophages and microglia. A pathway was considered significantly over-presented with FDR < 0.05.

We referred to a previously established method for brain tumor-infiltrating macrophages analysis by flow cytometry (34). Briefly, brain tumors were collected from mouse brains on day 12 post-tumor implantation for vector controls and 25 for Comt-KO group. Single-cell suspensions were prepared as described in single-cell RNA sequencing (scRNA-seq). Cells were then stained with LIVE/Dead Aqua viability dye in 1× PBS on ice for 30 minutes and washed with 1× PBS. Blocking buffer containing 1% Fc block, 5% normal mouse serum, and 5% normal rat serum was used to incubate cells at room temperature for 10 minutes, followed by cell surface staining at room temperature for 20 minutes with a master mix of fluorophore-conjugated primary antibodies against mouse CD45 (Clone 30F11; BioLegend; cat. #103106; RRID:AB_312971), CD11b (Clone M1/70; BioLegend; cat. #101233; RRID:AB_10896949), CD11c (Clone N418; BioLegend; cat. #117336; RRID:AB_2565268), Ly6C (Clone HK1.4; BioLegend; cat. #128018; RRID:AB_1732082), Ly6G (Clone 1A8; BioLegend; cat. #127622; RRID:AB_10643269), CD64 (Clone X54-5/7.1; BioLegend; cat. #139309: RRID:AB_2562694), F4/80 (Clone BM8; eBiosciences; cat. #17-4801-82; RRID:AB_2784648), and MHC class II (Clone M5/114.15.2; BD Biosciences; cat. #563415; RRID:AB_2738192). Cells were washed in 1× PBS, fixed with 0.4% paraformaldehyde in 1× PBS, and subjected to data acquisition with BD FACSAria III flow cytometer (RRID:SCR_016695). Data were processed and analyzed by FlowJo V10 software (RRID:SCR_008520).

Statistical analysis was conducted using GraphPad Prism 8.2.0 software (RRID:SCR_002798). Two-sided Student t test was used to compare two experimental groups. One-way ANOVA with Tukey post hoc test was applied to compare complex I enzymatic activities, ROS production, gene expression levels, and phagocytosis rates among multiple groups. Two-way ANOVA was applied to compare in vitro cell proliferation and in vivo tumor growth rates within two or more experimental groups. mtRNA and nucRNA levels in VC, COMT-KO, Act-D-treated VC and Act-D–treated COMT-KO, or NAC-treated VC and NAC-treated COMT-KO U87 cells were also analyzed using two-way ANOVA with Tukey post hoc tests. Log-rank (Mantel–Cox) test was used for mouse and human patient survival analysis. *, P < 0.05 was considered statistically significant. Please see Supplementary Table S3 for post hoc power analysis of the tumor growth delay experiments performed in this study. We also used the time for tumor size to reach 500 mm³ to calculate tumor growth delays (Supplementary Table S4).

Raw and metadata for bulk RNA-seq and scRNA-seq are deposited in the GEO database (RRID:SCR_005012) with the accession numbers GSE237053 and GSE238199. Analytic scripts for bulk RNA-seq and scRNA-seq are available upon request. All other raw data are available upon request from the corresponding author. In addition, the authors used publicly available data generated by others. Specifically, COMT expression data used in Fig. 1 were obtained from TCGA at http://www.cbioportal.org and GTeX at https://gtexportal.org/home.

We first compared the expression of COMT in brain tumor tissues and normal brain tissues using data from TCGA and the genotype-tissue expression (GTEx) project, respectively. Our analysis showed higher expression of COMT in both GBM and low-grade glioma (LGG) tissues compared to normal brain tissues (Fig. 1A and B). To test whether COMT affects brain tumor cell growth, we generated Comt knockout (KO) CT2A cells, an aggressive murine glioma cell line (Supplementary Fig. S1A). Deletion of Comt attenuated CT2A cell growth in vitro (Supplementary Fig. S1B). We also examined the sensitivity of Comt wild-type (wt) and Comt-deficient CT2A cells to radiation. We found that cell growth was delayed in control cells after exposure to radiation in a dose-dependent manner (Supplementary Fig. S1B). Tumor growth was further delayed in Comt-KO CT2A cells after radiation, indicating that targeting COMT could increase glioma cell sensitivity to radiation (Supplementary Fig. S1B). To confirm this result, we treated cells with entacapone, a clinically available COMT inhibitor, and measured in vitro cell growth alone or in combination with radiation. Consistently, COMT inhibition by entacapone further inhibited cell growth in addition to radiation (Supplementary Fig. S1C). As radiation can directly cause DNA damage, we then investigated whether inhibiting COMT leads to additional DNA breaks beyond radiation-induced DNA damage, which ultimately results in cell growth arrest. To this end, we compared the γH2AX foci, a sensitive marker for DNA double-stranded breaks, in control and COMT-deficient cells before and after radiation. We noticed that COMT inhibition alone increased the number of γH2AX foci. Upon radiation exposure, cells deficient in COMT showed a further increase in γH2AX foci, indicating additional DNA damage and reduced cell fitness (Supplementary Fig. S1D and S1E).

