Sex as a biological variable. Sex was not considered as a biological variable in this study.

Cell culture. THP-1 and Raw264.7 macrophage cell lines were cultured in RPMI 1640 medium (Gibco, 11875093). HMC3 and SIM-A9 microglia were cultured in Eagle’s Minimum Essential Medium (ATCC, 30-2003) and Dulbecco’s Modified Eagle Medium–Ham’s F12 medium (Gibco, 10565-018), respectively. 293T cell line was cultured in Dulbecco’s Modified Eagle Medium (Gibco, 11995-065). All these cell lines were cultured in indicated medium containing 10% FBS (Fisher Scientific, 16140071) and 1:100 antibiotic-antimycotic (Gibco, 15140-122) and were purchased from the ATCC. Raw264.7 and THP-1 cells were differentiated into macrophages by adding 200 ng/mL phorbol 12-myristate 13-acetate (Sigma-Aldrich, P8139) for 24 hours before further experiments were conducted (73). For stemness maintenance, CT2A cells were cultured in neural stem cell proliferation media (Millipore, SCM005) containing 20 ng/mL epidermal growth factor (PeproTech, AF-100-15) and basic fibroblast growth factor (PeproTech, 100-18B). Human patient-derived GSC2 and GSC272 cells were gifted by Frederick F. Lang (Brain Tumor Center, The University of Texas MD Anderson Cancer Center, Houston, Texas, USA). Mouse GBM tumor–derived GSC line QPP7 was provided by Jian Hu (The University of Texas MD Anderson Cancer Center). All human and mouse GSCs were cultured in neural stem cell proliferation medium containing 20 ng/mL epidermal growth factor and basic fibroblast growth factor. All cells were confirmed to be mycoplasma-free and were maintained at 37°C and 5% CO2. CM were collected from GSCs, or number-matched shC and GPNMB shRNA knockdown cells after culture for another 24 hours in growth factor–free or FBS-free culture medium.

Mice and intracranial xenograft tumor models. Female C57BL/6 (catalog 0000664) and nude (catalog 007850) mice at 4 to 6 weeks of age were purchased from The Jackson Laboratory. All animals were grouped by 5 mice per cage and maintained in IVC System (Lab Products, LLC) for a week before the experiment. GSC272 or CT2A cells were incubated with the CM of shC and shGPNMB macrophages or microglia for 12 hours. Then, the treated GSC272 cells or CT2A cells were mixed with macrophages or microglia expressing shC or shGPNMB by a 1:1 ratio. The intracranial xenograft tumor models were established as described previously (13, 14, 17). In brief, mice were anesthetized using isoflurane through IMPAC6 Anesthesia System. Then, a dental drill was used to open a small hole (1.2 mm anterior and 3.0 mm lateral to the bregma) in the skulls of mice. Mice were placed into the stereotactic apparatus, and 5 μL mixed cells in growth factor–free or FBS-free culture medium were injected into the right caudate nucleus 3 mm below the surface of the brain using a 10 μL Hamilton syringe with an unbeveled 30-gauge needle. The incision was closed using Vetbond glue (3M Science). Meloxicam (20 mg/kg, daily) was subcutaneously injected for pain relief for 3 days after the surgery. Then, mice were monitored for recording survival. Mice with neurological deficits or moribund appearance were sacrificed according to the IACUC protocol. At the end of the experiment, the brains of mice were isolated, either fixed in 4% paraformaldehyde after transcardiac perfusion with PBS for optimal cutting temperature–cryosectioning or processed to isolate tumor-derived immune cells for flow cytometry analysis.

Human samples. Deidentified human plasma (n = 40) and tumor samples (n = 30) from patients with GBM were obtained from the Northwestern Central Nervous System Tissue Bank (NSTB, Chicago, USA). Patients were diagnosed according to the WHO diagnostic criteria by an in-house neuropathologist at NSTB. Detailed patient information is provided in Supplemental Table 1. The healthy control plasma (n = 8) was purchased from Solomon Park Company.

