Skip to main content
NCBI home page
Discover Oncology logoLink to Discover Oncology
. 2026 Jun 6;17:1160. doi: 10.1007/s12672-026-05289-2

Single cell and spatial atlas of macrophage and astrocyte interactions in glioblastoma

Mengge Gao 1,2, Qiong Liu 1,2, Bo Yang 1,2, Xiaofang Liu 1,2,✉
PMCID: PMC13469014  PMID: 42249233

Abstract

Background

Glioblastoma (GBM) exhibits pronounced intratumoral heterogeneity, yet the interactions between macrophages and astrocyte-like (AC-like) cells remain incompletely understood.

Methods

We analyzed single-cell RNA sequencing data from 10 GBM patients, performing clustering with Seurat and Harmony, copy number variation to distinguish malignant cells, and trajectory inference to map lineage relationships. CellChat was applied to predict intercellular signaling, while CellTrek reconstructed spatial co-localization. Prognostic significance of identified key subpopulations was evaluated using Cox regression analysis.

Results

Thirteen cell subtypes were identified in GBM, including four macrophage clusters (e.g., Macro_CCL3, Macro_C1QL1) and eight AC-like clusters (e.g., AClike_RGCC, AClike_SLC7A5). Macro_C1QL1 and AClike_SLC7A5 were enriched in malignant regions and correlated with poor prognosis. Intercellular communication analysis indicated that Macro_C1QL1 and AClike_SLC7A5 cells engage in strengthened APP–CD74 signaling, emphasizing the involvement of tumor-specific CypA and APP pathways. Spatial mapping confirmed their co-localization in tumor-dense areas, supporting niche-specific interactions.

Conclusion

This study provides an integrated single-cell and spatial characterization of macrophage and AC-like cell crosstalk in GBM, identifying key signaling networks and potential therapeutic targets.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12672-026-05289-2.

Keywords: Glioblastoma, Single-cell and spatial transcriptomics, Tumor microenvironment, Macrophages, AC-like cell

Introduction

Glioblastoma (GBM) is the most aggressive and lethal primary brain tumor in adults, characterized by rapid progression, high recurrence, and poor prognosis despite multimodal therapies [1]. The median survival remains less than 15 months [2]. Recent advances in single-cell RNA sequencing (scRNA-seq) have revolutionized the understanding of GBM heterogeneity [3]. For instance, Patel et al. [4] first revealed extensive transcriptional diversity within individual GBM tumors, identifying multiple coexisting cellular states. These findings underscore the utility of scRNA-seq in delineating the intricate tumor ecosystem of GBM.

Among the diverse cell populations identified by single-cell studies, tumor-associated macrophages [5] (TAMs) and astrocyte-like (AC-like) [6] malignant cells play pivotal roles in GBM biology. Single-cell RNA sequencing analysis have revealed distinct macrophage subsets of both microglial and bone marrow origin, with immunosuppressive phenotypes contributing to tumor progression and therapy resistance [7]. Meanwhile, AC-like tumor cells, first defined by Neftel et al. [8], represent one of the major transcriptional states of GBM, displaying astrocytic gene signatures and being closely linked to mesenchymal transition and invasion. Single-cell transcriptomics reveals cellular heterogeneity but not functional interactions, while cell–cell communication analysis predicts signaling networks but requires precise cell subtyping. Integrating both approaches enables us to characterize GBM composition and dissect how specific macrophage and tumor subpopulations communicate to drive tumor progression, a necessity given the context-dependent and paracrine nature of these interactions that neither method alone can fully address.

Although both macrophages and AC-like tumor cells have been well characterized separately, their functional interactions within the GBM microenvironment remain largely unexplored. Previous studies have shown that TAMs can induce malignant cells to adopt an AC-like phenotype, and conversely, AC-like malignant cells can recruit and reprogram macrophages through secreted factors [9, 10]. However, these two cell subsets are still lacking in the single-cell interaction landscape of GBM. In particular, no study has systematically analyzed the spatial organization or communication between these two key cell types using integrated spatial transcriptomic approaches. This lack of spatial context limits our understanding of how macrophage–AC-like crosstalk shapes tumor evolution. Therefore, we performed a combined single-cell and spatial transcriptomic analysis to comprehensively map macrophage–AC-like interactions in GBM.

Materials and methods

Data acquisition

Single-cell RNA sequencing datasets were obtained from the Gene Expression Omnibus (GEO) under accession numbers GSE242044 [11] (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgiacc=GSE242044) and GSE103224 [12] (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi? acc=GSE103224). The combined dataset comprised 17 specimens from 15 patients, including 10 tumor tissue samples and 7 normal tissue samples. These two datasets were selected because they contain GBM transcriptomic profiles and include both tumor and adjacent non‑tumor tissues, whereas many other datasets lack normal controls or have very small sample sizes. In addition, the spatial transcriptomic data were obtained from GSE235672 [13]. The sequencing platform used in this dataset is 10x Visium, and we analyzed two tumor tissue samples. This dataset is derived from different samples than the single-cell RNA sequencing cohort.

Processing and computational analysis of single-cell data

Quality control, normalization, dimensionality reduction, and clustering of scRNA-seq data were performed using Seurat (v4.3.0.1) [14]. The Harmony algorithm (v0.1.1) [15] was applied to correct for technical artifacts and batch effects arising from multi-sample integration. Cell clusters were identified through principal component analysis (PCA) and visualized in two-dimensional space using uniform manifold approximation and projection (UMAP) [16]. To account for matching and batch effects, we included two matched tumor-normal samples and performed PCA before and after batch correction, confirming that batch effects were mitigated without losing biological differences. We used the FindAllMarkers function to calculate marker genes for cell subclusters (p < 0.05, logfc.threshold = 0.5).

