Preclinical Restenosis Models: Challenges and Successes

Abstract

Coronary artery disease remains a major problem for Western societies. The advent of percutaneous interventions, including stents has brought clinical care to a new level of efficacy, yet problems remain. Restenosis following stenting in human coronary arteries appears at last to be yielding to therapeutic strategies, especially drug eluting stents. Because therapeutic percutaneous coronary intervention is widely dominated by the intracoronary stent, restenosis therapies must include the stented coronary artery. Animal models and in particular the porcine coronary model seem to represent the human coronary artery reaction to stenting. It mimics several clinical conditions including thrombosis and neointimal formation. A key question in the era of intravascular technologies is how well this and other models can predict clinical events. This paper discusses the models and their application.

Introduction

Research in human coronary atherosclerosis is limited by an inability to control experiments and by the slow temporal lesion development. Fortunately, animal arterial injury models appear to yield comparable results to clinical trials, and can teach about the arterial response to injury. These models have become indispensable for understanding the interaction of the coronary artery with medical devices, and toward understanding neointimal genesis. They can likely function to test safety and efficacy of new devices. In these models the pathophysiologic aspects of disease can be simulated, variables can be controlled, and statistical data accrued in short time periods.

Many animal models have been used for restenosis studies. This variety comes because an ideal animal model does not exist. Each animal model has advantages and disadvantages. This chapter discusses the principal animal models described for restenosis studies, their characteristics, advantages and disadvantages compared with humans, and the considerations necessary for proximity to and ideal animal model and study design.

Preclinical Restenosis Models

Common animals models used for restenosis studies include rodents (rats, mice, rabbits), pigs, dogs, or primates. Frequent injury methods used in these models include mechanical injury (overstretch artery with noncompliant angioplasty balloons inflated to high pressures, very compliant, low-pressure balloons for denudation injury (Rogers et al., 1993), wire loops (Reidy and Schwartz, 1981; Walker et al., 1983; Lindner et al., 1991; Lindner and Collins, 1996; Lindner and Reidy, 1996) or directional atherectomy) or injury induced by agents such chemical-diet-, electrical injury (Carmeliet et al., 1997), heat (Douek et al., 1992), air desiccation, Fishman et al., 1975; Gellman et al., 1991; Sarembock et al., 1996 irradiation (Fajardo and Berthrong, 1988) or inducing severe inflammation with copper stents by foreign body implant (Ide et al., 1994; Schwartz et al., 1992a; Schwartz, 1994; Staab et al., 1997).

To enhance lesion formation or to reproduce conditions that predispose to the need for human arterial angioplasty (such as atheroma presence), before or after the “principal” injury, other authors have developed complementary injurious methods. Animals can be placed on a high-fat, high-cholesterol diet (chemical injury) or undergo other nondietary injury modes, alone or in tandem with the cholesterol diet. These create double or triple injury as models. It is unclear if such complementary injuries may positively or negatively affect final results of a study.

Rodents

Rat Carotid Artery Model

Extensive studies on the response to vascular injury were performed in the rat carotid artery model years before angioplasty became known. These studies, based on denudation injury with a very compliant, low-pressure balloon, identified the intimal layer as a key site in the proliferative response. In this model, both carotid arteries are typically used in the same animal (Figure 1). The rat carotid artery is injured either by air desiccation (Gellman et al., 1991; Sarembock et al., 1996) or by balloon endothelial denudation (Au et al., 1992; Clowes et al., 1991; Golden et al., 1990). A 2F Fogarty balloon is advanced through an incision in the external carotid artery to the common carotid artery. The balloon is inflated and drawn through the artery (while inflated) for multiple passes, generally 3 or more times. The balloon is deflated and removed, and the external carotid artery is ligated.

