Intracellular Osteopontin of Macrophages Promotes Carotid Plaques Formation by Inducing Foam Cells and Releasing Proinflammatory Cytokines

Abstract

Background

Atherosclerotic carotid plaque is the main cause of cerebrovascular diseases, but the molecular and cellular mechanisms involving in carotid plaque formation have not yet been fully uncovered.

Methods

Single cell RNA sequencing of human carotid plaques was used to identify major cell types. Cell‐cell communication and machine learning analyses were used to determine the key cell cluster involved in plaque development. In vitro experiments were conducted to explore the role of SPP1 (osteopontin) in foam cells formation. Clinical and histological analyses were also performed to validate the involvement of osteopontin in atherosclerosis.

Results

Eight distinguishable major cell types were identified in atherosclerotic carotid tissue. Cell‐cell communication and machine learning analyses identified SPP1 ^hi macrophage as the key cell cluster in plaque development, and functional annotation further highlighted the involvement of hypoxia response and cholesterol metabolism. Culturing mouse macrophages under 1% O₂ hypoxia condition with the addition of oxidized low‐density lipoprotein (ox‐LDL) significantly induced membrane translocation of CD36 and SR‐A, enhanced the expression of SPP1 (osteopontin), promoted foam cells formation, and activated NF‐κB pathway to release proinflammatory cytokines. All these changes of macrophages under hypoxia and ox‐LDL stimulations were effectively mitigated by inhibiting intracellular osteopontin. Clinical data and histological analysis of surgical resected carotid plaque tissue samples also confirmed the role of osteopontin in atherosclerosis.

Conclusions

Intracellular osteopontin of macrophages promotes plaque formation by inducing foam cells and releasing proinflammatory cytokines. Our data provide novel mechanistical insight and therapeutic strategy on atherosclerotic carotid plaque.

Nonstandard Abbreviations and Acronyms

ACP

atherosclerotic carotid plaques

CM

conditioned medium

iOPN

intracellular osteopontin

IPH

intraplaque hemorrhage

OPN

osteopontin

ox‐LDL

oxidized low‐density lipoprotein

scRNA‐seq

single‐cell RNA sequencing

sOPN

secreted osteopontin

VSMCs

vascular smooth muscle cells

Research Perspective.

What Is New?

• SPP1 ^hi macrophages are identified as a distinct subpopulation involved in the development of atherosclerotic carotid plaques.

• Upregulated intracellular osteopontin in macrophages under hypoxia and oxidized low‐density lipoprotein stimulations can directly promote foam cell formation through cytoskeleton and membrane structure rearrangement and indirectly induce the proliferation and phenotypic switching of vascular smooth muscle cells by releasing proinflammatory cytokines.

What Question Should Be Addressed Next?

• How does intracellular osteopontin modulate cytoskeletal organization, lipid protein distribution, and lipid metabolism?

• Does targeting intracellular osteopontin effectively mitigate atherosclerotic carotid plaque formation?

Atherosclerosis is a chronic inflammatory disease driven by dysregulated lipid metabolism. ¹ Activation of endothelial cells, along with internalization and deposition of lipids in the intima, is considered the initial step of atherosclerosis. Subsequently, modified low‐density lipoprotein (LDL) enhances a series of inflammatory responses, primarily involving the activation and recruitment of macrophages and T cells. Moreover, stimulated vascular smooth muscle cells (VSMCs) undergo complex transitions from the media to the intima, interacting with foam cells to compose fibrous cap of plaques. ² Atherosclerosis is initially classified into 4 stages by the World Health Organization: fatty streak, atheroma, fibrous plaque, and complicated lesion. ³ The American Heart Association recommends a new morphological classification based on 6 types of lesions, representing the progression of atherosclerotic lesions, ⁴ in which Intimal xanthoma or fatty streak is classified as the second level, and fibrous plaque belongs to the highest level, ⁴ indicating that not all cases of atherosclerosis will develop into plaques.

Most cardiovascular and cerebrovascular diseases are caused by atherosclerosis due to plaque rupture and blood vessels blockage. This implies that there are no noticeable clinical symptoms in atherosclerosis without plaque. Multiple postmortem examinations and imaging studies have also confirmed that many young‐ or middle‐aged individuals have asymptomatic or subclinical atherosclerotic lesions. ⁵ , ⁶ Therefore, exploring the cellular and biological processes of plaque formation is significant for the prevention and treatment of atherosclerotic disease.

OPN (osteopontin) is encoded by SPP1 gene and classified into secreted type (sOPN) and intracellular type (iOPN). As a secreted protein, sOPN has been already demonstrated to mediate inflammation activation, cell proliferation, and cell adhesion by binding to integrins and CD44. ⁷ , ⁸ , ⁹ In contrast, the biological function of iOPN has recently been in the spotlight. Previous studies have reported that iOPN is associated with cell motility, cytoskeletal rearrangement, and mitosis. ¹⁰ , ¹¹ , ¹² However, the precise role of OPN in atherosclerosis remains unclear.

In this study, single‐cell transcriptomic data combined with machine learning analyses identified the SPP1 ^hi macrophages as the key cell cluster in atherosclerotic carotid plaques. Our in vitro experiments further confirmed that macrophages under hypoxia and oxidized LDL (ox‐LDL) stimulations enhanced the expression of iOPN to promote foam cell formation and proinflammatory cytokine release, suggesting the critical role of iOPN in the development of atherosclerotic carotid plaques.

METHODS

This study uses data from publicly available versions of the Gene Expression Omnibus database, Decode database, and GTEx Portal database. All data and code that support the findings of this study are available from the corresponding author upon reasonable request.

