Learn more: PMC Disclaimer | PMC Copyright Notice
CXCL4 induces a unique transcriptome in monocyte-derived macrophages
Associated Data
Abstract
In atherosclerotic arteries, blood monocytes differentiate to macrophages in the presence of growth factors like macrophage colony-stimulation factor (MCSF) and chemokines like platelet factor 4 (CXCL4). To compare the gene expression signature of CXCL4-induced macrophages with MCSF-induced macrophages or macrophages polarized with IFN-γ/LPS (M1) or IL-4 (M2), we cultured primary human peripheral blood monocytes for six days. mRNA expression was measured by Affymetrix gene chips and differences were analyzed by Local Pooled Error test, Profile of Complex Functionality and Gene Set Enrichment Analysis. 375 genes were differentially expressed between MCSF- and CXCL4-induced macrophages, 206 of them overexpressed in CXCL4 macrophages coding for genes implicated in the inflammatory/immune response, antigen processing/presentation, and lipid metabolism. CXCL4-induced macrophages overexpressed some M1 and M2 genes and the corresponding cytokines at the protein level, however, their transcriptome clustered with neither M1 nor M2 transcriptomes. They almost completely lost the ability to phagocytose zymosan beads. Genes linked to atherosclerosis were not consistently up- or downregulated. Scavenger receptors showed lower and cholesterol efflux transporters higher expression in CXCL4- than MCSF-induced macrophages, resulting in lower LDL content. We conclude that CXCL4 induces a unique macrophage transcriptome distinct from known macrophage types, defining a new macrophage differentiation that we propose to call M4.
Introduction
The mononuclear phagocyte system is essential to the innate immune response and encompasses various types of constitutive tissue macrophages, e.g. Kupffer cells in the liver or alveolar macrophages in the lung. Under inflammatory conditions, macrophages can differentiate from peripheral blood monocytes under the influence of various growth factors, cytokines, or infectious agents (1). In atherosclerosis, macrophage differentiation is critically related to disease progression: During atherogenesis blood monocytes are thought to enter the arterial wall and differentiate into macrophages, which sustain an inflammatory milieu and promote plaque formation (2-5).
As demonstrated by in vitro and in vivo data, macrophages present in chronically inflamed tissues may assume different phenotypes. The best defined polarization types are M1 and M2 (6). According to the “classical” paradigm, M1 macrophages can be obtained through activation by interferon- γ (IFNγ), tumor necrosis factor-α (TNF-α), or lipopolysaccharide (LPS)), whereas the “alternative” M2 macrophages can be induced through activation by interleukin-4 (IL-4), IL-10, or IL-13 (7,8). The phenotypes of macrophages in vivo are incompletelz described and M1 and M2 are probably not the only macrophage phenotypes present in vivo.
In atherosclerosis, there is evidence for the presence of several different macrophage phenotypes within atherosclerotic plaques, some with features of M1 and M2 (9). In addition, other differentiation types like CD14-CD68+ and CD14+CD68- macrophages have been identified in coronary artery lesions (10). In vivo, differentiation of macrophages toward different phenotypes has been associated with certain drugs, growth factors, and other mediators. M2 differentiation is induced by PPARγ agonists (9), whereas MCSF preferentially induces CD14+CD68+ macrophages (10) and hemoglobin-haptoglobin promotes differentiation toward CD163highHLA-DRlow macrophages (11).
Only two growth factors are known to promote differentiation of monocytes into macrophages in vitro: macrophage colony-stimulating factor (MCSF) (12) and platelet factor-4 (CXCL4) (13). MCSF has been shown to induce a transcriptome that is similar to that of M2 macrophages (14). The physiological role and function of MCSF has been thoroughly studied. Knock-out mice lacking MCSF (CSF1) or its receptor (CSF1R) are protected from atherogenesis (15,16). By contrast, the role of the platelet chemokine CXCL4 is by far more enigmatic. CXCL4 strongly suppresses megakaryocyte differentiation (17), inhibits monocyte apoptosis and promotes macrophage differentiation (13). CXCL4 is released from platelets upon activation in micromolar concentrations and has a broad range of biological functions including induction of respiratory burst in human monocytes accompanied by secretion of several chemokines such as CCL3, CCL4, and CXCL8 (18-20). In vivo, presence of CXCL4 within atherosclerotic lesions has been shown to correlate with clinical parameters (21). Eliminating the PF4 gene coding for CXCL4 by homologous recombination has been shown to reduce lesion formation in a mouse model of atherosclerosis (22).
