To induce AAA, 8-week-old ApoE^−/− male mice (C57BL/6J, Beijing Vital River Laboratory Animal Technology Co., Ltd. Beijing) were anesthetized with 200 mg/kg ketamine and 10 mg/kg xylazine by intraperitoneal injection, and infused with 1,000 ng/kg/min Ang II (A9525-50MG sigma-Aldrich) or saline with alzet osmotic mini-pumps (Alzet model 2004, DURECT Corp., Cupertino, CA) for 28 days. To evaluate the effect of fibrocytes, 8 × 10⁶ fibrocytes were injected into mice via tail vein on days 7 and 21 after Ang II-infusion (1.6 × 10⁷ in total). GFP-labeled fibrocytes were injected with the same protocol to trace them in vivo. All animal experiments were approved by the Research Ethics Committee of Peking Union Medical College.

A single-cell cDNA library was generated as previously described (15). For the isolation of individual cells, both control and AAA mice were anesthetized with ketamine (200 mg/kg) and xylazine (10 mg/kg) by intraperitoneal injection. The normal aorta (one whole aorta from one mouse) or aneurysm tissue (one piece of the supra-renal part with aneurysm from one AAA mouse) were dissected and digested for 15 min at 37°C in PBS containing 200 U/ml collagenase I (SCR103, Sigma Aldrich, Germany), 0.05 U/ml elastase (E1250, Sigma Aldrich, Germany), 5 U/ml neutral protease (LS02111, Worthington, USA), and 0.3 U/ml deoxyribonuclease I (M6101, Promega, USA). The tissue digestion was stopped with DMEM containing 10% FBS (10099141, Gibco, USA.) and the mixture was filtered through a 40 μm cell strainer (15-1040, Biologix, China) to obtain single-cell suspensions. Using the method of 7-amino-actinomycin D (7-AAD) positive cells staining, living aortic cells (7-AAD negative cells) were sorted using a Moflo-XDP (Beckman, USA). Cell suspensions (~10,000 single cells) were next loaded on the Chromium Single Cell Controller (10X Genomics) to generate a single cell and gel bead emulsion (GEM). Single-cell sequencing library preparation was performed according to the instructions of Chromium single cell 3' library & Gel bead kit v2 (10X Genomics). Libraries were sequenced by Illumina Hiseq X Ten in paired-end to reach ~50,000 reads per single-cell (Novo Generation Bioinformatics Technology Co., Ltd.).

The 10X Genomics single-cell transcriptome sequencing data were filtered by removing bases with a mass less than 3 at the beginning and end of the reads, using a CellRanger software suite version 3.0.2 pipeline (16). The filtered reads were aligned to the MM10 mouse reference genome by STAR. For further analysis and statistics, based on the barcode and gene expression matrix, single-cell data were log-normalized and filtered by the R 3.6.0 package Seurat 3.2.3 (17, 18), with the following parameters: unique gene count per cell >500, cell counts per gene >3 (0.1% of the total cell amount), and percentage of mitochondrial genes <0.05. After the control and AAA samples were combined, the UMAP and automated cluster detection algorithms were performed stepwise. The resolution for cluster identification was set as 0.5. DEGs were identified by the Wilcox test with default parameters and p <0.05. DEGs from each cluster were sorted by log_e-fold change relative to the other clusters. Then, the DEGs of which log-fold change was more than 0 were selected and used for GO analysis by ClusterProfiler 3.14.3 (https://guangchuangyu.github.io/software/clusterProfiler/). The enriched function or pathway were further clustered to predict the most likely function of the specific cluster. Alternatively, to detect the possible polarization or differentiation process among different clusters, Monocle 2.14.0 (19) was used for the trajectory analysis. The co-expression level of n genes was calculated by e1×e2×⋯×enn (e_n means the expression of the nth gene) and the accuracy is confirmed by ROC curve. Velocyto.R v0.6 package were then used for calculating RNA velocity of each single cells and further visualization.

