Inhibition of Vascular Smooth Muscle Cell Proliferation and Neointimal Accumulation by Adenovirus-Mediated Gene Transfer of Cytosine Deaminase
Background Restenosis remains a significant problem after balloon angioplasty. Previous studies have demonstrated that recombinant adenoviruses are efficient vectors for gene transfer to the arterial wall and can be used to inhibit the proliferative aspect of restenosis. We sought to extend these observations using AdCMV.CD, an adenovirus that encodes cytosine deaminase (CD) and is capable of metabolizing 5-fluorocytosine (5-FC) to 5-fluorouracil.
Methods and Results Infection of vascular smooth muscle cells (VSMC) with AdCMV.CD increases by two to three orders of magnitude the growth-inhibitory effects of 5-FC. The degree of VSMC inhibition in vitro was a function of 5-FC concentration and the level of CD expression. Cells infected with AdCMV.CD exhibited a profound bystander effect on the growth of neighboring cells, which did not require direct cell-to-cell contact. The predominant effect of AdCMV.CD on growth of VSMC appeared to be cytostatic, not cytotoxic. Assessment of this strategy in a rabbit femoral artery model of balloon-induced injury demonstrated that compared with animals in either of two control groups, animals treated with the active combination of infection with AdCMV.CD and 1-week treatment with parenteral 5-FC had a significant reduction at 30 days in the neointimal-to-medial ratio.
Conclusions Our results suggest that adenovirus-mediated gene transfer of CD along with 5-FC administration may be a useful strategy to treat the proliferative aspects of restenosis.
Restenosis remains a significant limitation to the long-term success of percutaneous transluminal balloon angioplasty. Although both the cell types and mechanisms that contribute to restenosis remain controversial, the process undoubtedly depends on multiple factors. One aspect of the process appears to be the migration and proliferation of VSMC (for a review, see Reference 11 ). In an attempt to inhibit the proliferative aspect of restenosis, several recent reports have documented the use of recombinant adenovirus as a vector for gene delivery to areas of vascular injury.2 3 4 5 6 7 8 9 The adenovirally delivered genes have encoded for proteins that either directly or indirectly were cytotoxic,4 5 6 cytostatic,7 8 or antithrombotic.9
One gene therapy–based approach to achieve a local antiproliferative effect involves the delivery of a gene product capable of metabolizing a relatively innocuous prodrug to a more potent inhibitor of cell proliferation. The most widely used strategy involves HSV-tk along with the concomitant administration of the nucleoside analogue ganciclovir. Cells that express HSV-tk are capable of phosphorylating ganciclovir, which in turn allows for its incorporation into elongating strands of DNA. The incorporation of ganciclovir derivatives into DNA is thought to result in chain termination with subsequent growth arrest and ultimate cell death.10 11 In addition, the growth of neighboring cells that do not express HSV-tk can be inhibited because the phosphorylated form of ganciclovir can diffuse to adjacent cells. Recent results suggest that with this strategy, direct cell-to-cell contact is required for a maximal bystander effect.12 13
A similar strategy to inhibit proliferation of tumor cells has been recently demonstrated using the procaryotic gene CD.14 15 16 17 18 19 The CD gene product is capable of enzymatically metabolizing the prodrug 5-FC to the potent antimetabolite 5-FU. Production of 5-FU leads to inhibition of DNA and RNA synthesis, in large part due to the inhibition of thymidylate synthase activity. In this case, the prodrug 5-FC is a safe medication that is currently used for the treatment of severe mycotic infections. Similar to what has been demonstrated with HSV-tk and ganciclovir, gene transfer of CD to the injured arterial wall in the setting of systemic 5-FC administration would therefore be expected to produce high local levels of 5-FU. This in turn may have a beneficial effect by inhibiting local VSMC proliferation without producing any of the systemic symptoms associated with 5-FU treatment. In an effort to assess whether such an approach is feasible, we demonstrate that adenovirus-mediated CD gene transfer along with 5-FC administration is an effective strategy for inhibiting VSMC proliferation both in vitro and in vivo.
Primary cultures of VSMC were obtained through enzymatic digestion of a rabbit thoracic aorta. Cells, unless stated otherwise, were maintained in growth medium defined as M-199 medium (GIBCO-BRL) supplemented with 10% fetal bovine serum. Cells were used between passages 4 and 10 for in vitro studies.
