|Home | About | Journals | Submit | Contact Us | Français|
Recent evidence suggests that dendritic cells may play an important role in atherosclerosis. Based primarily on previous in vitro studies, we hypothesized that granulocyte macrophage colony-stimulating factor (GM-CSF)-deficient mice would have decreased dendritic cells in lesions.
To test this, we characterized gene targeted GM-CSF−/− mice crossed to hypercholesterolemic low-density lipoprotein receptor null mice. Our results provide conclusive evidence that GM-CSF is a major regulator of dendritic cell formation in vivo. Aortic lesion sections in GM-CSF−/− low-density lipoprotein receptor null animals showed a dramatic 60% decrease in the content of dendritic cells as judged by CD11c staining but no change in the overall content of monocyte-derived cells. The GM-CSF–deficient mice exhibited a significant 20% to 50% decrease in the size of aortic lesions, depending on the location of the lesions. Other prominent changes in GM-CSF−/− mice were decreased lesional T cell content, decreased autoantibodies to oxidized lipids, and striking disruptions of the elastin fibers adjacent to the lesion.
Given that GM-CSF is dramatically induced by oxidized lipids in endothelial cells, our data suggest that GM-CSF serves to regulate dendritic cell formation in lesions and that this, in turn, influences inflammation, plaque growth and possibly plaque stability.
A large body of evidence now implicates immune functions in the modulation of atherosclerosis.1–3 Advanced lesions contain both T and B lymphocytes as well as antigen-presenting dendritic cells, and experiments with mice deficient in various immune processes exhibit alterations in atherosclerotic lesion development. Recent studies have emphasized a prominent role of dendritic cells in atherosclerosis4–7 and in human lesions, dendritic cells accumulate in rupture-prone areas such as the plaque shoulder, appearing to co-localize with, as well as directly activate, T cells.5,6
In vitro studies have shown that GM-CSF stimulates the differentiation of dendritic cells from bone marrow precursors and monocytes,8 and an in vivo study in mice was also consistent with a role for GM-CSF in dendritic cell maturation.9 We hypothesized that the recruitment and accumulation of dendritic cells in atherosclerotic lesions is controlled in part by granulocyte macrophage colony-stimulating factor (GM-CSF). GM-CSF is induced in cultured aortic endothelial cells by oxidized low-density lipoproteins,10 a primary initiator of inflammation in atherosclerosis.11 Subsequent studies showed that GM-CSF is expressed in human atherosclerotic lesions.12
To test our hypothesis, we bred GM-CSF deficient mice to hypercholesterolemic low-density lipoprotein receptor-null mice (LDLR−/−), generating double knockout mice. The results revealed a dramatic impact on the content of dendritic cells in lesions, as the double knockout mice had only one-third as many dendritic cells as the LDLR−/− mice. Associated with the decreased lesional dendritic cell content were significant changes in the size and morphology of lesions as well as changes in immune responses to oxidized lipids.
