GEO (http://www.ncbi.nlm.nih.gov/geo) is a publicly accessible database hosting numerous high-throughput sequencing and microarray datasets contributed by global research institutions (38, 39). We utilized keywords such as subarachnoid hemorrhage and vascular dementia to search for gene expression datasets related to these conditions. The inclusion criteria required two independent expression profiles to originate from the same sequencing platform with the largest possible sample sizes. Moreover, the specimens included had to be derived from human sources. Ultimately, we downloaded three microarray datasets in FPKM format from the GEO database: GSE122063, GSE13353, and GSE161870. The GSE122063 dataset, sourced from frontotemporal lobe tissue, comprised gene expression profiles of 18 patients with vascular dementia (VaD) and 22 normal controls (40), which allows for a comparison of gene expression patterns that may be associated with the pathophysiology of VaD. GSE13353 contained 11 samples from ruptured aneurysm walls and 8 samples from unruptured aneurysm walls (41), which explores the molecular mechanisms underlying aneurysm rupture. The GSE161870 dataset assessed miRNA, including 2 samples from ruptured aneurysm walls and 2 control samples from intercostal arteries, allowing us to compare miRNAs’ expression in a vascular context.

The R package limma (42) was employed to calculate DEGs between various groups (42). An intersection analysis was conducted on genes upregulated and downregulated in both SAH and dementia, identifying common DEGs between these conditions. Probe sets lacking corresponding gene symbols were excluded, and for probe sets with the same gene symbol, their values were averaged. For mRNA, genes with an adjusted P-value < 0.05 were considered DEGs. For miRNA, only those with an adjusted P-value < 0.05 and |log2 fold change (FC)| ≥ 0.5 were identified as differentially expressed miRNAs.

The Gene Ontology (GO) database, developed by the Gene Ontology Consortium, offers straightforward annotations of gene products concerning their functions, biological pathways, and cellular locations (43, 44). The Kyoto Encyclopedia of Genes and Genomes (KEGG) Pathway database specializes in storing gene pathway information across various species. The “clusterProfiler” package was utilized for enrichment analysis, while the “org.Hs.eg.db” package facilitated the conversion between gene symbols and IDs. An adjusted P-value of < 0.05 was considered significant.

A random forest model was applied to rank the variables based on their importance within the model (45, 46). Univariate logistic regression was then utilized to analyze risk factors associated with dementia occurrence. LASSO regression analysis was conducted to screen disease-related common genes further, employing 10-fold cross-validation to prevent overfitting. Principal Component Analysis (PCA) was used to visualize the performance of the included variables in classifying outcomes.

The miRWalk platform facilitates the exploration of relationships between proteins of interest, including direct binding and co-regulated pathways, enabling the prediction of miRNAs that regulate core genes (47, 48). Cytoscape (http://www.cytoscape.org) was used to visualize this network, allowing observation of the interactions between core genes and miRNAs within the network.

To assess the predictive performance of core genes related to the disease outcomes, time-dependent Receiver Operating Characteristic (ROC) curves were plotted (49, 50). The area under the ROC curve (AUC) was calculated using the “pROC” package in R, and the visualization of the ROC curves was done with the ggplot2 package. The ROC curve evaluated the molecular classification ability concerning outcomes, where the AUC ranges from 0.5 to 1. An AUC close to 1 indicates a better diagnostic performance. An AUC between 0.5 and 0.7 suggests low accuracy, 0.7 to 0.9 indicates moderate accuracy, and above 0.9 signifies high accuracy.

Gene Set Enrichment Analysis (GSEA) was performed using the clusterProfiler package in R (51, 52). The species selected was “Homo sapiens,” and the reference gene set was chosen from the MSigDB Collections (http://www.gsea-msigdb.org/gsea/msigdb/collections.jsp), specifically “c2.cp.v7.2.symbols.gmt [Curated].” Pathways were considered significantly enriched if they met the criteria of a False Discovery Rate (FDR) < 0.25 and an adjusted p-value < 0.05.

