From August 2021 to November 2021, 21 atherosclerotic patients undergoing carotid atherectomy at Beijing Tiantan Hospital were enrolled in this study after informed consent provided. The detailed clinical data of the enrolled patients are shown in Table 1 . Plaques were obtained from each patient. According to the imaging results (CTA/MRA/CU), plaque vulnerability was determined based on imaging biomarkers (Intraplaque haemorrhage, Lipid-rich necrotic core, Neovascularisation, Carotid plaque thickness, Surface morphology and Carotid plaque volume) (14). Normal superficial temporal arteries were obtained from other 5 patients undergoing craniocerebral trauma surgery for RNA-seq analysis (See Figure 2 for details). This research was approved by the Institutional Review Board (IRB) and Ethics Committee of Beijing Tiantan Hospital. Written informed consent was obtained from all patients.

Peripheral blood and plaque tissues from 3 patients were used for RNA-seq analysis. In addition, superficial temporal artery samples from 5 patients undergoing craniocerebral trauma surgery were used as blank controls. The plaque tissues and artery samples were washed with DPBS (Sigma-Aldrich, United States) within 1 hour after surgical resection. Then, TRIzol method was also adopted. Sequencing libraries were generated using rRNA-depleted RNA with an NEBNext Ultra Directional RNA Library Prep Kit for Illumina (NEB, MA) following the manufacturer’s recommendations. We then performed the paired-end sequencing on illumina NovaSeq 6000(illumina, USA)as recommended by the supplier. After cluster generation, the libraries were sequenced on the Illumina HiSeq platform, and 150-bp paired-end reads were generated.

In order to obtain clean data, raw data in fastq format were first processed through in-house Perl scripts. Reference genome and gene model annotation files were downloaded directly from the genome website. The index of the reference genome was built using bowtie2 v2.2.8, and paired-end clean reads were aligned to the reference genome using Hisat2 v2.0.5. Hg19 RefSeq (RNA sequences, GRCh38) was downloaded from the UCSC Genome Browser (http://genome.ucsc.edu). The clean reads were aligned with both genome hg19 and transcript reference using STAR v2.2.1, and gene expression was calculated by RSEM v1.3.0 using FPKM (fragments per kilobase of exon per million fragments mapped). We then compared RNA-seq in peripheral blood between healthy individuals and atherosclerotic patients. P<0.05 indicated statistically significant difference in expression. R was used for analysis of the gene expression data.

Atherosclerotic plaque tissues were washed with Dulbecco’s phosphate-buffered saline (DPBS, Sigma-Aldrich, United States) within 1h after surgery. Each specimen was digested at 37°C for 1h using miscible liquids that contain collagenase type IV (GIBCO, Gaithersburg, United States), DNase (Sigma, DN25), hyaluronidase (Sigma, H3506), collagenase type XI (Sigma, C7657) and collagenase type II (Sigma, C6885). The mixture was filtered through a 70μm cell strainers with DPBS and washed with red blood cell (RBC) lysis buffer (BD Biosciences, United States). The dissociated cell suspension was then washed once with DPBS and resuspended in 1 mL of staining buffer (DPBS containing 5% fetal bovine serum, ScienCell, United States).

A panel of 39 antibodies designed to distinguish a broad range of immune cells was used. Antibodies were either purchased in a preconjugated form from Fluidigm (South San Francisco, United States) or purchased from Biolegend (San Diego, United States) in a purified form and conjugated inhouse using the Maxpar X8 Multimetal Labeling Kit (Fluidigm, United States) according to the manufacturer’s instructions. The antibodies and reporter isotopes are included in Supplementary Table 1 . The samples were then washed and stained with cisplatin-195Pt (Fluidigm, 201064) as a viability dye. Cell samples were then washed and stained with cell surface antibodies for 30 min on ice. Subsequently, the samples were permeabilized at 4°C overnight and stained with intracellular antibodies for 30 min on ice. The antibody-labeled samples were washed and incubated in 0.125 nM intercalator-Ir (Fluidigm, United States) diluted in phosphate-buffered saline (PBS, Sigma-Aldrich, United States) containing 2% formaldehyde and stored at 4°C until mass cytometry examination. Before acquisition, the samples were washed with deionized water and then resuspended at a concentration of 1 x 10⁶ cells/mL in deionized water containing a 1:20 dilution of EQ Four Element Beads (Fluidigm, United States). The samples were then examined by CyTOF2 mass cytometry (Fluidigm, United States).

CyTOF data were acquired in the form of.fcs files from the CyTOF2 system. The addition of EQ Four Element Beads allowed us to use the MATLAB-based normalization technique. The obtained data were uploaded to Cytobank. Firstly, beads are filtered and active cells were selected from the specific gate. Then, CD45+ cells were gated (see Supplementary Figure 1 for details). Further analysis using the automated dimensionality reduction algorithm FlowSom by R language. The results were shown by viSNE, a visual dimensionality reduction algorithm.

Plaques from 9 patients (4 vulnerable plaques and 5 stable plaques) were fixed overnight in 4% formalin (4°C) and embedded in paraffin blocks for paraffin sections. Three paraffin sections (4 um) were cut from each specimen. Hematoxylin and eosin (H&E) staining was performed. For immunofluorescence, paraffin sections were washed twice 15 min in PBS (Sigma-Aldrich, United States), permeabilized in 0.2%-0.5% Triton X-100 (Solarbio, China) and blocked in 5% normal donkey serum (Jackson Lab, United States) for 1 h and stained with primary antibody overnight. Primary antibody was detected using fluorescent-conjected second antibodies (ZSGB-BIO, China). Primary antibodies were: anti-CD68 (Abcam, United States), anti-CD86 (Abcam, United States), and anti-HLA_DR (Abcam, United States). Sections were mounted with fluorescence mounting medium (Glostrup, Denmark). Fluorescent images were acquired on a Zeiss LSM880 NLO microscope and Zeiss Axio Scope Al was used to obtain H&E images. Three fields were randomly selected from each staining and the number of fusion particles of CD68, CD86 and HLA_DR in each field was counted. The Wilcoxon rank test was used and P<0.05 indicated statistically significant difference.