Data Collection and Analysis

Single‐cell RNA sequencing (scRNA‐seq) data using Chromium instrument (10x Genomics) were obtained from 6 patients undergoing carotid endarterectomy. These samples were dissected into proximal adjacent portions of the carotid artery and atherosclerotic carotid plaques (ACP). The original single‐cell data were obtained from GSE155512 and GSE159677. ¹³ , ¹⁴ Bulk transcriptome data of human carotid atherosclerotic plaques and carotid atherosclerotic artery were obtained from GSE28829, GSE163154, and GSE100927. ¹⁵ , ¹⁶ , ¹⁷ Detailed information is presented in Table 1.

Table 1.

RNA Sequence Data Information

Platform	Samples	Type
GSE155512	Atherosclerotic carotid arteries 3	scRNA
GSE159677	Atherosclerotic carotid plaques 3; proximal adjacent portions of carotid artery 3	scRNA
GSE28829	Carotid plaque 29 (advanced 16; early 13)	Bulk RNA
GSE163154	Carotid plaque 43 (non‐IPH 16; IPH 27)	Bulk RNA
GSE100927	Carotid atherosclerotic artery:29; carotid artery control:12	Bulk RNA

IPH indicates intraplaque hemorrhage; and scRNA, single‐cell RNA sequencing.

We imported scRNA‐seq data into the R environment (version 4.1.2) and created objects to analyze the data by using the “Seurat” package. ¹⁸ Quality control on gene expression for each cell was performed by retaining cells with gene expression >200 and <5000 and removing cells with mitochondrial percentages >20%. We used RunHarmony function with default parameters in the “Harmony” package ¹⁹ and the top 20 principal components to integrate multiple scRNA‐seq data sets and eliminate batch effects. Top 2000 highly variable genes were selected for downstream analysis. Principal component analysis was used to map high‐dimensional single‐cell data to a 2‐dimensional space and then the “clustering trees ²⁰ “package was used to better determine resolution. Cell clusters were visualized using t‐distributed stochastic neighbor embedding and manually annotating them into distinguishable cell types based on Cellmark database and published studies. ¹³ , ²¹ , ²²

FindAllMarkers and FindMarkers functions in Seurat were used to analyze expression between different cell types and tissues, with parameter settings for a log‐fold threshold of 1 (avg_log2FC>1). Using the clusterProfiler package, ²³ we performed Gene Ontology and Kyoto Encyclopedia of Genes and Genomes analysis on differentially expressed genes. Hallmark pathways for cell clusters were analyzed using the GSEA package. Cell–cell communication analysis was performed using the “cellchat” package. ²⁴ All parameters were set according to the protocol.

Cell Culture and Treatments

J774.1 mouse macrophages (RIKEN BRC CELL BANK, Tsukuba, Ibaraki, Japan) were cultured in RPMI‐1640 medium (Fujifilm Wako, Japan) supplemented with 10% FBS (Corning, USA) and 1% penicillin–streptomycin (Fujifilm Wako, Japan). JCRB0150 mouse VSMCs (JCRB Cell Bank, Osaka, Japan) were cultured in DMEM medium (Fujifilm Wako, Japan) supplemented with 10% FBS and 1% penicillin–streptomycin. All cells were maintained at 37 °C in a humidified incubator with 5% CO₂.

For hypoxia treatment, macrophages were placed in the incubator (Sanyo, Japan) containing 1% O₂ and 94% N₂. To induce foam cell formation, macrophages were treated with 50 μg/mL ox‐LDL (Athens Research And Technology, GA, USA).

Lipofectamine 3000 (Thermo Fisher Scientific, Japan) and the Silencer Select siRNA system (s74323; Thermo Fisher Scientific, Japan) were used for SPP1 knockdown in macrophages. As a control, cells were treated with Silencer Negative Control siRNA (AM4611; Thermo Fisher Scientific, Japan). Antiosteopontin antibody (2 μg/mL, AF808; R&D Systems) was added to the culture medium for OPN neutralization, and normal goat IgG (2 μg/mL; AB‐108‐C; R&D Systems) was used as a negative control. After 48 hours, cells were collected for analysis.

To evaluate the paracrine effect of macrophages on phenotypic switching of VSMCs, we collected the conditioned medium (CM) of macrophages (2.5×10⁵ cells/mL) at 48 hours after culture under different conditions. A mixture of CM and fresh DMEM medium (1:1 ratio) was used to culture VSMCs.

Oil Red O Staining

According to a previous report, ²⁵ macrophages were fixed with a 4% formaldehyde solution. After washing, cells were incubated with Oil Red O staining solution for 15 minutes. Stained cells were imaged using the microscope (Biorevo BZ‐9000; Keyence).

Western Blot Analysis

Cells were collected and washed twice with PBS, then incubated with RIPA lysis buffer at 4 °C for 30 minutes. Protein concentration was determined using the BCA protein assay kit (Thermo‐Scientific). CM of macrophages (2.5×10⁵ cells/mL) were collected at 48 hours after culture under different conditions. Ponceau S staining was used to assess total protein levels in conditional medium. Samples were loaded on SDS‐PAGE gels for electrophoresis and transferred to PVDF membranes (Bio‐Rad Laboratories, USA). After 1 hour blocking with blocking solution (TAKARA), the membranes were incubated with primary antibodies (Table 2) overnight at 4 °C and followed by incubation with the corresponding secondary antibodies for 1 hour at room temperature. Visualization was performed using SuperSignal West Femto Maximum Sensitivity Substrate (Thermo‐Scientific), and detection was performed using an ImageQuant LAS 4000 mini (GE Healthcare Life Sciences, Chicago, IL, USA).

Table 2.

