Integrative transcriptomic and experimental analyses identifies SDF4 as an astrocyte-expressed regulator of glioma progression through PI3K-AKT signaling

Abstract

Background

Glioma is an aggressive brain tumor with poor prognosis and high recurrence, which significantly affects patients’ quality of life. Identifying new biomarkers and therapeutic targets is crucial for improving diagnosis and treatment. This study focuses on the SDF4 gene, which has been implicated in various cancers; however, its role in glioma remains underexplored due to limited research.

Methods

To comprehensively investigate the role of SDF4 in glioma, we integrated bulk, single-cell, and spatial transcriptomic analyses with experimental validation. This approach includes bioinformatics evaluations, RNA sequencing analysis, single-cell RNA sequencing (scRNA-seq), spatial transcriptomics, survival analyses, gene enrichment assessments, and in vitro experimental validation. Furthermore, we explored the relationship between SDF4 expression and specific clinical parameters such as tumor grade and patient prognosis. The expression levels of SDF4 in invasive glioma specimens were assessed through immunohistochemical techniques.

Results

Our investigation reveals that SDF4 is markedly upregulated in glioma and correlates with unfavorable prognostic outcomes. Through Least Absolute Shrinkage and Selection Operator (LASSO) Cox regression and weighted gene co-expression network analysis (WGCNA), SDF4 was identified as a pivotal prognostic biomarker; elevated levels of SDF4 were significantly linked to decreased overall survival rates. Functional enrichment analyses indicate that SDF4 regulates the cell cycle, mediates apoptosis, and remodels the extracellular matrix. Furthermore, single-cell and spatial transcriptomic studies have pinpointed SDF4 expression specifically within astrocytes, underscoring its essential function within the tumor microenvironment. In vitro experiments demonstrated that silencing SDF4 reduces glioma cell proliferation and promotes apoptosis; this effect is primarily mediated through the PI3K-Akt signaling pathway, as silencing effectively curtails its activation. Moreover, a notable elevation in SDF4 protein levels was observed in glioma patients. These elevated levels were associated with genetic mutations such as IDH1 and TERT, as well as clinical features including World Health Organization (WHO) classification and the methylation status of MGMT.

Conclusion

Our research underscores the critical role of SDF4 in glioma progression, highlighting its potential as a strategic target for glioma management and as a prognostic marker. Additionally, these findings pave the way for novel therapeutic strategies.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12885-026-16635-6.

Background

Gliomas, particularly glioblastomas, are among the most aggressive primary, and are characterized by diffuse infiltration, frequent recurrence, and poor prognosis [1, 2]. Despite maximal safe resection followed by radiotherapy with concurrent and adjuvant temozolomide, long-term survival remains limited [3–6]. Population-based data indicate that the five-year survival rates for glioblastoma is approximately 7.0% overall and only 5.6% among patients aged 40 years or older [7]. These outcomes underscore the need to identify molecular determinants of glioma progression and clinically relevant prognostic and therapeutic targets.

The glioma tumor microenvironment (TME) is a critical determinant of tumor growth, invasion, immune evasion, and treatment resistance [8–10]. It comprises malignant cells, resident and infiltrating immune cells, vascular and stromal cells, and extracellular matrix components that communicate through direct contact and soluble factors [11, 12]. Tumor-associated macrophages (TAMs) and microglia constitute major immune populations in glioma; tumor-derived signals recruit and reprogram these cells toward immunosuppressive, pro-angiogenic states, thereby promoting tumor progression and limiting antitumor immunity [13–16]. Accordingly, defining previously unrecognized tumor-derived mediators that coordinate malignant-cell behavior and immune-cell remodeling may reveal tractable vulnerabilities in glioma.

Stromal cell-derived factor 4 (SDF4), a calcium-binding member of the CREC family, is aberrantly expressed in several cancers and has been implicated in tumor-cell proliferation, migration, angiogenesis, and stress responses [17, 18]. Clinical studies have also suggested that elevated SDF4 expression or circulating levels may have diagnostic or prognostic value in certain malignancies [19]. However, evidence from other tumor types cannot be directly extrapolated to glioma. The expression pattern, cellular source, spatial distribution, prognostic significance, and biological function of SDF4 in the glioma microenvironment remain poorly defined. In particular, whether SDF4 links malignant glioma cells to microenvironmental remodeling and contributes to glioma progression has not been systematically investigated.

To address this gap, we integrated bulk transcriptomic analysis, single-cell RNA sequencing, spatial transcriptomics, cell–cell communication analysis, and in vitro functional validation to characterize SDF4 in glioma from population, cellular, spatial, and functional perspectives. We evaluated its association with clinicopathological features and prognosis, identified the principal cellular sources and spatial distribution of SDF4, examined its potential involvement in intercellular communication, and experimentally assessed its effects on glioma-cell proliferation, apoptosis, cell-cycle progression, and PI3K-AKT signaling. This evidence-driven framework was designed to determine whether SDF4 represents a biologically relevant prognostic biomarker and a candidate therapeutic target in glioma.

Materials and methods

Gene expression profiles of tumor and adjacent normal tumor tissues

The RNA sequencing datasets for both normal and tumor specimens were sourced from The Cancer Genome Atlas (TCGA) (http://cancergenome.nih.gov) and the Genotype-Tissue Expression (GTEx) (http://commonfund.nih.gov/GTEx/) initiatives [20, 21]. The prognostic validation cohort was extracted from the Chinese Glioma Genome Atlas (CGGA) database (http://www.cgga.org.cn) [22]. In addition, a total of 46 genes linked to stromal cell-derived factors, collectively referred to as Stromal Cell-Related Genes (SCRGs), were curated from GeneCards, with each exhibiting a relevance score exceeding 50 [23]. Moreover, the scRNA-seq dataset labeled as GSE138794 was retrieved from the GEO database, while the spatial transcriptomics dataset was acquired from GSE253080 [24, 25]. The dataset for Mendelian randomization was obtained from the IEU OpenGWAS database [26].

Weighted gene co-expression network analysis (WGCNA)

Gene co-expression network analysis was conducted using the WGCNA R package (version 1.72.3) [27]. The pickSoftThreshold function was applied to evaluate scale-free topology and mean connectivity across a range of candidate powers, and a soft-thresholding power of 8 was selected. A topological overlap matrix was then constructed to quantify network connectivity among genes with similar expression patterns. Finally, hierarchical clustering combined with the dynamic hybrid tree-cutting algorithm was used to identify co-expression modules.

Construction of the SCRGs risk model

Utilizing the TCGA database, nine potential SCRGs were identified through the application of least absolute shrinkage and selection operator (LASSO) Cox regression analysis [28]. This LASSO Cox regression analysis was employed to minimize redundancy and mitigate the risk of model overfitting. Following this, the three genes were chosen to construct a prognostic risk-scoring model to predict overall survival (OS) in glioma patients.

Gene enrichment analysis

Enrichment analyses were performed on the differentially expressed genes linked to glioma by employing Gene Ontology (GO) and the Kyoto Encyclopedia of Genes and Genomes (KEGG), utilizing the R package clusterProfiler for this purpose [29]. The samples were categorized into high-expression and low-expression cohorts according to the median expression level of the gene SDF4. Gene Set Enrichment Analysis (GSEA) was performed utilizing the clusterProfiler package [30]. Significant findings were identified as gene sets that demonstrated a normalized enrichment score (NES) exceeding 1, alongside a false discovery rate (FDR) of below 0.05.

