Transcriptome Analysis Reveals Non-Foamy Rather than Foamy Plaque Macrophages Are Pro-Inflammatory in Atherosclerotic Murine Models

Abstract

Rationale:

Monocyte infiltration into the subintimal space and their intracellular lipid accumulation are the most prominent features of atherosclerosis. To understand the pathophysiology of atherosclerotic disease, we need to understand the characteristics of lipid-laden foamy macrophages in the subintimal space during atherosclerosis.

Objective:

We sought to examine the transcriptomic profiles of foamy and non-foamy macrophages isolated from atherosclerotic intima.

Methods and Results:

Single-cell RNA-sequencing analysis of CD45⁺ leukocytes from murine atherosclerotic aorta revealed that there are macrophage subpopulations with distinct differentially expressed genes involved in various functional pathways. To specifically characterize the intimal foamy macrophages of plaque, we developed a lipid staining-based flow cytometric method for analyzing the lipid-laden foam cells of atherosclerotic aortas. We employed the fluorescent lipid probe BODIPY493/503 and assessed side-scattered light (SSC) as an indication of cellular granularity. BODIPY^hiSSC^hi foamy macrophages were found residing in intima and expressing CD11c. Foamy macrophage accumulation determined by flow cytometry was positively correlated with the severity of atherosclerosis. Bulk RNA-seq analysis showed that compared with non-foamy macrophages, foamy macrophages expressed few inflammatory genes but many lipid-processing genes. Intimal non-foamy macrophages formed the major population expressing interleukin-1β and many other inflammatory transcripts in atherosclerotic aorta.

Conclusions:

RNA-seq analysis of intimal macrophages from atherosclerotic aorta revealed that lipid-loaded plaque macrophages are not likely the plaque macrophages that drive lesional inflammation.

INTRODUCTION

Atherosclerosis is a chronic inflammatory disease and the leading cause of cardiovascular disease, which is a major contributor to mortality in Western countries¹. Atherosclerosis can lead to myocardial infarction, ischemic stroke, renal impairment and aneurysms with hypertension². An especially prominent feature of atherosclerotic lesions is the accumulation of lipids within cells in the aortic wall. As atherosclerosis progresses, macrophages (MØs) are recruited to the intima and take up modified low-density lipoprotein (LDL) particles via their scavenger receptors, leading to the progression of atherosclerotic lesions^3,4. Intracellular accumulation of lipid molecules results in the formation of lipid-laden cells called foam cells, which comprise the major cellular component of atherosclerotic lesions¹.

The number of foam cells in a lesion increases as atherosclerosis develops and decreases as regression occurs^5–7. Foam cells, especially MØ foam cells, participate in all developmental phases of a lesion (i.e., formation, maintenance and rupture) and affect the status of a lesion through cellular responses, such as cytokine secretion, apoptosis and necrosis^{8, 9}. Therefore, an important part of atherosclerosis research is examining the quantitative and qualitative changes in foam cells. MØ foam cell formation has been shown to be initiated by bone marrow-derived MØ (BMDM) infiltration that follows monocytosis or extramedullary hematopoiesis¹⁰. In steady-state mouse blood vessels, a small number of MØs reside in the intimal area, while the majority of such cells reside in the adventitia¹¹. Aortic MØs arise from CX3CR1⁺ embryonic precursors (yolk sac MØs) and bone marrow-derived monocytes that were briefly introduced in the postnatal period¹². Under steady state, there is no significant influx of MØs from blood monocytes¹¹. During lesion formation, however, the intima experiences infiltration/fixation of incoming monocytes and activation of monocyte proliferation¹³ caused by hypercholesterolemia and its inflammatory response^{11, 14–16}. Infiltration depends on the expression of CCR2, CCR5, and CX3CR1 on monocytes^{14, 15} and various adhesion molecules on endothelial cells¹⁷. MØs take up excess cholesterol; which is converted to cholesteryl ester that gathers into cytosolic lipid droplets to form foamy MØs¹⁸. Cholesterol accumulation induces MØs to undergo inflammatory responses¹⁹. Furthermore, decreased cellular cholesterol efflux also causes inflammatory responses and inflammasome activation²⁰. However, lipid accumulation in MØs was associated with suppression, rather than activation, of inflammatory gene expression²¹. Therefore, considering the phenotypical plasticity of MØs, further characterization of plaque MØs according to their lipid content is required.

To investigate lipid-enriched atherosclerotic lesions, researchers have used various staining methods. For example, the combination of oil red O staining and immunofluorescence-based analysis is commonly used to examine aortic sections and entire aortas²² and has come to be regarded as the standard approach for evaluating atherogenicity and characterizing foam cells in atherosclerotic lesions under different experimental conditions. However, it is difficult to define the cellular components of a lesion, as such characterization would require the analysis of numerous cellular biomarkers. For example, some markers (e.g., CD11c, MHCII, and CD11b) are shared by dendritic cells (DCs) and MØs, making it difficult to distinguish these cell groups. Thus, technically advanced, simpler, and more sensitive methods that combine objective measurements of lipid-rich atherosclerotic lesions with multiple cellular analyses are required. To overcome the limitations inherent to immunohistochemistry, multi-parametric flow cytometry has been used to analyze aortic immune cells from atherosclerotic mice^23–25.