The adenovirus AdCMV.CD containing the Escherichia coli CD gene and the control adenovirus AdCMV.βGAL encoding the E coli LacZ gene have been previously described.18 For each construct, the transgene is under the control of the CMV major immediate early promoter. The methods for preparing and titering viral stock using the 293 cell line (American Type Culture Collection) were as described previously.5
In Vitro Assessment of AdCMV.CD
For assessment of rabbit VSMC growth, cells were infected with the indicated moi with AdCMV.CD or, as a control, AdCMV.βGAL. At 3 days after infection (day 0), cells were trypsinized and reseeded sparsely at a density of ≈3×102 cells/cm2 in medium supplemented with the indicated concentration of 5-FC (Janssen), which was replenished every 48 hours. At the indicated time points, triplicate wells were harvested and counted. All in vitro experiments were performed at least three times.
To determine the potency of the bystander effect, cells were infected with 100 moi of AdCMV.CD or AdCMV.βGAL. At 3 days after infection, VSMC were mixed in the indicated proportions and incubated in medium containing 1 mmol/L 5-FC. Cell number from triplicate cultures was determined 10 days after the addition of the prodrug. For determination of whether cell-to-cell contact was needed for the observed bystander effect, cells infected 2 days previously with AdCMV.CD (100 moi) were incubated for 24 hours in medium containing 0.5% serum supplemented with 5-FC (1 mmol/L). This conditioned low-serum medium was then mixed in the indicated proportions with normal medium containing 0.5% serum and subsequently incubated with quiescent uninfected VSMC for 24 hours. Cells were then stimulated with 10% serum and 18 hours later assessed for the amount of DNA synthesis through incubation of cultures with 1 mCi/mL [3H]thymidine for 3 hours.
Cell cycle analysis of adenovirally infected VSMC was assessed by staining isolated nuclei with propidium iodine with the use of Cycle Test (Beckton-Dickinson) according to the manufacturer’s recommendation. Isolated propidium iodine stained nuclei were analyzed with an EPICS Elite-ESP cytometer (Coulter). Assessment of cell cycle distribution was performed with Multicycle Software (Phoenix Flow Systems).
In Vivo Assessment
To determine the pharmacokinetics of 5-FC in vivo, two New Zealand White rabbits, ≈3 kg in weight, received 200 mg/kg 5-FC via an implanted jugular vein catheter. The 5-FC was dissolved in lactated Ringer’s solution at a concentration of 10 mg/mL. Serial blood draws were made at the indicated times, and the serum level of 5-FC was determined at a commercial laboratory (Mayo Laboratories).
To determine the in vivo efficacy of CD expression and 5-FC administration on neointimal formation, New Zealand White rabbits underwent bilateral femoral artery injury with a 2F embolectomy catheter. Arterial segments ≈2 cm in length were injured through six sequential passes of the balloon catheter. The catheter was introduced via an arteriotomy in the distal femoral artery at the level of the knee. Immediately after injury, 300 μL of a viral solution containing 3×109 pfu of AdCMV.CD or AdCMV.βGAL was instilled through the arteriotomy after proximal and side branch vascular control was obtained with vascular clips. Fifty minutes later, the viral solution was removed, and the arteriotomy was closed. On the next day, treatment with 5-FC (200 mg/kg) or lactated Ringer’s solution was begun and continued for 7 days. Thirty days after injury, arterial segments were perfusion fixed at physiological pressure with phosphate-buffered solution containing 10% paraformaldehyde. N/M values were determined on the basis of paraffin-embedded sections by a blinded observer using digital planimetry as previously described.5 All animals received an implantable jugular catheter 2 weeks before femoral injury. A total of 15 animals (30 vessels) were used in the study to determine in vivo efficacy. There were no deaths; however, six vessels were excluded prospectively before assessment of N/M due to thrombus. The excluded arteries included two animals receiving AdCMV.βGAL/5-FC, three animals receiving AdCMV.CD/saline, and one animal receiving AdCMV.CD/5-FC.
For assessment of in vivo gene transfer efficiencies, arterial segments were infected as described above with 3×109 pfu of AdCMV.βGAL and processed using X-GAL (Gold Biotechnology) as a chromagenic substrate.2
All animals were studied under protocols approved by the Animal Care and Use Committee of the National Heart, Lung, and Blood Institute and in accordance with the Guide for the Care and Use of Laboratory Animals.
