Inhibitory Role of Plasminogen Activator Inhibitor-1 in Arterial Wound Healing and Neointima Formation
A Gene Targeting and Gene Transfer Study in Mice
Background Plasminogen-deficient mice display impaired vascular wound healing and reduced arterial neointima formation after arterial injury, suggesting that inhibition of plasmin generation might reduce arterial neointima formation. Therefore, we studied the consequences of plasminogen activator inhibitor-1 (PAI-1) gene inactivation and adenoviral PAI-1 gene transfer on arterial neointima formation.
Methods and Results Neointima formation was evaluated in PAI-1–deficient (PAI-1−/−) mice with perivascular electric or transluminal mechanical injury. PAI-1 deficiency improved vascular wound healing in both models: the cross-sectional neointimal area was 0.001±0.001 mm2 in PAI-1+/+ and 0.016±0.008 mm2 in PAI-1−/− mice within 1 week after electric injury (P<.02) and 0.055±0.008 mm2 in PAI-1+/+ and 0.126±0.006 mm2 in PAI-1−/− mice within 3 weeks after mechanical injury (P<.001). Proliferation of smooth muscle cells was not affected by PAI-1 deficiency. Topographic analysis of arterial wound healing after electric injury revealed that PAI-1−/− smooth muscle cells, originating from the uninjured borders, more rapidly migrated into the necrotic center of the arterial wound than wild-type smooth muscle cells. On the basis of immunostaining, PAI-1 expression was markedly upregulated during vascular wound healing. There were no genotypic differences in reendothelialization of the vascular wound. When PAI-1−/− mice were intravenously injected with replication- defective adenovirus expressing human PAI-1 (AdCMVPAI-1), plasma PAI-1 antigen levels increased in a dose-dependent fashion up to to 61±8 μg/mL with 2×109 plaque-forming units (pfu) virus. Luminal stenosis was 35±13% in control AdRR5-treated (2×109 pfu) and suppressed to 5±5% in AdCMVPAI-1–treated (6×108 pfu) PAI-1−/− mice (P<.002).
Conclusions By affecting cellular migration, PAI-1 plays an inhibitory role in vascular wound healing and arterial neointima formation after injury, and adenoviral PAI-1 gene transfer reduces arterial neointima formation in mice.
The plasminogen system has been implicated in several cardiovascular disorders, including wound healing after vascular reconstructions.1–12 The inactive proenzyme plasminogen is converted to its active derivative plasmin by two physiological PAs, t-PA and u-PA.13 Their action is controlled by irreversible PAIs, of which PAI-1 appears to be the predominant inhibitor.5 Plasmin has been implicated in wound healing via proteolysis of extracellular matrix components or activation of growth factors.13–15
Luminal stenosis, resulting from arterial neointima formation, limits the success of vascular reconstructions for treatment of atherothrombosis.6 Although vascular remodeling appears to be a major determinant of luminal stenosis after balloon angioplasty, intimal thickening can also contribute to the luminal narrowing, for example after placement of intraluminal stents. Intimal thickening results from accumulation of smooth muscle cells that proliferate in the media, migrate across the elastic membrane and subsequently contribute to intimal thickening by increased proliferation and matrix deposition.6,16 In a recent study, we have demonstrated that vascular wound healing and arterial neointima formation are significantly impaired in mice with a disruption of the plasminogen gene.17 In addition, u-PA was found to be the predominant activator mediating migration of smooth muscle cells.18 The role of PAI-1 in these studies was, however, not defined.
The present study in mice in which the gene encoding PAI-1 has been inactivated (PAI-1−/−)19 provides direct evidence, using two models of arterial injury, for a significant role of PAI-1 in arterial neointima formation via regulation of smooth muscle cell migration. In addition, we demonstrate that overexpression of PAI-1 in the plasma of mutant mice by intravenous injection of replication-defective adenovirus expressing human PAI-1 is able to suppress arterial neointima formation. These data might have relevance for the design of strategies to treat restenosis in patients but also suggest caution in the therapeutic use of PAI-1 inhibitors to improve thrombolysis and reduce atherosclerosis.
Arterial Injury Model
Six- to 8-week-old PAI-1+/+ and PAI-1−/− mice19 of either sex with a 75% C57Bl6/25% 129 genetic background were used. Groups of 8 to 14 animals were studied. Perivascular electric injury of left femoral arteries and analysis of neointima were performed as described elsewhere.20 Although neointima formation occurs in both femoral and carotid arteries, the response in the femoral arteries was analyzed because it occurs somewhat more rapidly and vigorously (P. Carmeliet, L. Moons, D. Collen, unpublished observations). Intraluminal mechanical injury of the carotid arteries was performed according to a recently described method by Lindner et al,21 with minor modifications. Briefly, the left internal carotid artery was exposed by blunt-end dissection, tied off distally, and looped proximally with 7–0 Mersilene nylon suture for vascular control during the procedure. A transverse arteriotomy was made in the proximal portion of the internal carotid artery, and a 380-μm flexible guide wire (C-SF 15–15, Cook Belgium NV) was passed to the aortic arch over ∼15 mm and withdrawn three times with a rotating motion. After removal of the guide wire, the proximal portion of the internal carotid artery was tied off, and the skin incision was closed as described above. This protocol allowed reproducible injury, resulting in rupture of the internal elastic lamina and neointimal accumulation of α-smooth muscle actin–positive cells within 3 weeks after injury. The carotid artery was used because of a technical limitation with insertion of the guide wire in the femoral artery.