The Hypercholesterolemic Rabbit Iliac Model

The rabbit atherosclerotic iliac restenosis model has been used commonly. Although the lesions of this model differ from human lesions, it provides valuable insights for understanding the mechanism of repair after injury to an abnormal artery and for testing restenosis therapies (Baumbach et al., 1997; Coats et al., 1996; Hansen et al., 1988; Jenkins et al., 1989; Kalinowski et al., 2001; Kanamasa et al., 2001; Nagae et al., 2001; Welt et al., 2000; Zou et al., 2000). Rabbit models are typically single, double- or even triple-injury models, and include biochemical injury with hypercholesterolemic diets is followed by mechanical injury into both femoral arteries with a balloon catheter or sometimes air desiccation. Four to 6 weeks after injury, lesions are evaluated for stenoses. If a significant lesion is found, an angioplasty (second mechanical injury) is performed under fluoroscopic guidance.

The Dog: Minimal Response to Injury

Dogs have been explored as an experimental model for restenosis mainly because of their size, cost, and ready availability. However, dogs have high fibrinolytic activity (Mason and Read, 1971), markedly different from the human coagulation system (Kirschstein et al., 1989). In addition, the canine vessel wall produces only a thin neointima when compared with other animal models (Schwartz et al., 1994) (Figure 2). These considerations make the dog a poor model for restenosis.

Porcine Coronary Injury Model (Figure 3)

The porcine heart and its coronary artery system have a size and anatomical structure very similar to that of humans (Ali et al., 1996; Schwartz, 1998; Schwartz et al., 1993).

The carotid arteries are typically used for arterial access in this model, although the femoral arteries may also be used without difficulty. Standard guide catheters and curves for human coronary angioplasty are used in both techniques for engagement of the left main or right coronary arteries, which is a great advantage of these models.

Mechanical injury by oversizing the artery and endothelial denudation alone has proven successful, but oversizing is a stronger stimulus for smooth muscle cell proliferation than endothelial denudation alone. Oversizing the coronary artery can be achieved using a coronary angioplasty balloon (Heras et al., 1989; Schwartz et al., 1990) or oversized stent implantation (McKenna et al., 1998). This model produces a neointimal response virtually identical to human restenotic neointima in terms of cell size, cell density, and histopathologic appearance (Schwartz et al., 1990, 1992, 1993, 1994). Specimens from balloononly injury typically show a single laceration of media, and specimens from oversized stent implantation show multiple injuries in the neighborhood of the stent wires, also very similar to human findings.

Nonhuman Primates

Nonhuman primates bear phylogenetic resemblance to humans, a potentially singular advantage. The temporal sequence of proliferative response and thrombotic activity in coronary arteries is not well related to human restenosis and other animals. Their limited availability, legal restrictions, ethical concerns, and high cost, make this animal model impractical. Few studies using primate arteries to balloon catheter injury have thus been reported (Geary et al., 1995, 1996, 1998; Hanson et al., 1991; Mondy et al., 1997).

Species-Specific Arterial Response to Injury?

Not surprisingly, species have individualized molecular and cellular arterial healing mechanisms, and so vascular lesions following injury differ immensely across these species. Rats, mice and porcine carotid injury models almost never form hemodynamically significant stenoses. In hypercholes-terolemic rabbits and porcine coronary arteries, macroscopic and hemodynamically significant stenoses may develop, but it does not occur systematically.

The arterial response to injury typically occurs in 6 phases: (1) arterial damage (endothelial denudation, internal elastic lamina fracture, media injury, adventitial injury), (2) platelet aggregation and thrombus formation, (3) elastic recoil, (4) inflammation, (5) smooth muscle cell migration, proliferation and extracellular matrix production (Clowes et al., 1989), the principal responsible of the neointimal thickening (Casscells, Clowes, and Schwartz, 1990; Fingerle et al., 1989; Hanke et al., 1990), and (6) arterial remodeling.

