To clarify the role of GAMs in gliomas, we performed immunohistochemistry (IHC) to detect the infiltration of GAMs in samples, including 35 gliomas and seven normal brain tissues. As shown in Figures 1A, B , the content of CD68-labeled GAMs in glioma samples was higher than that in normal brain tissue and positively correlated with glioma grade. Moreover, we found that the progression of glioma with high GAMs was more rapid compared to patients with low levels of GAMs, as evidenced by more severe edema (7.11 ± 1.58 vs. 5.59 ± 2.12) and earlier onset of symptoms (5.44 ± 4.86 days vs. 12.14 ± 3.13 days) ( Table S1 ).

We also analyzed RNA-seq or RNA-array data of thousands of glioma samples from several datasets for the correlation between the contents of GAMs and the malignancy of glioma. Consistent with the results of our clinical samples, the number of GAMs increased with the higher grade of glioma malignancy ( Figures 1C , S1A, F, I ). It is well known that wild-type isocitrate dehydrogenase 1 and 2 (IDH1 and IDH2, hereafter collectively referred to as IDH) and 1p19q non-deletion are two important genetic events in gliomas closely associated with rapid progression and high recurrence (14). Therefore, we examined the association of GAMs with these pathological features of gliomas. As shown in Figures 1D, E , the level of GAMs in gliomas with wild-type IDH or without 1p19q codeletion was higher than that of gliomas with mutant IDH or with codeletion of genome 1p-19q ( Figures S1D, E, H ). Meanwhile, our analysis showed that gliomas with higher GAMs infiltration had a worse prognosis in terms of overall survival and progression-free survival ( Figures 1F, G , S1B, G, J ). The recurrent gliomas also had higher GAMs infiltration than primary gliomas ( Figure S1C ). All these results suggested the critical role of GAMs in the malignant progression of gliomas.

To explore the potential biological effects of GAMs on glioma malignancy, the CIBERSORT algorithm (15) was used to measure the amount of GAMs in samples from four independent glioma datasets, and then the Gene Set Enrichment Analysis (GSEA) was performed for the levels of GAMs in the samples. The gene sets that are potentially related to the levels of GAMs in glioma samples were obtained from each dataset (TCGA-Seq: 5, CGGA-Seq: 23, Rembrandt-Array: 18, CGGA-Array: 18). To identify the most core biological functions to GAMs, taking intersection was performed from the obtained gene sets to select the common parts. As shown in the Venn diagram ( Figure 2A ), three gene sets were finally filtered, including “Interferon-Gamma-Response”, “Coagulation”, and “Epithelial-Mesenchymal-Translation” (EMT). Moreover, these three functional gene sets were significantly enriched in gliomas (from the TCGA dataset) with high GAMs ( Figure 2A , right). Because EMT is an essential process in tumor cell motility and migration that promotes tumor invasion, which is one of the most critical malignant phenotypes of glioma (16, 17), we further scored each of the glioma profiles for “EMT-ness” in multiple datasets by Creighton’s signature (18). The EMT scores showed a significant positive correlation with GAMs infiltration ( Figures 2B , S2A ). Moreover, the patients with high EMT scores suffered a worse survival than those with low EMT scores ( Figures 2C, D , S2B ). In addition, the EMT score was more prominent in the more malignant subtypes of glioma, including high grade, non-codeletion of 1p-19q, or IDH wild-type gliomas ( Figures 2E–G , S2C, D ). These results suggested that the infiltration level of GAMs in gliomas was significantly correlated with the EMT degree, which affected the malignant progression and the prognosis of gliomas.

To confirm the underlying relationship between GAMs and EMT processes in gliomas, we collected an independent sample set and examined the content of GAMs and the expression of EMT markers by IHC. As the degree of malignancy (WHO grade) increased, not only did the content of GAMs increase ( Figure 3A ), but the EMT markers exhibited significant changes, including the decrease of E-cadherin and the increase of N-cadherin and Vimentin ( Figures 3B–G ). Notably, as shown in Figures 3H–J , there were significant correlations between the content of GAMs and the expression of EMT markers. The expression of E-cadherin was negatively related to the level of GAMs infiltration ( Figure 3H ). On the contrary, the expression of N-cadherin and Vimentin were positively correlated with the infiltration of GAMs ( Figures 3I, J ). Interestingly, the correlation remained valid even in different regions in the same sample ( Figure S3 ). These results indicated a critical role for GAMs in the EMT process of glioma.

