Published online before print November 12, 2008
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
,||,¶,1
* The University of Queensland, Institute for Molecular Bioscience, Brisbane, Queensland, Australia;
Cooperative Research Centre for Chronic Inflammatory Diseases, The University of Queensland, Brisbane, Queensland, Australia;
Cytopia Pty. Ltd., Richmond, Victoria, Australia;
The Roslin Institute and Royal (Dick) School of Veterinary Studies, University of Edinburgh, Roslin, Scotland, United Kingdom;
|| The University of Queensland, Brisbane, School of Molecular and Microbial Sciences, Brisbane, Queensland, Australia; and
¶ The University of Queensland, Division of Molecular Cell Biology, Institute for Molecular Bioscience, and Centre for Microscopy and Microanalysis, Brisbane, Queensland, Australia
1 Correspondence: Institute for Molecular Bioscience, The University of Queensland, St. Lucia, Brisbane, Queensland, 4072, Australia. E-mail: m.sweet{at}imb.uq.edu.au
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Key Words: inflammation • chemokines • cholesterol
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Macrophage populations differentiated under the influence of different CSFs have distinct phenotypes and functions. For example, mouse macrophages differentiated in the presence of GM-CSF exhibit a more proinflammatory phenotype than those differentiated in CSF-1 [6 ]. Differential effects of GM-CSF versus CSF-1 are also apparent during human monocyte-to-macrophage differentiation [7 , 8 ], with CSF-1-induced human macrophage differentiation favoring an M2-polarized phenotype, as assessed by transcriptional profiling [9 ]. This phenotype is likely to predominate under homeostatic conditions, as CSF-1 is constitutively present in vivo (2–30 ng/ml in normal human serum) [10 , 11 ]. CSF-1 levels are elevated further in various pathological states [12 ], and several lines of evidence suggest that this growth factor has additional functions in regulating mature macrophage function in such settings.
In mice, CSF-1 production increases in response to TNF, IFN-
, and IgG aggregates [13
], modified low-density lipoproteins (LDL) [14
], and LPS [15
], thus implying some role in inflammation. Indeed, CSF-1 regulates several mature macrophage functions including chemotaxis, adherence, antimicrobial responses, and antibody-dependent cellular toxicity (for recent reviews, see refs. [12
, 16
]). For example, CSF-1 promotes cellular migration by acting directly as a chemotactic factor [17
] and by up-regulating the expression of the urokinase plasminogen activator (uPA) [18
]. It also primes mouse macrophages for enhanced TNF, IL-6, and IL-12 production in response to LPS [19
, 20
]. Such data imply a proinflammatory role for CSF-1, and in support of this conclusion, an anti-CSF-1 antibody reduced disease severity in a mouse model of rheumatoid arthritis, and administration of exogenous CSF-1 exacerbated disease [21
]. A recent study by Stanley and co-workers [22
] highlighted the importance of local production of CSF-1 in eliciting inflammatory responses. Macrophages from mice engineered to express CSF-1 in the same cells that express the CSF-1R produced higher levels of proinflammatory cytokine mRNAs and were primed to secrete elevated levels of these cytokines in response to LPS. This phenotype was not dependent on CSF-1 secretion and was actually enhanced compared with mice in which CSF-1 was systemically elevated [22
].
The extensive literature documenting the involvement of CSF-1 in inflammation [12 , 16 , 23 ] has led to the development of CSF-1R kinase inhibitors [20 , 24 25 26 27 28 ]. This strategy is based primarily on findings in mice, however; whether dysregulated levels of CSF-1 actually contribute causally to inflammatory disease progression in humans is unknown, as relatively few studies have focused on the acute effects of this factor on mature human macrophage function. Although some studies have performed detailed gene expression analyses of CSF-1-induced human monocyte-to-macrophage differentiation [8 , 9 , 29 ], none have specifically explored the biological effects of acute CSF-1 stimulation in mature macrophages. As the contribution of the CSF-1/CSF-1R signaling system to inflammatory disease in humans is unclear, and a new ligand for the CSF-1R was identified recently [5 ], we investigated the CSF-1 response in human monocyte-derived macrophages (HMDM) after acute CSF-1 stimulation. In so doing, we identified CSF-1-regulated genes and pathways that are likely to impact on inflammatory and cardiovascular disease.