GM-CSF null mice backcrossed nine generations to the C57BL/6J background were generously provided by Dr Bruce C. Trapnell (University of Cincinnati College of Medicine). These mice were then crossed to LDLR−/− mice on a C57BL/6J background to produce GM-CSF+/+, +/−, and −/− mice on an LDLR-null background. At approximately 10 weeks of age the mice were placed on an ad libitum Western type diet (TD 88137, Teklad) for 12 to 13 weeks before sacrifice. Plasma lipids were determined as described previously.12a Methods for the quantification of atherosclerotic lesions were as previously reported by and Mehrabian et al12b and Tangirala et al.12c Cryosections from the proximal aorta were stained for macrophages (rat anti-mouse MOMA-2, Beckman Coulter), smooth muscle cells (rabbit anti-mouse alpha smooth muscle actin, Spring Bioscience), dendritic cells (hamster anti-mouse CD11c, BD Biosciences), mature dendritic cells (rat anti-mouse CD86, BD Biosciences), GM-CSF (rat anti-mouse GM-CSF, US Biological), and T cells (rat anti-mouse CD4, BD Biosciences; rat anti-mouse CD8, BD Biosciences). Collagen content was determined using Masson trichrome staining and picrosirius red staining imaged by polarized light microscopy. By this technique, tightly packed collagen fibers appear yellow-red and less mature, whereas loosely packed collagen fibers appear green. Elastin fiber content was determined using elastica van Gieson staining. Blood leukocytes were incubated with an anti-mouse CD16/CD32 Fc gamma III/II receptor 2.462 blocking antibody (BD Pharmingen) before double staining with a PE labeled anti-mouse CD11c antibody (BD Pharmingen) and a FITC labeled anti-mouse I-A/I-E MHC II alloantigen antibody (BD Pharmingen). Plasma levels of EO6, autoantibodies to both copper and MDA oxidized LDL, and levels of IgG and IgM antibodies complexed with ApoB-100 containing lipoproteins were determined as previously described.12d–12f
To determine the effect of a GM-CSF deficiency on dendritic cell content of atherosclerotic lesions, GM-CSF−/− mice were intercrossed with LDLR−/− mice (both on a C57BL/6J background) to generate GM-CSF−/− LDLR−/−, GM-CSF+/− LDLR−/−, and GM-CSF+/+ LDLR−/− animals. The mice were fed a high-fat, high-cholesterol “Western” diet for 12 to 13 weeks and then euthanized. Aortic lesions were then examined in both the proximal aorta (by sectioning the aorta) and in the thoracic and abdominal region of the aorta (by en face staining for neutral lipids). Both male and female mice were studied, and plasma lipids, glucose, and insulin levels were determined.
Examination of the aortic root by immunostaining, using the GM-CSF−/− mouse as a negative control, clearly revealed the presence of GM-CSF (supplemental Figure I, available online at http://atvb.ahajournals.org). The GM-CSF was present throughout the intimal region, with no clear association with any cell type, as expected for a secreted protein.
GM-CSF is known to promote dendritic cell formation in vitro, and monocytes entering the subendothelial space can differentiate into either dendritic cells or resident macrophages. The factors regulating this transition are not well-characterized but are thought to be endothelial cell derived.13 We examined the possible effects on dendritic cell content by immunostaining lesions for the dendritic cell marker CD11c and the dendritic cell maturation marker CD86. A profound deficiency of dendritic cells in GM-CSF−/−LDLR−/− mice of both sexes was observed as compared with GM-CSF+/+LDLR−/− controls (Figures 1D and and2F,2F, and supplemental Figure IX). Dendritic cells comprised ≈16% of the lesion volume in GM-CSF+/+LDLR−/− mice but only 6% of the lesion volume in GM-CSF−/−LDLR−/− mice (P<0.0002) (Figure 2B). The subset of mature CD86+ dendritic cells represented approximately one-third of all lesional dendritic cells and decreased by ≈50% in lesions from GM-CSF−/− LDLR−/− mice (supplemental Figure IX). Overall, approximately half of all macrophages in lesions of GM-CSF+/+ LDLR−/− mice were dendritic cells, whereas only ≈12% of macrophages in lesions from GM-CSF−/− LDLR−/− mice on a 12-week western diet were such cells.
In light of decreased dendritic cells seen within the lesions of GM-CSF−/− LDLR−/− mice, we wanted to determine whether that was attributable to a systemic decrease in circulating levels of dendritic cells or MHC II+ cells. Using flow cytometry we measured the total number of CD11c+ and CD11c+ MHC II+ dendritic cells in the circulation. We could detect no differences in either the total number of circulating dendritic cells (represented by CD11c+ cells) or of mature dendritic cells (represented by CD11c+ MHC II+ cells). 5.73±0.52% of blood leukocytes in GM-CSF+/+LDLR−/− mice versus 4.43±0.32% of leukocytes from GM-CSF+/+ LDLR−/− mice stained positive for CD11c. Similarly, we could detect no differences in the total number of circulating MHC II+ cells using an antibody against an I-A MHC II alloantigen (Figure 1G and 1H), with 51.81±5.1% of leukocytes from GM-CSF+/+LDLR−/− mice and 47.83±2.9% of leukocytes from GM-CSF+/+LDLR−/− mice staining positive for that marker.
