|Home | About | Journals | Submit | Contact Us | Français|
The mechanisms of atherosclerotic plaque rupture are poorly understood. Urokinase-type plasminogen activator (uPA) is expressed at elevated levels by macrophages in advanced human plaques. Patients with evidence of increased plasminogen activation have an elevated risk of major cardiovascular events. We used atherosclerotic mice to test the hypothesis that increased macrophage uPA expression in advanced plaques would cause histologic features similar to those in ruptured human plaques.
Bone marrow from transgenic mice with increased macrophage uPA expression or nontransgenic controls (all Apoe−/−) was transplanted into 35-wk-old Apoe−/− recipients, and innominate lesions and aortae examined 8 – 13 wk later. Donor macrophages accumulated in innominate lesions adjacent to plaque caps and in aortae, increasing uPA expression at both sites. Recipients of uPA-overexpressing macrophages had an increased prevalence of intraplaque hemorrhage (61 vs 13%; P = 0.002) as well as increased lesion fibrin staining and fibrous cap disruption (P = 0.06 for both). Transplantation of uPA-overexpressing macrophages increased aortic matrix metalloproteinase (MMP) activity (40%; P = 0.02). This increase was independent of MMP-9.
In advanced plaques of Apoe−/− mice, macrophage uPA overexpression causes intraplaque hemorrhage and fibrous cap disruption, features associated with human plaque rupture. uPA overexpression also increases vascular MMP activity. These data provide a mechanism that connects macrophage uPA expression, MMP activity, and plaque rupture features in mice. The data also suggest that elevated plaque plasminogen activator expression and plasminogen activation in humans may be causally linked to plaque rupture and cardiovascular events.
Atherosclerotic plaque rupture with intraplaque hemorrhage, intraluminal thrombosis, and vascular occlusion is the most common cause of myocardial infarction in humans.1 Nonocclusive plaque rupture is a likely cause of embolic stroke,2 and may stimulate progression of arterial stenoses.3 Despite much work, the molecular and cellular mechanisms that cause plaque rupture remain poorly understood. Acquisition of a mechanistic understanding of plaque rupture could lead to novel diagnostic approaches that detect unstable plaques before they rupture and to the development of drugs that prevent plaque rupture.
Several causes of plaque rupture have been proposed, including: apoptosis of endothelial cells covering a plaque, exposing thrombogenic subendothelium;4 apoptosis of smooth muscle cells (SMC) in the plaque fibrous cap5, 6 or heightened endoplasmic reticulum stress within plaque SMC,7 both of which would weaken cap structure; rupture of plaque microvasculature, precipitating plaque hemorrhage;8–10 rheologic factors that alter plaque composition or otherwise precipitate thrombosis;9, 11 and increased plaque proteolytic activity that removes extracellular matrix (ECM), weakens plaque structure, and promotes SMC apoptosis.6, 12 Among these candidates, attention has focused primarily on increased proteolytic activity, most commonly attributed to matrix metalloproteinases (MMPs).4, 13
A role for MMP-mediated proteolysis in plaque rupture is supported by correlative data demonstrating that various MMPs—which supposedly can digest key structural ECM components—are produced within vulnerable regions of plaques.4, 13, 14 MMPs are translated as inactive proenzymes and require an activating protease to gain proteolytic function. Some MMPs are activated within the secretion pathway by proprotein convertase; however, for most MMPs, including essentially all MMPs implicated in plaque rupture, the activation mechanism is unknown.15 Plasmin, an abundant serine proteinase that could be activated within the plaque by macrophage-expressed urokinase plasminogen activator (uPA),16, 17 is often considered to be a critical activator of MMPs in the vessel wall.18, 19
Despite wide acceptance of this model,4, 13, 20 it has been surprisingly difficult to obtain experimental data that support a causal role for either MMPs or uPA in plaque rupture. Studies in atherosclerosis-prone mice that overexpress or lack MMPs paint a complex and sometimes contradictory picture of the role of MMPs in lesion composition, with almost none of these studies showing an effect of MMPs on plaque rupture, defined as cap disruption with intraplaque hemorrhage.20 Two exceptions are: a study in which adenovirus expressing MMP-9 led to intraplaque hemorrhage in advanced carotid lesions of Apoe−/− mice21 and another study in which macrophage-targeted expression of a constitutively active MMP-9 mutant—but not wild-type MMP-9—led to intraplaque hemorrhage and cap disruption, also in Apoe−/− mice.22 Although these studies indicate that MMP-9 can cause histologic features of plaque rupture, it remains unknown how endogenous proMMP-9 (or other proMMPs) would be activated within plaques. Similar studies in mice that either lack uPA or overexpress uPA in plaque macrophages have not yet revealed a role for this serine proteinase in plaque rupture,18, 23, 24 with one study suggesting that uPA contributes to plaque “stability.”25
Here we tested the hypothesis that increased expression of uPA in advanced atherosclerotic lesions will cause histologic features of plaque rupture. To test this hypothesis we introduced uPA-overexpressing macrophages into aged (35-wk) Apoe−/− mice, documented accumulation of these macrophages in advanced innominate artery plaques, and performed histologic analyses of the plaques. We also tested whether increased expression of uPA in this model was associated with increased total MMP activity and, if so, whether MMP-9 activity was increased.