Details of Antibodies

Name	Company	Dilution
OPN	Proteintech	1:1000 (WB)
ATP‐binding cassette subfamily G member 1	Proteintech	1:1000 (WB)
CD36	ThermoFisher	1:1000 (WB)
		1:200 (immunocytochemistry)
SR‐A	Santa Cruz Biotechnology	1:1000 (WB)
		1:200 (immunocytochemistry)
Phosphorylated MLC	Cell Signaling	1:1000 (WB)
MLC	Cell Signaling	1:1000 (WB)
Rho associated coiled‐coil containing protein kinase 1	Cell Signaling	1:1000 (WB)
Ras homolog family member A	Cell Signaling	1:1000 (WB)
Phosphorylated NF‐κB p65	Cell Signaling	1:1000 (WB)
NF‐κB p65	Cell Signaling	1:1000 (WB)
Phosphorylated IκBα	Cell Signaling	1:1000 (WB)
IκBα	Cell Signaling	1:1000 (WB)
Phosphorylated IKKα	Cell Signaling	1:1000 (WB)
IKKα	Cell Signaling	1:1000 (WB)
Alpha smooth muscle actin	Cell Signaling	1:1000 (WB)
Matrix metalloproteinase 9	Santa Cruz Biotechnology	1:1000 (WB)
COL1A1	Cell Signaling	1:1000 (WB)
Anti‐beta actin	Proteintech	1:50000 (WB)
Ki67	Abcam	1:250 (immunocytochemistry)
OPN	Servicebio	1:2000 (immunofluorescence)
CD68	Servicebio	1:2000 (immunofluorescence)

IκBα indicates inhibitor of nuclear factor kappa B alpha; IKKα, IκB kinase alpha; MLC, myosin light chain; NF‐κB, nuclear factor kappa B; OPN, osteopontin; and WB, Western blot.

Cell Eosin Staining

Macrophages were fixed with a 4% formaldehyde solution. After washing, cells were incubated 0.5% Eosin Y solution (MUTO PURE CHEMICALS CO., LTD) for 3 minutes. Stained cells were imaged using the microscope (Biorevo BZ‐9000; Keyence).

Cell Immunofluorescence Staining

Macrophages were plated in chamber slides for 48 hours culture under different conditions. After fixing with 4% paraformaldehyde, the cells were blocked for 1 hour in room temperature using Blocking One Histo (nacalai tesque; Japan). Next, cells were incubated with primary antibodies (Table 2) overnight at 4 °C and followed by incubation with the corresponding secondary antibodies for 1 hour at room temperature. Nuclei and F‐actin fibers were stained with DAPI or TRITC‐phalloidin (Vector Labs), respectively. Immunofluorescences of staining were detected using an inverted fluorescence microscope (Olympus FV10i, Olympus). For each staining, 10 images were captured from randomly selected fields at a magnification of ×60, and the average fluorescence intensity was measured using ImageJ software.

Mendelian Randomization Analysis

The relationship between SPP1 (exposure) and stroke (outcome) was evaluated by 2‐sample Mendelian randomization (MR) analysis. ²⁶ We used publicly available protein quantitative trait loci data (from Decode database: https://www.decode.com/) and expression quantitative trait loci data (from GTEx Portal database: https://www.gtexportal.org/home/) to identify single‐nucleotide variants (SNVs) associated with SPP1. SNVs need to meet the following criteria: strongly associated with exposure, only through exposure to affect outcomes; and independence from confounders that affect the “exposure–outcome” relationship. SNVs were screened with the threshold (P<5×10⁻⁸, linkage disequilibrium r ²<0.001, F‐statistics greater than 10) (Table S1). Then, we obtained publicly available genome‐wide association studies summary data for stroke (40 585 cases and 406 111 controls, exclusively derived from individuals of European descent) from IEU openGWAS (https://gwas.mrcieu.ac.uk). Details of genome‐wide association studies data are provided in Table S2.

The inverse variance weighted was chosen as our main MR analysis method. Heterogeneity was measured by the inverse variance weighted method. The presence of pleiotropy was identified and corrected by MR‐Egger intercept test. ²⁷

Histological Analysis on Human Atherosclerotic Carotid Tissue Samples

Atherosclerotic carotid plaques were obtained from patients undergoing carotid endarterectomy at the Second Affiliated Hospital of Nanchang University, China. The study was approved by Ethics Committees of the Second Affiliated Hospital of Nanchang University. An informed consent form was obtained from all participants. The exclusion criteria were current infection, autoimmune diseases, active or recurrent cancer, severe renal failure requiring dialysis, and peripheral arterial occlusive disease with rest pain. All studies were performed in accordance with the Declaration of Helsinki. The dissected plaques were washed several times with PBS to remove blood cells as much as possible. Tissue samples were fixed in formalin, embedded in paraffin, and cut into 5μm‐thick sections for experiments.

Tissue sections were stained with hematoxylin and eosin and examined and imaged using a microscope (OLYMPUS IX71, OLYMPUS). Immunohistochemical staining was performed to evaluate the expression of OPN and CD68 in plaque tissues. Briefly, paraffin tissue sections were dewaxed and dehydrated. After blocked with blocking buffer (Blocking One Histo, nacalai tesque, Japan), the slides were incubated with the CD68 antibody in 4 °C overnight, followed by incubation with the secondary antibody for 1 hour at room temperature. Next, slides were incubated with IF488‐Tyramide working solution (1:500; Servicebio) in the dark at room temperature for 10 minutes. After washing and performing antigen retrieval, the slides were incubated with OPN antibody at 4 °C overnight. Subsequently, the slides were incubated with the corresponding secondary antibody and Cy3‐Tyramide working solution (1:500; Servicebio). Stained sections were examined and imaged using the microscopes (Nikon Eclipse C1) and scanner (Pannoramic MIDI).