For flow cytometry analyses, single-cell suspensions of mouse aortae were prepared according to the previous protocol (20) and stained with following antibodies: 7AAD (00-6993-50, eBioscience, USA), Fluorescent isothiocyanate (FITC) anti-mouse-α-SMA (ab8211, Abcam, United Kingdom), PE-anti-mouse-CD34 (551387, BD, USA), FITC-anti-mouse-CD68 (MA1-82739, Invitrogen, USA), PE-anti-mouse-NK1.1 (ab269324, Abcam, United Kingdom), FITC-anti-mouse-CD19 (11-0193-82, eBioscience, USA), PE-anti-mouse-CD21 (12-0211-82, eBioscience, USA), PE-anti-mouse-CD45 (12-0451-82, eBioscience, USA), and Collagen I (ab88147, Abcam, United Kingdom; at room temperature for 30 min) with Alexa Fluor 488 goat anti-mouse antibody (A32723, Invitrogen, at room temperature for 30 min). Flow cytometry results were analyzed by using FlowJo 7.6 software (FlowJo, LLC., Ashland, Oregon). Statistical analysis was further performed by GraphPad Prism 8.0.

Bone marrow-derived fibrocytes were obtained by in vitro differentiation of spleen cells as previously described (21). Spleen cells were isolated from 8-week ApoE^−/− or generic GFP male mice as previously described (22). Briefly, spleens were separated, minced and digested at 37°C. Spleen cells were isolated using a 70 μm cell strainer (15-1070, Biologix, China) and cultured in 3 ml of RPMI-1640 (31800-500, Solarbio, China) with L-Glutamine, 20% fetal bovine serum (10099-141, Gibco, Australia), penicillin–streptomycin (P1400, Solarbio, China). After 3 days, the medium was changed into Fibrolife basal media (LM-0001, Lifeline Cell Technology, USA) with 25 ng/ml of murine macrophage colony-stimulating factor (M-CSF) (315-02, PeproTech, USA) and 50 ng/ml of murine interleukin (IL)-13 (210-13, PeproTech, USA) and cells were cultured in a humidified incubator containing 5% CO₂ at 37°C for 10 days. To detect whether cells were induced differentiation, we observed the morphological changes under microscopy (Nikon, Tokyo, Japan). Following 10 to 14 days, the adherent cells were harvested and used to injected to mice and immunofluorescence staining.

Human aortic tissue extracts were prepared from the abdominal aortic aneurysm patient (Peking Union Medical College Hospital and Beijing Anzhen Hospital) and from the body of the deceased donor with no detectable vascular disease (Peking Union Medical College Volunteer Corpse Donation Reception Station). Age and gender were matched (Supplementary Tables S1, S2). The human donor aortic tissues obtained were approved by the institutional review board of institute of basic medical sciences, Chinese academy of medical sciences. The abdominal aortic aneurysm tissue obtained were approved by the institutional review board of Peking Union Medical College Hospital and Beijing Anzhen Hospital.

The mouse aortae were embedded in optical cutting temperature (OCT) compound (4583, Sakura, Netherlands) and the frozen sections were used for H&E staining, Elastin staining or immunofluorescence staining. H&E staining and Elastin staining were performed by using H&E staining kit (G1120, Solarbio, China) elastic staining Kit (HT25A, Sigma-aldrich, Germany) according to the manufactures' instructions.

For immunofluorescence staining, the sections for mouse aortae and human aortae or cells seed in 35 mm glass bottom dish were fixed with 4% paraformaldehyde for 10 min and permeabilized with 1% Triton X-100 for 15 min. Then the sections blocked with 1% bovine serum albumin for 30 min at room temperature and incubated with mouse monoclonal collagen I (ab88147, Abcam, United Kingdom, 1:100 dilution) and rabbit polyclonal CD45 (ab10558, Abcam, United Kingdom, 1:100 dilution) antibodies overnight at 4°C. After washing with PBS three times, sections were incubated Alexa Fluor 488-conjugated goat anti-mouse IgG (H+L) secondary antibody (A-11008, Invitrogen, USA, 1:1,000), Alexa Fluor 594 conjugated goat anti-rabbit IgG(H+L) highly cross-adsorbed secondary antibody (A-11037, Invitrogen, USA, 1:1,000 dilution) for 40 min at room temperature. Finally, sections were mounted with ProLong Gold antifade reagent with DAPI (ab104139, Abcam, United Kingdom). Pictures were taken by confocal microscopy (LSM780, Zeiss).