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Total RNA isolation and quantitative (q)PCR
Cells for expression analysis were lysed in situ, and total RNA was prepared from 1 to 3 x 106 cells using the Qiagen RNeasy Mini kit with on-column DNaseI treatment, according to the manufacturer’s instructions. RNA was reverse-transcribed using Superscript III (Invitrogen Life Technologies) and an oligo-dT primer. Gene expression was quantitated by real-time PCR using SYBR Green PCR master mix (Applied Biosystems, Foster City, CA, USA) and an Applied Biosystems Prism 7000 or 7500 sequence detector. Amplification was achieved using an initial cycle of 50°C for 2 min and 95°C for 10 min, followed by 40 cycles of 95°C for 15 s and 50°C for 1 min. cDNA levels during the linear phase of amplification were normalized against hypoxanthine guanine phosphoribosyl transferase (HPRT) controls using the power delta Ct method. Data, presented as fold change compared with untreated cells (mean+SEM) were calculated from four to five independent HMDM preparations and a total of four to six BMM preparations from C57/Bl6J (two) and BALB/c (two to four) backgrounds. As no differences in CSF-1 responses were observed between BMM from C57/Bl6J and BALB/c mice, data from both mouse strains were combined. Statistical significance was calculated using a one-sample t-test to compare CSF-1-stimulated population means with controls (*, P<0.05; **, P<0.005; ***, P<0.0005). The Pearson correlation coefficient (with two-tailed test for significance; see 
Fig. 5C
) was calculated in Prism. Gene-specific primer pairs were designed with an optimal primer size of 20 bases, amplicon size of 100 bp, and annealing temperature of 65°C. Where possible, matched primers detecting mouse and human orthologs were designed to target homologous regions in the genes. CSF-1 isoform primer efficiency was tested over a cDNA dilution series, and all three primer pairs amplified their targets with approximately equal efficiency. Primer sequences used for RT-PCR amplification appear in Table 1
, where h and m denote human and mouse target genes, respectively.
 View larger version (19K): [in a new window] |
Figure 1. Classic CSF-1 responses in mouse BMM are not conserved in HMDM. HMDM (solid bars) or BMM (open bars), starved of CSF-1 overnight, were treated with CSF-1 over a time course. UPA (A) and APOE (B) mRNA expression was quantitated by qRT-PCR and expressed as fold change compared with unstimulated control (0 h). Plots represent mean fold change + SEM for five independent HMDM preparations and six independent BMM preparations (2xC57/Bl6J; 4xBALB/c). Statistical significance compared with unstimulated controls is indicated (***, P<0.0005). HMDM or BMM were plated overnight with (open bar) or without (solid bars) CSF-1, followed by stimulation with LPS for 24 h. Supernatants were harvested, and TNF (C) and IL-6 (D) levels were determined by ELISA. Plots are representative of greater than or equal to three independent HMDM and BMM experiments. (E) HMDM were CSF-1-starved overnight, followed by 5 min stimulation with CSF-1 or medium alone. Cell-surface CSF-1R expression was monitored by flow cytometry (profile representative of four independent experiments).