Tissue Harvest, Histology, and Immunocytochemistry
Tissue harvest, fixation, embedding, and sectioning were performed as described previously.17,20 Smooth muscle and inflammatory cells were immunostained for smooth muscle α-actin or CD45 as described previously.20 Macrophages were visualized with the use of a rat monoclonal antibody against Mac-3 (Pharmingen). Replication was determined by labeling cells with 5′-bromo-2′-deoxyuridine and immunostaining as described previously.20 Vitronectin was stained with a biotinylated cross-reacting anti-human mouse monoclonal antibody.22
PAI-1 staining was performed with a polyclonal rabbit anti-murine PAI-1 or a rabbit anti-human PAI-1 antiserum (with similar results).19 Colocalization of PAI-1 with smooth muscle cells or macrophages was established through the use of a double immunofluorescence labeling approach as described previously.23 With this procedure, PAI-1–positive cells appear green, α-actin–or Mac-3–positive cells appear red, and cells containing PAI-1/α-actin or PAI-1/Mac-3 appear yellow.
Morphometric Analysis and Reendothelialization
Morphometric measurements of cross-sectional areas, cell counts and proliferation rates, and topographic analysis of neointima formation in electrically injured arteries were performed as described previously.17,20 Morphometric analysis of mechanically injured arteries was performed on five sections, equally spaced across the injured segment. Reendothelialization was evaluated after staining of the denuded blood vessels with Evans blue as described previously.20
Zymographic analysis of PA activities in arterial extracts was performed as described previously.17 In situ zymography on 7-μm cryosections of arteries was performed by fibrin overlay using a gel prepared by clotting a mixture of human fibrinogen (final concentration, 4 mg/mL), plasminogen (final concentration, 10 μg/mL), and agarose (final concentration, 0.5%) with thrombin (final concentration, 0.3 NIH U/mL). The fraction of the lysis due to t-PA or u-PA activity was determined by including in the fibrin gel 20 μg/mL polyclonal neutralizing rabbit anti-murine t-PA–and/or u-PA–specific IgG, respectively.24 Purified murine t-PA and u-PA were used as standards.24 The amount of lysis of the gels (area×intensity) was quantitatively analyzed using the Quantimed 600 image analysis software and expressed in arbitrary units.
Adenovirus-Mediated PAI-1 Gene Transfer
The adenovirus precursor pACCMVPAI-1 was generated by ligating the 1.4-kb EcoRI/BglII fragment of pPAI-1RBR25 containing the entire coding sequence of human PAI-1 into EcoRI/BamHI–digested pACCMVpLpA. In this plasmid, the PAI-1 cDNA is positioned between the human CMV immediate-early enhancer/promoter and the SV40 t-antigen intron/polyadenylation signal. AdCMVPAI-1, control AdRR5 adenovirus (without transgene),26 and AdCMVLacZ virus (expressing the Escherichia coli β-galactosidase gene) were generated according to standard procedures and purified to stocks of >1010 pfu/ml as described previously.27
Primary human umbilical vein endothelial cells were cultured in serum and growth factors as described previously,25 infected for 1 hour with recombinant adenovirus stocks at a multiplicity of infection of 20 pfu/cell (2.5×106/6-cm plate), and subsequently cultured in culture medium for 48 hours. Cells were refed with 2 mL of Dulbecco’s modified Eagle’s medium/Medium 199 (1:1) supplemented with 0.5% bovine serum albumin, and the conditioned medium (4 hours at 37°C) was assayed for PAI-1 activity.28
Three days after electric injury, anesthetized mice were intravenously injected with 200 μL of the indicated dose of control (AdRR5) or recombinant adenovirus (AdCMVPAI-1, AdCMVLacZ) using a 2F catheter (Portex green, Baxter). On the seventh day (eg, 4 days after virus injection), 100 μL blood was sampled from the retroorbital plexus from virus-transduced mice in a recipient containing 0.1 vol of EDTA (100 μL of 10 mg/mL stock solution). PAI-1 antigen and activity were measured as described previously.29 On day 14, the mice were killed, and the injured arteries were analyzed as described above. In separate experiments, PAI-1−/− mice were injected with AdRR5, AdCMVLacZ, or AdCMVPAI-1 (1 to 2×109 pfu), and the liver and the injured left and uninjured right femoral arteries were harvested 4 days after infection for immunohistochemical staining of PAI-1 antigen or LacZ staining as described previously.27
Experimental values are expressed as mean±SEM. Statistical significant differences between groups were calculated by ANOVA followed by Bonferroni’s correction or Wilcoxon’s signed rank analysis, as indicated in the text.