Apoe−/− mice with macrophage-targeted overexpression of uPA (SR-uPA+/0 mice) were generated in our laboratory.23 The SR-uPA transgene includes a fragment of the human scavenger receptor (SR)-A promoter and the murine uPA gene. SR-uPA+/0Apoe−/− mice were bred with non-transgenic Apoe−/− mice to yield SR-uPA+/0 or non-transgenic SR-uPA0/0Apoe−/− littermate controls. Some of the bone marrow donors used in the initial lineage-tracing study were also transgenic for green fluorescent protein (GFP). Because we modeled our study after that of Gough et al.,2 we used a chow diet (Catalog # 5053, Labdiet). All animal protocols were approved by the University of Washington Office of Animal Welfare. See the online-only Data Supplement for further details.
To test if macrophage uPA overexpression induces histologic features of plaque rupture in advanced atherosclerotic lesions, BM from either SR-uPA+/0 or non-transgenic male SR-uPA0/0Apoe−/− donors was transplanted into lethally irradiated 35-wk old non-transgenic female Apoe−/− mice. Mice were killed 8 – 13 wk after BM transplant (BMT). BMT into older mice was necessary because germline transgenic SR-uPA+/0Apoe−/− mice die at 15–30 wk of age,23 before they can develop advanced atherosclerotic lesions.26
See the online-only Data Supplement.
See the online-only Data Supplement
Total RNA was extracted from innominate arteries. Aortae were removed and incubated overnight in M199 (GIBCO). Plasminogen activator (PA) activity in aortic-conditioned medium was measured.27, 28 See the online-only Data Supplement for further details.
Innominate arteries were dissected free, placed in formalin for over 48 h, embedded in paraffin, and sectioned starting at the proximal end. Arteries examined for evidence of plaque rupture features were all harvested 10 wk post-transplant. In a subset of mice (8 per group) we excised aortic roots, embedded them in OCT, and sectioned them.23 See the online-only Data Supplement for further details.
Entry of BM donor cells into recipient innominate lesions was investigated by staining serial innominate artery sections with antibodies to GFP (Molecular Probes) and to the macrophage antigen Mac-3 (PharMingen). We analyzed innominate lesion histology using established methods.22 Stains in each set included: hematoxylin and eosin (H & E), picrosirius red to detect collagen, Verhoeff-Van Gieson to visualize medial elastin, Carstairs to detect fibrin and red blood cells, fibrin(ogen) immunostain (Dako), and Mac-2 immunostain (Cedarlane Labs) to detect macrophages. Intraplaque hemorrhage, fibrin deposition, medial elastin breaks, and fibrous cap disruption were identified by an observer blinded to genotype, using published criteria.22 These criteria included: fibrous cap disruption was present if at least 3 sections separated by 50 μm showed discontinuity or disruption of the fibrous cap accompanied by disruption of the underlying plaque structure. Intraplaque hemorrhage was defined as extraluminal red blood cells in at least 3 sections separated by 50 μm. Fibrin deposition was identified by the presence of both red-orange Carstairs stain and fibrin(ogen) immunostain in the same location in a minimum of 3 step sections per artery. Intraplaque hemorrhages were assessed by a second blinded observer, whose assessments were nearly identical to those of the first observer. Sections of aortic roots were stained with H & E and Mac-2, and intimal lesion area, intimal Mac-2-stained area, and percentage Mac-2-stained area measured.23 The presence of intraplaque hemorrhage was evaluated on H & E-stained slides. See the online-only Data Supplement for further details.
