2.1 PD1− CD4 Tem and T-bet+ CD4 Tcm from RIA showed upregulated expression of the glycolysis-related rate-limiting enzymes PKM2 and LDH, whereas CD56^dim DN-NKT exhibited upregulation of multiple metabolic markers

To explore the differences in T cells between the UIA and RIA groups, we isolated T cell subtypes (CD4 T, CD8 T, and NKT) for Flowsom clustering analysis. The results revealed nine subgroups of CD4 T cells (Figure 1A), 11 subgroups of CD8 T cells (Figure 1C), and five subgroups of NKT cells (Figure 1E). Based on the expression patterns of CD45RA, CD45RO, and CCR7 (Figure 1B,D), CD4 T09 and CD8 T10 were considered Tn cells; five groups of CD4 T cells (CD4 T01, CD4 T02, CD4 T03, CD4 T04, and CD4 T06) and one group of CD8 T cells (CD8 T04) was identified as central memory T cells (Tcm); two groups of CD4 T cells (CD4 T05 and CD4 T07) and five groups of CD8 T cells (CD8 T01, CD8 T02, CD8 T03, CD8 T08, and CD8 T09) were designated as effector memory T cells (Tem). CD8 T05 and CD8 T06 were considered terminally differentiated effector memory cells (Temra). Among them, CD4 T01, CD8 T01, and CD4 T07 belonged to the PD1+ subgroup (Figure 1B,D). Additionally, based on the expression of CD25 and CD56 (Figure S3A), CD8 T04 was identified as a group of effector T cell (Teff). According to Foxp3 expression (Figure 1D), CD8 T09 was considered a group of regulatory T cells (Treg). For NK cells, they were broadly categorized into CD4+ T cell subgroups (NKT01 and NKT02), CD8+ T cell subgroups (NKT04 and NKT05), and double negative T cell (DN-NKT) subgroup (NK03) based on the expression patterns of CD4 and CD8 (Figure 1F). Among these, NKT02 and NKT04 belong to the high expression of CD56 subgroup, whereas other cells belong to the low expression subgroup (Figure 1F).

Whether in the UIA group or the RIA group, CD4 T05 (UIA: 19.48% and RIA: 24.33%) and CD8 T06 (UIA: 22.34% and RIA: 31.14%) were respectively the most abundant CD4 T (Figure 1G, top) and CD8 T cell subsets (Figure 1G, middle) in peripheral. Among NKT cells, NKT05 (33.54%) was the most abundant subset in the UIA group, whereas NKT04 (27.30%) was the most abundant in the RIA group (Figure 1G, bottom). Comparing the proportions of various T cell subsets between the two groups, we found that the RIA group has more PD1+ CD4 Tem (CD4 T07) (p = 0.00406 < 0.01), whereas the UIA group has more transitional memory subset (CD4 T08) (p = 0.0108 < 0.05) (Figure S3B–D).

Metabolic molecule comparative analysis revealed that CD56^dim DN-NKT (NKT03) was the subset with the greatest difference among T cells (Figure 1H). In the RIA group, NKT03 showed upregulated expression of GLUD (p = 0.0216 < 0.05), GLUT1 (p = 0.008 < 0.05), mTOR (p = 0.0183 < 0.05), and CPT1 (p = 0.000547 < 0.01). The CD4 Tcm (CD4 T04) exhibited upregulated LDH (p = 0.0475 < 0.05) in the RIA group, whereas CD98 (p = 0.046 < 0.05) was upregulated in the UIA group (Figure 1H). Furthermore, the RIA group showed upregulated PKM2 (p = 0.0137 < 0.05) in the CD4 Tem (CD4 T05) and upregulated SDH (p = 0.0328 < 0.05) in the CD56^bright CD8+ NK (NKT04) subset (Figure 1H). In addition, functional molecule comparisons revealed upregulated expression of the PD1 molecule in several T cell subsets in the RIA group (CD4 T01, CD4 T07, CD8 T01, and NKT03) (Figure 1H).