Previously, BODIPY493/503, a fluorescent lipid probe, was shown to stain cytosolic neutral lipids of cultured foamy MØs²⁶. We employed BODIPY493/503 and assessed side-scattered light (SSC) as an indication of cellular granularity to detect atherosclerotic foamy MØs in flow cytometric analysis. Before analyzing the characteristics of foamy MØs from atherosclerotic aortas, we first used single-cell RNA sequencing (scRNA-seq) to identify cell subtypes by their transcriptomic signatures and confirmed the heterogeneity of MØs in atherosclerotic aorta. We then characterized and sorted intimal foamy and non-foamy MØs from atherosclerotic aorta, using a newly established a lipid staining-based flow cytometric protocol. Finally, we used bulk RNA-seq to compare the transcriptomes of isolated foamy and non-foamy MØs.

METHODS

All expanded descriptions regarding animal models and methods are described in the online SUPPLEMENTAL MATERIAL.

The authors declare that all supporting data are available within the article and its Online Data Supplement files. All sequencing data sets in this article are deposited in international public repository, Gene Expression Omnibus (GEO), under accession ID GSE116239 for bulk RNA sequencing and GSE116240 for single cell RNA sequencings from mouse atherosclerotic aorta.

RESULTS

scRNA-seq reveals MØ subpopulations in murine atherosclerotic aorta.

To investigate heterogeneity of leukocytes in whole atherosclerotic aorta, we first performed unbiased classification of leukocytes present in whole atherosclerotic aorta by scRNA-seq analysis. We extracted live Propidium iodide (PI)⁻ CD45⁺ leukocytes from mouse (Ldlr^−/−) atherosclerotic vessels and used scRNA-seq to determine the leukocyte subpopulations according to gene expression patterns (GSE116240, Ldlr^−/− aorta dataset). Unsupervised graph-based clustering algorithm was used to define 12 clusters according to their gene expression profiles (Figure 1A). Expression of some genes was common among the clusters, but heterogeneity was evident from differentially expressed genes (DEGs) of which separated the clusters on the t-distributed stochastic neighbor embedding (t-SNE) plots (Figure 1B, Online Figure I A, Online Table I, and Online Data I). Cluster 11 was excluded from further analysis, as it was a non-leukocyte population (data not shown).

Figure 1. Single-cell (sc) RNA-seq reveals MØ subpopulations in murine atherosclerotic aorta.

(A) Left, scRNA-seq of CD45⁺ cells isolated from pooled whole aortas of Ldlr^−/− mice (n = 6) fed a WD for 12 weeks. Dimensionality reduction and identification of clusters of transcriptionally similar cells were performed in an unsupervised manner using Seurat package. Right, heatmap showing the top 30 differentially expressed genes for each leukocyte cluster. Normalized gene expression is shown. (B) Expression of principal hematopoietic markers in the 11 identified cell clusters shown as a t-SNE plot with colors corresponding to expression levels or shown as a distribution of gene expression levels in clusters.

To identify cells in the 11 leukocyte clusters, we examined well-known immune cell markers. First, MØs showed the largest cell numbers and had the most diverse subpopulations (Figure 1B) among CD45⁺ aortic leukocytes. Clusters 0–5, 7, and 8 exhibited high-level expression of MØ marker genes, such as those encoding Cd64 (Fcgr1) and Mafb (Figure 1B). Four MØ clusters (0, 3, 5, and 7) expressed high levels of Cd206 (Mrc1) while Lyve1 was highly expressed in clusters 0, 3, and 5 suggesting that these populations may be resident adventitial MØs (Figure 1B). Itgax was highly expressed in cluster 4 (Figure 1B). Next, we performed gene set enrichment analysis to observe functional enrichment of DEGs (all enriched genes and pathways are listed in Online Data I and II, Online Figure I, and Online Table I). Cluster 0 and 3 showed upregulated expression of genes related to endocytosis pathways. Cluster 1 showed increased expression of inflammatory genes related to nuclear factor kappa B (NF-κB), tumor necrosis factor (TNF), interleukin (IL)-17, cytokine and Toll-like receptor (TLR) signaling pathways. Cluster 2 genes were involved in DNA replication and ribosome pathways. Cluster 4 contained highly expressed genes associated with metabolic pathways including cholesterol metabolism, oxidative phosphorylation, and peroxisome proliferator-activated receptor (PPAR) signaling pathway. Cluster 5 contained inflammatory genes involved in cytokine/chemokine pathways similar to the gene expression pattern of cluster 1. Cluster 7 showed upregulated expression patterns in interferon-stimulated genes, including Mx1, Oasl1, and STAT1/2, suggesting interferon-responsive MØs (Online Table I). Cells in cluster 8 showed highly enriched cell cycle-related genes, such as Ccna2, Cdk1, and Cdk4, suggesting proliferating MØs.

Cluster 9 was enriched for T cell receptors (CD3e/d/g) and T cell subset markers (CD8a, and Foxp3), indicating that this cluster represents T cells (Figure 1B, Online Figure I A, and Online Table I). Clusters 6 and 10 showed high-level expression of genes required for DC differentiation, such as Flt3 and Zbtb46 (Figure 1B, and Online Table I). CD11c (Itgax) and some MHC II molecule (H2-DM/-O)-encoding genes (Figure 1B) were also expressed those clusters, and cluster 10 was the only group in which the gene for CD103 (Itgae) was expressed (Online Figure I A, and Online Table I), indicating that cluster 6 and 10 are classical DC2 (cDC2) and cDC1, respectively.

Lipid staining-based flow cytometric method identifies lipid-laden foam cells of atherosclerosis.

To confirm and characterize the cells of the scRNA-seq-identified cluster that appeared to represent intimal foamy MØs, we set out to develop a new flow cytometric method for detecting lipid-laden cells in aortic single-cell suspensions. To first examine the use of the neutral lipid stain BODIPY493/503 in atherosclerotic aortas, we performed en face co-staining using oil red O and BODIPY493/503 (Figure 2A) and found that the atherosclerotic lesions formed at aortic intimal surfaces were specifically stained by both oil red O and BODIPY493/503.