The U251 cell line was obtained from the American Type Culture Collection (ATCC, Rockville, MD, USA). These cells were maintained in DMEM (Gibco, Grand Island, NY, USA) supplemented with 10% fetal bovine serum (FBS, BI, Kibbutz Beit Haemek, Israel) at a temperature of 37°C in an atmosphere of 5% CO₂. The DOK3 gene was introduced into U251 cells through lentiviral vectors for gain-of-function experiments. The targeting sequences utilized were sourced from Open Biosystems.

Following the manufacturer’s instructions, total RNA was extracted using the Cell Total RNA Isolation Kit (FOREGENE, Chengdu, China). RNA was reverse transcribed into cDNA using the StarScript II First-strand cDNA Synthesis Mix (GenStar, Beijing, China). Quantitative real-time PCR (qRT-PCR) was performed using the CFX96 Touch™ Real-Time PCR Detection System (BIO-RAD, USA). The results were analyzed using the 2^−ΔΔCt method, with GAPDH mRNA as the internal control.

Cell proliferation was assessed using the CCK-8 assay. Cells were seeded in a 96-well plate at a density of 1×10⁶ cells/ml and treated with various concentrations of CVB-D (0, 15, 30, 60, 120, 240 µmol/l) for 24, 48, and 72 hours. The condition resulting in a significant decrease in cell viability was 240 µmol/l for 72 hours, which was the highest concentration and longest duration used in the experiment. Each group consisted of six replicates and was performed in three independent experiments. After treatment, wells were washed with PBS (0.01 M), and 10 µl of CCK-8 solution was added to 90 µl of serum-free medium. The cells were incubated in the dark at 37°C for 1-2 hours. Absorbance was measured at 450 nm using a microplate reader (Tecan Group, Ltd.). For the clonogenic assay, T98G and Hs683 cells were treated with different concentrations of CVB-D (0, 5, 10, 20, 40, 80 µmol/l) for 24 hours. Cells were seeded at 500 cells per well in a 6-well plate and cultured for 10 days post-treatment. Subsequently, the cells were fixed with 4% paraformaldehyde (room temperature for 10 minutes) and stained with 0.05% crystal violet solution. Statistical analysis was conducted using GraphPad Prism 8 (GraphPad Software, Inc.).

Cells from the four groups were collected and centrifuged at 1,000 rpm for 5 minutes, then fixed in 70% pre-cold ethanol at -20°C for 24 hours. The samples were rinsed with phosphate-buffered saline (PBS) and incubated at 37°C with 1 mg/ml RNase (Sigma) for 30 minutes. After suspension in PBS, they were fixed with 70% ethanol for a minimum of 6 hours, followed by another PBS wash. Subsequently, the samples were resuspended in 1 ml of Propidium Iodide (PI) staining solution (50 µg/ml PI, Sigma), incubated in the dark at 4°C for 30 minutes, and analyzed by fluorescence-activated cell sorting (FACS) using a flow cytometer (FACScan, Becton Dickinson).

GBM cells were fixed with 4% paraformaldehyde for 30 minutes at room temperature. The cells were then permeabilized with 0.3% Triton X-100 in PBS for 15 minutes and blocked with normal goat serum for 1 hour at room temperature. After blocking, the cells were incubated overnight at 4°C with rabbit anti-NMDAR2B antibody (Cell Signaling, 1:300). The following day, the cells were washed and incubated with goat anti-rabbit Alexa Fluor 555-conjugated secondary antibody (Thermo Scientific, 1:500) for 1 hour at room temperature. The cell nuclei were stained with Hoechst 33258 (MO, USA). Images were captured using a confocal laser scanning microscope (Zeiss LSM710, Germany).

The release of lactate dehydrogenase (LDH) from neurons was measured using a commercial LDH assay kit. LDH activity was determined using an LDH cytotoxicity detection kit (Nanjing, China), based on a coupled enzyme reaction that converts tetrazolium salt to formazan. Absorbance was measured at 450 nm, and results were expressed as a percentage of LDH release relative to the control group. All experiments were conducted in triplicate.

