Cholesterol Accumulation Regulates Expression of Macrophage Proteins Implicated in Proteolysis and Complement Activation
Objective—Cholesterol accumulation by macrophages plays a key role in atherogenesis. To begin to develop a global picture of this process, we used proteomics and transcriptomics to analyze foam cells generated with acetyl-low-density lipoprotein, a classic ligand for scavenger receptors.
Methods and Results—Tandem mass spectrometry and stringent statistical analysis revealed that foam cells differentially expressed 15 of 542 proteins (2.8%) detected in macrophage-conditioned medium. Apolipoprotein E was one of the most upregulated proteins, confirming that proteins involved in lipid metabolism are important targets for regulation by sterol accumulation. However, levels of proteins linked to complement activation and lysosomal proteolysis also changed markedly. Transcriptional analysis demonstrated that 698 of 19 700 genes (3.5%) were regulated in foam cells, including many genes important in sterol metabolism. We also found that cholesterol accumulation regulated genes implicated in complement activation but failed to affect genes linked to proteolysis and macrophage polarization. Changes in protein levels in macrophage-conditioned medium were largely independent of changes in mRNA levels.
Conclusion—Loading sterol into macrophages regulates levels of complement proteins and lysosomal proteases—key players in the immune system and plaque rupture. Posttranscriptional mechanisms are likely important for controlling levels of most of the proteins detected in macrophage medium.
Scavenger receptors expressed by macrophages play key roles in host defense by clearing bacterial pathogens and apoptotic cells.1 However, scavenger receptors also mediate the uptake of modified lipoproteins during atherosclerosis.2–5 When macrophages take up and degrade more lipoprotein-derived cholesterol than they can excrete, they convert free cholesterol to cholesteryl ester.2 Macrophages loaded with cholesteryl ester-laden lipid droplets—termed foam cells because of their microscopic appearance—are the hallmark of early atherosclerotic lesions.2 Moreover, genetic, biochemical, and clinical studies provide compelling evidence to indicate that macrophages engorged with cholesterol are of central importance in both the initiation and progression of atherosclerotic lesions.2–5
Macrophages do not become foam cells in vitro if incubated with low-density lipoprotein (LDL), a major risk factor for atherosclerosis, because cholesterol uptake downregulates the LDL receptor.2 In contrast, when the lysine residues of apoB, the major LDL protein, are acetylated chemically (acetyl-LDL), the modified lipoprotein binds to macrophage receptors that are not regulated by intracellular sterol content.2 Thus, macrophages incubated with acetyl-LDL in vitro rapidly endocytose the lipoprotein to become foam cells laden with cholesteryl ester.
Studies of mouse macrophages exposed to acetyl-LDL have provided important insights into the role of sterol metabolism in foam cell biology.2,3 To begin to develop a global view of this model system, we used transcriptomics and proteomics to investigate mRNA and protein expression by mouse macrophages incubated with acetyl-LDL. Our results indicate that proteins that are differentially expressed in the medium of macrophage foam cells are linked to 3 functional modules6: lipid metabolism, lysosomal biology, and complement activation. These observations raise the possibility that sterol accumulation by macrophages makes previously unsuspected contributions to the regulation of complement activation and proteolysis in atherosclerotic lesions and other inflammatory tissues.
Materials and Methods
An expanded Methods section is available in the online-only Data Supplement.
Macrophages harvested from the peritoneum of C57Bl/6J mice 5 days after injection of thioglycolate were plated, washed, and cultured in medium supplemented with 20 µg/mL of LDL or acetyl-LDL.9
Proteins were harvested from macrophage-conditioned medium, digested with trypsin, and subjected to liquid chromatography-electrospray ionization-tandem mass spectrometry (LC-ESI)-mass spectrometry/mass spectrometry (MS/MS) analysis.9 MS/MS spectra were searched against the mouse International Protein Index database10 (version 2006/04/18). Proteins were quantified by spectral counting, using dual statistical criteria and replicate analyses of each sample.9,11–13
Total cellular RNA (RNeasy Mini Kit, Qiagen) was analyzed with Affymetrix Mouse Gene 1.0 ST arrays by the Center for Array Technologies (University of Washington). Raw microarray data were processed with Affymetrix Expression Console Software, using robust multi-array average normalization (http://affymetrix.com). P values were calculated with a modified t test and an empirical Bayes method.14 P values were adjusted for multiple comparisons, using the Benjamini-Hochberg method.15 Microarray data are available from the Gene Expression Omnibus Database (accession number GSE19926).