|
 View larger version (20K): [in a new window] |
Figure 2. CSF-1 down-regulates CXCR4 expression in HMDM (A). HMDM were CSF-1-starved overnight in the presence of 5 µM GW2580 (GW.), imatinib mesylate (Imat.), or vehicle (Veh.) followed by stimulation with CSF-1 for 24 h. CXCR4 mRNA expression was determined by qRT-PCR (average of triplicates+SD). (B) CSF-1-mediated survival in the presence of GW2580 was monitored in Ba/F3 cells engineered to express full-length human CSF-1R or the intracellular domain of human c-Kit or Flt3 spliced to the CSF-1R ligand-binding domain. Ba/F3 cells were incubated for 48 h in the presence of CSF-1 and increasing concentrations of GW2580 or DMSO vehicle before cell survival was measured by MTT assay. Data are expressed as percent vehicle survival and represent mean and SEM from three independent experiments. (C) HMDM were CSF-1-starved overnight in the presence of 5 µM GW2580 (open bars) or DMSO vehicle (solid bars) as indicated, followed by stimulation with CSF-1 for the indicated times. CXCR4 mRNA expression was determined by qRT-PCR (average of triplicates+SD). (D) Fold change in CXCR4 expression after 24 h CSF-1 stimulation was calculated using HMDM from 14 independent donors (P<0.0001). (E) Fold change in CXCR4 expression over a time course of CSF-1 stimulation was calculated in HMDM from four independent donors (solid bars) and four BMM preparations (2xBALB/c; 2xC57Bl6J; open bars). Statistical significance compared with unstimulated controls is indicated (**, P<0.005; ***, P<0.0005).
|
 View larger version (32K): [in a new window] |
Figure 3. CSF-1 regulates chemokine expression in HMDM. (A) HMDM (solid bars) or BMM (open bars) were CSF-1-starved overnight, followed by stimulation with CSF-1 for the indicated times. CCL2 (i), CCL7 (ii), CXCL2 (iii), and CXCL10 (iv) expression was quantitated by RT-PCR and expressed as fold change compared with unstimulated control (0 h). Plots represent mean fold change + SEM for HMDM from five independent donors (solid bars) and four independent BMM preparations (2xC57/Bl6J; 2xBALB/c; open bars). Statistical significance compared with unstimulated controls is indicated (*, P<0.05; **, P<0.005; ***, P<0.0005). (B) HMDM were CSF-1-starved overnight in the presence of 5 µM GW2580 (open bars) or DMSO vehicle (solid bars), followed by stimulation with CSF-1 for the indicated times. CXCL10 (i) and CCL7 (ii) expression was quantitated by RT-PCR. (iii) Cell culture supernatants were collected from HMDM that were CSF-1-starved or stimulated with CSF-1 for 24 h in the presence of 5 µM GW2580 or DMSO vehicle and assayed for CXCL10 by ELISA.
|
 View larger version (32K): [in a new window] |
Figure 4. CSF-1 regulates cholesterol metabolism in HMDM, which were CSF-1-starved overnight, followed by stimulation with CSF-1 for the indicated times. HMGCR (A), MVD (B), DHCR7 (C), and ABCG1 (D) expression was quantitated by RT-PCR and expressed as fold change compared with unstimulated control (0 h). Plots represent mean fold change + SEM for HMDM from five independent donor preparations. Statistical significance compared with unstimulated controls is indicated (*, P<0.05; **, P<0.005; ***, P<0.0005). (E) CSF-1-starved HMDM from two donors were stimulated with CSF-1 or cholesterol or both for 6 h. HMGCR and DHCR7 expression was quantitated by RT-PCR. Data represent mean fold change + range for HMDM from two independent donor preparations. (F) HMDM from two donors were CSF-1-starved or treated with 5 µM GW2580 or DMSO vehicle overnight, followed by stimulation with CSF-1 for 6 h. Cells were fixed and stained with Filipin and Oil Red O to quantify cellular cholesterol. Filipin fluorescence intensity per cell was calculated and expressed as mean + SEM fluorescence units (FLU)/cell for each donor (*, P<0.05; **, P<0.005; n.s., not significant).