Lineage trajectory reconstruction

Developmental trajectories and cellular differentiation states were inferred using pseudotemporal ordering. CytoTRACE2 (v1.0.0) [17] was employed to estimate differentiation potential, while Slingshot (v2.8.0) [18] was used to reconstruct branching lineage paths.

Inference of somatic copy number alterations

The inferCNV package [19] was utilized to predict large-scale chromosomal copy number variations from gene expression data. Cells were classified as malignant or non-malignant based on the magnitude and pattern of inferred CNV profiles. We used T cells and B cells as reference cells. A cell was classified as malignant if its inferCNV score (mean absolute CNV signal) exceeded the 95th percentile of the distribution of scores from the reference cells. Denoising was performed using the “denoise” function with the Hidden Markov Model method. The denoise cutoff was set to 0.15 for amplification and deletion events, respectively, and the window length was set to 100 (the actual value was slightly adjusted based on the number of genes on each chromosome).

Functional annotation and pathway analysis

Differentially expressed genes for each cell cluster were analyzed for functional enrichment using clusterProfiler [20]. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses revealed biological processes and signaling pathways characteristic of distinct cellular subpopulations.

Survival analysis

We performed survival analysis using the TCGA-GBM dataset (n = 173). The gsva() function was used to calculate GSVA scores for Macro_C1QL1 and AClike_SLC7A5 for each patient. Patients were divided into high and low groups using thresholds of 5.5549 and 0.065, respectively. The log-rank test was used to assess whether there was a significant difference in survival between the two groups.

Intercellular signaling network analysis

Cell–cell communication was systematically investigated with CellChat (v2.1.2) [21]. By identifying significant ligand–receptor interactions across cell types, signaling networks within the tumor microenvironment were comprehensively characterized. Ligand–receptor pairs were prioritized based on a combination of statistical significance (p < 0.001) and communication probability, with emphasis placed on interactions that ranked highest across these metrics. At the signaling pathway level, networks were further ranked according to their overall communication strength and statistical significance (adjusted p < 0.01), and those with the highest ranking were selected for detailed downstream analysis.

Spatial transcriptomic mapping

CellTrek (v0.0.94) [22] was applied to deconvolute spatial transcriptomics data using scRNA-seq references, enabling the reconstruction of spatial localizations and co-localization patterns of distinct cell types in GBM and adjacent parenchyma. We used the dist_thresh = 0.55 parameter to define the distance threshold; only when the distance between a single cell and a spatial spot is less than this threshold is it considered mappable. The top_spot = 5 parameter limits each cell to mapping to at most the top 5 most similar spatial spots. The repel_r = 20 parameter defines the repulsion radius (units are typically pixels or micrometers in spatial coordinates). When multiple cells map to the same spatial location, a minimum distance constraint is imposed to avoid excessive overlap; a value of 20 indicates that two mapping points must maintain a distance of 20 units from each other.

Results

The single-cell landscape of glioblastoma

To explore cellular heterogeneity in GBM, we examined scRNA-seq data from 17 specimens from 15 patients, including 10 tumor tissue samples and 7 normal tissue samples. Following stringent quality control, dimensionality reduction, and batch-effect correction, we identified 13 major cell populations, visualized via UMAP (Fig. 1A). Based on canonical marker gene expression, these clusters were annotated as AC-like cell, OPC-like cell, MES-like cell, oligodendrocyte, microglia, T cell, B cell, fibroblast, endothelial cell, macrophage, NPC-like cell, mast cell, and plasma cell. Tumor tissues exhibited substantial enrichment of AC-like, MES-like, endothelial, and fibroblast populations, along with multiple immune subsets such as T cells, B cells, mast cells, and macrophages (Fig. 1B).

Fig. 1.

Fig. 1

Open in a new tab

Single-cell transcriptomic profiling reveals cellular heterogeneity in glioblastoma and adjacent normal tissues. A UMAP projection illustrating the composition and distribution of 13 distinct cell clusters (FindAllMarkers function: p < 0.05, logfc.threshold = 0.5). B Comparative analysis of cell cluster abundances between tumor and normal tissue conditions. C Bubble plot displaying canonical marker genes across the 13 major cell types. D Proportion of the 13 cell types between tumor and normal tissue groups. E Feature plots showing expression of representative marker genes for each cell type on UMAP (FindAllMarkers function: p < 0.05, logfc.threshold = 0.5)

To validate cluster identities, we assessed the expression of established marker genes. Bubble plots illustrated representative markers for each population, confirming their annotation (Fig. 1C). The tumor tissues exhibited greater cellular subpopulation diversity than normal tissues, with AC-like cells being the predominant subset among all subpopulations (Fig. 1D). In normal tissues, although the proportions of AC-like, MES-like, and OPS-like subsets were lower than those in tumor tissues (Figure S1A), these three subgroups rank among the top three in the cell subpopulations of normal tissues (Fig. 1D). Spatial feature mapping within the UMAP further highlighted distinct transcriptional profiles across lineages, reinforcing subtype-specific gene expression patterns (Fig. 1E). Together, these results establish a high-resolution single-cell atlas of GBM and matched normal brain tissues.