We next examined Comt deletion’s influence on tumor growth in vivo. First, we established subcutaneous vector control (VC) and Comt-KO CT2A tumors in syngeneic C57BL/6 mice and evaluated their response to RT following the schedule in Fig. 1C. Comt deletion was sufficient to suppress tumor growth by itself with three out of six mice cured (Fig. 1D). Furthermore, RT cured all mice with Comt-KO CT2A tumors (Fig. 1D). Next, we established orthotopic brain tumors by intracranially injecting VC and Comt-KO glioma cells. As shown in Fig. 1E, radiotherapy was administered 9 days post intracranial injection of CT2A cells. Consistent with the subcutaneous tumor model results, one out of five mice bearing intracranial Comt-KO CT2A tumors survived long-term (>90 days). Importantly, Comt deletion significantly enhanced intracranial tumor sensitivity to RT, leading to four out of five mice remaining tumor-free for long term (Fig. 1F; Supplementary Fig. S1F).

To understand the biological pathways affected by COMT inhibition, we first performed Gene Ontology (GO) analysis comparing TCGA GBM patients with the top 10th percentile COMT expression levels to those with the lowest 10th percentile (35). Our analysis showed that mitochondria-related genes were enriched in patients with high COMT expression (Fig. 2A). In particular, gene signatures involved in mitochondrial electron transport, NADH to ubiquinone transition, and mitochondrial respiratory chain complex I assembly were the top three pathways enriched in COMT high expressors (Fig. 2A and B), indicating a positive correlation between COMT and gene signatures related to mitochondrial functions in brain tumor cells. To further examine the relationship between COMT and mitochondrial function, we generated COMT-KO U87 (Supplementary Fig. S2A) human glioma cells and performed RNA-seq analysis. GSEA showed that OXPHOS was significantly downregulated due to the deletion of COMT (Fig. 2C and D) in U87 glioma cells, consistent with GBM patient data (Fig. 2A and B). Moreover, our RNA-seq data also showed that KO of COMT interfered with mitochondrial RNA granules for mitoribosome assembly and was associated with reduced expression of mitochondrial-encoded genes for the mitochondrial electron transport chain (ETC) subunits (Supplementary Fig. S2B and S2C). Therefore, it is likely that COMT deficiency has a major impact on the whole ETC complex at both transcriptional and protein levels. The membrane-bound (MB)-COMT has previously been indicated to localize to the plasma membrane and the outer membrane of mitochondria (36). To determine its localization in glioma cells, we co-stained COMT and mitochondria with an anti-COMT antibody and MitoTracker, respectively, in U87 and CT2A cells. We validated the specific COMT recognition of the anti-COMT antibody using control and COMT-KO U87 cells (Supplementary Fig. S2D). We found a substantial fraction of COMT was colocalized with mitochondria in both cell lines (Fig. 2E; Supplementary Fig. S2E). Next, we tested whether COMT KO influences complex I activity and electron transport by complex I enzyme activity. We extracted mitochondria from control and COMT-KO U87 cells and measured the oxidation rate of NADH to NAD⁺. The normalized oxidation rate was much lower in COMT-KO cells than in control cells, suggesting the impairment of complex-I activity (Fig. 2F). Comt deletion reduced complex-I activity in CT2A cells (Supplementary Fig. S2F). Furthermore, treating U87 cells with COMT inhibitors entacapone or tolcapone led to a significant decrease in the enzymatic activities of mitochondrial complex I, similar to COMT-KO (Fig. 2G).

