Two male patients with SAH were enrolled in this study. Paired CSF and PB samples were collected from each patient on day 7 after onset through lumbar drainage. The inclusion criteria for patients with SAH are as follows: (1) age ≥18 years old; (2) sudden headache accompanied by epilepsy and other clinical manifestations; (3) aneurysm subarachnoid hemorrhage was diagnosed through computed tomography examination and digital subtraction angiography on admission. The exclusion criteria were listed as follows: (1) patients with systemic inflammatory diseases such as intracranial infection, urinary tract infection, and other infectious diseases; (2) subarachnoid hemorrhage caused by other diseases, such as arteriovenous malformation, hemorrhagic stroke, and brain trauma; (3) patients using immunosuppressive drugs. This study was approved by the ethics committee of the First Affiliated Hospital of Guangxi Medical University and was conducted in accordance with the Declaration of Helsinki. Informed consent was obtained from the legally authorized representative. In addition, scRNA-seq data for CSF from eight healthy individuals were downloaded from the Gene Expression Omnibus database (GSE134578), which served as healthy control (HC) (Gate et al., 2020). The baseline demographic and clinical data for the healthy controls and patients with SAH were shown in Supplementary Tables S1 and S2.

CSF and PB samples were processed within half an hour after collection. MC was isolated from CSF and PB samples by using Ficoll-Paque PLUS (Solarbio Biotech, China). Briefly, samples diluted with phosphate-buffered saline (PBS) (Solarbio Biotech, China) were layered onto the Ficoll-Paque PLUS, followed by centrifugation and isolation of the buffy coat. The MC suspensions isolated from the buffy coat and mixed with cryoprotectant were frozen for 24 h in an ultralow temperature freezer and then transferred to liquid nitrogen for storage. The MC suspensions were thawed and quality-controlled prior to scRNA-seq.

ScRNA-seq was performed on MC suspensions through the 10x Genomics Chromium platform according to the manufacturer’s instructions. In this study, eligible MC suspensions were loaded on a 10× Chromium microfluidics system to generate single-cell nanoliter-scale gel beads in emulsions (GEMs). Next, mRNA released from MC was reverse-transcribed into cDNA in every GEM. Finally, the cDNA from different GEMs were pooled and amplified by PCR to generate a cDNA library, which was sequenced using an Illumina NovaSeq 6,000.

Cell ranger (10X Genomics) was performed to process raw data into gene expression matrices. Subsequently, Seurat was used for data filtering, normalization, dimensionality reduction, cell clustering (Butler et al., 2018). Low-quality cells were removed using the following criteria: (1) gene numbers <200 or >4,500; (2) total UMI counts <1,000 or >20,000; and (3) percentage of mitochondrial genes >10. Differentially expressed genes (DEGs) across clusters were explored using the Seurat function “FindAllMarkers” with parameters: min. pct = 0.25, logfc. threshold = 0.25. T-SNE (T-distributed Neighbor Embedding) was used to classify and visualize cell subsets (Kobak and Berens, 2019).

SingleR is an R package for automated cell type annotation in scRNA-seq data (Aran et al., 2019). In this study, SingleR and marker genes were used together to identify cell type identification. First, SingleR was performed to display a preliminary cell type annotation, which was subject to further manual identification based on the expression of marker genes and the following reference datasets: CellMarkrer, PanglaoDB, and Human Cell Atlas. Briefly, it is well known that the marker genes CD14 and CD16 are used for monocyte identification (Ziegler-Heitbrock et al., 2010; Ma et al., 2022; Villani et al., 2017; Wu et al., 2023; Shi et al., 2021). In scRNA-seq studies, CD3D and CD79A/CD79B have been recognized as reliable gene markers widely used for T (Menon et al., 2023; Wang et al., 2021; Zernecke et al., 2023; Zhu et al., 2020) and B cells (Tkachenko et al., 2023; Xu et al., 2023a) identification, respectively. In addition, NKG7 and GZMA are the markers for natural killer cells (NK) identification (Yao et al., 2023; Chen X. et al., 2023) and pro-platelet basic protein (PPBP) for platelet identification (Lee et al., 2023).