While the transcriptomes of MCSF-induced macrophages and their M1 or M2 polarizations have been extensively studied (14), the published data on the phenotype of CXCL4-induced macrophages is scarce. CXCL4 has been shown to induce macrophages expressing CD86, but not HLA-DR on the cell surface (13). We recently showed that CXCL4 strongly suppresses expression of the hemoglobin-haptoglobin receptor CD163 (23). Both findings sugges that the CXCL4 macrophage is distinct from its MCSF counterpart. However, thus far a comprehensive transcriptome analysis of the CXCL4-induced macrophage phenotype has not been undertaken. Furthermore, it remains unclear whether the CXCL4 macrophage is relevant for atherogenesis and can be attributed to any of the known polarization patterns.
We hypothesized that the transcriptome of CXCL4-induced macrophages may be unique and different from MCSF or other known polarization types. We therefore conducted a comprehensive analysis of the CXCL4 macrophage transcriptome and compared it to its MCSF counterpart, speculating that this analysis may give insight into mechanisms by which CXCL4 macrophages may promote disease progression in atherosclerosis.
Materials and methods
Monocyte-derived human macrophages
With approval from the institutional review board, peripheral blood mononuclear cells were isolated from human peripheral blood using Histopaque (Sigma, St.Louis, MO) followed by negative isolation with magnetic beads (Stem cell, Vancouver, Canada). Monocyte purity was 96.2 ± 0.2 % as assessed by CD14 expression. After red blood cell lysis and several wash steps with 1 mM EDTA, monocytes were essentially free from platelet contamination as demonstrated by virtual absence of CD41 positivity in flow cytometry (data not shown). Monocytes were cultured in macrophage serum-free medium (Gibco, Carlsbard, CA) supplemented with Nutridoma SP (Roche, Indianapolis, IN) and penicillin/streptomycin (Sigma, St. Louis, MO) for six days in the presence of 100 ng/ml rhMCSF (Peprotech, Rocky Hill, NJ) or 1 μM rhCXCL4 (Peprotech). The concentration of 1 μM rhCXCL4 was chosen because this concentration was previously demonstrated to be sufficient to induce macrophage differentiation from monocytes (13). Furthermore, our own preliminary experiments confirmed that after six days, this concentration induced expression of typical macrophage markers like CD11b or CD68 to a similar extent as MCSF (Fig. 1 and data not shown).
(A) Morphology of macrophages after six days in culture. Bar indicates 50 μm. (B) Gene and protein expression of lineage marker genes PTPCR (CD45), CD14 (CD14), ITGAM (CD11b) in both macrophage types (differences not significant by Local Pooled Error test (LPE)). (C) The upper graph shows log2 transformed intensity of all expressed genes in M0 macrophages plotted against the intensity in M4 macrophages (r=0.934, P<0.0001). The lower plot shows the same data including only genes with a FDR<0.05 as determined by Local Pooled Error (LPE) test. (D) Heatmap showing all significantly differentially regulated genes (FDR<0.05 by LPE test). Gene expression was normalized and standardized (Gene list in Supplementary table S1). Red indicates high, green low gene expression. Genes and conditions were allowed to freely cluster in the y and x axis, respectively.
oxLDL-induced foam cell formation and phagocytosis assays
For foam cell formation assays, macrophages were exposed to 10 μg/ml DiI-labeled acetylated or oxidized LDL (Biomedical Technologies, Stoughton, MA) for four hours at 37°C. Subsequently, cells were washed and fluorescence intensity was assessed in a flow cytometer (FACScalibur, Becton Dickinson, San Jose, CA). Untreated macrophages served as negative control.
Phagocytosis was assessed using M0 and M4 macrophages as phagocytes and zymosan beads (Alexa Fluor 488, Invitrogen, Carlsbad, CA) as targets, at a ratio of ten zymosan beads to one macrophage. Macrophages were incubated at 37° for 1 hour with opsonized or non-opsonized zymosan beads. Opsonization was carried out by incubation at 37° for 1 hour with autologous serum followed by three 3 PBS wash steps using low speed centrifugation (1500g, 15 minutes). The extent of phagocytosis was analyzed by flow cytometry, using untreated macrophages (no beads) as control.
Affymetrix gene chip experiments
For each condition RNA was isolated from macrophages derived from two donors using columns including a DNAse-step followed by reverse transcription (all reagents from Qiagen, Valencia, CA). RNA was labeled and hybridized to Affymetrix HG133 Plus 2.0 arrays as described previously (24). For each donor, RNA was hybridized to a separate gene array. Signal intensity values were obtained from the Affymetrix MicroArray Suite software (MAS 5.0). The data set and technical information according to the Minimum Information about a Microarray Experiment (MIAME) requirements are available at the Gene Expression Omnibus (GEO) website (www.ncbi.nlm.nih.gov/geo), accession number GSE <to be assigned>.
