If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
A comparative gene expression matrix in Apoe-deficient mice identifies unique and atherosclerotic disease stage-specific gene regulation patterns in monocytes and macrophages
Department of Cardiology and Angiology, University Heart Center Freiburg-Bad Krozingen, Faculty of Medicine, University of Freiburg, Hugstetter Street 55, Freiburg, Germany
Department of Cardiology and Angiology, University Heart Center Freiburg-Bad Krozingen, Faculty of Medicine, University of Freiburg, Hugstetter Street 55, Freiburg, Germany
Department of Cardiology and Angiology, University Heart Center Freiburg-Bad Krozingen, Faculty of Medicine, University of Freiburg, Hugstetter Street 55, Freiburg, Germany
Department of Cardiology and Angiology, University Heart Center Freiburg-Bad Krozingen, Faculty of Medicine, University of Freiburg, Hugstetter Street 55, Freiburg, Germany
Department of Cardiology and Angiology, University Heart Center Freiburg-Bad Krozingen, Faculty of Medicine, University of Freiburg, Hugstetter Street 55, Freiburg, Germany
Department of Cardiology and Angiology, University Heart Center Freiburg-Bad Krozingen, Faculty of Medicine, University of Freiburg, Hugstetter Street 55, Freiburg, Germany
Department of Cardiology and Angiology, University Heart Center Freiburg-Bad Krozingen, Faculty of Medicine, University of Freiburg, Hugstetter Street 55, Freiburg, Germany
Department of Cardiology and Angiology, University Heart Center Freiburg-Bad Krozingen, Faculty of Medicine, University of Freiburg, Hugstetter Street 55, Freiburg, Germany
Department of Cardiology and Angiology, University Heart Center Freiburg-Bad Krozingen, Faculty of Medicine, University of Freiburg, Hugstetter Street 55, Freiburg, Germany
Department of Cardiology and Angiology, University Heart Center Freiburg-Bad Krozingen, Faculty of Medicine, University of Freiburg, Hugstetter Street 55, Freiburg, Germany
Department of Cardiology and Angiology, University Heart Center Freiburg-Bad Krozingen, Faculty of Medicine, University of Freiburg, Hugstetter Street 55, Freiburg, Germany
Department of Cardiology and Angiology, University Heart Center Freiburg-Bad Krozingen, Faculty of Medicine, University of Freiburg, Hugstetter Street 55, Freiburg, Germany
Department for Vascular and Endovascular Surgery, Technical University Munich, Arcisstr. 21, Munich, GermanyDeutsches Zentrum für Herz-Kreislaufforschung (DZHK), Berlin, GermanyDepartment of Medicine, Karolinska Institutet and University Hospital, Eugeniavägen 3, Stockholm, SwedenPartner Site Munich Heart Alliance, Arcisstr. 21, Munich, Germany
Department of Cardiology and Angiology, University Heart Center Freiburg-Bad Krozingen, Faculty of Medicine, University of Freiburg, Hugstetter Street 55, Freiburg, Germany
Corresponding author. University Heart Center Freiburg-Bad Krozingen, Dept. of Cardiology and Angiology, Hugstetter Street 55, 79106, Freiburg, Germany.
Department of Cardiology and Angiology, University Heart Center Freiburg-Bad Krozingen, Faculty of Medicine, University of Freiburg, Hugstetter Street 55, Freiburg, Germany
Corresponding author. University Heart Center Freiburg-Bad Krozingen, Dept. of Cardiology and Angiology, Hugstetter Street 55, 79106, Freiburg, Germany.
Department of Cardiology and Angiology, University Heart Center Freiburg-Bad Krozingen, Faculty of Medicine, University of Freiburg, Hugstetter Street 55, Freiburg, GermanyInstitute for Experimental Cardiovascular Medicine, University Heart Center Freiburg-Bad Krozingen and Faculty of Medicine, University of Freiburg, Elsaesser Street 2Q, Freiburg, Germany
Atherosclerosis is a systemic and chronic inflammatory disease propagated by monocytes and macrophages. Yet, our knowledge on how transcriptome of these cells evolves in time and space is limited. We aimed at characterizing gene expression changes in site-specific macrophages and in circulating monocytes during the course of atherosclerosis.
Methods
We utilized apolipoprotein E-deficient mice undergoing one- and six-month high cholesterol diet to model early and advanced atherosclerosis. Aortic macrophages, peritoneal macrophages, and circulating monocytes from each mouse were subjected to bulk RNA-sequencing (RNA-seq). We constructed a comparative directory that profiles lesion- and disease stage-specific transcriptomic regulation of the three cell types in atherosclerosis. Lastly, the regulation of one gene, Gpnmb, whose expression positively correlated with atheroma growth, was validated using single-cell RNA-seq (scRNA-seq) of atheroma plaque from murine and human.
Results
The convergence of gene regulation between the three investigated cell types was surprisingly low. Overall 3245 differentially expressed genes were involved in the biological modulation of aortic macrophages, among which less than 1% were commonly regulated by the remote monocytes/macrophages. Aortic macrophages regulated gene expression most actively during atheroma initiation. Through complementary interrogation of murine and human scRNA-seq datasets, we showcased the practicality of our directory, using the selected gene, Gpnmb, whose expression in aortic macrophages, and a subset of foamy macrophages in particular, strongly correlated with disease advancement during atherosclerosis initiation and progression.
