Abstract

   Neutrophil extracellular traps (NETs) released by neutrophils can
   exacerbate inflammation, leading to rupture of intracranial aneurysms
   (IA). This study aims to explore potential NETs-related genes in IA.
   RNA sequencing data for IA were downloaded from the Gene Expression
   Omnibus database. NETs-related genes were screened using weighted gene
   co-expression network analysis (WGCNA) and differentially expressed
   gene (DEG) analyses. Gene Ontology (GO) and Kyoto Encyclopedia of Genes
   and Genomes (KEGG) enrichment analyses were performed. LASSO Cox
   regression analysis identified optimal genes for model construction.
   Immune cell infiltration in IA was studied using CIBERSORT. Five
   NETs-related hub genes were identified in IA, involved in pathways like
   neutrophil chemotaxis, Toll-like receptor signaling, and regulation of
   inflammatory response. A risk score model was developed based on TLR7,
   TLR2, IL1B, ENTPD4, and FPR1. Immune cell infiltration analysis showed
   significant variations between low-risk-IA and high-risk-IA groups.
   Monocytes and neutrophils infiltration proportions were significantly
   positively correlated with the risk score. The ROC analysis showed AUC
   values exceeding 0.85 for both training and validation sets, confirming
   the model’s excellent performance. A novel NETs-related diagnostic
   signature for IA was created, offering new insights into the
   pathogenesis and diagnosis of IA.

Supplementary Information

   The online version contains supplementary material available at
   10.1186/s12883-025-04068-w.

   Keywords: Intracranial aneurysm, NETs, Hub genes, Risk score, Immune
   cell infiltration

Introduction

   Intracranial aneurysm (IA) is an angiogenic disease triggered by
   expansion of the arterial wall due to elastic membrane damage and
   destruction of the media within the arterial wall [[36]1]. This
   condition affects approximately 2–5% of the general population [[37]2].
   IA has the potential to cause non-traumatic subarachnoid hemorrhage
   (SAH), a condition that can lead to severe central nervous system and
   other organ damage, potentially resulting in life-threatening outcomes
   [[38]3, [39]4]. Fortunately, with advancements in medical imaging
   technology and nerve intervention, unruptured intracranial aneurysms
   (UIA) are being detected more frequently, significantly enhancing
   patient survival rates [[40]5]. Despite these improvements, patients
   with IA still encounter uncommon but significant challenges following
   endovascular treatment, including bleeding and recurrence [[41]6,
   [42]7]. Given that most IA cases are asymptomatic until rupture occurs,
   early detection of IA is crucial for monitoring its progression and
   implementing preventive measures [[43]8]. This requires a deeper
   understanding of the underlying molecular mechanisms involved.

   The pathological mechanisms underlying the formation and progression of
   IA remains incompletely understood. Current knowledge has indicated
   that the recruitment and infiltration of immune cells play a vital part
   in IA [[44]9, [45]10]. For instance, the accumulation of immune cells,
   such as macrophages [[46]11], and immune-related molecules, including
   interleukin [[47]12] and chemokines [[48]13], could lead to
   dysregulation of the local immune microenvironment. Interventions
   targeting the immune inflammatory response and immune microenvironment
   have been shown to significantly reduce secondary brain damage, thereby
   improving outcomes for patients with SAH [[49]14]. Despite these
   findings, the exact role of immune cells and immune-related molecules
   in IA has not fully understood. Further exploration is warranted to
   gain a more comprehensive understanding. Neutrophils, the most abundant
   immune cells in peripheral blood, have garnered increasing attention
   for their role in disease development [[50]15–[51]17]. Recent research
   has further revealed that neutrophils in IA patients have been
   activated peripherally, leading to changes in their RNA expression
   profile [[52]18]. Nevertheless, the exact role of neutrophils in IA
   remains incompletely understood, necessitating further investigation.