The heterogeneity and lineage potential of macrophage

Based on the above analysis, we observed that macrophages were the most abundant immune cell population and showed significant differences between tumor and normal tissues. Previous studies have reported that macrophages play important roles in GBM, though the underlying mechanisms remain unclear [23, 24]. Therefore, we further characterized their heterogeneity by performing subclustering analysis. High-resolution clustering of macrophages identified 4 distinct subpopulations, designated Macro_SCGB1D2, Macro_PIGC, Macro_CCL3, and Macro_C1QL1, which were which were annotated by identifying marker genes for each subpopulation using the FindAllMarkers function and visualized using UMAP (Fig. 2A). Tumor tissues showed pronounced enrichment of the Macro_SCGB1D2 subset, while Macro_CCL3 was predominantly detected in normal tissues (Fig. 2B). The enrichment of the Macro_SCGB1D2 subset in tumor tissues suggests that this subset may be associated with the pro-tumor phenotype of TAMs. SCGB1D2, a member of the secretoglobin family, has been reported in previous studies to be involved in immune regulation and tissue remodeling [25, 26]. Its enrichment in tumors may indicate that this subset promotes the formation of an immunosuppressive microenvironment through the secretion of related factors, or interacts with tumor cells to support tumor progression. In contrast, the Macro_CCL3 subset is predominantly distributed in normal tissues. The preferential distribution of this subset in normal tissues suggests that it may be involved in maintaining tissue homeostasis and immune surveillance [27].

Fig. 2.

Fig. 2

Open in a new tab

Identification and characterization of macrophage subpopulations in glioblastoma. A UMAP visualization of macrophage clustering, identifying four macrophage subtypes (FindAllMarkers function: p < 0.05, logfc.threshold = 0.5). B Bar plot indicating the proportion of each macrophage subtype in tumor and normal tissues. C Bubble plot showing marker gene expression across macrophage subtypes. D Lineage trajectories of macrophage inferred by Slingshot, revealing two distinct differentiation branches. Red indicates the starting point of the trajectory, while blue represents the endpoint. The cell type annotations are consistent with those shown in Fig. 1A. E–H Correlation analysis of four macrophage subsets with effector CD8 + T cells. I Functional enrichment analysis of Macro_C1QL1. J Relationship between Macro_C1QL1 and patient overall survival (log-rank p = 0.029)

Bubble plots illustrated representative marker expression patterns, supporting the classification of each macrophage subset (Fig. 2C). Pseudotemporal trajectory analysis revealed two branching differentiation paths, suggesting developmental plasticity among macrophage populations during tumor progression, suggesting that they possess high developmental plasticity and potential for state transitions (Fig. 2D). Given that CD8 + T effector cells are central to anti-tumor immunity and have been linked to TAMs-related dysfunction [28], we next examined the correlation between each macrophage subpopulation and CD8 + T effector gene signatures. Correlation analysis further indicated that Macro_C1QL1 marker levels were inversely associated with CD8⁺ T effector gene signatures (Fig. 2E) (PMID: 38442575), Combined with its complement activation-related functions, this suggests that this subset may inhibit T cell anti-tumor immune responses through complement-mediated immunoregulatory mechanisms. In contrast, the markers of Macro_CCL3, Macro_PIGC, and Macro_SCGB1D2 showed positive correlations with CD8⁺ T effector activity, suggesting that these subsets may have a synergistic relationship with the anti-tumor immune response (Figs. 2F–H).

Gene Ontology analysis highlighted enrichment of the Macro_C1QL1 subset in processes related to cell growth, regulation of protein stability, and positive control of the cell cycle (Fig. 2I). Clinically, high Macro_C1QL1 signature scores were significantly associated with poor patient prognosis, with elevated scores predicting reduced overall survival (log-rank p = 0.029; Fig. 2J), suggesting that this subset not only possesses immunosuppressive potential but may also serve as an independent risk marker for poor prognosis in patients with glioblastoma. Based on these findings, Macro_C1QL1 emerges as a tumor-enriched, immunosuppressive macrophage subset associated with poor prognosis.

The heterogeneity of AC-like cell in the tumor microenvironment

The multi-omics study conducted by Xiong et al. [6] classified GBM into four subtypes: MES-like, AC-like, OPC-like and NPC-like, each reflecting distinct cellular states along the glioma differentiation hierarchy. The CNV analysis revealed pronounced genomic instability in NPC-like and AC-like cells, which exhibited elevated CNV burdens relative to other subtypes (Fig. 3A), suggesting that these populations may have undergone greater genomic divergence, potentially driven by accumulated oncogenic insults and contributing to intratumoral heterogeneity. We then performed survival analysis and found that AC-like cells with high CNV scores were significantly associated with poor prognosis in GBM patients (log-rank p = 0.04; Figure S1B), whereas OPC-like cells showed no significant association with prognosis (log-rank p = 0.14; Figure S1C). Based on genetic characteristics and clinical significance, we focused on the analysis of the AC-like cell subpopulation. Pseudotemporal trajectory reconstruction delineated a branched differentiation continuum across these four subtypes, with AC-like cells occupying the terminal position, suggesting a mature or end-state phenotype in GBM evolution (Fig. 3B), implying that they represent a more differentiated, possibly end-state phenotype in GBM evolution. Unsupervised subclustering of AC-like cell identified eight distinct subpopulations—AClike_RGCC, AClike_SLC7A5, AClike_ERMN, AClike_HLADRB5, AClike_IL1R1, AClike_CD52, AClike_MGP, and AClike_IGLJ3—each demonstrating differential spatial distributions between tumor and normal tissues (Fig. 3C). Tumor samples showed significant enrichment of the AClike_SLC7A5 subset, whereas AClike_RGCC was predominantly localized in normal tissue (Fig. 3D). These AC-like cell subpopulations were validated by their unique marker gene expression profiles (Fig. 3E).

Fig. 3.