Conclusions
Our study provides a unique toolset to explore gene regulation of macrophage-related biological processes in and outside the atheromatous plaque at early and advanced disease stages.
], manifesting itself locally in atheromatous plaque growth in arterial vessels, but also systemically with increased inflammatory markers linked to elevated cholesterol levels [
]. Atheromatous plaque macrophages propagate disease progression primarily originating from monocytes and with unknown relations to macrophages in other compartments, such as the peritoneal cavity [
]. Single-cell RNA sequencing (scRNA-seq) provided major insights into macrophage subset identification in different tissues, including the atherosclerotic aorta [
], but typically focused on one disease stage and compartment. In this work, our objective was to compare bulk gene expression profiles of blood Ly6Chigh monocytes, aortic macrophages and peritoneal macrophages retrieved from atherosclerosis-prone mice before and during atherosclerosis development at early and advanced disease stages. We aimed to compose a unique database for therapeutic target identification and pathomechanistic investigation in experimental atherosclerosis research.
2. Materials and methods
2.1 Animals and diet
Female Apoe-deficient mice (Apoe−/−, B6.129P2-Apoetm1Unc) were purchased from The Jackson Laboratory (Bar Harbor, ME, USA) and housed under specific pathogen-free conditions. At the age of eight weeks, one group of animals was sacrificed as the baseline group while the remaining mice were placed on a high-cholesterol diet (HCD, 1.25% w/w cholesterol). A second group of animals was sacrificed after four weeks of HCD to represent the early atherosclerotic disease stage, whereas the third group of animals was sacrificed after 24 weeks of HCD and having developed advanced atherosclerosis.
2.2 Histology
Murine aortic roots were embedded in OCT Tissue Tek (Sakura Finetek, Tokyo, Japan) and cut into serial cryostat sections (5 μm) starting at the level of the aortic valve. Sections were stained with primary antibodies anti-CD68 (clone FA-11, BioRad AbD Serotec, Purchheim, Germany), anti-Glycoprotein Nmb (GPNMB) (BS-2684R, Bioss Antibodies Inc., Woburn, Massachusetts, USA), the secondary antibodies rabbit anti-rat AF647 (ab169349, Abcam, Berlin, Germany), alpaca anti-rabbit AF488 (ChromoTek GmbH, Planegg-Martinsried, Germany), and DAPI Mounting Medium (Carl Roth, Karlsruhe, Germany) according to the manufacturers’ instructions. Abdominal aortas were pinned for Oil-red O (ORO, Sigma Aldrich, St. Louis, MO, USA) en face staining. Fluorescent images were recorded with the Axioplan 2 imaging fluorescence microscope (Carl Zeiss MicroImaging GmbH, Göttingen, Germany). Images were analyzed with Image Pro Premier 9.2 (Media Cybernetics, Silver Springs, USA).
2.3 Cholesterol assay and enzyme-linked immunosorbent assay (ELISA)
Murine plasma cholesterol levels were measured using Cholesterol FS 10′ Multi-purpose kit (DiaSys Diagnostic Systems GmbH, Holzheim, Germany) according to the manufacturer's instructions. Murine plasma samples for mouse GPNMB ELISA (ab270892, Abcam Berlin, Germany) were diluted 1:10 and the assay was performed following manufacturer's instructions.
2.4 Stable and unstable human plaque
Human plaque specimens were provided from patients with carotid artery disease undergoing carotid endarterectomy at the Department of Vascular and Endovascular Surgery at the Klinikum rechts der Isar of the Technical University Munich, Munich, Germany. All patients provided written informed consent, and the study has been approved by the local Ethics committee at the Klinikum rechts der Isar of the Technical University Munich. Plaque stability was assessed using the Rothwell/Redgrave criteria based on fibrous cap thickness (>200 μm stable, <200 μm unstable) on hematoxylin & eosin stained slides as previously shown [
Aortic macrophages, peritoneal macrophages, and pooled monocytes from blood and spleen were sorted directly into Buffer RLT (QIAGEN, Stockach, Germany) plus 1% β-mercaptoethanol (AppliChem, Darmstadt, Germany). After further homogenizing the cell lysates with QIAshredder (QIAGEN, Stockach, Germany), we extracted the RNA from the lysates with RNeasy Micro kit (QIAGEN, Stockach, Germany). We prepared RNA library with the Ovation SoLo RNA-seq systems (NuGEN, Crailsheim, Germany). The library was sequenced on a NextSeq instrument with 75 bp paired-end reads using NextSeq 500 High Output v2 kit (Illumina, San Diego, CA, USA). At least 2.5 million reads were acquired from each bulk sample.
2.6 Bioinformatic processing of RNA-seq data
The transformation of fastq files to gene counts was processed in the Galaxy platform (https://usegalaxy.eu/). After ensuring adequate quality using FASTQC (Galaxy Version 1.7), cutadapt (Galaxy Version 1.16.5) was used to trim adapters and end bases with Phred score lower than 20. Applying the criteria, all samples had less than 2% reads trimmed. Trimmed reads with length shorter than 20 bp were discarded. The trimmed reads were aligned to the mouse genome (mm10) using RNA STAR (Galaxy Version 2.7.2b) with default settings. At least 1.7 million uniquely aligned reads were acquired from each sample. FeatureCounts (Galaxy Version 1.6.4+galaxy1) was applied to count the features from the forward stranded bam files using default settings. All samples had at least 1.1 million reads assigned to a feature. The gene count files were downloaded and imported into R (Version 4.0.4) for downstream analyses.