   The release of neutrophil extracellular traps (NETs), a defense
   mechanism employed by neutrophils [[53]19], comprise a complex network
   of chromatin DNA, histones, and neutrophil granule proteins [[54]20].
   Initially reported in 2004, NETs have demonstrated a crucial influence
   in bacterial killing by neutrophils [[55]21]. Furthermore, NETs have
   been implicated in the development and progression of diseases,
   particularly in those diseases relating to vascular and immune systems
   [[56]22]. A recent report has suggested a strong association between
   NETs and atherosclerotic disease, with their components serving as
   diagnostic biomarkers for this condition [[57]23]. Kato et al. have
   demonstrated that NETs formation exacerbates sepsis [[58]24].
   Similarly, Korai et al. have found that neutrophils promote the rupture
   of IA through NET formation (NETosis) [[59]25]. A recent animal study
   has further revealed that inhibiting NETosis has significantly reduced
   aneurysmal rupture in estrogen-deficient mice compared to controls
   [[60]26]. Despite accumulating pro-inflammatory evidence for NETs, the
   precise mechanism of their action in IA remains enigmatic. Therefore,
   it is imperative to identify convenient and cost-effective biomarkers
   associated with NETs for IA.

   In this study, we aim to investigate the hub genes associated with NETs
   in IA and construct a risk score model. Using comprehensive
   bioinformatics tools, our objective is to unravel the functional
   pathways of these hub genes and comprehensively explore the immune cell
   infiltration of IAs. Our findings are expected to offer a novel
   signature for diagnosing and mitigating IA, thereby contributing to
   enhanced patient outcomes.

Materials and methods

Subjects

   The microarray dataset [61]GSE75436 and mRNA expression profile dataset
   [62]GSE36791 were downloaded from the Gene Expression Omnibus (GEO)
   database ([63]https://www.ncbi.nlm.nih.gov/geo/). There were 30 samples
   (15 IAs and 15 controls) in the [64]GSE75436 dataset, which was
   obtained from the Affymetrix Human Genome U133 Plus 2.0 Array platform.
   The [65]GSE36791 dataset comprised 18 blood samples from healthy
   individuals and 43 blood samples from patients with IAs, which was
   processed on the Illumina HumanHT-12 V4.0 expression bead chip
   platform.

Weighted gene co-expression network analysis

   Weighted gene co-expression network analysis (WGCNA) employs a
   structured approach to group genes exhibiting similar expression
   patterns, transforming gene expression into modules. The “WGCNA”
   package in R software was utilized to build and visualize the network
   [[66]27]. The network dendrogram was generated for visualization,
   meanwhile modules were represented by distinct colors, and modules with
   high similarity were combined. To identify the modules most closely
   relating to IA, the Pearson correlation analysis was conducted to
   estimate the relationship among gene modules.

Differentially expressed gene analysis

   The differentially expressed genes (DEGs) were screened using “limma”
   package of R softeare [[67]28] with the cut-off criteria of|log[2]FC|>
   1 and p.adjust < 0.05.

Functional enrichment analysis

   The Gene Ontology (GO) terms, Kyoto Encyclopedia of Genes and
   Genomes(KEGG) pathway enrichment analysis and Gene Set Enrichment
   Analysis (GSEA) were conducted using the “clusterProfiler” R function
   package [[68]29]. The “DOSE” package of R langue was used to perform
   disease ontology (DO) analysis. The p-value < 0.05 was considered
   statistically significant.

Protein–protein interaction (PPI) networks analysis

   The STRING database ([69]https://string-db.org/, version 11.0) was
   employed to delve into the interactions among the proteins encoded by
   the NETs-related genes, and the Cytoscape (version 3.7.2) was utilized
   to visualize the resulting protein-protein interaction (PPI) network
   [[70]30]. The online database GeneMANIA ([71]http://www.genemania.org)
   was utilized to gain insights into the intricate interactions between
   proteins and assess potential biological pathways.

Predictive model construction

   To further identify the genes associated with NETs, the LASSO Cox
   regression analysis was performed using the R package “glmnet” (version
   4.1-4) [[72]31]. The risk score for each sample was calculated as:
   graphic file with name d33e344.gif

   Where Coef[i] was the risk coefficient for each gene, and x[i] was the
   expression of the gene [[73]32]. Patients were subsequently categorized
   into low-risk and high-risk groups based on the median risk score.

Immune cell infiltration analysis

   The “CICERSORT” R package was used to characterize the composition of
   immune cells for the high-risk group and low-risk group, as well as the
   IAs group and control group. The proportions of a total of 22 immune
   cells were calculated for each sample. In addition, four types of
   immune cell infiltration were estimated using the XCELL database
   ([74]https://xcell.ucsf.edu/).