|
 View larger version (27K): [in a new window] |
Figure 5. Autocrine CSF-1 variably regulates HMDM gene expression. (A) HMDM were incubated overnight in the presence of 5 µM GW2580 or DMSO vehicle, with or without CSF-1 stimulation. CXCR4 expression was quantitated by RT-PCR. The two donor profiles were selected for their distinct profiles from eight similar experiments. (B) Relative CSF-1 expression (all isoforms) in HMDM from Donors 1 and 2 depicted in A. (C) CSF-1 expression (relative to HPRT) in CSF-1-starved HMDM was plotted against CXCR4 repression (relative expression of CXCR4/HPRT in CSF-1-stimulated/unstimulated HMDM) observed in CSF-1-stimulated HMDM from 13 donors, yielding a highly significant (P =0.001) Pearson correlation coefficient of r = 0.8. (D) HMDM were incubated overnight in the presence of 5 µM GW2580 or DMSO vehicle, with or without CSF-1 stimulation. Surface CXCR4 was analyzed by flow cytometry. Profiles are representative of four independent experiments. (E) Diagram depicting the three CSF-1 protein isoforms and their encoding transcripts. Sites of proteolytic cleavage (arrows) and glycosaminoglycan addition (GAG+), involved in post-translational processing of CSF-1 precursors, and the transmembrane domain (TM) are indicated. (F) cDNA from CD14+ monocytes and differentiated MDM (matched samples from three donors) was used to quantitate the expression of the CSF-1 mRNA isoforms A, B, and C. Relative expression in HMDM compared with monocytes was calculated for each donor and expressed as mean fold change + SEM for the three experiments.
|
|
View this table: [in a new window] |
Table 1. Real Time PCR Primers
|
Measurement of proinflammatory cytokine release by ELISA
BMM or HMDM were plated in 24-well plates at 5 x 105 cells per well in 1 ml complete medium with or without CSF-1 (104U/ml). The next morning, cells were stimulated with 10 ng/ml LPS, and supernatants were collected after 24 h. IL-6, TNF, and IFN-inducible protein 10 (IP-10) levels were estimated in triplicate using sandwich ELISAs (BD PharMingen, San Diego, CA, USA).
Flow cytometry
For CSF-1R cell-surface analysis, HMDM were CSF-1-starved overnight, followed by stimulation with 104U/ml CSF-1 for 5 min. HMDM were harvested rapidly by gentle scraping in ice-cold PBS/0.5% BSA/0.1% NaAzide (PBA), blocked with 10% human serum, and 1 x 106 cells were stained with PE-conjugated mouse anti-human CSF-1R antibody or matched isotype control (R&D Systems, Minneapolis, MN, USA). For CXCR4 surface analysis, HMDM were starved of CSF-1 overnight or stimulated with 104U/ml CSF-1 in the presence of 5 µM GW2580 or DMSO vehicle. Cells (1x106) were harvested in PBA, blocked with 10% human serum, and stained with 0.5 µg anti-CXCR4 (1G4, R&D Systems) or anti-V5 (Serotec, UK) as an isotype control, followed by staining with a FitC-conjugated goat anti-mouse secondary antibody. Immunostained samples were analyzed using a FACSCalibur cytometer (Becton Dickinson, San Jose, CA, USA).
Free cholesterol quantification
HMDM (0.5x105/cm2) were seeded on glass coverslips and CSF-1-starved or treated with GW2580 for 16–20 h, followed by stimulation with CSF-1 for 6 h as indicated. Cells were fixed in 4% paraformaldehyde, permeabilized in PBS/0.5% BSA with 0.1% saponin, followed by sequential staining with 1 mg/ml Filipin in PBS/0.5% BSA (Sigma-Aldrich) and Oil Red O (65% in isopropanol, Sigma-Aldrich). Images (10–20; >100 cells) per treatment were collected randomly and analyzed using ImageJ (http://rsbweb.nih.gov/ij/). Filipin fluorescence intensity per cell was calculated by dividing total fluorescence (integrated density) by the number of cells per frame (manually counted) for 10–20 randomly selected frames per treatment (>100 cells). Fluorescence/cell was averaged over all frames per treatment and expressed as mean + SEM.