2.7 Bioinformatic analysis of differentially expressed genes, regulation patterns, and gene ontology (GO) enrichment
We utilized dplyr (Version 1.0.7) to compute the gene count files. Expressed genes were defined as genes expressed in at least 3 out of 4 biological replicates. Differentially expressed genes (DEGs) were called by DESeq2 (Version 1.28.1). Significantly regulated genes were genes that increased or decreased significantly as opposed to earlier disease stages. DEGs of each cell type were grouped into early, late, and transitorily regulated genes based on their regulation patterns. The grouped DEGs were used for the enrichment analysis of biological pathways with clusterProfiler (Version 3.16.1). Statistical significance of DEGs and enriched pathways were defined as adjusted p-value <0.05 post-Bonferroni correction.
2.8 Bioinformatic inspection of Gpnmb expression in scRNA-seq datasets
Single-cell gene expression matrices and metadata were retrieved from Gene Expression Omnibus (GEO) dataset GSE131776 or offered by the authors [
]. The Apoe−/− mouse datasets were analyzed jointly with Seurat (version 4.0.3). In short, the raw counts were logarithmically normalized with the scale factor of 10,000; thereafter, the 2000 most variable genes identified with variance stabilization transformation were used in principle component analysis. With RunUMAP, the first 20 principle components were applied for dimensionality reduction. Graph-based clustering was performed by FindNeighbors and FindClusters with a resolution of 0.65 and 1.2 for the GSE131776 and Winkels et al. datasets, respectively [
]. Cell types were appointed to the clusters based on the markers suggested in both publications. The combined matrices and metadata with cell type information were made into a Seurat object and analyzed according to the standard Seurat workflow with all parameters kept the same except that the resolution was fine-tuned to 0.8 for clustering. The two datasets were nicely integrated that the macrophage and T cell clusters from both sources lapped over. For the simplicity of presentation, we dismissed the subcluster annotation of smooth muscle cells and T cells. The macrophage subset was extracted from the Seurat object and underwent the standard Seurat workflow with the aforementioned parameters. Subtypes of macrophage were identified and annotated according to the known markers—inflammatory (Tnf, Il1b, and Ccl3), resident-like (Folr2, Timd4, Lyve1)), and foamy (Lgals3, Cd9, and Trem2)) [
]. We further analyzed the combined macrophage matrix and metadata with FindIntegrationAnchors and IntegrateData functions, following the Seurat v3 Integration workflow. From GSE131776, we acquired the human right coronary artery single-cell dataset and integrated the data by RunHarmony (v1.0) to remove the inter-individual variables. Thereafter, we used the same analysis workflow and settings that we applied to the mouse datasets. Comparison of gene expression in Gpnmb+ and Gpnmb− foamy macrophages was performed using permutation test. From GSE159677, we acquired single-cell macrophage data of human carotid plaque retrieved from atherosclerotic core and proximal adjacent tissue. The data was analyzed using the same workflow applied to GSE131776 dataset.
2.9 Data visualization
All cell types acquired at different disease phases were clustered based on principle component variances and visualized with plotPCA of DESeq2 package (Version 1.28.1). Number of genes present in various compartments throughout the disease course was visualized by Sankey diagram using plotly (Version 4.9.4.1) and by Venn diagram using colorfulVennPlot (Version 2.4). The color scale of grid plot was made with ggplot2 (Version 3.3.5). Logarithmic mean counts of highly differentially regulated genes were shown in heatmaps with pheatmap (Version 1.0.12). Graphs of single-cell datasets were programmed with Seurat (Version 4.0.3). The dimensional reduction plots were generated with DimPlot. The Expression of features and Gpnmb/GPNMB was visualized by FeaturePlot and VlnPlot.
2.10 Data availability
The datasets supporting the conclusions of this article are available in the GEO database repository, GSE213189, GSE131776, GSE159677 and GSE213189, and enclosed within the article and its Supplementary Table A.2.
3. Results
3.1 Compartment-determined myeloid cell transcriptomic profiles in Apoe-deficient mice
At eight weeks of age with modestly elevated cholesterol levels on a chow diet, aortas from Apoe−/− mice were free of atherosclerotic lesions in en face lipid staining of the aorta. One month after the start of HCD, we observed an 7.7% (95% confidence interval (CI): 5.7%–9.0%) lipid coverage in the aortic inner layer, which indicated an early formation of atheroma. Atherosclerosis advanced as HCD feeding was continued for another 5 months accelerating hypercholesterolemia and covering about 30.7% (95% CI: 29.3%–31.4%) of the aortic surface with lipid-rich plaques (Fig. 1A–C).
Fig. 1Severity of atherosclerosis of the studied mice and the isolation of monocytes and macrophages for bulk RNA-seq.
