A Quantitative In Vitro Model of Smooth Muscle Cell Migration Through the Arterial Wall Using the Human Amniotic Membrane
Objective— The development of intimal hyperplasia involves smooth muscle cell (SMC) migration into the intima and proliferation. Matrix metalloproteinases and their tissue inhibitors play important roles in this process. In this study, we describe a novel in vitro model for studying SMC migration through the vessel wall.
Methods and Results— Human aortic SMCs (hASMCs) labeled with 125I-iododeoxyuridine or unlabeled were grown on the stromal aspect of the human amniotic membrane. Mechanical damage to endothelial cells grown on the basement membrane and addition of growth factors or platelets were characterized for their effect on SMC migration into the stroma both by histological methods and by measuring the radioactivity associated with the membrane after removal of noninvasive SMCs. To assess the reliability of the model, the cells were infected with a recombinant adenovirus encoding the tissue inhibitor of metalloproteinase-1 (TIMP-1). Addition of a platelet-derived growth factor gradient stimulated hASMC infiltration into the stroma. This effect was abolished with TIMP-1-transduced hASMC, confirming that TIMP-1 overexpression blocks SMC invasion of the stroma.
Conclusions— This in vitro model of SMC migration in the vessel wall provides an inexpensive, quantitative, and reliable tool to study the molecular and cellular mechanisms of intimal hyperplasia.
Smooth muscle cell (SMC) proliferation and migration from the tunica media to the intima are key mechanisms of intimal hyperplasia and atherosclerosis.1,2⇓ These cell functions are controlled by several growth factors and require the action of extracellular matrix (ECM)-degrading proteinases. Platelet-derived growth factor (PDGF) and basic fibroblast growth factor (FGF-2) stimulate SMC migration and proliferation in tissue culture3,4⇓ and in injured vessels.5–8⇓⇓⇓ However, growth factors alone do not stimulate SMC migration or proliferation in organ cultures9,10⇓ or uninjured vessels.11 A variety of extracellular proteinases are involved in vascular remodeling.12 Among them, the matrix metalloproteinases (MMPs) have important roles.12,13⇓ In extracellular spaces, MMP activity is controlled by tissue inhibitors of metalloproteinases (TIMPs), which are secreted by a variety of cells, including SMC.12,14⇓ Retrovirus-mediated overexpression of TIMP-1 in SMC reduces SMC invasion of ECM in vitro and migration into the intima in vivo.15–17⇓⇓
A variety of experimental models has been developed to study neointima formation in vivo or in vitro. Animal models afford reproducing human lesions but have major disadvantages, including high cost, the high variability intrinsic with animal models, species variability, and difficulty to quantitate the results in an objective, observer-independent manner. Conversely, in vitro models afford quantitative characterization of select pathogenetic components of intimal hyperplasia, eg, smooth muscle cell proliferation or migration, but are limited by their artificial conditions. The structure and composition of the substrates used for studying ECM degradation or cell migration in vitro (eg, Matrigel) differ considerably from their in vivo counterparts. We attempted to develop a simple and reproducible model for intimal hyperplasia in vitro using a natural ECM. In this study, we describe an in vitro model of the arterial wall using the human amniotic membrane18,19⇓ to characterize SMC migration through a natural collagenous stroma. To test the reliability of our model, we characterized the effect of factors that affect formation of intimal hyperplasia in vivo, including the inhibitory effect of TIMP-1 overexpression. The results show that this in vitro set up can reliably be used to reproduce the structure of a vessel wall in vitro and to quantitatively study SMC migration through a natural ECM.