Results are presented as mean± SEM. Values of P<.05 were considered significant. Assessment of in vivo efficacy was made using an unpaired two-tailed t test between groups. An ANOVA of the three separate groups was significant (P<.01).
AdCMV.CD Inhibits VSMC Proliferation
In an attempt to assess the ability of adenovirus-mediated gene transfer of CD along with 5-FC treatment to inhibit VSMC proliferation, primary rabbit VSMC were infected with AdCMV.CD or, as a control, AdCMV.βGAL. As seen in Fig 1⇓, rabbit VSMC infected with AdCMV.CD and treated with 5-FC failed to proliferate over a time course of 18 days. In contrast, cells infected with AdCMV.CD and not treated with 5-FC or cells infected with AdCMV.βGAL with or without 5-FC treatment grew at similar rates. These results are similar to what has been observed using adenovirus-mediated HSV-tk gene transfer and concomitant ganciclovir treatment4 5 6 and argue that although each alone is ineffective, the combination of CD expression and 5-FC administration results in potent inhibition of VSMC proliferation.
We next sought to assess whether the degree of inhibition of VSMC proliferation was a function of 5-FC concentration and the level of CD expression. Such relationships would be expected given that the level of production of the active agent 5-FU presumably reflects the amount of substrate (5-FC) and the level of activity of the converting enzyme (CD). Cells were infected with 25, 50, or 100 moi of AdCMV.CD or, as a control, AdCMV.βGAL (100 moi) and then exposed to varying concentrations of 5-FC (0.01 to 10 mmol/L). It has been previously demonstrated for both AdCMV.CD and other adenoviruses that infection with increasing moi leads to proportional increases in transgene expression and enzymatic activity.18 As seen in Fig 2⇓, cells infected with AdCMV.CD are sensitive to 5-FC at concentrations as low as 0.01 mmol/L. Control-infected cells, in contrast, show only a mild growth-inhibitory effect at 5-FC concentrations of ≥10 mmol/L. At any given concentration of 5-FC, a higher moi of AdCMV.CD, and therefore a higher level of CD expression, resulted in greater growth inhibition.
In Vivo Bystander Effect
One potential advantage of a prodrug-based gene therapy approach is the potential for bystander effects, in which the growth of neighboring cells is inhibited, presumably by local diffusion of the active agent. To assess whether VSMC expressing CD could influence the growth of neighboring nontransduced cells, we mixed VSMC infected with AdCMV.CD or AdCMV.βGAL in various proportions and incubated them in the presence of full growth medium supplemented with 5-FC (1 mmol/L). As seen in Fig 3⇓, consistent with an appreciable bystander effect, there is an ≈50% reduction in total cell number when as few as 5% of the cells express CD.
We next sought to determine whether the observed bystander effect required direct cell-to-cell contact. VSMC infected with AdCMV.CD were incubated for 24 hours in low serum (0.5%) medium containing 1 mmol/L 5-FC. This conditioned medium was then placed in varying dilutions on quiescent, uninfected cells. Twenty-four hours later, cells were serum-stimulated, and [3H]thymidine incorporation was assessed. As seen in Fig 4⇓, conditioned medium obtained from AdCMV.CD-infected cells efficiently inhibited the ability of uninfected VSMC to respond to serum stimulation. Conditioned medium from cells similarly infected with AdCMV.βGAL had no effect (data not shown). These results suggest that cell-to-cell contact is not required for the observed bystander effects.
Evidence for a Cytostatic Effect of AdCMV.CD
Although 5-FU can induce the regression of a variety of different tumors in vivo, the effect of this agent on VSMC is not well characterized. Exposure of cells previously infected with AdCMV.CD to 5-FC produced a marked reduction in cell numbers (see Fig 1⇑); however, the growth-inhibited cells remained attached and morphologically showed no evidence consistent with necrosis or apoptosis (Fig 5A⇓ and 5B⇓). Indeed, in VSMC exposed to growth medium supplemented with 1 mmol/L 5-FC for 5 days, a direct assessment of cell viability of ≈500 random cells revealed no significant difference in the percentage of nonviable trypan blue–positive cells (AdCMV.CD infected with 1 mmol/L 5-FC, 4.9%; AdCMV.βGAL with 1 mmol/L 5-FC, 5.6%). Assessment of cell cycle distribution did reveal, however, that 5-FC treatment of AdCMV.CD-infected cells produced a higher percentage of cells in the G1 phase of the cell cycle and considerably fewer in S phase (Fig 5E⇓ and 5F⇓), which is consistent with a cytostatic effect. The reversible nature of the effect was further demonstrated through removal of 5-FC, return to normal growth medium, and subsequently assessment of cell growth. As shown in Fig 5C⇓ and 5D⇓, cells previously growth-inhibited were capable of resuming cell division after removal of the prodrug. These results argue that the predominant mechanism of 5-FC growth inhibition of AdCMV.CD-infected cells is via a cytostatic mechanism.