2.2 GATA3+ B cells from RIA exhibited downregulation of PKM2

To explore potential differences in B cells between RIA and UIA, B cells were isolated and subjected to Flowsom clustering analysis. A total of eight distinct cell subsets were identified based on the expression of surface molecules (Figure 2A). Among them, B08 was classified as a CD19−CD20+ cell subgroup, whereas the other subgroups belonged to the CD19+CD20+ category (Figure 2B). Further subdivision of B cells was achieved using additional cell surface markers (Figure 2B).

Comparing the proportions of various B cell subsets between the two groups, it was observed that B01 was the most abundant subgroup in both UIA (42.18%) and RIA (42.46%) (Figure 2C). The least cell subset in the UIA group was B03 (2.10%), whereas B05 held that position in the RIA group (1.65%) (Figure 2C). Comparative analysis between the two groups did not reveal any statistically significant differences (Figure S3E).

Comparative analysis of metabolism molecules among various cell subsets between the two groups revealed that GATA3+ B cells (B04) from the UIA group expressed higher levels of the PKM2 compared with the counterpart in RIA group (p = 0.0446 < 0.05) (Figure 2D). The functional molecule comparison also revealed that PD1+ B cells (B02) and T-bet+ B cells (B07) from RIA respectively showed higher expressions of IFNγ (p = 0.0417 < 0.05) and PD1 (p = 0.0404 < 0.05) (Figure 2D).

2.3 CD56^dim NK cells from RIA showed upregulated expression of LDH

To explore differences in NK cells between the UIA and RIA groups, NK cells were isolated and subjected to Flowsom clustering analysis, revealing five distinct subsets (Figure 2E). Based on the expression of CD56 molecule, NK04 and NK05 belonged to the CD56^bright subgroup, whereas the remaining cells were categorized as CD56^dim subgroup (Figure 2F).

The proportions of various NK subsets were similar between the two groups, with NK01 constituting the majority (UIA: 69.82%; RIA: 70.75%) and NK05 being the least represented (UIA: 2.13%; RIA: 1.66%) (Figure 2G). Comparative analysis between the two groups did not reveal any statistically significant differences (Figure S3F).

Comparative analysis of metabolic profiles of NK subsets between the two groups revealed that CD56^dim NK (NK03) from the RIA group expressed higher levels of the LDH (p = 0.0155 < 0.05) and Ki67 (p = 0.0449 < 0.05) (Figure 2H). CD14+CD56^bright NK (NK04) from the UIA group displayed increased expression of CD36 (p = 0.0491 < 0.05) (Figure 2H). Functional molecule comparison revealed that NK03 from the RIA group exhibited high expression of IFNγ (p = 0.0424 < 0.05) and IL10 (p = 0.0267 < 0.05) (Figure 2H).

2.4 mDC and iMo from RIA showed upregulated expression of LDH and CPT1, whereas ncMo upregulated LDH and PKM2. In addition, a special subset of CD127+ cMo from RIA group upregulated mTOR

To explore differences in monocytes between the UIA and RIA groups, monocytes were isolated and Flowsom clustering analysis was performed to explore the difference between UIA and RIA groups. The results found 11 distinct subsets (Figure 3A). Based on the expression patterns of CD14 and CD16, the subsets were categorized as CD14+CD16+ intermediate monocytes (iMo) (M01 and M08), CD14−CD16+ nonclassical monocytes (ncMo) (M03) and CD14+CD16− classical monocytes (cMo) (M06, M07, M10, and M11) (Figure 3B). Additionally, M02 and M05 expressed CD1c, resembling two subgroups of myeloid dendritic cells (mDCs) (Figure 3B). M09 expressed CD123, suggesting a potential subgroup of plasmacytoid dendritic cells (pDCs) (Figure 3B).

Comparing the proportions of various monocyte subsets between the two groups, it was observed that M06 represented the most abundant cell subset (UIA: 42.10%; RIA: 47.26%), whereas M10 constituted the least abundant subset (UIA: 1.11%; RIA: 0.36%) (Figure 3C). Comparing the proportions of various monocyte subsets between the two groups, we found that M02 (p < 0.001), M05 (p < 0.001), and M10 (p < 0.001) were enriched in the UIA group, whereas M06 (p = 0.00371 < 0.01) and M09 (p = 0.0104 < 0.05) were enriched in the RIA group (Figure S3G).