Development of nomograms

Nomograms incorporating clinical features and the SDF4 models were developed using the “rms” R package to predict OS in glioma patients. In order to assess the predictive accuracy of the nomograms, a comprehensive evaluation was conducted utilizing time-dependent calibration curves. This approach facilitated the comparison between predicted outcomes and actual survival data across different time intervals. Furthermore, a univariate Cox regression analysis was performed to determine the potential of the SDF4 model as an independent prognostic marker for OS in glioma patients. In addition, receiver operating characteristic (ROC) curves were utilized to calculate the area under the curve (AUC) value, offering a quantitative measure of the diagnostic performance and relevance of the nomogram in predicting patient outcomes.

Mendelian randomization (MR) analysis

GWAS summary data for SDF4 expression (eqtl-a-ENSG00000078808) were obtained from the IEU OpenGWAS database. Independent SDF4-associated SNPs were selected as instrumental variables after excluding linkage disequilibrium-correlated, ambiguous, and weak instruments. Effect alleles were harmonized using the harmonise_data function in the TwoSampleMR package [26]. The MR function combined five algorithms (MR egger, weighted median, inverse variance weighted (IVW), simple mode, and weighted mode to perform MR analysis. In addition, sensitivity analyses were performed to verify the robustness of the findings. Heterogeneity, horizontal pleiotropy, and robustness were evaluated using Cochran’s Q test, the MR-Egger intercept test, funnel plots, and leave-one-out analysis.

ScRNA-seq data integration and analysis

The scRNA-seq dataset identified as GSE138794 was analyzed using the Seurat package in the R programming environment [31]. A thorough quality control procedure was implemented to filter the cells, which included verifying that the mitochondrial UMI ratio was maintained below 10%. Following the selection of cells, it is essential to conduct harmony integration [32], a process that facilitates the identification of genes with significant variability (Figure S1). The RunPCA function was utilized to perform Principal Component Analysis (PCA), which enabled the reduction of data dimensionality (Figure S2). The dataset that had been normalized was then merged employing the LogNormalization approach. After this procedure, cell clustering was achieved by leveraging the “FindNeighbors” and “FindClusters” functions. A resolution parameter of 0.5 was selected to enhance the efficacy of cell clustering (Figure S3). Furthermore, a bubble chart was generated to illustrate the expression levels of marker genes among different clusters (Figure S4). The annotations for cell types were obtained by identifying cell cluster markers from the CellMarker 2.0 database (http://bio-bigdata.hrbmu.edu.cn/CellMarker/) [33].

Spatial transcriptome analysis

The spatial transcriptomics dataset employed in this investigation was subsequently examined and represented using the Seurat R package. The histogram corresponding to each sample (GSM8013395 to GSM8013410) demonstrates the distribution of nCount_Spatial values across all spots. Concurrently, the spatial map illustrates the distribution of diverse cell types, with distinct colors representing different cell clusters (Figure S5). Genes associated with mitochondria, ribosomes, and red blood cells were excluded from the analysis, along with any genes exhibiting expression levels below 10 spots. Subsequently, the dataset underwent normalization through the SCTransform method, which was succeeded by dimensionality reduction via PCA [34]. The identification of distinct cell populations was performed utilizing HE staining sections and genes that displayed variability across clusters. The analysis of the spatial transcriptomics data was conducted employing the SpatialDimPlot and SpatialFeaturePlot functions.

Patients and tissue specimens

Tissue samples were obtained from individuals diagnosed with various cancers, including glioma, laryngeal cancer, gastric cancer, liver cancer, breast cancer, prostate cancer, colon adenocarcinoma, bladder urothelial carcinoma, and esophageal cancer, from the Department of Pathology at the Second Affiliated Hospital of Zhejiang University School of Medicine. This research adhered rigorously to the ethical principles established in the Declaration of Helsinki. The exclusion criteria included: (i) the presence of autoimmune diseases or other significant health issues; (ii) the existence of severe comorbidities; and (iii) a history of immunosuppressive therapy. The study protocols were approved by the Ethics Committee of the Second Affiliated Hospital of Zhejiang University School of Medicine in Hangzhou, China (approval number: 2025–2030). A total of 46 patients diagnosed with glioma were selected for inclusion in this study. Each patient had previously undergone surgical resection at this facility, and tumor specimens, along with their associated medical records, were collected. The Ethics Committee granted approval for this research and waived the necessity for informed consent from the participating patients.

Immunohistochemistry (IHC) analysis

Immunohistochemical analyses were conducted following established methodologies [35]. The primary antibodies utilized in this investigation comprised SDF4 (dilution 1:1,000; Proteintech), Ki-67 (dilution 1:1,000; Zhongshan Golden Bridge Biotechnology Co.), GFAP (dilution 1:1,000; Zhongshan Golden Bridge Biotechnology Co.), Olig2 (dilution 1:1,000; Zhongshan Golden Bridge Biotechnology Co.), IDH1 (dilution 1:1,000; Zhongshan Golden Bridge Biotechnology Co.), and P53 (dilution 1:1,000; Zhongshan Golden Bridge Biotechnology Co.). The staining outcomes were meticulously assessed by a trio of pathologists, who allocated scores according to predetermined criteria. Tumor intensity was classified into four categories: 0 (negative), 1 (weak positive), 2 (moderate positive), or 3 (strong positive); tumor extent was categorized as 0 (0–10%), 1 (10%-25%), 2 (26%-50%), 3 (51%-75%), or 4 (76%-100%). The comprehensive score was calculated by multiplying the intensity score by the extent score.

Cell culture and transfection

The H4 and SW1783 glioma cell lines were obtained from the BeNa Culture Collection. These cell lines were cultured in DMEM-H (Hyclone) and L-15 medium (Procell Life Science & Technology), respectively. Both media were supplemented with 10% fetal bovine serum (Yousi Biotechnology) and 1% antibiotics (Gino Biotechnology). Small interfering RNAs (siRNAs) designed to target SDF4, along with a negative control, were synthesized by ELK Biotechnology. Transfection was conducted using Lipofectamine 2000 (Invitrogen) after the cells were plated and reached 70% confluence. Following transfection, the cells were incubated at 37 °C for 48 h prior to analysis. The specific sequences were as follows: SDF4-targeting siRNAs (siR-SDF4#1: 5’-TAGCCAACAGGGAGGAGAATGAGAT-3’; siR-SDF4#2: 5’-CGAGGAACTCAAAGTGGATGAGGAA-3’; siR-SDF4#3:5’-TGGAAATGAGTAGCCAGGAAGTTCA-3’) and negative control siRNA (siR-SDF4-NC: 5’-TAGAGGAGCCAACAGGGAATGAGAT-3’).

Colony formation assay

Approximately 500 transfected cells were inoculated into a 6-well plate and allowed to culture for a duration of two weeks. Following this incubation period, the cells were fixed using a 4% paraformaldehyde solution, subsequently stained with crystal violet, and washed with PBS. Finally, colonies were visualized and quantified using an Olympus IX51 microscope.