We sought to determine some of the biological and immunologic ramifications of a lesional dendritic cell deficiency by examining both the T cell content within the lesion and plasma levels of autoantibodies to oxidized lipids. CD4+ helper T cell content decreased ≈50% within lesions of GM-CSF−/−LDLR−/− mice (supplemental Figure IX). CD8+ cytotoxic T cells were found in very small numbers within lesions of GM-CSF+/+LDLR−/− mice and were virtually undetectable in lesions from GM-CSF−/−LDLR−/− mice (data not shown). This decrease in lesional T cell content, coupled with the decreased dendritic cell content, suggested that there might exist an overall decrease in the inflammatory state of the lesion in GM-CSF−/−LDLR−/− mice. However, we could not detect a difference in aortic mRNA levels of either the inflammatory cytokines IL-6 or MCP-1 (supplemental Figure VIII). We also hypothesized that the decreased number of dendritic cells in lesions of GM-CSF−/−LDLR−/− mice may result in less activation of B cells responsible for the production of autoantibodies to the immunogenic oxidized lipids present in lesions. Indeed, we observed ≈2.5-fold decrease in plasma levels of IgG autoantibodies recognizing malondialdehyde (MDA)-LDL and a 3-fold decrease in plasma levels of IgG autoantibodies recognizing copper oxidized LDL (supplemental Figure V). We also observed a 2-fold increase in the levels of IgM antibodies complexed with ApoB-100 containing lipoproteins (supplemental Figure V). We could detect no change in plasma levels of oxidized lipids bound to ApoB-100 containing lipoproteins as determined with an assay using an anti-EO6 antibody or in plasma levels of IgM autoantibodies recognizing MDA-LDL and copper-oxidized LDL (data not shown).
No significant differences in the content of smooth muscle cells (supplemental Figure II), macrophages (Figure 1A to 1C), or myeloid cells were observed in GM-CSF−/−LDLR−/− mice. We also performed semi-quantitative staining for collagen and elastin content in the lesions. In contrast to the previously reported findings from the normal aorta,14 we observed no significant differences in lesional collagen staining or aortic gene expression of α-1 (I), α-1 (VIII) procollagen between GM-CSF+/+LDLR−/−, and GM-CSF−/−LDLR−/− female mice (Figure 3A to 3F, and supplemental Figure VIII). Using picrosirus red staining, we observed that the less mature, green stained collagen was the primary form of collagen in the lesions of both groups of mice. We did, however, observe a striking decrease of elastin staining in the GM-CSF−/−LDLR−/− male and female mice. In aortic lesions of GM-CSF+/+LDLR−/− mice, elastin staining typically encompassed almost the entire circumference of the aortic sections (82%±3%) (Figure 3G to 3H). However, most sections of aortic lesions from GM-CSF−/−LDLR−/− mice included large areas of the media devoid of elastin staining (Figure 3I to 3J). In those sections, the percent of circumference staining for elastin varied from 16% to 79%, with an average of 53%±4% (Figure 3K).
Therefore, we examined the gene expression levels of several metalloproteinases (MMPs) able to degrade collagen and the expression of the elastin protein (tropoelastin). We also examined the expression of an inhibitor of MMPs, TIMP1. We reasoned that the decrease we observed in arterial elastin content could be attributed to increased expression of MMPs, decreased expression of tropoelastin, or decreased expression of an inhibitor of MMP activity. No differences were detected between GM-CSF+/+LDLR−/− and GM-CSF−/− LDLR−/− mice in the aortic expression of the elastases MMP3, MMP12, the gelatinase MMP9, tropoelastin, or TIMP1, as measured by quantitative polymerase chain reaction (supplemental Figure VII).