The above results suggested a potential correlation between GAMs and the process in gliomas. Therefore, in vitro experiments were performed to investigate the regulatory effects of GAMs on the EMT process of glioma cells. GAMs (CD68⁺) were isolated from glioma samples and cultured to obtain conditioned media ( Figure S4 ). Then, we co-cultured glioma cells with the conditioned media to investigate the effects of soluble factors secreted by GAMs on the EMT of glioma cells. As shown in Figures 4A, B , conditioned media from GAMs significantly upregulated N-Cadherin and Vimentin, whereas downregulated E-Cadherin’s expression of RNA and protein in glioma cells. Moreover, the supernatant of GAMs significantly promoted the migration of glioma cells ( Figure 4C ). These results demonstrated that infiltration of macrophages could increase the malignant progression of gliomas by promoting the EMT process of glioma cells.

All transcriptome datasets in this study, including RNA-seq (TCGA-Seq, CGGA-Seq) and Array (CGGA-Array, Rembrandt-Array), and corresponding clinical information were obtained from The Cancer Genome Atlas (TCGA, https://portal.gdc.cancer.gov/), Chinese Glioma Genome Atlas (CGGA, http://www.cgga.org.cn/), and Gene-Expression Omnibus (GEO-GSE108474, https://www.ncbi.nlm.nih.gov/geo/). These data were preprocessed by R software (v5.1) with the Limma package, including data washing and normalization (36). The content of GAMs in the tumors was assessed by the CIBERSORT algorithm (15). The “EMT score” of samples was calculated by the method reported by Chad J. Creighton (18). Gene Set Enrichment Analysis (GSEA) was performed by GSEA software (v3.1), and hallmark gene sets were used as reference (c1.all.v7.0.symbols.gmt) (37). |NES|>1, False discovery rate (FDR)< 0.25 and p-value< 0.05 were considered significantly enrichment.

This study was approved by the Ethics Committee of Sanbo Brain Hospital, Capital Medical University (SBNK-YJ-2020-001-01). Primary glioma tissues were collected from 35 patients at the Sanbo Brain Hospital, Capital Medical University. Diagnosis of gliomas was established histologically according to the 2016 WHO classification. Seven normal brain tissues were obtained from the patients undergoing emergency cranial surgery, where some of the brain tissue needed to be removed to ensure that the surgery could be performed properly to save lives. According to the preoperative MRI (enhanced T1WI, T2WI) of the patients, the maximum diameter and vertical diameter of the tumor in axial position (marked as a and b), as well as the maximum height in coronal position (marked as c), were measured respectively. According to the formula: V = 4 3 π × a b c , the size of the lesion (V edema + V tumor) by T2WI, and the size of the tumor itself (enhanced T1WI, V tumor) were calculated, respectively. The edema index (EI) was calculated according to the formula: E I = V ( e d e m a ) + V ( t u m o r ) V ( t u m o r ) (38). Human glioma cell line T98G was obtained from the Chinese Academy of Sciences cell bank. In this study, we established two glioma cell lines from two glioblastoma patients named XL and DZH. Primary GBM tumor cell lines XL and DZH (uniformly expressing the glial fibrillary acidic protein, GFAP⁺) were obtained by culture and expansion of ex vivo tumor specimens in the cell culture medium. Cells were used after 10–15 passages. All the glioma cell lines were cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum (FBS), 100 units/ml penicillin, and 100 ng/ml streptomycin. All cells were incubated at 37°C in an atmosphere of 5% CO₂.

Paraffin sections (4 μm) from samples were deparaffinized in 100% xylene and re-hydrated in descending ethanol series and water according to standard protocols. Heat-induced antigen retrieval was performed in 10 mM citrate buffer for 2 min at 100°C. Endogenous peroxidase activity and non-specific antigens were blocked with peroxidase blocking reagent containing 3% hydrogen peroxide and serum, followed by incubation with rabbit anti-human CD68 antibody (1: 200) (Abcam, United Kingdom), rabbit anti-human E-cadherin antibody (1: 150) (Zhongshanjinqiao, China), rabbit anti-human N-cadherin antibody (1: 150) (Zhongshanjinqiao, China), rabbit anti-human Vimentin antibody (1: 150) (Zhongshanjinqiao, China), overnight at 4°C. After washing, the sections were incubated with the biotin-labeled rabbit anti-goat antibody for 30 min at 37°C. The peroxidase reaction was developed using 3,3-diaminobenzidine (DAB) chromogen solution in the DAB buffer substrate. Sections were visualized with DAB, counterstained with hematoxylin, mounted in neutral gum, and analyzed using a bright field microscope.