In Vivo Assessment
Based on our in vitro data, the effective concentration of 5-FC needed to inhibit VSMC proliferation appeared to be in the range of 0.1 to 1.0 mmol/L (see Fig 2⇑). We therefore determined whether such a concentration could be obtained in vivo. The administration of 5-FC in humans is usually achieved through oral or parenteral administration in divided dosages with a total daily dose of 50 to 200 mg/kg. For ease of administration and because previous data suggest that bolus infusions of 5-FC were more effective than continuous administration,15 we chose to treat our experimental rabbits with a single daily infusion of 5-FC. Pharmacokinetic data of serum levels of 5-FC after a single bolus of 200 mg/kg 5-FC demonstrated that effective levels of the prodrug (>0.1 mmol/L) could be achieved for 8 to 10 hours after infusion (Fig 6⇓). This dose was therefore chosen for the subsequent in vivo experiments.
To assess the ability of AdCMV.CD to inhibit VSMC in vivo, we performed balloon injury of the rabbit femoral artery. Injured arterial segments were exposed immediately after injury to solutions containing 3×109 pfu of AdCMV.CD or AdCMV.βGAL. Analysis of the level of gene transfer suggested that in some regions of the artery exposed to virus, a significant percentage of medial cells demonstrated expression of the transgene (Fig 7⇓). In general, however, after AdCMV.βGAL infection, evidence of successful gene transfer was localized to discrete regions comprising <50% of the ≈2 cm of exposed arterial wall. Within those positive regions, gene transfer rates were high, especially to the most luminal medial smooth muscle cells (see Fig 7⇓). In contrast, other regions of the artery similarly exposed to AdCMV.βGAL showed essentially no evidence of gene transfer (data not shown). As such, the overall transfection efficiency varied significantly depending on the region of the artery from which analyzed sections were obtained. It is currently unclear what underlies the nonuniform efficiency in gene transfer, but it may relate to differences in the degree of injury of the vessel wall. The level of medial gene transfer was significantly higher than we previously reported in the rat carotid artery,2 which may reflect anatomic differences, especially in the number of internal elastic laminae between the elastic rat carotid and muscular rabbit femoral arteries.
For the purpose of our study, animals were divided into two control and one therapeutic treatment group. One control group was infected with AdCMV.βGAL and treated with 5-FC; a second control group received AdCMV.CD and saline treatment; and the final group received the presumably therapeutic combination of AdCMV.CD and 5-FC. Treatment with 5-FC was via a jugular catheter and consisted of a single daily bolus of 200 mg/kg for 7 consecutive days. Thirty days after injury, arterial segments were perfusion-fixed and harvested to assess the degree of neointimal accumulation. Quantification of the N/M by a blinded observer revealed a significant reduction in N/M of animals receiving CD and 5-FC. The magnitude of the reduction was ≈45% compared with either of the two control groups (Fig 8A⇓). Fig 8B⇓ to 8D shows histological sections representative of uninjured arteries and the degree of neointimal formation 30 days after injury from two of the three groups of animals.
Analysis of serum electrolytes and renal and hepatic function revealed no abnormalities over the course of the study. Three animals who received 5-FC (two who received AdCMV.CD, one who received AdCMV.βGAL) developed transient reversible neutropenia (lowest WBC nadir, 1.11×103/μL) at day ≈10.