Figure 2. BODIPY493/503-based lipid staining and flow cytometry define lipid-laden cells from atherosclerotic aortas.

(A) Lipid staining of atherosclerotic lesions. The lesions were first stained with BODIPY493/503 (green) and imaged, then subsequently stained with oil red O and imaged (red). (B) Gating strategy of lipid probe-based flow cytometry for detecting aortic foam cells. (C) Live/dead staining using Zombie Aqua showed that most autofluorescent cells were dead. (D) Cells with high granularity (SSC^hi) were strongly stained with BODIPY493/503. Aortic cells from B6, normal diet (ND)-fed Ldlr^−/−, Western diet (WD)-fed Ldlr^−/−, and ApoE^−/− mice were analyzed. This result is representative of at least three independent experiments. (E) Aortic SSC^hiBODIPY^hi cells were found in atherosclerotic intimal tissue but not in adventitia or normal aorta. The separation of adventitia from aorta was performed by partial enzyme digestion. SSC^hiBODIPY^hi foam cells were counted by flow cytometry (n = 9 per group).

To examine whether BODIPY493/503 can be used in flow cytometry to detect lipid-laden foam cells, we stained single-cell suspensions of mouse atherosclerotic aortas with BODIPY493/503 and Zombie Aqua viability dye (Figure 2B, C). After gating out the autofluorescent dead cells and debris, we found that aortic cells have higher granularity (SSC^hi) and positivity for BODIPY493/503 staining (BODIPY^hi; Figure 2B). These SSC^hiBODIPY^hi cells were only found in atherosclerotic aortas from western diet (WD) fed Ldlr^−/− and ApoE^−/− mice (Figure 2D). To define the location of SSC^hiBODIPY^hi cells in the atherosclerotic aorta, aortic adventitia was separated by partial enzyme digestion and peeling (Online Figure II). SSC^hiBODIPY^hi cells were not found in the adventitia of atherosclerotic aorta and normal aorta (Figure 2E). The number of SSC^hiBODIPY^hi cells was positively correlated with the anatomical distribution of atherosclerotic lesions (Online Figure III).

Assessment of atherosclerosis in mouse models, using lipid probe-based flow cytometry.

We next examined whether atherosclerosis can be quantitatively assessed by our flow cytometric approach in Ldlr^−/− and ApoE^−/− mice fed a WD for 4, 8, or 12 weeks. In WD-fed ApoE^−/− and Ldlr^−/− mice, total cholesterol and LDL levels were dramatically increased (Online Figure IV A). As shown in Online Figure III B, atherosclerotic lesions in the aorta were marginal at 4 weeks but markedly increased at after 8 and 12 weeks. In flow cytometric analysis, the number of SSC^hiBODIPY^hi foam cells in these mice gradually increased (Figure 3A, B). We compared atherosclerosis assessment using our flow cytometric method with the classic measurement of atherosclerotic lesions by oil red O staining and found that the two methods were comparable in assessing the severity of atherosclerosis (Figure 3A, B). The aortic foam cells were divided into CD45⁺ and CD45^– populations (Online Figure V A). Of the total foam cells in ApoE^−/− and Ldlr^−/− mice fed a WD for 12 weeks, 90.5 ± 1.6% and 80.8 ± 5.5% were CD45⁺, respectively. A prolonged WD decreased CD45⁺ cells to 67.2 ± 6.2% of total foam cells. In CD45⁺ foam cells, 98.7 ± 0.5% (ApoE^−/−; 12 weeks WD), 96.9 ± 2.5% (Ldlr^−/−; 12 weeks WD), and 93.9 ± 2.5% of cells (Ldlr^−/−; 33 weeks WD) were CD11b⁺CD64⁺ (Online Figure V B). We then compared the time-dependent changes in abundance of CD11b⁺CD64⁺SSC^hiBODIPY^hi foamy MØs and total aortic MØs (Figure 3C–E). Whereas the number of foamy MØs was markedly increased in WD-fed mice (up to 100–1000-fold), the total number of aortic MØs only showed a slight change (Figure 3D, E, left). We further assessed atherosclerotic lesions in a transgenic mouse model expressing a gain-of-function D374Y mutant form of human proprotein convertase subtilisin/kexin type 9 (D374Y-hPCSK9) with elevated plasma total cholesterol, triglyceride levels and atherosclerotic lesions (Online Figure IV A, C)^27–30. The number of SSC^hiBODIPY^hi foam cells was significantly increased in D374Y-hPCSK9 transgenic mice (10,574 ± 3738 cells/aorta) compared with that in wild type (WT) controls (343 ± 136 cells/aorta; Online Figure IV C). During foam cell formation, cellular granularity and size can increase through the accumulation of lipid droplets. Since flow cytometric analysis can be used to monitor the relative granularity and size of cells through analysis of forward side cytometer (FSC) and SSC parameters, respectively, we compared the granularity and size of foam cells in mice fed a WD for 4, 8 and 12 weeks, and divided the foam cells into four groups (Online Figure VI A, B). In ApoE^−/− mice, small foam cells with high granularity (P4) were markedly increased between 4 and 8 weeks (Online Figure VI A). Similarly, Ldlr^−/− mice fed a WD, showed increased levels of small foam cells with high granularity (Online Figure VI B). These results indicate that aortic phagocytic cells take up and accumulate more lipids in their cytosols during atherosclerosis progression.

Figure 3. The frequency of SSC^hiBODIPY^hi cells is positively correlated with severity of atherosclerosis.