Computational analysis of human GBM datasets. For analysis of human GBM samples, we downloaded the gene expression data from TCGA-GBM, TCGA-LGG, Rembrandt (36), and Gravendeel (51) datasets from GlioVis (http://https://gliovis.bioinfo.cnio.es/). The enrichment scores of infiltrating macrophages and microglia in each sample were investigated by using the single-sample GSEA (ssGSEA) based on the list of 33 human macrophage signature genes (48) and the 9 human microglia signature genes (49), respectively. This analysis was conducted by using ssGSEA algorithm via the “GSVA” package in R software (v.4.3.3).

The single-cell sequencing data of tumors from patients with GBM were downloaded and analyzed by using the BBrowser BioTuring under Talk2Data platform based on established methodology (74). Data from public repositories were integrated and used in this study, including EGAS00001004871 (37) and EGAS00001004422 (44) from the European Genome-phenome Archive (EGA) and GSE131928 (41), GSE148842 (42), GSE89567 (43) and GSE182109 (47) from the Gene Expression Omnibus (GEO). Based on GPNMB expression in macrophages and microglia, patients were regrouped into GPNMB-low and GPNMB-high subgroups. Cancer cells or GSCs from each group were selected for GSEA analysis.

Plasmids and viral transfections. shRNAs targeting human GPNMB and CD44 and mouse Gpnmb in the pLKO.1 vector (Sigma-Aldrich, SHC001) were used. Lentiviral particles were generated as we described previously (14, 17). In brief, 8 μg of the shRNA plasmid, 4 μg of the psPAX2 plasmid (Addgene, 12260), and 2 μg of the pMD2.G plasmid (Addgene, 12259) were transfected into 293T cells plated in 100 mm dishes using Lipofectamine 2000 (Invitrogen, 13778150). Supernatant with lentiviral particles was collected and filtered at 48 and 72 hours after transfection. Cells were infected with viral supernatants containing 10 μg/mL polybrene (Millipore, TR-1003-G). After 48 hours, cells were selected by puromycin-containing (10 μg/mL; Millipore, 540411) medium for 1 week. Then, cells were cultured in low-puromycin-containing (2 μg/mL) medium to keep selecting pressure. The knockdown efficiency was tested for the expression of GPNMB and CD44 using immunoblots. The following human and mouse shRNA sequences were used in this study, GPNMB, #1, TRCN0000178982; #2, TRCN0000151637; #3, TRCN0000154721; and #4, TRCN0000179643; CD44, #1, TRCN0000308110; #2, TRCN0000296191; and #3, TRCN0000289233; and Gpnmb, #1, TRCN0000294593; #2, TRCN0000307370; #3, TRCN0000294525; and #4, TRCN0000294523.

Immunofluorescence. Immunofluorescence was conducted following a previously outlined standard protocol (14, 17). Briefly, cryosection slides were left at room temperature for 30 minutes and then fixed in 10% paraformaldehyde for another 30 minutes before permeabilization. Subsequently, 0.25% Triton X-100 in PBS was applied for 30 minutes at room temperature to facilitate cell membrane permeabilization. Following 3 PBS washes, the sections were blocked with 5% goat serum for 30 minutes. Primary antibody incubation was carried out for 1 hour at room temperature and then overnight at 4°C. Unbound primary antibodies were removed by three 3-minute PBS washes, and secondary antibody cocktails were prepared and applied to the sections for 1-hour incubation. Cell nuclei were counterstained using DAPI/anti-fade mounting medium (Vector Laboratories, H-1200-10). Immunofluorescence images were captured using the Nikon AX/AX R Confocal Microscope System equipped with an apo 60 1.40 Oil 160/0.17 objective. The protein signal’s relative intensity was quantified using ImageJ (NIH). Antibodies specific to CX3CR1 (Invitrogen, 702321), F4/80 (Cell Signaling Technology, 30325S), Ki67 (Cell Signaling Technology, 9129S), CC3 (Cell Signaling Technology, 9661S), SOX2 (Abcam, ab97959), GPNMB (Proteintech, 66926-1-Ig), Ly6C (BioLegend, 128024), and Ly6G (BioLegend, 127606) were used.