Statistical Analysis

Machine learning was performed by splitting 75% of the data as the training data set and 25% of the data as the test data set. Using the “glmnet” package, ²⁸ least absolute shrinkage and selection operator regression analysis on linear relationships between gene candidates and disease risk was done under the optimal lambda value of 95 determined by cross‐validation analysis. Using the “xgboost” package, ²⁹ Extreme Gradient Boosting analysis on the nonlinear relationships between gene candidates and disease risk was done by setting grid parameters through expand.grid function. We used receiver operating characteristic curve analysis to assess the reliability of the predictive model. Least absolute shrinkage and selection operator is a linear regression method that performs feature selection by imposing L1 regularization, making it effective for identifying key variables. In contrast, Extreme Gradient Boosting is a tree‐based ensemble learning method that captures nonlinear relationships and interactions among features. By using both methods, we aimed to enhance the robustness of our analysis.

Statistical analyses were performed using GraphPad Prism 10. Data are presented as mean±SD. For in vitro experiments involving >2 groups, 2‐way ANOVA followed by Tukey's multiple comparisons test was used. Data were obtained from 4 independent experiments, with no repeated measures on the same samples. For comparisons between 2 groups, including in vitro experiments, bulk transcriptome data analysis, and human tissue histological analysis, unpaired Student's t test was applied. A value of P<0.05 was considered statistically significant. The specific statistical tests used for each experiment are detailed in the figure legends.

RESULTS

Single‐Cell Transcriptomic Analysis Combined With Machine Learning Elucidates the Correlation Between SPP1 and Carotid Plaque

We collected scRNA‐seq data of 57 158 cells and obtained 45 688 cells for final analysis after batch effect removal and data normalization (Figure S1). Using the clustering tree to determine the optimal resolution value of 0.5, all cells were divided into 23 clusters based on their respective top genes (Figure S1; Table S3) and distinguished into 8 cell types: T cell (CD2, CD3D, CD3E, CD3G, IL‐7R), B cell (CD79A, CD79B, MS4A1), natural killer cell (NKG7, GNLY), mast cell (MS4A2, CPA3, KIT, TPSAB1), endothelial cell (CD34, PECAM1, VWF), smooth muscle cell/SMC (ACTA2, MYH11, TAGLN), fibroblast (DCN, PODN, LUM), and macrophage (CD14, CD68, FCGR2A, LYZ) (Figure 1A; Figures S1, S2A). We found that the proportion of each cell type, particularly the macrophage and endothelial cell, was different between ACP and proximal adjacent portions of the carotid artery (Figure 1B; Figure S2B,C).

Figure 1. Cell communication and machine learning highlight the unique role of SPP1 in atherosclerotic carotid plaques.

A, tSNE plot displaying the cell atlas identification of 45 688 cells in atherosclerotic tissue. B, Proportion of each cell type in ACP and PAA. C, Venn diagram showing the overlap of 82 genes identified by 2 machine learning methods. D, PPI network displaying central genes. E, Significant signaling pathways were ranked based on their differences of information flow within the inferred networks between ACP and PAA. F, SPP1 expressions were analyzed in atherosclerotic arteries, advanced plaques, IPH, and their control samples using transcriptomes data. Analyzed with Student's unpaired t test. *P<0.05. ACP indicates atherosclerotic carotid plaques; EC, endothelial cell; IPH, intraplaque hemorrhage; LASSO, least absolute shrinkage and selection operator; NK, natural killer; PAA, proximal adjacent portions of carotid artery; PPI, protein–protein interaction; SMC, smooth muscle cell; tSNE, t‐distributed stochastic neighbor embedding; and XGBoost, Extreme Gradient Boosting.

Machine learning was performed to explore the key genes involved in plaque formation (Table S4). Using training data set to construct the least absolute shrinkage and selection operator linear model and Extreme Gradient Boosting nonlinear model, both models exhibited excellent predictive validities in the training and testing data sets (area under the curve >0.9) (Figure S2E,F). The 82 genes crossing over 2 models were used to construct an interaction network (Figure 1C,D). Using the maximum clique centrality method, we identified FN1, ELN, SPP1, and CTGF as the key genes (Figure 1D).

Cell–cell communication analysis also revealed the unique signal of SPP1 in ACP (Figure 1E), suggesting its potential role in carotid plaque formation. According to bulk transcriptome data, ¹⁵ , ¹⁶ , ¹⁷ enhanced expression of SPP1 was found in carotid atherosclerotic artery, advanced plaques, and intraplaque hemorrhage (IPH) samples (Figure 1F) and showed significant predictive potentials (Figure S2D).

SPP1 ^hi Macrophages Exert Distinct Effects in the Formation and Progression of Carotid Atherosclerosis

In deep cell–cell communication analysis showed that SPP1 almost exclusively expressed in macrophages in ACP (Figure 2A), and macrophages in ACP send the majority of SPP1 outgoing signals (Figure S3A). In contrast, SPP1 incoming signals are picked up by all cell types, including macrophage, SMC, endothelial cell, fibroblast, and T cell in ACP rather proximal adjacent portions of the carotid artery (Figure S3B), probably through CD44 and integrin receptors (Figure S3C). Interestingly, the expression levels of SPP1 are largely varied among macrophages even in ACP (Figure 2A).

Figure 2. Functional characteristics of SPP1 ^hi macrophage.

A, The expression distribution of signaling genes involved in the inferred SPP1 signaling network. B, Heatmap showing the average gene expression in each macrophage subcluster (top 5 genes sorted by fold change). C, tSNE plot showing the distribution of reclustered cell populations with macrophage and SPP1 distribution in PAA and ACP. D, SPP1 ^hi macrophage score were analyzed in atherosclerotic arteries, advanced plaques, IPH, and their control samples using transcriptomes data. E, F, GO and KEGG analysis of macrophage subtypes based on specific upregulated genes in each cluster. Analyzed with Student's unpaired t test. *P<0.05. ACP indicates atherosclerotic carotid plaques; EC, endothelial cell; GO, Gene Ontology; IPH, intraplaque hemorrhage; KEGG, Kyoto Encyclopedia of Genes and Genomes; MHC, major histocompatibility complex; NK, natural killer; PAA, proximal adjacent portions of carotid artery; SMC, smooth muscle cell; and tSNE, t‐distributed stochastic neighbor embedding.