Fig. 3

Open in a new tab

Dissection of AC-like cell heterogeneity and functional states in glioblastoma. A Violin plots showing CNV score distributions across four epithelial cell clusters. B Cell trajectory analysis depicting the progression path with AC-like cells as the endpoint. Red indicates the starting point of the trajectory, while blue represents the endpoint. The cell type annotations are consistent with those shown in Fig. 3A. C UMAP plot showing clustering of AC-like cells into eight distinct subsets (FindAllMarkers function: p < 0.05, logfc.threshold = 0.5). D Bar plot comparing the abundance of each AC-like cell subset in tumor and normal samples. E Bubble plot illustrating marker gene expression profiles of AC-like cell subsets. F Functional enrichment results for AClike_SLC7A5 subsets. G Relationship between AClike_SLC7A5 and patient overall survival (log-rank p = 0.0015)

Functional pathway enrichment analysis indicated that the AClike_SLC7A5 subset was closely associated with oxidative phosphorylation, ATP metabolism, regulation of cell development, and epithelial-specific processes (Fig. 3F). Clinically, high abundance of AClike_SLC7A5 was correlated with significantly worse overall survival in GBM patients (log-rank p = 0.00015), underscoring its prognostic relevance and potential role in disease progression (Fig. 3G). Collectively, these findings elucidate the cellular and functional heterogeneity of AC-like cell and identify SLC7A5-enriched AC-like subpopulation as a clinically significant subgroup in GBM.

Collectively, these findings not only reveal the functional heterogeneity of AC-like cells but also identify the SLC7A5-enriched subpopulation as a clinically significant subgroup in GBM. Its association with metabolic reprogramming and poor prognosis positions it as a potential therapeutic target and a candidate biomarker for risk stratification.

Tumor-specific macrophage and AC-like cell interaction

Profiling of intercellular communication revealed pronounced remodeling of signaling networks between tumor and normal tissues. Tumor samples displayed markedly increased activity in pathways such as JAM, APP, and GRN (Fig. 4A), while normal tissues were primarily characterized by stronger engagement of JAM, GRN, and APP signaling. Although the above pathways are significantly enriched in both tumor and normal tissues, their distribution across different cell subpopulations may differ significantly (Fig. 4B). In the normal state, only the Macro_CCL3 macrophage was found to interact with all other AC-like subpopulations except AClike_IGLJ3 (Fig. 4C). This restricted interaction pattern suggests a tightly regulated communication network in the normal state, possibly reflecting homeostatic surveillance and maintenance of tissue integrity. In contrast, tumor tissues exhibited a broader interaction landscape, where all four macrophage subsets communicated with all eight AC-like subtypes (Fig. 4D). This expansion of cellular crosstalk implies a loss of regulatory constraints in the tumor microenvironment and points to a globally activated signaling state that may drive diverse tumor-supportive functions, including immune suppression, metabolic cooperation, and stromal remodeling.

Fig. 4.

Fig. 4

Open in a new tab

Tumor-specific macrophage and AC-like cell communication landscape in glioblastoma. A, B Number of predicted cell-cell interactions between macrophage and AC-like cell subsets in tumor and normal tissues. C, D Circle plot showing the interaction between AC-like cell and macrophage subpopulations in tumor and normal tissues. E Bubble plot showing cell interactions between Macro_C1QL1 subtypes and AClike_SLC7A5. F, G Differences in ligand-receptor distribution of CypA and APP signaling pathway network in tumor and normal conditions

Ligand–receptor analysis further identified a tumor-specific communication axis between macrophages and AC-like cells, predominantly mediated by interactions between Macro_C1QL1 and AClike_SLC7A5 clusters through pairs such as PPIA-BSG and APP-CD74 which have the highest statistical significance (p < 0.001) and the greatest communication probability (Fig. 4E). Furthermore, prior studies have shown that APP–CD74 is involved in neuroinflammatory pathways, supporting its biological relevance [29, 30]. Consistently, CypA and APP signaling pathways ranked among the most significantly altered (adjusted p < 0.01) and exhibited the highest interaction scores across all predicted networks, while both CypA and APP have been implicated in immune regulation [29, 31], supporting their prioritization for further analysis. Based on the above analysis, we suggest that the CypA and APP signaling pathways may play important roles in mediating the tumor-promoting effects of Macro_C1QL1 and AClike_SLC7A5. Detailed examination of CypA (Fig. 4F) and APP (Fig. 4G) signaling networks revealed distinct ligand-receptor distribution patterns distinguishing tumor from normal tissues. Notably, the Macro_C1QL1 to AClike_SLC7A5 connection represented the strongest tumor-specific interaction, but absent in normal tissues. Collectively, these findings demonstrate an extensively reprogrammed communication landscape in GBM, driven by macrophage and AC-like crosstalk and context-dependent pathway activation.

Collectively, these findings demonstrate an extensively reprogrammed communication landscape in GBM, driven by context-dependent pathway activation and the emergence of specialized crosstalk between macrophage subsets and AC-like malignant cells. This reprogramming likely contributes to the establishment of an immunosuppressive and tumor-permissive microenvironment, and the identified tumor-specific ligand-receptor pairs, particularly the Macro_C1QL1-AClike_SLC7A5 axis, may represent promising therapeutic targets for disrupting pathogenic cellular communication in GBM.

Spatial transcriptome characteristics of macrophage and AC-like cell subsets

Spatial transcriptomic analysis of GBM specimens revealed distinct topological reorganization within the tumor microenvironment. UMAP visualization illustrated clear regional segregation between histologically distinct malignant areas across two representative GBM cases (Figs. 5A, B), indicating that distinct transcriptional programs are spatially compartmentalized, a feature likely driven by local microenvironmental cues and contributing to functional heterogeneity within the tumor. Integration of single-cell and spatial transcriptomic datasets enabled high-resolution mapping of cellular distributions, uncovering region-specific localization patterns of major cell classes (Figs. 5C, D). This spatial segregation suggests that different cell types occupy specialized niches, where local signaling gradients and physical interactions may orchestrate context-dependent functional states.

Fig. 5.