Immunoblotting. Protein expression in cells was assessed through immunoblotting analysis according to our previously established protocol (14, 17). Briefly, cells were lysed on ice using RIPA lysis buffer (Thermo Scientific, 89900) supplemented with Protease Inhibitor Cocktail (Thermo Scientific, 78429) and Phosphatase Inhibitor Cocktail (Cell Signaling Technology, 5870S). The protein concentration was determined using the BCA Protein Assay Kit (Thermo Fisher Scientific, PI23225). The protein solution was mixed with LDS sample buffer (GenScript, M00676) and heated at 95°C for 10 minutes. Subsequently, the protein samples were loaded onto SurePAGE gels (GenScript, M00653) and transferred to 0.2 μm nitrocellulose (NC) membranes (Bio-Rad, 1620112) using a predefined standard protocol for 30 minutes in the Trans-Blot Turbo system (Bio-Rad). The NC membranes were then blocked with 5% dry milk in TBST for 1 hour at room temperature and incubated overnight at 4°C with primary antibodies (diluted 1:1,000). After 3 washes, the membranes were incubated with HRP-conjugated anti-mouse (Cell Signaling Technology, 7076) or anti-rabbit (Cell Signaling Technology, 7074S) secondary antibodies for 1 hour at room temperature. Following another round of washing, the membranes were incubated with ECL substrate and imaged using the ChemiDoc Touch Imaging System (Bio-Rad). The antibodies used were sourced from specific companies and included β-actin (Cell Signaling Technology, 3700S), GPNMB (Proteintech, 66926-1-Ig), CD133 (Biosis, BS-4770R), SOX2 (Abcam, ab97959), P-RSK2 (Cell Signaling Technology, 3556S), RSK2 (Cell Signaling Technology, 9355S), P-PYK2 (Cell Signaling Technology, 3291S), PYK2 (Cell Signaling Technology, 3090S), and CD44 (Cell Signaling Technology, 37259S).

Brain tumor immune cell isolation. At the endpoint of experiments, tumor-bearing C57BL/6 mice were euthanized to retrieve their brains. Tumor-derived immune cells in the brain were isolated using the percoll density gradient cell separation method as we previously described (14, 17). In brief, after perfusion with PBS, brains were homogenized on ice with precold 10 mL HBSS (Gibco, 14025092). Then, cells were spun down at 500 g for 10 minutes at 4°C and were resuspended in 30% Percoll (GE Healthcare, 17-0891-01). The solution was gently laid on top of the 70% Percoll and centrifuged at 1,200g for 30 minutes at 4°C with accelerator 7 and breaker 0. After removing myelin and debris, the interphase was collected and centrifuged at 500 g for 10 minutes at 4°C. The tumor-derived cell pellet was resuspended for further analysis.

Primary BMDMs and microglia isolation. For mouse primary BMDM isolation, healthy C57BL/6 mice were euthanized by CO2 inhalation followed by cervical dislocation. The abdomen and hind legs of mice were sterilized with 70% ethanol before autopsy. An incision was made in the midline of the abdomen to expose the hind leg, and the skin and muscle of hind legs were removed as much as possible using scissors. The femur and tibia were isolated followed by 70% ethanol sterilization and 5 minutes of PBS washing. Ophthalmic scissors were used to carefully remove epiphyses of tibia and femur. Then, 5 mL RPMI containing 2% FBS and 1% antibiotic-antimycotic was injected from one end of the tibia and femur to flush out bone marrow cells. The collected cells were filtered through a 70 μm cell strainer and spun down at 500 g for 10 minutes at 4°C and were resuspended in RPMI containing 10% FBS, 1% antibiotic-antimycotic, and 1 ng/mL M-CSF for differentiation. Fresh media was added to the plate on day 3 to facilitate BMDM expansion. BMDMs were fully differentiated on day 6 and analyzed by flow cytometry analysis. For mouse primary microglia isolation, the percoll density gradient cell separation method (14, 17) was used, which was indicated above.