Primary GAMs were prepared from high-grade fresh glioma tissue (WHO III-IV) as previously described (39–41). In brief, cells were isolated from gliomas by density-gradient centrifugation followed by immunomagnetic beads with CD11b and were suspended in the conditioned culture medium and plated at a final density of 5×10⁵ cells/cm². GAMs supernatant was harvested after 24 h for further studies. Conditioned media were prepared as follows: 24 h after seeding 1×10⁶ XL cells or onto 100-mm dishes, the standard culture medium was exchanged for 8 mL of high-glucose DMEM with 10% FBS and GlutaMax (a medium used for a GAMs culture). Conditioned media were harvested after 24 h from 85 to 90% confluent cultures and centrifuged at 300g for 10 min to remove cell debris. Freshly prepared conditioned mediums were added to GAMs cultures.

For the cell migration assay, 1×10⁴ cells in 100 μL DMEM medium without FCS were seeded on a fibronectin-coated polycarbonate membrane insert in a Transwell apparatus (Costar, MA). In the lower chamber, 600 μL DMEM with 10% FBS was added as a chemoattractant. After the cells were incubated for 12 h at 37°C in a 5% CO₂ atmosphere, the insert was washed with PBS, and cells on the top surface of the insert were removed with a cotton swab. Cells adhering to the lower surface were fixed with methanol, stained with crystal violet, and counted under a microscope in five predetermined fields (200×). All assays were independently repeated at least three times.

Total RNA was extracted from the cells with the TRIZOL reagent (Invitrogen, Camarillo, CA, USA). The total RNA (1 mg) from each sample was reverse-transcribed into cDNA with the M-MLV reverse transcriptase (Promega, Madison, WI, USA) according to the manufacturer’s instructions. Real-time qPCR was performed with SYBR Green (TransGen Biotech, Osaka, Japan) on an ABI 7500 Real-Time PCR System (Applied Biosystems, Foster City, CA, USA) according to the manufacturer’s protocol. The primers were designed using Primer Premier 5.0 software (PREMIER Biosoft International, Palo Alto, CA, USA) and are listed in Table 1 . The amplification was conducted under the following conditions: 95°C for 30s, 95°C for 5s, 55°C for 10s, and 72°Cfor 45s for 35 cycles. In each PCR reaction, nuclease-free water (TIANGEN, Beijing, China) was substituted for the cDNA as a negative control. The relative gene expression was determined using the 2^-ΔΔCt method according to the manufacturer’s recommended protocol.

E-cadherin, N-Cadherin, and vimentin expression were investigated with immunocytochemistry. Cells grown on glass coverslips were fixed in 3% H2O2 methyl alcohol for 5 min. After three washes in PBS, 0.2% Triton-X 100 was added to the cells. Cells were blocked with 10% goat serum and then double-stained with rabbit anti-human E-cadherin antibody (1: 150) (Zhongshanjinqiao, China), rabbit anti-human N-cadherin antibody (1: 150) (Zhongshanjinqiao, China), rabbit anti-human Vimentin antibody (1: 150) (Zhongshanjinqiao, China), overnight at 4°C. After washing, the coverslips were incubated with a biotin-labeled rabbit anti-goat antibody for 30 min at 37°C. The peroxidase reaction was developed using 3,3-diaminobenzidine (DAB) chromogen solution in the DAB buffer substrate. Finally, all cell nuclei were stained with hematoxylin. Coverslips were then mounted, and the slides were examined with a microscope

The immunohistochemically stained tissue sections were reviewed and scored separately by two pathologists blinded to the clinical parameters. For immunohistochemically, the expression of CD68, E-cadherin, N-cadherin, and Vimentin were scored semi-quantitatively based on staining intensity and distribution using the immunoreactive score (IRS). Briefly, Immunoreactive score (IRS) = SI (staining intensity) PP (percentage of positive cells).SI was assigned as: 0= negative; 1= weak; 2= moderate; 3= strong. PP is defined as 0 = 0%; 1 = 0-25%; 2 = 25-50%; 3 = 50-75%; 4 = 75-100%. For categorization of the continuous CD68 values into low and high, we chose a commonly used cutoff point for the measurements (range 0–12, cut point ≤ 3 versus > 3). For immunocytochemistry, the IOD (integrated optical density) and whole-cell area in each view field (area sum) were analyzed by IPP (Image-Pro Plus) software. The expression of CD68, E-cadherin, N-cadherin, and Vimentin were scored semi-quantitatively based on average IOD (average IOD= IOD/area sum). Both in immunohistochemistry and immunocytochemistry, Cells stained with indicated antibodies were calculated per field of view, with at least 5 view fields per section evaluated at 400× magnification.

All quantified data represented an average of at least triplicate samples. SPSS 13.0 and Graph Pad Prism 5.0 were used for statistical analysis. Data are presented as mean ± SD. One-way ANOVA or two-tailed Student’s t-test was used for comparisons between groups. Pearson correlation analysis was used to evaluate the linear correlation between the two variables. Kaplan-Meier curve and Log-rank test were applied for survival analysis. Differences were considered statistically significant when P< 0.05.