In both males and females there was ≈20% decrease in lesion development in the aortic root of GM-CSF−/−LDLR−/− as compared with GM-CSF+/+LDLR−/− mice (Figure 4A to 4B) (GM-CSF−/−LDLR−/− females, 242,955±13,869 μm2 versus GM-CSF−/−LDLR−/− females, 196 855±53 238 μm2; and GM-CSF+/+LDLR−/− males, 221 316±20 565 μm2 versus GM-CSF−/−LDLR−/− males, 173 134±13 345 μm2). Heterozygous mice exhibited an intermediate lesion size (GM-CSF+/−LDLR−/− females, 212 187±20 262 μm2 and GM-CSF+/−LDLR−/− males, 204 000±15 881 μm2) (Figure 4A, 4B). As compared with the aortic root, there was a larger effect of the GM-CSF deficiency on atherosclerotic lesions in the thoracic and abdominal aortae of female mice as measured by en face staining of lipid (Figure 4C). Interestingly, this ≈50% decrease in lesions was sex-biased, because males did not show significant evidence of an effect of GM-CSF deficiency on lesion development.
The fasting levels of plasma lipids (total cholesterol, high-density lipoprotein cholesterol, triglycerides, unesterified cholesterol, and free fatty acids) were determined in the mice before euthanization. As shown in supplemental Table I and supplemental Figure X, plasma lipid levels, relative particle and triglyceride distributions were very similar between GM-CSF+/+LDLR−/− and GM-CSF−/−LDLR−/− mice within each sex. After correction for multiple comparisons, no significant differences were observed among the various groups. Consistent with previous observations, there were clear differences in lipid levels between sexes, with females exhibiting somewhat lower levels of all lipid classes. The difference for triglyceride levels between sexes was particularly prominent (supplemental Table I).
Male LDLR−/− mice maintained on a western diet become insulin-resistant and diabetic, and we examined whether GM-CSF might influence these traits. Male mice tended to have higher plasma insulin levels than female mice, consistent with previous findings15 but the insulin levels were not significantly altered by the GM-CSF deficiency (data not shown). Interestingly, glucose levels appeared to be decreased in females and not males as a result of the GM-CSF deficiency. The decreased atherosclerotic lesions seen in GM-CSF−/−LDLR−/− females could result, in part, from reduced glucose levels.
Circulating plasma levels of M-CSF were measured by enzyme-linked immunosorbent assay to determine if M-CSF levels may have increased or decreased in response to the GM-CSF deficiency. No change in M-CSF levels was detected between GM-CSF+/+ LDLR−/− and GM-CSF−/−LDLR−/− animals (1655±104 pg/mL versus 1561±32 pg/mL).