Fig. 5

Open in a new tab

Spatial transcriptomics confirms colocalization of macrophage and AC-like cell subsets in glioblastoma. A, B Dimensionality reduction and clustering of spatial transcriptomics in tumor tissues from two glioblastoma patients. C, D Distribution of single cells and spatial spots in tumor tissues from two glioblastoma patients. E, F Spatial deconvolution of macrophage and AC-like cell in two glioblastoma patients. Red circles highlight representative regions of Macro_C1QL1 and AClike_SLC75A co-localization observed in glioblastoma patients, illustrating the spatial reorganization of these cell populations in the tumor microenvironment. G, H Network diagram showing spatial co-localization of macrophage and AC-like cell subpopulations in tumor tissues. I, J Heatmap showing spatial neighborhoods of macrophage and AC-like cell subpopulations in tumor and normal tissues

Spatial deconvolution further resolved the topographic organization of macrophage and AC-like cellular subsets, demonstrating fundamentally different spatial patterning between tumor and normal tissue contexts (Figs. 5E, F). Network analysis of cellular co-localization revealed extensive rewiring of macrophage–AC-like cell interactions in tumor regions, with emergence of specialized spatial neighborhoods unique to the malignant state (Figs. 5G, H). This reorganization implies that tumor progression drives the formation of spatially confined functional units, in which macrophage and malignant cell subsets co-localize to establish local microenvironments that may facilitate immune modulation, metabolic exchange, and pro-tumorigenic signaling. Quantitative evaluation of spatial adjacency through heatmap visualization confirmed ecosystem-level remodeling in GBM, characterized by systematic alteration of spatial relationships between macrophage and AC-like subpopulations compared to normal tissue counterparts (Figs. 5I, J).

Spatial transcriptomic analysis reveals that GBM progression drives a marked reorganization of the tumor microenvironment, characterized by the spatial segregation of macrophage and AC-like cell subsets into distinct, functionally specialized niches that remodel cell-cell interactions compared with normal tissue.

Discussion

In this study, we performed an integrative single-cell and spatial transcriptomic analysis to comprehensively characterize the cellular heterogeneity and intercellular communication landscape of GBM. By delineating 13 major cell populations and focusing on macrophages and AC-like malignant cells, we uncovered profound transcriptional and spatial diversity within the tumor ecosystem. Among macrophage subsets, we identified a tumor-enriched Macro_C1QL1 population characterized by complement activation and cell growth signatures, whereas among AC-like cells, the AClike_SLC7A5 subgroup exhibited metabolic activation and poor prognostic implications. Notably, our integrative analysis revealed a tumor-specific interaction axis between Macro_C1QL1 and AClike_SLC7A5 mediated by PPIA-BSG and APP-CD74 signaling, suggesting a previously unrecognized immunometabolism crosstalk contributing to tumor progression. The combined single-cell and spatial approach thus provide a high-resolution view of GBM cellular ecosystems and their spatial organization, advancing the current understanding of tumor–immune communication in the brain tumor microenvironment.

Recent single-cell studies have emphasized that TAMs in GBM exhibit marked heterogeneity and context-dependent immunoregulatory functions. For instance, Liu et al. [32] identified VSIG4S100A10TAM that drive M2 polarization and impair anti-PD-1 responses, while Chen et al. [33] described an EZH + TAM subset linked to cell-cycle activation and poor prognosis. Our data extend these findings by revealing four distinct TAM subsets, especially Macro_C1QL1, characterized by complement activation and protein stability pathways. Unlike the classically defined M2-like or metabolic TAMs such as ABCA1⁺ cells [34], Macro_C1QL1 displayed proliferative signatures and a strong inverse correlation with CD8⁺ T-cell effector genes, suggesting a non-canonical immunosuppressive phenotype. C1QL1 is a member of the C1q family, and previous studies have suggested that this family of proteins is involved in synaptic pruning and immune regulation [35, 36]. We speculate that the Macro_C1QL1 subpopulation may inhibit CD8 + T cell activation or effector function, for example, by interfering with T cell receptor signaling or inducing T cell exhaustion, through the secretion of C1QL1 or other synergistic molecules. Moreover, studies have reported that C1q family molecules can regulate the immunosuppressive phenotype of macrophages and indirectly suppress T cell responses [37, 38]. Thus, this inverse correlation may provide new insights into macrophage-mediated immune evasion. Functionally, its high abundance predicted unfavorable outcome, supporting the emerging concept that specific transcriptional programs—rather than simple M1/M2 polarization—govern TAM-mediated immune evasion in GBM. Additionally, recent studies have shown that tumor-specific CD4 + T cells expressing CXCL13 are associated with the transcriptional states of macrophage [39], suggesting that future investigations into different immune cell populations may further elucidate the complex immune landscape of GBM.

GBM exhibits substantial intratumoral heterogeneity, with tumor cells classified into MES-like, AC-like, OPC-like, and NPC-like state [6, 40]. Previous studies indicate that AC-like cells occupy metabolically active region [10], which is consistent with our research. Unlike the previous single AC-like subtype, our study further delineated eight AC-like subpopulations, in particular identifying the AClike_SLC7A5 subset as specifically enriched in tumor tissue, exhibiting high oxidative phosphorylation, ATP metabolism, and epithelial-associated programs. Clinically, high abundance of AClike_SLC7A5 correlated with worse overall survival, highlighting its prognostic significance and potential role in GBM progression. Moreover, environmental carcinogens such as benzo[a]pyrene may further influence GBM progression via TP53 modulation in MES-like and AC-like cells and in CD8⁺ exhausted T cells [41]. Together, these results integrate cellular heterogeneity, metabolic reprogramming, and extrinsic factors, providing a comprehensive view of GBM plasticity and potential targets for immunotherapeutic interventions.