For animal experiments, the exact number of mice used in each experiment is reported in the figure legends. Normality tests were conducted first. Then, comparisons between two groups were made with an unpaired two-tailed t-test or non-parametric test. Comparisons among groups with two factors were made with two-way ANOVA, and p < 0.05 was considered statistically significant. Data were presented as mean ± standard error of the mean (SEM).

To explore the cell types involved in AAA, apolipoprotein E–deficient (ApoE^−/−) male mice were implanted with Ang II osmotic pump for 4 weeks to induce AAA or underwent sham surgery, and the tissues were collected for scRNA-seq (Figure 1A). After alignment (Supplementary Figure S1A) and quality control (Supplementary Figures S1B,C), 7,914 and 9,338 cells from control and AAA groups were combined and included in the subsequent analysis. Unsupervised clustering by Seurat R package revealed 15 distinct clusters (Figure 1B). In accordance with the canonical markers, the clusters were fibroblasts (Cluster 0, 2, 3; Cd34⁺Pdgfra⁺), macrophages [Cluster 1, 4; Cd68⁺Ptprc⁺ (encoding Cd45)], SMC [Cluster 5; Acta2⁺ (encoding α-SMA) Cd34⁻Pdgfra⁻], endothelial cells (ECs) [Cluster 6; Pecam1⁺ (encoding Cd31), T cells (Cluster 7; Cd3e⁺Cd8a⁺), B cells (Cluster 8; Cd19⁺), lymphatic endothelial cells (Cluster 12; Lyve1⁺Pecam1⁺), and proliferating cells (Cluster 9; Top2a⁺Mki67⁺)] (Figure 1C). Consistent with previous studies, among all cell types, the proportion of immune cells significantly increased in the AAA group, especially macrophages and B cells (Figures 1D,E). By contrast, fibroblasts and SMCs were reduced after 4-week of Ang II-exposure (Figures 1D,E).

To assess the robustness of the dataset and explore the cellular mechanisms of AAA formation, differentially expressed gene (DEG) were determined by comparing each cell type between AAA and control groups and Gene Ontology (GO) analyses were performed. We found that the functions of “positive regulation of cytokine production,” “regulation of inflammatory response,” and “antigen processing and presentation” (Figure 1F; Supplementary Figure S2A) was upregulated in macrophages in AAA. Moreover, B cells in this dataset were predicted to be activated to plasma cells with the high expression level of Cr2 (encoding Cd21) and Ighm, and the upregulated genes of B cells were mainly enriched in the function of “immunoglobulin mediated immune response.” As for T cells, “natural killer cell activation” was upregulated with increased expression of Klrb1c and Il2rb in AAA pathogenesis. To validate the results from scRNA-seq analyses, the cells were analyzed by flow cytometry, which confirmed the infiltration of macrophages, and the activation of B cells and NK T cells in AAA (Supplementary Figures S3A–C).

Fibroblasts were composed of three clusters (Clusters 0, 2 and 3). The highly expressed genes (HEGs) among these clusters and potential functions of these cells were analyzed. The typical fibroblast functions such as “extracellular matrix organization” and “extracellular structure organization” were enriched in all of the three clusters. HEGs of Cluster 0 were mainly enriched in the function of “transmembrane receptor protein serine/threonine kinase signaling pathway” (Supplementary Figure S2B). By contrast, the specifically enriched GO functions of Clusters 2 and 3 were “connective tissue development” and “wound healing” (Supplementary Figures S2C,D). Further analysis based on the gene expression patterns determined Cluster 0 as the inactivated fibroblasts, with higher expression of Cd34 but no expression of Acta2 (Supplementary Figure S2E). By contrast, Cluster 3, with higher expression level of Acta2 and lower expression level of Cd34 compared with Cluster 0, was regarded as myofibroblasts (11). Cluster 2 could be the intermediate types between Cluster 0 and 3. In addition, scRNA-seq data showed that the proportion of Acta2⁺ myofibroblasts increased significantly (Figure 1G) in AAA, which was validated by flow cytometry analysis (Supplementary Figure S3D).