The Gene Ontology knowledgebase and Database for Annotation, Visualization, and Integrated Discovery were used to functionally annotate proteins and genes.16 Enriched functional categories were identified relative to the entire mouse genome.
Protein and Gene Interaction Networks
Protein/gene interaction networks were created by pathway analysis,17 using Ingenuity System software, BIND, DIP, MIPS, IntAct, BioGRID, and MINT. Protein and gene networks were derived by pathway analysis of direct and indirect physical, transcriptional, and enzymatic interactions derived from a global molecular network. All nodes in the protein network were proteins detected by MS/MS analysis of macrophage-conditioned medium.
The Shed and Secreted Macrophage Proteome Is Enriched in Specific Functional Modules
We used tandem mass spectrometry to compare the proteomes of conditioned medium harvested from control and cholesteryl ester-laden macrophages generated by incubating cultured mouse peritoneal macrophages with LDL or acetyl-LDL. Acetyl-LDL treatment doubled the cells’ free cholesterol content and markedly increased their cholesteryl ester mass (for macrophages exposed to LDL or acetyl-LDL, 19.3±2.1 and 38.3±3.8 μg free cholesterol per mg protein and 0.8±0.5 and 42.3±4.3 μg cholesteryl ester per mg protein, respectively). The cells were washed and incubated with serum-free medium for 6 hours, and the macrophage-conditioned medium was subjected to liquid chromatography-electrospray ionization-tandem mass spectrometry to identify proteins.
To ensure high confidence identification of proteins, we processed our MS/MS search results with 2 Bayesian algorithms.18,19 We further required that each protein be detected in every analysis of 1 biological condition, which markedly diminishes the false-positive rate for protein detection.20 This approach generated a list of 542 proteins identified with high confidence in macrophage-conditioned medium.
To organize the shed and secreted macrophage proteome into functional modules,6 we used Gene Ontology annotations, which associate gene products with biological processes, cellular components, and molecular functions (Figure 1A). This approach identified significant enrichment (relative to the entire mouse genome) of a wide variety of functional modules: cytoskeletal regulation (P<10−4), lysosomal biology (P<10−4), oxidation (P<10−4), cell motility (P<10−4), coagulation (P<10−4), complement regulation (P=10−4), regulation of apoptosis (P=10−4), immunity (P=10−3), and extracellular matrix (P=0.03). Although we detected 14 proteins that mapped to lipid metabolism, that module was not significantly enriched in macrophage-conditioned medium (P=0.1).
Analysis of Macrophage Proteome Reveals a Network of Potential Interactions
We constructed a relational network based on interactions17 of the 542 proteins detected with high confidence in macrophage-conditioned medium. Protein interactions could be direct or indirect, and were physical, transcriptional, or enzymatic. The single, spanning network comprised 238 proteins (nodes) and 505 interactions (edges; Figure 1B).
The number of functional interactions exhibited by individual proteins in the network varied widely, indicating that the overall architecture of the network was heterogeneous.21 Thus, the 6 most highly connected proteins (Figure 1B and 1C) accounted for 30% of the interactions, whereas each of 125 proteins had only 1 or 2 interactions (25% of the total). Topological analysis of this network demonstrated that, consistent with many biological networks, it is scale free.21 Five of the 6 most highly connected proteins (coagulation factor II, fibronectin, insulin-like growth factor 1, growth factor receptor bound protein 2, and Ras homolog gene family member A) were themselves densely interconnected (Figure 1C), suggesting that they were important hubs in the network’s overall topology.
Many of the proteins in the different functional classes mapped to specific regions of the protein interaction network.22 Two examples of such subnetworks were proteins involved in the immune response (Figure 1B and 1D; purple nodes; 100% mapped to subnetwork) and those involved in complement regulation (Figure 1B and 1D; pink nodes; 100% mapped to subnetwork). These observations raise the possibility that proteins that are shed and secreted from macrophages interact to form functionally important modules in the extracellular milieu.