Previous studies have shown that various cell types within GBM, including macrophages, microglia [24], oligodendrocyte [42], and other cell subpopulations [43, 44], interact extensively with each other within the tumor microenvironment (TME). These interactions have been reported to influence tumor progression, therapeutic resistance, and immune evasion. For example, TAMs and microglia are critical for GBM progression, and recent work demonstrates that lysis oxidase (LOX) and olfactomedin like-3 (OLFML3) mediate reciprocal regulation between macrophages and microglia, influencing immune infiltration and therapeutic response [45]. In addition, previous studies have demonstrated that combining VEGF pathway inhibitors with imipramine can remodel the tumor vasculature and reprogram immunosuppressive TAMs into immunostimulatory cells [46]. Inhibiting BACE1 has been shown to reprogram tumor-promoting macrophages into tumor-suppressive macrophages and enhance their phagocytosis of glioma cells [47]. These findings provide a strong rationale for investigating the crosstalk between TAMs and specific malignant cell subsets in our study. However, the specific interactions between macrophages and AC-like cells remain unexplored. In our study, we identified the AClike_SLC7A5 subpopulation, which is enriched in tumor regions and exhibits high oxidative phosphorylation and epithelial-like programs. Spatial transcriptomic analysis revealed its co-localization with macrophage-rich areas, suggesting that interactions between macrophages and AClike_SLC7A5 cells may contribute to local immunosuppression and tumor progression. These findings provide novel insights into the cellular heterogeneity and tumor microenvironment dynamics in GBM. Importantly, the identified AClike_SLC7A5 subpopulation and its crosstalk with macrophages via CypA/APP signaling pathways may serve as potential biomarkers for patient prognosis or therapeutic targets. Future translational studies, including the development of inhibitors targeting these signaling axes or the use of spatial signatures for patient stratification, could pave the way for more personalized immunotherapeutic strategies in GBM.

This study has several noteworthy limitations. The cohort’s limited ethnic diversity and the underrepresentation of certain brain tumor subtypes may restrict the broader applicability of our results. Although our analysis concentrated on GBM, future studies should encompass additional brain malignancies, including neuroblastoma and oligodendroglioma. Moreover, while multi-omics integration allowed us to uncover previously unrecognized macrophage and AC-like subpopulations, their precise functional contributions to tumor progression remain to be experimentally validated, for instance, through organoid co-culture or targeted genetic perturbation. These constraints underscore the need for larger, multi-center cohorts and mechanistic investigations to facilitate the clinical translation of therapies targeting these newly defined cell populations.

Conclusion

In summary, our study constructs a detailed single-cell and spatial atlas of macrophage and AC-like subset interactions in GBM, uncovering key communication networks and distinct cellular niches that may contribute to immunosuppression and tumor progression. These insights offer a valuable framework for future mechanistic studies and lay the groundwork for the development of therapeutic strategies that focus on specific cellular crosstalk within GBM.

Supplementary Information

Below is the link to the electronic supplementary material.

Supplementary Material 1. (11.3KB, xlsx)
Supplementary Material 2. (99.4KB, docx)

Acknowledgements

Not applicable.

Author contributions

Conceptualization, Mengge Gao; Investigation, Qiong Liu; Software, Bo Yang; Formal Analysis, Mengge Gao; Writing—original draft, Mengge Gao; Writing—review & editing, Xiaofang Liu; Supervision, Xiaofang Liu; Visualization, Mengge Gao; Funding acquisition, Mengge Gao.

Funding

This study was supported by grant from the Huadu District People’s Hospital of Guangzhou Internal Research Fund Project (grant number: 2024C01).

Data availability

Single-cell RNA sequencing datasets were obtained from the Gene Expression Omnibus (GEO) under accession numbers GSE242044 [11] (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgiacc=GSE242044) and GSE103224 [12] (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi? acc=GSE103224). All data generated in this study are included in this published article. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Declarations

Ethical approval and consent to participate

The data utilized in this study were sourced from publicly accessible databases and were managed under approved ethical exemptions.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Footnotes