In terms of vascular cells, SMCs involved in the function of “regulation of apoptotic signaling pathway” were upregulated in AAA, which was consistent with the increased SMC death in AAA (Supplementary Figure S2E). Furthermore, ECM-related functions were enriched in SMCs in the AAA group, suggesting that the SMCs underwent phenotypic transformation in AAA (Supplementary Figure S2E). In addition, GO analysis of DEGs of ECs suggested that the upregulated functions were related to leukocyte migration and EC proliferation (Supplementary Figure S2F), which was consistent with the previous reported function of ECs in AAA formation.

As the largest and most varied population in our scRNA-seq data, we focused on macrophages (Clusters 1 and 4) and re-clustered them by Seurat (Figures 1C–E). We observed four main clusters (Clusters 0–3) (Figure 2A). Using canonical markers and HEGs to analyze the macrophage subtypes, we identified Trem2⁺Acp5⁺ osteoclast-like macrophages (Cluster 0), Mrc1⁺Cd163⁺ M2-like macrophages (Cluster 1), Il1b⁺Ccr2⁺ M1-like macrophages (Cluster 2), and Cd34⁺Col1a2⁺ bone marrow-derived fibrocyte (21, 23) (Cluster 3) (Figure 2A; Supplementary Figure S4A). To define the specific gene expression features and potential functions, HEG and GO analyses were performed on three subtypes of macrophages (Figure 2B). Osteoclast-like macrophages (Cluster 0) expressing high levels of Trem2, Tyrobp, Plin2, and Lpl were enriched in the function of “apoptotic cell clearance” and “lipid storage,” suggesting that osteoclast-like macrophages could be involved in lipid deposition and the phagocytosis of apoptotic cells or lipid. In M2-like macrophages (Cluster 1), C-C motif chemokine (CCL) genes (Ccl8, Ccl6, Ccl24, Ccl9 and Ccl12) were highly expressed, and their enriched function was “mononuclear cell migration,” implying that M2-like macrophages might contribute to the recruitment of mononuclear cells by producing chemokines in AAA pathogenesis. Furthermore, the enriched function of “tissue remodeling” related genes were also highly expressed in M2-like macrophages, which is consistent with the previous reported phenotypic characteristics of M2 macrophages (24). In M1-like macrophages (Cluster 2), HEGs were mainly enriched in the function of “regulation of cytokine biosynthetic process” and “antigen processing and presentation.” Combining with the high expression of Il1b and Ccr2, these cells have been predicted to be the most important pro-inflammatory cells in AAA pathogenesis.