Cells and Culture Media
Human aortic smooth muscle cells (hASMC; passage 2-8; Clonetics, San Diego, Calif) were grown in SMC basal medium supplemented with 10 ng/mL human epidermal growth factor (hEGF), 2 ng/mL human fibroblast growth factor-2 (hFGF), 0.39 μg/mL dexamethasone (Clonetics), 50 μg/mL gentamycin, 50 μg/mL amphotericin-B, and 5% FBS (Gibco BRL, Life Technologies, Inc). Bovine smooth muscle cells (BSMCs; passage 3-10) and bovine aortic endothelial cells (passage 8-14), kindly provided by Dr D.B. Rifkin (NYU School of Medicine, New York, NY), were grown in DMEM containing 5% DCS (Gibco BRL).
Coculture of Endothelial and Smooth Muscle Cells on the Human Amniotic Membrane
Human amniotic membranes were prepared as described18–20⇓⇓ with several modifications. The amniotic membrane of fresh placentas obtained from cesarean sections was separated from the chorion by blunt dissection under sterile conditions. A Teflon ring (16 mm ID, 22 mm OD, 9.5 mm high; Rockefeller University Instrument Shop) was fastened to the amnion with a Viton O-ring (C.E. Conover), the fetal (epithelial) aspect of the membrane facing the inside of the ring. The amnion was separated from the placenta and washed twice in PBS containing 1000 U/mL penicillin, 40 mg/mL streptomycin sulfate, and 2.5 μg/mL amphotericin-B (PBS-PSF). Additional rings (as many as 100 per membrane) were fastened to the membrane with the same orientation as the first one. The rings were separated from each other, washed in PBS-PSF, and incubated in NH4OH 0.25 mol/L at room temperature for 2 hours to lyse the epithelial cell layer. Subsequently, the remnants of the epithelium were scraped off with a rubber policeman. The membrane-ring setups were incubated in PBS-PSF at 37°C overnight to remove traces of NH4OH and either used immediately or stored at 4°C in DMEM. A schematic diagram of the membrane-ring setup is shown in Figure 1. For SMC migration assays, the membrane-ring setups were placed into 6-well culture plates with the stromal aspect of the amnion facing up. A sterile silicone rubber ring (23 mm OD, 14.5 mm ID, 3 mm high) was glued on top of each Teflon ring with sterilized nontoxic silicone lubricant (Dow Corning Co) to create a culture chamber. The bottom of this culture chamber is the stromal aspect of the amnion, the wall is the inner wall of the silicone rubber ring (Figure 1A). Unlabeled or 125I-dUR-labeled hASMC or BSMC (3×105 cells/0.2 mL of medium) were seeded into these culture chambers and allowed to attach to the amnion stroma for 1 hour at 37°C. The silicone rings were removed and placed on the bottom of the plastic wells, and the membrane-ring setups were turned over and placed onto the silicone rings. One milliliter of growth medium containing 3×105 BAEC was added onto the basement membrane inside the Teflon rings, and 3 mL of complete medium was added to the outside of the rings so that no hydrostatic pressure would be exerted on the amniotic membrane (Figure 1B). After 2 days of incubation at 37°C, when the endothelial cells formed a confluent monolayer, either 10 ng/mL of PDGF (Sigma Chemical Co) or 2×109 human platelets or 10 ng/mL of FGF-2 (kindly provided by Dr D.B. Rifkin, NYU School of Medicine, New York, NY) or an equivalent volume of control medium was added into the upper compartment of the culture chambers. In some experiments, the endothelial cells were omitted, and 1 mL of medium without cells was added onto the amnion basement membrane. After the indicated incubation times, the cultures were fixed in 3% phosphate-buffered glutaralaldehyde for histological examination. When 125I-dUR-labeled SMCs were used, the cells were labeled with 125I-dUR and the amnion cultures were processed as described19,20⇓ (please see the online supplement, available at http://atvb.ahajournals.org).
Construction of Recombinant Adenoviral Vectors and Characterization of Transduced Cells
Confluent hASMCs were transduced with recombinant adenoviruses encoding TIMP-1 (Ad.TIMP-1) or β-galactosidase (Ad.βgal)21,22⇓ and characterized for TIMP-1 expression by Northern and Western blotting and by reverse gelatin zymography, as described (please see the online supplement).
