Transcriptome Analysis Identified SPP1+ Monocytes as a Key in Extracellular Matrix Formation in Thrombi

Thrombi follow various natural courses. They are known to become harder over time and may persist long-term; some of them can also undergo early spontaneous dissolution and disappearance. Hindering thrombus stability may contribute to the treatment of thrombosis and the prevention of embolisms. However, the detailed mechanisms underlying thrombus maturation remain unknown. Using RNA sequencing, we revealed the transcriptional landscape of thrombi retrieved from the cerebral vessels and identified SPP1 as a hub gene related to extracellular matrix formation. Immunohistochemistry confirmed the expression of osteopontin in monocytes/macrophages in the thrombi, particularly in older thrombi. Single-cell RNA sequencing of thrombi from the pulmonary artery revealed increased communication between SPP1-high monocytes/macrophages and fibroblasts. These data suggest that SPP1-high monocytes/macrophages play a crucial role in extracellular matrix formation in thrombi and provide a basis for new antithrombotic therapies targeting thrombus maturation. Teaser SPP1+ monocytes play a key role in thrombus maturation, which can be a potential target for novel antithrombotic therapies.


Introduction
Thromboembolic diseases account for 1 in 4 deaths worldwide (1).These diseases can be classified into two groups, based on whether the thrombus is formed under high-or low-pressure systems (2), with intracardiac (cardiac atria) and venous thrombi accounting for most of the latter.Cardiogenic embolism, which can be caused by the embolism of an intracardiac thrombus, is the most severe type of ischemic stroke.
Prevention of cardiogenic embolism is of great social interest because the number of patients with atrial fibrillation, the largest risk factor for cardiogenic embolism, is rapidly increasing (3).Venous thrombi primarily form in the lower extremities and can be symptomatic.In some patients, they can embolize the pulmonary artery, which can be fatal.Anticoagulants are primarily used to prevent and treat thromboembolic diseases under low-pressure systems.Anticoagulants can decrease the risk of cardioembolic stroke in patients with atrial fibrillation; however, an approximately 40% stroke risk remains even upon treatment with anticoagulants (4).In addition, patients may experience bleeding complications (5).Therefore, a more effective antithrombotic therapy with a low risk of bleeding is desirable.
Thrombi detected in patients can either disappear or persist.More than half (63-89%) of thrombi found in atrial appendages disappear without symptomatic embolic events, while some patients develop cerebral embolism (6)(7)(8)(9).Moreover, 35% of surgery-associated venous thrombi resolve spontaneously within 72 hours, while some extend to involve the proximal veins (10,11).Furthermore, half of pulmonary embolisms resolve within a few weeks, while thrombi persist; chronic thromboembolic pulmonary hypertension (CTEPH) occurs in approximately 5% of patients after pulmonary embolism (11,12).Recently, dynamic changes occurring in thrombi over time have been noted (13,14), which are presumably involved in determining the natural course of thrombi.As time passes after a deep venous thrombus is formed, the proportion of red blood cells (RBCs) decreases, and the level of extracellular matrix (ECM) proteins, such as collagen, increases (14).Thrombi can become stiffer and more resistant to thrombolysis over time (15,16).We have previously reported that older thrombi are more resistant to reperfusion therapy in patients with cerebral embolism (17).Hence, it can be assumed that there are conflicting processes within a formed thrombus: some to make it stiff and stable and others to dissolve it.However, the mechanisms underlying these processes are not completely understood.
Here, we report the transcriptional differences between thrombi retrieved from cerebral vessels and peripheral blood.Our data revealed several major processes occurring in the thrombus, including ECM formation and inflammatory and antiinflammatory responses.We identified SPP1 as an upregulated hub gene for ECM formation in the thrombus.Furthermore, using single-cell RNA sequencing data, our results suggest that SPP1-high monocytes/macrophages (MCs/MPs) are key players in ECM formation in thrombi.Collectively, our data provide a basis for developing new antithrombotic therapies that modify the natural history of thrombi.