To better explore the unique role of SPP1 in plaque development, we divided macrophages into 7 distinguishable clusters based on their gene expression properties (Figure 2B; Table S5), and SPP1 is highly and extensively expressed in cluster 1 macrophages. Moreover, macrophages are differentially distributed between proximal adjacent portions of the carotid artery and ACP, and cluster 1 macrophages are predominantly found in ACP (Figure 2C; Figure S3D). All these data clearly indicate the enrichment of cluster 1 macrophages with high SPP1 expression (hereafter SPP1 ^hi macrophage) in ACP (Figure 2B,C), suggesting the critical role of SPP1 ^hi macrophage in plaque development. Using the top 10 genes as references (Table S5), the scores of SPP1 ^hi macrophage were significantly higher in carotid atherosclerotic arteries, advanced plaques, and IPH comparing to their controls (Figure 2D). These data provide clear evidence about the distinct role of SPP1 ^hi macrophages in plaque formation.

To explore the relevant molecular mechanisms by which SPP1 ^hi macrophages promote plaque formation, we performed functional and pathway enrichment analyses. Differing from other clusters, the cluster 1 SPP1 ^hi macrophages are functionally responsible for lipoprotein particle stimulation and oxidative stress (Figure 2E), through distinctive pathways (Figure 2F).

The traditional view classifies macrophages into M1 proinflammatory (CCL2, CCL7, CXCL3) and M2 anti‐inflammatory (FN1, IL4, IGF) subtypes. ³⁰ , ³¹ Of note, SPP1 ^hi macrophages highly expressed both M1 and M2 markers (Table S5), suggesting SPP1 ^hi macrophages do not to be defined by traditional M1/M2 paradigm. Recent studies have demonstrated the role of TREM2+ macrophages in atherosclerosis. ³² However, we found that TREM2 is highly expressed in either cluster 1 (SPP1 ^hi macrophages) or cluster 0 (resident‐like macrophages) (Figure S4A). Additionally, based on their respective top genes, clusters 0, 3, and 5 correspond to the resident‐like subtype, associated with macrophage activation and humoral immune response; clusters 2 and 4 correspond to the inflammatory subtype, involved in regulating the inflammatory response; and cluster 6 corresponds to the proliferating subtype, functionally responsible for mitotic cell cycle phase transition (Figure 2E; Figure S4A).

Considering the possibility of sample heterogeneity, we performed analysis for further validation via PlaqView website (https://plaqviewv2.pods.uvarc.io/). ³³ Slenders et al. analyzed scRNA‐seq data from 38 human carotid plaques and identified CD14+CD68+ macrophages I as a crucial cell subpopulation in carotid plaques ³⁴ (Figure S4B). Interestingly, CD14 + CD68+ macrophages I faithfully replicate the characteristics of SPP1 ^hi macrophages, because CD14 + CD68+ macrophages I extensively express the top 10 genes in SPP1 ^hi macrophages (Figure S4C), functional involves in phospholipid metabolic process, LDL particle stimulus, TNF (tumor necrosis factor) signaling pathway, and cytoskeletal regulation (Figure S4D) and express both M1 and M2 markers (Figure S4E). Therefore, SPP1 ^hi macrophages represent a distinct subpopulation that is integrally involved in the development of atherosclerotic carotid plaques.

Upregulated iOPN in Macrophages Under Hypoxia and Ox‐LDL Stimulations Promote Foam Cell Formation Through Cytoskeleton and Membrane Structure Rearrangement

As functional annotation analysis highlights the response of SPP1 ^hi macrophages to lipoprotein particle stimulus and oxidative stress (Figure 2E,F), we cultured mouse macrophages under hypoxia and ox‐LDL stimulations. The expression of iOPN was significantly upregulated in macrophages under either hypoxia or ox‐LDL treatment, and further enhanced by combined stimulations (Figure 3A). The expression of several proteins involving in lipid uptake (CD36, SR‐A) and cholesterol efflux (ABCG1 [ATP‐binding cassette subfamily G member 1]) was also significantly enhanced under hypoxia and ox‐LDL stimulations, although CD36 was not induced by hypoxia alone (Figure 3A). In parallel with the upregulated iOPN, the ratio of Oil Red O‐positive stained cells was also increased with hypoxia and ox‐LDL treatments (Figure 3B). The knockdown of SPP1 in macrophages effectively mitigated the hypoxia and ox‐LDL‐induced expression of CD36, SR‐A, and ABCG1 (Figure 3C) and foam cell formation (Figure 3D). All these in vitro experimental data confirmed the role of OPN in promoting form cell formation of macrophages.

Figure 3. The responses of mouse macrophages to hypoxia and ox‐LDL stimulations.

A, Western blot analysis on the expression of OPN, CD36, SR‐A, and ABCG1 in macrophages treated with 1% O₂ or 50 μg/mL ox‐LDL for 48 hours. B, Oil Red O staining on foam cells formation at 48 hours after treatments. C, Western blot analysis on the expression of OPN, CD36, SR‐A, and ABCG1 in macrophages with SPP1 knockdown under hypoxia and ox‐LDL stimulations. D, Oil Red O staining on foam cells formation in macrophages with SPP1 knockdown under hypoxia and ox‐LDL stimulations. All data are presented as mean±SD from 4 independent experiments and analyzed with 2‐way ANOVA followed by the Tukey's multiple comparisons test. *P<0.05. ABCG1 indicates ATP‐binding cassette subfamily G member 1; OPN, osteopontin; and ox‐LDL, oxidized low‐density lipoprotein.