Dual Statistical Criteria Identify Proteins Differentially Expressed by Macrophage Foam Cells
To identify proteins that differed in relative abundance in media of cholesteryl ester-loaded and control macrophages,9 we used a log likelihood test (G-test) together with a 2-tailed t test to find significant differences in spectral counts (the sum of all peptides detected by MS/MS) for each protein.11–13 Permutation analysis23,24 revealed that requiring G>1.5 (G-test) and α<0.05 (t test) yielded the maximum number of true-positive protein identifications with an acceptable false discovery rate of 7%. Using these stringent analytical criteria, we identified 15 proteins that were differentially expressed in medium from macrophages loaded with acetyl-LDL (Table).
As assessed by these dual criteria, the fraction of differentially expressed proteins (2.8%, Table) and genes (3.5%, microarray analysis; Table I in the online-only Data Supplement) in macrophages were similar after sterol loading by exposure to acetyl-LDL.
Biochemical Validation of Differential Protein Expression by Macrophage Foam Cells
We investigated the effectiveness of our analytical strategy by biochemically quantifying changes in relative abundance for 3 of the 15 macrophage proteins that met both of our statistical criteria (Table). Immunoblot analysis of medium isolated from macrophages incubated with acetyl-LDL demonstrated significant changes in levels of all 3 proteins (Figure 2A and 2B): apolipoprotein E, P=0.007; cathepsin L, P=0.005; and cathepsin D, P=0.016.
Differentially Expressed Macrophage Proteins Map to Modules Implicated in Lysosomal Biology, Lipid Metabolism, and Complement Regulation
We used proteins that were differentially expressed by macrophage foam cells to identify a subnetwork of the full protein interaction network (Figure 1B) that is regulated when cholesterol accumulates in macrophages. Inclusion of first neighbors yielded a spanning subnetwork centered on the differentially expressed proteins (Figure 2C). Importantly, many of the first neighbors included in the subnetwork (α-2-macroglobulin, complement component C1Q β chain, C1Q γ chain, heat shock protein-5, integrin β-2, galectin-3, and Y box protein-1) were differentially expressed by macrophage foam cells when only the t test or G test was used to assess significance, suggesting that they might also be regulated by sterol loading.
Gene Ontology analysis of the network of sterol-regulated proteins identified significant enrichment in 3 functional modules: lysosomal biology (P<10−4), complement activation (P<10−4), and lipid metabolism (P=0.03; Figure 2C). Taken together, our observations indicate that proteins linked to lipid metabolism, lysosomes, and complement activation are differentially expressed by macrophage foam cells.
Statistical and Fold-Change Criteria Identify Genes Differentially Expressed by Macrophage Foam Cells
To investigate the role of altered mRNA levels in regulating protein expression by foam cells, we analyzed total RNA extracted from control macrophages and macrophages exposed to acetyl-LDL. For a change to be considered significant, we required genes to show an expression change of ≥1.5-fold with P<0.05 as assessed by Student t test (adjusted for multiple hypothesis testing). On the basis of these criteria, we identified 449 genes that were upregulated and 249 genes that were downregulated (Table I in the online-only Data Supplement). Consistent with previous studies,25 genes involved in cholesterol synthesis were a major target for downregulation in macrophages exposed to acetyl-LDL.
CD5L, a secreted protein that is a member of the scavenger receptor cysteine rich superfamily, showed the greatest increase in both mRNA levels (8-fold) and protein levels (Table) when macrophages were exposed to acetyl-LDL. CD5L is also highly prevalent on B-1 cells, but we failed to detect mRNA characteristic of these cells (such as CD20) in our microarray data; we also failed to detect the protein in the conditioned medium of elicited macrophages.9
Multiple genes that regulate complement activation were also upregulated in macrophages incubated with acetyl-LDL. Four striking examples were C1qa, C1qb, and C1q, which are key early components in the classic pathway for complement activation, and Cfp (properdin), a component of the alternative pathway for complement activation.