RNA expression in thrombi differs from that in blood
Mechanical thrombectomy is an endovascular surgery procedure that has rapidly developed over the last decade to recanalize occluded cerebral vessels in patients with acute ischemic stroke.In this study, we compared the RNA expression profile in three thrombi retrieved from cerebral vessels via mechanical thrombectomy with that in simultaneously sampled blood (Fig. 1A).All patients were diagnosed with a cardiogenic embolism.The heatmap in Fig. 1B shows a substantial difference in RNA expression between the thrombus and blood.Unbiased clustering detected sample differences (Fig. 1C).Compared to that in the blood, a total of 1,121 genes were significantly upregulated and 693 were downregulated in the thrombus (Fig. 1D).The top 20 upregulated differentially expressed genes (DEGs) included pro-inflammatory chemokines such as CXCL8 and CCL2, consistent with a previous report (Fig. 1E) (18).
We also found that genes related to the ECM, such as FN1 and SPP1, were upregulated in the thrombi.

Gene sets related to ECM were enriched in thrombi
The results of the gene set enrichment analysis are shown in Fig. 2A as network plots.
Biological process analysis revealed clusters, such as tissue morphogenesis and ECM organization.Cellular component analysis revealed clusters of ECM, cell adhesion, and mitochondrial components.Molecular function analysis revealed clusters such as chemokine activity, cell adhesion, and ECM constituents.In all analyses, pathways related to the ECM were significantly enriched.Genes related to ECM were highly expressed in thrombi (Supplementary Fig. S1)

Protein-protein interaction (PPI) analysis identified SPP1 as one of the hub genes
Among the 1,121 upregulated DEGs, 1,038 genes were identified in the STRING database after excluding the mitochondrial genes.We then identified 23 hub genes using CytoHubba (Supplementary Fig. S2).Unbiased MCL clustering identified 278 clusters using 963 genes.The top four clusters, which included the most genes among the identified clusters, are visualized in Fig. 2B.Cluster 1 included 55 genes associated with ECM formation.Cluster 2 included 31 genes related to the anti-inflammatory response.Cluster 3 included 26 genes related to the ERBB signaling pathway.Cluster 4 included 25 genes related to pro-inflammatory cytokines.Fifteen hub genes were included in the top four clusters.
We identified five hub genes (FN1, MMP2, IGF1, SERPINE1, and SPP1) in the ECM formation cluster, as this cluster included the most genes and pathways related to ECM were highly enriched in the gene set enrichment analysis (Fig. 2A).FN1 encodes for fibronectin, a glycoprotein involved in cell adhesion, migration, wound healing, and embryonic development.Fibronectin is vital for bleeding control (19).
MMP2 encodes matrix metalloproteinase-2, which is involved in the degradation and remodeling of various ECM components.It was reported that MMP2 inactivation prevented thrombosis and prolonged bleeding time (20).IGF-1 (insulin-like growth factor 1) is essential in the insulin signaling pathway and is a key growth factor involved in various processes such as cell proliferation, maturation, differentiation, and survival (21).It also regulates fibroblast growth and extracellular matrix deposition (22).SERPINE1 encodes plasminogen activator inhibitor-1; its deficiency causes abnormal bleeding (23).SPP1 encodes osteopontin (OPN), a secreted multifunctional glycophosphoprotein that plays an important role in physiological and pathophysiological processes (24).OPN drives immune responses under ischemic conditions and induces neutrophil and macrophage infiltration (25)(26)(27).It was recently reported that OPN is one of the ligands for integrin α9β1, a potential target for preventing arterial thrombosis (28)(29)(30).Therefore, we decided to investigate whether OPN is present in thrombi retrieved from cerebral vessels and to determine how OPN is associated with the clinical characteristics of patients who underwent mechanical thrombectomy.