Each of these factors contributes to restenosis following angioplasty alone (Faxon and Currier, 1995; Landzberg et al., 1997). However, following stent placement, endothelial damage, thrombosis, inflammation and intimal hyperplasia appears to be the predominant pathology (Braun-Dullaeus et al., 1998; Gershlick and Baron, 1998). No single model appears to have all component processes identical to humans. Species-specific differences in arterial healing must be considered in study design and interpretation. Failure to account for these can cause confusion, potential data misinterpretation, or errors. Restenosis pathophysiology is described elsewhere in this book. The purpose of the present section is to highlight pathophysiologic differences between species.

Arterial Damage and Injury Score

All types of arterial injury begin with endothelial denudation, and progress to deeper injury. The degree and susceptibility to deeper injury exhibits species specificity. In the rat carotid model, injury typically shows endothelial denudation, remaining intact other arterial structures such the internal elastic lamina, media, and external elastic lamina (Figure 1) (Lindner et al., 1989). This mild injury contrasts to deeper arterial injury usually observed in the rabbit iliac and porcine coronary arteries (Figures 3 and 4), where internal elastic lamina and medial dissection is similar to the endothelial and medial damage following human percutaneous coronary intervention.

The type of injury may also induce different arterial injury degree even in the same animal. For example, arterial injury in rats is different, using a wire loop were only endothelial denudation is seen. This is compared with electric injury where wide necrosis zones from intima to adventitia can be seen. This different injury could stimulate different arterial healing phases in the same animal models.

Severity of mechanical injury across animal models might account for variability in neointimal hyperplasia. A porcine coronary injury score (Figures 4 and 5) based on the integrity of the structural components of the vessel wall has resulted from such observations. This progressively relates superficial vessel wall damage (injury score of zero) to newly formed neointima that is very thin, as occurs with appropriately sized stenting. Stenoses develop progressively only when stent wires fracture the internal elastic lamina (score 1), or lacerate the media (score 2) or the external elastic lamina (score 3). It is unknown whether the elastin membranes influence the biomolecular aspects of neointima formation or if it can be regarded only as a marker for injury severity. There is evidence that the internal elastic membrane may function as a barrier for the diffusion of macromolecules from the lumen and as a base for the attachment of endothelial cells (Sims, 1989).

The injury score can be used to compare studies and quantitate the response to potential therapies (Schwartz et al., 1994, 1996b) (Huber et al., 1993). The peripheral arteries have been used similarly in examining the arterial response to injury (Yang et al., 1996; Schwartz et al., 1996a; Kullo et al., 1997).

Thrombus Formation—The Importance of the Thrombotic and Fibrinolytic Response

After arterial injury, the clotting and fibrinolytic mechanisms are activated. This response to vessel wall injury is substantially different across species (Kirschstein et al., 1989; Mason and Read, 1971; Schwartz et al., 1992b, 1993), and different amounts of mural thrombus can be seen depending the animal model used. In the rat carotid and canine models, significant fibrin-rich thrombus is rarely if ever found. Conversely, in the rabbit iliac and porcine model, macroscopic thrombus does occur and has been characterized in several reports. Primates have comparable fibrinolytic or hemostatic systems to humans. For example, baboons are prone to acute stent thrombosis within the first 3 days after the procedure, which is significantly different than the acute stent occlusion in pig arteries, which usually occurs within 6 hours (Mason and Read, 1971; Schwartz 1998; Schwartz and Holmes, 1994).

Mural thrombus provides a scaffold for medial smooth muscle cell colonization. According to this concept, the amount of mural thrombus might govern the total neointi-mal burden. This could explain, among other factors, why rat carotid and dog coronary arteries may not generate substantial neointimal volume (and macroscopic stenoses) in distinction to the rabbit and porcine models. Against this theory, is that numerous anticoagulation trials have failed to impact restenosis, including warfarin, heparin, and direct thrombin inhibitors. However, whether these regimes eliminate local thrombus formation is unknown.

Differences in mural thrombus formation after angioplasty must be present in the design and interpretation of antithrom-botic agents in various models. Animal models with greater tendency to thrombus formation (pigs, for example) can be more sensitive to antithrombotic agents than humans. It is thus possible that antithrombotic agents are effective for pigs but not for humans.