Statistical analysis was performed by the ANOVA test for comparison of multiple groups. P<0.05 was considered significant.
Characterization of the In Vitro Reconstituted Arterial Wall
The fetal aspect of the human amnion membrane is covered with a monostratified cuboidal epithelium. After lysis with NH4OH and scraping as described in Methods, virtually all epithelial cells were removed and the basement membrane was completely denuded with no microscopically detectable damage (Figures 2A through 2C). The amnion basement membrane, ≈10 μm in thickness, is similar in structure and composition to a vascular basal lamina.18 The constituents of the 500-μm-thick stroma include collagen types I, III, V, VII, and XII, fibronectin, and laminin-1 and -5.18,23–27⇓⇓⇓⇓⇓ Movat staining for elastin showed that the stroma of the amniotic membrane contains a meshwork of elastic fibers (Figures 2A and 2B). The stroma also contains occasional fibroblasts that were not lysed by treatment with NH4OH. However, consistent with previous reports,28 by electron microscopy no signs of viability were observed in these cells (data not shown).
hASMCs seeded onto the stromal aspect of the amnion spread on the substrate within 1 hour and remained attached after the amnion-ring setup was turned over. The cells proliferated and formed a confluent monolayer of noninvasive cells within 72 hours after seeding (Figure 2C). Incubation for as many as 12 days did not affect the integrity of the monolayer, and occasional hASMCs were rarely seen invading into the stroma. BAECs seeded onto the basement membrane also formed a noninvasive, sealed monolayer (Figure 2D). Similar results were obtained with human aortic endothelial cells (data not shown).
To characterize factors that stimulate SMC invasion into the stroma, we tested several experimental conditions (Table). Human or bovine aortic SMCs grown on the amnion could be stimulated to migrate into the stroma by addition of 10 ng/mL of PDGF in the upper compartment to create a gradient. In contrast, addition of PDGF to the lower compartment was ineffective. A similar effect was observed when 2×109 human platelets were added to the endothelial cell monolayer. Addition of 10 or 50 ng/mL of FGF-2 either to the upper or the lower compartment did not increase migration. Addition of 10% FBS to the growth medium was also ineffective. To simulate intimal injury, we scratched the endothelial cell monolayer with the blunt edge of a forceps and measured BSMC migration into the amniotic stroma after 6 days by histological methods and by labeling with 125I-dUR. Injury of the endothelium alone did not affect BSMC migration. When the endothelium was injured in the presence of either 10 ng/mL of PDGF or 2×109 platelets, migration was similar to but not significantly higher than in the absence of injury. Regardless of the experimental setting, human or bovine SMCs were unable to migrate through the basement membrane into the endothelial cell monolayer.
These results showed that endothelial damage per se does not stimulate smooth muscle cell migration, FGF-2 released from damaged endothelial cells or added exogenously does not affect SMC migration, and PDGF is a major inducer of SMC migration in the stroma. Because under our experimental conditions SMC migration was not affected by the presence of an endothelial cell layer on the basement membrane, subsequent experiments were done in the absence of endothelial cells.