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Identification of novel CSF-1 target genes in human macrophages
To identify genes that require constant CSF-1 signaling in mature human macrophages, we used a microarray approach. Gene expression in HMDM, prepared from three independent blood donations stimulated with CSF-1 for 6 h compared with untreated controls, was analyzed on an Illumina Sentrix6 bead array. Notwithstanding the fact that autocrine production of CSF-1, a factor that is highly variable between HMDM from different individuals (see below), could affect constitutive expression of CSF-1 target genes, we saw robust changes in gene expression in response to exogenous CSF-1, which coordinately regulated 120 genes in HMDM from all three donors (87 induced; 33 repressed; Supplemental Table 1
). These genes were analyzed for GO over-representation using DAVID [30
]. Significantly enriched biological processes, functions, and cellular components are listed in Table 2
. CSF-1 altered transcription of genes involved in several biological processes in mature macrophages, but two prominent effects were upon the expression of the chemokine/chemokine receptor repertoire and genes encoding enzymes of the cholesterol biosynthetic pathway.
|
View this table: [in a new window] |
Table 2. Enriched among CSF-1-Regulated Genes in HMDM
|
In addition to the chemokine receptor CXCR4, CSF-1 regulated several chemokine mRNAs in HMDM. We confirmed the microarray data by real-time PCR over a CSF-1 time course. CSF-1 acutely up-regulated mRNAs encoding the C–C chemokines CCL2/MCP-1 and CCL7/MCP-3 (Fig. 3A ). In contrast to the heightened expression of chemokines associated with monocyte and lymphocyte recruitment (CCL2, CCL7), CXCL2/MIP-2, a neutrophil chemoattractant, was repressed in HMDM over the same time course (Fig. 3A) . CXCL10/IP-10, a chemokine classically associated with Th1 polarization and recruitment [34 ], was also CSF-1-inducible in HMDM (Fig. 3A) . For each of these chemokines, the CSF-1-regulated response in HMDM was antagonized by GW2580 (Fig. 3B , and data not shown). Inducible IP-10 expression, as well as antagonism by GW2580, was also apparent at the protein level (Fig. 3B) , although some donor variability was observed for this response. We conclude that CSF-1 reprograms chemokine responsiveness (e.g., CXCR4) as well as the chemokine-secretion profile of HMDM. Some of these novel CSF-1 responses were also observed in mouse BMM, namely CCL2 and CCL7 induction (Fig. 3A) , but there was also considerable divergence. For example, CXCL2 was CSF-1-repressed in HMDM but induced in BMM. Conversely, CXCL10 was CSF-1-inducible in HMDM but down-regulated modestly in BMM (Fig. 3A) .
CSF-1 regulates cholesterol biosynthesis and lipid metabolism in HMDM
The most significantly over-represented ontology among CSF-1-regulated genes was that of lipid metabolism and, in particular, cholesterol biosynthesis. CSF-1 significantly increased mRNA expression of at least 10 enzymes of the cholesterol biosynthetic pathway in mature human macrophages (Table 3
). These include HMGCR, which catalyzes the rate-limiting step in cholesterol biosynthesis, as well as MVD, IDI1, and FDPS, which are involved in conversion of C2 acetyl CoA to squalene, the first isoprenoid intermediate committed to cholesterol synthesis [35
]. Downstream of squalene, CSF-1 up-regulated transcripts for SQLE, CYP51A1, EBP, NSDHL, as well as DHCR7 and DHCR24 that catalyze the final step of cholesterol synthesis [35
]. We validated some of these findings by real-time PCR (HMGCR, MVD, and DHCR7; Fig. 4 A-C
). CSF-1 also transiently repressed ABCG1 mRNA, which encodes the ATP-binding cassette protein G1 that mediates reverse cholesterol transport to lipid-rich apolipoproteins in macrophages [36
] (Fig. 4D)
. Cholesterol biosynthesis is regulated tightly by a negative-feedback loop, in which excess cholesterol leads to transcriptional repression of cholesterol biosynthetic enzymes [37
]. Consequently, the addition of exogenous cholesterol blocked CSF-1-mediated up-regulation of HMGCR and DHCR7 (Fig. 4E)
. Furthermore, acute CSF-1 stimulation increased levels of free cholesterol greater than or equal to twofold in HMDM, and this response was blocked by the CSF-1R kinase inhibitor GW2580 (Fig. 4F)
.