Statistical analysis

   All statistical analyses were conducted using R software (version
   4.2.0). The Wilcox test was applied to assess the difference in
   continuous variables between the two groups. Statistical significance
   was considered at p < 0.05.

Results

Identification of IA-related modules using WGCNA

   To identify potential modules associated with IA, we performed WGCNA on
   the [75]GSE75436 dataset. The soft power of β = 4 (Fig. [76]1A) was
   determined as soft-thresholding. The results showed that the brown
   module (r=-0.84, p < 0.001) and greenyellow module (r = 0.87,
   p < 0.001) were significantly related to IA (Fig. [77]1B). There was a
   significant correlation between the module membership and gene
   significance of the brown and greenyellow modules (Fig. [78]1C, D).
   Therefore, these two modules were selected for downstream analysis,
   containing a total of 2,249 genes (Table [79]S1).

Fig. 1.

   [80]Fig. 1
   [81]Open in a new tab

   Co-expression analysis of NETs-related genes in [82]GSE75436 cohort.
   (A). Determination of the soft-thresholding (power threshold β = 4).
   (B). Heatmap of the association of gene modules with IA and control
   groups. Numbers in each cell indicate correlation and significance. (C,
   D). Scatter plots showing the correlation of genes with the brown
   module (C) and greenyellow module (D). (E, F). The top 10 GO enrichment
   (E) and the top 20 KEGG pathway enrichment (F) of brown module genes.
   (G, H). The top 10 GO enrichment (G) and the top 20 KEGG pathway
   enrichment (H) of brown greenyellow genes

   GO and KEGG enrichment analyses of genes in the brown and greenyellow
   modules were performed to further determine the biological functions of
   these modules in IA. The results showed that brown module genes were
   significantly enriched in GO terms such as muscle contraction,
   contractile fiber, and immune receptor activity (Fig. [83]1E),
   interestingly, they were also involved in the biological process of
   regulation of neutrophil migration (Table [84]S2). The KEGG enrichment
   results showed that the genes in brown module were significantly
   involved in 61 pathways such as arginine and proline metabolism and
   ECM-receptor interaction (Fig. [85]1F). The greenyellow module genes
   demonstrated significant enrichment in GO terms such as negative
   regulation of protein localization, endoplasmic reticulum lumen, and
   extracellular matrix structural constituent conferring tensile strength
   (Fig. [86]1G). Additionally, they were also prominently involved in 16
   KEGG pathways like protein digestion and absorption (Fig. [87]1H). More
   details of the enrichment were shown in Table [88]S2 and Table [89]S3.

Seven candidate genes related to NETs in IA were screened

   To further screen NETs-related candidate genes in IA, we first
   identified DEGs between IA and control samples. Using the criteria of
   “|log[2]FC|> 1 and p.adjust < 0.05”, a total of 2,823 DEGs were
   identified in the IA group compared to the control group (Fig. [90]2A),
   including 1,534 up-regulated genes and 1,289 down-regulated genes.
   Moreover, a previously published study has identified 69 genes as
   initial biomarkers for NETs [[91]33], which were taken as the
   NETs-related genes. Then, in order to identify NETs-related candidate
   genes in IA more accurately, the cross-analysis was conducted, which
   incorporated 2,249 IA-related module genes (basing on WGCNA), 2,823
   IA-related DEGs, and 69 NETs-related genes. Next, seven overlapping
   NETs-related candidate genes were generated and displayed in the Venn
   diagram (Fig. [92]2B).

Fig. 2.