Indicators of oxidative stress, including superoxide dismutase (SOD), glutathione peroxidase (GSH-Px), reactive oxygen species (ROS), and malondialdehyde (MDA), were measured using commercial assay kits according to the manufacturer’s instructions. Samples were analyzed with a SpectraMax M2 spectrophotometer (Molecular Devices). SOD activity was expressed as U/mg protein (450 nm), GSH-Px activity as U/mg protein (412 nm), ROS levels as pmol/mg protein/min (excitation wavelength = 488 nm, emission wavelength = 520 nm), and MDA levels as nmol/mg protein (523 nm). Inflammatory cytokines, including TNF-α, IL-1β, and IL-6, were measured using ELISA kits after 24 hours of treatment, following the manufacturer’s guidelines. Data were analyzed using the SpectraMax M2 spectrophotometer (Molecular Devices).

Statistical analyses for this study were performed using R software (version 4.0.2, https://www.r-project.org/). Correlation analysis was conducted using Pearson’s method. Categorical data were analyzed with the chi-square test, while continuous data were assessed using either the t-test or the Wilcoxon test. Sankey diagrams were created using the ggalluvial package [version 0.12.3]. A P-value of less than 0.05 was deemed statistically significant for all analyses.

Figure 1 explores the shared molecular mechanisms between subarachnoid hemorrhage (SAH) and vascular dementia (VaD), focusing on the roles of DOK3 and PAPOLA in apoptotic pathways. Using datasets from GSE122063, GSE13353, and GSE161870, differentially expressed genes (DEGs) were identified through comparative analyses, including volcano plots and Venn diagrams. Functional enrichment analysis revealed their involvement in apoptosis and neurodegenerative processes. Core genes were prioritized through LASSO-logistic regression and machine learning, with heatmaps depicting their expression profiles. miRNA-mRNA interaction networks and circular plots illustrated regulatory mechanisms, while gene set enrichment analysis (GSEA) and pan-cancer analyses provided insights into broader pathways and immune infiltration. In this study, we utilized machine learning techniques to identify 15 common DEGs between SAH and dementia, discovering 494 DEGs in GSE122063, 1436 in GSE13353, and 115 miRNAs in GSE161870 ( Figures 2A–C ). Venn diagram analysis ( Figures 2D, E ) revealed 10 upregulated and 5 downregulated DEGs shared between the conditions, including C5AR1, DOK3, KHDRBS1, etc. Functional enrichment analysis ( Figure 2F ) revealed vital pathways such as mRNA processing regulation, neurotransmitter transport control, and neutrophil degranulation, indicating significant roles of these genes in nucleotide activity, transporter activity, and immune response modulation. The random forest model identified key DEGs crucial for classifying SAH and dementia, with METRNL, KHDRBS1, SLC6A1, TNFAIP8L1, and PAPOLA being the primary discriminators for SAH, and METRNL, NT5DC1, ZNF627, TUBG2, and TNFAIP8L1 crucial for dementia classification ( Figures 2G, H ). These genes are significantly associated with dementia ( Figure 2I ) and may have important implications for disease progression and pathophysiology.

We applied logistic regression and Lasso logistic regression analyses to identify dementia-associated genes. Univariate logistic regression analysis showed that 13 out of 15 differentially expressed genes, such as ZNF627, KHDRBS1, and SLC6A1 ( Figure 3A ), were significantly linked to dementia. Lasso logistic regression further refined this list to seven genes: DOK3, LONRF3, MILR1, PAPOLA, SLC6A1, STK11IP, and ZNF627 ( Figure 3B ). Principal Component Analysis (PCA) demonstrated a clear separation between dementia patients and controls, proving the classification effectiveness of these genes ( Figure 3C ). Additionally, we constructed a miRNA-mRNA interaction network for the seven core DEGs identified in SAH and dementia using the miRWALK website, highlighting DOK3 and PAPOLA as hub genes with the most connections ( Figure 3D ), showcasing their potential in predicting dementia and their molecular mechanisms.