ELISA and cytokine bead arrays
Protein concentration of selected cytokines was measured in cell culture supernatants using ELISA (CCL18, R&D Systems, Minneapolis, MN; CCL22, Cell Sciences, Canton, MA) or cytokine bead arrays (IL-1β, IL-6, IL-8, IL-10, IL-12p70, TNF; Becton Dickinson, San Jose, CA) according to the manufacturers’ instructions. Supernatants were pooled over six days and diluted where necessary to obtain concentrations within the range of the assays.
Flow cytometry
For flow cytometry, cells were treated with Fc block (Miltenyi) and subsequently stained with antibodies against CD36 (clone CB38, BD Biosciences) and SR-A (clone 351615, R&D Systems). For SR-A staining, a FITC-labeled secondary antibody was used. Appropriate isotype controls were used in all experiments. Fluorescence was measured on a FACSCalibur flow cytometer (Becton Dickinson). Fluorescence was assessed as background-corrected mean fluorescence (MFI).
Local Pooled Error test (LPE)
For statistical analysis, the open source statistical software package R (www.r-project.org) was used including the Local Pooled Error (LPE) test for differential expression discovery between two conditions (25). Gene chip data were analyzed as described previously (24). Briefly, after exclusion of non-expressed genes, data were normalized and log2 transformed to achieve normal distribution. The LPE test is statistically powerful for identifying differentially expressed genes between low-replicated microarray data. It pools probe sets with similar expression levels providing a statistic for each probe set. The absolute value of the LPE-statistic is larger for more significantly differentially expressed probe sets. A false discovery rate (FDR) was calculated to discover probe sets differentially expressed with FDR<0.05 (26). Heatmaps were constructed using R in a way that allows all conditions and genes to freely cluster both in the x (condition) and the y axis (gene).
Profile of Complex Functionality (ProfCom)
To assess functional networks regulated in each macrophage type, profiled complex functionality was analyzed employing the ProfCom software, a web-based tool for the interpretation of genes that were identified to be functionally linked by experiment (27). This tool compares the proportion of genes related to specific gene ontology (GO) categories among the genes found regulated to the proportion of genes related to the same category within the gene ontology reference genes and corrects for multiple testing.
Gene Set Enrichment Analysis (GSEA)
Gene set enrichment was analyzed using an open access software for gene set enrichment analysis (GSEA) (28) to assess potential similarities between the CXCL4-induced gene expression profile and the known M1 and M2 signatures. The latter were extracted from gene expression data of monocyte-macrophage differentiation and polarization published by Mantovani et al. (14) (GEO data set 2430). Using the LPE test, genes differentially expressed between the M1 and M2 data set were identified and used as M1 and M2 gene sets, respectively. Overexpression of the M1 and M2 gene sets was tested by GSEA in the MCSF and CXCL4 gene expression data. GSEA calculates an enrichment score, which indicates the degree of overrepresentation of these gene sets, and estimates its significance with adjusting for multiple hypothesis testing.
Modified Principal Component Analysis (PCA) and hierarchical clustering
A modified PCA was performed on the previously published M1 and M2 (14) as well as on the new CXCL4 gene expression data normalized to the corresponding MCSF gene expression sets. This normalization step avoided bias due to inter-experimental variance. Firstly, PCA was performed including all genes that were significantly overexpressed (as determined by LPE) in M1 relative to M2. Subsequently, a second PCA was performed including all genes that were overexpressed in M2 relative to M1. The first principal components from each of these analyses (independent by definition) were used to define a new coordinate space in which CXCL4 gene expression data was plotted.
Hierarchical clustering was used to determine the level of similarity between the three normalized groups (29). All genes were included in the analysis, and the results are displayed in a dendrogram. Distance was determined by average linkage, wherein the distance between two groups, A and B, is determined according to the equation: , where ni is the number of members in group i and dij is the Euclidean distance between two points, i and j.
Results
MCSF and CXCL4 induce macrophages with a similar transcriptome
Preliminary experiments confirmed that after six days in culture both MCSF- as well as CXCL4-induced macrophages (which we suggest to call M0 and M4, respectively) displayed a morphology characteristic of macrophages (Fig. 1A). Accordingly, gene and protein expression of the classical lineage markers (CD45, CD14, and CD11b) was comparable in both macrophage types, indicating that the cells studied were fully differentiated macrophages (Fig. 1B).