   [93]Fig. 2
   [94]Open in a new tab

   Identification of candidate genes associated with NETs in IA. (A).
   Volcano plots of DEGs. (B). Venn diagram of the WGCNA module genes,
   DEGs, and published NETs-related genes. (C, D). The top 10 GO
   enrichment (C) and the top 20 KEGG pathway enrichment (D) of candidate
   genes. (E). PPI network interaction map of candidate genes. (F).
   Network interaction map of candidate genes based on GENEMANIA

   To obtain the potential functional information underlying the seven
   NETs-related candidate genes, the GO and KEGG enrichment analyses were
   performed on the above candidate genes. The GO enrichment results
   showed that the seven candidate genes were enriched in immune-related
   biological processes, including positive regulation of interleukin-8
   production, neutrophil chemotaxis, and regulation of inflammatory
   response (Fig. [95]2C, Table [96]S4). Besides, the KEGG pathway
   enrichment analysis identified immune-related pathways, notably the
   Toll-like receptor signaling pathway and neutrophil extracellular trap
   formation (Fig. [97]2D, Table [98]S4). We thus speculated that these
   NETs-related candidate genes may play an important role in
   immune-mediated processes and deserve further attention.

   Going further, we performed PPI analysis using the online database
   STRING, which revealed nine interactions among the seven NETs-related
   candidate genes (Fig. [99]2E). We then employed the GeneMANIA network
   to investigate potential PPI patterns, aiming to deepen our biological
   understanding of the functional correlations among these seven
   NETs-related candidate genes. The results identified the top 20
   proteins that interact with these candidate genes (Fig. [100]2F).

Construction and verification of NETs-related diagnostic model in IA

   To explore the diagnostic and predictive value of these candidate genes
   in IA and enhance their clinical applicability, the machine learning
   algorithm was used to construct a risk score model. Firstly, we have
   analyzed the correlation among the seven candidate genes, while no
   significant strong correlation (R > 0.8) among them was found (Fig.
   [101]S1). We thereby conducted LASSO Cox regression analysis on seven
   candidate genes, of which five genes (corresponding to the lowest
   likelihood bias and best imitative effect) were identified as potential
   optimal genes to build diagnostic model for IA, including TLR7, TLR2,
   IL1B, ENTPD4, and FPR1 (Fig. [102]3A, B). Subsequently, a risk score
   model was constructed, the risk score was calculated based on these
   five genes’ expressions: Risk Score = (TLR7*0.4189663)+
   (TLR2*0.3784702)+ (IL1B*0.6492899)+ (ENTPD4*0.1678660)+ (FPR1*
   0.2000896) (Fig. [103]3C).

Fig. 3.

   [104]Fig. 3
   [105]Open in a new tab

   Construction and verification of risk score model in IA. (A). The cross
   validation curve of the model. X-axis referred to logarithm λ of
   penalty parameter, and Y-axis represented likelihood bias. The smaller
   the likelihood bias, the better the imitative effect. (B). The path
   diagram of regression coefficient. Upper X-axis referred to the number
   of nonzero coefficient, Y-axis referred to the value of coefficient,
   and lower X-axis referred to log (λ). Each curve represented the
   trajectory of the coefficient of independent variable. (C). Coefficient
   of the five selected hub genes. (D). The expression of hub genes in
   high-risk and low-risk groups in the [106]GSE75436 dataset. (E, F). The
   expression of hub genes in IA and control groups in the [107]GSE75436
   dataset (E) and [108]GSE36791 dataset (F). (G, H). Box plots of the
   risk score in IA and control groups in the [109]GSE75436 dataset (G)
   and [110]GSE36791 dataset (H)

   The samples in [111]GSE75436 dataset were stratified into high-risk and
   low-risk groups using the median risk score value as the cutoff, in
   order to validate the predictive accuracy of the risk score model. We
   initially compared the expression levels of the five genes between the
   high-risk and low-risk groups. Our observations revealed significant
   differences, the high-risk group exhibiting elevated expression levels
   compared to the low-risk group (Fig. [112]3D). Furthermore, we analyzed
   the expression levels of these genes between IA and control samples in
   both the [113]GSE75436 and [114]GSE36791 datasets. In the [115]GSE75436
   cohort, all five genes were significantly over-expressed in IAs
   compared to controls (Fig. [116]3E). In the [117]GSE36791 cohort, all
   genes except TLR7 also demonstrated significantly higher expression in
   the IA group (Fig. [118]3F). These findings suggested a consistent
   expression pattern among the majority of these genes in IA, further
   supporting that they could serve as the NETs-related hub gens for IA.
   In addition, in both cohorts, IA patients had significantly higher risk
   scores than the control group (Fig. [119]3G, H), implying a relatively
   reliable performance of our risk score. Overall, high risk samples were
   at a higher predisposition for developing IA.