|
View this table: [in a new window] |
Table 3. Enzymes of the Cholesterol Biosynthetic Pathway and their Fold Change in Expression in HMDM after 6 h CSF-1 Stimulation
|
Mammalian cells produce three biologically active CSF-1 isoforms [a soluble, circulating glycoprotein; a membrane-tethered glycoprotein; and an extracellular matrix (ECM)-anchored proteoglycan], and the primers used in Figure 5B detected all three isoforms. As emerging literature describes distinct functions for different CSF-1 isoforms [3 ], we designed real-time PCR primers to distinguish among the transcripts encoding all CSF-1 protein isoforms: NM_000757.3/NM_17221.1 encoding secreted glycoprotein and secreted proteoglycan CSF-1 (CSF-1-A), NM_172211.1 encoding cell-surface glycoprotein CSF-1 (CSF-1-C), and NM_172210.1 encoding a third CSF-1 that has not been investigated experimentally (CSF-1-B; Fig. 5E ). It is not yet clear how CSF-1-B processing occurs; it may exist as a cell-surface and secreted protein. The first five exons of the CSF-1 transcript encode biologically active CSF-1, whilst splicing in Exon 6 determines how the cytokine is processed, and alternative use of Exons 9 and 10 produce distinct 3' untranslated regions (UTRs) that control message stability [38 ]. The expression of alternative 3' UTRs was not considered here. Although the canonical CSF-1-A transcript is generally thought to be the dominant CSF-1 isoform, all three CSF-1 transcripts were expressed in HMDM and were strongly up-regulated (more than tenfold) during monocyte-to-macrophage differentiation (Fig. 5F) . Although there are further levels of regulation involved in translating and processing CSF-1 into its biologically active forms, this observation supports a role for autocrine/paracrine CSF-1 in human macrophage maturation and function.
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Certain proinflammatory effects of CSF-1, apparent in mouse BMM (e.g., amplification of LPS-induced TNF and IL-6), were not observed in HMDM (Fig. 1 C and D) . These effects are well established in the mouse, in vitro and in vivo, in a range of macrophage populations [19 , 20 , 42 ]. Although CSF-1 did not prime TNF and IL-6 production in HMDM, several of the CSF-1-regulated chemokines that we identified in HMDM have well-documented roles in inflammation. Acute CSF-1 stimulation increased expression of the monocyte- and lymphocyte-attracting chemokines CCL2, CCL7 (both of which signal via CCR2), and CXCL10, while repressing transcription of the neutrophil chemoattractant CXCL2 (Fig. 3) . CSF-1 could thus be anticipated to alter the balance of immune effector cells in tissue microenvironments, favoring monocytes and lymphocytes over neutrophils. CCL2/MCP-1 plays a major, nonredundant role in monocyte recruitment into tissues during inflammation and has been implicated in a wide range of inflammatory diseases characterized by monocyte-rich infiltrates, including atherosclerosis, rheumatoid arthritis, and multiple sclerosis (reviewed in ref. [43 ]). Similarly, CCL7/MCP-3 was required for monocyte mobilization from bone marrow and recruitment to inflammatory sites in mice [44 ]. CXCL10/IP-10, a potent chemoattractant for monocytes and activated T lymphocytes, enhances inflammatory cell adhesion to the vascular endothelium, further facilitating their recruitment to sites of inflammation [45 ]. Although low levels of IP-10 are detected in normal human serum, increases in circulating IP-10 are associated with a number of chronic inflammatory diseases including rheumatoid arthritis, atherosclerosis, and diabetes [46 47 48 ]. It is therefore possible that this increase is influenced by an elevation in local or circulating CSF-1, which is also commonly associated with such conditions [12 , 49 ].