(A) Representative pictures of ORO-stained aorta of mice undergoing zero, one, and six months of HCD feeding. (B and C) Quantification of lipid deposition in aorta by area covered by ORO, and of total cholesterol in plasma. Results are presented as mean ± SEM, *p < 0.05 denotes statistically significant differences between groups, n = 4 per group, One-way ANOVA with Holm Sidak Post-hoc testing. (D) Representative dot plots of flow cytometry bulk-sorting gates. The framed area in each plot encircles sorted cells with their respective mean percentages ± SEM. 451.8 ± 115, 1688 ± 337.8 and 5912 ± 1836, *, §aortic macrophages were sorted at baseline, early and advanced disease stages, respectively. Counts are presented as mean ± SEM, n = 4 per group, *, §p-value < 0.05 denotes statistically significant differences between baseline and early (*), and early and advanced (§) groups, One-way ANOVA with Holm-Sidak post-hoc testing. For blood/splenic monocytes and peritoneal macrophages, 50.000 cells each were sorted. Four mice were used per condition.
To understand the transcriptomic regulation of atheroma-associated macrophages, we performed bulk RNA-seq of F480high macrophages isolated from thoracoabdominal aortas of mice representing different stages of atherosclerosis: baseline (absence of disease), early and advanced disease. From the same animals, we collected another two cell populations as comparators: First, pooled Ly6Chigh monocytes isolated from blood and spleen, which represent major precursor cells of atheromatous plaque macrophages that proliferate in situ [
]. Second, peritoneal macrophages which settled in a compartment distant from the aortic lesion. These cells are still exposed to systemic hypercholesterolemia and regularly being used for foam cell formation studies [
]. For the simplicity of presentation, and given that gene expression profiles of splenic reservoir monocytes overlap with those of circulating blood monocytes [
], the source of monocytes is labeled as blood in the rest of the text. The three myeloid cell populations sorted from four Apoe−/− mice per group at the three time points of atherosclerosis development were subjected to RNA-seq (Fig. 1D, Supplementary Fig. 1).
The RNA-seq results revealed an overall high-quality data with a mean Phred score of above 25 for all nucleobases of the trimmed reads. Raw sequencing reads were converted to gene counts following genome alignment and feature counting. With principal component analysis, we identified three clusters which were distinctively separated according to the compartments where the cells were retrieved from. Sub-grouping associated with disease stages was observed in macrophages in particular (Fig. 2A, Supplementary Fig. 2). The finding suggested that the compartmentalized environment played a dominant role in shaping cell biology despite systemic exposition to hypercholesterolemia throughout atherosclerotic disease development. Yet, comparing gene expression profiles between all three cell types at three disease stages identified hundreds to a few thousand DEGs even within the same compartment (Supplementary Fig. 3). We therefore asked: to what extent a common hypercholesterolemic challenge would incite similar or unique gene expression changes in the three myeloid cell populations during atherosclerosis initiation and progression.
Fig. 2Gene expression and regulation according to disease stages and physical compartments of monocyte/macrophages.
(A) Principle component analysis (PCA) plot of all samples undergoing bulk RNA-seq. (B) Sankey plot depicting quantitative changes of expressed genes in monocytes or macrophages of single or shared compartments at different stages of atherosclerosis. (C)Venn diagrams showing the number of expressed genes (upper) and DEGs (lower) in monocytes or macrophages of single or shared compartments at different stages of atherosclerosis. A, aorta; B, blood; DEG, differentially expressed gene; P, peritoneum; n = 4 per group.
3.2 Shared and differentially regulated genes in monocytes and macrophages during atherosclerosis
The studied monocytes and macrophages of all disease stages expressed overall 14362 genes among 24421 annotated genes. Most of the genes were commonly expressed by all three cell types, irrespective of disease stages, accounting for the housekeeping genes required for basic cellular functions (Fig. 2B and C). At baseline, only few genes were expressed exclusively in aortic macrophages, whereas thousands of genes were expressed uniquely or shared in blood monocytes and peritoneal macrophages. During the course of atherosclerosis, however, we observed a continuous increase of genes expressed in the aortic macrophage cell population; in contrast, the number of genes exclusively present in blood monocytes or peritoneal macrophages declined, many of which were later commonly shared by aortic macrophages. These observations are in line with the prevailing concept that infiltrating monocytes are the primary source of atheromatous plaque macrophages (Fig. 2B). At the advanced disease stage, 609 genes were uniquely expressed in aortic macrophages. Testimony to a compartment-specific transcriptomic regulation, this number exceeded those in monocytes and peritoneal macrophages (Fig. 2C). The compartment-specific transcriptomic regulation was even better illustrated by the number of DEGs during atherosclerosis development (Fig. 2C). Supplementary Fig. 4 illustrates how the number of genes in the Venn diagrams were counted.
Hundreds of DEGs were uniquely regulated in each of the three compartments, whereas only a maximum of 41 DEGs were commonly shared (Fig. 2C, Supplementary Table A.1). The number of DEGs at different disease stages also varied noticeably between the compartments (Fig. 2C). In both macrophage populations, more genes were upregulated than downregulated early during atherosclerosis development; whereas in blood monocytes or during disease progression, up- and downregulation of genes was more evenly distributed (Supplementary Fig. 5A). Supplementary Fig. 5B shows a selected panel of genes that were highly expressed in a stage-specific manner, exemplifying variable patterns of gene regulation, which show consistent or transient up- or downregulation during disease progression. Next, we categorized all DEGs according to their regulation patterns as being regulated early, transitorily or late during atherosclerosis development to explore disease stage- and compartment-specific functions based on gene expression changes.