Effect of TIMP-1 Overexpression on Vascular Remodeling
To characterize the role of TIMP-1 in SMC migration in our model, we transduced hASMC with Ad.TIMP-1 or, as a control, with Ad.βgal. By Northern blotting, cells transduced with Ad.TIMP-1 expressed high amounts of recombinant 1.1 kb TIMP-1 mRNA. Western blotting with TIMP-1 antibody and reverse gelatin zymography showed considerably higher levels of TIMP-1 in the conditioned medium of Ad.TIMP-1-transduced hASMCs than in control or Ad.βgal-infected cells (online Figure I, available at http://atvb.ahajournals.org). Thus, hASMC transfection with Ad.TIMP-1 resulted in overexpression of functional TIMP-1. The growth rate of virus-transduced cells was slightly but not significantly higher than that of nontransduced cells (data not shown). To characterize the effect of TIMP-1 gene transfer on SMC migration, transduced hASMCs were seeded onto the stromal aspect of the amniotic membrane and migration was measured. TIMP-1-transduced SMCs, as well as nontransduced or control Ad.βgal-infected cells, formed a confluent monolayer of noninvasive cells within 72 hours after seeding. In the presence of 10 ng/mL PDGF, untransduced or Ad.βgal-transduced hASMCs migrated into the collagenous stroma to the same extent. Cells were found at varying depth in the stromal meshwork, with some cells adhering to the basement membrane (Figures 3A and 3C). Conversely, only occasional superficial Ad.TIMP-1-transduced hASMCs were detected in the stroma in the presence of a PDGF gradient (Figure 3B). In 10 random microscopic fields (×200), only single TIMP-1-transduced hASMCs were found in the stroma after 3 or 6 days of incubation. In contrast, up to 20-fold more nontransduced or Ad.βgal-transduced hASMCs were present in the membrane under the same conditions (P<0.001) (Figure 3D). To quantitate SMC migration in an observer-independent manner, 125I-dUR-labeled hASMCs were grown on the amnion stroma for 3 days and the radioactivity associated with the membrane was measured after removal of the noninvasive monolayer as described19,20⇓ (please see the online supplement). As shown in Figure 4, with nontransduced hASMCs in the absence of PDGF, the radioactivity associated with the membrane was 14.2±3.77% of the total radioactivity (mean±SEM). This value increased 2.1-fold to 30.6±3.12% (P<0.05) in the presence of 10 ng/mL PDGF, indicating that a large number of hASMCs migrated into the amnion stroma. With Ad.TIMP-1-transduced cells, the radioactivity associated with the membrane was 6.9±5% of the total radioactivity, ie, significantly lower than with control, nontransduced (P<0.01) or Ad.βgal-transduced hASMCs (25.6±5.6%, P<0.05).
A variety of in vivo and in vitro models have been developed to study smooth muscle cell migration during the development of intimal hyperplasia. However, all suffer from major disadvantages, including the high cost and variability of animal models, difficulty to quantitate the results in an objective, observer-independent manner, or the artificial conditions of in vitro models. In an attempt to develop a simple and reproducible model for intimal hyperplasia, we adapted an in vitro assay that had been developed to study tumor invasion and angiogenesis using a natural basement membrane and stroma.18–20⇓⇓ Indeed, neointima formation involves tissue remodeling processes that share several features with tumor invasion and angiogenesis.19,20⇓ The data reported here show that the human amniotic membrane affords engineering an in vitro model of the vessel wall to study vascular remodeling in vitro. Our in vitro model has several advantages. First, it is inexpensive. One human amniotic membrane provides a large surface of basement membrane and stroma that can be used for assaying as many as 100 samples.18–20,28⇓⇓⇓ Cocultures of endothelial and smooth muscle cells on the amniotic membrane can be maintained for at least 14 days without significant decrease in cell viability. Second, it can provide quantitative and reproducible results rapidly. Early events leading to intimal hyperplasia, such as SMC migration, can be quantified in as short a time as 3 days. Third, our model reproduces major components of the arterial wall in vitro using a natural collagenous and elastic stroma. Most in vitro models for tissue remodeling use substrates for degradation or cell migration whose structure and composition differ considerably from in vivo conditions. Occasional fibroblasts present in the amnion stroma are killed by a harsh NH4OH treatment, which also inactivates growth factors and other molecules that can affect a variety of endothelial or SMC