Experimental samples, including aortic-conditioned media and human neutrophil lysates (obtained from normal human volunteers according to an approved institutional protocol) were incubated with OmniMMP Fluorogenic Substrate (Biomol) at 37 °C in a black 96-well assay plate (Corning). Substrate cleavage was measured with a fluorometer (BioTek) using an excitation filter of 328 nm and emission filter of 393 nm.
Aortic explant culture media, aortic extracts,29 and neutrophil lysate were separated under non-reducing conditions by 10% SDS-PAGE in gels containing 0.5 mg/ml gelatin. Neutrophil lysate was used as a positive control for MMP-9 gelatinolytic activity.30 See the online-only Data Supplement for further details.
MMP-9 in aortic-conditioned media and neutrophil lysate was removed and identified by immunoprecipitation and western blotting using protein G Sepharose beads (Invitrogen) and an antibody to rat MMP-9 that also binds human and mouse MMP-931 (Millipore # AB19016; 1:100 dilution) or an antibody to GAPDH (Sigma). See the online-only Data Supplement for further details.
Data are mean ± SD or median (25–75% range). A priori hypotheses were tested with the unpaired t test or by Mann-Whitney rank-sum test when group data were non-normally distributed or if group variances were unequal. Survival rate curves were generated with the Kaplan-Meier method and compared with the log-ranktest. Fisher’s exact test was used to compare the prevalence of intraplaque hemorrhage, fibrin deposition, and fibrous cap disruption.
Flow cytometry of peripheral blood of recipients of GFP+/0 BM, performed 12–13 wk after BMT, showed successful reconstitution (83 ± 5% GFP+ cells, n = 7; data not shown). Immunohistochemical staining of innominate arteries of GFP+/0 BM recipients revealed abundant, foamy intimal cells that expressed both GFP and Mac-3 (Figure 1).
uPA mRNA was elevated in innominate arteries of recipients of SR-uPA+/0 BM [2.2 (2.1 – 2.3) vs 0.03 (0.02 – 0.08) arbitrary units in recipients of non-transgenic BM; P = 0.001; Figure 2A]. PA activity in medium conditioned by aortae of recipients of SR-uPA+/0 BM was similarly increased [1.5 (1.3 – 1.6) vs 0.04 (0.03 – 0.09) IU/mg protein in medium conditioned by aortae from recipients of non-transgenic BM; P = 0.001; Figure 2B].
Engraftment of SR-uPA+/0 BM did not affect the total number or percentage of peripheral blood monocytes (P ≥ 0.3 for both; Table 1). Recipients of SR-uPA+/0 BM had modestly higher plasma total cholesterol (38%) and slightly lower body weights (8%) than recipients of non-transgenic BM (P ≤ 0.01 for both; Table 1).
Recipients of SR-uPA+/0 BM began to die suddenly starting 6 wk after BMT. By 10 wk after BMT, 32% of the SR-uPA+/0 recipients had died suddenly compared to 0% of the recipients of non-transgenic BM (P = 0.001; Figure I in the online-only Data Supplement). The sudden death phenotype is also present in SR-uPA+/0 Apoe−/− mice and younger Apoe−/− recipients of SR-uPA+/0 BM.23, 32 The premature deaths of younger mice with SR-uPA+/0 BM appear to be due to occlusive coronary atherosclerosis without plaque rupture (no ruptured aortic root or coronary artery plaques were seen in younger mice);23, 32 however, the proximate cause of death in these mice and in the older mice in the present study (including possible arrhythmias, thromboembolism, or acute pump failure) remains unclear.