EdU assay

Cell proliferation was evaluated utilizing the EdU Cell Proliferation Assay Kit (RiboBio Co.) in cells exhibiting logarithmic growth. Following a 24-hour period after transfection, cells were plated at a density of 1 × 10⁵ per well within 6-well plates and allowed to incubate overnight. Subsequently, 100 µl of a 50 µM EdU solution was introduced and incubated at 37℃ for a duration of 2 h. The cells were then fixed using 50 µl of 4% paraformaldehyde for 30 min, followed by neutralization with glycine and permeabilization with 0.5% Triton X-100 for 20 min. For the staining process, 100 µl of Apollo and Hoechst solutions were added in succession, with each solution incubated for 30 min in a dark environment. Fluorescent images were captured, and the proliferation rate was quantified through ImageJ software by analyzing the number of EdU-positive cells.

Cell cycle analysis

Cell cycle was detected using the Cell Cycle Detection Kit (40301ES60, Yeasen Biotech). After harvesting the cells, they were fixed with 70% ethanol at 4℃ overnight. Subsequently, the cells were washed, and a staining working solution was prepared by adding 10 µL propidium iodide (PI) stock solution (40301-B) and 10 µL RNase A (40301-A) to 0.5 mL staining buffer (40301-C). Each cell sample was incubated with 0.5 mL of the prepared staining working solution at 37℃ for 30 min in the dark. Immediately after staining, cell cycle was detected using a CytoFLEX flow cytometer (Beckman Coulter, San Diego, California, USA), and the detection results were analyzed with CytExpert software.

Apoptosis assay

Cell apoptosis was detected using the Cell Apoptosis Detection Kit (Cat. No.: 556547, Shanghai Weijing Biotechnology Co., Ltd. (Wegene Bio)). After harvesting the cells, they were resuspended in binding buffer. Subsequently, 5 µl Annexin V-FITC and 5 µl propidium iodide (PI) were added sequentially, followed by incubation for 15 min in the dark. The apoptosis rate was detected by flow cytometry (Beckman Coulter, USA).

Quantitative reverse transcription polymerase chain reaction (qRT-PCR)

The TRIpure Total RNA Extraction Reagent (ELK Biotechnology Co.) was employed for the isolation of RNA from either tissues or cells. Subsequently, reverse transcription was performed to synthesize complementary DNA (cDNA) utilizing the EntiLink™ kit (ELK Biotechnology Co.) under conditions of 42 °C for 30 min, followed by a heat inactivation step at 85 °C for 5 min. Quantitative reverse transcription polymerase chain reaction (qRT-PCR) was executed using EnTurbo™ SYBR Green PCR SuperMix (ELK Biotechnology Co.) in conjunction with the QuantStudio 6 Flex system (Invitrogen). Specific experimental conditions were adhered to, and the expression levels of SDF4 were quantified relative to β-actin using the 2^(-ΔΔCt) methodology. The primer sequences for both SDF4 and β-actin were supplied as part of the experimental protocol. The primer sequences for SDF4 and β-actin were as follows: human SDF4 (forward: 5’-TGTCTTGGGACGAGTATAAGGTG-3’; reverse: 5’-GTTTCCTCATCCACTTTGAGTTC-3’); human actin (forward: 5’-GTCCACCGCAAATGCTTCTA-3’; reverse: 5’-TGCTGTCACCTTCACCGTTC-3’).

Western blot

Proteins were isolated utilizing RIPA Buffer (Aspen Biotechnology) and subsequently quantified employing a BCA Kit (Aspen Biotechnology). A total of 40 µg of protein per lane was subjected to separation via 10% SDS-PAGE, followed by transfer onto PVDF membranes. The membranes were incubated in 5% non-fat milk for one hour to block non-specific binding sites, and subsequently exposed to primary antibodies overnight at 4 °C. The antibodies included Cyclin D1 (dilution 1:3000; Abcam), CDK4 (dilution 1:1000; CST), CDK6 (dilution 1:1000; Proteintech), P21 (dilution 1:1000; CST), P27 (dilution 1:1000; CST), BAX (dilution 1:2000; CST), Bcl-2 (dilution 1:1000; Abcam), Cleaved caspase-3 (dilution 1:500; Affbiotech), SDF4 (dilution 1:500; Santa Cruz Biotechnology), p-PI3K (dilution 1:1000; Affbiotech), PI3K (dilution 1:2000; CST), p-AKT (dilution 1:1000; CST), AKT (dilution 1:2000; CST), and β-actin (dilution 1:10000; Beijing TDY Biotechnology Co.) targeting various proteins of interest. The following day, membranes underwent washes with TBST and were then incubated with HRP-conjugated secondary antibodies (dilution 1:10000; Aspen Biotechnology) for one hour, according to the source of the primary antibodies. Finally, signal visualization was performed using a Canon Lide 110 scanner.

Statistical analysis

Statistical analyses were performed using R software (version 4.4.0). Continuous data are presented as mean ± SD, and all in vitro experiments were independently repeated at least three times. Normality and variance homogeneity were assessed using the Shapiro–Wilk and Levene’s tests, respectively. Two-group comparisons were conducted using an unpaired two-tailed Student’s t-test or Wilcoxon rank-sum test, while multiple-group comparisons used one-way ANOVA with Tukey’s post hoc test or the Kruskal–Wallis test with Dunn’s test, as appropriate. Correlations were assessed using Spearman’s and distance correlation analyses. Survival was evaluated using Kaplan–Meier curves and the log-rank test. All tests were two-sided, with P < 0.05 considered statistically significant.

Results

Development of a prognostic risk model and identification of key genes.

LASSO Cox regression identified 11 genes with potential prognostic relevance (Figs. 1A and B), of which nine were retained for multivariate Cox analysis (Fig. 1C). Higher expression of CFH, VEGFA, SDHD, F3, and SDF4 was independently associated with increased mortality risk, whereas GDNF was protective. The expression heatmap further indicated that most candidate genes were associated with adverse clinicopathological features (Fig. 1D).

Fig. 1.

Identification of prognostic genes associated with stromal cell-derived factor in glioma. A Ten-fold cross-validation curve for the LASSO Cox model. Dashed lines indicate the minimum-error and 1-standard-error λ values, and the numbers indicate retained variables. B Coefficient trajectories of candidate stromal cell-related genes across the log(λ) sequence. C Multivariable Cox regression forest plot showing hazard ratios (HRs), 95% confidence intervals, and P values. HR > 1 indicates increased mortality risk, whereas HR < 1 indicates a protective association. D Heatmap of candidate-gene expression and clinicopathological features, including WHO grade, IDH status, 1p/19q codeletion, and survival outcome. Expression values are row-scaled Z scores. E WGCNA module–trait relationship heatmap. The turquoise module showed the strongest association with tumor status (r = 0.94, P < 0.001). F Venn diagram identifying SDF4 and VEGFA as overlapping genes between the LASSO candidates and the turquoise module. SDF4 was prioritized because its role in glioma remains comparatively underexplored. Survival associations were assessed using Cox proportional-hazards regression. LASSO, least absolute shrinkage and selection operator; WGCNA, weighted gene co-expression network analysis; IDH, isocitrate dehydrogenase

WGCNA identified the turquoise module as the module most strongly associated with glioma status (r = 0.94, P < 0.001), comprising 2,345 genes (Fig. 1E). Intersection of these genes with the 11 LASSO-derived candidates identified SDF4 and VEGFA as shared genes (Fig. 1F). Because VEGFA is already well characterized in glioma angiogenesis and progression, SDF4 was prioritized based on research novelty and study scope rather than statistical superiority.