The results presented here indicate that the growth of rabbit VSMC in vitro can be inhibited by the combination of CD expression and 5-FC treatment; the level of growth inhibition observed was dependent on the concentration of 5-FC and the level of CD expression. In addition, a prominent bystander effect was observed with a >50% reduction in total cell number noted when as few as 5% of the cells expressed CD. The mechanism responsible for in vitro growth inhibition was predominantly cytostatic, not cytotoxic. In vivo assessment revealed a significant reduction in the N/M of animals infected with AdCMV.CD and receiving 5-FC compared with two groups of control animals. As such, these results expand the potential gene therapy strategies to inhibit the proliferative aspect of restenosis.
In the present study, the degree of reduction of the N/M was ≈45%. These results are similar to what has been observed with a strategy that consisted of the cytotoxic combination of HSV-tk gene transfer and ganciclovir treatment in both the rat5 6 and pig4 model of vascular injury. In addition, a similar degree of inhibition has been observed with a cytostatic approach that involved an adenovirus encoding either a mutant Rb gene product or the cell-cycle regulator p21.7 8 The inability to date to achieve a reduction significantly >50% perhaps suggests that any adenovirus-mediated gene transfer approach targeting proliferation may have similar limitations. In this regard, it is interesting to note that analysis of the derivation of neointimal cells in the rat injury model has suggested that 50% of the cells arise through proliferation and 50% arise through migration.20 Because it is unknown whether any of the strategies to date affect migration of cells, the observed ceiling level of efficacy may represent the degree of neointimal thickening that occurs when proliferation is inhibited and other processes such as migration and thrombosis are unaffected. Other explanations are certainly possible as to why we observed only a 50% reduction in the N/M, including insufficient levels of initial gene transfer, rapid washout in vivo of locally produced 5-FU, or insufficient duration of 5-FC treatment.
Comparison of efficacy between HSV-tk– and CD-based approaches is difficult. Our results suggest that in rabbit cells, CD has a more pronounced bystander effect than we previously observed in rat VSMC using HSV-tk.5 Another potential advantage of the current strategy compared with HSV-tk–based approaches is that 5-FC is easily administered in oral form, whereas ganciclovir requires intravenous therapy. It is important to note that both ganciclovir and 5-FC can produce unwanted side effects. For 5-FC, the most prominent effects include gastrointestinal disorders such as diarrhea (not observed) and transient, reversible bone marrow suppression (seen in three animals).
It is currently unclear which, if any, of the various transgene strategies that involve adenovirus-mediated gene transfer will be the most useful in the treatment of human restenosis. Cytostatic gene transfer approaches previously described in which the transgene was p21 or a mutant Rb have the advantage of presumably not inducing a large inflammatory response caused by cytotoxic cell death. On the other hand, the most commonly described cytotoxic approach that involves HSV-tk and ganciclovir offers the potential advantage of bystander effects, which serve to magnify the antiproliferative effects of the strategy. This may be particularly important in human atherosclerotic lesions, in which the lesion is more complex and the level of gene transfer may be significantly reduced compared with normal arteries.21 22 23 As such, a potential advantage of the CD-based strategy over an HSV-tk–based strategy is that the approach described here is predominantly cytostatic and thus may be less likely to induce an in vivo inflammatory response while still maintaining a bystander effect. Nevertheless, at this time, any advantage of a CD-based strategy over HSV-tk, or other previously described approaches, is purely theoretical and would require a subsequent comparison in the same animal model of the two strategies to compare directly the degree of inflammation and efficacy. In addition, because both remodeling and neointimal accumulation appear to contribute to late lumen loss, the use of CD- or HSV-tk–based therapy may prove most efficacious in situations in which the proliferative aspect predominates. Such a clinical situation may be found in restenosis occurring after mechanical stenting.
In summary, we describe a new prodrug strategy using adenovirus-mediated gene transfer to inhibit VSMC proliferation in vitro and in vivo. Further studies in normal and atherosclerotic vessels will be useful to assess the relative efficacy of this strategy compared with previously described approaches. Ultimately, combination approaches that target the varying processes involved in restenosis, including proliferation, remodeling, and thrombosis, may prove to be the most fruitful.
Selected Abbreviations and Acronyms
|AdCMV.CD||=||adenovirus encoding cytosine deaminase|
|HSV-tk||=||thymidine kinase gene of the herpes simplex virus|
|moi||=||multiplicity of infection|
|VSMC||=||vascular smooth muscle cell(s)|
- Received October 7, 1996.
- Revision received January 16, 1997.
- Accepted January 22, 1997.