3.3 Monocyte/macrophage populations feature divergent gene regulation patterns during atherosclerosis
Fig. 3 summarizes the overall regulation pattern of each gene in each compartment, which are listed in detail in Supplementary Table A.2. Aortic macrophages carried 11845 expressed genes, among which 3245 genes were differentially regulated throughout the disease course. To our surprise, although we identified thousands of DEGs in aortic macrophages, only ten of them were continuously upregulated during atheroma initiation and progression. All of these genes were lowly expressed or downregulated in blood monocytes and peritoneal macrophages, demonstrating a unique gene regulation in the atheroma lesion. The majority of DEGs in macrophages (n = 1729) isolated from aortas with early atherosclerotic lesions showed an increased expression which did not change significantly during disease progression. Conversely, 80 genes were downregulated early on and remained suppressed in macrophages from advanced atheroma. Together, more than half of the DEGs in aortic macrophages were regulated early during atherogenesis. Twenty-two percent of aortic macrophage DEGs were regulated transitorily, meaning that they showed either an increased (n = 207) or decreased (n = 508) expression level at the early disease stage compared to baseline, which was subsequently reverted as no significant gene expression change was detected at the advanced disease stage relative to baseline. Another twenty-two percent of aortic macrophage DEGs were regulated late, meaning that they showed an increased (n = 450) or decreased (n = 135) expression level at the advanced but not the early disease stage compared to baseline. Only 126 aortic macrophage DEGs were significantly regulated during progression of atherosclerosis from early to the advanced disease stage. In contrast, 43% of peritoneal macrophage DEGs and 52% of blood monocyte DEGs featured a late regulation pattern (Fig. 3).
Fig. 3Frequency of regulation patterns of the DEGs and the enriched biological processes based on disease stages.
Bar charts showing the frequency of regulation patterns of the DEGs of each monocyte or macrophage compartment. Colored arrows indicate the regulation direction and the statistical significance, based on which we could assign each DEG a regulation pattern. The regulation patterns were further categorized into three groups, regulated early, late, and transitorily. Genes belonging to the same group of each monocyte or macrophage compartment were used for biological process enrichment analysis. Statistically enriched biological processes that were uniquely identified in a specific disease stage were listed in the shaded boxes right next to the gene group. Statistical significance of DEGs and enriched pathways were defined as adjusted p-value < 0.05 post-Bonferroni correction (n = 4 per group). Number in the brackets next to the GO terms indicate the ratio of DEGs to all genes composing the GO term.
Thus, during atherogenesis, nearly 80% of gene expression changes in aortic macrophages occur within 4 weeks of disease onset in Apoe−/− mice on HCD. About a quarter of these regulated genes reverted their expression changes during further disease progression. Blood Ly6Chigh monocytes regulated hundreds of genes as cholesterol started to accumulate in the circulation, but most of their expression changes occurred gradually over time and reached statistical significance at the advanced disease stage only. In peritoneal macrophages, which situated distant from the atherosclerotic aorta, most genes were regulated transitorily. In aortic macrophages, however, the most frequent regulation pattern was an early upregulation of DEGs which remained uplifted during disease progression. In peritoneal macrophages, which situated distant from the atherosclerotic aorta, transitory regulation was the most frequent in this cell type, whereas in aortic macrophages, the most frequent regulation pattern was an early upregulation of DEGs which remained uplifted during disease progression. As shown in Fig. 2c, only 41 out of 4290 genes were regulated in common by all three cell populations during initiation of atherosclerosis, and nearly none later on. Notably, gene regulation patterns differed even in those few shared DEGs between compartments (Supplementary Table A.1). Given the obvious differences in gene regulation patterns between the cell populations and considering the small overlap in shared DEGs, we interrogated the functional implications of gene expression changes in each compartment.
3.4 Compartments determine monocyte and macrophage functions in atherosclerosis
DEGs grouped as regulated early, transitorily and late were enriched in variable biological processes according to GO term analysis. Our data revealed that in the aorta, DEGs regulated early enriched biological processes that related to general cellular responses to stress, such as apoptosis and autophagy, as well as to extracellular matrix remodeling of the plaque microenvironment with collagens and metalloproteinases. Common macrophage features, such as inflammation, lipid handling, and macrophage activation were manifested by DEGs regulated transitorily or late (Fig. 3). On the contrary, biological processes associated with immune activation were enriched in Ly6Chigh monocytes early on, where leukocyte migration and myeloid cell differentiation were key procedures that monocytes require to infiltrate and form atheroma lesions. Peritoneal macrophages, distant from the lesion site, had a mild response during atheroma initiation shown by the low number of early-regulated DEGs with enriched biological process unspecific to immune reactions. However, a systemic impact was revealed by their late-regulated DEGs, which involved NFκB signaling, macrophage activation, and TNF family cytokine production (Fig. 3).
The gene regulation patterns and enriched biological processes suggest that compartments determined monocyte and macrophage reactions in a disease stage-specific manner. During atheroma initiation, circulating monocytes prepared to migrate and differentiate. At the same time, macrophages in the aorta showed cellular response to stress as well as secretory factors for extracellular matrix organization, an essential atherogenic procedure which promotes cell proliferation and migration [
]. All three cell types showed inflammatory reactions by genes regulated transitorily or late, revealing a sustained systemic inflammation during atheroma progression. Nevertheless, we observed a more intensive immune response in the atheroma lesion as opposed to the systemic compartments (Supplementary Fig. 6).