Metabolic molecule comparative analysis revealed that mDC (M05) was the subgroup with the greatest difference among monocytes (Figure 3D). M05 from RIA group showed upregulated expression of LDH (p = 0.00211 < 0.01) and CPT1 (p = 0.00232 < 0.01), whereas M05 from UIA group showed upregulated expression of GLUT1 (p = 0.0417 < 0.05). mTOR was also found to be upregulated in the CD127+ cMo (M10) subset from the RIA group (p = 0.038 < 0.05) (Figure 3D). Furthermore, we also observed upregulated expression of metabolic molecules in the ncMo (M03) and iMo (M08) subsets from the RIA group (Figure 3D). Specifically, M03 showed an upregulation of LDH (p = 0.0308 < 0.05) and PKM2 (p = 0.000231 < 0.01), whereas M08 exhibited an upregulation of LDH (p = 0.0108 < 0.05) and CPT1 (p = 0.0011 < 0.01) (Figure 3D). Functional molecule comparisons revealed upregulation of the PD1 molecule in the M02 (p = 0.0378 < 0.05) and M11 (p = 0.0365 < 0.05) from RIA group (Figure 3D). Higher expression of TNF-α was also observed in M05 (p = 0.0224 < 0.05) and M08 (p = 0.0224 < 0.05) with significantly different metabolic markers (Figure 3D).

2.5 CD101+HLADR+ neutrophils were enriched in the RIA group and various subsets from RIA group downregulation of anti-inflammatory factors

To explore differences in neutrophils between the UIA and RIA groups, isolated granulocytes were analyzed using a particular antibody panel. The Flowsom clustering was employed to investigate subsets, revealing a total of 12 cell clusters (Figure 4A). Among these, N03, N04, and N09 expressed CD33 and CD14, resembling polymorphonuclear myeloid-derived suppressor cells (PMN-MDSCs) (Figure 4B,C). N01 and N12 exhibited HLA-DR expression, suggesting distinctive functional statuses. N05, N06, and N08 respectively showed CD34, CD117, and CD123 expression, implying their association with immature subsets (Figure 4B,C). Conversely, mature markers CD10 and CD101 were respectively detected on N07 and N01 surfaces, characterizing these clusters as mature subsets. N02 exhibited typical neutrophil surface markers (CD66b+CD11b+CD15+CD14−CD16+) (Figure 4B,C). Additionally, N10 showed CD56 expression, delineating distinct cellular subset (Figure 4B,C).

Regardless of the RIA or UIA groups, N02 constituted the majority of the cell subset (UIA: 64.76%; RIA: 70.04%). Meanwhile, N04 represented the least abundant subset in both groups (UIA: 0.56%; RIA: 0.04%) (Figure 4D). Comparative analysis of subset proportions between the two groups revealed N01 was predominantly observed in the RIA group (p = 0.00297 < 0.01), whereas N04 exhibited higher occurrence in the UIA group (p = 0.0038 < 0.01) (Figure 4E).

Functional analysis revealed a more active state in the UIA group. N03 exhibited the most significant functional differences between two groups, with all functional molecules expressed at higher levels in UIA group (Figure 4F). N04 from the UIA group expressed higher levels of IL4 (p = 0.0281 < 0.05) and TGFβ (p = 0.0374 < 0.05) compared with the RIA group. N05 from the UIA group exhibited elevated expression of ARG1 (p = 0.031 < 0.05) (Figure 4F). Additionally, elevated TGFβ expression was observed in N01 (p = 0.00619 < 0.01), N06 (p = 0.000527 < 0.001), and N08 (p = 0.00789 < 0.01) from the UIA group (Figure 4F).

2.6 Distinct immune cell interactions between UIA and RIA group

To investigate differences in the associations between immune cell subsets in UIA and RIA groups, we performed correlation analyses on the proportions of major immune cell subsets discovered by cyTOF (including major cell types, high and low metabolic subsets, clusters with substantial metabolic differences).