The expression levels of SDF4 in multiple tumor types are significantly correlated with patient prognosis

Pan-cancer analysis showed that SDF4 expression was significantly higher in tumor tissues than in corresponding normal tissues in 18 cancer types, including bladder urothelial carcinoma (BLCA), breast invasive carcinoma (BRCA), cholangiocarcinoma (CHOL), colon adenocarcinoma (COAD), lymphoid neoplasm diffuse large B-cell lymphoma (DLBC), esophageal carcinoma (ESCA), glioblastoma multiforme (GBM), head and neck squamous cell carcinoma (HNSC), kidney renal clear cell carcinoma (KIRC), kidney renal papillary cell carcinoma (KIRP), brain lower-grade glioma (LGG), liver hepatocellular carcinoma (LIHC), pancreatic adenocarcinoma (PAAD), prostate adenocarcinoma (PRAD), skin cutaneous melanoma (SKCM), stomach adenocarcinoma (STAD), thyroid carcinoma (THCA), and thymoma (THYM) (Figs. 2A and B). Conversely, reduced SDF4 expression was observed in six cancer types, suggesting context-dependent roles across malignancies. SDF4 expression was also associated with immune-cell infiltration (Fig. 2C). Univariate Cox analysis of TCGA data across 33 cancer types further showed that high SDF4 expression was significantly associated with poorer overall survival in ACC, BLCA, CESC, KICH, LGG, and LIHC (Figs. 2D–F).

Fig. 2.

The expression levels and survival analysis of SDF4 across different tumor types. A Violin plots comparing SDF4 mRNA expression between tumor and normal tissues across TCGA cancer types, with GTEx normal tissues included where applicable. SDF4 expression was significantly higher in BLCA, BRCA, CHOL, COAD, DLBC, ESCA, GBM, HNSC, KIRC, KIRP, LGG, LIHC, PAAD, PRAD, SKCM, STAD, THCA, and THYM, but lower in ACC, KICH, LAML, OV, READ, and TGCT. B Radar plot summarizing SDF4 expression across cancer types. C Heatmap showing correlations between SDF4 expression and immune-cell infiltration; red and blue indicate positive and negative correlations, respectively. D Univariate Cox regression forest plot showing hazard ratios (HRs) and 95% confidence intervals for overall survival. E Heatmap summarizing survival-associated HRs across cancer types. F Kaplan–Meier curves for ACC, BLCA, CESC, KICH, combined GBM/LGG, and LIHC, stratified into SDF4-high and SDF4-low groups using the prespecified cutoff. Tumor-versus-normal comparisons were performed using the statistical test described in Methods. Immune-cell associations were assessed by correlation analysis, survival curves by the log-rank test, and HRs by univariate Cox regression. *P < 0.05, **P < 0.01, and ***P < 0.001. Cancer abbreviations follow TCGA nomenclature

GO, KEGG, and GSEA of SDF4 expression in glioma

GO enrichment analysis suggested that SDF4-related genes were mainly involved in cell migration, endothelial-cell proliferation, fibroblast differentiation, extracellular-matrix organization, calcium-dependent secretion, vesicle regulation, and cell–matrix interactions (Figs. 2A–C; Table 1). KEGG analysis further indicated enrichment in calcium signaling, PI3K-AKT signaling, extracellular matrix–receptor interaction, and neuroactive ligand–receptor interaction pathways (Fig. 2D; Table 1).

Table 1.

Supplementary information of GO and KEGG analysis

Otology	ID	Description	P. val
BP	GO:0043062	extracellular structure organization	9.79346E-15
BP	GO:0060326	cell chemotaxis	5.97629E-06
BP	GO:0045785	positive regulation of cell adhesion	0.005917966
BP	GO:0048762	mesenchymal cell differentiation	0.000292898
BP	GO:0045766	positive regulation of angiogenesis	0.002729777
BP	GO:0017156	calcium-ion regulated exocytosis	0.01133921
CC	GO:0031983	vesicle lumen	0.003622797
CC	GO:0005604	basement membrane	0.000101302
CC	GO:0031253	cell projection membrane	0.001047144
CC	GO:0098878	neurotransmitter receptor complex	4.17468E-05
CC	GO:0008021	synaptic vesicle	0.00190618
MF	GO:0061134	peptidase regulator activity	0.001677162
MF	GO:0005178	integrin binding	0.000888901
MF	GO:0070851	growth factor receptor binding	0.009390267
MF	GO:0050840	extracellular matrix binding	0.00124884
KEGG	hsa05202	Transcriptional misregulation in cancer	6.16484E-05
KEGG	hsa04151	PI3K-Akt signaling pathway	0.000881709
KEGG	hsa04512	ECM-receptor interaction	4.00148E-08
KEGG	hsa04020	Calcium signaling pathway	0.005471226
KEGG	hsa04080	Neuroactive ligand-receptor interaction	7.05824E-14

Consistently, enrichment analysis comparing the SDF4-high and SDF4-low groups showed that high SDF4 expression was associated with PI3K-AKT signaling, focal adhesion, cancer-related pathways, apoptosis, programmed cell death, and cell-cycle regulation (Figs. 2E and F). The heatmap identified five PI3K-AKT pathway genes with significant expression changes, defined by |log fold change| > 1 and P < 0.05 (Fig. 2G).

Clinical prognostic value of SDF4 in glioma

Marked discrepancies were identified in various parameters, including WHO grade, IDH status, 1p/19q codeletion, histological type, age, OS, disease-specific survival (DSS), and progression-free interval (PFI), when comparing glioma patients exhibiting high versus low levels of SDF4 expression (Figs. 3A-H; Table 2). The robustness of our risk model was further substantiated through ROC curve analysis (Figs. 3I-K). Additionally, we examined OS outcomes in patients stratified by low and high SDF4 expression levels across distinct phenotypes. The findings indicated that increased SDF4 expression correlated with diminished survival rates among patients classified as G2, G3, those with IDH mutations, and individuals with 1p/19q non-codeletion, as well as across all age groups diagnosed with astrocytoma or glioblastoma (Figs. 3L-T).

Fig. 3.

Enrichment analysis of SDF4. A–C Circular Gene Ontology enrichment plots for biological process, cellular component, and molecular function based on genes differentially expressed between the SDF4-high and SDF4-low groups. Outer points show gene-level log2 fold changes, and inner sectors indicate pathway-level Z scores. D Circular KEGG enrichment plot showing representative SDF4-associated pathways. E GSEA curves for focal adhesion–PI3K/AKT/mTOR signaling, PI3K-AKT signaling, pathways in cancer, focal adhesion, cell cycle, cell-cycle checkpoints, apoptosis, and programmed cell-death pathways. F Ridge plots summarizing enrichment of these pathways. G Samples ranked by SDF4 expression, with the heatmap showing standardized expression of representative PI3K-AKT pathway genes, including PIK3R6, PTEN, PDK1, AKT1, and MTOR. These analyses indicate that high SDF4 expression is associated with growth-, survival-, cell-cycle-, and adhesion-related transcriptional programs but do not establish direct pathway activation. SDF4-high and SDF4-low groups were defined using the median expression level. Significant enrichment was defined as NES > 1 and FDR < 0.05. BP, biological process; CC, cellular component; MF, molecular function; KEGG, Kyoto Encyclopedia of Genes and Genomes; GSEA, gene set enrichment analysis; NES, normalized enrichment score; FDR, false-discovery rate

Table 2.