To further analyze the changes of macrophage subtypes between control and AAA groups, DEG and GO analyses were performed on each macrophage subtype. In osteoclast-like macrophages, the genes upregulated in AAA were enriched in the function of “NADH dehydrogenase complex assembly,” “ATP metabolic process,” and “glycolytic process” (Figure 2C; Supplementary Figure S4B). In M2-like macrophages, the function of “positive regulation of cytokine production” and “tissue remodeling” were upregulated (Figure 2C). On the other hand, upregulated pro-inflammatory factors, such as Il1b and Thbs1, were enriched in the function of “positive regulation of cytokine production” in M1-like macrophages. In addition, the expression of genes related to “apoptotic cell clearance” were also increased in M1-like macrophages. More interestingly, we found that the mode of ATP metabolism in osteoclast-like macrophages might be shifted to glycolysis (Supplementary Figure S4B), which is typical of M1 macrophages (25). Moreover, the upregulated functions of M2-like macrophages were similar to the signature functions of M1-like macrophages. These data together suggest a connection between the three subtypes, which was clarified by trajectory analysis with the Monocle2 R package. As shown in Figure 2D, we found the tendency of transition among different macrophages. As shown by the RNA velocity, we found that osteoclast-like macrophages and M2-like macrophages tended to polarize to M1-like (Supplementary Figures S5A,B). Key genes related to the macrophage polarization were significantly changed among these three subtypes of macrophages. For example, Klf4 (26) was highly expressed in M2 compared to M1 cells and was reported to play a crucial role in macrophage polarization (Figure 2E). Deficiency of Klf4 could evoke a partial loss of M2 but gain of M1. Moreover, the expression of Cd36 declined from M2 to M1 (Figure 2E), which was proven to be essential for fatty acid uptake and metabolism of M2 macrophages (27). By contrast, the expression of Pkm and Pgam1 related to glycolysis and gluconeogenesis, increased during the M2 to M1 transition (Figure 2E). The high levels of Pkm and Pgam1 in M1 macrophages indicate that M1 macrophages are glycolytic (25), which was consistent with our GO analysis above (Figure 2C). Taken together, we identified three subtypes of macrophages in aortic and aneurysmal tissues and we thus proposed that the osteoclast-like and M2-like macrophages might polarize toward M1 type in AAA pathogenesis.

While analyzing the subtypes of macrophages, a remarkable population of Cd34⁺Col1a2⁺ bone marrow-derived fibrocytes (Cluster 3) were found in control aorta and AAA tissues, and tended to increase in AAA (Figure 2A; Supplementary Figures S6A,B). When performing HEG analysis on fibrocytes compared with the three subtypes of macrophages, we found that the expression pattern of fibrocytes was exceedingly distinct from macrophages (Figure 3A). ECM-related genes were highly expressed in fibrocytes, such as collagen (Col1a1, Col1a2, Col3a1, Col5a, Col4a1, etc.), elastin (Eln), matrix metalloproteinases (Mmp2, Mmp3, etc.), and periostin (Postn) (Figure 3A). GO analysis of HEGs suggested fibrocyte functions such as ECM organization, collagen related process and response to growth factor and wounding (Figure 3B), similar to the functions of fibroblasts (Supplementary Figures S2B–D). Further trajectory analysis of fibrocytes and macrophages revealed that there might be no connection or transition between the fibrocytes and three subtypes of macrophages (Figure 3C). Taken together, we identified fibrocytes in AAA tissues, whose gene expression patterns and functions were different from those of macrophages.