Flow cytometry. The single-cell suspensions were incubated with fixable viability dye (Invitrogen, 5211229035) on ice for 10 minutes. After washing with FACS buffer, cells were incubated with the TruStain FcX (anti-mouse CD16/32) Antibody (BioLegend, 103132) and True-Stain Monocyte Blocker (BioLegend, 426102) in 5% BSA PBST for 30 minutes on ice to block Fc receptors and nonspecific binding of the cyanine acceptor fluorophores. Different antibody cocktails, including Percp/Cy5.5 anti-mouse CD45 (BioLegend, 103132), PE/Cy7 anti-mouse/human CD11b (BioLegend, 101216), BV650 anti-mouse CD11c (BioLegend, 117339), Alexa Fluor 647 anti-mouse CD206 (BD Bioscience, 565250), PE anti-mouse CD68 (BD Bioscience, 566386), BV421 anti-mouse CX3CR1 (BD Bioscience, 567531), Alexa Fluor 700 anti-mouse Ly6C (BioLegend, 128024), and FITC anti-mouse Ly6G (BioLegend, 127606), were added to the samples and incubated for 30 minutes on ice. Intracellular protein staining was performed followed by cell surface marker staining. After washing with PBS, cells were incubated with fixation buffer (BioLegend, 420801) for 20 minutes. Cells were then permeabilized by permeabilization buffer (0.1% Triton X-100 in PBS). GPNMB primary antibody (Proteintech, 66926-1-Ig) was added to cell suspension for 1 hour on ice, followed by incubation with Alexa Fluor 594 anti-mouse secondary antibody (Cell Signaling Technology, 8890S) for 30 minutes. Cells were again incubated with fixation buffer overnight. Samples were read through the BD FACSymphony or BD LSRFortessa flow cytometer and analyzed in FlowJo v10.8.1.

Proliferation (CFSE) assay. Cell proliferation was assessed using the CellTrace carboxyfluorescein succinimidyl ester (CFSE) Cell Proliferation Kit (Invitrogen, C34554). Briefly, 1 × 106 cells were collected and incubated with CFSE working solution (1:1,000) for 20 minutes at 37°C. The staining was stopped by adding complete cell culture media. After washing, cells were continuously cultured for 3–5 days with GPNMB recombinant protein treatment (R&D Systems, 2550-AC) or macrophage/microglia-derived CM in the dark and used for flow cytometry analysis. The percentage of CFSE-positive peaks over the undivided peak (generation 0) was analyzed using FlowJo v10.8.1.

Tumorsphere formation assay. The tumorsphere formation assay was performed as previously described (17). Briefly, GSCs were treated with GPNMB recombinant protein in the presence or absence of PF-271 (Selleck Chemicals, S2672) or SL0101 (MedChemExpress, HY-15237) or macrophage/microglia-derived CM. Cells were seeded into a 96-well plate at 100 cells/well with NSC medium, and tumorsphere numbers in each well were imaged and quantified after 2 weeks.

Incucyte live-cell assay. Cells were seeded in a 96-well plate (Corning, 3599) for culturing overnight. Cells were then treated with GPNMB recombinant protein in the presence or absence of PF-271 or SL0101 before reading in the IncuCyte Zoom Live cell analysis system (Sartorius). The cell proliferating rate was calculated by using the following formula: (cell confluency at indicated time – cell confluency at 0 hour)/cell confluency at 0 hour. The cells were monitored in the IncuCyte system for 48 hours.