Innominate artery intimal, medial, lumen, and total vessel areas did not differ between recipients of SR-uPA+/0 and recipients of non-transgenic BM (P ≥ 0.3 for all except lumen area, which was slightly (15%) smaller in SR-uPA+/0 mice with borderline significance; P = 0.08; Table I in the online-only Data Supplement). There were also no differences between the groups in lesion collagen area and macrophage content (either total or %; Table 2 and Figures IB–C in the online-only Data Supplement), or in the number of elastin breaks in the underlying media (Table I, Figures IIF and III in the online-only Data Supplement).
Innominate intraplaque hemorrhage was significantly more prevalent in recipients of SR-uPA+/0 BM [11 of 18 (61%) in SR-uPA+/0 recipients vs 3 of 23 (13%) in recipients of nontransgenic BM; P = 0.002; Figure 3A–C, F and Table 2]. Because neither we nor several other groups that have reported murine plaque hemorrhage have observed microvessels in innominate plaques of Apoe−/− mice,33 we attribute plaque hemorrhage to entry of luminal blood through a disrupted plaque cap. Fibrin accumulation (Figure IID–E in the online-only Data Supplement) and fibrous cap disruption (Figure 3D–E) were also increased in recipients of SR-uPA+/0 BM, although statistical significance was borderline (P = 0.06 for both; Table 2). Macrophages were always present at sites of fibrous cap disruption (Figure 3D–E).
Intimas of aortic root lesions were not significantly larger in recipients of SR-uPA+/0 BM (5.9 ± 0.38 vs 4.9 ± 0.54 μm2× 105; P = 0.2). However, aortic root intimas of recipients of SR-uPA+/0 BM had a significantly higher macrophage content (0.62 ± 0.056 vs 0.37 ± 0.060 μm2× 105; P = 0.01), and a higher percentage of Mac-2-stained area (11 ± 0.85 vs 7.4 ± 0.75%; P = 0.01) than were present in recipients of nontransgenic BM. Intraplaque hemorrhage was present in 2 of 8 aortic root lesions in recipients of SR-uPA+/0 BM and 1 of 8 lesions in recipients of nontransgenic BM (P = 1.0). These hemorrhages were small, with far fewer extravasated red cells than in the innominates.
Because innominate arteries are too small to generate sufficient conditioned media for the MMP activity assay, we used aortic-conditioned media to investigate whether transplantation of SR-uPA+/0 macrophages increased vascular MMP activity. Total MMP activity was 40% higher in medium conditioned by aortae of SR-uPA+/0 BM recipients (18 ± 4.0 vs 13 ± 2.9 ΔRFU/min/μg protein for aortic-conditioned medium from SR-uPA0/0 BM recipients; P = 0.02; Figure 4). Control experiments showed that neither uPA, uPA with added plasminogen, nor plasmin cleaved the MMP substrate (data not shown).
Gelatin zymography of aortic-conditioned media and aortic extracts of BMT recipients revealed variable levels of active MMP-9 with no consistent difference between the groups (Figure 5A and data not shown). To assess secreted MMP-9 activity directly, we immunoprecipitated the enzyme from aortic-conditioned media and quantified the activity recovered. Immunoprecipitation of MMP-9 removed all MMP activity from MMP-9-rich neutrophil lysate (Figure IV in the online-only Data Supplement) and also removed substantial immunoreactive MMP-9 from aortic-conditioned media (Figure 5B). However, immunoprecipitation of MMP-9 from aortic-conditioned media did not alter total MMP activity (Figure 5C).
We used a BMT model and older Apoe−/− mice to investigate whether macrophage uPA overexpression in advanced atherosclerotic lesions induces histologic features of plaque rupture. Our major findings were: (1) Donor BM-derived macrophages entered mature innominate lesions and became foam cells; (2) Recipients of SR-uPA+/0 BM expressed significantly higher levels of uPA in innominate artery and aorta; (3) Recipients of SR-uPA+/0 BM had a higher prevalence of intraplaque hemorrhage, plaque fibrin accumulation, and fibrous cap disruption; (4) MMP activity was elevated in aortae of recipients of SR-uPA+/0 BM; and (5) MMP-9 does not contribute to the increase in aortic MMP activity. These data support a causal role for macrophage uPA overexpression in atherosclerotic plaque rupture and suggest that macrophage uPA overexpression induces plaque rupture by activation of MMP(s) other than MMP-9.