Association of SDF4 expression with clinicopathological characteristics in glioma patients from the TCGA database

Characteristics	Low expression of SDF4	High expression of SDF4	P value
n	349	350
WHO grade, n (%)			< 0.001
G2	172 (27%)	52 (8.2%)
G3	130 (20.4%)	115 (18.1%)
G4	14 (2.2%)	154 (24.2%)
IDH status, n (%)			< 0.001
WT	27 (3.9%)	219 (31.8%)
Mut	317 (46%)	126 (18.3%)
1p/19q codeletion, n (%)			< 0.001
Non-codel	184 (26.6%)	336 (48.6%)
Codel	165 (23.8%)	7 (1%)
Gender, n (%)			0.670
Female	146 (20.9%)	152 (21.7%)
Male	203 (29%)	198 (28.3%)
Age, n (%)			< 0.001
<= 60	309 (44.2%)	247 (35.3%)
> 60	40 (5.7%)	103 (14.7%)
Histological type, n (%)			< 0.001
Astrocytoma	88 (12.6%)	108 (15.5%)
Oligoastrocytoma	86 (12.3%)	49 (7%)
Oligodendroglioma	161 (23%)	39 (5.6%)
Glioblastoma	14 (2%)	154 (22%)
OS event, n (%)			< 0.001
Alive	286 (40.9%)	141 (20.2%)
Dead	63 (9%)	209 (29.9%)
DSS event, n (%)			< 0.001
No	288 (42.5%)	146 (21.5%)
Yes	58 (8.6%)	186 (27.4%)
PFI event, n (%)			< 0.001
No	241 (34.5%)	112 (16%)
Yes	108 (15.5%)	238 (34%)

We subsequently integrated the WHO grade, IDH status, 1p/19q co-deletion, patient age, and SDF4 expression levels to create a nomogram designed to forecast survival outcomes (Fig. 3U). Notably, SDF4 expression has demonstrated potential in enhancing the accuracy of survival probability estimations at the 1-, 3-, and 5-year intervals. Additionally, a calibration chart was utilized to evaluate the precision of the predictions produced by the model (Figs. 3V-X). The time-dependent ROC curves indicated AUC values exceeding 0.75 for the 1, 3, and 5-year durations, underscoring the model’s strong performance (Fig. 3Y). Univariate Cox regression analysis identified WHO grade, IDH status, 1p/19q co-deletion, age, and SDF4 expression as independent prognostic factors for glioma patients. Moreover, univariate Cox regression analysis reaffirmed WHO grade, IDH status, 1p/19q co-deletion, age, and SDF4 expression as independent prognostic indicators (Fig. 3Z).

Verification of the SDF4 by CGGA dataset

We employed the CGGA dataset for the purpose of external validation. Furthermore, we conducted an analysis of the expression profiles of SDF4 alongside risk scores, survival durations, and the distribution of survival statuses within the CGGA dataset. The findings indicated that the expression levels of SDF4 in the high-risk cohort were markedly elevated compared to those in the low-risk cohort, which was associated with unfavorable prognoses (Fig. 4A). The KM curve analysis illustrated a significant correlation between heightened SDF4 expression levels and adverse clinical outcomes (Fig. 4B). Moreover, the data demonstrated that increased SDF4 expression was linked to diminished survival rates among patients classified as 1p/19q non-codel, IDH wild-type, and G4, specifically in those diagnosed with glioma (Figs. 4C-F). Four SNPs were retained as instrumental variables. The five MR methods showed broadly consistent effect directions, suggesting a potential association between genetically predicted SDF4 expression and glioma risk (Fig. 4G). The funnel and SNP-level forest plots showed no marked asymmetry or substantial inconsistency among the instrumental variables (Figs. 4H and I).

Fig. 4.

The relationship between the expression levels of SDF4 and various clinical characteristics. A–H Violin plots comparing SDF4 expression across WHO grade, IDH status, 1p/19q codeletion status, histological subtype, age, and survival-related events. Higher SDF4 expression was generally associated with more aggressive clinicopathological features. I–K ROC curves evaluating the ability of SDF4 to distinguish G4 from G2/G3 tumors, IDH-mutant from IDH-wild-type tumors, and 1p/19q-codeleted from non-codeleted tumors; AUCs and 95% CIs are shown. L–T Subgroup Kaplan–Meier analyses comparing overall survival between SDF4-high and SDF4-low groups across WHO grade, IDH status, 1p/19q codeletion, and age strata. U Nomogram integrating WHO grade, IDH status, 1p/19q codeletion, age, and SDF4 expression to predict 1-, 3-, and 5-year overall survival. V–X Calibration plots for the nomogram. Y Time-dependent ROC curves at 1, 3, and 5 years. Z Univariate and multivariable Cox regression analyses showing HRs and 95% CIs. Group comparisons were performed using the statistical tests specified in Methods. Survival curves were compared using the log-rank test, and Cox regression was used to estimate HRs and 95% CIs. *P < 0.05, **P < 0.01, and ***P < 0.001. DSS, disease-specific survival; PFI, progression-free interval; ROC, receiver operating characteristic; AUC, area under the curve

ScRNA-seq profiling and clustering

tSNE analysis classified the scRNA-seq data into 14 distinct cell clusters (Fig. 5A). Cell-type annotation showed an increased proportion of astrocytes in the SDF4-high group (Fig. 5B). Cluster-specific genes were visualized using heatmaps and dot plots (Figs. 5C and D), while FeaturePlot illustrated the expression of representative markers, including FERMT1, SHD, IFITM10, VIPR2, MAG, PLP1, ID4, SPARCL1, GFAP, C1orf61, CCL3, CCL4, C1QB, and TYROBP (Fig. 5E).

Fig. 5.

Validation of the SDF4 prognostic model utilizing the CGGA dataset. A Distribution of risk score, survival status, and SDF4 expression in the CGGA validation cohort, with patients stratified into low- and high-risk groups. B Kaplan–Meier analysis showing shorter overall survival in the SDF4-high group. C–E Associations of SDF4 expression with 1p/19q codeletion status, IDH status, and WHO grade. F Distribution of major clinicopathological characteristics in the SDF4-high and SDF4-low groups. G–I Mendelian randomization analyses, including scatter, funnel, and forest plots based on IVW, MR-Egger, weighted-median, simple-mode, and weighted-mode methods. Because of the limited number of instrumental variables, the MR findings should be considered exploratory. Group comparisons and survival analyses were performed using the statistical tests described in Methods. MR estimates are presented with 95% CIs. CGGA, Chinese Glioma Genome Atlas; MR, Mendelian randomization; SNP, single-nucleotide polymorphism; IVW, inverse-variance weighted

Samples were stratified into SDF4-high and SDF4-low groups using the median SDF4 expression level across the analyzed samples as the cutoff. Compared with the SDF4-high group, the SDF4-low group showed reduced proportions of astrocytes and microglia and an increased proportion of oligodendrocytes (Figs. 6A–E). Malignant and non-malignant scores derived from TCGA glioma-versus-normal differentially expressed genes were used to distinguish malignant from non-malignant cells (Figs. 6F and G). Glial cells were subsequently extracted and re-clustered into four subpopulations (Fig. 6H). hdWGCNA was then performed, with a soft-thresholding power of 3 selected at a scale-free topology fit index of 0.8 (Fig. 6I). Seven co-expression modules were identified and characterized by eigengene expression, hierarchical clustering, and gene–trait relationships (Figs. 6J–L). Notably, SDF4 was located in the turquoise module and was predominantly expressed in astrocytes.

Fig. 6.