In order to explore the composition of immune microenvironment in plaques, RNA sequencing data from plaques of 3 atherosclerotic patients and data from superficial temporal artery of 5 patients undergoing craniocerebral trauma surgery were analyzed. A total of 16036 differentially expressed genes (defined as |log2FoldChange|>1 and FDR <0.05) were identified. Among them, 3656 genes were upregulated in the atherosclerosis plaques ( Figure 1A ). To elucidate the functional implications of the differentially expressed genes, we performed pathway enrichment analyses of the upregulated and downregulated genes. For genes upregulated in plaques, NF-kappaB signaling and B cell activation were found to be significantly by Gene Ontology analysis and T cell receptor signaling pathway and B cell receptor signaling pathway were found to be significantly by Kyoto Encyclopedia of Genes and Genomes analysis. These results suggested immune responses involved T lymphocyte and B lymphocyte were existed in plaques ( Figure 1C ). Some of the up-regulated genes in plaques were also enriched in the biological processes associated with TNF signaling pathway. This suggested the TNF-a mediated immune responses within the plaques ( Figure 1D ) (Complete GO and KEEG enrichment data were shown in Supplementary Tables 2 and 3 ).

Finally, based on these RNA-Seq data, we used xCell (R package, Aran Dvir, 2017) to infer the type and proportion of cells contained in atherosclerotic plaques. The results showed that endothelial cells, smooth muscle cells and fibrous cells were composed of superficial temporal artery ( Figure 1E ). In addition to these cells, atherosclerotic plaques also contain a large number of macrophages, dendritic cells, T and B lymphocytes ( Figure 1E ). The results of immune infiltration analysis provided guidance for the subsequent selection of CyTOF antibody profiles.

According to the imaging (CTA/MRA/UTA) results, we divided the enrolled specimens into 12 stable plaques and 9 vulnerable plaques (Typical H&E images of stable and vulnerable plaques were shown in Supplementary Figure 2 ). Specific metal antibodies were selected on the basis of RNA-seq analysis (Detailed information of antibodies panel was shown in Supplementary Table 1 ). Flowsom, an unsupervised clustering method, was used to analyze the cellular composition of atherosclerotic plaques from 21 patients ( Figure 2 ). Finally, 15 clusters were found. Each cell type was identified by specific markers on the its surface ( Figure 3A ). We found that macrophages and lymphocytes were the main components of plaques ( Figure 3B ). Macrophages were composed of 6 clusters, accounting for 63.9%. Lymphocytes consisted of T lymphocytes and B lymphocytes. While 7 clusters made up T lymphocytes, accounting for 31.9%. B lymphocytes were composed of two clusters, accounting for 4.2% ( Figure 3B ). T-SNE was employed to convert high-dimensional CyTOF data and was used to map the immune compartments of all samples.

The composition of the two plaques was compared to explore differences in immune composition between stable and vulnerable plaques. Overall, each cell subset was distributed in both types of plaques. T cells were predominant in stable plaques, accounting for 57.2%. Macrophages accounted for 39.1%. However, macrophages were the most abundant cell group in vulnerable plaques, accounting for 56.4%. T cells only accounted for 36.7% ( Figure 4C ). These seemed to suggest differences in immune environments between the two plaques.

The type of T cells in plaques was diverse. CD4+T cells dominate the plaques. There were 5 clusters of CD4+T lymphocytes, including memory cells and effector cells. Comparing the percentage of cell subsets in the two plaques, CD4+T cells of cluster10 were the only group of cells more abundant in stable plaques (p<0.05) ( Figures 4B, D ). CD45RO+ and CD127, surface markers of memory cells, co-exist on the surface of cells from this cluster (15, 16). There was no difference in the content of other subsets between the two types of plaque. However, Cluster2 (CD11c+CD4+T cells) in vulnerable plaques secreted a higher level of IFN-r than in stable plaques (p< 0.05) ( Figure 4E-1). It was suggested that this group of cells in vulnerable plaques might also play a unique function.

Two types of B cells existed in samples. These B cells have different surface molecular patterns. Cluster9 expressed CD19+, whereas Cluster8 expressed CD20 but not CD19 ( Figure 3D ). Cluster9 also expressed surface markers HLA_DR and CCR7, which were involved in cell activation and migration, suggesting that this cluster may be a group of activated B subsets cells (17, 18). There was no difference in the number of these clusters between the two types of plaque. Functional analysis showed that cluster8 expressed a higher level of PD-1 molecule in vulnerable plaques (p<0.05) ( Figure 4E-2 ). Cluster9 showed a higher level of TNF-a and IL-6 (p<0.05) ( Figure 4E-3 and 4E-4 ).

Macrophages were the most abundant cluster in plaques and contained multiple cell subsets ( Figure 4C ). However, compared with stable plaques, cluster14 was the only cluster that was more abundant in vulnerable plaques ( Figures 4B, D ). These clusters expressed CD68 and HLA_DR, typical macrophage surface markers (19). In addition, CD86, corresponding to M1-like macrophages, existed on the surface of cluster14 (20, 21). Polychromatic immunofluorescence showed that there were more M1 macrophages in vulnerable plaques ( Figures 5A, B ). Functional analysis showed that this cluster also expressed IL-6 and IFN-r (pro-inflammatory cytokines), and functional analysis did not reveal a significant difference ( Figure 4E-4 ).