Compared to circulating monocytes and peritoneal macrophages, aortic macrophages had higher expression of genes associated with cell proliferation and immune activation, such as cytokines and toll-like receptors (Fig. 4A). In general, monocytes had lower expression of chemokines and genes related to lipid handling than macrophages did. Having said that, aortic and peritoneal macrophages showed a distinctive regulation of phagocytic receptors. For instance, aortic macrophages had a higher expression of Trem2, whereas peritoneal macrophages expressed more Msr1 (Fig. 4A). Both of these two genes function in lipid uptake and thus contribute to atherogenesis [
]. However, our data illustrated a distinctive expression and regulation of these two genes in compartmentally segregated macrophages from the same atherosclerotic organisms, elucidating macrophage reactions depending on their microenvironment. Indeed, although macrophage activation (GO:0042116) was an enriched biological process in all three investigated cell types, only few involved genes were actually shared by all cell types. Fig. 4B illustrates the DEGs and their regulation patterns in the three cell types, presenting disparities of gene regulation between monocytes and macrophages from different compartments. In addition, most of genes involved in aortic macrophage activation were constantly upregulated, whereas many DEGs controlling macrophage activation in circulating monocytes and peritoneal macrophages were transitorily regulated. A similar pattern of gene regulation discrepancy was observed in the production of IL-1β (GO:0032611) (Supplementary Fig. 7), suggesting a continuous pro-inflammatory reaction in the aorta as opposed to a fluctuating inflammatory response in compartments distant from the atheroma lesion.
Fig. 4Aortic macrophages uniquely upregulated genes associated with inflammation and lipid handling.
(A) Heatmap showing mean expression of genes relevant to macrophage function in atherosclerosis (n = 4 per group). * and # denote adjusted p-value < 0.05 (post-Bonferroni correction) compared to baseline and early disease stage of the same cell type, respectively. (B and C) Network plot showing genes that enriched biological process of macrophage activation in each monocyte or macrophage compartment. Arrows alongside the genes indicate their regulation pattern.
3.5 Multifaceted validation of Gpnmb expression and its association with atheroma development
To validate our RNA-seq directory and to demonstrate the applicability of our RNA-seq directory in generating pathomechanistic and therapeutic insights, we inspected the expression of Gpnmb in murine and human atheroma, using published scRNA-seq datasets. We selected Gpnmb, because it was one of the ten DEGs that significantly up-regulated between each of the disease stages. Gpnmb was previously reported in association with neuroinflammation and liver damage involving organ-specific macrophages; however, controversial data assert that Gpnmb functions to promote as well as resolve inflammation [
]. The features mentioned above made Gpnmb a suitable candidate for demonstrating the practicality of our directory using multifaceted molecular biology tools.
During the course of atherosclerosis, the total counts of Gpnmb remained low with subtle changes in blood monocytes and peritoneal macrophages, whereas aortic macrophages up-regulated Gpnmb by 220 and 50 folds during atheroma initiation and progression, respectively (Fig. 4A). The disease stage-, and compartment-dependent elevation of Gpnmb was confirmed with real-time polymerase chain reaction (PCR). (Supplementary Fig. 8A). The profound change of gene expression suggested its positive association with atheroma development, which was also reflected by an increase in Gpnmb protein levels in the plasma of Apoe−/− mice as atherosclerosis progressed (Supplementary Fig. 8B).
To further understand the correlation of Gpnmb and atherosclerosis, we investigated scRNA-seq datasets acquired from atherosclerotic aortas of Apoe−/− mice [
]. We observed that Gpnmb was almost exclusively expressed in fibroblasts and some macrophages. A similar cell type-specific expression was also shown in human plaques (Fig. 5A and B). Subsequently, we had a closer look at macrophage subtypes classified by the hallmark genes for foamy, inflammatory, and resident-like macrophages, as previously described [
Letter by Cochain et al regarding article, “Transcriptome analysis reveals nonfoamy rather than foamy plaque macrophages are proinflammatory in atherosclerotic murine models. vol. 123. Circulation Research,
2018
] (Supplementary Fig. 9). The analysis revealed that Gpnmb was distinctively expressed by Trem2+ foamy macrophages in both mouse and human plaque (Fig. 5C and D). Furthermore, the data suggested that an escalation of Gpnmb expression within a heterogeneous pool of aortic macrophages likely resulted from a growing fraction of foamy macrophages within advancing atherosclerotic lesions as opposed to an unanimous upregulation of Gpnmb by all aortic macrophages. In line, thin-cap unstable atheroma with a large lipid core expressed numerically higher tissue Gpnmb levels compared to thick-fibrous cap atheroma in humans (Supplementary Fig. 10).
Fig. 5Regulation of GPNMB in atherosclerotic aorta of mouse and human.