We used GM-CSF−/−LDLR−/− mice to examine the regulation and function of dendritic cells in atherosclerotic lesions. GM-CSF is induced in endothelial cells by oxidized lipids10 and our present studies with GM-CSF−/−LDLR−/− mice have clearly shown that GM-CSF is abundantly present in atherosclerotic lesions. Consistent with previous in vitro studies implicating GM-CSF in dendritic cell differentiation,4,8 our results indicate a critical role for GM-CSF specifically in lesional dendritic cell accumulation, because the GM-CSF−/− LDLR−/− mice had only approximately one-third as many dendritic cells and approximately one-half as many mature dendritic cells in atherosclerotic lesions as wild-type mice but had no change in lesional macrophages or myeloid cells. We were able to support our hypothesis that GM-CSF regulates dendritic cell formation within a lesion by showing that the total number of circulating dendritic cells did not change in GM-CSF−/− LDLR−/− mice (Figure 1), suggesting that the decrease of dendritic cells within GM-CSF−/−LDLR−/− lesions was not attributable to decreased numbers of dendritic cells entering the lesion. Recent studies have shown that retention of dendritic cells within a lesion is triggered by their interaction with platelet activating factor or other inflammatory lipids.7 In the absence of inflammatory lipids, dendritic cells were able to emigrate from the lesion and migrate to lymph nodes, resulting in lesion size reduction.7 The normal role of dendritic cells is to sample the environment, display antigens and modulate T cell activation, usually after migrating back to the lymph nodes. Thus, the decline in lesional dendritic cell content in GM-CSF−/−LDLR−/− mice could result in decreased activation of T cells, and this could retard lesion growth. The decline in lesional dendritic cells coincided with decreased numbers of CD4+ T cells in lesions and with changes in levels of IgM antibodies complexed with ApoB100 containing lipoproteins, suggesting an overall decrease in the inflammatory state of the lesion as well as an enhanced ability of the immune system to eliminate oxidized lipids. It has been shown in humans that dendritic cells accumulate in the rupture-prone shoulders of lesions along with T cells, suggesting that they may contribute to plaque destabilization.6 Activated dendritic cells express the inflammatory modulator 5-lipoxygenase,16 which has been associated with both atherogenesis and aortic aneurysm formation.17
We observed that hyperlipidemic animals with a GM-CSF deficiency exhibited ≈20% to 50% decrease in the size of aortic lesions. The magnitude of the effect of GM-CSF on atherosclerosis is much smaller than that of M-CSF, a growth factor that is also induced by oxidized lipids in endothelial cells and whose function overlaps with that of GM-CSF. The mechanism underlying the very large decline in M-CSF−/− mice lesions is thought to relate to decreased monocyte/macrophage recruitment, survival or proliferation within the lesion.18 The presence of normal levels of M-CSF could be partially compensating for any macrophage survival and growth deficiencies caused by the absence of GM-CSF. Also, the lack of GM-CSF and presence of M-CSF could drive monocytes/macrophages toward a more phagocytic nature that would more efficiently uptake cholesterol in the artery wall, increasing foam cell formation. Diabetes is a risk factor for atherosclerosis in humans and diabetic individuals have increased levels of advanced end products of glycation that interact with receptors on endothelial cells to increase inflammation.19 To the extent that these effects of hyperlipidemia are mitigated by decreased glucose levels in GM-CSF−/− mice, they could be responsible for some component of the decreased atherosclerosis seen in these animals. With respect to the sex specific differences in lesion size, we have observed in this study that female animals exhibited less variable lesion data than male animals. Thus, it is possible we would be able to detect a significant decrease in aortic root lesions in male GM-CSF−/−LDLR−/− animals by increasing out sample size. However, with respect to differences we observed in lesion data measured in the ascending and descending aorta by en face, we hypothesize that the distribution or kinetics of lesion formation in this compartment may differ between male and female mice.
