Fig 2: ERK/AP-1 axis promotes the expression of CSTA in GAMs. (A) In the GSE242543 cohort, the distribution of differentially expressed genes (DEGs) between human peripheral blood macrophages with and without IL-4 stimulation, with screening criteria set as |LogFC| ≥ 1 and FDR < 0.05. (B) KEGG functional enrichment analysis results of the DEGs above. (C) The intersection of prediction results for potential transcription factors of CSTA from hTarget, FMO_JASPA, KnockTF, GTRD, and Chip_Atlas databases, retaining three potential transcription factors: TP53, FOS (c-FOS), and CTCF. D, E. Correlation tests between c-FOS and CSTA in TCGA pan-cancer (D) and GTEx normal human tissue (E) RNA-seq cohorts, where red dots indicate a significant association (p < 0.05 and correlation coefficient ≥ 0.25). F. Motif diagram of the potential binding base sequence of c-FOS in the CSTA promoter region. G. Results of ChIP-PCR for c-FOS and CSTA promoter fragments in BV2 and HMC3 cells, statistically analyzed based on relative expression compared to the Input group. H. Co-localization of c-FOS and CSTA by immunofluorescence (IF) staining in BV2 and HMC3 cell climbing sheets after IL-4-induced polarization. I. Effect of c-FOS knockdown on CSTA protein levels in IL-4-induced BV2 cells. J. Changes in protein levels of CSTA, c-JUN, c-FOS, and phosphorylated signaling molecules in BV2 cells after separate activation of MAPK, JAK/STAT, and RAP1 signaling pathways. K, L. Changes in protein levels of CSTA, c-JUN, c-FOS, and phosphorylated signaling molecules in BV2 cells after simultaneous MAPK inhibition and IL-4 induction (K), as well as after simultaneous c-FOS knockdown and MAPK activation (L). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

Fig 3: ITGB4 is the functional receptor of CSTA. (A) Affinity IP-mass spectrometry to identify glioma plasma membrane proteins potentially binding to CSTA; performed in CT2A membrane protein extracts, including 6 datasets (3 IP vs. 2 IgG, one-to-one comparisons). After intersection, Robust Rank Aggregation (RRA) ranking was applied (top to bottom = high to low RRA rank), with color intensity indicating LogFC levels in corresponding comparisons. (B) Molecular docking prediction of murine CSTA and ITGB4, enlarged area shows potential binding sites and corresponding amino acid residues. (C) Alignment of ITGB4 amino acid residues 80–100 across species, red box = residues 87–91 co-predicted to interact with CSTA in humans and mice via molecular docking, blue = conserved residues. (D) IP validation of CSTA-ITGB4 interaction in CT2A and LN229 cells. (E) IP validation of CSTA-ITGB4 interaction in CT2A and LN229 cells with ITGA6 knockdown. (F) Brief summary of ITGB4 truncation mutant design: residues 1–683 aa encode the extracellular domain, and residues 684–1822 aa encode the transmembrane and intracellular domains. (G) Pull down assay results of His-tagged human and murine ITGB4 truncation mutants and Fc-CSTA recombinant protein. H, I. Binding efficiency of Fc-CSTA to different ITGB4 AAPM mutants in CT2A (H) and LN229 (I) cells. J. Phosphorylation levels of signaling molecules in ITGB4-related pathways (MAPK, PI3K/AKT, NF-κB) in CT2A cells after CSTA stimulation, and effects of AAPM #88 and sh-ITGB4. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

Fig 4: The CSTA-ITGB4-TGFB1 loop and the biomarker potential of CSTA. A, B, C. IF staining of TGFB1 and macrophage/microglia marker IBA1 in brain sections from CT2A/BV2 mixed orthotopic GBM xenograft models (A), with statistics on the proportions of TGFB1+ (B) and IBA1+ (C) cells among total cells across groups. D, E. IF staining of IBA1 and M2 polarization marker ARG1 (D), and statistics on the proportion of IBA1 + ARG1+ cells relative to total cells in the visual field across groups (E). F. Correlation between TGFB1 + and IBA1 + ARG1+ cell proportions, with each point representing values from an entire field. G. CSTA concentrations in cerebrospinal fluid (CSF) samples from patients with different glioma grades. H. CSTA concentrations in preoperative CSF samples from primary vs. recurrent glioma patients. I, J. ROC analysis of CSF CSTA concentrations in glioma vs. normal patients (I) and in LGG (G2, G3) vs. GBM (G4) patients (J), with corresponding AUC values of 0.9891 and 0.9589. K. Paired t-test of CSF CSTA concentrations in patients with different glioma grades before and 1 week after surgery. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

Fig 5: CSTA is a secretory signal of M2-like GAMs. (A) The neural network was constructed using the random forest method, comprising 16 immune signature genes as the input layer, 5 hidden layers, and an output layer. (B) UMAP illustrating the spatial distribution patterns of distinct cell subpopulations in GBM scRNA-seq data from GSE182109. (C) Differential expression profiles of immune signature genes across various cell subpopulations, where bubble size denotes the positive rate within each subpopulation and color intensity indicates the expression level. (D) Peak plot showing the distribution of different GAMs subpopulations along the pseudotime axis. (E) Differential expression patterns of M1/M2 macrophage markers along the pseudotime axis. (F) Differences in basal CSTA protein levels between GAMs and glioma cell lines. G, H. Differences in RNA (G) and protein (H) levels of CSTA, CD206, and ARG1 in BV2 cells with or without IL-4 treatment. I, J. Immunofluorescence (IF) co-localization results of ARG1 (I) and CD206 (J) with CSTA in GBM patient tissue sections. K, L. Correlation analysis of fluorescence intensity in co-localized cells between ARG1 (K) and CD206 (L) with CSTA in GBM patient tissue sections. M, N. ELISA assay showed the changes in CSTA concentration in the supernatants of BV2 (M) and HMC3 (N) cells with IL-4 induction time. O. Western blot (WB) results showing the effect of CSTA knockdown on protein levels of M2-like polarization markers ARG1 and CD206 in BV2 cells. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001