Osteopontin is expressed by monocytes/macrophages in thrombi
We first validated the expression of SPP1 via RT-qPCR using five pairs of samples, adding two other pairs to the samples used for RNA sequencing.In the thrombi retrieved from the cerebral artery, the expression of SPP1 and its known receptor (CD44) was elevated compared to that in the blood.Additionally, TIMP1 which is modulated by OPN and related to ECM (31), was also elevated.
We then performed immunohistochemical staining of paraffin-embedded thrombus samples using three anti-OPN antibodies.All antibodies validated the presence of OPN in thrombi (Supplementary Fig. S3).All the observed samples were positive for OPN, but the extent varied (Fig. 3B, 3C).In one thrombus, some regions showed strong positivity, whereas others did not (Fig. 3D).We also observed OPN+ cell aggregation (where OPN+ cells were observed as a cluster) when the thrombus was strongly positive for OPN.Thus, we classified the thrombi into OPN-high or OPN-low according to the presence or absence of OPN+ cell aggregation.Double staining revealed that OPN was not expressed by neutrophils that are defined as cells with a lobulated nucleus and positive for neutrophil elastase (Fig. 3E).OPN was mainly expressed by MC/MPs in the thrombi (Fig. 3F).

Osteopontin expression is more observed in older thrombi than in fresh thrombi
Next, we analyzed 66 thrombi retrieved during mechanical thrombectomy for cerebral embolisms (Supplementary Fig. S4).Among the 66 samples, 40 were OPN-high and 26 were OPN-low.We then compared the features of the thrombi based on OPN expression.Fresh thrombi were less common among OPN-high samples (46% vs. 3%, P < 0.001; Fig. 4A).No significant differences were observed in the proportions of red RBCs, fibrin, or platelets.The density of MCs/MPs was higher in the OPN-high samples (Fig. 4B).As the density of MC/MPs is higher in older thrombi than in fresh ones (17,32), these observations support the idea that OPN-high thrombi tend to be older than OPNlow thrombi.

Osteopontin expression and clinical characteristics
To determine the association between OPN expression in thrombi and the clinical characteristics of patients, we compared patient backgrounds according to OPN expression (Table 1).The proportion of patients with atrial fibrillation was marginally higher (54% vs. 78%, P = 0.060), the level of brain natriuretic peptide was marginally higher (82.5 vs. 225 pg/mL, P = 0.054), and the cardiothoracic ratio was significantly higher (57% vs. 62%, P = 0.023) in patients with OPN-high thrombi.The proportion of stroke subtypes also differed significantly across OPN expression (P = 0.043, Fig. 5A); the proportion of cardioembolic stroke was higher among patients with OPN-high thrombi.We then investigated the effect of OPN expression on reperfusion quality.
There was no significant difference in the time to reperfusion after arterial puncture across OPN expression levels (P = 0.26, Fig. 5B), and no significant difference was observed in the number of passes before successful reperfusion (P = 0.17, Fig. 5C).

SPP1-high monocytes/macrophages in thrombi are related to ECM formation
After confirming that SPP1 was expressed by MC/MPs in stroke thrombi, we investigated whether SPP1 was also expressed in thrombi retrieved from patients with other thromboembolic diseases and the role of SPP1+ MC/MPs in thrombi.We used single-cell RNA sequencing data of thrombi from patients with CTEPH (33) to characterize MCs/MPs that express SPP1 in the thrombus.After unbiased clustering (Supplementary Fig. S5), we successfully identified MCs/MPs (Fig. 6A, 6B).SPP1 was mainly expressed by MC/MPs in these samples.Four subclusters of MCs/MPs were identified (Fig. 6C), and SPP1 was listed as one of the top ten highly expressed genes in subcluster 2 (Fig. 6D).The expression of hub genes identified via PPI analysis is shown in Fig. 6E.CD68 was highly expressed in subcluster 2, whereas CXCL8 and CCL3 were highly expressed in subcluster 1. Gene set enrichment analysis revealed that pathways related to the extracellular matrix are upregulated in subcluster 2 (Fig. 6E, red circle).
Pathways related to lysosomes (Fig. 6E, pink and light blue circles) were also upregulated in subcluster 2. In addition, pathways related to inflammation, such as the cellular response to tumor necrosis factor, response to interleukin-1, and chemotaxis, were enriched in subcluster 1 (Supplementary Fig. S6A).
To understand the role of subcluster 2 (SPP1-high MC/MPs) in thrombus formation, we performed a ligand-receptor interaction analysis using CellChat.Dense communication between MCs/MPs and fibroblasts, which are the major source of ECM, was inferred in the thrombi from patients with CTEPH (Fig. 7A).Among the immune cells found in the thrombus, MCs/MPs in subcluster 2 had the highest number and strength of inferred ligand-receptor interactions as the sender between fibroblasts (Fig. 7B).When MCs/MPs were assigned to the sender, the ligand-receptor pairs SPP1-ITGAV_ITGB1, SPP1-ITGA8_ITGB1, and SPP1-ITGAV_ITGB5 were inferred to be the main communicators between subcluster 2 MCs/MPs and fibroblasts (Fig. 7C, 7D).In addition, ligand-receptor pairs of PDGFB-PDGFRB were inferred between subcluster 1 MC/MPs and fibroblasts (Supplementary Fig. S6B).