Single-cell transcriptome profiles of glioma samples. A t-SNE maps of the GSE138794 scRNA-seq dataset showing 14 clusters in the SDF4-high and SDF4-low groups. B Cell-type annotation identified 11 major populations, including astrocytes, chondrocytes, ciliated cells, epithelial cells, fibroblasts, mature oligodendrocytes, microglia, neural progenitor cells, oligodendrocytes, progenitor cells, and T cells. C, D Heatmap and dot plot showing representative marker-gene expression across annotated cell types. E Feature plots of FERMT1, SHD, IFITM10, VIPR2, MAG, PLP1, ID4, SPARCL1, GFAP, C1orf61, CCL3, CCL4, C1QB, and TYROBP, supporting the assigned cell identities. Cells were quality controlled, normalized, integrated, and clustered using Seurat, and cell identities were assigned based on canonical markers and CellMarker 2.0. t-SNE, t-distributed stochastic neighbor embedding; scRNA-seq, single-cell RNA sequencing

Cellular communication analysis

Four glial cell types were re-clustered and visualized using t-SNE (Fig. 7A). Cell–cell communication analysis of the MIF pathway identified microglia as major signal senders, whereas astrocytes and oligodendrocytes primarily acted as receivers (Figs. 7B and C). Compared with the SDF4-high group, the SDF4-low group showed reduced network connectivity, particularly between astrocytes and microglia and between astrocytes and mature oligodendrocytes (Figs. 7D and E). Conversely, astrocytes in the SDF4-high group displayed more numerous and stronger interactions with other glial cells (Figs. 7F and G), suggesting that elevated SDF4 expression is associated with enhanced glial intercellular communication.

Fig. 7.

The hdWGCNA suggests that SDF4 is predominantly expressed in astrocytes. A, B Absolute and relative proportions of annotated cell types in the SDF4-high and SDF4-low groups. C–E Summary of cell-type composition and cell numbers across the scRNA-seq dataset. F, G Malignant and nonmalignant cell classification and their relative proportions in the two SDF4 groups. H t-SNE map of astrocytes, mature oligodendrocytes, microglia, and oligodendrocytes retained for hdWGCNA. I Soft-threshold diagnostics; a power of 3 was selected at a scale-free topology fit of approximately 0.8. J–M Identification, clustering, and cellular distribution of seven hdWGCNA modules. The turquoise module was enriched in astrocytes and contained SDF4, supporting an astrocyte-associated transcriptional context for SDF4. Cell-composition comparisons were performed at the biological-sample level using the test specified in Methods. hdWGCNA findings represent descriptive network associations and do not establish lineage-specific causality. hdWGCNA, high-dimensional weighted gene co-expression network analysis; kME, module eigengene connectivity

Spatial transcriptomics reveals that SDF4 is primarily expressed in astrocytes

Spatial transcriptomic data were obtained from the publicly available GEO dataset GSE253080 and analyzed using the 10× Visium platform. Distinct spatial clusters were identified based on gene-expression profiles (Figs. 8A and B). Overall transcript abundance and the number of detected genes across spatial locations were visualized in Figs. 8C and D. Spatial mapping further showed the distribution of SDF4 and major glial cell types (Figs. 8E and F), indicating that SDF4 expression was predominantly associated with astrocyte-enriched regions.

Fig. 8.

Cell communication analysis. A t-SNE map of astrocytes, mature oligodendrocytes, microglia, and oligodendrocytes used for cell–cell communication analysis. B, C Predicted MIF signaling network and communication-probability heatmap among the four glial populations. Arrows indicate signaling direction, and edge width reflects predicted communication strength. D, E Circle plots comparing the overall number and strength of inferred interactions between the SDF4-high and SDF4-low groups. F, G Cell-type-specific MIF signaling networks in the two groups. The SDF4-high group showed denser and stronger predicted communication, particularly involving astrocytes. These findings represent computational predictions and were not directly validated by cell-type-specific or in vivo experiments. Cell–cell communication was inferred using CellChat based on ligand–receptor expression. Edge widths represent model-derived communication probabilities. MIF, macrophage migration inhibitory factor

Protein expression of SDF4 in human tissues

To evaluate SDF4 protein expression in human malignancies, immunohistochemical staining was performed using tumor tissue specimens collected from the Department of Pathology, the Second Affiliated Hospital, Zhejiang University School of Medicine. Relatively high SDF4 protein expression was observed in several tumor types, including laryngeal cancer, gastric cancer, liver cancer, breast cancer, prostate cancer, colon adenocarcinoma, bladder urothelial carcinoma, and esophageal cancer (Fig. 9). This finding suggests that SDF4 plays a crucial role in the progression of several malignancies. Therefore, further research is necessary to elucidate its fundamental mechanisms and evaluate its potential as a target for therapeutic interventions. GFAP is recognized as a definitive marker for astrocytes, whereas oligodendrocyte lineage is specifically identified through Olig2. The mutation status of IDH1, particularly the R132H variant, is indicated by the IDH marker. The tumor suppressor protein p53 plays a critical role in indicating the mutation status of its corresponding gene. Ki-67 serves as a marker for cellular proliferation, reflecting the extent of proliferative activity. In distinguishing astrocytoma from oligodendroglioma, the characteristic expression profiles are as follows: astrocytoma demonstrates pronounced GFAP expression, significant positivity for p53, and diminished Olig2 expression; in contrast, oligodendroglioma is characterized by low GFAP expression, reduced p53 levels, and heightened Olig2 expression. When differentiating glioblastoma from oligodendroglioma, glioblastoma is identified by robust GFAP expression, negative IDH1 status, and a Ki-67 index surpassing 10%, while oligodendroglioma is characterized by weak GFAP expression, positive IDH1 status, and a Ki-67 index below 5% [1]. In our examination of glioma, we found that SDF4 expression levels were markedly elevated in glioblastoma compared to oligodendroglioma and astrocytoma (Fig. 10A). Furthermore, we explored the relationship between SDF4 expression and various clinical parameters in glioma patients. Our analysis indicated that elevated SDF4 levels were associated with higher WHO grades, the presence of IDH1 wild type, TERT mutations, and an unmethylated status of MGMT (Figs. 10B-E). In conclusion, the increased expression of SDF4 is not only associated with the initiation and progression of various cancers but also shows promise as a vital biomarker.

Fig. 9.

Spatial transcriptomic atlas of glial cell types in glioma tissue. A, B Spatial distribution of the 12 unsupervised clusters identified in the GSE253080 glioma spatial-transcriptomics dataset. C, D Spatial distributions of total transcript counts and detected gene numbers per spot. E Spatial expression pattern of SDF4. F Spatial maps of astrocytes, mature oligodendrocytes, microglia, and oligodendrocytes. The overlap between SDF4-rich and astrocyte-enriched regions supports an astrocyte-associated spatial localization of SDF4, although protein-level spatial validation is required. Spatial data were normalized and analyzed using Seurat. Color intensity represents transcript abundance, detected gene number, gene expression, or cell-type score. Because individual spatial spots may contain multiple cells, these findings should be interpreted as transcriptomic inferences

Fig. 10.