Vascular Wound Healing in Wild-Type Mice
Electric Injury Model
Because the electric injury model differs from the mechanical injury model,21 we briefly discuss the characteristics. Fig 1⇓ schematically represents the healing response to electric injury, which denudes the endothelium and destroys all cells in the media (<1% surviving medial cells).20 This resulted in a wound-healing response characterized by transient thrombosis, removal of necrotic debris, reendothelialization, repopulation of the media by smooth muscle cells, and the formation of a neointima with only a few layers within 1 week (Fig 2b⇓) and multiple layers of smooth muscle cells within 3 weeks after injury (Fig 2d⇓ and 2e⇓).20 A neoadventitia formed that was rich in leukocytes and fibroblast-like cells.
Mechanical Injury Model
Mechanical injury resulted in less severe medial cell necrosis (52±8% surviving medial cells, P<.05 versus uninjured arteries) than electric injury. Because residual viable smooth muscle cells persisted across the entire injured segment, neointima formation occurred more uniformly across the injured segment, as described previously for similar mechanical injury models.21 Fig 2g⇑ and 2h⇑ displays a representative neointimal lesion containing smooth muscle cells in a wild-type carotid artery within 3 weeks after mechanical injury. Because the electric injury model permits the differentiation between proliferation and migration of smooth muscle cells and is technically easier than the mechanical injury model,20 it was used more extensively throughout this study.
Vascular Wound Healing in PAI-1−/− Mice
Electric Injury Model
After electric injury, the removal of the necrotic debris, repopulation of the media by smooth muscle cells, and accumulation of neointimal cells were accelerated in PAI-1−/− mice during the first 2 weeks after electric injury. Compared with PAI-1+/+ arteries, which revealed only a small neointima at the uninjured borders within 1 week after injury (Fig 2b⇑), a larger neointima containing more cells was already present in PAI-1−/− arteries at the uninjured borders (Fig 2c⇑) and progressed further into the necrotic center (see below). In addition, the adventitia appeared larger and hypercellular (not shown). Quantitative measurements of the cross-sectional neointimal area within the boundary of the internal elastic lamina and the lumen (Fig 3a⇓), intima-to-media ratio (Fig 3b⇓), percent luminal stenosis (Fig 3c⇓), and number of neointimal cells (Fig 3d⇓) revealed that neointima formation was significantly accelerated in PAI-1−/− mice compared with PAI-1+/+ mice within the first weeks after electric injury. There were no significant genotype-specific differences in the neointimal area or cell counts beyond 3 weeks after injury (Fig 3⇓), indicating that PAI-1 deficiency accelerated but did not increase the extent of neointima formation in the electric injury model.
Mechanical Injury Model
PAI-1 deficiency also improved neointima formation after mechanical injury. On histological analysis, the neointima appeared much larger and contained more cells (Fig 2i⇑). Morphometric analysis revealed that within 3 weeks after injury, the neointimal cross-sectional area (Fig 4a⇓), intima-to-media ratio (Fig 4b⇓), percent luminal stenosis (Fig 4c⇓), and number of neointimal cells (Fig 4d⇓) were higher in PAI-1−/− than PAI-1+/+ arteries. The mechanisms underlying the improved neointima formation in PAI-1−/− mice were further investigated using the electric injury model.
Proliferation of Smooth Muscle Cells
To evaluate whether deficiency of PAI-1 affects cellular proliferation, incorporation of 5′-bromo-2′-deoxyuridine into replicating cells was determined (Table 1⇓). Proliferation of cells in electrically injured arteries 2 weeks after injury is a better index of smooth muscle cell proliferation because the media and neointima are largely devoid of leukocytes beyond 1 week after injury.20 As shown in Table 1⇓, medial and neointimal cell proliferation did not different significantly between PAI-1+/+ and PAI-1−/− arteries.
Migration of Smooth Muscle Cells
After electric injury, smooth muscle cells migrated within the media and alongside the lumen from the uninjured borders into the necrotic center, resulting in the formation of a neointima, first at the uninjured borders and subsequently with progression into the center of the wound.20 Progression of the neointima from the borders to the center was quantified through measurement of the luminal narrowing (percent stenosis) at equally spaced positions across the injured segment. Within 1 week after injury, significant neointima formation was initiated at the borders of the injury in PAI-1−/− arteries, but formation was only minimal in PAI-1+/+ arteries (P<.05 versus PAI-1+/+ by ANOVA) (Fig 5a⇓). Within 3 weeks after injury, the neointima had progressed into the center of the injury in both PAI-1−/− and PAI-1+/+ arteries (P=NS) (Fig 5b⇓).