SPP1-high monocytes/macrophages are expanded in murine venous vessel walls after thrombosis induction
As MC/MPs are presumed to migrate into the thrombus through the vessel wall (34), we hypothesized that SPP1-high MC/MPs may be present in the vessel wall after thrombosis.We investigated the expression of SPP1 by MC/MPs in the venous vessel walls of mice with and without deep vein thrombosis (DVT) (Supplementary Fig. S7A-S7D).Our results showed that the subcluster of MC/MPs with high SPP1 expression was expanded in the DVT group (Supplementary Fig. S7E).

Discussion
In this study, we compared RNA expression in thrombi from acute cerebral infarctions with that in simultaneously collected peripheral blood.In addition to pro-inflammatory responses, which have been reported previously (18), our findings revealed an antiinflammatory response and ECM formation in thrombi.SPP1 is one of the hub genes upregulated in thrombi, and the expression of its translation product, OPN, was observed particularly in older thrombi retrieved from the cerebral vessels.Furthermore, single-cell RNA sequencing data from thrombi in patients with CTEPH revealed that SPP1-high MC/MPs likely contributed to ECM formation.These results elucidate the transcriptional landscape that occurs after thrombus formation and could contribute to the development of new antithrombotic drugs that target thrombus maturation.SPP1+ macrophages have recently been identified as key cell types promoting tissue fibrosis (35,36).In line with these reports, the results of gene set enrichment analysis suggest that subcluster 2 MCs/MPs (SPP1-high) are related to the ECM in the thrombi of patients with CTEPH.This was supported by the results of our ligandreceptor interaction analysis.Integrin αV, an OPN receptor expressed by fibroblasts, is a key molecule in fibrotic diseases (37,38).The interaction between OPN and integrins αVβ1 and αVβ5 has been reported to contribute to the adhesion of smooth muscle and fibroadipogenic progenitor cells (36,39); OPN is also chemotactic for smooth muscle cells (40).Hence, SPP1-high MC/MPs may promote ECM formation by recruiting fibroblasts to the thrombus.Nonetheless, our results do not exclude the possibility that other subclusters of MCs/MPs are involved in ECM formation in the thrombus.Subcluster 1 MCs/MPs have been shown to promote fibrogenesis through PDGFB-PDGFRB interactions.In addition, the THBS1-CD47 interaction, which has been reported to promote fibrosis (41), was also found in other subclusters.Further investigations are required to clarify the role of heterogeneous MC/MPs in ECM formation in thrombi.OPN+ cells were not evenly distributed throughout the thrombus in thrombi retrieved from cerebral vessels.They were often densely aggregated in certain areas, particularly around the periphery of the thrombus.These aggregations of OPN+ cells are presumed to occur sometime after thrombus formation, as they are less common in fresh thrombi.In DVT, macrophages are presumed to increase migrating from the vascular wall to the thrombus during venous thrombosis (17,34,42).The present study showed that SPP1-high MC/MPs were expanded in the venous wall after thrombosis induction.Moreover, a recent study revealed that SPP1+ macrophages are expanded in the left atrial tissue of patients with persistent atrial fibrillation and could be targets for immunotherapy in atrial fibrillation (43).Therefore, some of the OPN+ MCs/MPs observed in thrombi from patients with cardioembolic stroke may have migrated from the left atrial wall.The observation in the present study that patients without atrial fibrillation had fewer OPN-high thrombi supports this hypothesis.Hence, the migration of OPN+ MCs/MPs may be a potential target for new antithrombotic therapies.
Intracardiac thrombi are frequently encountered in clinical practice, and the prevalence of left atrial appendage thrombus formation is estimated to be between 5% and 27% in patients who have not previously received anticoagulant therapy (44).
Patients with intracardiac thrombi are at an elevated risk of developing cardiogenic embolisms.Thus, convenient and cost-effective screening methods for intracardiac thrombi are required.Recently, OPN has been recognized as a biomarker for vascular diseases (24).The plasma concentration of OPN was found to be higher in patients with atrial fibrillation than in those without and is correlated with the extent of atrial fibrosis (45).Using unbiased proteomics, Mühlen et al. reported that OPN levels in the urine could be a biomarker for venous thrombosis and pulmonary embolism (46).It was also reported that plasma OPN levels are higher among patients with venous thrombosis than in those without (47).OPN is cleaved by thrombin into two halves.In this study, we showed that cleaved OPN (N-half) was present in a thrombus retrieved from a cerebral vessel.Considering that the presence of cleaved OPN implies a thrombogenic state, the level of OPN, especially cleaved OPN, in the peripheral blood may have potential value as a biomarker to predict intracardiac thrombus and future embolic events in patients with atrial fibrillation.
In the present study, thrombi retrieved from the cerebral vessels were found to contain neutrophils, consistent with previous studies (32,48,49).However, neutrophils could not be identified in the single-cell RNA sequencing data of thrombi harvested from patients with CTEPH.Experimental findings in a mouse inferior vena cava ligation model indicated that neutrophils are abundant in the initial stages of thrombus formation; however, their proportion decreased over time, with the proportion of MC/MPs increasing later on (50).Therefore, neutrophils may disappear from the CTEPH thrombus with age.
This study has several limitations.First, some RNA integrity numbers (RIN) were not sufficiently high in the thrombus samples.RINs in thrombus samples are commonly low, and acellular RINs tend to have lower RNA quantity and quality (51).
We chose samples with a relatively high RIN for RNA sequencing and validated the results using qPCR after adding other samples; however, careful interpretation of the results is needed.Second, the sample size of the clinical data may not have been sufficiently large to detect differences in reperfusion quality.
In conclusion, transcriptional responses, including ECM formation, inflammatory response, and anti-inflammatory response, were identified in thrombi in this study.Our results collectively suggest that SPP1+ MC/MPs play a key role in ECM formation in thrombi and may be a potential target for new antithrombotic therapies that modify thrombus maturation.