- Copyright © 1997 by American Heart Association
Leu MW, Roubin GS, King SB. Restenosis after coronary angioplasty: potential biologic determinants and role of intimal hyperplasia. Circulation. 1989;79:1374-1387.
Guzman RJ, Lemarchand P, Crystal RG, Epstein SE, Finkel T. Efficient and selective adenovirus-mediated gene transfer into vascular neointima. Circulation. 1993;88:2838-2848.
Lee SW, Trapnell BC, Rade JJ, Virmani R, Dichek DA. In vivo adenoviral vector-mediated gene transfer into balloon-injured rat carotid arteries. Circ Res. 1993;73:797-807.
Ohno T, Gordon D, San H, Pompili VJ, Imperiale MJ, Nabel GJ, Nabel EG. Gene therapy for vascular smooth muscle cell proliferation after arterial injury. Science. 1994;265:781-784.
Guzman RJ, Hirschowitz EA, Brody SL, Crystal RG, Epstein SE, Finkel T. In vivo suppression of injury-induced vascular smooth muscle cell accumulation using adenovirus-mediated transfer of the herpes simplex virus thymidine kinase gene. Proc Natl Acad Sci U S A. 1994;91:10732-10736.
Chang MW, Ohno T, Gordon D, Lu MM, Nabel GJ, Nabel EG, Leiden JM. Adenovirus-mediated transfer of the herpes simplex virus thymidine kinase gene inhibits vascular smooth muscle cell proliferation and neointima formation following balloon angioplasty of the rat carotid artery. Mol Med. 1995;1:172-181.
Chang MW, Barr E, Seltzer J, Jiang YQ, Nabel GJ, Nabel EG, Parmacek MS, Leiden JM. Cytostatic gene therapy for vascular proliferative disorders with a constitutively active forma of the retinoblastoma gene product. Science. 1995;267:518-522.
Chang MW, Barr E, Lu MM, Barton K, Leiden JM. Adenovirus-mediated over-expression of the cyclin/cyclin-dependent kinase inhibitor p21 inhibits vascular smooth muscle cell proliferation and neointima formation in the rat carotid artery model of balloon angioplasty. J Clin Invest. 1995;96:2260-2268.
Borrelli E, Heyman ER, Hsi M, Evans RM. Targeting of an inducible toxic phenotype in animal cells. Proc Natl Acad Sci U S A. 1988;85:7572-7576.
Mullen CA, Kilstrup M, Blaese RM. Transfer of the bacterial gene for cytosine deaminase to mammalian cells confers lethal sensitivity to 5-fluorocytosine: a negative selection system. Proc Natl Acad Sci U S A. 1992;89:33-37.
Huber BE, Austin EA, Good SS, Knick VC, Tibbels S, Richards CA. In vivo antitumor activity of 5-fluorocytosine on human colorectal carcinoma cells genetically modified to express cytosine deaminase. Cancer Res. 1993;53:4619-4626.
Huber BE, Austin EA, Richards CA, Davis ST, Good SS. Metabolism of 5-fluorocytosine to 5-fluorouracil in human colorectal tumor cells transduced with the cytosine deaminase gene: significant antitumor effects when only a small percentage of tumor cells express cytosine deaminase. Proc Natl Acad Sci U S A. 1994;91:8302-8306.
Mullen CA, Coale MM, Lowe R, Blaese MR. Tumors expressing the cytosine deaminase suicide gene can be eliminated in vivo with 5-fluorocytosine and induce protective immunity to wild type tumor. Cancer Res. 1994;54:1503-1506.
Clowes AW, Schwartz SM. Significance of quiescent smooth muscle migration in the injured rat carotid artery. Circ Res. 1985;56:139-145.
Feldman LS, Steg PG, Zheng LP, Chen D, Kearney M, McGarr SE, Barry JJ, Dediev J-F, Perriraudet M, Isner JM. Low-efficiency of percutaneous adenovirus-mediated arterial gene transfer in the atherosclerotic rabbit. J Clin Invest. 1995;96:2662-2671.
Frances BA, Mazur W, Ali NM, Geske RS, Finnigan JP, Rodgers GP, Roberts R, Raizner AF. Percutaneous transluminal in vivo gene transfer by recombinant adenovirus in normal porcine coronary arteries, atherosclerotic arteries, and the models of coronary restenosis. Circulation. 1994;90:2403-2413.