Identification and enrichment analysis of ORGs in IA

Differential expression analysis was conducted on the GSE75436 dataset using the Limma method, revealing significant differences between the IA and STA groups. Compared to those in the STA group, 1315 upregulated genes and 1127 with downregulated genes were identified in the IA group (Fig. 1A, B). The intersection of DEGs with ORGs revealed 63 DEORGs (Fig. 1C). A list of differential genes can be found in Additional file 3.

Fig. 1

Identification and enrichment analysis of ORGs. A Volcano plot of DEGs in the GSE75436 dataset, with filtering criteria of |log2FC|> 1 and p < 0.05. B Heatmap showing the top ten genes with upregulated or downregulated expression. C Identification of 63 DEGs related to OS. D, E GO analysis of DEGs. F Chordal diagram about KEGG analysis. G Hallmark pathway analysis

Functional analysis of DEORGs was performed using multiple gene sets. GO analysis was performed separately on the upregulated and downregulated DEORGs, with redundant terms removed using the REVIGO website. The results showed that both the upregulated and downregulated DEORGs were mainly enriched in the Biological Process (BP) terms of response to stress, response to oxidative stress, and response to chemical. However, they differed in the Cellular Component (CC) and Molecular Function (MF) categories (Fig. 1D, E). KEGG analysis revealed enrichment of these genes in pathways such as malaria, arachidonic acid metabolism, and toxoplasmosis (Fig. 1F). Hallmark pathway enrichment analysis revealed close associations between DEORGs and pathways such as apoptosis, IL-6/JAK/STAT3 signaling, and myogenesis (Fig. 1G). The enrichment results can be found in Additional file 4.

Immune infiltration and WGCNA

To explore the relationship between immunity and IA formation, we constructed a landscape of GSE75436 immune infiltration data (Fig. 2A). We analyzed the percentages of different immune cells in the CIBERSORT algorithm in both IA and Control tissues, and the results showed that M2 macrophages were well represented in both the IA and Control groups, and that there were more in the IA group. In addition, mast resting cells and T_Cells_gamma_delta were the most common in IA, and neutrophils and T_Cells_gamma_delta were the most common in control (Fig. 2B). Five types of immune cells showed significant differences in abundance between the IA and control samples (p < 0.05). These included naïve CD4 + T-cells, monocytes, M1 macrophages, and resting and activated mast cells (Fig. 2C). The xCell algorithm was used to calculate the xCell score through transcriptomic data to analyze the immune microenvironment. The immune score, stroma score, and microenvironment score were significantly different between IA tissues and normal tissues (Fig. 2D). The immunization score is provided in Additional file 5: Table S5.

Fig. 2

Immunoinfiltration analysis and WGCNA revealed the role of immune cells in IA. A Immunoinfiltration landscape in the GSE75436 dataset. B Percentage of immune cells in IA. C Abundance levels of various immune cells in the IA and control groups. D Immune scores, stromal scores, and microenvironment scores derived from the xCell algorithm. E Selection of the optimal soft-thresholding power (β). F Identification of 10 modules presented as a clustering tree. G Correlations between modules and immune cells. H The correlation between the light green module and Mast_cells_resting, and the correlation between the pink module and M2 macrophages. I Extraction of 538 hub genes from the light green and pink modules intersecting with DEORGs, resulting in 16 key genes

For WGCNA, The optimal soft threshold power (β = 22) was chosen according to the construction of a scale-free network (Fig. 2E). We identified 10 coexpression modules from 30 samples of 20,020 genes, which are shown in different colors (Fig. 2F). Subsequently, we analyzed the correlation between the module and immune-infiltrating cells by Pearson correlation analysis and found that the pink module was most strongly correlated with resting mast cells (Cor = 0.63, P = 2.2e − 4) and that the light green module was strongly correlated with M2 macrophages (Cor = 0.62, P = 2.3e − 4) (Fig. 2G). Significant correlations between GS (Gene Significance) and MM (Module Membership) are showcased within the pink and light green modules (Fig. 2H). As a result, these modules, associated with immune infiltrating cells, were deemed pivotal and earmarked for further scrutiny. The hubgenes for the two modules are found in Additional file 6. A sum of 538 hub genes, exhibiting GS values exceeding 0.20 and MM values surpassing 0.80, were culled from these modules for subsequent analysis. We intersected these genes with the DEORGs and identified 16 hub genes: CRYAB, KCNA5, MSRB2, TLR4, AIF1, BID, APOE, TREM2, HMOX1, RBPMS, TLR6, MAPK13, ALOX5, CD38, BTK and GPX1 (Fig. 2I).