Sample collection and total RNA isolation
This study was approved by the Institutional Review Board of the Toyonaka Municipal Hospital, and all participants provided written informed consent.The clinical backgrounds of the participants are shown in Supplemental Table S1.Thrombi retrieved from five patients with acute ischemic stroke were stored in RNAlater (Thermo Fisher, Waltham, MA, USA) overnight at 4°C immediately after mechanical thrombectomy.
The thrombi were then transferred to a -80°C freezer until subsequent analyses.Whole blood was sampled from the patients via the femoral arterial sheath during the mechanical thrombectomy and stored using a PAXgene RNA blood collection tube (762165, BD Biosciences, San Jose, CA, USA) in a -80°C freezer until the subsequent steps.Total RNA was extracted from each thrombus using a ReliaPrep RNA tissue MiniPrep System (Z6110, Promega, Madison, WI, USA) and from whole blood using a PAXgene Blood RNA Kit (762164, Qiagen, Hilden, Germany) according to the respective manufacturer's instructions.

Library preparation and sequencing
The purity and quantity of the isolated total RNA were assessed using a NanoDrop 2000 spectrophotometer (Thermo Fisher).The RIN was assessed using Bioanalyzer 2100 (Agilent, Santa Clara, CA, USA).The three thrombus RNA samples with the highest RINs and their paired blood RNA samples were subjected to subsequent analyses.The RINs for the thrombus samples were 8.0, 6.8, and 4.5, and those for the paired blood samples were 7.2, 7.1, and 7.7.The total RNA samples were subjected to library preparation using an Illumina TruSeq stranded Total RNA Library Prep Kit with Ribo-Zero Globin (Illumina, San Diego, CA, USA) according to the manufacturer's protocol.
RNA libraries were subjected to 100-bp paired-end sequencing on a NovaSeq 6000 system (Illumina) with a median of 40 million reads.