Exploration of biological significance of NETs-related diagnostic model

   To gain deeper insights into the functional role of hub genes related
   to NETs in IA, we delved into the biological implications of the risk
   score model. Initially, we conducted DEG analysis comparing high-risk
   and low-risk groups, and a total of 2,481 DEGs were identified in
   high-risk group, compared to low-risk group (Fig. [120]4A). DO
   enrichment analysis on the 2,481 DEGs showed that they were
   significantly enriched in 485 DO terms such as demyelinating disease
   and aortic aneurysm. The top 10 most significant DO terms were shown in
   Fig. [121]4B, and all the enrichment results were shown in Table
   [122]S5.

Fig. 4.

   [123]Fig. 4
   [124]Open in a new tab

   Exploration of the biological significance of the risk score model.
   (A). Volcano plots of DEGs in high-risk and low-risk groups. (B). The
   top 20 DO enrichment of DEGs between high-risk and low-risk groups. (C,
   D). The top 5 activated (C) and suppressed (D) pathways according to
   p < 0.05 based on GSEA. (E). Heatmap of differentially enriched KEGG
   pathway between high-risk group and low-risk group based on GSVA

   Then, GSEA enrichment analysis revealed that 128 KEGG pathways were
   notably enriched under the screening criteria of|NES|>1 and p < 0.05
   (Table [125]S6). Notably, several pathways, including asthma,
   intestinal immune network for IgA production, leishmaniasis, systemic
   lupus erythematosus, and neutrophil extracellular trap formation, among
   others, were significantly activated in high-risk group (Fig. [126]4C,
   [127]S2). On the contrary, cardiac muscle contraction, dilated
   cardiomyopathy, hypertrophic cardiomyopathy, motor proteins, and
   regulation of lipolysis in cardiac muscle contraction adipocytes, among
   other related pathways, were significantly inhibited in patients with
   IA (Fig. [128]4D, [129]S2).

   In addition, we also conducted GSVA analysis to explore the biological
   mechanism leading to differences between the high-risk and low-risk
   groups. We found that some signaling pathways were significantly
   diversity between the two groups. For instance, the samples in the
   high-risk group were involved in glycan biosynthesis and toll like
   receptor signaling pathway (Fig. [130]4E).

Distinct immune cell infiltration characteristics between different risk
score groups

   To investigate the association between risk scores and immune status in
   patients with IA, the CIBERSORT algorithm was utilized to calculate the
   relative abundance of 22 distinct types of immune cells. We combined
   the [131]GSE75436 and [132]GSE36791 datasets, called combined-dataset.
   All IA samples from the combined-dataset were categorized into
   high-risk-IA and low-risk-IA groups based on the median risk score. The
   results revealed noteworthy disparities in the proportion of 14
   distinct types of immune infiltrating cells between the high-risk-IA
   and low-risk-IA groups (Fig. [133]5A). Notably, the high-risk-IA group
   exhibited a significantly elevated proportion of T cell CD4 + naive, NK
   cell resting, monocyte, and neutrophil compared to the low-risk-IA
   group. In contrast, the proportion of lymphocytes, such as B cell
   plasma and T cell CD4 + memory resting, were significantly lower in the
   high-risk-IA group. This observation aligns with the recently
   identified phenomenon of elevated inflammatory cells and reduced
   lymphocytes in the peripheral blood of patients with IA.

Fig. 5.

   [134]Fig. 5
   [135]Open in a new tab

   The landscape of immune cell infiltration in IA patients. (A). The
   immune cell infiltration in the high-risk-IA and low-risk-IA groups.
   (B). The immune cell infiltration in the combined-IA and
   combined-control groups. (C). Correlation analysis of six immune cells
   and risk score. (D, E). The proportion of four classes of immune cells
   in the high-risk-IA and low-risk-IA groups (D) and the combined-IA and
   combined-control groups (E) based on XCELL

   Moreover, the differences in immune cell infiltration between the
   combined-IA (IA samples in the combined-dataset) and combined-control
   (control samples in the combined-dataset) groups were also explored.
   Significant disparities were observed in the infiltration of nine types
   of immune cells between the combined-control and combined-IA groups
   (Fig. [136]5B). Consistent with previous findings, the combined-IA
   group exhibited a notably higher proportion of monocytes and
   neutrophils compared to the combined-control group.