Since complex I of ETC generates four protons per NADH oxidized, it contributes to the generation of pH gradient (ΔpH) and the mitochondrial membrane potential (MtMP, ΔΨm; ref. 37). As such, we tested the impact of COMT inhibition on MtMP using the well-established JC-1 assay (38). We observed a significantly higher green/red fluorescence ratio in COMT-KO U87 cells, suggesting increased mitochondrial membrane permeabilization and depolarization (Fig. 2H and I; Supplementary Fig. S2G and S2H). It was previously shown that reactive oxygen species (ROS) were produced by mitochondrial ETC, where complex I is mainly responsible (39). It was also shown that COMT, as an essential metabolic enzyme in the catecholamine cycle, is involved in down-regulating ROS production induced by catechol estrogens (CE) or DA (40). Consistent with earlier findings, we found substantially higher ROS production after COMT deletion or inhibition (Fig. 2J and K; Supplementary Fig. S2I and S2J). Therefore, our data supported the critical role of COMT in maintaining regular mitochondrial functions.

To further characterize the molecular pathways that may underlying in the sensitivity of COMT-deficient tumors to RT, we analyzed VC and COMT-KO U87 cell RNA-seq data and discovered that the expression of antiviral double-stranded RNA (dsRNA)-sensing related genes, including MAVS, IRF7, and OAS2, were highly enriched in COMT-KO U87 cells (Fig. 3A). qRT-PCR analysis further confirmed the elevation of OAS2 and IRF7 mRNA expression in COMT-KO U87 cells (Fig. 3B and C). We also compared the protein expression levels of genes involved in dsRNA sensing in control and COMT-KO cells. DsRNA sensors RIG-I and MDA5, as well as their downstream factors, including MAVS and IRF7, were upregulated (Fig. 3D). Importantly, the level of phospho-TBK1 (p-TBK1), representing TBK1 activation (41), was also increased in COMT-KO cells (Fig. 3D). COMT inhibition by entacapone showed similar results (Fig. 3D). Consistently, we observed cytoplasmic accumulation of double-stranded RNA (dsRNA) in COMT-deficient cells by immunofluorescence staining with the J2 antibody, which has been shown to reliably detect dsRNA (Fig. 3E; ref. 42). In contrast, cytoplasmic dsDNA levels were comparable between control and COMT-KO cells (Supplementary Fig. S3A). These results suggest that COMT deficiency-induced type I interferon (IFN) response in COMT-KO cells is mainly caused by elevated levels of cytoplasmic dsRNA but not dsDNA.

Loss of MtMP is frequently associated with increased mitochondrial membrane permeability (43). Since mtDNA is circular and mitochondrial RNA transcription is bidirectional, this may lead to dsRNA formation. Therefore, we hypothesized that the observed cytoplasmic dsRNA is likely the mtRNA released in COMT-deficient cells. To examine this hypothesis, we isolated the cytoplasmic fraction from VC and COMT-KO U87 cells (see Fig. 3F; Supplementary Fig. S3B for the validation data of our fractionation method). We then extracted RNA and DNA from the cytosolic fractions, respectively, and performed RT-PCR and qPCR analysis using primers that amplify nuclear (nuc) and mitochondrial (mt) encoded genes, respectively. We found increased accumulation of cytoplasmic mtRNA (Fig. 3G) but not nucRNA (Supplementary Fig. S3C), mtDNA (Supplementary Fig. S3D), or nucDNA (Supplementary Fig. S3E) in COMT-KO cells relative to control cells. To further confirm mitochondrial dsRNA (mtdsRNA) as the primary source of increased cytoplasmic dsRNA, we treated U87 cells with 0.5 μg/mL Act-D. This transcription inhibitor causes rapid loss of mitochondrial dsRNA but not nuclear RNAs (25). COMT-KO–induced cytoplasmic dsRNA accumulation was abolished after Act-D treatment, as illustrated by J2 staining (Fig. 3E). In addition, Act-D treatment significantly reduced the cytosolic mtRNA levels in COMT-KO U87 cells, making it comparable to Act-D–treated control cells while nucRNA levels were not affected (Fig. 3F and G; Supplementary Fig. S3C). To examine whether mtdsRNA leakage contributes to the activation of cytosolic antiviral RNA sensing signaling, we depleted mtdsRNA by Act-D treatment and found that increased expression of OAS2 and IRF7 due to COMT deletion was abrogated after Act-D treatment (Fig. 3H and I). This suggests that mtdsRNA is the predominant cytoplasmic dsRNA that activates cytosolic antiviral RNA sensing pathway, which promotes type I IFN production and innate immune response.