The statistically significant correlation results showed that the associations among cells were more complex and diverse in the UIA group. Specifically, in the UIA group, the correlations were mainly associated with pDCs, CD56^dim NK, CD8 Temra, and cMo. CD56^dim NK were positively correlated with CD8 Treg, CD8 Temra, and CD4 Tn, whereas negatively correlated with PMN-MDSCs (N03, N04, and N09). CD8 Temra were positively correlated with CD56^bright NK and NKT Hm (high metabolic subsets, NKT02 and NKT04), and negatively correlated with mDCs and cMo. pDCs were positively correlated with CD4 Tn, PMN-IM (immature PMNs, N05, N06, and N08), and negatively correlated with PMN-MDSCs, NKT Lm (low metabolic subsets within NKT, excluding NKT02 and NKT04), NKT03 (clusters with highly divergent metabolic profiles), and CD19−CD20+ B cells. cMo were positively correlated with CD4 Tn, CD4 Temra, and PMN-MDSCs. Additionally, we observed negative correlations between NKT Hm and PMN-MDSCs and between CD8 Tcm and CD8 Teff. PMN-IM showed negative correlations with PMN-MDSCs (Figure 5, left).

In the RIA group, the correlations mainly involved ncMo, CD8 Temra, and NKT Hm. Specifically, CD8 Treg and NKT Hm exhibited positive correlations with ncMo, whereas CD8 Temra, CD4 Temra, and CD4 Tcm showed negative correlations. CD8 Temra were positively correlated with PMN-IM and NKT03 but negatively correlated with NKT Hm. NKT Hm showed negative correlations with pDCs and positive correlations with iMo. Furthermore, PMN-MDSCs displayed negative correlations with CD4 Temra and NKT Lm but a positive correlation with CD8 Tn (Figure 5, right).

2.7 RNA-seq analysis of intracranial aneurysm tissues

To further analyze the metabolic changes of immune cells in intracranial aneurysm tissues, we analyzed RNA-seq data from 11 RIA tissues and 8 UIA tissues from the GEO database (Figure 6A). GO and KEGG enrichment analysis revealed that differentially expressed genes (DEGs) were mainly involved in immune processes related to monocytes and DC cells (Figure 6C). Additionally, DEGs were also enriched in immune processes such as T cell activation, B cell activation regulation, and tumor necrosis factor production (Figure 6D). These results were consistent with the findings of CyTOF analysis, highlighting the roles of monocytes, DCs, T cells, and B cells in the rupture of intracranial aneurysms. Furthermore, we focused on the differential expression of genes related to cell metabolism between the two groups. The results showed that metabolism-related genes were mainly enriched in pathways such as fatty acid metabolism, glycolysis/gluconeogenesis, and oxidative phosphorylation (Figure 6E). Additionally, DEGs were also found in the pathways of tryptophan metabolism, as well as in the metabolism pathways of alanine, aspartate, and glutamate (Figure 6E). Further Gene Set Enrichment Analysis (GSEA) confirmed the upregulation of these DEGs in the RIA group (Figure 6F,G,H).