Further screening of hub genes and identification of their functions

We analyzed the correlations among these 16 hub genes (Fig. 3A) and generally found a strong correlation among them. Subsequently, we explored the PPIs of these 16 DEORGs (Fig. 3B). MCC algorithm in the CytoHubba plugin to was employed to evaluate these interacting proteins, with the top six being TLR4, APOE, AIF1, HMOX1, ALOX5, and TRPM2 (Fig. 3C). Additionally, we used various algorithms, including MNC, degree, and EcCentricity, to assess hub genes. These scores can be found in Additional file 7.

Fig. 3

Further validation, external verification, and functional analysis of hub genes. A Correlations among the 16 hub genes. B Using was Cytoscape software to construct A PPI network. C Score of MCC algorithm in CytoHubba. D Determination of the soft-thresholding power (lambda = 0.15) and selection 4 genes through LASSO analysis. E The Venn diagram illustrates the intersection of LASSO with various PPI algorithms. F ROC curve of TLR4 and ALOX5 in the GSE75436 dataset and the external validation dataset. An AUC > 0.7 was considered to indicate good predictive performance. G In GSE75436, TLR4 and ALOX5 low expression groups were enriched in the myogenesis pathway

Next, LASSO analysis was performed on these 16 genes, with an ideal lambda value set at 0.1475, resulting in the identification of 4 genes, ALOX5, TREM2, TLR4, and RBPMS (Fig. 3D). LASSO results can be found in Additional file 8. The results of LASSO-Cox were consistent with the PPI algorithms, revealing 2 hub genes, namely, TLR4 and ALOX5 (Fig. 3E). These two genes may be key genes involved in the pathogenesis of IA. Next, we validated their predictive ability.

We performed internal and external dataset validation of the hub genes employing ROC curves to evaluate their diagnostic precision (Fig. 3F). In GSE75436, the AUC for TLR4 was 0.96, and that for ALOX5 was 0.98. GSE15629 is a dataset containing data on 14 patients with IA and 5 controls, with AUCs of 0.84 and 0.71 for TLR4 and ALOX5, respectively. We also unexpectedly found that TLR4 and ALOX5 exhibited good predictive performance in the GSE36791 dataset, with AUCs of 0.80 and 0.71, respectively. GSE36791 is a serum dataset containing data on 43 patients with ruptured IA and 18 controls, suggesting that these two hub genes may also have diagnostic effects on ruptured aneurysms. Sg-GSEA shows the impact of TLR4 and ALOX5 expression levels on IA modulation. Low TLR4 and ALOX5 expression was enriched in the myogenic pathway. (Fig. 3G). Sg-GSEA results can be attached in Additional file 9.

Exploring miRNA regulatory networks and potential small molecule drugs

We used online websites to predict the upstream TFs, interacting miRNAs, and potential targeted drugs for TLR4 and ALOX5. Among them, EGR1, SP1, HDAC2, and TP53 are TFs that can regulate ALOX5, while ZNF160, IRF3, and IRF8 are TFs that can regulate TLR4. There was no overlap in the TFs regulating each gene. For small-molecule drugs, resatorvid and eritoran tetrasodium are the most highly scored TLR4 inhibitors, while diethylcarbamazine and zileuton are the most strongly scored ALOX5 inhibitors. Methotrexate (MTX) is a common inhibitor of both genes and may have potential therapeutic effects. miRNAs such as hsa-miR-146a-5p and hsa-let-7d-5p coregulate ALOX5 and TLR4 (Additional file 1:Fig. S1A-D). TF, miRNA network and drug results can be found in Additional files 10 and  11.