The classification performance of seven core DEGs identified in SAH and dementia was evaluated, with ROC curve analysis demonstrating their strong discriminative capabilities, particularly PAPOLA in dementia with an AUC of up to 0.989, followed by DOK3, LONRF3, and MILR1, all exceeding 0.900 ( Figures 3G, H ). In SAH, these genes also showed good classification capabilities, with AUC values ranging from 0.864 to 0.966, highlighting PAPOLA, DOK3, and SLC6A1 as significant predictors ( Figures 4A, B ). We also explored the regulation of DOK3 and PAPOLA by specific miRNAs, finding DOK3 negatively correlated with SLC6A1 and ZNF627 but positively with LONRF3, MILR1, PAPOLA, and STK11IP; PAPOLA showed negative correlations with ZNF627 and SLC6A1 but positive with LONRF3, MILR1, STK11IP, and DOK3 ( Figures 4C, D ). Additionally, GSEA indicated that upregulation of DOK3 and PAPOLA affected crucial biological pathways potentially linked to processes like immune response modulation and apoptosis, with implications for conditions such as dementia ( Figures 4E, F ).

In our pan-cancer analysis, we examined the expression of DOK3 and PAPOLA across various cancer types. Figure 5A depicts the distribution of copy number variation (CNV) rates across 20 cancer types, with different colors representing each type, revealing significant CNV variations that may influence tumor progression. The differential expression analysis in Figure 5B highlighted the significant log2 fold change and -log10 FDR values for DOK3 and PAPOLA, emphasizing their potential as biomarkers. Further correlation analyses ( Figures 5C–E ) demonstrated how genetic and epigenetic factors impact gene expression. Additionally, the expression of DOK3 and PAPOLA showed significant positive or negative correlations with various immune cell infiltration levels ( Figures 5F–H ). These findings were reinforced by the EPIC method in Figure 5I , indicating that PAPOLA might play a dual role in influencing immune cell presence in tumors. The consistent findings across different methodologies for DOK3 and PAPOLA emphasize their importance in shaping the immune landscape of various cancers, suggesting that these genes could be critical targets for therapeutic interventions to modulate tumor immunity. This comprehensive analysis provides a solid foundation for future studies to elucidate the precise mechanisms through which DOK3 and PAPOLA influence immune infiltration in cancer.

In this study, we investigated the role of DOK3 and PAPOLA in tumor prognosis using U251MG glioblastoma cells. Kaplan-Meier survival analysis revealed that high DOK3 expression was significantly associated with poor prognosis, including worse overall survival (OS), disease-specific survival (DSS), and progression-free interval (PFI), compared to low expression ( Figure 6A ). Patients with low PAPOLA expression also demonstrated better survival rates, further suggesting the prognostic importance of these genes in glioblastoma. Meta-analysis of multiple datasets confirmed that DOK3 is a significant risk factor for glioblastoma survival (HR > 1), with moderate heterogeneity across studies ( Figure 6B ). PCR validation showed that DOK3 was successfully overexpressed in the DOK3 OE group ( Figure 6C ). Functionally, DOK3 overexpression led to increased cell proliferation, as indicated by a significant rise in colony numbers in the plate colony formation assay ( Figure 6D ), and enhanced cell migration in the Transwell migration assay ( Figure 6E ), indicating that DOK3 plays a crucial role in promoting tumor cell aggressiveness. Furthermore, GSEA and KEGG pathway analysis revealed that cell cycle-related pathways were enriched in the high DOK3 expression group, with metabolic pathways such as galactose metabolism activated and lysine degradation pathways inhibited, further suggesting that DOK3 impacts both cellular proliferation and metabolism ( Figure 6F, G ). These results collectively demonstrate the critical involvement of DOK3 and PAPOLA in glioblastoma prognosis and tumor progression, making them potential therapeutic targets for future interventions.

The pathway and immune microenvironment analysis revealed significant correlations between immune-related scores and gene expression levels within the cohort, as shown in Figure 7A . The Spearman correlation of Tumor Immune Profile (TIP) scores highlighted strong interactions between various immune processes, with thicker lines representing stronger correlations. Notably, in the DOK3 high-expression group, immune stimulation genes, immunosuppressive genes, chemokines, and human leukocyte antigen (HLA) expressions were generally elevated, as depicted in the heat maps ( Figure 7B ). These findings demonstrate that DOK3 overexpression is associated with an overall increase in immune activity. Additionally, the regulatory landscape of immunomodulators, including somatic copy number alterations (SCNA) and epigenetic factors, underscores the complex immune landscape influenced by DOK3 expression ( Figure 7C ). Further exploration of the relationship between genomic instability markers (e.g., mutation rates, homologous recombination defects) and immune responses revealed distinct patterns of immune engagement, with higher MeTIL scores observed in the high chemokine expression group. In contrast, groups with low expression of CYT, IFNγ, T cell-inflamed markers, and TLS showed lower MeTIL scores ( Figures 7D, E ). These findings underscore the role of DOK3 in modulating immune responses and its potential impact on tumor progression and immune infiltration. Integrating genomic status with immune profiles provides a comprehensive view of the tumor microenvironment and highlights potential therapeutic targets.