To further analyze the metabolic changes of immune cells in RIA tissues, we conducted immune infiltration analysis of RNA-seq data. First, the CIBERSORT algorithm was used to assess the composition of immune cells in tissues (Figure 7A). The results showed that the UIA group contained more Tregs, Follicular helper T cell (Tfh), B naive cells, and activated mast cells, whereas the RIA group contained more B memory cells, neutrophils, eosinophils, naive CD4 T cells, M0 macrophages, and resting NK cells (Figure 7B). We then analyzed the correlation between immune cells and common metabolic processes. The statistically significant differences were mainly from the RIA group. Among them, lipid metabolism-related genes were positively correlated with B memory cells (p = 0.0063 < 0.05, R = 0.76), M0 macrophages (p = 0.02, R = 0.69), and CD4 T naive cells (p = 0.0016 < 0.05, R = 0.83) from the RIA group, whereas negatively correlated with activated DC cells (p = 0.0027 < 0.05, R = −0.81), B naive cells (p = 0.023 < 0.05, R = −0.67), resting CD4 T memory cells (p = 0.0051 < 0.05, R = −0.77), and CD8 T cells (p = 0.019 < 0.05, R = −0.71) from the RIA group (Figure 7C). Furthermore, we found that tricarboxylic acid cycle (TAC) cycle-related genes were negatively correlated with resting NK cells (p = 0.01 < 0.05, R = −0.73) from the RIA group, whereas positively correlated with activated NK cells (p = 0.025 < 0.05, R = 0.67) from the RIA group (Figure 7C). Similarly, M2 macrophages (p = 0.019 < 0.05, R = 0.71) from the RIA group were positively correlated with TAC cycle-related metabolic genes (Figure 7C). We also compared the metabolic profiles of neutrophils from the RIA and UIA groups. The results showed that TAC cycle-related genes were positively correlated with neutrophils from the UIA group and negatively correlated with neutrophils from the RIA group. To be specific, a comparison revealed that neutrophils from the RIA group were negatively correlated with OGDH (p = 0.025 < 0.05) (Figure 7D). In addition, the glucose metabolism-related gene PDHA2 was negatively correlated with neutrophils from the UIA group (p = 0.037 < 0.05) and positively correlated with neutrophils from the RIA group (p = 0.037 < 0.05) (Figure 7D).

These results suggests that there are differences between the local environment and the peripheral immune environment. When peripheral immune cells infiltrate local lesions, corresponding metabolic reprogramming may occur, altering their functions and thus affecting the stability of aneurysms.

5.1 Human specimens and ethics statements

Between October 2022 and April 2023, 72 cases of IA patients were included according the inclusion criteria (55 cases of UIAs and 17 cases of RIAs). Ruptured aneurysms were determined based on the presence of SAH according to image. Inclusion criteria: (a) Patients diagnosed with intracranial aneurysms by digital subtraction angiography, computed tomography angiography (CTA), or magnetic resonance angiography. (b) Patients with ruptured aneurysms who presented symptoms and were admitted within 6 h. Exclusion criteria: (a) Patients with tumor diseases; (b) patients with chronic or acute systemic inflammatory diseases; (c) patients using immunosuppressive medications; (d) patients who have received chemotherapy, radiotherapy or treatments potentially impairing the systemic immune system; (e) patients with liver or kidney impairments. All blood samples were collected upon admission. Among them, 58 patients had complete preoperative imaging data. The clinical information of 72 enrolled patients was presented in Table S1 and 58 patients with complete imaging data were presented in Table S2. Figure 8 illustrated the concise workflow of this study.

5.2 Isolation and characterization of single cells from PBMCs and PMNs

Fresh blood samples obtained upon admission were collected in EDTA anticoagulation tubes. Within a 2-h timeframe, the blood samples were overlaid with 5.0 mL of Polymorphprep (Serumwerk Bernburg AG; 1895). Following this, the tubes were centrifuged at 500×g for 30 min at a temperature ranging from 18 to 22°C. After centrifugation, the upper band (containing PBMCs) and the lower band (containing PMNs) were carefully aspirated and transferred separately into two centrifuge tubes. These cellular fractions were then resuspended in isotonic saline (0.9% NaCl) and underwent an additional round of centrifugation.

After centrifugation, the cells were washed using RBC lysis buffer (Solarbio; R1010), followed by another round of centrifugation. Subsequently, each tube received 0.5 mL of cisplatin (Fluidigm; 201064) for cell viability labeling. The reaction was halted after a 2-min incubation by adding 1 mL of 2% FBS (Gibco; 10091−148). Following another centrifugation step and removal of the supernatant, the cells were fixed with 1 mL of 1.5% paraformaldehyde (Biosharp; BL539A) for 15 min. Neutralization was achieved using 2 mL of 2% FBS, followed by another centrifugation step. Finally, the cells were resuspended in a cell preservation solution (Bioteh; C41100) for subsequent analysis.