Because COMT-deficient cells had low complex I gene expression levels and function (Fig. 2E and F) and complex I is involved in antioxidant response, we examined ROS production in control and COMT-KO U87 cells. Our analysis showed significantly increased ROS levels in COMT-KO cells. Furthermore, adding the mtDNA transcription inhibitor Act-D did not substantially influence ROS levels. In contrast, N-acetyl-L-cysteine (NAC), an antioxidant that can quench ROS, significantly but not completely inhibited ROS production in COMT-KO cells (Supplementary Fig. S3F). Similarly, Act-D treatment did not rescue MtMP reduction in COMT-KO U87 cells, while NAC treatment partially rescued MtMP reduction (Supplementary Fig. S3G and S3H).

We next evaluated whether increased ROS production contributes to mtRNA leakage. We treated cells with NAC and fractionated the cells to get the cytosolic extracts free of mitochondria and nuclei (Supplementary Fig. S3I). We then isolated RNA from the cytosolic fraction for qRT-PCR using primers targeting mitochondrial genes. Primers targeting nucleic genes were also used for comparison. We found that ROS reduction significantly attenuated the release of mtRNA from mitochondria (Supplementary Fig. S3J). In contrast, cytoplasmic nuclear RNA levels were not affected by NAC treatment (Supplementary Fig. S3K). In addition, ROS inhibition by NAC reduced the mRNA levels of OAS2 and IRF7 (Supplementary Fig. S3L and S3M). Furthermore, the protein expression levels of MDA5/RIG-I/MAVS/IRF7 were also reduced significantly by NAC production (Supplementary Fig. S3N). In addition, the activation of TBK1 (p-TBK1) was also inhibited by NAC treatment (Supplementary Fig. S3N). These data suggest that COMT-KO-promoted ROS production was responsible for mtdsRNA leakage to the cytoplasm that stimulated antiviral RNA sensing signaling. High levels of ROS can hinder cell proliferation (44). However, treatment with NAC to inhibit ROS did not reverse the in vitro cell growth delay mediated by Comt-KO (Supplementary Fig. S3O). This indicates that increased ROS levels were not the primary mediators of cell growth defects.

Because type I IFNs are key downstream effectors of the dsRNA-sensing pathway, we next determined type I IFN production in COMT-KO cells and its role in tumor growth suppression. GSEA analysis of our RNA-seq data showed that IFNα/β signaling was highly enriched in COMT-KO U87 cells (Fig. 4A; Supplementary Fig. S4A). RNA sensing and IFN-stimulated gene (ISG) signatures were also upregulated in COMT-KO cells (Fig. 4B; Supplementary Fig. S4B). We further confirmed increased expression of IFNα, IFNβ, RIG-I, ISG15, and IFIT1 in COMT-KO U87 cells by qRT-PCR (Fig. 4C). These data suggest that COMT-KO stimulates the production of type I IFN and ISGs. To determine if the production of type I IFNs was responsible for COMT-KO–mediated tumor suppression, we blocked IFNα/β signaling in vivo by intraperitoneal injection of an αIFNAR1 antibody following a schedule shown in Supplementary Fig. S4C. Notably, αIFNAR1 treatments abrogated the tumor growth suppression and worsened survival outcome of mice bearing COMT-KO tumors (Fig. 4D and E). Anti-IFNAR1 antibody treatment reduced the survival of Comt-KO tumor–bearing mice to a level similar to those with control tumors (Fig. 4D and E), indicating the key role of type I IFNs in Comt-KO mediated antitumor responses. On the other hand, administration of neutralizing antibodies that depleted CD4⁺ T cells and NK cells did not significantly affect the growth rate of Comt-KO tumors, while CD8⁺ T-cell depletion only slightly promoted tumor growth (Supplementary Fig. S4C–S4E). We also treated control and Comt-KO cells with the anti-IFNAR1 antibody and measured cell growth in vitro but found no growth difference (Supplementary Fig. S4E). The discrepant results between in vitro and in vivo glioma cell growth suggest the key role of type I IFN signaling in nontumor cells within the tumor microenvironment.