Previous study has reported that OPN is involved in cytoskeleton regulation, ³⁵ and actin cytoskeleton remodeling in macrophages can increase the clustering of CD36 on the cell membrane, thereby promoting foam cell formation. ³⁶ We noticed the round shape of macrophages with SPP1 knockdown (Figure S5A). Therefore, we further investigated the detailed molecular mechanisms by focusing on cytoskeleton remodeling. As expected, hypoxia and ox‐LDL stimulations significantly increased the levels of cytoskeletal‐related proteins, including p‐MLC/MLC (phosphorylated myosin light chain), ROCK1 (Rho associated coiled‐coil containing protein kinase 1), and RhoA (Ras homolog family member A) in macrophages (Figure 4A). However, SPP1 knockdown only partially but significantly alleviated the enhanced expression of p‐MLC, ROCK1, and RhoA in macrophages under hypoxia and ox‐LDL stimulations (Figure 4A). More interestingly, SPP1 knockdown significantly increased the ratio of p‐MLC/MLC in macrophages (Figure 4A). As previous study has reported that sOPN is present in various body fluids such as serum, milk, and urine, ³⁷ we surmise the involvement of sOPN in culture medium. We could detect OPN in the basic medium containing 10% FBS (Figure 4B). Hypoxia and ox‐LDL stimulations to macrophages not only enhanced the expression of intracellular OPN (Figure 3C, but also dramatically induced the secretion of OPN into the conditioned medium (Figure 4B). Naturally, SPP1 knockdown effectively mitigated the enhancement of either iOPN or sOPN under hypoxia and ox‐LDL stimulations (Figures 3C, 4B).

Figure 4. Intracellular osteopontin promotes foam cell formation of macrophages through cytoskeletal rearrangement.

A, Protein expression levels of p‐MLC, MLC, ROCK1, and RhoA in macrophages with SPP1 knockdown under hypoxia and ox‐LDL stimulation. B, Western blot analysis was performed to detect osteopontin in medium, and ponceau S staining was done for total protein normalization. C, The protein expression of OPN, p‐MLC, MLC, ROCK1, and RhoA were examined in macrophages treated with OPN antibody for 48 hours, with SPP1 knockdown under hypoxia and ox‐LDL stimulation. D, E, immunofluorescence staining of CD36, SR‐A, and phalloidin in macrophages with SPP1 knockdown under hypoxia and ox‐LDL stimulation. Scale bars: 20 μm. All data are presented as mean±SD from 4 independent experiments and analyzed with 2‐way ANOVA followed by the Tukey's multiple comparisons test. *P<0.05. CM indicates conditioned medium; MLC, myosin light chain; OPN, osteopontin; ox‐LDL, oxidized low‐density lipoprotein; p‐MLC phosphorylated myosin light chain; RhoA, Ras homolog family member A; and ROCK1, Rho associated coiled‐coil containing protein kinase 1.

To confirm the role of exogenous OPN in medium, we neutralized the sOPN using antibody. We found that the neutralization of exogenous OPN significantly decreased the p‐MLC/MLC ratio in macrophages with SPP1 knockdown (Figure 4C). As the neutralization of exogenous OPN did not significantly affect the expression of intracellular OPN and cytoskeletal‐related proteins in macrophages (Figure S5B), we performed a further confirmation experiment on the direct regulatory role of iOPN in cytoskeleton rearrangement and lipid uptake. We observed that the neutralization of exogenous OPN did not remarkably affect the enhancement of CD36 and SR‐A in macrophages under hypoxia and ox‐LDL stimulations (Figure S5C). Immunofluorescence staining further indicated the formation of stress fibers and the induced colocalization of CD36 and SR‐A with F‐actin bundles in macrophages under hypoxia and ox‐LDL treatments, but these alternations disappeared in macrophages with SPP1 knockdown (Figure 4D,E). Previous studies have confirmed that the translocation of CD36 and SR‐A are crucial for lipid uptake. ³⁸ , ³⁹ Therefore, it seems that upregulated iOPN in macrophages under hypoxia and ox‐LDL stimulations promotes foam cell formation through cytoskeleton and membrane structure rearrangement.

Increased iOPN in Macrophages Under Hypoxia and Ox‐LDL Stimulations Activates the Nuclear Factor Kappa B Pathway to Release Proinflammatory Cytokines

Although sOPN has been demonstrated to induce IL‐1 (interleukin‐1), IL‐12, IFN‐γ (interferon‐gamma), and TNF‐α expression through integrins and CD44, ⁹ , ⁴⁰ , ⁴¹ the role of iOPN in inflammatory processes remains unclear. Our in vitro experiments showed that the relative expression of p‐IKKα (phosphorylated IκB kinase alpha), p‐IκBα (phosphorylated inhibitor of nuclear factor kappa B alpha), and p‐p65 (phosphorylated p65) in macrophages were significantly increased under hypoxia and ox‐LDL treatments, which was effectively attenuated by SPP1 knockdown (Figure 5A). Additionally, SPP1 knockdown alone also led to a slight suppression of p‐IKKα, p‐IκBα, and p‐p65 (Figure 5A). Moreover, hypoxia and ox‐LDL stimulations significantly induced the release of IL‐6, TNF‐α, and CCL2 (C‐C motif ligand 2) from macrophages but was completely canceled by SPP1 knockdown (Figure 5B). Our data suggested that upregulated iOPN in macrophages under hypoxia and ox‐LDL stimulations activated NF‐κB (nuclear factor kappa B) pathway to induce the release of proinflammatory cytokines.

Figure 5. Intracellular osteopontin activates NF‐κB pathway to induce the release of inflammatory cytokines from macrophages under hypoxia and ox‐LDL stimulations.