One of the key assumptions of this study is that the hyperlipidemic mouse innominate artery is an informative experimental setting for elucidating the molecular and cellular pathophysiology of human plaque rupture. The mouse innominate model was initially widely embraced because—unlike other animal models of atherosclerosis—it exhibits fibrous cap disruption and intraplaque hemorrhage, two key histologic features of human plaque rupture.26, 34 Although this model remains widely used, 22, 35, 36 it has also been criticized because it does not replicate other well-described histologic features of human plaque rupture including occlusive thrombi, eroded endothelium, and luminal eruption of calcific nodules.33 Despite these limitations, akin to those of all murine models of human atherosclerosis,37 the mouse innominate is currently the best available animal model for unraveling the molecular and cellular events that lead to human plaque rupture.
Our finding that BMT of uPA-overexpressing macrophages and their accumulation in plaque shoulder areas led to fibrous cap disruption and intraplaque hemorrhage (Figure 3, Table 2) provides a mechanistic connection between histologic data examining plasminogen activation in human plaques and data from clinical studies. uPA is highly expressed in macrophage-rich areas adjacent to the shoulder regions of advanced human atherosclerotic lesions.16, 17, 38 “Culprit” human carotid plaques have increased expression of plasminogen activators and elevated plasmin activity compared to uncomplicated plaques,39 and both elevated plasminogen activator mRNA and increased uPA receptor expression by macrophages correlate with intraplaque hemorrhage and rupture in acutely symptomatic human carotid plaques.40, 41 In the clinical realm, low plasma HDL is associated with elevated plasminogen activation by human peripheral blood monocytes (likely due to upregulation of monocyte uPA receptors),42 a small study revealed elevated plasma uPA in patients with plaque rupture visualized by intravascular ultrasound,43 and 3 large clinical trials associated increased plasminogen activation (measured as plasma plasmin-antiplasmin complexes) with risk of major cardiovascular events.44–46 The present study, by linking plaque plasminogen activation and histologic features of plaque rupture, connects these histological and clinical observations.
Fitting the results of the current study with data from other studies in which uPA expression was manipulated in atherosclerotic mice is more of a challenge. Nevertheless, in all cases apparent discrepancies can be explained. In an early study of atherosclerosis in uPA knockout mice, no effects of uPA deficiency on cap disruption and intraplaque hemorrhage were reported;18 however, only aortic root lesions were examined and neither cap disruption nor intraplaque hemorrhage are typically found in the mouse aortic root. In a more recent study of older, atherosclerotic uPA-null mice, innominate lesions were examined, but neither cap disruption nor plaque hemorrhage were among the end points reported, and might have been overlooked.25 In our own work, both SR-uPA+/0 transgenic mice and younger recipients of SR-uPA+/0 BM had accelerated aortic atherosclerosis, whereas in the present study we found no effect of uPA overexpression on innominate or aortic root plaque volume (Table I in the online-only Data Supplement). This apparent discrepancy may indicate a stage-specific role for uPA in lesion growth, with little effect of uPA on growth of advanced lesions in these 35–45 wk-old mice.26
Our study provides biochemical data that support a step-wise connection between increased plasminogen activation, MMP activation, and plaque rupture. These connections have long been hypothesized,19, 47 but heretofore lacked experimental support. A recent report established an association between plasmin, MMP-9, and experimental murine aortic aneurysms.31 However, despite numerous experiments in atherosclerotic MMP-knockout and MMP-overexpressing mice,48 only scant data support the long-hypothesized role of MMPs in plaque rupture. The study that most convincingly links MMP activity and murine plaque rupture relied on intraplaque expression of a constitutively active MMP-9 mutant, leaving open the question of why, in the same study, overexpression of wild-type, activatable MMP-9 in rupture-prone innominate plaques did not cause plaque rupture.22 Here we provide a possible answer to this conundrum by showing that overexpression of wild-type, activatable uPA increases MMP activity significantly and precipitates plaque rupture via an MMP-9-independent pathway. Thus, overexpression of active MMP-9 might cause plaque rupture experimentally, but the native pathway through which plaques rupture may be via uPA, plasmin, and a different MMP. Future work will be aimed at identifying the MMP(s) activated by uPA in this model and connecting the activity of these MMP(s) with cap disruption and intraplaque hemorrhage. Finally, more widespread use of the highly quantitative total MMP activity assay (Figure 4) would facilitate comparisons among studies of vascular MMP overexpression, potentially clarifying why plaque rupture features are found in some studies but not in others.20