   Finally, we identified six overlapping immune cell infiltration that
   exhibited significant differences in both high-risk-IA vs. low-risk-IA
   and combined-IA vs. combined-control. In order to further confirm the
   relationship between the six overlaping immune cell infiltration and
   risk score, we conducted correlation analysis, revealing a significant
   positive correlation between monocytes and neutrophils with risk score.
   Conversely, T cell CD4 + memory resting, T cell follicular helper,
   macrophage M1, and macrophage M2 exhibited a notable negative
   correlation with risk score (Fig. [137]5C).

   In addition, we utilized XCELL to analyze the proportions of four
   distinct immune cell types, including dendritic cells, lymphocytes,
   macrophages, and mast cells, and found that macrophages had a
   significant difference in combined-IA compared with the
   combined-control group. However, no significant differences in other
   immune cells were observed (Fig. [138]5D, E).

The risk score model exhibited a good diagnostic performance for IA

   To quantitatively evaluate the diagnostic and predictive capabilities
   of the risk score model, we conducted an ROC analysis using the
   [139]GSE75436 and [140]GSE36791 datasets. The AUC of the model in the
   [141]GSE7536 cohort and the [142]GSE36791 cohort were 0.996 and 0.851
   (Fig. [143]6A, B), respectively, indicating that the risk score model
   exhibited a good diagnostic performance.

Fig. 6.

   [144]Fig. 6
   [145]Open in a new tab

   Clinical value evaluation and drug prediction of risk score module. (A,
   B). ROC curves of the risk score in the [146]GSE75436 dataset (A) and
   [147]GSE36791 dataset (B). (C). Five hub genes target drug network
   predicted by the DGIdb database

   Consequently, we proceeded to delve into the drug molecules potentially
   associated with these five hub genes. Utilizing the DGIdb database, we
   analyzed the gene-drug interactions and discovered 4, 37, 6, and 12
   drugs related to TLR2, IL1B, FPR1, and TLR7, respectively
   (Fig. [148]6C, Table [149]S7). However, it was unfortunate that no
   targeted drugs were identified for the prediction of ENTPD4.

Discussion

   IA is a rare but potentially life-threatening cerebrovascular disease.
   Before rupture, IA is often asymptomatic, necessitating regular imaging
   studies for early detection [[150]34, [151]35]. Immune inflammation
   plays a pivotal role in the progression and rupture of IA [[152]36].
   Biomarkers derived from the transcriptome of circulating neutrophils
   have shown promise in detecting unruptured IA [[153]18]. NETs were
   shown to regulate neutrophil development and were involved in
   inflammation through various mechanisms [[154]37]. However, a limited
   insight of mechanisms underlying the NETs of IA has hindered progress
   in diagnostic and therapeutic techniques. Consequently, identifying
   biomarkers associated with NETs in IA remains an urgent research
   priority. Here, we have developed a novel diagnostic signature related
   to NETs in IA.

   Initially, we utilized transcriptome data downloaded from GEO to
   identify seven candidate genes related to NETs in IA. These genes were
   involved in biological functions such as interleukin-8 production,
   neutrophil chemotaxis, regulation of inflammatory response, and
   Toll-like receptor signaling pathway. The Toll-like receptor (TLR), a
   well-characterized pattern recognition receptor, has been shown to play
   a crucial role in pathogen recognition and the elicitation of the
   innate immune response [[155]38]. It was reported that interleukin 8
   (IL-8) was a pro-inflammatory cytokine that mainly functioned in the
   chemotaxis of neutrophils in the blood [[156]39]. Therefore, we
   postulated that these candidate genes associated with NETs played
   pivotal roles in immune-mediated processes in IA.