Medial cell repopulation was evaluated by counting the medial cell nuclei at the left border, center, and right border of the injured segment (locations 1, 5, and 10 in the arteries of Fig 1⇑), revealing that the cells migrated further and were found at earlier times after injury in the center of the wound in PAI-1−/− than in PAI-1+/+ arteries (Table 2⇓). This process is schematically represented in Fig 1⇑.
Electric injury completely denuded the injured segment of intact endothelium as revealed by Evans blue staining immediately after injury.20 Reendothelialization was initiated from the uninjured borders, progressed into the necrotic center, and almost complete within 1 week after injury in PAI-1+/+ mice (n=7 mice) (Fig 6⇓). Endothelial regrowth was not affected, either in extent or in rate, by PAI-1 deficiency (n=5 mice) (P=NS).
Because deficiency of PAI-1 protects against thrombosis and facilitates lysis of 125I-fibrin-labeled pulmonary plasma clots,19 the incidence and extent of thrombosis were semiquantitatively evaluated. At 1 week after injury, 5 of 34 sections (15%) from wild-type arteries (n=7) contained residual mural thrombus that stenosed the lumen for 5% to 25%. In contrast, only 2 of 40 PAI-1−/− arteries (n=8; 5%) contained a mural thrombus (P=.08 versus wild-type by Wilcoxon’s signed rank analysis), indicating a trend toward increased clot lysis in PAI-1−/− mice.
Expression of PAI-1 During Arterial Wound Healing
PAI-1 staining was evaluated in uninjured and electrically injured arteries (n=3) within 1 and 2 weeks after injury, which is the time frame of active cellular migration. Fig 7a⇓ summarizes the semiquantitative analysis and indicates the topographic locations across the injured arteries.
In most uninjured arteries, PAI-1 immunoreactivity was undetectable (Fig 7b⇑); only occasionally could a low level of PAI-1 immunoreactivity be detected in medial smooth muscle cells of quiescent arteries.
At 1 week after injury, expression of PAI-1 was induced in cells that accumulated in the media, neointima, and, to a lesser degree, adventitia. In sections adjacent to the injury in which smooth muscle cells start to proliferate and migrate into the wound (location I in Fig 7a⇑), PAI-1 immunoreactivity was present in smooth muscle cells in the media (Fig 7c⇑). At the site of maximal neointima formation (location II), PAI-1 staining was present in most of the cells in the intima and the media (Fig 7d⇑). Regenerating endothelial cells expressed low levels of PAI-1. PAI-1 staining was also detectable, albeit at a lower level, in fibroblast-like cells in the enlarged adventitia closest to the external elastic lamina (Fig 7d⇑) and in areas of leukocyte infiltration (cfr infra). A similar PAI-1–staining pattern was observed at the leading front of migrating cells (location III) (not shown). Double immunofluorescence staining revealed colocalization (yellow) of PAI-1 (green) within α-actin smooth muscle cells (red) in the media and neointima (Fig 7e⇑) and within Mac-3–positive macrophages (red), predominantly located in the adventitia (Fig 7f⇑). There was no PAI-1 immunoreactivity detected in injured PAI-1−/− arteries (Fig 7g⇑) or in wild-type arteries when the primary antibodies were omitted (not shown). Overall, a similar staining pattern and intensity profile were observed in injured arteries within 2 weeks after injury. Immunoreactivity for vitronectin, which interacts with PAI-1,30,31 was detected in the adventitia of electrically injured arteries of either genotype only within 2 and 3 weeks after injury. This suggests that PAI-1 and vitronectin in the media and intima did not colocalize, either spatially or temporally, during cell migration (not shown). Taken together, these data indicate that during vascular wound healing, PAI-1 is produced by smooth muscle, endothelial, inflammatory, and fibroblast-like cells that repopulate the vascular wound. Induction of PAI-1 in activated endothelial, inflammatory, and smooth muscle cells during wound healing has been reported previously.2,32,33
Zymographic Analysis of PAs in PAI-1−/− Mice
To evaluate whether deficiency of PAI-1 caused any changes in t-PA or u-PA expression in uninjured arteries, t-PA and u-PA activities were zymographically quantified in extracts of PAI-1+/+ and PAI-1−/− arteries. There was a slight increase in t-PA–mediated lysis in PAI-1−/− arteries (0.99±0.06; n=3) compared with PAI-1+/+ arteries (0.60±0.03; n=5; P<.05 by ANOVA), whereas u-PA–mediated lysis in PAI-1+/+ arteries (1.18±0.13; n=5) was similar to that in PAI-1−/− arteries (0.90±0.12; n=3; P=NS).
PA activities of injured arteries were evaluated by in situ zymography on sections, allowing precise discrimination between uninjured and injured regions as deduced from histological analysis on adjacent sections. Because the absolute levels of lysis varied between experiments and the specific activities of t-PA and u-PA in this assay were significantly different (due to their different affinity for and activation by fibrin), PA activities were compared by expressing them as a ratio of lysis present in injured versus noninjured segments per artery. Lysis was quantified within 1 week after injury, when genotypic differences in neointima formation were significant.