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

References

  • 1.Alexander BM, Cloughesy TF. Adult Glioblastoma. J Clin oncology: official J Am Soc Clin Oncol. 2017;35(21):2402–9. [DOI] [PubMed] [Google Scholar]
  • 2.Zhao X, Jakobsson V, Tao Y, Zhao T, Wang J, Khong PL, et al. Targeted radionuclide therapy in glioblastoma. ACS Appl Mater Interfaces. 2024;16(31):40391–410. [DOI] [PubMed] [Google Scholar]
  • 3.Hernández Martínez A, Madurga R, García-Romero N, Ayuso-Sacido Á. Unravelling glioblastoma heterogeneity by means of single-cell RNA sequencing. Cancer Lett. 2022;527:66–79. [DOI] [PubMed] [Google Scholar]
  • 4.Patel AP, Tirosh I, Trombetta JJ, Shalek AK, Gillespie SM, Wakimoto H, et al. Single-cell RNA-seq highlights intratumoral heterogeneity in primary glioblastoma. Sci (New York NY). 2014;344(6190):1396–401. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Pombo Antunes AR, Scheyltjens I, Lodi F, Messiaen J, Antoranz A, Duerinck J, et al. Single-cell profiling of myeloid cells in glioblastoma across species and disease stage reveals macrophage competition and specialization. Nat Neurosci. 2021;24(4):595–610. [DOI] [PubMed] [Google Scholar]
  • 6.Xiong A, Zhang J, Chen Y, Zhang Y, Yang F. Integrated single-cell transcriptomic analyses reveal that GPNMB-high macrophages promote PN-MES transition and impede T cell activation in GBM. EBioMedicine. 2022;83:104239. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Ochocka N, Segit P, Walentynowicz KA, Wojnicki K, Cyranowski S, Swatler J, et al. Single-cell RNA sequencing reveals functional heterogeneity of glioma-associated brain macrophages. Nat Commun. 2021;12(1):1151. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Neftel C, Laffy J, Filbin MG, Hara T, Shore ME, Rahme GJ, et al. An Integrative Model of Cellular States, Plasticity, and Genetics for Glioblastoma. Cell. 2019;178(4):835–e849821. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Al-Dalahmah O, Argenziano MG, Kannan A, Mahajan A, Furnari J, Paryani F, et al. Re-convolving the compositional landscape of primary and recurrent glioblastoma reveals prognostic and targetable tissue states. Nat Commun. 2023;14(1):2586. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Hu H, Huang Y, Li J, Liu R, Li H, Cai M, et al. Glioma promotes macrophage immunosuppressive phenotype through ANXA1 in a methionine metabolism-dependent manner. Discover Oncol. 2025;16(1):1269. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Xie Y, Yang F, He L, Huang H, Chao M, Cao H, et al. Single-cell dissection of the human blood-brain barrier and glioma blood-tumor barrier. Neuron. 2024;112(18):3089–e31053087. [DOI] [PubMed] [Google Scholar]
  • 12.Yuan J, Levitin HM, Frattini V, Bush EC, Boyett DM, Samanamud J, et al. Single-cell transcriptome analysis of lineage diversity in high-grade glioma. Genome Med. 2018;10(1):57. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Mei Y, Wang X, Zhang J, Liu D, He J, Huang C, et al. Siglec-9 acts as an immune-checkpoint molecule on macrophages in glioblastoma, restricting T-cell priming and immunotherapy response. Nat cancer. 2023;4(9):1273–91. [DOI] [PubMed] [Google Scholar]
  • 14.Satija R, Farrell JA, Gennert D, Schier AF, Regev A. Spatial reconstruction of single-cell gene expression data. Nat Biotechnol. 2015;33(5):495–502. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Korsunsky I, Millard N, Fan J, Slowikowski K, Zhang F, Wei K, et al. Fast, sensitive and accurate integration of single-cell data with Harmony. Nat Methods. 2019;16(12):1289–96. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Li X, Zhou Y, Jiang Q, Huang J, Liu Z, Li Y, et al. A novel scoring model based on RNA modification writers can predict the prognosis and guide immunotherapy in gastric cancer. Funct Integr Genom. 2023;23(2):162. [DOI] [PubMed] [Google Scholar]
  • 17.Kang M, Armenteros JJA, Gulati GS, Gleyzer R, Avagyan S, Brown EL et al. Mapping single-cell developmental potential in health and disease with interpretable deep learning. bioRxiv: the preprint server for biology 2024.
  • 18.Street K, Risso D, Fletcher RB, Das D, Ngai J, Yosef N, et al. Slingshot: cell lineage and pseudotime inference for single-cell transcriptomics. BMC Genomics. 2018;19(1):477. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Chen K, Wang Y, Hou Y, Wang Q, Long D, Liu X, et al. Single cell RNA-seq reveals the CCL5/SDC1 receptor-ligand interaction between T cells and tumor cells in pancreatic cancer. Cancer Lett. 2022;545:215834. [DOI] [PubMed] [Google Scholar]
  • 20.Wu T, Hu E, Xu S, Chen M, Guo P, Dai Z, et al. clusterProfiler 4.0: A universal enrichment tool for interpreting omics data. Innov (Cambridge (Mass). 2021;2(3):100141. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Jin S, Guerrero-Juarez CF, Zhang L, Chang I, Ramos R, Kuan CH, et al. Inference and analysis of cell-cell communication using CellChat. Nat Commun. 2021;12(1):1088. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Wei R, He S, Bai S, Sei E, Hu M, Thompson A, et al. Spatial charting of single-cell transcriptomes in tissues. Nat Biotechnol. 2022;40(8):1190–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Lin H, Liu C, Hu A, Zhang D, Yang H, Mao Y. Understanding the immunosuppressive microenvironment of glioma: mechanistic insights and clinical perspectives. J Hematol Oncol. 2024;17(1):31. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Khan F, Pang L, Dunterman M, Lesniak MS, Heimberger AB, Chen P. Macrophages and microglia in glioblastoma: heterogeneity, plasticity, and therapy. J Clin Investig 2023; 133(1). [DOI] [PMC free article] [PubMed]
  • 25.Zafrakas M, Petschke B, Donner A, Fritzsche F, Kristiansen G, Knüchel R, et al. Expression analysis of mammaglobin A (SCGB2A2) and lipophilin B (SCGB1D2) in more than 300 human tumors and matching normal tissues reveals their co-expression in gynecologic malignancies. BMC Cancer. 2006;6:88. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Yu M, Wang P, Liu R, Li M. Characterization of immune landscape and prognostic value of IL-17-related signature in uterine corpus endometrial carcinoma. Clin translational oncology: official publication Federation Span Oncol Soc Natl Cancer Inst Mexico 2026. [DOI] [PubMed]