When comparing the overall gene expression signature of M0 and M4 macrophages, the two gene expression patterns were found to be similar and highly correlated (r=0.934, P<0.0001, Fig. 1C upper panel). Out of 26,051 probe sets expressed above the detection limit in at least one macrophage sample, 460 annotated probe sets were significantly up- or downregulated with a FDR<0.05 corresponding to a total number of 375 regulated genes (Fig. 1C lower panel). Two hundred six of these genes displayed higher and 169 lower expression levels in M4 macrophages as compared to M0 macrophages (Fig. 1D). A list of differentially expressed genes is given in Supplementary table 1.
CXCL4-induced macrophages overexpress genes implicated in the immune response, antigen-processing and presentation, and lipid metabolism
Based on the genes found to be differentially expressed between M0 and M4 macrophages by LPE test, we sought to identify functional processes as defined by gene ontology that were transcriptionally overrepresented in M4 macrophages. Applying profiling of complex functionality (ProfComp) analysis (27), we found a number of biological processes that were associated with genes expressed in M0 or M4 macrophages (Fig. 2A-F). Most prominently, both M0 and M4 macrophage overexpressed genes related to the inflammatory and the immune response. In M4 macrophages, CCL18 and TNFSF10 (TRAIL) were overexpressed, whereas in M0 macrophages AIF1 (allograft inflammatory factor), ALOX5 (arachidonate 5-lipoxygenase), and IL1RN (interleukin-1 receptor antagonist) were found in the inflammation and immune response gene sets (in all cases P<0.05).
Bars indicate the percentage (A, C, E) or the absolute number (B, D, F) of genes attributed to a certain GO category within all genes of the GO data set (empty bars), genes overexpressed in M0 macrophages (black bars), or genes overexpressed in M4 macrophages (grey bars). Data are arranged by biological process (A, B), cellular component (C, D), and molecular function (E, F). * P<0.05 adjusted for multiple testing.
Genes involved in antigen processing and presentation were significantly overrepresented in M4 macrophages (P<0.05), including HLA-DRB1, HLA-DRB3, HLA-DRB4, and HLA-DQA1 (coding for MHC class II), as well as the costimulatory surface molecule CD86. Interestingly, several genes implicated in lipid metabolism and transport were also found overexpressed (P<0.05), including APOC2, APOE, and SORL1 (sortilin-related receptor, low-density lipoprotein receptor class A repeat-containing protein). By contrast, genes overexpressed in MCSF-induced macrophages were more likely to be implicated in chemotaxis (represented by the chemokines CCL3,CCL7, and the chemokine receptor CCR1, P<0.05) or cell adhesion (as indicated by the integrin genes ITGAV, ITGA6, ITGB8B, and the COL6A gene coding for the extracellular matrix component collagen 6A, P<0.05).
M4 macrophages do not display a clear M1 or M2 pattern and their transcriptome is distinct from the M1 or M2 transcriptomes
As reported previously, M0 macrophages display a gene expression pattern very similar to that of M2 macrophages, whereas the gene signature of M1 macrophages is distinct (14). To better understand the characteristics of M4 macrophages, we sought to assess whether the M4 macrophage transcriptome is comparable to either of these polarization types. When looking at selected genes characteristic for M1 or M2 polarization, it became clear that a large number of polarization marker genes were not differentially expressed between M0 and M4 macrophages. This was true for a number of cytokines (IL6, IL12, TNF (TNF-α) (all M1), IL10 (M2)), many chemokines (CCL2, CCL5 (both M1), CCL1 (M2)), several surface receptors (CCR7, TLR2, TLR4 (all M1)), or specific enzymes (NOS2 (iNOS) (M1), ARG1 (arginase-1) (M2)) (8). On the other hand, a small number of marker genes displayed significant differential expression between M0 and M4 macrophages, however, there was no clear pattern for preferential expression of M1 or M2 markers in either of the macrophage types (Fig. 3A). Table 1 shows all M1 and M2 genes significantly overexpressed in either M0 or M4 macrophages. Measuring protein levels of cytokines released into cell culture supernatants largely confirmed this pattern with IL-6, TNF (both M1), CCL18, and CCL22 (both M2) levels being higher in M4 macrophages, and IL-10 (M2) levels higher in M0 macrophages (Fig. 3B). No differences were seen for the levels of IL-1β, IL-8, and IL-12p70 (data not shown).