The levels of SDF4 protein expression in different human cancerous tissues

A significant decrease in SDF4 expression substantially hinders the proliferative ability of glioma cells in vitro

In order to examine the influence of SDF4 on the aggressive growth of glioma cells, we initially developed siRNAs specifically targeting SDF4 (siR-SDF4#1, #2, #3) and subsequently assessed their effectiveness in reducing SDF4 expression in H4 and SW1783 cell lines (Figs. 11A, B). Our results led us to select two cell lines with optimal knockdown efficiency (siR-SDF4#2; siR-SDF4#3) for further investigation into the role of SDF4 in glioma cellular progression. To clarify the function of SDF4 in glioma advancement, we began by analyzing the impact of SDF4 knockdown on the proliferative ability of glioma cells through colony formation assays. Compared to the control group, the experimental group with SDF4 knockdown exhibited a significant decrease in the rate of cell growth, suggesting a reduced capacity for proliferation (Figs. 11C, D). To further validate the influence of SDF4 on glioma cell proliferation, we employed the EdU assay to evaluate the effect of SDF4 knockdown on DNA replication activity. The findings indicated that SDF4 knockdown substantially limited the DNA replication proficiency of glioma cells (Figs. 11E, F). In summary, these experimental results indicate that the silencing of SDF4 hinders the proliferative potential of glioma cells.

Fig. 11.

Immunohistochemical analysis of SDF4 expression in glioma and its correlation with clinical parameters. A Representative IHC staining for SDF4, GFAP, Olig2, IDH1 R132H, p53, and Ki-67 in glioblastoma, oligodendroglioma, and astrocytoma specimens. Brown DAB staining indicates antigen positivity. SDF4 staining was visually strongest in the representative glioblastoma specimen. B–E Comparison of SDF4 IHC scores according to WHO grade, IDH1 status, TERT status, and MGMT promoter methylation status. Each point represents one patient specimen. WHO-grade comparisons were performed using one-way ANOVA with Tukey’s post hoc test, and two-group comparisons used the test specified in Methods. *P < 0.05 and **P < 0.01. GFAP, glial fibrillary acidic protein; IDH1, isocitrate dehydrogenase 1; TERT, telomerase reverse transcriptase; MGMT, O6-methylguanine-DNA methyltransferase; IHC, immunohistochemistry

Suppressing SDF4 enhances apoptosis of glioma cells and impedes their progression through the cell cycle

To investigate whether the diminished malignant proliferation capacity of glioma cells due to SDF4 knockdown is linked to cell cycle arrest or apoptosis, we utilized PI fluorescence staining alongside Annexin V-FITC, analyzed through flow cytometry, to evaluate the effects of SDF4 knockdown on both the cell cycle and apoptosis in glioma cells. Our results demonstrated that, compared to the control group, the apoptosis rate in H4 cells post-SDF4 knockdown increased from 3.56% to 17.13%, while in SW1783 cells, the apoptosis rate escalated from 5.32% to 18.29% (Figs. 12A, B). This consistent pattern indicates that the rates of apoptosis in the SDF4 knockdown groups are markedly higher than those seen in the control group. Additionally, we noted that SDF4 knockdown leads to cell cycle arrest at the G0/G1 phase in glioma cells (Figs. 12C, D). Collectively, these findings underscore the critical role of SDF4 in promoting cell cycle arrest at the G0/G1 phase and enhancing apoptosis.

Fig. 12.

The influence of SDF4 on the growth of glioma cells. A, B qRT-PCR analysis of SDF4 expression in H4 and SW1783 cells transfected with control siRNA or three SDF4-targeting siRNAs. Expression was normalized to β-actin using the 2^−ΔΔCt method. siR-SDF4#2 achieved the strongest knockdown and was selected for subsequent experiments. C, D Representative colony-formation assays and quantification showing reduced clonogenic growth after SDF4 knockdown. E, F Representative EdU images and quantification showing decreased DNA synthesis in both cell lines. Hoechst-stained nuclei are blue, and EdU-positive nuclei are red. Scale bar = 50 μm. Data are presented as mean ± SD from independent biological experiments. Multi-group comparisons were performed using one-way ANOVA with Tukey’s post hoc test, and two-group comparisons used the test specified in Methods. *P < 0.05; ns, not significant. EdU, 5-ethynyl-2′-deoxyuridine

Fig. 13.

SDF4 influences the cell cycle progression and apoptosis in glioma cells. A, B Representative Annexin V–FITC/PI flow-cytometry plots and quantification showing increased apoptosis after SDF4 knockdown in H4 and SW1783 cells. C, D Cell-cycle analysis showing an increased G0/G1 fraction and reduced S-phase fraction, indicating G0/G1 arrest. E Immunoblotting showed increased Bax and cleaved caspase-3 and decreased Bcl-2 after SDF4 knockdown. F SDF4 knockdown reduced Cyclin D1, CDK4, and CDK6 and increased p21 and p27. G KEGG enrichment analysis identified PI3K-AKT signaling as a candidate pathway for validation. H SDF4 knockdown reduced p-PI3K and p-AKT levels, whereas total PI3K and AKT were comparatively unchanged, supporting an association between SDF4 and PI3K-AKT phosphorylation rather than a direct activation mechanism. Flow-cytometry data are presented as mean ± SD from independent experiments and were analyzed using the two-group test specified in Methods. *P < 0.05. Immunoblots are representative of independent experiments. FITC, fluorescein isothiocyanate; PI, propidium iodide; KEGG, Kyoto Encyclopedia of Genes and Genomes; p-PI3K, phosphorylated PI3K; p-AKT, phosphorylated AKT

SDF4 modulates the expression of genes associated with the cell cycle and apoptosis in glioma cells and is associated with alterations in the PI3K-Akt signaling pathway to facilitate its own biological functions

To clarify the function of SDF4 in the regulation of apoptosis and the progression from the G0/G1 phase of the cell cycle, we performed protein immunoblotting assays to evaluate the expression of potential regulatory proteins implicated in these mechanisms. Significantly, after the silencing of SDF4, there was a pronounced increase in the expression levels of the pro-apoptotic proteins Bax and Cleaved caspase-3 in brain glioma cells compared to the control group. In contrast, the levels of the anti-apoptotic protein Bcl2 were significantly reduced (Fig. 12E). Additionally, the expression of two key regulatory proteins critical for the transition from the G0/G1 phase to the S phase, specifically CDK4, CDK6, and Cyclin D1, demonstrated a notable decrease, while the expression of the CIP/KIP family of cyclin-dependent kinase inhibitors, P21 and P27, increased markedly (Fig. 12F). The results obtained from the Western blot analysis were in alignment with those acquired from flow cytometry.

In order to elucidate the molecular mechanisms by which SDF4 modulates cell cycle regulation and apoptosis-related proteins in glioma cells, we performed a comparative analysis of expression profiles between SDF4 knockdown cells and control vector cells, employing high-throughput transcriptome sequencing technology. Furthermore, we examined the biological implications of these proteins through KEGG pathway enrichment analysis (Fig. 12G). Remarkably, the PI3K-Akt signaling pathway exhibited significant enrichment scores and statistical significance, suggesting its potential involvement in the biological processes linked to SDF4. Following this, we assessed the protein levels of essential components within the PI3K-Akt signaling pathway using Western blot analysis. Our findings indicated that, relative to the control group, the knockdown of SDF4 led to a substantial reduction in the protein expression levels of P-PI3K and p-AKT, while the levels of total PI3K and total AKT showed minimal variation (Fig. 12H). These results imply that SDF4 is crucial for the promotion of biological functions through the activation of the PI3K-Akt signaling pathway, and its knockdown significantly hampers the activation of this pathway.