(A, B) Comparison of atherosclerosis assessment by en face oil red O staining and lipid probe-assisted flow cytometry in ApoE^−/− (n = 6) and Ldlr^−/− (n = 7) mice. Left, increase in SSC^hiBODIPY^hi foam cells (red, percentage of foam cells; blue, number of foam cells) during a WD for 4, 8 or 12 weeks. Right, similarity between atherosclerosis assessment by flow cytometry and en face oil red O staining. The size of the oil red O stained area (red) and number of foam cells (blue) were normalized to between 0 and 1. *P < 0.01 and **P < 0.001. (C) Gating strategy to identify foamy MØs from whole atherosclerotic aorta.

(D, E) Comparison of abundance changes in aortic foamy MØs and total aortic MØs in ApoE^−/- (n = 6) and Ldlr^−/- (n = 7) mice. Left (dot graph), fold changes compared with cells at the 0-time point (t₀ = 0 weeks; base = 1). The number of foamy MØs (red) dramatically increased during WD. Right (bar graph), percentages of foamy (red) and non-foamy (blue) MØs in total aortic singlets at each time point (4, 8, and 12 weeks). *P < 0.05, **P < 0.01 and *** P < 0.001.

Evaluating the therapeutic potential of anti-atherosclerotic treatments using lipid probe-based flow cytometry.

We examined whether our method can be used to evaluate the therapeutic potential of anti-atherogenic drugs. We treated Ldlr^−/− mice for 10 weeks with atorvastatin or rosuvastatin, which are two representative HMG-CoA reductase inhibitors used to treat atherosclerosis³¹. The administration of statins effectively attenuated atherosclerosis (Online Figure VII) and the number of foam cells significantly decreased by atorvastatin (2454 ± 1109 cells/aorta) and rosuvastatin (2172 ± 704 cells/aorta) treatment compared with that of control Phosphate-buffered saline (PBS)-injected mice (6356 ± 1032 cells/aorta; Online Figure VIII A). Foamy MØ numbers also significantly decreased by atorvastatin (1908 ± 982 cells/aorta) and rosuvastatin (1693 ± 621 cells/aorta) treatment compared with that of PBS controls (4828 ± 802 cells/aorta; Online Figure VIII A). We next analyzed the foam cells of a plaque regression model by injecting ApoE^−/− mice with adeno-associated viral vectors encoding apoE (APOE-AAV)⁶ (Online Figure VIII B, C). The number of foam cells and foamy MØs were markedly decreased in the APOE-AAV group compared with that in the control group (Online Figure VIII B). 2-Hydroxypropyl-ß-cyclodextrin (CD) has been reported to induce the plaque regression by removing cholesterol crystals^32–34, lowering intracellular cholesterol levels in vivo³⁵, and inducing MØ reprogramming and liver X receptor (LXR)-dependent atheroprotection³⁶. Ldlr^−/− mice with advanced atherosclerosis (WD for 25 weeks) were injected with CD (2 g/kg body weight) or PBS (vehicle control) for 8 weeks on a WD. CD treatment did not alter body weight (data not shown) but significantly decreased the number of aortic foam cells (1766 ± 529 cells/aorta) compared with that in PBS controls (3138 ± 1114 cells/aorta; Online Figure VIII D). These results indicate that our flow cytometric method can be used to evaluate the anti-atherogenic potential of drug candidates.

CD45⁺SSC^hiBODIPY^hi foam cells mostly originate from MØs.

Although most CD45⁺ foam cells were CD11b⁺CD64⁺ MØs, we examined whether other immune cells can participate in foam cell formation. As we reported previously³⁷, aortic CD45⁺ immune cells included MØs (CD11b⁺CD64⁺), plasmacytoid DCs (pDCs; PDCA1⁺B220^intLy6c⁺CD11b⁻CD64⁻), DCs (CD64⁻CD11c⁺MHCII⁺), monocytes (CD64⁻CD11c⁻CD11b⁺Ly6c^lo-hi), neutrophils (CD11b⁺ CD64⁻Ly6G⁺), T cells (CD11b⁻CD64⁻CD3⁺), and B cells (CD11b⁻CD64⁻MHCII⁺B220⁺; Figure 4A). Unlike MØs, other immune cells were SSC^lo and BODIPY^lo except very few DCs and monocytes (Figure 4B). In the t-SNE plot produced by fluorescence-activated cell sorting (FACS) data, most CD45⁺ foam cells were found in the CD11b⁺CD64⁺ MØ population (Figure 4C and Online Figure V B). Collectively, these results support the notion that most atherosclerotic foamy leukocytes originate from MØs. Furthermore, SSC^hiBODIPY^hi foamy MØs showed relatively higher levels of CD11c than did non-foamy MØs (Figure 5A). t-SNE analysis confirmed that foamy MØs are CD11c⁺ MØs (Figure 5B). Importantly, the cytosol of aortic SSC^hiBODIPY^hi MØs sorted out from atherosclerotic aorta contained abundant lipid droplets, whereas cytosol lipid droplets in SSC^loBODIPY^lo MØs were scant (Figure 5C). CD11c⁺ foamy MØs were in the atherosclerotic lesion area (Figure 5D) and immunostaining CD206 and CD11c showed that most CD11c⁺ cells were found in the intimal lesion area while CD206⁺ cells were located in the adventitia (Figure 5E, F, Online Video I, and Online Figure II C). When we analyzed MØ populations in CD-treated mice, we found that CD treatment effectively decreased both foamy MØs and non-foamy MØ levels. CD11c⁺ non-foamy MØs were more susceptible than CD206⁺ non-foamy MØs to CD treatment (Online Figure VIII E). We then separated aortic intimal tissue and adventitia and analyzed their MØ populations using flow cytometry (Figure 5G). The adventitia of atherosclerotic aortas contained CD206⁺ MØs (Figure 5E, G), whereas CD11c⁺ MØs were mostly found in intimal tissue (Figure 5G). We next examined the proportional changes of intimal foamy (CD64⁺CD11b⁺SSC^hiBODIPY^hi) and non-foamy (CD64⁺CD11b⁺SSC^loBODIPY^lo) MØs during atherosclerosis progression (Online Figure IX A). Intimal MØs were gradually increased in aortic intima of Ldlr^−/− mice fed WD for 12 weeks and 24 weeks (2001 ± 936 vs 4817 ± 550 cells per aorta, respectively, P = 0.017, Online Figure IX B). The percentage of foamy MØs in total intimal MØs was not markedly increased at 24 weeks compared to 12 weeks (12 weeks, 18.5 ± 11%; 24 weeks, 23.8 ± 4.9%, P = 0.26, Online Figure IX B).