Pseudotime trajectory analysis was performed to explore cell differentiation and development through the R software Monocle (Qiu et al., 2017). Monocytes with different colors were ordered on the pseudotime trajectory based on the differentially expressed genes. Monocle was used to construct monocyte lineage differentiation trajectory and identify differentially expressed gene modules that regulated the differentiation process.

The Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis were performed to study the heterogeneity of signaling pathways and biological functions across monocyte subsets. The GO annotation includes three categories: cellular component (CC), molecular function (MF), and biological process (BP). KEGG is a database resource for exploring high-level biological functions and signaling pathways. Gene set variation analysis (GSVA) was performed to study the activation of signaling pathways among monocyte subsets in an unsupervised way, showing the enrichment degree of target pathways across groups.

Transcription factors are a class of proteins that regulate the transcription of the target gene. DoRothEA was used to evaluate the activation of transcriptional factors in different cell subsets (Garcia-Alonso et al., 2019). Cell-cell communication interaction can coordinate cell differentiation, immune regulation, and other life activities. Cellchat was used to explore cell-cell communication based on the expression of ligands, receptors, and their cofactors (Jin et al., 2021). The communication interaction networks between monocyte subsets and other immune cell types were observed through Cellchat.

In this study, 4 samples were obtained from two male patients with SAH, including 2 CSF and 2 PB samples. Each sample was sequenced individually using 10x Genomics scRNA-seq (Figure 1A), and the gene expression matrices were pooled for downstream biological analysis. Finally, after the quality control, a total of 17,242 cells were obtained, including 7,224 cells from CSF and 10,018 cells from PB (Supplementary Figure S1). These 17,242 cells were divided into 19 (0–18) clusters by using T-SNE (Figure 1B). And then 19 cell clusters were further identified as four immune cell types and platelet, including monocytes (clusters 0, 1, 3, 11, 12, 15 and 16) with marker genes CD14 and CD16 (FCGR3A) (Ziegler-Heitbrock et al., 2010; Ma et al., 2022; Villani et al., 2017; Wu et al., 2023; Shi et al., 2021), T cells (clusters 2, 5, 6, 7, 10, 13 and 17) with marker gene CD3D (Menon et al., 2023; Wang et al., 2021; Zernecke et al., 2023; Zhu et al., 2020), B cells (clusters 8 and 9) with marker gene CD79A (Tkachenko et al., 2023; Xu et al., 2023a) and natural killer (NK) cells (clusters 4 and 18) with marker gene NKG7 (Figure 1C) (Yao et al., 2023; Chen X. et al., 2023). In addition, cluster 14 was defined as platelet (Lee et al., 2023), which was excluded from the subsequent analysis. The violin plot showed that each marker gene was specifically highly expressed in corresponding cell clusters (Figure 1D). The Feature plot displayed the relative distribution of marker genes across all clusters (Figure 1E). The histogram showed the total number and percentage of each immune cell type from PB and CSF (Figure 1F). Briefly, the 17,242 cells were identified as monocytes, T cells, B cells, NK cells, and platelets. Only monocytes were extracted for downstream analysis to explore the heterogeneity of transcriptome and phenotype across monocyte subsets after SAH.

Monocytes have multiple subtypes that differ in biological function, morphology, and transcriptional profile (Williams et al., 2023). To study monocyte heterogeneity, we further re-clustered monocytes. In this study, 2091 and 5,868 monocytes were obtained from CSF and PB samples, respectively (Figure 2A). Based on the expression levels of CD14 and CD16, these monocytes were traditionally divided into three subsets: classical monocytes (CM), intermediate monocytes (IM) and nonclassical monocytes (NCM) (Figure 2B). CM (CD14⁺⁺ CD16⁻) highly expressed CD14 but no CD16, IM (CD4⁺⁺CD16⁺) represented high expression of CD14 and low expression of CD16 while NCM (CD14⁺CD16⁺⁺) showed high expression of CD16 together with low CD14 (Ziegler-Heitbrock et al., 2010). The violin plot showed the expression of marker genes CD14 and CD16 in each cell cluster (Figure 2C). The histogram showed the total number and percentage of monocyte subsets from PB and CSF (Figure 2D).