(A) Gene expression of selected markers for M1 and M2 polarization (TNFSF10 = TRAIL, PTX3 = pentraxin-3, MRC1 = mannose receptor). *** FDR<0.001 by LPE test. (B) Gene expression (upper row) and protein levels (bottom row) of cytokines related to M1 or M2 polarization. *** FDR<0.001 by LPE test (gene expression), * P<0.05 by paired t test (protein levels), dotted line indicates detection limit of the assay. (C) Enrichment plot of M1 or M2 gene sets in M0 versus M4 macrophages. All genes were ranked using the GSEA difference of class metric, and for each gene in the M1 or M2 gene set, the enrichment score was calculated and plotted against the position of the genes within the rank ordered data set. In both cases, no significant enrichment was found.
Table 1
Probe sets and genes attributed to (A) M1 polarization or (B) M2 polarization (22) with significantly differential expression between M0 and M4 macrophages according to LPE test. Log2 normalized expression data, z stats and FDR as calculated by LPE test.
| (A) | ||||||||
|---|---|---|---|---|---|---|---|---|
| Probe sets/genes related to M1 polarization | ||||||||
| Probe set | Gene symbol | Annotation | M0 #1 | M0 #2 | M4 #1 | M4 #2 | z stats | FDR |
| 205686_s_at | CD86 | CD86 | 9.72 | 10.07 | 11.66 | 11.57 | -6.97 | <0.0001 |
| 210895_s_at | CD86 | CD86 | 11.27 | 11.66 | 12.89 | 13.01 | -6.74 | <0.0001 |
| 209728_at | HLA-DRB4 | major histocompatibility complex, class II, DR beta 4 | 11.86 | 11.57 | 12.90 | 13.29 | -6.28 | 0.0017 |
| 205685_at | CD86 | CD86 | 8.93 | 9.34 | 10.64 | 10.82 | -5.84 | 0.0039 |
| 213831_at | HLA-DQA1 | major histocompatibility complex, class II, DQ alpha 1 | 9.86 | 2.03 | 10.77 | 5.92 | -5.62 | 0.0075 |
| 204670_x_at | HLA-DRB1 | major histocompatibility complex, class II, DR beta 1 | 12.41 | 11.94 | 13.09 | 13.58 | -5.22 | 0.0105 |
| 204972_at | OAS2 | 2’-5’-oligoadenylate synthetase 2 | 8.14 | 7.64 | 9.23 | 9.52 | -4.72 | 0.0188 |
| 202688_at | TNFSF10 | TNF-related apoptosis inducing ligand TRAIL | 6.12 | 5.83 | 7.83 | 8.12 | -4.63 | 0.0202 |
| 208306_x_at | HLA-DRB1 | major histocompatibility complex, class II, DR beta 1 | 12.66 | 12.65 | 13.34 | 13.98 | -4.39 | 0.0284 |
| 215193_x_at | HLA-DRB1 | major histocompatibility complex, class II, DR beta 1 | 12.45 | 12.16 | 13.14 | 13.38 | -4.37 | 0.0298 |
| 206157_at | PTX3 | Pentraxin 3 | 10.76 | 10.85 | 9.65 | 9.69 | 4.32 | 0.0322 |
| (B) | ||||||||
| Probe sets/genes related to M2 polarization | ||||||||
| Probe set | Gene symbol | Annotation | M0 #1 | M0 #2 | M4 #1 | M4 #2 | z stats | FDR |
| 207861_at | CCL22 | CCL22 | 11.32 | 11.14 | 14.40 | 15.28 | -11.47 | <0.0001 |
| 32128_at | CCL18 | pulmonary and activation-regulated chemokine (PARC) | 5.34 | 4.84 | 9.29 | 9.94 | -7.92 | <0.0001 |
| 204438_at | MRC1 | Mannose receptor | 10.64 | 10.67 | 11.99 | 12.58 | -7.13 | <0.0001 |
| 223280_x_at | MS4A6A | membrane-spanning 4-domains subfamily A, member 6A | 7.55 | 5.53 | 9.68 | 8.58 | -7.02 | <0.0001 |
| 224356_x_at | MS4A6A | membrane-spanning 4-domains, subfamily A, member 6A | 7.78 | 5.93 | 9.80 | 8.81 | -6.98 | <0.0001 |
| 204112_s_at | HNMT | histamine N-methyltransferase | 12.05 | 12.45 | 10.51 | 10.77 | 6.84 | <0.0001 |
| 201427_s_at | SEPP1 | selenoprotein P, plasma, 1 | 10.16 | 9.03 | 8.33 | 3.89 | 7.64 | <0.0001 |
| 209555_s_at | CD36 | CD36 | 14.32 | 14.50 | 13.02 | 12.54 | 9.08 | <0.0001 |
| 206488_s_at | CD36 | CD36 | 14.37 | 14.47 | 13.13 | 12.42 | 9.15 | <0.0001 |
| 211719_x_at | FN1 | Fibronectin 1 | 11.35 | 12.01 | 9.45 | 6.72 | 12.59 | <0.0001 |
| 212464_s_at | FN1 | Fibronectin 1 | 11.62 | 11.88 | 10.06 | 6.25 | 12.69 | <0.0001 |