(A and B) Uniform Manifold Approximation and Projection (UMAP) plot showing atheroma-associated cells and Gpnmb expression in mouse and human. (C and D) UMAP plot showing subtypes of macrophages and the expression of Gpnmb in mouse and human. (E and F) Stacked violin plots showing the expression of genes related to efferocytosis and inflammation in Gpnmb+ and Gpnmb– macrophages. (G) Representative histology image with co-staining of nucleus, CD68+ macrophages, and Gpnmb in murine aortic root lesions. Bar chart showing the proportion of Gpnmb+ CD68+ macrophages of three mice, from each of which, 10 to 11 sections were used for quantification (mean ± SEM). (H) UMAP plot showing GPNMB expression in subtypes of macrophages retrieved from atheromatic core or proximal adjacent human plaque. CD, cluster of differentiation; DAPI, 4′,6-diamidino-2-phenylindole; Star signs indicate p-value < 0.05 by permutation test.
On a single-cell level, Gpnmb-expressing foamy macrophages expressed higher levels of lipid flux genes (e.g. Trem2, Abca1, Abcg1, Lrp1, and Cd36) but lower levels of pro-inflammatory genes (e.g. Nfkbia, Tnf, Nlrp3, and Il1b) as opposed to murine foam cells not expressing Gpnmb. Peroxisome proliferator activator activated receptor gamma (Pparg), a ligand-activated transcription factor that reduces inflammatory response and stimulates cholesterol efflux [
] was overexpressed in Gpnmb+ foamy macrophages (Fig. 5E). A similar phenotype of GPNMB+ foamy macrophages was observed in humans, in particular with regard to the lipid flux features and suppressed pro-inflammatory markers (Fig. 5F). Histological staining of aortic roots from atherosclerotic Apoe−/− mice verified the expression of Gpnmb protein in a small fraction of CD68+ macrophages residing in the plaque (Fig. 5G) in a range similar to the fraction of Gpnmb+ macrophages detected in the scRNA-seq dataset (Fig. 5C). Comparing scRNA-seq data [
] of human macrophage populations isolated from the atherosclerotic core to those retrieved from the adjacent, less diseased section of the same atheromatous plaque, we observed elevated GPNMB expression in the atherosclerotic core predominately by Trem2+ foamy macrophages (Fig. 5H). Thus, to the extent that different regions within the same human plaque can reflect atherosclerotic disease progression, the stage-dependent rise in Gpnmb expression in our murine aortic macrophage time line analysis well aligned with the human data.
4. Discussion
Our directory provides an overview of transcriptomic regulation of macrophages in the course of atherosclerosis. Monocytes and macrophages collected via fluorescence-activated cell sorting (FACS) in our directory were defined based on the long-established surface markers, which streamlines the cross-comparison of our dataset with most published works as well as the extrapolation of our findings. To acquire a comprehensive view of gene regulation in a defined cell group, bulk RNA-seq is the current method of choice as it retains most transcriptomic information in a cost-effective fashion. Nevertheless, we are aware of the compromise of identifying cell subpopulations by our method, whereas alteration of macrophage biology within atheroma were still evidently manifested in our dataset.
4.1 Evident and specific capture of compositional change of aortic macrophage population by bulk RNA-seq
The continual discovery of diverse origins of aortic macrophages has been fascinating the community over decades [
Polymerase chain reaction-based method for quantifying recruitment of monocytes to mouse atherosclerotic lesions in vivo: enhancement by tumor necrosis factor-α and interleukin-1β.
] recorded in our directory. Resident-like macrophages, on the other hand, plausibly proliferated in proportion to the atheroma plaque advancement, indicated by a rather steady expression of its hallmark genes (Lyve1, Folr2, Timd4) throughout the disease course (Fig. 4A). We previously demonstrated that local proliferation of macrophages also contribute to atheroma expansion in Apoe−/− mice that received two to six month of HCD [
], and accordingly cell proliferation markers were upregulated in aortic macrophages (Fig. 4A). Vascular smooth muscle cell-transdifferentiated macrophage-like cells were absent in our macrophage pool due to the exclusion of Cd45-negative cell populations during FACS sorting. The hallmark cell type of atherosclerosis, Trem2+ foamy macrophages, undoubtedly emerged in the growing atheroma [
]. Importantly, the aforementioned features were only revealed in macrophages isolated from aorta but not in those collected from peritoneum, which demonstrates an adequate capture of macrophage transcriptomic regulation in lesional and distant compartments both being exposed to hypercholesterolemia. Because the full murine aorta was subjected to bulk RNA-seq, macrophages presenting with resident cell-like features may stem from both the adventitia and intima. However, also human atherosclerotic plaque free of adventitial tissue contain resident-like macrophages, and we therefore consider resident-like macrophages in the intimal plaque as relevant for disease. The advancement of single-cell technologies enables the subtyping of atheroma macrophages to an unprecedented granularity [
Single-cell analysis of fate-mapped macrophages reveals heterogeneity, including stem-like properties, during atherosclerosis progression and regression.
], which supports an in-depth understanding of macrophage evolution in the atheroma lesion. However, surface markers of many of these macrophages are not fully established, thus limits the use of bulk RNA-seq to study specific subtypes of macrophages. On the other hand, it is more cost-efficient to study larger number of samples using bulk RNA-seq. Considering study granularity and affordability, we consider that bulk and single-cell RNA-seq methods compensate each other and should be exploited according to the research questions.