A, Analysis of protein expression levels of p‐IKKα, IKKa, p‐IκBα, IκBα, p‐p65, and p65 (as well as concentrations of IL‐6, TNF‐α, and CCL2 in conditioned media (B)) in macrophages with SPP1 knockdown under hypoxia and ox‐LDL stimulation. C, CM collected from macrophages under different treatment and cocultured with VSMCs for 24 hours. Western blot analysis of MMP9, collagen I, and α‐SMA in VSMCs. D, Immunofluorescence staining for Ki‐67 (green: Ki‐67, blue: DAPI, scale bars: 100μm) and quantification of Ki‐67‐positive cells in VSMCs. All data are presented as mean±SD from 4 independent experiments and analyzed with 2‐way ANOVA followed by Tukey's multiple comparisons test. *P<0.05. α‐SMA indicates alpha smooth muscle actin; CCL2, C‐C motif ligand 2; CM, conditioned media; IL‐6, interleukin‐6; MMP9, matrix metalloproteinase 9; NF‐κB, nuclear factor kappa B; ox‐LDL, oxidized low‐density lipoprotein; p‐IKKα, phosphorylated IκB kinase alpha; p‐IκBα, phosphorylated inhibitor of nuclear factor kappa B alpha; p‐p65, phosphorylated p65; TNF‐α, tumor necrosis factor alpha; and VSMC, vascular smooth muscle cells.

Synthetic phenotype switching of VSMCs is essential for plaque formation, and our data analysis showed the relationship between macrophages and VSMCs (Figure S3C). VSMCs cultured with CM from macrophages under hypoxia and ox‐LDL stimulations significantly increased the expression of MMP9 (matrix metalloproteinase 9) and collagen I but decreased the expression of α‐SMA (alpha smooth muscle actin) (Figure 5C), suggesting the switching from a contractile to a synthetic phenotype. Additionally, Ki‐67 staining revealed that CM from macrophages under hypoxia and ox‐LDL stimulations significantly enhanced the proliferation of VSMCs (Figure 5D). It seems that the upregulated iOPN in macrophages under hypoxia and ox‐LDL stimulations activates the NF‐κB pathway to release proinflammatory cytokines, which thereby lead to the biological/phenotypic changes of VSMCs.

Clinical Data Confirm the Association Between SPP1 and Atherosclerotic Diseases

Many genetic variations closely associated with specific traits, along with the publication of large‐sample genome‐wide association studies, have enabled researchers to accurately analyze the causal relationship between risk factors and diseases. As atherosclerotic carotid plaques are the most important risk factor leading to stroke, we used 2‐sample MR randomization analyses to explore the association between SPP1 and stroke. Inverse variance weighted analysis indicates a causal relationship between SPP1 and stroke, at either protein expression level (odds ratio [OR], 1.076 [95% CI, 1.009–1.148]; P value=0.025) or genetic expression level (OR, 1.085 [95% CI, 1.032–1.141]; P value=0.001) (Figure 6A). Additionally, there was no evidence of heterogeneity and pleiotropy between SPP1 and stroke at protein quantitative trait loci and expression quantitative trait loci (Figure S6A, Tables 3, S6).

Figure 6. The association between SPP1 expression and the risk of atherosclerotic disease.

A, Visualization of the MR analysis of SPP1 on stroke. B, Representative images of immunofluorescence staining on OPN/ SPP1 (red) and CD68 (green) in each tissue sample of 4 early plaques and 5 advanced plaques. Nuclei were stained by DAPI (blue), and quantitative data on OPN and CD68 expression are shown in the bar graph. Analyzed with Student's unpaired t test. *P<0.05. eQTL indicates expression quantitative trait loci; MR, Mendelian randomization; OPN, osteopontin; OR, odds ratio; and pQTL, protein quantitative trait loci.

Table 3.

Results for Mendelian Randomization Analysis

Exposure	Outcome	Method	No. SNV	Beta	SE	P value	OR	OR Lower 95% CI	OR Upper 95% CI
pQTL‐SPP1	Stroke	Mendelian randomization Egger	6	0.065124382	0.059534383	0.335456004	1.067291768	0.949743806	1.199388416
pQTL‐SPP1	Stroke	Weighted median	6	0.06655881	0.037195608	0.073546141	1.06882382	0.993675503	1.149655349
pQTL‐SPP1	Stroke	Inverse variance weighted	6	0.073434399	0.032829765	0.02529786	1.076197934	1.009129481	1.14772387
pQTL‐SPP1	Stroke	Simple mode	6	0.05956519	0.043352896	0.227849363	1.06137495	0.974913543	1.155504292
pQTL‐SPP1	Stroke	Weighted mode	6	0.0689444	0.041155858	0.154745557	1.071376639	0.988347344	1.161381078
eQTL‐SPP1	Stroke	Inverse variance weighted	2	0.081739845	0.025676265	0.001455168	1.08517346	1.031913005	1.141182864

OR indicates odds ratio; pQTL, protein quantitative trait loci; eQTL, expression quantitative trait; SNV, single‐nucleotide variant; and SPP1.

To further confirm our results, atherosclerotic carotid plaques were obtained from 9 patients undergoing carotid endarterectomy. According to the guidelines of the American Heart Association committee on differentiating atherosclerotic plaques, ⁴² the 9 carotid plaque samples were divided into early plaques (N=4) and advanced plaques (N=5) (Figure S6B). Immunofluorescence staining revealed that, despite a similar number of macrophages, OPN expression was detected more extensively in advanced plaques than in early plaques (Figure 6B), suggesting the role of SPP1 ^hi macrophages in carotid plaque formation and progression.

DISCUSSION

Carotid atherosclerosis can induce partial or complete occlusion of the vessel, thereby restricting blood flow to the brain. Even more important, plaque formation and acute rupture lead to ischemic stroke due to local blockage and thrombosis. Severe sequelae and life‐threatening consequences further highlight the importance of understanding the formation and development of atherosclerotic carotid plaques.