Monocle was performed to map the differentiation trajectory of monocyte subsets and study the differentially expressed gene modules to explore the mechanisms of monocyte differentiation. Monocyte differentiation began on the right of the trajectory and progressed gradually to the left, with a minor branch at the bottom right. (Figure 3A). Monocyte subsets with different colors were mapped onto the trajectory, representing a gradual lineage differentiation transition from CM to IM and NCM (Figure 3B).

The heatmap showed that two key gene modules were significantly differentially expressed along the differentiation process of monocyte subsets (Figure 3C). Gene module one, containing the S100 gene family, showed high expression in CM but low expression in IM and NCM. GO enrichment analysis found that gene module one was related to the regulation of inflammatory response, the production of interleukin, and other biological processes (Figure 3C). Conversely, gene module two, including the CXCL gene family, was gradually highly expressed in IM and NCM along with differentiation, associated with immune cell migration and leukocyte chemotaxis (Figure 3C).

The differentiation trajectory had two branches, including the major branch (red Arrow) and the minor branch (black Arrow) (Figure 3D). Most of IM and NCM were mapped onto the major branch, while CM was mapped onto the minor branch. The heatmap revealed two differentially expressed gene modules along the differentiation trajectory (Figure 3E). Gene module one was highly expressed in monocytes differentiated towards the minor branch, which was related to regulation of inflammatory response and positive regulation of defense response (Figure 3E). However, Gene module two was significantly upregulated in monocytes differentiated towards the major branch, which was associated with immune cell chemotaxis and response to chemokine, etc (Figure 3E).

These monocyte subsets showed different gene expression patterns along the differentiation process, displaying significant transcriptome profile and phenotype heterogeneity. Together, these results suggested the differentially expressed gene modules may play a key role in the differentiation and biological functions of monocyte subsets.

GSVA was performed to study how the monocytes mediated neuroinflammation. First, the findings indicated that Toll-like receptor (TLR), Nod-like receptor (NLR), and other inflammation-related signaling pathways were more highly activated in IM from CSF compared with PB (Figure 4A, Arrow). Additionally, further comparative analysis demonstrated that these signaling pathways are significantly upregulated in IM from CSF (Figure 4B, Arrow). To explore whether these pathways were involved in neuroinflammation after SAH, we downloaded scRNA-seq data regarding CSF from eight healthy individuals to serve as healthy controls (HC). The frequency and percentage of monocyte subsets from CSF in both HC and SAH groups were shown in Supplementary Figure S2. Similarly, compared with HC, these pathways were significantly activated (Figure 4C, Arrow) and upregulated (Figure 4D, Arrow) in IM from CSF after SAH. In particular, the TLR and NLR pathways exhibited the highest activation and upregulation among these signaling pathways, which have been proven to be involved in neuroinflammation after SAH (Peng et al., 2024; Peng et al., 2023; Lai et al., 2023; Wang W. et al., 2023). In addition, the activation of transcription factors in IM increased after SAH (Figure 4E). To explore the changes in gene expression in IM, we conducted a comparative analysis of differentially expressed genes (DEGs) in HC and SAH. We found that many inflammation-related genes, especially those from the CCL and CXCL gene families, were significantly upregulated in SAH compared to HC (Supplement DEGs). GO enrichment analysis based on the DEGs showed that various critical biological processes related to neuroinflammation were enriched in IM after SAH (Figure 4F). To conclude, these findings suggest that IM may play a key role in regulating neuroinflammation by mediating the TLR and NLR pathways after SAH.

Immune cells maintain immune system homeostasis through cell communication interaction. Cellchat was performed to study the cell communication network between various immune cell types. There was abundant cell communication interaction between the immune cells in CSF, especially monocyte subsets (Figure 5A). Likewise, communication interaction among various immune cell types was also complex and abundant in PB (Figure 5B). Compared with PB, the number and strength of communication interaction between monocyte subsets were significantly increased in CSF (Figure 5C). Notably, the communication interaction patterns of ligand-receptor pairs among these immune cell types in CSF and PB were different (Figure 5D, E). The macrophage migration inhibitory factor (MIF)- (CD74+CXCR4) and MIF- (CD74+CD44) pairs were highly activated among various immune cell types in both CSF and PB (Figure 5D, E, red box). In short, the communication interactions of monocyte subsets exhibited significant differences in CSF and PB, possibly due to varying immune microenvironments in SAH.