| 210495_x_at | FN1 | Fibronectin 1 | 11.56 | 12.13 | 9.63 | 6.72 | 12.99 | <0.0001 |
| 216442_x_at | FN1 | Fibronectin 1 | 11.42 | 12.01 | 9.46 | 4.17 | 14.06 | <0.0001 |
| 209924_at | CCL18 | pulmonary and activation-regulated chemokine (PARC) | 6.25 | 5.71 | 7.93 | 9.33 | -6.31 | 0.0017 |
| 219666_at | MS4A6A | membrane-spanning 4-domains, subfamily A, member 6A | 7.14 | 6.29 | 9.31 | 8.30 | -5.73 | 0.0039 |
| 208422_at | MSR1 | Scavenger receptor-A | 7.93 | 12.27 | 5.76 | 11.11 | 5.72 | 0.0039 |
| 227265_at | FGL2 | fibrinogen-like protein 2 | 8.66 | 7.25 | 10.55 | 8.80 | -5.55 | 0.0075 |
| 204834_at | FGL2 | fibrinogen-like protein 2 | 8.50 | 7.86 | 10.42 | 9.26 | -5.49 | 0.0075 |
| 211732_x_at | HNMT | histamine N-methyltransferase | 10.11 | 10.63 | 8.79 | 9.51 | 4.44 | 0.0259 |
| 228772_at | HNMT | histamine N-methyltransferase | 8.42 | 8.81 | 7.07 | 7.33 | 4.00 | 0.0463 |
To generate larger gene sets for M1 or M2 polarization, we compared gene array data sets for M1 and M2 polarized macrophages published by Mantovani et al. (14). These gene expression data were derived from human monocyte-derived MCSF macrophages, which were either treated with LPS and IFN-γ (M1) or IL-4 (M2) (14). Genes with a FDR<0.05 as determined by LPE testing between M1 and M2 (Supplementary Table S2) were included in the gene sets (Supplementary Table S3). Using these gene sets, we performed gene set enrichment analysis (GSEA) for M1 and M2 genes. This demonstrated no significant overexpression of either of the gene sets in M0 or M4 macrophages (FDR=0.98 (M1 gene set) and 1.0 (M2 gene set), respectively, Fig.3 C). This finding establishes that M4 macrophages are neither M1 nor M2 but represent a distinct phenotype.
To test whether the M4 macrophage transcriptome is similar to any of the classical polarization patterns (M0, M1, M2), we used a modified Principal Components Analysis (PCA) on the normalized M1 and M2 data (14). At first, PCA was performed including all genes that were significantly overexpressed (as determined by LPE) in M1 relative to M2 (n=2431). A second PCA was performed that included all genes overexpressed in M2 relative to M1 (n=3944). Based on the principal components from each of these analyses, a new coordinate space was defined in which the M4 gene expression data were plotted. The M4 macrophages did not cluster with either M1 or M2 macrophages (Fig. 4A). To corroborate this finding, we employed hierarchical clustering, now including all genes of the M1, M2, and M4 macrophage expression data. This analysis confirmed that the M4 transcriptome significantly differs from M0, M1 or M2 macrophages and represents a unique macrophage phenotype. In fact, M1 and M2 are more similar to each other than to M4 (Fig. 4B).
(A) Modified Principal Components Analysis of M1 and M2 gene expression data (as described in Methods). Briefly, M4 gene expression data were plotted into a coordinate space defined by M1 and M2 gene expression data. (B) Hierarchical clustering of the normalized M1, M2, and M4 gene expression data. All genes were included in the analysis, and the results are displayed as dendrogram. Root = M0. #1, #2, and #3 indicate donor-specific replicates for each condition.
Phagocytotic function
One function of macrophages is to phagocytose pathogens and foreign materials (29). Phagocytosis was recently shown to be inhibited in M2-polarized macrophages (30). To test the phagocytosis function, we incubated M0 and M4 macrophages with zymosan beads with and without serum opsonization. Although about 20% of M0 macrophages phagocytosed zymosan beads, this function was almost completely suppressed in M4 macrophages (Fig. 5A and B).