Here, we report a subset of SPP1 ^hi macrophages enriched in atherosclerotic plaque tissues can be classified as foam cell types and exhibit high expression of lipid metabolism‐related genes, such as FABP4, FABP5, and CD36. Functional and pathway analyses also indicate the association between SPP1 ^hi macrophages and lipid metabolism. In addition, SPP1 ^hi macrophages also exhibit NF‐κB pathway activation and increased proinflammatory cytokines release. The mainstream identifies activated macrophages as M1 (inflammatory property) macrophages and M2 (anti‐inflammatory and wound‐healing property) macrophages. Foam macrophages have been classified as noninflammatory and may exhibit an M2 phenotype in atherosclerosis. ⁴³ , ⁴⁴ However, we found that SPP1 ^hi macrophages exhibit both characteristics of M1 and M2, representing a novel subpopulation that is not captured by traditional M1/M2 classification. The PLIN2 ^hi /TREM1 ^hi macrophages in atherosclerosis displaying lipid accumulation and inflammation transcriptomic characteristics and associating with a high risk of cerebrovascular events have been recently identified. ⁴⁵ Therefore, because there is no impassable barrier between foam macrophages and inflammatory macrophages, certain specialized populations of macrophages, such as SPP1 ^hi macrophages and PLIN2 ^hi /TREM1 ^hi macrophages, may promote atherosclerotic plaque formation by regulating metabolic processes, leading to lipid accumulation and foam cell formation, while simultaneously exacerbating atherosclerosis by activating and maintaining inflammation.

OPN encoded by SPP1 is a multifunctional secreted glycoprotein originally described as T lymphocyte activation protein and expressed in macrophages in several tissues. ³⁷ Most studies have focused on the extracellular effects of secreted OPN, such as angiogenesis, cell proliferation, phagocytosis, and antiapoptosis. ⁷ , ⁸ , ⁹ In contrast, intracellular osteopontin has been demonstrated to be associated with osteoblast differentiation, natural killer cell development, and inflammatory pathways activation. ⁴⁶ , ⁴⁷ It has been reported that OPN transgenic mice fed an atherogenic diet develop more severe aortic atherosclerosis compared with wild‐type mice, accompanying by increased OPN expression in foam cells within plaques. ⁴⁸ In apoE−/− (apolipoprotein E) mice exposed to angiotensin II, OPN deficiency not only significantly reduces atherosclerotic lesions but also leads to fewer macrophages and foam cells in the plaques. ⁴⁹ However, the precise role and mechanisms of iOPN in atherosclerosis are still poorly understood.

In this study, we have demonstrated that iOPN promotes foam cell formation of macrophages under hypoxia and ox‐LDL stimulations through cytoskeleton and membrane structure rearrangement. Kang et al. have demonstrated that OPN‐deficient erythroblasts exhibit defects in F‐actin filaments through the phosphorylation or activation of multiple proteins, including Rac‐1 GTPase and the actin‐binding protein, but exogenous OPN treatment restores all these changes. ³⁵ However, our data showed that the enhanced iOPN in macrophages rather than the increased sOPN under hypoxia and ox‐LDL stimulation induced foam cell formation via upregulated the expression of cytoskeleton‐related proteins. Notably, the relationship between cytoskeletal rearrangement and macrophage phenotype has been previously reported, yet little is known about the role of the cytoskeleton in foam cell formation. ⁵⁰

VSMCs switching to the synthetic phenotype with increased proliferative ability are considered a key contributor to both the early and late stages of atherosclerosis. In primary mouse aortic VSMCs, OPN induced synthetic phenotypic switching by downregulating smooth muscle actin and calponin expression through extracellular signaling pathways. ⁵¹ Moreover, during the early stages of atherosclerosis, OPN expression coincides with the induction of VSMC proliferation, and anti‐OPN treatment has been shown to inhibit this response. ⁵² We observed an increased release of proinflammatory cytokines from macrophages under hypoxia and ox‐LDL stimulations strongly promoted the proliferation and transition of VSMCs to a synthetic phenotype. This suggests that SPP1 ^hi macrophages may alter the biological properties of VSMCs through paracrine effects.

This study has some limitations. First, we found that iOPN regulates cytoskeletal organization, increases lipid protein expression, promotes foam cell formation, and induces the colocalization of lipid protein with F‐actin bundles, suggesting a potential association between iOPN‐induced cytoskeletal changes and lipid protein distribution. However, further study is needed to better understand how iOPN modulates cytoskeletal organization, lipid protein distribution, and lipid metabolism. Second, the MR analysis in this study was conducted using European ancestry cohorts, which may limit the applicability of our findings. Third, we used 2 methods (weighted median method and MR‐Egger regression) to reduce bias from horizontal pleiotropy and other potential confounders, residual confounding due to environmental factors (eg, diet, lifestyle) cannot be completely eliminated. Validation in independent cohorts with non‐European populations is needed to confirm the robustness and wild applicability of our findings. Finally, it will be better to perform in vivo experiments to further confirm the role of intracellular OPN in carotid plaque formation.

Conclusions

In summary, we have identified the SPP1 ^hi macrophages as a distinct subpopulation involving in the development of atherosclerotic carotid plaques. Our in vitro experiments have further demonstrated that upregulated intracellular osteopontin in macrophages under hypoxia and ox‐LDL stimulations can directly promote foam cell formation through cytoskeleton and membrane structure rearrangement and indirectly induce the proliferation and phenotypic switching of VSMCs by releasing proinflammatory cytokines. Targeting intracellular OPN in macrophages may represent a novel effective therapeutic approach for atherosclerotic carotid plaques.

Sources of Funding

This work was supported by the Grant‐in‐Aid for Scientific Research from the Japan Society for the Promotion of Science; the Program of the Network‐type Joint Usage/Research Center for Radiation Disaster Medical Science, and the Collaborative Research Program from the Atomic Bomb Disease Institute of Nagasaki University. The funder had no role in the study design, data collection, data analysis, and preparation of the article. Nie Han is awarded by China Scholarship Council (No. 202106820015).

Disclosures

None.

Supporting information

Tables S1–S6

Figures S1–S6