GM-CSF is expressed in smooth muscle cells of normal human arteries,12 and GM-CSF−/− mice have been reported to have alterations of the artery wall structure even in the absence of atherosclerosis.14 These alterations include decreased medial collagen content and decreased, disorganized collagen fibrils. In situ hybridization data showed that the GM-CSF deficiency resulted in decreased numbers of vascular cells expressing collagen and decreased levels of collagen expression, particularly type VIII collagen. We examined collagen and elastin gene expression and distribution in the artery wall and lesions of GM-CSF–deficient animals. Collagen staining was predominantly seen within the atherosclerotic lesions and aortic media using either Masson’s trichrome staining or picrosirus red staining. However, no differences in less mature, loosely packed collagens, tightly packed collagens, overall collagen content or aortic expression of α-1 (I) and α-1 (VIII) procollagen were observed. In our experience, elastic fiber staining intensity and quantity is highly variable within the aortas of LDLR−/− mice fed a high-fat diet, but the percent of the luminal circumference outlined by elastic fiber is relatively consistent. We used the measure of luminal circumference outlined by elastic fibers to detect changes in elastic fiber staining. Using that standard, GM-CSF−/−LDLR−/− mice exhibited a large decline in medial elastic fiber staining. This decline was much greater in magnitude than the decrease in atherosclerosis seen in these animals. Previous studies have indicated that GM-CSF treatment of macrophages induces MMP-120 (although previously expression of this collagenase could not be detected in the mouse aorta17) and MMP-12 (macrophage elastase).21 Thus, the absence of GM-CSF would be predicted to decrease MMP expression, not increase it. In any case, we did not detect a difference in the expression of a panel of matrix-degrading molecules consisting of collagenases, elastases, a tissue inhibitor of elastases or tropoelastin gene expression within the aortas of GM-CSF–deficient animals, indicating that other post-transcriptional factors related to GM-CSF are responsible for the abnormal elastin distribution. Based on the expression pattern of GM-CSF and its effect on the artery wall, it is plausible that arterial aneurysms would be more prevalent in GM-CSF−/− animals; however, we observed no increase in the number of aneurysms within the aortic root (data not shown).
Sugiyama et al22 showed that GM-CSF was expressed in areas of advanced human plaques that were prone to rupture and that GM-CSF was capable of inducing macrophage myeloperoxidase in vitro. Hypochlorous acid, a potent oxidizer produced by myeloperoxidase, can activate matrix metalloproteinase23,24 that degrade the extracellular matrix (ECM) and itself degrade collagen.25,26 However, in contrast to human lesions, myeloperoxidase is not expressed at high levels in murine atherosclerosis27 and is therefore not likely to play a role in the pathology that we have seen in GM-CSF−/− mice.
Exogenous infusion of GM-CSF and M-CSF has been shown to decrease plasma cholesterol. This effect is thought to be mediated by increased macrophage mediated cholesterol uptake.28 Consistent with this, increased cholesterol levels were indeed observed in M-CSF−/−LDLR−/− mice.18 However, GM-CSF−/−LDLR−/− exhibited no changes in total cholesterol, although female mice showed a small decline in high-density lipoprotein cholesterol. The discrepancy in total cholesterol levels between these two knockout mice could be indicative of an intrinsic change in the character of macrophages that result from a GM-CSF deficiency compared with an M-CSF deficiency. It could also be reflective of the fact that M-CSF null animals have a 50% decline in circulating monocytes, a decrease that may hinder cholesterol clearance, whereas GM-CSF−/− mice do not have a monocyte deficiency29,30
In conclusion, our results have revealed a novel regulatory mechanism controlling the composition and inflammatory nature of atherosclerosis. Thus, in response to the accumulation of oxidized lipids in the artery wall, GM-CSF is dramatically induced and regulates the recruitment or differentiation of dendritic cells in the lesion. These dendritic cells within the lesion appear to significantly impact the T cell content of lesions, resulting in a weakened B cell-mediated response against oxidized low-density lipoproteins. Dendritic cell content is likely to be important with respect to the activation of lymphocytes, the formation of foam cells and, perhaps, the stability of the lesion.
We acknowledge the generous help of Dr Joseph Witztum at the University of California, San Diego, in determining levels of autoantibodies to oxidized lipids.
Sources of Funding
This work was supported by National Institutes of Health Grant PO1 HL-30568, a Bristol-Meyers Squibb Freedom to Discover Award, and the Laubish Fund at the University of California, Los Angeles. Flow cytometry was performed in the UCLA Jonsson Comprehensive Cancer Center (JCCC) and Center for AIDS Research Flow Cytometry Core Facility that is supported by National Institutes of Health awards CA-16042 and AI-28697, and by the JCCC, the UCLA AIDS Institute, and the David Geffen School of Medicine at UCLA.