Metabolic assays. L-Lactate levels were measured using the Glycolysis Assay Kit (Sigma-Aldrich, MAK439) according to the manufacturer’s instructions. Briefly, cells were seeded into a 6-well plate and treated with GPNMB recombinant protein in the presence or absence of PF-271 or SL0101 or macrophage/microglia-derived CM for 24 hours. Then, cells were cultured for another 24 hours in growth factor–free or FBS-free culture medium, and the glycolytic activity (L-lactate level) of media was measured at 565 nm wavelength. On the other hand, for the glucose metabolism, SeahorseXFe96 analyzer (Agilent) was used to determine ECARs. Glucose metabolism of indicated control and/or treated/modified cells was measured using the Seahorse XF Cell Mito Stress Test Kit (Agilent Technologies, 103015-100) in a Seahorse XFe96 on Seahorse XFe96/XF Pro FluxPak plates (Agilent Technologies, 103792-100). Briefly, cells were cultured in 6-well plate treated with GPNMB recombinant protein in the presence or absence of PF-271 or SL0101 or macrophage/microglia-derived CM before being plated onto seahorse cell culture plates coated with Cell-Tak (Corning, 354240). Cells were analyzed under basal conditions, and in response to sequential injections of glucose (25 mM), oligomycin (2 μM), FCCP (0.5 μM), and rotenone, antimycin A, and 2-Deoxy-D-glucose (1 μM, 1 μM and 50 mM) were used. Data were collected using Agilent Seahorse Wave 2.6 software.

Enzyme-linked immunoassay (ELISA). The levels of GPNMB in human plasma or CM of GSC272 educated HMC3 or THP-1 cells were measured by ELISA using a commercial GPNMB kit (MyBioSource, MBS7607219) following the manufacturer’s instructions. OD value was measured by a Biotek Synergy 2 SL Microplate Reader.

Human phospho-kinase array. After starvation for 24 hours, GSC2 cells were treated with 100 ng/mL GPNMB protein for 1 hour in an incubator. Protein lysis of GPNMB-treated GSC2 and control cells was prepared separately according to the instructions for the Proteome Profiler Human Phospho-Kinase Array Kit (Bio-Techne, ARY003C). Briefly, 400 mg protein in 300 mL lysis buffer from each group was mixed with Array Buffer (Bio-Techne, ARY003C) and incubated with both membranes A and B overnight at 4°C. Membranes were then washed in wash buffer and incubated with Detection Antibody Cocktail A or B (Bio-Techne, ARY003C) for 2 hours at room temperature, respectively. After washing, membranes were placed in an 8-well Multi-dish (Bio-Techne, ARY003C) containing the diluted Streptavidin-HRP (Bio-Techne, ARY003C) for 30 minutes on a rocking platform shaker. Chemiluminescent Reagent Mix (Bio-Techne, ARY003C) was prepared and added to membranes. Signal was determined using the ChemiDoc Imaging System (Bio-Rad) and quantified using ImageJ (NIH).

Statistics. Statistical analyses were performed with 2-tailed Student’s t tests for comparisons between 2 groups and 1-way ANOVA tests for comparisons among more than 2 groups. For the Incucyte recorded proliferation assay, a 2-way ANOVA test was used to determine the statistical difference between each proliferation curve. Data are represented as mean ± SEM. Correlation analysis was conducted using Pearson’s test to determine the Pearson correlation coefficient (R value) and P value. The survival analysis for animal models and patients with GBM was determined by conducting the log-rank (Mantel-Cox) test or Gehan-Breslow-Wilcoxon test. All statistical analyses were performed using GraphPad Prism 10. P values of less than 0.05 were considered significant.

Study approval. All animal experiments were performed with the approval of the IACUCs at Cleveland Clinic and Northwestern University.

Data availability. The data used to support the findings of this study are available within this article and within the Supporting Data Values file. The human TCGA GBM data (TCGA-GBM, TCGA-LGG, Rembrandt, and Gravendeel datasets) are available at GlioVis: https://gliovis.bioinfo.cnio.es/ The scRNA-Seq data of tumors from patients with GBM are available in public repositories, including EGAS00001004871 and EGAS00001004422 from EGA and GSE131928, GSE148842, GSE89567 and GSE182109 from GEO.