  • 27.Ntanasis-Stathopoulos I, Fotiou D, Terpos E. CCL3 Signaling in the Tumor Microenvironment. Adv Exp Med Biol. 2020;1231:13–21. [DOI] [PubMed] [Google Scholar]
  • 28.Kersten K, Hu KH, Combes AJ, Samad B, Harwin T, Ray A, et al. Spatiotemporal co-dependency between macrophages and exhausted CD8(+) T cells in cancer. Cancer Cell. 2022;40(6):624–e638629. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Wang C, Chen S, Guo H, Jiang H, Liu H, Fu H, et al. Forsythoside A Mitigates Alzheimer’s-like Pathology by Inhibiting Ferroptosis-mediated Neuroinflammation via Nrf2/GPX4 Axis Activation. Int J Biol Sci. 2022;18(5):2075–90. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Stoll AC, Kemp CJ, Patterson JR, Howe JW, Steece-Collier K, Luk KC, et al. Neuroinflammatory gene expression profiles of reactive glia in the substantia nigra suggest a multidimensional immune response to alpha synuclein inclusions. Neurobiol Dis. 2024;191:106411. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Jin S, Zhang M, Qiao X. Cyclophilin A: promising target in cancer therapy. Cancer Biol Ther. 2024;25(1):2425127. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Liu Z, Yang Y, Fang H, Cen B, Fan Y, Li J, et al. Single-cell and spatial analyses reveal the effect of VSIG4(+)S100A10(+)TAMs on the immunosuppression of glioblastoma and anti-PD-1 immunotherapy. Int J Biol Macromol. 2025;308(Pt 3):142415. [DOI] [PubMed] [Google Scholar]
  • 33.Chen X, Chen Y, Chen X, Wei P, Lin Y, Wu Z, et al. Single-cell RNA sequencing reveals intra-tumoral heterogeneity of glioblastoma and a pro-tumor subset of tumor-associated macrophages characterized by EZH2 overexpression. Biochim et Biophys acta Mol basis disease. 2022;1868(12):166534. [DOI] [PubMed] [Google Scholar]
  • 34.Wang S, Li L, Zuo S, Kong L, Wei J, Dong J. Metabolic-related gene pairs signature analysis identifies ABCA1 expression levels on tumor-associated macrophages as a prognostic biomarker in primary IDH(WT) glioblastoma. Front Immunol. 2022;13:869061. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Ding C, Dong J, Pan Z, Liu S, Song Q, Yang G et al. Glioblastoma-secreted C1QL1 orchestrates tumor microtube expansion and neural synaptic pruning to drive malignant synapse formation and recurrence. Cancer Discov 2026. [DOI] [PubMed]
  • 36.Weng KQ, Liu JY, Li H, She LL, Qiu JL, Qi H, et al. Identification of Treg-related prognostic molecular subtypes and individualized characteristics in clear cell renal cell carcinoma through single-cell transcriptomes and bulk RNA sequencing. Int Immunopharmacol. 2024;130:111746. [DOI] [PubMed] [Google Scholar]
  • 37.Revel M, Sautès-Fridman C, Fridman WH, Roumenina LT. C1q+ macrophages: passengers or drivers of cancer progression. Trends cancer. 2022;8(7):517–26. [DOI] [PubMed] [Google Scholar]
  • 38.Zhang S, Peng W, Wang H, Xiang X, Ye L, Wei X et al. C1q(+) tumor-associated macrophages contribute to immunosuppression through fatty acid metabolic reprogramming in malignant pleural effusion. J Immunother Cancer 2023; 11(8). [DOI] [PMC free article] [PubMed]
  • 39.Veatch JR, Lee SM, Shasha C, Singhi N, Szeto JL, Moshiri AS, et al. Neoantigen-specific CD4(+) T cells in human melanoma have diverse differentiation states and correlate with CD8(+) T cell, macrophage, and B cell function. Cancer Cell. 2022;40(4):393–e409399. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Liu M, Ji Z, Jain V, Smith VL, Hocke E, Patel AP, et al. Spatial transcriptomics reveals segregation of tumor cell states in glioblastoma and marked immunosuppression within the perinecrotic niche. Acta Neuropathol Commun. 2024;12(1):64. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41.Yi L, Wang W, Sun Z, Chen Y, Xiong Z, Ma L, et al. Deciphering the carcinogenic role of benzo[a]pyrene in glioblastoma: Insights from network toxicology, single-cell transcriptomics, and Mendelian randomization. Ecotoxicol Environ Saf. 2025;304:119155. [DOI] [PubMed] [Google Scholar]
  • 42.Hide T, Komohara Y, Miyasato Y, Nakamura H, Makino K, Takeya M, et al. Oligodendrocyte Progenitor Cells and Macrophages/Microglia Produce Glioma Stem Cell Niches at the Tumor Border. EBioMedicine. 2018;30:94–104. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43.Long X, Zhang Z, Li Y, Deng K, Gao W, Huang M et al. ScRNA-seq reveals novel immune-suppressive T cells and investigates CMV-TCR-T cells cytotoxicity against GBM. J Immunother Cancer 2024; 12(4). [DOI] [PMC free article] [PubMed]
  • 44.Luo F, Mei Y, Li Y, Yang J, Xi S, Cao E et al. CAF-derived LRRC15 orchestrates macrophage polarization and limits PD-1 immunotherapy efficacy in glioblastoma. Neuro-oncology 2025. [DOI] [PMC free article] [PubMed]
  • 45.Liu Y, Wu J, Najem H, Lin Y, Pang L, Khan F et al. Dual targeting macrophages and microglia is a therapeutic vulnerability in models of PTEN-deficient glioblastoma. J Clin Investig 2024; 134(22). [DOI] [PMC free article] [PubMed]
  • 46.Chryplewicz A, Scotton J, Tichet M, Zomer A, Shchors K, Joyce JA, et al. Cancer cell autophagy, reprogrammed macrophages, and remodeled vasculature in glioblastoma triggers tumor immunity. Cancer Cell. 2022;40(10):1111–e11271119. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 47.Zhai K, Huang Z, Huang Q, Tao W, Fang X, Zhang A, et al. Pharmacological inhibition of BACE1 suppresses glioblastoma growth by stimulating macrophage phagocytosis of tumor cells. Nat cancer. 2021;2(11):1136–51. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Material 1. (11.3KB, xlsx)
Supplementary Material 2. (99.4KB, docx)

Data Availability Statement

Single-cell RNA sequencing datasets were obtained from the Gene Expression Omnibus (GEO) under accession numbers GSE242044 [11] (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgiacc=GSE242044) and GSE103224 [12] (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi? acc=GSE103224). All data generated in this study are included in this published article. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Articles from Discover Oncology are provided here courtesy of Springer

RESOURCES