Figure 4. CD45⁺SSC^hiBODIPY^hi leukocytes originate from MØs.

(A) Gating strategy to identify immune cell populations, including MØs, DCs, monocytes, T cells, Tregs, B cells, and neutrophils from whole atherosclerotic aorta. (B) Representative plots of SSC and BODIPY levels from aortic immune cells in atherosclerosis. (C) t-SNE analysis of FACS data. The colored clusters correspond to each leukocyte population. CD45⁺SSC^hiBODIPY^hi foam cells (red) mostly overlapped with the MØ population. The results are representative of at least three independent experiments.

Figure 5. SSC^hiBODIPY^hiCD11c⁺ MØs contain cytosolic lipid droplets and reside in the atherosclerotic intima.

(A) Flow cytometric detection of foamy MØs from Ldlr ^−/− mice fed a WD for 12 weeks. Foamy MØ (red box) CD11c levels were higher than in non-foamy MØs (blue box). These plots represent of data from six atherosclerotic aortas. (B) t-SNE analysis of FACS data. The majority of foamy MØs (red) overlapped with CD11c⁺ MØs (blue). These figures are representative of three experiments. (C) Morphology of aortic SSC^hiBODIPY^hi and SSC^loBODIPY^lo MØs. SSC^hiBODIPY^hi and SSC^loBODIPY^lo cells were sorted from atherosclerotic aortas (n = 3), stained with Hema 3 (left) or oil red O (middle), and analyzed by optical or transmission electron microscopy. (D) Immunofluorescence staining of CD11c and MOMA-2. Lesional CD11c⁺ and MOMA-2⁺ cells (red) were co-stained with BODIPY493/503 (green). (E) Immunofluorescence staining of CD11c (red) and CD206 (blue) in an atherosclerotic aortic arch. CD11c (red) staining was mostly detected in the aortic lesion and some adventitial areas. The green signal is autofluorescence. (F) Whole-mount immunostaining of CD11c in intimal lesions. (G) Comparison of CD206 and CD11c expression in adventitial MØ (blue) and intimal foamy MØs (red).

RNA-seq uncovers distinct gene expression between intima foamy and non-foamy MØs in atherosclerosis.

To analyze the transcriptomic profile of intimal foamy and non-foamy MØs, live (PI⁻) SSC^hi BODIPY^hi foamy MØs and SSC^lo BODIPY^lo non-foamy MØs were simultaneously isolated from adventitia-removed atherosclerotic aortas (Figure 6A). We performed RNA-seq on the two populations (GSE116239) and found that non-foamy and foamy MØs had different degrees of variation in their expressed genes (Figure 6B, C, and Online Data III). Non-foamy and foamy MØs had different numbers of DEGs, with 580 genes for foamy MØs and 748 genes for the non-foamy MØs (adjusted P value [P adj] < 0.05; Figure 6C, D). We then analyzed the expression of genes related to cholesterol and fatty acid transport, cholesterol uptake, and pro- and anti-inflammatory responses and found that the two populations showed markedly different gene expression profiles (Figure 6E, Online Data III, and Online Figure X)³⁸. Furthermore, the expression of some genes related to efferocytosis, such as Cd36, Mertk, and Nr1h3 was found elevated in foamy MØs, whereas genes like Ccl2, Ccr2, and Abca7 were downregulated. The inflammatory genes such as I1b, Nfkbia, Tlr2, and Tnf were mostly upregulated in non-foamy MØs. In contrast, resolving/regression-related genes were upregulated in foamy MØs compared with that in non-foamy MØs (Online Figure X). Next, we compared the gene expression signatures of foamy and non-foamy intimal MØs with those of each scRNA-seq cluster. The top 100 DEGs enriched in foamy MØs were mostly found in cluster 4, whereas many of those from non-foamy MØs were specifically expressed in cluster 1 (Figure 7A and Online Data III). This suggests that scRNA-seq cluster 4 represents the intima foamy MØs and cluster 1 corresponds to intima non-foamy MØs.

Figure 6. Transcriptome profiling reveals distinct gene expression between intimal foamy and non-foamy MØs in atherosclerosis.

(A) FACS sorting of live (PI-) SSC^loBODIPY^lo and SSC^hiBODIPY^hi MØs from the aortic tissues without adventitia of ApoE ^−/− mice (n = 6) with fed a WD for 28 weeks. (B) Principal component analysis (PCA) of variances in the bulk RNA-seq dataset of intima foamy and non-foamy MØs. PCA plot was generated using the 500 most variable genes. (C) Differential expression between foamy and non-foamy MØs represented as a heatmap (genes are sorted by t-statistic). GSEA plots for two representative KEGG pathways (cytokine-cytokine receptor interaction for non-foamy MØs and lysosome for foamy MØs) are shown next to the heatmap. (D) Volcano plot of genes enriched in each cell group. Genes with adjusted P values < 0.05 (red) are arranged by their P values and fold changes (log2) (E) Heatmap of representative genes involved in foam cell formation during atherosclerosis, as assessed by a comparison between intimal non-foamy and foamy MØs. Genes were enriched in each non-foamy (NF) or foamy (F) MØ groups with an adjusted P value < 0.05. The fold change of each gene (log2FC) and adjusted P value (P adj) were shown with the heatmap.