4.2 Selection of suitable macrophage targets with minimal extra-aortic effects
Using anti-inflammatory agents to combat atherosclerosis is currently an area of research [
]. Our data also supports this strategy as many pro-inflammatory cytokines (Il1a, Il1b, Il6, Tnf) were expressed more strongly by macrophages in aorta than by those in peritoneum. Facilitation of reparative macrophages also appears propitious, especially in the advanced atheroma. We found that the anti-inflammatory cytokines, Il13 and Il33 [
] reached their highest expression in the early phase and dropped in the advanced atheroma (Fig. 4A), indicating that failing to sustain anti-inflammatory effects might deteriorate atherosclerosis. According to our dataset, we propose a more cautious monitoring of using therapeutics that target LDL receptors on monocytes and macrophages, because these cells downregulate LDL receptor expression in the course of atherosclerosis, which could possibly result in treatment ineffectiveness overtime. However, targeting scavenger receptors appears as more adequate approach for targeting lipid-laden macrophages, the hallmark cell type of atheroma [
Absence of CD36 protects against atherosclerosis in ApoE knock-out mice with no additional protection provided by absence of scavenger receptor A I/II.
]. Among the long list of scavenger receptors, our dataset provides some hints that assist target prioritization. Trem2 appears the most favorable target because of its time-dependent upregulation in aortic macrophages and low expression in peritoneal macrophages (Fig. 4A). Conversely, Cd36 and Msr1 show also strong expression in extra-aortic macrophages (Fig. 4A).
4.3 Validation of RNA-seq target candidates with multifaceted complementary approaches
Although RNA-seq is a powerful tool that enables high-throughput assessment of thousands of gene regulations from one sample, validation of the chosen candidates is crucial owing to the fact that transcriptome does not perfectly stand for proteome. In this article, we deployed multifaceted tools to verify the expression of Gpnmb in atheroma, intending to demonstrate the potential of using RNA-seq data to identify targets of interest. Real-time PCR confirmed the quantitative accuracy of Gpnmb expression captured by RNA-seq (Supplementary Fig. 8A). Published scRNA-seq datasets allowed us to have a glimpse of Gpnmb expression in a wide-range of cell types as well as subtypes of macrophages. Knowing that Gpnmb is primarily expressed in a subset of foamy macrophages, we confirmed its positive association with atheroma aggravation. In line, we detected Gpnmb protein in a subset of atheromatous plaque macrophages by histology, and we detected a continuous rise in plasma Gpnmb levels as disease progressed (Supplementary Fig. 8B). However, the atheroma single-cell expression data document that Gpnmb is also expressed by fibroblasts besides macrophages, and therefore global Gpnmb deficiency or Gpnmb silencing experiments may not adequately and selectively explore the specific role of Gpnmb+ macrophages in atherogenesis. Till the time when we composed this article, the role of Gpnmb in inflammation remains debatable and is well summarized in a recent review [
]. Some evidence suggests that Gpnmb supports inflammation resolution, while others demonstrate that Gpnmb could act pro-inflammatory, associating with enlarged plaque size specifically at the aortic roots in Apoe-deficient mice [
]. Using Gpnmb as a marker to distinguish Gpnmb+ and Gpnmb– foamy macrophages, we were able to confirm that the latter express relatively high levels of inflammatory genes, and thus challenge the simplistic view that foam cells are non-inflammatory, as was previously suggested [
]. With the facilitation of RNA-seq datasets, we have gained precious knowledge for the future research, and we hope that our directory will be of value for the community in the search of regulators of monocyte and macrophage biology and potential therapeutic targets in the context of atherosclerosis.
Financial support
This work was supported by the Foundation of Studying Civilization Diseases to I.H., the German Research Foundation (HI1573/2, HI1573/9, HI1573/10). I.H. is a member of the Collaborative Research Centre SFB1425 (grant 422681845).
Author contributions
CH acquired, analyzed and interpreted the data, and assisted in writing the manuscript, AK, CAE, TV, CS, AE, KK, BD, NH acquired and analyzed data. PS, TH, DW, CzM, BG, CSR, LM, DW provided materials and helped to interpreted the data and to edit the manuscript. TD and IH conceived and supervised the project, analyzed and interpreted the data, wrote the manuscript, and acquired funding.
Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgements
We thank staff members of the Lighthouse core facility at University of Freiburg, Marie Follo, Jan Bodinek-Wersing, and Dieter G. Herchenbach, for their help and guidance with cell sorting. We thank Vladimir Benes, Head of Genomic Core EMBL, Heidelberg, for his support and advice with sequencing.
Appendix A. Supplementary data
The following are the Supplementary data to this article.
Letter by Cochain et al regarding article, “Transcriptome analysis reveals nonfoamy rather than foamy plaque macrophages are proinflammatory in atherosclerotic murine models. vol. 123. Circulation Research,
2018
Polymerase chain reaction-based method for quantifying recruitment of monocytes to mouse atherosclerotic lesions in vivo: enhancement by tumor necrosis factor-α and interleukin-1β.
Single-cell analysis of fate-mapped macrophages reveals heterogeneity, including stem-like properties, during atherosclerosis progression and regression.
Absence of CD36 protects against atherosclerosis in ApoE knock-out mice with no additional protection provided by absence of scavenger receptor A I/II.
00104-1&id=ga1.jpg){kind=link}
00104-1&id=gr1.jpg){kind=link}
00104-1&id=gr2.jpg){kind=link}
00104-1&id=gr3.jpg){kind=link}
00104-1&id=gr4.jpg){kind=link}
00104-1&id=gr5.jpg){kind=link}