The ratio of lysis observed in injured versus noninjured sections within 1 week after injury was 1.02±0.09 (n=17) in PAI-1+/+ arteries and 2.3±0.3 (n=17; P<.05 versus PAI-1+/+) in PAI-1−/− arterial sections. Because lysis of the fibrin overlay was inhibited >95% by inclusion of neutralizing t-PA antibodies in the overlay, t-PA activity after injury was similar in PAI-1+/+ arteries and slightly increased in PAI-1−/− arteries. Accurate quantification of u-PA activity could not be reliably performed due to incomplete immunoneutralization of t-PA activity, which lyses fibrin overlays much more efficiently than u-PA (data not shown).
Adenoviral PAI-1 Gene Transfer
Rationale and Study Protocol
To confirm the inhibitory role of PAI-1 on neointima formation, we studied the effect of adenoviral PAI-1 gene transfer on arterial neointima formation. PAI-1−/− mice were used as recipients for PAI-1 gene transfer for two reasons: (1) it permits a greater genetic shift in the proteolytic balance without possible interference by endogenous PAI-1, and (2) neointima formation is greater and occurs at earlier times after injury in PAI-1−/− than in PAI-1+/+ mice (7 to 14 days), coincident with the time window of transgene expression after adenoviral gene transfer using first-generation adenovirus vectors (7 to 10 days). Systemic injection of AdCMVPAI-1 virus was used instead of local PAI-1 gene transfer in the injured artery, for technical reasons and because it allows the study of the effect of increased plasma PAI-1 levels on neointima formation (see below). Therefore, PAI-1−/− mice were injected intravenously with AdCMVPAI-1 or control AdRR5 virus 3 days after electric injury (eg, at a time just before the first cells start to migrate), and neointima formation was quantified after 2 weeks. Plasma PAI-1 levels were determined 4 days after viral infection, which corresponds to the time of maximal PAI-1 gene expression (as revealed from initial experiments).
PAI-1 Expression After Adenoviral PAI-1 Gene Transfer
Intravenous injection of 109 pfu AdCMVLacZ virus in PAI-1−/− mice (n=5) resulted in preferential expression of LacZ in the liver (as revealed by the widespread blue staining; Fig 8a⇓), consistent with numerous previous studies,34 In contrast, transduced cells were not observed in the uninjured aorta (Fig 8b⇓) or femoral artery (Fig 8c⇓), or in the femoral artery that was electrically injured 3 days before virus injection (Fig 8d⇓). Thus, systemic virus injection failed to locally transduce vascular cells but preferentially targets expression of the transgene to the liver.
AdCMVPAI-1 efficiently infected cultured human umbilical vein endothelial cells; synthesis of active PAI-1 was <7.5 fmol/h when they were infected with control AdRR5 virus and 14 pmol/h (1870-fold increase) when infected with AdCMVPAI-1 virus. Intravenous injection of PAI-1−/− mice with 2 × 109 pfu AdCMVPAI-1 resulted in PAI-1 immunoreactivity in ∼30% of hepatocytes but not in uninfected control animals or animals infected with AdRR5 (Fig 7h⇑). Production of human PAI-1 in the liver significantly increased plasma PAI-1 levels; human plasma PAI-1 antigen levels increased from undetectable (<0.8 ng/mL) after AdRR5 adenovirus injection to 0.8±0.2, 13±7, and 61±8 μg/mL at 4 days after intravenous injection with 2×108, 6×108, and 2×109 pfu AdCMVPAI-1. These plasma levels of human PAI-1 are significantly higher than the usual murine PAI-1 levels in the plasma from wild-type mice (∼2 ng/mL).19,35 Active PAI-1 levels in mice injected with 2×108, 6×108, and 2×109 pfu AdCMVPAI-1 were 80±8, 130±80, and 3000±140 ng/mL, respectively, whereas they were undetectable in uninfected or AdRR5- infected mice (<0.8 ng/mL). Through immunostaining, human PAI-1 immunoreactivity was detected within 1 week after electric injury in the neointima from PAI-1−/− mice that received a viral dose of 109 pfu AdCMVPAI-1 but not of AdRR5 (Fig 8e⇑ and 8f⇑). Taken together, systemic injection of AdCMVPAI-1 did not transduce the vascular cells directly but increased production of PAI-1 in the transduced hepatocytes, with resultant augmented levels of PAI-1 in the plasma and its deposition in the extracellular matrix of the neointima in the injured artery.
No side effects of liver inflammation (histologically analyzed) or toxicity (revealed by measurements of the liver enzymes alanine aminotransferase and aspartate aminotransferase) resulted from the highest dose of adenoviral PAI-1 gene transfer (not shown). There also were no signs of thrombosis in AdCMVPAI-1–infected mice.