Several examples exist in the literature, such prostacyclin (Colombo et al., 2003), aspirin (Clopath, 1980) and hirudin (Abendschein et al., 1996; Buchwald et al., 1996; Gallo et al., 1998) or low molecular weight heparin (Buchwald et al., 1996), where studies performed in pigs, show some efficacy of hemostatic interventions in the reduction of restenosis that failed in humans (Bittl et al., 1995; Serruys et al., 1995; Johnson et al., 1999). However, as discussed later, different endpoints in the animal studies may cause differences from human studies. Since mural thrombus may provide a scaffold for medial smooth muscle cell colonization, the tendency to thrombus formation may be an explanation why some antiproliferative therapies demonstrate significant inhibition of neointimal hyperplasia in some animal models that do not translate to clinical trials. In summary, more or less thrombus formation must be present when choosing an animal model for antithrombotic therapy testing and in the extrapolation of the results in humans.

Inflammation

Few studies document the role of inflammation in restenosis, although it is key. In addition, inflammation resolution is also important since it may produce a scar fibrosis, and resultant negative remodeling.

In the rat carotid model of injury there is remarkably little inflammatory response to injury. Hypercholesterolemic rabbits, porcine and nonprimate models show robust inflammatory reactions to injury (Figure 5), with early mononuclear cell infiltration from the lumen into the thrombus (Schwartz et al., 1992b; Miyauchi et al., 1998). In the porcine model inflammation is positively related to neointimal thickness (Kornowski et al., 1998). Human studies of stented arteries also show acute inflammation early after implantation, especially when stenting is associated with medial injury or lipid core penetration (Komatsu et al., 1998; Farb et al., 1999; Grewe et al., 2000). Macrophage infiltration in atherectomy tissue and the activation status of blood monocytes correlate with an increased rate of restenosis (Moreno et al., 1994; Pietersma et al., 1995).

Smooth Muscle Cell Proliferation and Migration

Although rat and mouse studies initiated the concept of proliferation, such proliferation and migration from media to intima, is considered a prominent feature in all animal models. Despite the observation that SMC neointima formation causes instent restenosis, the role of cell proliferation within the neointima remains controversial.

It is uncertain if there is a species-specific cell proliferation and migration. Cell proliferation and migration in rats mice and pigs begins early after denudation (1 or 2 days) and proceeds for the following 14 to 30 days (peaking in 2–3 weeks); (Fingerle et al., 1989; Schwartz, 1992b; Zempo et al., 1996). The Rabbit iliac model shows proliferation over the same period with peak at 8 days (Stadius et al., 1994). In non-human primates, proliferation is increased at 4 and 7 days but later declines to control rates (Geary et al., 1996). Regardless of the fact that animal models show a hyperplastic response to injury, kinetics of cell proliferation in human vessels does not appear well defined. Human lesions show a comparative hypocellular response with abundant matrix.

Elastic Recoil and Remodeling

Acute elastic recoil immediately following balloon deflation and late vascular constriction (negative remodeling); (Lafont et al., 1995; Mintz et al., 1996; Bauters and Isner, 1997; Schwartz et al., 1998) occur with PTCA alone, and are important aspects of restenosis pathophysiology in the prestent era. Coronary stents acting as a mechanical scaffold within the vessel, eliminating elastic recoil and vessel contracture (Kay et al., 2000; Shah et al., 2002). Hypercholesterolemic rabbit (Kakuta et al., 1998; Kalef-Ezra et al., 2002), porcine coronary artery (Waksman et al., 1997) and nonhuman primates (Coats et al., 1997) exhibit remodeling behavior in much the same fashion as man (Schwartz et al., 1998). Mouse arteries tend to enlarge after angioplasty and appear naturally prone to positive remodeling (Carmeliet et al., 1997; de Smet et al., 1998).