Figure 7. Intimal non-foamy MØs, rather than foamy MØs, are pro-inflammatory.

(A) Top 100 (left) and bottom 100 (right) genes sorted by t-statistic from bulk RNA-seq were used to identify clusters corresponding to foamy and non-foamy MØs in the scRNA-seq dataset. Averaged normalized expression of these gene is shown as a tSNE plot with colors corresponding to averaged normalized expression. (B) Enriched KEGG pathways comparing cluster 1 with cluster 4 in the scRNA-seq dataset (left) or comparing foamy MØs with non-foamy MØs (right). All pathways listed are statistically significant in both comparisons (adjusted P value < 0.01). (C, D) Analysis of IL-1β signaling pathway genes in bulk RNA-seq (C, heatmap) and scRNA-seq (D, plots). The genes were enriched in each NF or F MØ groups (adjusted P value < 0.05) or each cluster (adjusted P value < 0.01; logFC > 0).

Intimal non-foamy MØs, and not foamy MØs are pro-inflammatory.

In gene set enrichment analysis of DEGs, foamy MØs expressed fewer inflammation-related genes than intimal non-foamy MØs, whereas genes related to lipid metabolism and transport pathways, including cholesterol metabolism (P adj = 9.7e-05), and PPAR signaling pathways (P adj = 5.3e-05), such as Abca1, Fabp4, Lipa, and Mertk, were highly expressed (Figure 6E, 7B). Moreover, foamy MØs showed increased expression of genes related to oxidative phosphorylation (P adj = 3.7e-05), lysosome (P adj = 3.7e-05) and proteasome (P adj = 7.8e-04; Figure 7B,). In contrast, intimal non-foamy MØs were enriched in genes involved in inflammatory processes, including cytokine-cytokine receptor interaction (P adj = 2.6e-05), NF-κB (P adj = 3.7e-05), IL-17 signaling (P adj = 2.6e-05), TLR signaling (P adj = 2.2e-04), and TNF signaling pathways (P adj = 2.6e-05; Figure 6E, 7B, and Online Figure XI A).

Recent successful clinical trials on IL-1β targeted therapy prompted us to compare the expression of inflammasome-related genes in foamy and non-foamy MØs³⁹. In bulk RNA-seq analysis, inflammasome-related genes were upregulated in non-foamy MØs rather than in foamy MØs (Figure 7C). Likewise, the expression of Il1β and Nlrp3 was mostly enriched in scRNA-seq cluster 1 (Figure 7D, and Online Figure XI B). Thus, we attempted to analyzed IL-1β mRNA expression in human and mouse atherosclerotic lesions by in situ hybridization. We found that IL-1β mRNA expression was downregulated in foamy CD68⁺ MØs compared to that of non-foamy MØs in human atheroma (Figure 8A, B). Likewise, foam cells poorly expressed IL-1β mRNA in mouse atherosclerotic plaque (Figure 8C). Moreover, quantitative PCR (qPCR) analysis on sorted foamy and intimal non-foamy MØs demonstrated that non-foamy MØs expressed Il1β, whereas foamy MØs did not (Figure 8D). These results suggest that intimal non-foamy MØs, rather than foamy MØs, are inflammatory in atherosclerotic lesions.

Figure 8. IL-1β mRNA expression in human and mouse atheroma.

(A) Localization of IL-1β mRNA in human atheroma (n = 4). CD68 (DAB staining) and H&E staining was performed to identify foamy MØ-rich and fibrous cap regions with MØs. In situ hybridization for DapB and human PPIB were performed as negative and positive controls, respectively. Stars indicate foam cells and arrow heads mark IL1β-stained cells. (B) The percentage of IL-1β mRNA-positive cells in CD68-positive areas of fibrous cap or foam cell rich areas (five different areas, 8.4×10⁴ μm² each) in human atheroma (n=4). (C) IL-1β mRNA expression in mouse atheroma. The stars indicate foam cells and the arrowhead indicates Il1b-stained cells. These figures are representative of three experiments. (D) Single-cell qPCR analysis of IL-1β mRNA expression in foamy or intimal non-foamy MØs sorted from mouse atherosclerotic aorta. Left, mRNA expression was normalized to Gapdh mRNA levels (cells from 4 mice; N.D, not detected). Right, gel analysis of qPCR end products; Il1b and Gapdh.

DISCUSSION

The formation of foam cells and their role in atherosclerosis are important topics of discussion when we seek to understand atherosclerosis pathogenesis and in exploring new therapeutic targets. Immune cells that are recruited to the subintimal space appear to create an inflammatory environment that affects lesion formation and deterioration^{5, 8, 9}. Therefore, the targeting of inflammatory cells is potentially an effective therapeutic regimen for attenuating atherosclerosis. Although statins are currently used to control blood lipid levels as a major treatment for atherosclerosis, a recent clinical trial of IL-1β-targeting antibody showed promising results in attenuating atherosclerosis³⁹. This indicates that atherosclerosis can be treated by an inflammation-regulating drug without controlling blood lipid levels. Clearly, we need to better understand the role of inflammatory cells in atherosclerosis.