Macrophages differentiated with MCSF (M0) or CXCL4 (M1) were exposed to zymosan beads (A) or zymosan beads opsonized with FCS (B) as described in Materials and Methods. The phagocytotic capacity of M0 and M4 macrophages was assessed by flow cytometry. Representative histograms are shown in (A) and (B), results of two independent experiments are presented as bar graphs in (C). ** P<0.01.
Atherosclerosis-related genes in M4 macrophages and potential functional implications
To understand the potential relevance of CXCL4-induced macrophages in atherogenesis, we further investigated the list of differentially expressed genes. As indicated by gene ontology analysis, these included several chemokines, matrix metalloproteases and two members of the cathepsin family, but also some genes implicated in lipid metabolism and foam cell formation. Most of these gene groups did not display a consistent pattern, indicating that both pro-atherogenic and anti-atherogenic genes were expressed in M4 macrophages. Thus, while MMP7 showed higher expression in M4 than M0 macrophages, MMP8 and MMP12 expression were higher in M0 macrophages (Fig. 6A).
(A) Expression of atherosclerosis-related genes with significantly different expression level in M0 and M4 macrophages. *FDR<0.01, ***FDR<0.001. (B) Heatmap of genes implicated in foam cell formation. Red indicates high, green low gene expression. Genes and conditions were allowed to freely cluster in the y and x axis, respectively.(C) Representative histograms of surface expression of scavenger receptor-A (SR-A) and CD36 in M0 (solid line) and M4 (dotted line) macrophages derived from the same donor. Isotype control shown in gray. (D) and (E) Mean fluorescence intensity of DiI-labeled acetylated (D) or oxidized (E) LDL in M0 (solid line) and M4 (dotted line) macrophages after 4 hours exposure to 10 μg/ml LDL as determined by flow cytometry. Representative histograms and a bar graph summarizing the flow cytometric data are shown (* P<0.05 by t test, mean ± SEM, n=3-6).
Strikingly, genes implicated in foam cell formation (i.e. the scavenger receptors CD36 and MSR1 (SR-A) and the cholesterol efflux transporter ABCG1) displayed highly differential expression between M0 and M4 macrophages (ABCA1 mRNA was neither expressed in M0 nor in M4 macrophages). While M0 macrophages expressed higher levels of CD36 and MSR1 mRNA, ABCG1 mRNA expression was higher in M4 macrophages. This gene expression pattern suggested that M4 macrophage would be less likely to take up modified LDL and more likely to promote cholesterol efflux (Fig. 6A and B). At the protein level, CD36 expression was slightly lower in M4 macrophages, whereas no significant difference was seen for SR-A expression (Fig. 6C). To assess the functional relevance of this finding, we studied uptake of DiI-labeled acetylated LDL (acLDL) or oxidized LDL (oxLDL) by M0 and M4 macrophages. After four hours exposure to 10 μg/ml DiI-labeled acLDL or oxLDL, M4 macrophages displayed significantly reduced content of modified LDL as assessed by flow cytometry (Fig. 6D and E). This suggests that the gene signature actually translates into cellular function and that the prevailing gene expression pattern of M4 macrophages tends to result in reduced foam cell formation.
Discussion
This paper reports the first comprehensive analysis of the transcriptome of monocyte-derived macrophages induced by the chemokine platelet factor-4 (CXCL4). Our study demonstrates that (1) CXCL4 induces a macrophage phenotype that is distinct from that induced by MCSF, (2) the CXCL4-induced transcriptome shares similarities with, but is also distinct from each of the classical M1 and M2 phenotypes, and (3) the transcriptome of CXCL4-induced macrophages is not clearly pro- or anti-atherogenic. Based on its unique properties, we suggest calling the macrophage polarization induced by CXCL4 ‘M4 macrophages’.
Our knowledge about heterogeneity of polarized macrophages has increased significantly over the past several years. Thus, in addition to the “classical” M1 macrophages (characterized by high expression of pro-inflammatory cytokines, iNOS expression and production of reactive oxygen species) and M2 macrophages (expressing high levels of mannose receptor, dectin-1 and arginase), a number of M2 subsets has been characterized (31). These include M2 macrophages activated by IL-4 or IL-13 (now termed M2a), macrophages activated by immune complexes (termed M2b), and macrophages polarized with glucocorticoids or IL-10 (M2c) (32). The MCSF-induced transcriptome as well as the corresponding M1 and M2 (more specifically M2a) transcriptomes have been studied by Mantovani et al. using Affymetrix gene chips (14). These experiments showed a close similarity between MCSF-induced and M2a macrophages. Furthermore, they demonstrated differences between M1 and M2a in genes involved in metabolic activities as well as genes coding for chemokines (14).