Inhibition of Neointima Formation by PAI-1 Gene Transfer
PAI-1−/− mice were injected intravenously with increasing doses of AdCMVPAI-1 or AdRR5 at 3 days after electric injury, and neointima formation was quantified after 2 weeks. Intravenous injection of AdCMVPAI-1, at doses ranging from 2×108 to 2×109 pfu, inhibited neointima formation in PAI-1−/− mice in a dose-related manner (Fig 8g⇑). Half-maximal and maximal inhibition rates were obtained at a dose of 2×108 and 6×108 pfu AdCMVPAI-1, respectively, which are ≥1 order of magnitude lower than the toxic dose of adenovirus (>1010 pfu/mouse). Neointima formation was reduced by >85% by comparison with untreated or AdRR5-treated control mice, whereas the administration of an equivalent dose of AdRR5 was ineffective. A representative example of an AdCMVPAI-1–treated artery is shown in Fig 2f⇑.
Role of PAI-1 in Smooth Muscle Cell Migration and Tissue Remodeling
The present study demonstrates that PAI-1 is an inhibitory regulator of arterial neointima formation during vascular wound healing and functions by affecting migration rather than proliferation of smooth muscle cells (Fig 1⇑). This hypothesis is consistent with previous reports that both the plasmin inhibitor tranexamic acid and metalloproteinase inhibitors reduce smooth muscle cell migration in the rat carotid artery without affecting smooth muscle cell proliferation.7,36 Apart from controlling cellular migration, PAI-1 may also affect tissue remodeling during arterial wound healing. Removal of the necrotic debris and formation of a neoadventitia were impaired in u-PA–deficient18 and plasminogen-deficient arteries17 but accelerated in PAI-1–deficient arteries (present study), implicating a role for u-PA/PAI-1 in remodeling of the vessel wall. That both wild-type and PAI-1−/− arteries eventually develop a similar degree of neointima may suggest that the PA/PAI-1 system is not the sole proteinase system involved in smooth muscle cell migration. To what extent matrix metalloproteinases participate in this process remains to be determined.
PAI-1 could affect vascular wound healing by inhibiting u-PA–mediated pericellular plasmin proteolysis or by inhibiting u-PAR–31 or integrin-30 mediated cell adhesion and migration through interaction with vitronectin. The precise contribution of each of these mechanisms is at present unknown. Although u-PAR is expressed by migrating smooth muscle cells in the electrically injured artery, its deficiency apparently does not affect cellular migration or neointima formation (P. Carmeliet, L. Moons, M. Dewerchin, S. Rosenberg, J.-M. Herbert, F. Lupu, and D. Collen, unpublished observations). In addition, although vitronectin is present in the injured artery, it is mostly present in the adventitia and is more abundant beyond the second week after injury (eg, after maximal cell migration has occurred; not shown). Thus, although we cannot exclude that PAI-1 inhibits cellular migration by affecting cell adhesion,30,31 its poor temporal and spatial colocalization with vitronectin, the apparent redundancy of u-PAR, and the finding that deficiency of plasminogen17 also impairs neointima formation to the same extent as deficiency of u-PA18 argue for a role of PAI-1 in controlling proteolysis rather than adhesion.
Expression of PAI-1 During Vascular Wound Healing
PAI-1 expression was minimal in the uninjured femoral artery but significantly induced during wound healing by cells at the migrating front and at sites of tissue remodeling. This is somewhat surprising because migrating cells require increased pericellular proteolysis. This apparent paradox can be explained by the observation that expression of u-PA mRNA and immunoreactivity and, more importantly, that net u-PA–dependent fibrinolytic activity in injured arteries increased by >100-fold. This therefore indicates that u-PA synthesis increased more than PAI-1. Possibly, the increased PAI-1 levels in the vascular wound may limit excessive pericellular proteolysis, which might result from uncontrolled u-PA expression.
Zymographic analysis indicated that t-PA activity was somewhat higher in PAI-1−/− than in PAI-1+/+ arteries before and after injury. However, it is unlikely that t-PA is responsible for the facilitated cellular migration in the mutant mice. Indeed, u-PA increased much more (>100-fold) than t-PA (less than twofold) after injury, and a deficiency of u-PA, but not t-PA, impaired smooth muscle migration and arterial neointima formation.18 It is at present unclear what role t-PA plays in vascular wound healing, although it is presumed to participate in passivation of the injured vessel lumen.1,2
Role of PAI-1 in Endothelial Cell Regeneration
Endothelial cell regrowth was not accelerated by PAI-1 deficiency. This is somewhat surprising in view of the induced expression of PAs and PAIs in regenerating endothelium in vivo and in vitro after injury (present study and References 2, 33, and 372 33 37 ). This may suggest that the plasminogen system is redundant, possibly because other proteinase systems (maybe matrix metalloproteinases) play a more important role. Alternatively, endothelial cells may require more (plasmin) pericellular proteolysis when they cross anatomic barriers (eg, during the formation of new blood vessels) than when they migrate alongside the internal elastic lamina of a denuded artery. Indeed, pathological neovascularization during tumor formation appears to be impaired in mice lacking u-PA–mediated plasmin proteolysis.38
Arterial Injury Models in Mice: Relevance for Human Restenosis?