scRNA-seq analysis

Two IA samples and 3 STA samples were subjected to single-cell sequencing. After screening based on the quality control criteria described in the methods, 23,342 cells were selected. After the samples were merged, 20 different cell clusters were identified by unsupervised Seurat clustering (Fig. 4A). Based on the expression levels of typical cell type-specific markers, dot plots were drawn, and the cells were manually annotated (Fig. 4B), including VSMCs (MYH11, ACTA2, MYL9, and TAGLN), T- and NK cells (NKTR, CD3E, TRAC, and TRBC2), Mos/Mφs (CSF1R and CD14), neutrophils (S100A9, CSF3R, and FCGR3B), DCs (CD74, IRF8, and HLA-DRA), endothelial cells (PECAM1, VWF, and FLT1), fibroblasts (DCN, PDGFRA), Schwann cells (MPZ, PLP1, and PMP22), and mast cells (MS4A2). The cell clusters were ultimately classified into 9 types, and UMAP was used to display the cell distribution (Fig. 4C). Groupwise analysis revealed an obvious increase in the numbers of immune cells such as T- and NK cells, neutrophils, and Mos/Mφs in the IA group. Mapping ALOX5 and TLR4 expression onto the UMAP plot revealed that they were mainly distributed in the Mo/Mφ cluster and neutrophil cluster of the IA group (Fig. 4D). We further visualized the expression of these two hub genes in various cell clusters using violin plots and dot plots. The average expression level of ALOX5 was 1.93 in mast cells and 1.05 in neutrophil clusters, significantly higher than in VSMC clusters (−0.75), fibroblast clusters (−0.76), and endothelial cell clusters (−0.74). For TLR4, the average expression level was 2.12 in neutrophil clusters and 1.18 in Mo/Mφ clusters, significantly higher than in VSMC clusters (−0.64) and fibroblast clusters (−0.52) (Fig. 4E, F).

Fig. 4

scRNA-seq analysis of TLR4 and ALOX5 expression in patient IA. A UMAP visualization of 20 clusters identified through unsupervised clustering. B Dot plot of characteristic marker genes used for cell type identification. C UMAP showing annotated cells. D Groupwise expression of TLR4 and ALOX5 in the IA and STA groups. E Dot plot showing the expression of TLR4 and ALOX5 in each cell cluster. F Violin plot showing the expression of TLR4 and ALOX5 in different cells

Expression of ALOX5 and TLR4 in IA patients and elastase-treated IA mice

To validate the expression of the hub genes ALOX5 and TLR4 in IA patient tissues, we performed IHC and IF staining on IA samples and STA samples, which were used as controls. TLR4 and ALOX5 expression were mainly distributed in the cytoplasm, and their average optical density (AOD) in IA tissues was significantly greater than that in control tissues (p < 0.05, Fig. 5A, B). In addition, IF staining revealed a notable increase in the number of cells positive for TLR4 and ALOX5 expression in the IA tissues (Fig. 5C).

Fig. 5

Validation of hub gene expression in patients. A Immunohistochemical validation of TLR4 and ALOX5 expression in the IA and STA groups. (Scale bar: 100 μm). B Statistical analysis of the AOD of TLR4 and ALOX5 expression. C Immunofluorescence of TLR4 and ALOX5 expression (scale bar: 100 μm). **p < 0.01; ***p < 0.001

We used elastase induction combined with hypertension to construct a mouse IA model (Fig. 6A). Microscopic observation of mouse Willis’ ring morphology revealed local protrusions in blood vessels, which is indicative of aneurysm formation (Fig. 6B). Compared with that in control brain vessels, IF staining revealed a significant increase in the expression of TLR4 and ALOX5 at IA sites (Fig. 6C). Subsequently, brain vessels from the Willis’ rings of the mice were extracted for WB detection, and the results showed a significant increase in the protein expression levels of TLR4 and ALOX5 in the IA group (p < 0.05, Fig. 6D, E).

Fig. 6

Validation of TLR4 and ALOX5 expression in the mouse IA model. A Model construction process. B Typical images of aneurysm formation in the Willis’ rings observed under a microscope. C Immunofluorescence validation of TLR4 and ALOX5 expression in IA tissues (scale bar: 50 μm). D WB analysis of TLR4 and ALOX5 protein expression (n = 3/per group). E Statistical analysis of grayscale values. **p < 0.01; ***p < 0.001