Differentially Expressed Macrophage Genes Map to Modules Implicated in Lipid and Protein Metabolism, Complement Activation, and Oxidative Phosphorylation
We used Gene Ontology annotation to organize acetyl-LDL regulated genes into functional modules. This approach identified significant enrichment in a wide variety of modules: complement regulation (P<10−4), ribosomes (P<10−4), lipid metabolism (P<10−4), proton transporting ATP synthase complex (P<10−4), intracellular organelles (P<10−4), oxidoreductase activity (P=0.01), immune response (P=0.02), chemotaxis (P=0.02), and phagocytosis (P=0.04). Mapping differentially expressed genes onto the Kyoto Encyclopedia of Genes and Genomes integrated database26 indicated that 3 major metabolic pathways were regulated in macrophage foam cells: steroid biosynthesis, the large and small ribosomal subunits, and oxidative phosphorylation (Figures I–III in the online-only Data Supplement). These observations suggest that the major impact of loading sterol into macrophages is on protein and lipid metabolism, mitochondrial function, and complement activation. It is noteworthy that we also detected enrichment of complement proteins in conditioned medium of macrophage foam cells,9 and that recent studies implicate oxidative phosphorylation in the function of alternatively activated macrophages.27
We constructed an interaction network based on pathway analysis of the 698 genes differentially expressed in macrophage foam cells. The single, spanning network contained 179 interactions between 144 nodes (cholesterol and 143 genes; Figure 3). Many of the genes clustered into specific regions of the gene interaction network. For example, Ndufs2, Ndufs2, Ndufs8, and Ndffs5 are all mitochondrial oxidoreductases. This approach indicates that multiple genes linked to lipid metabolism, oxidoreductase activity, ribosomes, and complement activation could potentially interact with one another to form a functional network.
Cholesterol Loading Fails to Affect Macrophage Polarization But Suppresses Lipopolysaccharide Induction of Inflammatory Genes
The release of proinflammatory cytokines is markedly increased when peritoneal macrophages deficient in ATP-binding cassette subfamily A member 1 (ABCA1) and ATP-binding cassette subfamily G member 1 are stimulated with lipopolysaccharide (LPS), leading to the proposal that cholesterol loading renders the cells proinflammatory.28,29 However, we observed a different response in peritoneal macrophages incubated with acetyl-LDL: sterol loading blunted the response of the macrophages to LPS without altering the overall pattern of LPS-induced gene expression (Figure 4D). Moreover, we observed no consistently altered expression of pro- (M1) or anti-inflammatory (M2) genes in macrophages exposed to acetyl-LDL (Figure 4B and 4C). The pattern of gene expression we observed in LPS-stimulated peritoneal macrophages was remarkably similar to that reported for macrophages derived from bone marrow cells.30 These data indicate that loading macrophages with cholesterol derived from acetyl-LDL does not consistently affect polarization or enhance the inflammatory response when macrophages are subsequently challenged with LPS.
Differential Protein Expression in Macrophage-Conditioned Medium Is Largely Independent of Differences in mRNA Expression
We used 2 complementary approaches to explore the relationship between mRNA and protein levels in macrophage-conditioned medium. First, to provide a general overview, we performed this analysis using all proteins in macrophage-conditioned medium that exhibited an average of >2 spectral counts by MS/MS analysis in both control and macrophage foam cells (Figure 5A). We used this criterion because it is difficult to use spectral counting to quantify low-abundance proteins.11,13 For the 313 proteins that met this criterion, no correlation was observed between the expression of proteins in conditioned medium of control and acetyl-LDL exposed cells and their cognate mRNAs (P=0.84, r2<0.01).