Given the evidence supporting a proatherogenic function for CSF-1 in mice [40 , 50 , 51 ] and disparate data suggesting CSF-1 is atheroprotective in rabbits [52 53 54 ], it is interesting that a common thread linking the CSF-1-responsive genes, identified in this study about human macrophages, is their association with atherosclerosis. All three CSF-1-inducible chemokines identified here have been implicated previously in atherosclerosis, with the evidence for CCL2 involvement being particularly strong [43 , 46 , 55 ]. The recruitment of circulating monocytes to atherosclerotic lesions was absolutely dependent on the receptor for CCL2, CCL7, and CCR2 [56 ]. In addition to inducing proatherogenic chemokines, CSF-1 down-regulated CXCR4 mRNA and cell-surface protein expression (Figs. 2 and 5) . CXCR4 is the receptor for SDF/CXCL12, and interference with the SDF-1/CXCR4 axis aggravated atherosclerosis through enhanced leukocyte recruitment to plaques [57 ]. Thus, CSF-1 has pronounced proatherogenic effects on the macrophage chemokinome.
Atherosclerosis is a disease of hypercholesterolemia, in which lipid accumulation in vessel walls leads to monocyte infiltration and differentiation into lipid-laden foam cells. Proteoglycan-CSF-1 has been shown to bind LDL with high affinity [58 ], and it has been suggested that CSF-1 embedded in the ECM is involved in lipoprotein retention, leading to increased susceptibility to oxidation, increased aggregation, and increased lipid internalization by macrophages [59 ]. CSF-1 can also act more directly in this process; it enhanced the uptake and degradation of modified LDL and cholesterol esterification by human macrophages in vitro [60 ]. Our data further implicate CSF-1 in lipid homeostasis, suggesting this cytokine promotes cholesterol biosynthesis and/or its intracellular retention. HMDM accumulated free cholesterol in response to acute (6 h) CSF-1 stimulation (Fig. 4F) , contemporaneous with the induction of cholesterol biosynthetic genes and repression of the cholesterol exporter ABCG1 (Fig. 4 A-D , and Table 3 ).
As free cholesterol is a toxic metabolite [61 ], cellular cholesterol homeostasis is finely regulated by multiple negative-feedback mechanisms at the level of uptake, synthesis, esterification, and export [37 ]. The balance between free and esterified cholesterol is progressively unbalanced during atherogenesis, and the resulting excess of free cholesterol may heighten inflammatory responses and contribute to apoptosis of lesional cells and ultimately to plaque rupture [61 , 62 ]. Intriguingly, in addition to genes controlling cholesterol biosynthesis, CSF-1 induced HSPA5/GRP78, a molecular chaperone whose induction is a hallmark of the unfolded protein response (UPR) [63 , 64 ] in all three donors in our microarray study (Supplemental Table 1 ). In macrophages, the UPR is activated by free cholesterol accumulation [65 , 66 ] and during lesion development in atherosclerotic mice [65 ]. Thus, elevated levels of CSF-1 may contribute to disease progression by inducing free cholesterol accumulation in macrophages, which in turn, constitutes a source of endoplasmic reticulum stress.