Recently, two intriguing studies using scRNA-seq demonstrated the heterogeneity of leukocyte populations in atherosclerotic aorta containing MØs, DCs, T cells, B cells, NK cells and granulocytes^{40, 41}. However, in our scRNA-seq data, MØs were found to compose the largest immune cell population in atherosclerotic aorta, whereas the identification of B cells, NK cells, and neutrophils were failed. In our FACS analysis, MØ was also the largest cell population in aortic CD45⁺ immune cells (Figure 4A, Online Figure XII). Actually, the composition of CD45⁺ leukocyte populations sorted from tissue can vary depending on how the cells are isolated from the tissue, i.e. type and concentration of digestive enzymes, digestion time, and pore size of strainer. Since the aim of this study was to examine the transcriptome of MØ populations in aortic tissue, we used a higher concentration of the enzyme mixture and longer digestion time than previous studies (Online table II). This means that our cell extraction method is more optimized for isolating MØs that are more firmly attached to the aortic tissue than other immune cells. Since relatively small population may not be recovered in scRNA-seq analysis owing to technical limitation, our isolation method appears to be responsible for the difference in immune cell composition between our scRNA data and two previous data.

Various cellular markers for MØs overlap with those of DCs, even though the two cell types are completely different in their functions. For example, CD11c (Itgax)-positive DCs are involved in the formation of early atherosclerotic lesions⁴² while proinflammatory MØs and foamy MØs also express Itgax on the protein and mRNA level. In the absence of a clear method for separating foamy and non-foamy MØs from atherosclerotic aorta, most previous studies used in vitro cultures of MØ cell lines. Although such cultures reasonably mimic the foam cells of atherosclerosis, they do not recapitulate various in vivo microenvironmental factors that affect the formation or pathophysiology of foam cells in atherosclerotic lesions, as cell phenotypes can vary depending on the culture conditions. Indeed, in vivo studies have shown the changes in gene and protein expressions of foam cells directly extracted from the peritoneum of atherosclerotic mice²¹. Laser capture microdissection (LCM) was found to recapitulate the specific propensity of in situ foam cells in lesions⁴³. However, since many marker genes overlap between cell types, this method did not enable researchers to distinguish among the myeloid cells. Moreover, recent reports have indicated that smooth muscle cells (SMCs) can generate lesional foam cells that express some MØ markers⁴⁴. Foamy peritoneal MØs were previously shown to be enriched for molecules involved in physical interaction and vesicular transport rather than lipid metabolism or proinflammatory responses⁴⁵. Lipid-laden peritoneal MØs exhibited up-regulation of genes related to angiogenesis, lipid metabolism, and extracellular matrix organization as well as down-regulation of genes involved in inflammatory sterol metabolism, genes through desmosterol-induced activation of the LXR pathway²¹. In animal experiments, foamy MØs recruited to subcutaneously inserted sponges showed up-regulation of genes related to connective tissue development and function, cell growth, cell proliferation, and cholesterol metabolism (e.g., Abca1, Pparγ, Rxra, Rxrb, and Srebp1)⁴⁶. LCM followed by transcriptome analysis showed that lesional foam cells were enriched for CD68 and CD14, negative for the SMC markers, alpha actin and MYH11, and exhibited increased expressions of TNF, IL-1β, SR-A, ABCA1, and ADFP⁴⁷. In an LCM experiment comparing mice fed a high-cholesterol diet for 2 or 14 weeks, lesional foam cells showed up-regulation of CXCL13 in the 2-week group and up-regulation of GBPS in the 14-week group with no overall change in inflammatory genes expression⁴³. Notably however, these LCM experiments failed to clearly separate lesional foamy MØs from other cell types.

In this study, foamy MØs showed a strikingly reduced expression of inflammatory genes and increased expression of genes related to cholesterol uptake, processing, and efflux whereas intimal non-foamy MØs expressed elevated levels of genes encoding cytokines that are related to the recruitment of leukocytes to lesions and exacerbation of inflammation (e.g., Il1β and Nlrp3), indicating that the inflammasome is activated. These results suggest that newly recruited MØs promote lesion development by producing proinflammatory molecules. Considering the success of a recent clinical trial of an anti-IL-1β antibody canakinumab³⁹, it appears that highly proinflammatory intimal non-foamy MØs may critically contribute to the development and progression of plaque, while foamy MØs may try to defuse plaque progression by exhibiting enhanced lipid uptake and efflux until they undergo apoptotic cell death. To confirm the transcriptomic profiles of lesional MØs and further characterize total foam cell populations containing CD45⁺ and CD45⁻ foam cells, we performed another scRNA-seq analysis (GSE116240, ApoE^−/− aorta dataset) on isolated live total foam cells from atherosclerotic aorta. The gene expression profiles of SSC^hiBODIPY^hi foam cells (n = 809) sorted from pooled atherosclerotic aortas of ApoE^−/− mice (n = 6) were successfully analyzed and the cells were divided into two large groups containing 8 clusters (Online Figure XIII A). Cluster 2, 3, 4, 5, and 6 expressed Ptprc (CD45), and Mafb, indicating these clusters are foamy MØs (Online Figure XIII B). In agreement with our previous data, the clusters of foamy MØs (cluster 2 ~ 6) expressed low levels of inflammatory genes like Il1b, Nlrp3, Tnfsf9, and Nfkb1 (Online Figure XIII C), but showed increased expression levels of Lgals3, Ctsb, Itgax, and Trem2 (Online Figure XIII D). Furthermore, we confirmed the expressions of the top 100 enriched genes of non-foamy or foamy MØs in scRNA-seq data of total foam cells (Online Figure XIV A). Whereas foamy MØ genes were enriched in cluster 2 ~ 6, non-foamy MØ genes were poorly expressed in these clusters. In accordance with our previous data, the functional enrichment analysis on scRNA data of total foam cells demonstrated that foamy MØs had increased expression of genes related to oxidative phosphorylation, cholesterol metabolism, lysosome, and proteasome (Online Figure XIV B). Previously, MØs contributing to lesion formation have been known to be mostly derived from circulating precursors and increase during disease progression⁴⁸. But recently local proliferation of MØs importantly contributes to lesion formation during the atherogenesis^49,50. Interestingly, there was a MØ subpopulation having increased expression of genes related cell cycle and proliferation in our scRNA-seq data of CD45⁺ aortic cells (Figure 1A, cluster 8, and Online table I). In addition, a small foam cell subpopulation expressing high level of cell cycle / proliferation-related genes was also identified in scRNA-seq analysis on total mouse foam cells (Online Figure XV A and B, cluster 6). 5.8 ± 2.8 % of total foamy MØs expressed KI-67 in foam cell-rich area of human atheroma. In fibrous cap area, 9.54 ± 0.54% non-foamy macrophages were Ki-67-positive, which was slightly higher than the percentage of foamy macrophages. (Online Figure XV C). Thus, although macrophages appear to proliferate constantly irrespective of lipid accumulation, proliferation was more during the non-foamy stage. Thus the proliferation of macrophages appears to happen constantly, regardless of their lipid accumulation, but it seems to occur more during non-foamy stage. However, further studies are required to understand the exact origin and characteristics of plaque macrophages. For instance, additional scRNA-seq analysis on lesional MØs isolated from early or advanced plaques would be useful to dissect MØ subsets in lesions and understand the phenotypic changes of MØs during disease progression. Furthermore, fate mapping experiments may be useful for tracing lesional MØs and examining the origin of foamy and non-foamy MØs, i.e. whether foamy MØs are generated mostly from non-foamy MØs or a specific subset of MØs.