The primary immune cells in the brain are microglia and perivascular macrophages (45). They work as primary phagocytes in homeostatic maintenance and response to pathological conditions, such as brain injuries and neurodegenerative diseases (46). To evaluate if COMT inhibition influences the phagocytic abilities of macrophages, we performed in vitro phagocytosis assay by coculturing CT2A cells with a C57BL/6J-derived macrophage cell line at a ratio of 5:1. We found that CT2A cells with Comt-KO were significantly more susceptible to phagocytosis by macrophages (Fig. 4F and G; Supplementary Video S1) when compared with control CT2A cells (Fig. 4F and G; Supplementary Video S2). Furthermore, we discovered that RT further enhanced the susceptibility of both control and Comt-KO CT2A cells to macrophage phagocytosis (Fig. 4G). Consistently, chemical inhibition of COMT by entacapone or tolcapone also promotes tumor cell phagocytosis (Fig. 4H). In further experiments, we determined the potential involvement of type I IFNs in glioma cell phagocytosis. To this end, we incubated Comt-KO CT2A cells in the presence of αIFNAR1 and macrophages (activated via treatment with 1 μg/mL lipopolysaccharides (LPS) and 20 ng/mL IFNγ). Compared to sham-treated Comt-KO cells, those treated with αIFNAR1 were phagocytosed at a much lower frequency by macrophages (Fig. 4I). These data suggest COMT inhibition can boost macrophage phagocytosis via type I IFN signaling.

To further characterize the tumor immune microenvironment (TIME) in COMT-deficient tumors, we sorted CD45⁺ live immune cells from intracranially engrafted control and Comt-KO brain tumors for scRNA-seq. A total of 8,538 cells passed quality control and were used for data analysis. We first identified major tumor-infiltrated immune cell populations, including tumor-associated macrophages (TAM), microglia (MG), dendritic cells (DC), T cells, and other cell populations, using cell-type specific markers (Fig. 5A and B; ref. 47). Among all the immune cells, TAM and MG were the most abundant populations that are differentially distributed in VC and Comt-KO tumors. TAMs were dominant in VC tumors, whereas MG were enriched in Comt-KO tumors (Fig. 5C and D). By contrast, the composition of other infiltrating cell populations, such as DCs, T cells, and NK cells, were comparable (Fig. 5D). We further confirmed this finding with flow cytometry analysis (Fig. 5E). Positive staining for CD45⁺ CD11b⁺ was used to gate for myeloid cells. We then excluded monocytes and neutrophils by gating on Ly6C⁻Ly6G⁻ cells. Following a previous report, CD64⁺ cells were targeted for all macrophage populations, with CD45^high as TAM and CD45^low as MG (48). Consistent with results from scRNA-seq, we found more infiltrating microglia but fewer TAMs in Comt-KO brain tumors (Fig. 5E and F). In addition, scRNA-seq data showed high transcriptional expression of H2ab1 and Itgax encoding MHC-II and CD11c, respectively, in TAM and high expression of Mertk in MG (Supplementary Fig. S5A). Flow cytometry analysis of surface protein expression levels also corroborated the scRNA-seq data (Supplementary Fig. S5B).

To gain further insights into the functions of the TAMs and MGs, we chose the top 200 differentially expressed genes (DEG) from scRNA-seq TAM and MG clusters and performed GO analysis. We found that TAMs enriched for genes involved in the negative regulation of both innate and adaptive immune responses (Supplementary Fig. S5C). This is in agreement with the prevailing view that macrophages are often educated by tumor microenvironment to evade immune surveillance (49). In contrast, MG gene signatures were enriched in microglial cell migration and activation. Upregulation of Golgi-to-lysosome pathway also indicated elevated lysosomal activity (Supplementary Fig. S5C). To gain further insights into the functions of infiltrating TAMs and tissue-resident MGs, we further analyzed the heterogeneity of myeloid cells by identifying subclusters using more cell-specific markers (Fig. 5G and H; Supplementary Fig. S5D). From the results of subclustering (Fig. 5G) and trajectory analysis (Supplementary Fig. S5E), we found monocytes were differentiated into four subsets of TAM, including DC-like TAM (50), replicating TAM (51), arginase1 (Arg1)-expressing TAM (52), and IFN-producing TAM (Supplementary Fig. S5E; ref. 53). Our observation that much fewer TAMs were detected in Comt-KO brain tumors compared to VC tumors also supports the current understanding that TAM are predominantly immunosuppressive and promotes tumor progression (54). We also performed MG gene module analysis and identified two subsets of MG: homeostatic microglia (MG0) and disease-associated microglia (DAM; ref. 40), using markers previously identified in CNS neurodegeneration studies (Fig. 5I and J). Downregulation of DAM markers on the top row of Fig. 5J indicates the homeostasis of MG while upregulation of DAM markers in the middle and bottom rows, including Cd68 that is specific for phagocytic MG, suggests MG activation (Fig. 5J; Supplementary Fig. S5F). In contrast to extremely low percentages of infiltrated TAM subsets in Comt-KO tumors, both MG0 and DAM were enriched in Comt-KO tumors at two different time points (day 12 and day 25; Fig. 5H). These data thus suggest that COMT deficiency in tumors promotes MG proliferation and activation in the TIME.