functions. In contrast, Matrigel, the widely used in vitro reconstituted extracellular matrix, contains several growth factors. The amnion membrane has a dense basement membrane and an abundant collagenous stroma with elastin fibers, which confer great resistance to traction and resilience on the tissue and make it similar to an artery. In contrast, Matrigel lacks the resilience and consistency typical of the arterial wall. Finally, the simplicity of our model allows studying SMC migration without the many confounding factors of in vivo models, such as the presence of blood components, endothelial-white blood cell interactions, and hemodynamic influences. Our in vitro model lacks several factors, including blood cells or cholesterol, that modulate vascular remodeling in vivo. However, these factors can be characterized by adding them exogenously. In our model, the presence of confluent endothelial cells on the basement membrane did not affect SMC migration. Likewise, injury of the endothelial cell monolayer or addition of FGF-2 did not induce SMC migration into the stroma. This observation contrasts with previous findings that endothelial cells affect SMC morphology, proliferation, and protein synthesis.29,30⇓ A stimulatory effect of endothelial cells on SMCs has been proposed based on observations that in animal models intimal hyperplasia can occur with an intact endothelium31 and in humans it can be more severe in areas of intact endothelium relative to partially denuded areas.32 However, addition of endothelial cell-conditioned medium to or cocultivation of SMC with endothelial cells inhibits SMC proliferation.33,34⇓ Endothelial cell-derived FGF-2 acts as a potent mitogen but is not chemotactic for SMC in vitro,3,8,35⇓⇓ although it can stimulate SMC migration in vitro and in vivo.36 In our model, neither injury of the endothelial cell monolayer and consequent FGF-2 release nor addition of FGF-2 induced SMC migration. Conversely, in the presence of a PDGF gradient or platelets, SMCs migrated into the amnion stroma, confirming that PDGF and platelets have a strong stimulatory effect on SMC migration. Although we incubated our cultures on the amnion for as long as 10 days in the presence of 50 ng/mL PDGF, we never observed SMC migration through the basement membrane. Possible explanations for this observation include that the amnion basement membrane is thicker and denser than a vessel’s basal lamina, SMC migration through the basal lamina requires additional factors (eg, cytokines, proteinases, and/or proteinase inhibitors) missing in our model, or basement membrane integrity was maintained under our experimental conditions, whereas SMC migration through the basal lamina may require damage of this structure. When we injured the endothelial cell layer with the blunt edge of a forceps, histological examination showed an intact basement membrane. In contrast, arterial injury often entails damage of the basal lamina, as is the case for the widely used balloon injury model of the rat carotid artery or in humans during percutaneous transluminal coronary angioplasty. Several studies have shown the role of TIMP-1 or other anti-MMP agents in controlling SMC migration.16 Therefore, we used Ad.TIMP-1-transduced SMCs that overexpress TIMP-1 to test the reliability of our model. Consistent with previous findings, upregulation of TIMP-1 expression by adenovirus gene transfer completely inhibited SMC migration into the amnion stroma, showing the reliability of our model. In conclusion, our in vitro model provides a convenient system for studying the individual contribution of several factors to SMC migration and other aspects of vascular remodeling, as well as a rapid, reliable, and inexpensive method for the preliminary characterization of treatments aimed at preventing intimal hyperplasia in vivo.
This work was supported by funds from the Seymour Cohn Cardiovascular Surgery Research Laboratory and by grants from the NIH (SBIR CA80476-01) and from the United States Department of Defense (DAMD17-99-1-9324) to P.M.
- Received November 14, 2002.
- Accepted March 20, 2003.
- ↵Newby AC, George SJ. Proposed roles for growth factors in mediating smooth muscle proliferation in vascular pathologies. Cardiovasc Res. 1993; 27: 1173–1183.
- ↵Lindner V, Lappi DA, Baird A, Majack RA, Reidy MA. Role of basic fibroblast growth factor in vascular lesion formation. Circ Res. 1991; 68: 106–113.
- ↵Jawien A, Bowen-Pope DF, Lindner V, Schwartz SM, Clowes AW. Platelet-derived growth factor promotes smooth muscle migration and intimal thickening in a rat model of balloon angioplasty. J Clin Invest. 1992; 89: 507–511.