Differentially expressed genes and gene set enrichment analysis
DEGs between the thrombi and blood were identified using an integrated browser application, iDEP1.1 (56).Minimal counts of 1.5/million in at least two libraries were set in the preprocessing data interface, and the count data were subjected to a variancestabilizing transform for clustering.A heat map was generated to include the top 2,000 genes.Unsupervised hierarchical clustering was performed using 1-Pearson correlation and average linkage.DEGs were identified using the iDEP built-in DESeq2 package (57) with a threshold of false discovery rate < 0.01 and a minimum absolute value of fold-change > 4.
Gene set enrichment analysis was performed using the DEGs defined above to determine the most enriched gene ontology (GO) pathways in terms of biological processes, cellular components, and molecular functions.Genes with a false discovery rate > 0.6 were removed from the pathway analysis.The top 20 pathways were displayed as a network.Nodes were connected if they shared 20% or more genes.All parameters not specified above were left as the default values.Pathway clusters were manually annotated.

Protein-protein interaction (PPI) analysis
The PPIs between all DEGs upregulated in the thrombus, except for mitochondrial genes, were analyzed using the STRING database (v12.0)(58).A minimal interaction score of 0.4 was set as the default.Hub genes were identified using Cytoscape (v3.10.1)software and the CytoHubba plugin (v0.1) (59).DEGs with the highest Matthews Correlation Coefficient scores were considered hub genes.The DEGs were clustered using MCL with a default inflation parameter of 3; GO enrichment analysis was performed on these clusters.Clusters were manually annotated based on the enriched pathways and visualized using Cytoscape software.

Quantitative real-time PCR analysis
We performed RT-qPCR to amplify and detect SPP1 and its receptor CD44 from five pairs of thrombus and blood RNA samples, comprising three pairs used for RNA-seq analysis and two additional pairs.First, cDNA was prepared by reverse-transcribing 400 pg of total RNA using Superscript III and random primers (Thermo Fisher Scientific).
The resulting cDNA was used for real-time PCR analysis using a Platinum SYBR Green qPCR SuperMix (Thermo Fisher Scientific).Next, 100 reverse transcription products and standard plasmids were subjected to real-time PCR analysis (QuantStudio 7 Flex Real-Time PCR System; Applied Biosystems) using human β-actin as an internal control.The following PCR program was used: 10 min of denaturation at 95°C, and then 40 cycles of 95°C for 15 s, 58°C for 30 s, and 72°C for 30 s.The primers used are listed in Supplementary Table S2.

Subjects for the histological analysis of the thrombus
We examined 168 consecutive patients who underwent mechanical thrombectomy for acute ischemic stroke between January 2015 and December 2019 at two tertiary referral hospitals with comprehensive stroke centers in Japan (Osaka University Hospital, Osaka; Osaka General Medical Center, Osaka).Thrombus specimens were available for 76 patients.Patients with left ventricular assist devices, atherosclerotic intracranial stenosis, or cerebral artery dissection were excluded.Finally, 66 patients with thrombi retrieved during mechanical thrombectomy for cerebral embolisms were included in this study (Supplementary Fig. S4).Clinical data were also collected from the included patients.The detailed methods for clinical data collection were as previously described (17).