To further examine the activity of MG in brain tumors, we stained tumor sections for expression levels of ionized calcium binding adaptor molecule 1 (IBA1) and CD68. IBA1 is expressed in both resting and activated macrophages and microglia (55) while CD68 is a lysosomal marker that can be expressed by both TAM and MG. However, its expression is substantially increased in DAM, which indicates phagocyte activation in response to proinflammatory stimulation (56). We used both markers to indicate the activation of phagocytes in the brain tumor microenvironment. We found characteristic “cavities” in Comt-KO tumors with cells surrounding the cavities that were IBA1⁺CD68⁺, indicating the strong activation of phagocytic cells (Fig. 5K, right). However, this phenotype was not observed in VC tumors (Fig. 5K, left). This result suggests that glioma cells with Comt deletion are more susceptible to microglial phagocytosis in vivo.

Our results suggest that COMT inhibition may be an effective approach to enhance glioma RT. To evaluate the merit of this idea, we first used a subcutaneous mouse glioma model and tested the efficacy of the COMT inhibitor entacapone in combination with RT following a schedule depicted in Fig. 6A. Our results indicated that entacapone treatment in combination with RT significantly delayed CT2A tumor growth (Fig. 6B). In another experiment, we showed that adding a second radiation dose in combination with entacapone treatment further suppressed the tumor growth of mice compared to those receiving only one fraction of RT (Supplementary Fig. S6A and S6B). Fractionated radiotherapy is a common practice in medical clinics. It allows the normal tissues to recover between fractions and reduce toxicity (57). We next assessed the effectiveness of combining COMT inhibitors with fractionated radiation to delay the growth of glioma tumors (Fig. 6C and D). The results showed that combination treatments with entacapone and five fractions of 2 Gy radiation significantly delayed the CT2A tumor growth rate compared to those receiving either entacapone or radiation treatment alone (Fig. 6D). Please see Supplementary Table S3 for tumor growth delay calculations for experiments using subcutaneous tumor models.

To simulate the brain tumor microenvironment more realistically, we tested the efficacy of tolcapone in combination with RT in the CT2A orthotopic mouse glioma model. Tolcapone was chosen here because of its ability to penetrate the BBB. Following a treatment schedule shown in Fig. 6E, we showed that tolcapone treatment alone did not suppress tumor growth significantly in the intracranial tumor model. However, its combined use with RT significantly extended the lifespan of host mice (Fig. 6F; Supplementary Fig. S6C), with 50% (7/15) of treated mice remaining alive at 60 days post-tumor cell inoculation, significantly better than those treated with RT alone (1/15; Fig. 6F; Supplementary Fig. S6C).

Current glioma treatments include surgical resection, RT, and TMZ chemotherapy (5). However, about 90% of patients with GBM show recurrence despite combined RT and TMZ treatment. Therefore, there is an urgent need to develop novel therapeutic agents.