In this study, we evaluated the effects of LPS treatment and DOK3 overexpression (OE) on U251MG glioblastoma cells, focusing on inflammatory cytokine expression, oxidative stress markers, and tumor-related processes. As shown in Figure 8A , qPCR analysis revealed that LPS treatment significantly upregulated the expression of pro-inflammatory cytokines TNF-α, IL1β, and IL6 compared to the control group (p < 0.001). However, DOK3 OE significantly attenuated this increase (p < 0.001), indicating its potential anti-inflammatory effects. Further, Figure 8B demonstrates that superoxide dismutase (SOD) activity was markedly reduced following LPS treatment (p < 0.001), while DOK3 OE partially restored SOD activity to near-control levels (p < 0.001), highlighting its role in combating oxidative stress. The measurement of malondialdehyde (MDA) levels, a marker of lipid peroxidation, showed a significant increase in the LPS group (p < 0.001), as shown in Figure 8C , whereas DOK3 OE resulted in a dramatical reduction in MDA levels (p < 0.01). Similarly, Figure 8D indicates that LPS treatment significantly decreased glutathione peroxidase (GSH-Px) activity (p < 0.001), but DOK3 OE restored this activity (p < 0.001), suggesting improved antioxidant defense. The reactive oxygen species (ROS) level comparison in Figure 8E further confirmed that LPS treatment significantly elevated ROS levels (p < 0.001), while DOK3 OE mitigated this increase (p < 0.001), reinforcing its role in reducing oxidative damage. Additionally, Figure 8F illustrates the quantification of cytokines (TNF-α, IL1β, IL6) via ELISA, which revealed a significant elevation of these cytokines in the LPS group, but DOK3 OE substantially reduced their levels (p < 0.01), demonstrating an anti-inflammatory effect. Finally, Figure 8G shows Pearson’s correlation analysis, where DOK3 expression was positively correlated with various tumor-related processes such as apoptosis (R = 0.65, p = 2.2e-16), cell cycle (R = 0.54, p = 2.1e-06), and proliferation (R = 0.42, p = 1.6e-08), suggesting that DOK3 may play a crucial role in regulating these processes in glioblastoma cells. These results collectively indicate that DOK3 overexpression can counteract LPS-induced inflammatory and oxidative stress responses while also regulating critical tumor-related pathways.

In the present study, we explored the role of DOK3 in modulating inflammation and oxidative stress in BV2 cells. LPS treatment markedly reduced cell viability, as demonstrated by the CCK-8 assay, while DOK3 overexpression (OE) partially restored cell viability, indicating its protective effect against LPS-induced cytotoxicity ( Figure 9A , p<0.001). Furthermore, the LDH release assay showed increased LDH levels in LPS-treated cells, indicative of cell membrane damage, whereas DOK3 overexpression significantly reduced LDH release, suggesting a reduction in cellular injury ( Figure 9B , p<0.001). Flow cytometry analysis revealed that DOK3 overexpression mitigated LPS-induced apoptosis, demonstrating its anti-apoptotic effect ( Figure 9C , p<0.001). Immunofluorescence analysis further confirmed that DOK3 overexpression reduced the expression of key inflammatory markers, including IL-1β ( Figure 9D ), NLRP3 ( Figure 9E ), and TNF-α ( Figure 9F ), following LPS treatment. Additionally, flow cytometric analysis of ROS levels showed that LPS significantly increased ROS production, while DOK3 overexpression effectively decreased oxidative stress in BV2 cells ( Figure 9G ). Collectively, these findings indicate that DOK3 overexpression exerts significant anti-inflammatory and protective effects by reducing apoptosis, inflammation, and oxidative stress, thus highlighting its potential therapeutic applications in inflammation-related diseases and cancer research.