Bone marrow–derived fibrocytes are differentiated from circulating CD14⁺ peripheral blood mononuclear cells (PBMCs) and co-express markers of hematopoietic stem cells (CD34⁺), monocyte lineage (CD11b⁺), leukocyte (CD45⁺), and stromal cells (collagen I⁺). These characterizations are used to identify fibrocytes (23, 28). In our scRNA-seq analysis, marker genes distinguishing fibrocytes from macrophages were retrieved and combined with the canonical markers to identify fibrocytes from other cell types. Among all cell types, the fibrocyte features distinct from macrophages were remarkably similar with fibroblasts, but the fibrocytes also exhibited leukocyte-like features. This suggests that fibroblasts represent a unique cell type (Figure 4A). For screening the markers of fibrocytes, we calculated the co-expression levels (details in Methods) of the fibroblast features (i.e., Cd34, Col1a2, Meg3, Aebp1, Dcn, Igfbp7, and Nbl1) and the leukocyte-like features (Cd45) and assessed the accuracy and specificity of different feature combinations by ROC curve. The result showed that the combination of Cd45 and Col1a2 showed a high specificity (>95%) for identifying fibrocytes with a co-expression level of more than 0.25 (Figure 4B). Fibrocytes were newly identified by co-expression of Ptprc and Col1a2, and distributed among macrophage, fibroblast and smooth muscle cell clusters (Supplementary Figures S6C,D). Furthermore, we adopted and analyzed a previous single cell RNA-seq dataset from human ascending thoracic aortic aneurysm (29) (ATAA) to validate our current findings in mice. By calculating the co-expression level of PTPRC and COL1A2, fibrocytes were also detected in the ATAA dataset. Fibrocytes were mainly distributed among macrophage, fibroblast and smooth muscle cell clusters in the embedding space (Supplementary Figures S7A,B), which is consistent with our results. In addition, the proportions of fibrocytes tended to increase in ATAA patients (Supplementary Figure S7C). Next, the transcriptomic changes of fibrocytes in AAA were analyzed. In terms of the classic markers of fibrocytes, the hematopoietic stem cell- and immune cell-like features (i.e., Cd34, Ptprc, Itgam) were decreased but the stromal feature (Col1a2) was increased (Figure 4C). DEG analysis of our data showed the upregulation of Ccl8, Gm2564, Aif1, and Lgals3 in fibrocytes, which were enriched in the function of “monocyte chemotaxis.” The expression levels of Timp1, Tyrobp, Mif, Col1a1, and C3 were increased, and they were enriched in the function of “response to wounding” (Figures 4D,E). These data together suggest the response and recruitment of fibrocytes to the AAA lesion. In addition, ECM and collagen-related functions were upregulated in fibrocytes during AAA formation (Figure 4E), which were similar to fibroblast functions (Supplementary Figures S2B–D). Further combining with the ATAA single cell dataset, DEGs of fibrocytes caused by aortic aneurysms were overlapped with those in our dataset and determined. Specifically, Timp1, Sfrp2, and Sfrp4 were upregulated in both human and mouse samples (Figure 4F). These genes are known to be closely related to AAA formation (30–32). Further trajectory analysis between fibroblasts and fibrocytes showed fibrocytes were similar with inactivated fibroblasts and might be differentiated toward myofibroblasts (Supplementary Figure S5D), consistent with their role as the fibroblast progenitor. Taken together, we hypothesized that fibrocytes are recruited to AAA tissues and tend to transform toward fibroblasts to influence AAA pathogenesis by mediating ECMs remodeling.

In order to validate the existence and localization of fibrocytes, co-immunofluorescence staining was performed and the results showed that numerous CD45⁺COL1⁺ fibrocytes were significantly increased in both mouse (Figures 5A,B) and human (Figures 5C,D) AAA tissues, suggestive of fibrocytes mainly in the adventitia. Co-localization of CD45 and elastin layers showed that CD45⁺ cells were mainly localized in the adventitia (Supplementary Figure S8), further indicating the localization of fibrocytes in the adventitia. Flow cytometry data further confirmed the increase of fibrocytes in mouse AAA tissues (Supplementary Figure S9).

Fibrocytes have been reported to play essential roles in wound healing, atherogenesis, and lung diseases (21, 23, 33, 34), but their role in AAA is not understood. To test the hypothesized roles of fibrocytes in AAA formation, bone marrow-derived fibrocytes were administered into Ang II-treated mice on days 7 and 21 by tail-vein injection (Figure 6A). Bone marrow-derived fibrocytes were produced from spleen monocytes and cultured with IL-13 (50 ng/ml) and M-CSF (25 ng/ml) for 10 days (Supplementary Figure S10). GFP-labeled fibrocytes from GFP-transgenic mice were then injected into Ang II-infused mice (non-GFP) to trace the cells in vivo. By co-staining GFP with fibrocyte markers, GFP-labeled fibrocytes were found to be recruited to AAA tissues (Supplementary Figure S11). More importantly, reconstitution of fibrocytes significantly attenuated AAA formation, demonstrated by a decrease in AAA incidence and mortality (Figures 6B,C), reduced lesion diameters (Figures 6D–F) and diminished elastin degradation (Figures 6G,H). Taken together, our data suggest that fibrocytes attenuate Ang II-induced AAA formation.