Similar to MCSF, the platelet chemokine CXCL4 has been demonstrated to prevent monocyte apoptosis and promote macrophage differentiation from human peripheral blood monocytes (13). Surprisingly, the phenotype of these CXCL4-induced macrophages has not been studied in detail. Our data suggest that CXCL4 induces a macrophage phenotype that shares similarities with both M1 and M2 macrophages. Thus, some M1- as well as M2-related genes are overexpressed in M4 macrophages as compared to MCSF-induced macrophages. This finding was confirmed for a number of cytokines on the protein level. Most importantly, an unbiased analysis using different approaches like gene set enrichment, modified principal component, and hierarchical clustering analysis all confirmed the uniqueness of the CXCL4-induced macrophage transcriptome.
Platelets as well as monocytes and monocyte-derived macrophages are present within atherosclerotic lesions, and it is now clear that both contribute to lesion formation (5). The platelet chemokine CXCL4 is known to promote atherosclerosis as demonstrated in CXCL4-deficient PF4-/- mice. On the Apoe-/- background, the PF4-/- mice showed about 60% reduction of lesion size in the aorta (33). One way by which CXCL4 may contribute to atherogenesis is by promoting macrophage differentiation from monocytes present in the arterial wall. It had been speculated that CXCL4 may induce a macrophage polarization favorable for the development of atherosclerotic lesions. Our in vitro data suggest that CXCL4 alone is not sufficient to promote atherosclerosis, because compared to MCSF-induced M0 macrophages CXCL4-induced M4 macrophages express a number of atherosclerosis-related genes at higher and others at lower levels.
Thus, compared to M0 macrophages, CXCL4 induced high expression of the matrix metalloproteases 7 and 12, whereas MMP-8 was only expressed at low levels. Even though all three MMPs have been clearly implicated in atherosclerosis (34), the gene expression data need to be interpreted with caution, because the activity of MMPs is regulated by complex mechanisms involving proteolytic cleavage by cathepsins, MMPs, and serine proteases (34). Accordingly, changes in gene expression do not necessarily indicate changes in activity in vivo (34). While genes coding for the apolipoproteins APOC2 and APOE were only expressed at low levels in CXCL4 macrophages, two members of the proteolytic cathepsin family (B and K) showed high gene expression levels. Several cathepsins have been found to be overexpressed in atherosclerotic lesions and contribute to atherogenesis through different mechanisms including effects on lipid metabolism, inflammation and MMP activity (35).
Most strikingly, when looking at expression levels of genes implicated in foam cell formation, CXCL4 macrophages showed low levels of scavenger receptors necessary for uptake of modified LDL and at the same time higher levels of the cholesterol efflux transporter ABCG1. The finding that exposure of M4 macrophages to acetylated or oxidized LDL resulted in less intracellular cholesterol content than in M0 macrophages from the same donor suggests that these findings on the gene expression level translate into relevant functional differences.
In vivo, entire platelets with their granule contents are present in atherosclerotic lesions and not isolated CXCL4. Thus, CXCL4 may synergize with other platelet elements to induce a pro-atherosclerotic macrophage phenotype that is believed to be lacking in PF4-/- mice. In fact, it has been demonstrated in Apoe-/- mice that pharmacological inhibition of heterodimerization of CXCL4 with CCL5, which is also released from activated platelets, resulted in significant reduction of lesion formation (36). We recently showed that CXCL4-induced macrophage lack expression of the hemoglobin scavenger receptor CD163 and that in human atherosclerotic lesions expression of CXCL4 and CD163 are inversely correlated (23). This supports the notion that the M4 macrophage phenotype can actually be identified within human atherosclerotic lesions and may have pathophysiological relevance in atherosclerosis.
In summary, our data provide new insight into the process of macrophage differentiation. By comparing the transcriptome of MCSF- and CXCL4-induced macrophages in vitro, we identify M4 macrophages and provide novel starting points for further atherosclerosis- and other disease-related research.
Acknowledgments
We thank Keely Arbenz-Smith for excellent technical assistance.
This work was supported by a grant from the Deutsche Forschungsgemeinschaft (GL599/1-1) to C.A.G. and NIH grant 58108 to K.L.