In our scRNA-seq data of total foam cells, clusters 0 and 1 expressed Tagln and Acta2, while cluster 7 had high expression levels of Pecam1 (Online Figure XIII B). This result suggests that CD45⁻ foam cells would be derived from SMCs and endothelial cells. Since previous reports suggested that vascular SMCs can also generate foam cells in atherosclerotic plaque^51,52, we stained foam cells with SMC markers to define SMC derived-foam cells. In flow cytometric analysis, we found that foam cells included CD45⁻ cells expressing smooth muscle alpha actin (SMA; Online Figure XVI A). In atherosclerotic plaque, some BODIPY⁺ cells also expressed SM22α, another SMC marker (Online Figure XVI B, C). These results indicate that although a large proportion of foam cells is derived from MØs, SMCs also contribute to foam cell formation.

Collectively, we first established a lipid probe-based flow cytometric method for analyzing foamy MØs of the atherosclerotic intima to further enable transcriptomic profiling of intimal foamy and non-foamy MØs. Our RNA-seq analysis revealed for the first time that MØ subtypes of atherosclerotic aorta exhibit distinct transcriptomic profiles according to their lipid content. Although we showed attenuated IL1β expression in foamy MØs in human atherosclerotic lesions, the transcriptomic profile of human lesional MØs remains to be elucidated. Actually, the preparation of foamy and non-foamy MØs for scRNA-seq or bulk RNA-seq analyses from human atherosclerotic aorta is still highly challenging. But it is very important to demonstrate whether the transcriptomic profiles of human foamy and non-foamy MØs are comparable to those of murine counterparts, which will provide new immunological insights regarding the innate immune responses that take place in atherosclerosis.

NOVELTY AND SIGNIFICANCE.

What Is Known?

• Macrophages process excessive cholesterol and induce inflammatory responses via Toll-like receptors and inflammasome activation during atherogenesis.

• Macrophages are the major contributors of foam cells in atherosclerotic lesions.

• Generation of foamy macrophages suppresses the inflammatory status of cells via activation of the liver X receptor.

What New Information Dose This Article Contribute?

• We developed a straightforward method to readily differentiate foam cells in plaques from other macrophages in the aortic wall.

• The foamy macrophages in atherosclerotic lesions do not express RNAs that are primarily associated with inflammation; instead, they express genes related to lipid processing.

• Intimal non-foamy macrophages form the major population expressing interleukin-1β and other inflammatory transcripts in atherosclerotic aorta.

Macrophages in atherosclerotic plaques internalize excessive lipids, become foamy macrophages, and produce proinflammatory molecules. However, excessive uptake of exogenous lipids has been shown to suppress inflammatory responses in macrophages. Therefore, understanding the characteristics of foamy plaque macrophages is important for understanding their roles in atherosclerosis. In this study, we used single-cell RNA sequencing to unbiasedly classify aortic macrophages in atherosclerosis. We first developed a FACS-based method to distinguish foamy macrophages from other macrophages in atherosclerotic aorta. Then, using this newly developed method, we sorted out the intimal non-foamy and foamy macrophages and performed transcriptomic analysis using RNA sequencing. These two RNA sequencing approaches revealed that while non-foamy macrophages predominantly expressed genes related to immune response, the foamy macrophages mainly expressed genes related to lipid processing instead of inflammation. Similarly, the foamy macrophages in human atheroma showed scant expression of interleukin-1β, an inflammatory cytokine. These results suggest that intimal non-foamy macrophages, rather than foamy plaque macrophages, perform pro-inflammatory roles in the atherosclerotic aorta.

Nonstandard Abbreviations and Acronyms:

APOE-AAV

adenovirus encoding apoE

scRNA-seq

single-cell RNA sequencing

MØs

macrophages

WD

Western diet

P adj

adjusted P value