Our study revealed a critical role of COMT in mitochondrial homeostasis. A previous study has suggested that 22q11.21 deletion in patients, including COMT deletion, is associated with mitochondrial dysfunction (58). In our study, we found that COMT inhibition impaired the activity of complex I of the electron transport chain, which promotes ROS generation, lowering of the MtMP, and mtdsRNA leakage to the cytoplasm that stimulated antiviral RNA sensing pathway and type I IFN production. The striking effects of COMT inhibition on MtMP and other mitochondrial functions have not been reported previously. It highlights that COMT, a key enzyme involved in catecholamine metabolism, may be involved in fundamental mitochondrial functions in ways we do not fully understand yet. Our data also suggest that COMT deficiency impairs mitochondrial biosynthesis at both mRNA and protein levels by interfering with ETC subunit expression and mitochondrial RNA granules. Other changes mediated by COMT deficiency may also contribute to mitochondrial dysfunction. For example, COMT inhibitors prevent DA from being degraded. The increased DA levels can, in turn, upregulate the activation of another DA degrader monoamine oxidase (MAO) to accelerate DA turnover. Deamination of DA by MAO has produced excessive ROS, leading to oxidative damage within and outside mitochondria (59). Indeed, we cannot exclude that the higher ROS production we observed in COMT-KO cells is partially due to MAO activation. Further investigation is needed to understand the mechanism behind COMT deficiency’s role in causing mitochondrial dysfunction.

What are the translational prospects of our finding? Multiple COMT inhibitors, including entacapone, tolcapone, and opicapone, are FDA-approved to be used in conjunction with levodopa and carbidopa to manage PD. Our finding of COMT’s potential role in regulating glioma growth and its response to RT provides a strong rationale for evaluating COMT inhibitors in glioma radiotherapy. Although we do not yet have any such data, it is possible COMT inhibitors can further enhance the efficacy of combined RT and TMZ therapy, given that it works through very different mechanisms from either RT or TMZ. Among the available COMT inhibitors, tolcapone, as the only one that can penetrate the BBB, is probably best suited for combinatorial evaluation with RT and/or TMZ. In support of this idea, tolcapone was shown to inhibit neuroblastoma growth (60).

Our study also provides the foundation for several avenues of further pursuit. One question is whether increased DA levels caused by COMT inhibition can influence the immune system. DA has been shown to positively activate T cells and macrophages via dopamine receptors (61). DA can also mediate innate immune response in the brain through NLRP3 inflammasome priming in macrophages (61). In addition, neurons and glial cells, including microglia, also express COMT in the CNS (62). Under brain injury or pathological conditions, COMT expression is increased in microglia (63). However, whether COMT in microglia is involved in type I IFN–mediated microglia activation for glioma cell phagocytosis is unknown. COMT-induced local and systemic changes of other catecholamines, including epinephrine and norepinephrine, may also influence immunoregulation. Furthermore, it is also important to evaluate how COMT affects peripheral tissues like the liver and kidney, which have a widespread enzyme expression. Studies have shown that COMT inhibition is important for promoting renal sodium excretion in response to acute high-sodium loading, maintaining blood pressure and renal dopaminergic tone (64). Moreover, high doses of COMT inhibitors can impede liver fatty acid metabolism, and tolcapone, in particular, can cause dose-dependent hepatotoxicity by increasing liver aminotransferase (65). On the other hand, COMT inhibitors have been used safely for decades to treat PD. Therefore, it may be possible to find an effective COMTi dose window that may be effective in glioma treatment with minimal toxicity.

In summary, our study identifies a novel role for COMT in maintaining mitochondrial homeostasis and provides preclinical evidence that COMT inhibition may be a promising approach to enhance glioma RT through a novel immunological mechanism. Because COMT inhibitors have been used extensively to treat PD with an excellent safety record, our study, therefore, provides a strong rationale to evaluate COMT inhibitors in conjunction with RT in glioma and other malignancies.

M. Jiao reports a patent for DU8427PROV pending. C.-Y. Li reports grants from NIH during the conduct of the study; in addition, C.-Y. Li has a patent for DU8427PROV pending to Duke University. No disclosures were reported by the other authors.

M. Jiao: Conceptualization, data curation, software, formal analysis, supervision, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. C.J. Pirozzi: Investigation, methodology, writing–review and editing. C. Yu: Software, investigation, methodology, writing–review and editing. X. Bao: Investigation. M. Hu: Investigation, methodology. D. Pan: Investigation, methodology. S. Littleton: Investigation. N. Reynolds: Investigation. D.R. Saban: Investigation. F. Li: Resources, supervision, investigation. C. Li: Conceptualization, resources, supervision, funding acquisition, writing–original draft, project administration, writing–review and editing.

This work was funded by the National Institute of Health under award numbers CA272591 and CA251439 (F. Li and C.-Y. Li). We thank the Duke University Light Microscopy Core Facility for professional help with confocal microscopy. We further thank colleagues from Duke Center for Genomic and Computational Biology for their expert assistance with bulk RNA-seq and single-cell RNA-seq analyses.