Discussion

Glioma is the most common and one of the most aggressive primary malignant tumors of the central nervous system, characterized by diffuse infiltrative growth, a high recurrence rate, and poor long-term prognosis [1, 2]. Despite the widespread implementation of standard treatment strategies centered on maximal safe surgical resection followed by radiotherapy, chemotherapy, and targeted therapy, the 5-year overall survival rate of patients with glioblastoma remains below 10%, imposing a substantial burden on clinical management and society [3, 4, 7]. Conventional clinicopathological indicators, including WHO grade and IDH status, provide only limited prognostic stratification, and the molecular mechanisms underlying glioma progression remain incompletely understood. Therefore, the identification of reliable prognostic biomarkers and therapeutically actionable targets is of considerable clinical importance. By integrating transcriptomic analyses with in vitro functional validation, the present study identified SDF4 as a novel prognostic biomarker enriched in astrocytic cells. SDF4 was shown to regulate glioma cell proliferation, cell-cycle progression, and apoptosis, potentially through modulation of the PI3K-AKT signaling pathway.

Increasing evidence indicates that stromal cell-derived secreted proteins are critical regulators of tumor microenvironment remodeling, malignant progression, and therapeutic resistance in glioma [8, 9, 14]. Several secreted factors, including vascular endothelial growth factor A (VEGFA), secreted protein acidic and rich in cysteine (SPARC), and tenascin-C, have been reported to promote glioma proliferation, invasion, and angiogenesis by regulating tumor–stromal interactions [36, 37][. SDF4, a member of the CREC family of calcium-binding proteins, has been implicated in tumor growth, stromal remodeling, and chemoresistance in several solid tumors, including colorectal and nasopharyngeal carcinomas [17–19]. However, its expression pattern, cellular origin, and biological functions in glioma have not been systematically characterized. In contrast to VEGFA, a classical angiogenic factor that has been extensively investigated, the integrative analyses performed in this study revealed a distinctive enrichment of SDF4 in astrocytic cells within the glioma microenvironment, suggesting that SDF4 may exert stromal regulatory effects through a non-endothelial cellular source during glioma progression.

Consistent findings from two independent glioma cohorts, TCGA and CGGA, demonstrated that elevated SDF4 expression was significantly associated with higher WHO grade, IDH-wild-type status, 1p/19q non-codeletion, and shorter overall survival. Multivariable Cox regression analysis further showed that SDF4 remained an independent prognostic factor after adjustment for established clinicopathological variables. A nomogram integrating SDF4 expression with conventional clinical parameters showed favorable calibration for predicting 1-, 3-, and 5-year survival. Notably, the addition of SDF4 improved both the area under the time-dependent receiver operating characteristic curve and the concordance index, indicating that SDF4 provides prognostic information beyond that conveyed by traditional clinicopathological characteristics. These results support the potential utility of SDF4 as a complementary biomarker for improving risk stratification in patients with glioma.

Single-cell RNA sequencing, spatial transcriptomics, and high-dimensional weighted gene co-expression network analysis consistently indicated that SDF4 was predominantly enriched in astrocytic cell subsets within the glioma microenvironment and was associated with enhanced MIF-mediated intercellular communication among glial cells. These findings provide new insights into the cellular origin of SDF4 and its potential role in microenvironmental regulation. Nevertheless, the inferred cellular localization and communication patterns were derived from computational analyses and therefore require further experimental confirmation. In vitro functional experiments using glioma cell lines showed that SDF4 knockdown markedly suppressed cell proliferation, induced G0/G1 cell-cycle arrest, and promoted apoptosis, thereby providing cellular-level evidence supporting a tumor-promoting role of SDF4 in glioma.

Pathway enrichment analyses identified several signaling processes associated with SDF4, including focal adhesion, extracellular matrix remodeling, cell-cycle regulation, and the PI3K-AKT signaling pathway. The PI3K-AKT pathway was prioritized for experimental validation for three principal reasons. First, the PI3K-AKT cascade is a well-established central regulator of glioma cell proliferation, cell-cycle progression, and resistance to apoptosis, and is therefore highly consistent with the phenotypic changes observed following SDF4 knockdown [38, 39]. Second, aberrant activation of PI3K-AKT signaling is a frequent molecular event in glioma and is closely associated with poor prognosis and therapeutic resistance, making this pathway a major focus of targeted therapeutic research in glioma [40, 41]. Third, among the tumor-related pathways identified in the present study, PI3K-AKT signaling showed the highest enrichment significance, providing an objective rationale for its prioritization. Western blot analysis demonstrated that SDF4 knockdown significantly reduced the phosphorylation levels of PI3K and AKT, supporting a functional association between SDF4 expression and PI3K-AKT pathway activity. Other enriched pathways, such as focal adhesion and extracellular matrix remodeling, are also involved in glioma invasion and tumor–stroma interactions [42]. However, these pathways were not functionally investigated in the present study and therefore represent important directions for future mechanistic research.

From a clinical perspective, the present findings support SDF4 as a potential prognostic biomarker for risk stratification in glioma. Beyond its biomarker value, SDF4 may also represent a potential therapeutic target. Although direct targeting of secreted calcium-binding proteins remains technically challenging, emerging therapeutic strategies, including neutralizing antibodies and small-molecule inhibitors, may provide feasible approaches for further translational investigation [43, 44]. Several issues nevertheless warrant additional study. First, the astrocyte-specific functions of SDF4 and its paracrine effects on the tumor microenvironment should be validated using co-culture systems and in vivo models. Second, rescue experiments and pathway-intervention studies are required to establish a causal regulatory relationship between SDF4 and PI3K-AKT signaling. Third, the prognostic value of SDF4 should be further evaluated in larger prospective clinical cohorts.

In conclusion, this study identifies SDF4 as a novel prognostic biomarker enriched in astrocytic cells in glioma. Elevated SDF4 expression was associated with aggressive clinicopathological characteristics and unfavorable patient outcomes. Functionally, SDF4 knockdown inhibited glioma cell proliferation, induced G0/G1 cell-cycle arrest, and promoted apoptosis, and these effects were accompanied by reduced PI3K-AKT pathway activity. Collectively, these findings expand the current understanding of the molecular mechanisms underlying glioma progression and suggest that the SDF4-PI3K-AKT regulatory axis may serve as a potential target for prognostic stratification and therapeutic intervention in glioma.

Authors' contributions

CW, BS, and SC performed the experiments, analyzed the data, and wrote the paper. WK, ZY, FZ, YH, HW, LW, JL, and CZ helped to perform the experiments and analyze the data. JZ, ZJ, NW, and ZZ designed the research, analyzed the data, and wrote and revised the paper. All authors confirm the authenticity of all the raw data. All authors have read and approved the final manuscript.

Funding

This study was funded by National Natural Science Foundation of China Youth Fund Project (No. 82504271), Medical and Health Science Program of Zhejiang Province (No2025HY0381, No. 2024KY1213, No. 2024KY1225), Zhejiang Provincial Natural Science Foundation (No. LQ24H160009), Zhejiang Provincial Traditional Chinese Medicine Science and Technology Project (No. 2025ZR123), and Research Projects of Zhejiang Chinese Medical University (No. 2023 JKJNTZ10).

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

Ethical approval was received for the study by the Clinical Research Ethics Committee of The Second Affiliated Hospital, Zhejiang University School of Medicine (approval no. 2025–2030; Hangzhou, China). Informed consent was waived by the Clinical Research Ethics Committee of The Second Affiliated Hospital, Zhejiang University School of Medicine. All methods were carried out in accordance with relevant guidelines and regulations.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.