PAI-1 inhibited neointima formation in both the electric and mechanical injury models. The rate versus the extent of neointima formation in electrically versus mechanically injured PAI-1−/− arteries may relate to the amount of injury and cell death, the degree and timing of proliferation or migration, thrombus formation, inflammation, and the amount and composition of the deposited matrix, which may differ significantly between the two models.20 Although the mechanical injury model more closely reflects the injury sustained by vessels in human subjects after angioplasty, the electric injury model may better reflect the healing process during conditions of more severe inflammation, cell death, or tissue necrosis. Leukocytes, cell death, and matrix deposition are characteristic features of advanced atherosclerotic plaques and have previously been suggested to determine the cellular responses during restenosis.6 The topographic healing and cellular migration pattern in electrically injured arteries are similar to those observed after venous grafts, laser-induced thermal injury, necrotizing transluminal ligation injury, and end-to-end microvascular anastomosis (see Reference 2020 for references).
It remains to be determined how PAI-1 gene transfer will affect arterial wound healing in other animal models or in patients. Indeed, its overall effect may depend on the level, timing, and cellular distribution of recombinant PAI-1 gene expression; presence of inflammation and necrosis and the expression pattern of endogenous plasminogen activators in the injured vessel; degree to which intimal thickening depends on cellular migration; and cellular requirement for pericellular plasmin proteolysis. Intravascular seeding of retrovirally transduced smooth muscle cells overexpressing PAI-1 in balloon-injured rat carotid arteries inhibits neointima formation (Clower A.W. et al, personal communications), whereas a virus-encoded serine proteinase inhibitor, SERP-1 (inhibiting plasminogen activators and plasmin), also blocks atherosclerotic plaque development after balloon angioplasty in the rabbit femoral artery.39 Furthermore, although most currently available (gene) therapies aim to reduce intimal cell accumulation to improve blood flow, the level of PAI-1–mediated inhibition of intimal thickening may have to be controlled because insufficient repopulation may leave the arterial wound weak and unstable. Conversely, our finding that inhibition of PAI-1 may aggravate intimal thickening cautions against the (excessive) use of PAI-1 inhibitors used to improve thrombolysis and reduce atherosclerosis during periods of healing after vascular reconstructions.
In conclusion, this study provides evidence for an inhibitory role of PAI-1 in vascular wound healing and arterial neointima formation due to inhibition of cellular migration.
Selected Abbreviations and Acronyms
|PAI||=||plasminogen activator inhibitor|
|pfu||=||plaque forming unit(s)|
|t-PA||=||tissue plasminogen activator|
|u-PA||=||urokinase plasminogen activator|
This work was supported by the Flanders Interuniversity Institute for Biotechnology and by NIH SCOR grant HL-17669. The authors gratefully acknowledge the technical assistance of B. Amarneh, A. Bouché, I. Cornelissen, O.J. Cooper, M. De Mol, B. Hermans, G. Luyckx, A. Manderveld, and A. Van den Boomen and the help of M. Deprez for artwork.
- Received March 25, 1997.
- Revision received July 23, 1997.
- Accepted August 14, 1997.
- Copyright © 1997 by American Heart Association
Clowes AW, Clowes MM, Au YP, Reidy MA, Belin D. Smooth muscle cells express urokinase during mitogenesis and tissue-type plasminogen activator during migration in injured rat carotid artery. Circ Res. 1990;67:61–67.
Reidy MA, Irvin C, Lindner V. Migration of arterial wall cells: expression of plasminogen activators and inhibitors in injured rat arteries. Circ Res. 1996;78:405–414.
Carmeliet P, Collen D. Role of the plasminogen/plasmin system in thrombosis, hemostasis, restenosis and atherosclerosis: evaluation in transgenic animals. Trends Cardiovasc Med. 1995;5:117–122.
Okada SS, Groβmyer SR, Barnathan ES. Contrasting effects of plasminogen activators, urokinase receptor, and LDL-related protein on smooth muscle cell migration and invasion. Arterioscler Thromb Vasc Biol. 1996;16:1269–1276.
Fearns C., Samad F, Loskutoff DJ. Synthesis and localisation of PAI-1 in the vessel wall. In: Advances in Vascular Biology. Camberwell, Victoria, Australia: Gordon and Breach Publishers; 1995.
Libby P, Schwartz D, Brogi E, Tanaka H, Clinton SK. A cascade model for restenosis: a special case of atherosclerotic progression. Circulation. 1992;86 (suppl. III):III-47–III-52.
Villaverde CA, Miralles M, Salvat N, Juan O, Tibau N, Pena E, Vigo P. Regulation of plasmin-generating activity in human arteries. Fibrinolysis. 1996;10(suppl 3):148a. Abstract.