Immunohistochemical staining and histological analysis
Thrombus samples were fixed in 10% neutral-buffered formalin and embedded in paraffin.To identify the presence of osteopontin (OPN), a product of SPP1, three primary antibodies were used: mouse monoclonal antibodies (10011, IBL, Gunma, Japan), rabbit polyclonal antibodies (25715-1-AP, Proteintech, Rosemont, IL, USA), and rabbit polyclonal antibodies against the N-terminal region of human OPN (18625, IBL).Immunohistochemical staining was performed using a Roche Ventana BenchMark GX autostainer (Ventana Medical Systems, Tucson, AZ, USA), according to the manufacturer's instructions.We stained human gallbladder and kidney samples as positive controls to determine the optimal antibody concentrations.The stained slides were captured as digital images using a VS200 Slide Scanner (Olympus, Tokyo, Japan).
The age, size, and components of the thrombi and the extent of NETosis were evaluated as previously described (17).Thrombus age was evaluated based on hematoxylin and eosin (H&E) staining and positivity for anti-alpha-smooth muscle actin (60).Thrombus size and RBC proportion were measured using H&E staining, fibrin was detected using phosphotungstic acid-hematoxylin staining, and platelets were subjected to immunohistochemical staining for CD42b.The density of MC/MPs was determined via staining for CD163, and the extent of NETosis was determined using H3Cit staining.A thrombus was considered OPN-high when the aggregation of OPN+ cells was observed.
In addition, four samples were subjected to double-staining with anti-OPN and antineutrophil elastase antibodies or with anti-OPN and anti-CD163 antibodies.

Single-cell RNA sequencing data analysis of CTEPH thrombi
We obtained publicly available single-cell RNA sequencing data of thrombi collected from patients with CTEPH, which were deposited by Rajagopal et al. (PRJNA929967) (33).Filtered matrices were loaded into the R package Seurat (v.5.0) (61), and cells with less than 200 features, more than 9,000 features, less than 300 UMIs, more than 60,000 UMIs, or more than 15% mitochondrial gene fractions were removed.Normalization was performed using the R package sctransform (62), followed by the integration workflow of Seurat.Principal component analysis was performed on the integrated data, and the top 30 principal components were used to cluster the cells.The FindNeighbors and FindClusters functions in Seurat were used to identify cell clusters with a resolution of 1.0.The FindAllMarkers function in Seurat was used to identify the markers for each cluster.Clusters were manually annotated using known lineage markers and several clusters were combined as needed.MC/MPs were defined as cells with a high expression of CD14 or CD163.MC/MPs were subjected to subclustering using a resolution of 0.2.DEGs from each subcluster were calculated using the FindAllMarkers function, and the top 10 upregulated genes were used for the DoHeatmap function.
Pathway enrichment analysis was performed using the enrichGO function in the R package clusterProfiler (63).We set the adjusting P method to FDR, the threshold to 0.05, the minimum gene set size to 10, and the maximum gene set size to 600.Pathways that included fewer than three genes were excluded.The top 40 pathways with the smallest P values were visualized using the Emapplot function.The bound pathways were automatically clustered using the Emapplot function.
The receptor-ligand interaction between each subcluster of MCs/MPs and fibroblasts was analyzed using CellChat (v.2.1) (64).Normalized data were used for each condition, and cell types with fewer than 10 cells were excluded.Interaction strength refers to the probability of communication between a given ligand and receptor.
It was calculated as the degree of cooperativity/interactions derived from the law of mass action and the degree to which the ligands and receptors are expressed.

Single-cell RNA sequencing data analysis of cells from murine venous walls
We obtained single-cell RNA sequencing data from cells of the vein wall of a mouse DVT model, which were deposited by Zhou et al. (PRJNA916965) (65).Cells with less than 800 features, more than 8,000 features, less than 1000 UMIs, more than 40,000 UMIs, or more than 20% mitochondrial gene fractions were removed from the analysis.Data analysis was performed as above, except that the top 15 principal components were used and that clustering was performed with a resolution of 0.2.Identified MCs/MPs were subjected to subclustering using the top 15 principal components and a resolution of 0.6.represents the number of shared genes.The pathways are manually clustered and annotated.(B) Protein-protein interaction network.Genes belonging to the top 4 clusters out of the 963 upregulated genes identified in the STRING database are visualized.
Clustering was performed using MCL and each cluster is represented by different colors.
The thickness of each edge represents the level of confidence.Hub genes are identified through their Matthews Correlation Coefficients and are highlighted in bold circles.
Gene set enrichment analysis results according to each cluster are shown as a table on the right and are arranged according to strength.Cluster annotation was performed manually.FDR, false discovery rate.

Fig 5 .
Fig 5. Associations between clinical characteristics and OPN expression in the

Fig. 6 .
Fig. 6.Analysis of single-cell RNA sequencing data of thrombi from patients with