5.3 Mass cytometry

To assess distinctions among PBMCs, we designed a marker set comprising 36 antibodies (markers of distinguishing various PBMC subsets, markers of metabolic profiles, and markers of function). Additionally, to investigate the heterogeneity of PMNs in peripheral blood, we utilized a panel of 24 antibodies (markers of distinguishing PMN subsets and markers of function). These antibodies, purchased in a purified state from BioLegend (San Diego, USA), were subsequently labeled in-house using the Maxpar X8 Multimetal Labeling Kit (Fluidigm, USA) in strict accordance with the instructions. For a comprehensive list of the antibodies used, along with their corresponding isotopes, please refer to Table S3 (PBMCs panel) and Table S4 (PMNs panel).

5.4 CyTOF data analysis

The CyTOF data were obtained in the. fcs file format using the CyTOF2 system. To facilitate normalization, EQ Four Element Beads were incorporate and a MATLAB-based technique was applied. Subsequently, the data were uploaded to Cytobank (https://premium.cytobank.cn). Initially, bead filtering was conducted, followed by the application of specific gating criteria to isolate single cells. Viable cells were subsequently identified based on 193Ir staining, and CD45+ cells were gated to separate PBMCs and PMNs (refer to Figure S1 for details). Subsequent data analysis was conducted utilizing the automated dimensionality reduction algorithm known as Flowsom, implemented within the R programming environment. Visualization of the outcomes was accomplished through tSNE, a visual dimensionality reduction algorithm.

5.5 Vascular reconstruction and measurement of morphological parameters

We obtained preoperative CTA data in DICOM format from a high-resolution CTA workstation (Siemens, Berlin, Germany) and transformed them into slice DICOM data with a slice thickness of approximately 0.5 mm. Afterward, we imported this dataset into Mimics 19.0 (Mimics Research 19.0; Materialize, Belgium) for subsequent reconstruction and analysis. Radiological assessments were performed by an experienced neurosurgeon (ML) utilizing high-resolution CTA scans. Measurements of L, d, H, angles (flow angle, aneurysm angle, angle of blood vessel, and mother–child angle), and NSI were derived from the CTA scans, as depicted in Figure S2.

5.6 RNA-seq data analysis

The RNA-seq data were mainly derived from the GSE13353 dataset in the GEO database, which includes RNA-seq data from 11 RIAs and eight UIAs (detailed information can be found at https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE13353). The limma package was used for data normalization. DEGs were selected under the criteria of |log2FoldChange| > 1 and FDR < 0.05. The “org.Hs.eg.db” was used for gene ID conversion, and GO and KEGG enrichment analysis was performed using the “clusterProfiler” R package. Additionally, GSEA analysis of DEGs was conducted using GO and KEGG datasets as references.

A bioinformatics algorithm called CIBERSORT was used to evaluate immune cell infiltrations. The leukocyte gene signature matrix LM22 with 1000 permutations was used to calculate the putative abundance of immune cells (21). The data with a CIBERSORT value of p < 0.05 were filtered and retained for the following analysis. Thus, a matrix of immune cells fractions was generated. SLC2A1, PFKFB2, PKM, PDHA1, PDHA2, DLAT, DLD, LDHA, and LDHB were selected as genes of glucose metabolism; IDH1, IDH2, OGDH, SDHA, SDHB, SDHC, and SDHD were selected as genes of the TAC cycle; CPT1A, CD36, and HADHA were selected as genes of lipid metabolism; SLC38A2, SLC1A5, and SLC3A2 were selected as genes of amino acid metabolism. The “ggcorrplot” R package was used to visualize the correlation between these genes and 22 types of immune cells.

5.7 Statistical analysis

Continuous clinical variables were presented as mean ± interquartile range and categorical variables were expressed as counts (n) and percentages (%). Categorical variables were analyzed using chi-square tests, normally distributed continuous variables were assessed using t-tests and non-normally distributed continuous variables were evaluated using Wilcoxon tests. Scatter plots were utilized for visualizing the comparison of subset proportions between the two groups. Intergroup differences were determined using Wilcoxon tests. Following the normalization of subset proportions, Pearson correlation coefficients were computed to examine relationships between different cell subsets. Statistical analyses were performed using R software and significance was defined as a p value below 0.05.