- ↵Ferns GA, Raines EW, Sprugel KH, Motani AS, Reidy MA, Ross R. Inhibition of neointimal smooth muscle accumulation after angioplasty by an antibody to PDGF. Science. 1991; 253: 1129–1132.
- ↵Soyombo AA, Angelini GD, Bryan AJ, Newby AC. Surgical preparation induces injury and promotes smooth muscle cell proliferation in a culture of human saphenous vein. Cardiovasc Res. 1993; 27: 1961–1967.
- ↵Pepper MS. Role of the matrix metalloproteinases and plasminogen activator-plasmin system in angiogenesis. Arterioscler Thromb Vasc Biol. 2001; 21: 1104–1117.
- ↵Hasenstab D, Forough R, Clowes AW. Plasminogen activator inhibitor type 1 and tissue inhibitor of metalloproteinases-2 increase after arterial injury in rats. Circ Res. 1997; 80: 490–496.
- ↵Forough R, Koyama N, Hasenstab D, Lea H, Clowes M, Nikkari ST, Clowes AW. Overexpression of tissue inhibitor of matrix metalloproteinase-1 inhibits vascular smooth muscle cell functions in vitro and in vivo. Circ Res. 1996; 79: 812–820.
- ↵Strauss BH, Robinson R, Batchelor WB, Chisholm RJ, Ravi G, Natarajan MK, Logan RA, Mehta SR, Levy DE, Ezrin AM, Keeley FW. In vivo collagen turnover following experimental balloon angioplasty injury and the role of matrix metalloproteinases. Circ Res. 1996; 79: 541–550.
- ↵Mignatti P, Tsuboi R, Robbins E, Rifkin DB. In vitro angiogenesis on the human amniotic membrane: requirement for basic fibroblast growth factor-induced proteinases. J Cell Biol. 1989; 108: 671–682.
- ↵Aplin JD, Campbell S, Allen TD. The extracellular matrix of human amniotic epithelium: ultrastructure, composition and deposition. J Cell Sci. 1985; 79: 119–136.
- ↵Gerecke DR, Olson PF, Koch M, Knoll JH, Taylor R, Hudson DL, Champliaud MF, Olsen BR, Burgeson RE. Complete primary structure of two splice variants of collagen XII, and assignment of alpha 1(XII) collagen (COL12A1), alpha 1(IX) collagen (COL9A1), and alpha 1(XIX) collagen (COL19A1) to human chromosome 6q12-q13. Genomics. 1997; 41: 236–242.
- ↵Liotta LA, Goldfarb RH, Brundage R, Siegal GP, Terranova V, Garbisa S. Effect of plasminogen activator (urokinase), plasmin, and thrombin on glycoprotein and collagenous components of basement membrane. Cancer Res. 1981; 41: 4629–4636.
- ↵Russo RG, Liotta LA, Thorgeirsson U, Brundage R, Schiffmann E. Polymorphonuclear leukocyte migration through human amnion membrane. J Cell Biol. 1981; 91: 459–467.
- ↵Hofstra L, Tordoir JH, Kitslaar PJ, Hoeks AP, Daemen MJ. Enhanced cellular proliferation in intact stenotic lesions derived from human arteriovenous fistulas and peripheral bypass grafts: does it correlate with flow parameters? Circulation. 1996; 94: 1283–1290.
- ↵Castellot JJ Jr, Addonizio ML, Rosenberg R, Karnovsky MJ. Cultured endothelial cells produce a heparinlike inhibitor of smooth muscle cell growth. J Cell Biol. 1981; 90: 372–379.
- ↵Jackson CL, Raines EW, Ross R, Reidy MA. Role of endogenous platelet-derived growth factor in arterial smooth muscle cell migration after balloon catheter injury. Arterioscler Thromb. 1993; 13: 1218–1226.