Our study has limitations. First, the 30-fold overexpression of uPA in mouse innominates and aortae (Figure 2) is higher than the level of uPA overexpression we achieved in aortae of younger mice with macrophage-targeted uPA overexpression23, 32 and could be viewed as nonphysiologic. It would be interesting and valuable to test whether lower levels of uPA overexpression in mature plaques, over longer periods of time, also caused plaque rupture, but this is not possible with the current model. Second, in contrast to our previous studies in which the SR-uPA transgene did not affect plasma cholesterol,23, 24, 32 we also observed a modest (40%) increase in plasma cholesterol in recipients of SR-uPA+/0 BM. However, plasma cholesterols of mice with histologic features of plaque rupture did not differ significantly from cholesterols of mice without plaque rupture (data not shown), and the increased plasma cholesterol in SR-uPA+/0 BM recipients did not affect innominate intimal volume (Table I in the online-only Data Supplement). Both of these observations argue that differences in plasma cholesterol between the groups was not responsible for the increased intraplaque hemorrhage, fibrin deposition, and elevated MMP activity in recipients of SR-uPA+/0 BM. Third, the features of plaque rupture we observed were not associated with luminal thrombosis and vascular occlusion. As mentioned above, absence of thrombotic occlusion is a well-known limitation of the mouse innominate model.33
In summary, our data support a causal connection between elevated lesion uPA expression, increased vascular MMP activity, disruption of plaque structural integrity, and intraplaque hemorrhage. The clinical relevance of these data—generated in an animal model—is supported by human pathologic and clinical studies that associate increased plasminogen activation with unstable plaques and major cardiovascular events. Future studies in this model might identify key substrates (such as matrix molecules) whose cleavage leads to plaque rupture. We also speculate that clinical strategies aimed at limiting vascular wall plasminogen activation might inhibit plaque rupture, thereby preventing both myocardial infarctions and strokes.
Rupture of previously stable atherosclerotic plaques causes unstable angina, myocardial infarctions, and strokes. However, the mechanisms that cause plaque rupture are not yet understood. If these mechanisms could be identified, therapies might be developed that prevent plaque rupture. The most widely held mechanistic hypothesis regarding plaque rupture is that it is caused by increased activity of plaque proteolytic enzymes including urokinase, plasmin, and matrix metalloproteinases. This hypothesis is supported by histologic analyses of human plaques that show active proteases in ruptured plaques and by clinical studies that associate higher levels of proteolytic activity (in plasma) with major cardiovascular events. However, these studies do not prove causality and a large number of animal studies have not revealed a reliable cause-and-effect relationship between plaque proteolytic activity and histologic features of plaque rupture. Here we generated a new animal model in which we introduce macrophages that express high levels of urokinase into advanced atherosclerotic plaques in mice. Arteries from these mice have elevated proteolytic activity, including high levels of both plasminogen activator and matrix metalloproteinase activity. Compared to control mice, mice with uPA-overexpressing macrophages have a significantly higher (61 vs 13%) prevalence of intraplaque hemorrhage as well as a higher prevalence of plaque cap disruption and fibrin(ogen) staining (features found in ruptured human plaques). These data provide experimental support for a mechanistic connection between macrophage uPA expression, matrix metalloproteinase activity, and plaque rupture. These data also support a causal link between elevated plaque plasminogen activator expression, plaque rupture, and cardiovascular events in humans.
We thank Dr. April Stempien-Otero, Dr. Casilde Sesti, Dr. Stephen Farris, Isaac Emery, Roderick Browne, and Jerry Ricks for technical assistance. Margo Weiss helped with manuscript preparation. Elaine Raines provided insightful suggestions and advice, and we thank Dr. Michael Rosenfeld for helpful discussions.
This work was supported by a grant from the National Heart, Lung, and Blood (HL080597) to D.A.D. M.J. was supported in part by a grant to the University of Washington from the Howard Hughes Medical Institute through the Undergraduate Science Education Program.