Schneiderman J, Sawdey MS, Keeton M, Bordin GM, Bernstein EF, Dilley RB, Loskutoff DJ. Increased type 1 plasminogen activator inhibitor gene expression in atherosclerotic arteries. Proc Natl Acad Sci U S A. 1992;89:6998–7002.
Lupu F, Bergonzelli GE, Heim DA, Cousin E, Genton CY, Bachmann F, Kruithof EKO. Localization and production of plasminogen activator inhibitor-1 in human healthy and atherosclerotic arteries. Arterioscl Thromb. 1993;13:1090–1100.
Fuji S, Sawa H, Saffitz JE, Lucore CL, Sobel BE. Induction of endothelial cell expression of the plasminogen activator inhibitor type 1 gene by thrombosis in vivo. Circulation. 1992;86:2000–2010.
Lee E, Vaughan DE, Parikh SH, Grodzinsky AJ, Libby P, LarkMW, Lee RT. Regulation of matrix metalloproteinases and plasminogen activator inhibitor-1 synthesis by plasminogen in cultured human vascular smooth muscle cells. Circ Res. 1996;78:44–49.
Collen D, Lijnen HR. Basic and clinical aspects of fibrinolysis and thrombolysis. Blood. 1991;78:3114–3124.
Carmeliet P, Moons L, Herbert JM, Crawley J, Lupu F, Lijnen R, Collen D. Urokinase but not tissue plasminogen activator mediates arterial neointima formation in mice. Circ Res.. 1997;81:829-839.
Carmeliet P, Stassen J-M, Schoonjans L, Ream B, van den Oord JJ, De Mol M, Mulligan RC, Collen D. Plasminogen activator inhibitor-1 gene-deficient mice, II: effects on hemostasis, thrombosis and thrombolysis. J Clin Invest. 1993;92:2756–2760.
Lindner V, Fingerle J, Reidy MA. Mouse model of arterial injury. Circ Res. 1993;73:792–796.
Preissner KT, Grulich-Henn J, Ehrlich HJ, Declerck P, Justus C, Collen D, Pannekoek H, Müller-Berghaus G. Structural requirements for the extracellular interaction of plasminogen activator inhibitor-1 with endothelial cell matrix-associated vitronectin. J Biol Chem. 1990;265:18490–18498.
Lupu F, Heim DA, Bachmann F, Hurni M, Kakkar VV, Kruithof EKO. Plasminogen activator expression in human atherosclerotic lesions. Arterioscl Thromb Vasc Biol. 1995;15:1444–1455.
Reido FX, Munford RS, Campbell WB, Reisch JS, Chien KR, Gerard RD. Deacylated lipopolysaccharide inhibits plasminogen activator inhibitor-1, prostacyclin, and prostaglandin E2 induction by lipopolysaccharide but not by tumor necrosis factor-alpha. J Immunol. 1990;144:3506–3512.
Gerard RD, Meidell RS. Adenovirus vectors. In: Hames BD, Glover D, eds. DNA Cloning: A Practical Approach: Mammalian Systems. Oxford, UK: Oxford University Press; 1995.
Declerck P, Alessi MC, Verstreken M, Kruithof EKO, Juhan-Vague I, Collen D. Measurements of PAI-1 in biological fluids with a murine monoclonal antibody-based enzyme linked immunoassay. Blood. 1988;71:220–225.
Deng G, Curriden SA, Wang S, Rosenberg S, Loskutoff DJ. Is plasminogen activator inhibitor-1 the molecular switch that governs urokinase receptor-mediated cell adhesion and release. J Cell Biol. 1996;134:1563–1571.
Herz J, Gerard RD. Adenovirus-mediated transfer of low density lipoprotein receptor gene acutely accelerates cholesterol clearance in normal mice. Proc Natl Acad Sci. U S A. 1993;90:2812–2816.
Bendeck MP, Zempo N, Clowes AW, Galardy RE, Reidy MA. Smooth muscle cell migration and matrix metalloproteinase expression after arterial injury in the rat. Circ Res. 1994;75:539–545.
Bacharach E, Itin A, Keshet E. In vivo patterns of expression of urokinase and its inhibitor PAI-1 suggest a concerted role in regulating physiological angiogenesis. Proc Natl Acad Sci U S A. 1992;89:10686–10690.
Sabapathy KT, Pepper MS, Kiefer F, Möhel-Steinlein U, Tacchini-Cottier F, Fetka I, Breier G, Risau W, Carmeliet P, Montesano R, Wagner EF. Polyoma middle-T induced vascular tumor formation: the role of the plasminogen activator/plasmin system. J Cell Biol. 1997;137:953–963.
Lucas A, Liu L, Macen J, Nash P, Dai E, Stewart M, Graham K, Etches W, Boshkov L, Nation PN, Humen D, Lundstrom-Hobman M, McFadden G. Virus-encoded serine proteinase inhibitor SERP-1 inhibits atherosclerotic plaque development after balloon angioplasty. Circulation.. 1996;94:2890-2900.