   To identify the hub genes related to NETs in IA, we employed LASSO
   machine-learning techniques to construct a model and identify more
   precise biomarkers. Ultimately, TLR7, TLR2, IL1B, ENTPD4, and FPR1 were
   identified as hub genes for IA, and a risk score model was then
   formulated. Notably, the five hub genes were overexpressed in the
   high-risk group compared to the low-risk group. We hypothesized that
   the elevated expression of these five hub genes could serve as a
   crucial factor in the pathogenesis of IA. To corroborate our findings,
   we validated the consistency of these expression patterns in an
   independent dataset. Previously, TLR2 was found to be significantly
   up-regulated in IA patients, indicating its crucial role in the
   development of IA among the Chinese population [[157]40]. More
   recently, TLR2 was discovered to be a central gene in
   atherosclerosis-associated IA, primarily promoting its pathogenesis by
   activating the inflammatory response and apoptosis mediated by the
   TLR2-Myd 88-NF-κB pathway [[158]41]. IL1B has also emerged as a
   potential biomarker for IA [[159]42], with its expression level
   positively correlating with immune cells such as CD8 T cells [[160]43].
   Du et al. previously proved that FPR1 was up-regulated in IA [[161]44],
   and that its upregulation induces Mac-1-mediated neutrophil synthesis,
   thereby exacerbating the progression of abdominal aortic aneurysm (AAA)
   [[162]45]. In summary, our study has reliably identified NETs-related
   genes. Furthermore, we discovered that IA patients from both cohorts
   exhibited significantly higher risk scores compared to controls,
   indicating that individuals with higher risk scores might be more prone
   to develop IA. These reminded us to intervene in time for patients with
   high risk scores.

   In addition, we conducted a detailed analysis of immune cell
   infiltration to elucidate the intricate relationship between the risk
   score model and immune cells. In both the high-risk-IA and combined-IA
   groups, we observed an abnormal elevation of infiltrating immune cells,
   including T cell CD4 + naive, NK cell resting, monocyte, and
   neutrophil, which aligns with previous research findings. NETs played a
   pivotal role in inflammation and tissue remodeling, leading to an
   increased risk of aneurysm rupture [[163]25]. Meanwhile, it has been
   previously documented that under the background of IA, endothelial
   progenitor cells have exerted crucial role in protecting vascular
   integrity via replacing injured and dysfunctional endothelial cells
   [[164]46]. Consequently, in IA, hemodynamic forces trigger endothelial
   damage and neutrophil activation, leading to the secretion of
   pro-inflammatory cytokines, neutrophil recruitment, and subsequent NETs
   formation. Additionally, the enhanced migratory capacity of NK cells in
   IA patients’ peripheral blood promoted their aggregation and activation
   within the aneurysm wall, thereby contributing to disease progression
   [[165]47]. Monocytes were dysregulated in IA patients, favoring a
   pro-inflammatory response [[166]48]. Correlation analysis further
   reinforced this finding, revealing a positive association between
   monocyte and neutrophil infiltration and the risk score. These findings
   underscored the intimate association between inflammation, immune
   responses, and IA, highlighting the potential of immune-based therapies
   in managing this condition.

Conclusions

   Our study successfully constructed a risk score model and identified
   five NETs-related hub genes including TLR7, TLR2, IL1B, ENTPD4, and
   FPR1 for IA. All five hub genes were significantly up-regulated in both
   the high-risk group and IA patients and involved in inflammatory
   response regulation. In conclusion, the risk score model could serve as
   a promising diagnosis signature for IA. Our study offered novel
   insights into the pathogenesis and diagnosis of IA, especially
   involving NETs related roles.

Electronic supplementary material

   Below is the link to the electronic supplementary material.
   [167]12883_2025_4068_MOESM1_ESM.pdf^ (5.8KB, pdf)

   Supplementary Material 1: Figure S1. The correlation results among
   seven candidate genes.
   [168]12883_2025_4068_MOESM2_ESM.jpg^ (349.6KB, jpg)

   Supplementary Material 2: Figure S2 The top 10 activated and suppressed
   pathways according to NES1 based on GSEA in the GSE75436 cohort.
   [169]Supplementary Material 3^ (133.5KB, xls)
   [170]Supplementary Material 4^ (456KB, xls)
   [171]Supplementary Material 5^ (122.5KB, xls)
   [172]Supplementary Material 6^ (153KB, xls)
   [173]Supplementary Material 7^ (192.5KB, xls)
   [174]Supplementary Material 8^ (86KB, xls)
   [175]Supplementary Material 9^ (34.5KB, xls)

Acknowledgements