The local environment may further magnify the importance of CSF-1 in atherosclerosis. The proatherogenic effects of CSF-1 are likely to be mediated by local CSF-1 (presumably the cell-surface or ECM-anchored isoform) [12 , 59 ]. Oxidized LDLs induced CSF-1 expression in aortic endothelial and smooth muscle cells [67 ], and we show here that human macrophages themselves express mRNAs for secreted glycoprotein, secreted proteoglycan, and cell-surface CSF-1 and up-regulate all three isoforms during maturation (Fig. 5) . Thus, multiple cell types present in atherosclerotic lesions express CSF-1, which can act on HMDM to elicit proatherogenic responses. Indeed, macrophages differentiated in the presence of CSF-1 had a similar immunophenotype to those present in atherosclerotic lesions [8 ].
CSF-1 regulation of CXCR4 expression is of interest for several reasons other than the relevance to atherogenesis. Although CXCR4 down-regulation upon monocyte-to-macrophage differentiation has been reported [68 ], our studies show that it is also acutely repressed by CSF-1 in mature macrophages and that this suppressive effect is reversible (Fig. 5) . Although SDF/CXCL12 signaling via CXCR4 has been classically associated with homeostatic leukocyte homing, it can also regulate the survival and proliferation of several cell types, including myeloid cells [69 ]. As SDF/CXCL12 mRNA is also expressed by HMDM (unpublished data), and HMDM respond to SDF [68 ], it is plausible that the SDF/CXCR4 axis acts as an autocrine loop for cell survival in HMDM, when CSF-1 levels are limiting. A further consequence of CSF-1-mediated CXCR4 down-regulation may be the mobilization of blood monocytes. Disruption of SDF-1/CXCR4 signaling in the bone marrow, by blocking the action of ligand or receptor, promotes hematopoietic stem-cell mobilization into the blood [70 ]. As in vivo administration of CSF-1 results in a dramatic increase in the pool of circulating blood monocytes [71 ], CSF-1 suppression of CXCR4 expression may contribute to monocyte mobilization.
In summary, we have identified novel CSF-1 target genes in mature human macrophages that provide insight into the function of this cytokine in health and disease. Our findings suggest that in humans, CSF-1 provides a molecular link between inflammation and cardiovascular disease by regulating pathways that stimulate cholesterol accumulation and a proatherogenic chemokine environment. CSF-1 antagonists may thus have future applications in the treatment of cardiovascular disease. Apart from the obvious relevance of our findings to human disease, this study illustrates several differences between BMM and HMDM, arguably the most commonly used ex vivo models of human and mouse macrophages. A growing body of evidence suggests that innate immune cells and their responses differ between humans and mice in numerous respects [72 , 73 ]. Whether some or all of the differences that we observe truly reflect species-specific effects, the heterogeneity of macrophages derived from different origins or both is unknown. Nonetheless, they highlight the importance of establishing the relevance of mouse models for the study of human inflammatory diseases.
Received August 25, 2008; revised September 25, 2008; accepted October 14, 2008.
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
) system studied using the colony stimulating factor-1-deficient op/op mouse Physiol. Rev. 76,927-947, tumor necrosis factor-
, IgG aggregates, and cAMP in mouse mesangial cells J. Immunol. 150,1971-1978[Abstract] J. Leukoc. Biol. 53,707-714[Abstract] inducible protein 10 kD (IP-10): a novel, pleiotropic member of the C-X-C chemokine superfamily Cytokine Growth Factor Rev. 8,207-219[CrossRef][Medline] inducible protein-10 expression in rheumatoid arthritis Arthritis Res. Ther. 5,R74-R81[CrossRef][Medline]-inducible chemokine IP-10/CXCL10 are augmented in both newly diagnosed Type I diabetes mellitus patients and subjects at risk of developing the disease Diabetologia 45,1107-1110[CrossRef][Medline], tumor necrosis factor receptor superfamily member 9, and macrophage colony-stimulating factor Arthritis Rheum. 58,2257-2267[CrossRef][Medline]
B Arterioscler. Thromb. Vasc. Biol. 15,1591-1598-stimulated signaling and human immunodeficiency virus-1 entry in human monocyte-derived macrophages Immunology 114,565-574[CrossRef][Medline]
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