Second, we examined the relationship between the proteins and mRNAs that exhibited the largest differences in relative values between control and cholesterol-loaded cells (Figure 5B). For this analysis, we included the following: (1) all genes that exhibited a significant, >2-fold change at the mRNA level (as assessed by the t test) and that were detected by MS/MS analysis, and (2) all 15 proteins that were differentially expressed as assessed by both the G test and t test. For this analysis we excluded CD5L, because the protein was not detected in the medium of control macrophages, yielding a ratio of differential protein expression that was infinite. The correlation between protein level and mRNA level was borderline significant (P=0.069, r2=0.16). Collectively, these observations indicate that mRNA can be an important regulator of protein expression if its relative level increases significantly, but that there is little overall relationship between changes in mRNA levels and protein levels in macrophage-conditioned medium. However, we did observe that similar functional categories—lipid metabolism and the complement pathway—were regulated by sterol loading at both the protein and mRNA levels.
Macrophages incubated with acetyl-LDL are a widely used model system for studying foam cell biology.2 To begin to develop a global view of the foam cell, we used the 15 proteins that were differentially expressed by cultured macrophage foam cells and their first neighbors to construct a subnetwork of shed and secreted proteins. This approach linked functional modules involved in lysosomal proteolysis, lipid metabolism, and complement regulation to cholesterol deposition in macrophages.6
In macrophages, lysosomal digestion of endocytosed lipoproteins promotes cholesterol accumulation,2,31 and proteolysis of prosaposin by cathepsins regulates the expression of ABCA1.31 Thus, lysosomal proteases are likely to be important in macrophage cholesterol homeostasis. Studies of hypercholesterolemic mice have implicated lysosomal proteases and cystatin C in atherogenesis and degradation of structural proteins in vascular lesions.32 Proteolysis and regulation of cholesterol accumulation in macrophages may play a critical role in triggering plaque rupture, the major cause of myocardial infarction and sudden death in humans with coronary artery disease.33
Quantification of transcribed mRNA has been widely used to identify pathways induced by targeted perturbations of cells.34 We therefore complemented our proteomic studies by using microarrays to analyze mRNA from control and cholesteryl ester-laden macrophages. These studies confirmed previous observations that multiple genes involved in cholesterol biosynthesis are downregulated in cholesteryl ester-laden macrophages and that certain genes, such as Abca1, that promote cholesterol excretion are upregulated.3,5,25
Unexpectedly, we also found that loading cholesterol into macrophages regulated the expression of multiple complement regulatory proteins at both the mRNA and protein levels. It is noteworthy that several of those proteins, including C1q and C3, are of central importance in triggering complement activation, and that activated components of the complement pathway have been detected in advanced human atherosclerotic lesions.36 Altered release of lysosomal proteases and complement regulatory proteins by macrophage foam cells may therefore promote plaque rupture, ischemic injury, and activation of the coagulant cascade—key pathogenic events in human atherosclerosis.
It is generally believed that the conversion of macrophages to foam cells is proinflammatory, but several lines of evidence indicate that this is not the case for peritoneal macrophages exposed to acetyl-LDL. First, we found no evidence for increased expression of inflammatory proteins induced by cholesterol loading of macrophages at both the mRNA and protein level. Second, we observed no consistent differences in the expression of a wide range of genes used to monitor the pro- (M1) and anti-inflammatory (M2) phenotypes.37 Third, we demonstrated that sterol loading blunted the response of the macrophages to LPS without altering the overall pattern of LPS-induced gene expression. Collectively, these observations strongly suggest that macrophages are not polarized toward a proinflammatory state when they are cholesterolloaded by incubation with acetyl-LDL.
A strength of our proteomics approach is the unbiased nature of the analysis and the emphasis on protein (as opposed to mRNA) expression. A limitation of untargeted MS/MS is the inability to detect low-abundance proteins, such as cytokines.13 However, we also observed no consistent increases in mRNA levels for a wide range of inflammatory cytokines or chemokines. In particular, we observed no increased expression of classic inflammatory genes, such as intercellular adhesion molecule–1, tumor necrosis factor, or macrophage chemotactic protein-1.
Different results have been reported for ABCA1- and ABCA1/ATP-binding cassette subfamily G member 1-deficient macrophages, which have elevated levels of cholesterol and clearly exhibit a proinflammatory phenotype.28,29 It is important to note that deficiency of those proteins impairs the efflux of free cholesterol from the macrophage plasma membrane. The resulting increase in free cholesterol has been proposed to alter cell signaling by proinflammatory membrane-associated receptors, such as Toll-like receptor 4.37,38
Modified lipoproteins increase cellular cholesterol levels by a completely different mechanism. In that system, endocytosis delivers lipoproteins to lysosomes, which release free cholesterol that is then delivered to the endoplasmic reticulum.9,25 The increased level of free cholesterol in the endoplasmic reticulum activates acetyl-coenzyme A acetyltransferase. This enzyme converts free cholesterol to cholesteryl ester, which is subsequently stored in lipid droplets in the cells. Under these conditions, one would expect little or no change in the free cholesterol content of the plasma membrane and no activation of cell surface receptors, such as TLR4. These observations suggest that ABCA1/ATP-binding cassette subfamily G member 1 deficiency and endocytosis of modified LDL increase the concentration of free cholesterol in different cellular pools, and that this in turn is likely to have different effects on the phenotype of macrophages.
Changes in mRNA levels do not necessarily predict changes in protein levels.35,39 We therefore examined the relationship between changes in mRNA and protein levels induced by loading cholesterol into macrophages. Overall there was no correlation between the 2 even if we limited the analysis to genes that exhibited a significant change. However, we observed a large increase in both mRNA and CD5L protein in the medium of macrophages incubated with acetyl-LDL. We also found that 2 functional modules, lipid metabolism and complement activation, were regulated at both the mRNA and protein level, suggesting coregulation by pre- and posttranscriptional mechanisms. However, posttranscriptional mechanisms seem to be important for the regulation of most of the proteins we detected in the medium of macrophages. Collectively, these observations suggest that mRNA can play a major role in regulating macrophage protein expression if its relative level is markedly altered or if multiple genes in a pathway are coordinately regulated.25,34
In summary, our studies indicate that the complement pathway and lysosomal proteins—key components of the immune system and potential players in plaque rupture—are regulated when mouse macrophages are loaded with cholesterol by exposure to acetyl-LDL. We also observed the upregulation of multiple genes linked to oxidative phosphorylation and ribosomal protein synthesis, but not in genes linked to macrophage polarization. Our findings have identified previously unsuspected links among macrophage sterol accumulation, complement regulation, and proteolysis, all of which are implicated in atherogenesis.
Sources of Funding
This research was supported by grants from the National Institutes of Health (HL108897, HL092969, HL112625). Microarray analyses were performed by the Center for Array Technologies (University of Washington).
The online-only Data Supplement is available with this article at http://atvb.ahajournals.org/lookup/suppl/doi:10.1161/ATVBAHA.112.300383/-/DC1.
- Received May 17, 2012.
- Accepted September 13, 2012.
- © 2012 American Heart Association, Inc.
- Greaves DR,
- Gordon S
- Ory DS
- Mendez AJ,
- Oram JF,
- Bierman EL
- Old WM,
- Meyer-Arendt K,
- Aveline-Wolf L,
- Pierce KG,
- Mendoza A,
- Sevinsky JR,
- Resing KA,
- Ahn NG
- Benjamini Y,
- Hochberg Y
- Heinecke NL,
- Pratt BS,
- Vaisar T,
- Becker L
- Kanehisa M,
- Goto S,
- Kawashima S,
- Nakaya A
- Zhu X,
- Lee JY,
- Timmins JM,
- Brown JM,
- Boudyguina E,
- Mulya A,
- Gebre AK,
- Willingham MC,
- Hiltbold EM,
- Mishra N,
- Maeda N,
- Parks JS
- Yvan-Charvet L,
- Welch C,
- Pagler TA,
- Ranalletta M,
- Lamkanfi M,
- Han S,
- Ishibashi M,
- Li R,
- Wang N,
- Tall AR
- Haidar B,
- Kiss RS,
- Sarov-Blat L,
- Brunet R,
- Harder C,
- McPherson R,
- Marcel YL
- Liu J,
- Sukhova GK,
- Sun JS,
- Xu WH,
- Libby P,
- Shi GP
- Moreno PR,
- Falk E,
- Palacios IF,
- Newell JB,
- Fuster V,
- Fallon JT
- Oram JF,
- Heinecke JW
- Gygi SP,
- Rochon Y,
- Franza BR,
- Aebersold R