Regardless that remodelling is lost in the stent era, animal models prone to positive or negative remodeling must be a consideration today since positive vascular remodeling occurs after baremetal stent implantation (Shah et al., 2002), after catheter-based radiation followed by conventional stent implantation (Kay et al., 2000) and but not demonstrated, after drug-eluting stent implantation.

The ideal experimental model to assess restenosis treatment would be one that reliably predicts the risks and outcome of human clinical trials. Although such an ideal model does not yet exist, experimental studies are ongoing. Many pharmacologic agents, such as antiplatelet, anticoagulants, ACE inhibitors and antiproliferative drugs have been tested successfully in animal models failed in human clinical trials. The marked disparity of results between animal model research and clinical trials has led to skepticism about the validity of animal models in restenosis research.

The failure of animal studies to predict efficacy in preventing human restenosis is potentially attributable to two general factors. Species differences may in part be responsible, a factor not easily modified except with transgenic animals. Second there are several modifiable factors, not taken into account, which are able to approach the currently models to the ideal animal model.

Unknowns in Models

The importance of several biologic factors variability between species in restenosis or in the positive or negative predictive values are unknown. The impact of concomitant atherosclerosis in animal models is not well defined. Restenosis can be studied on either normal or previously injured arteries. The most commonly employed technique is the arterial injury over a normal coronary artery. The normal coronary artery of a young rat, rabbit, or pig differs distinctly from the atherosclerotic coronary artery of an older patient. Arteries of these animal models, even those of with hyperlipidemic diets (developing during a few weeks instead of decades as in humans), do not show densely fibrous and acellular plaques with ulceration, calcification, thrombosis, or hemorrhage into the vessel wall, all features of human restenosis. The impact of this atherosclerotic environment on restenosis and whether the use of models that produce atherosclerosis will have advantages over nonatherosclerotic models is also unknown.

Another consideration is the impact of protective molecules against atherosclerosis in restenoisis models such cholesteryl ester transferase or the high levels of HDL and lows levels of LDL in mice. Cholesteryl ester transferase is present in humans, swine, and rabbits and is deficient in dogs and rodents. This enzyme explains in part the difficulty in inducing atheroma lesions in these latter animals (Tall, 1986; Narayanaswamy et al., 2000).

Conclusions—The Ideal Model

Many “catastrophic” articles in the last decade have been written about the failure of animal models in predict results in human trials. These have generated distrust in the current animal models, but little true data exist about the true place for these models and their relation to humans. There is no doubt that animal models have significantly advanced our understanding of the mechanisms of restenosis formation and have served to improve the therapeutic options.

For the moment, a single ideal global model does not exist, but promising research is ongoing in this area. At present, compromises in choosing an animal model are inevitable. A detailed species understanding for limitations and strength features must be considered for the specific purpose of the study (for example thrombosis vs. migration), for the results interpretation and for human extrapolation.

Taking into account all data commented in the present chapter, the “ideal model” should instead be considered of an “ideal study.” A rational approach to the ideal study is important:

1. Arterial response to injury (pathogenesis) studies: Multiple animal models are valid. Since this model serves to create hypothesis and subsequently new treatment strategies, minimal arterial changes may be sufficient. New hypothesis and treatment strategies should be confirmed in animal models that demonstrate good predictive values, such as the coronary porcine model.

2. Safety studies: prior comments about artery type, devices, dosage and drug timing, animal pathophysiologic differences must be considered. More close alignment with human stenting is strongly recommended. For this purpose, not all animal models would be valid.

3. Efficacy studies: The fundamental parameter besides safety studies is the amount of neointimal hyperplasia. Models with more neointima are typically the best models. It is of paramount importance to assess histologic information along the same human endpoints such as angiographic restenosis or IVUS. All results should be regularly reported and carefully evaluated.

To date, considering all models, the porcine overstretch coronary artery appears closest to humans for global use since its coronary anatomy, physiology and pathophysiology. In addition, it permits using of the same devices used in humans and produces the thickest neointima in response to injury.

Animal models will continue to provide more complete understanding of restenosis and to find the improved therapies for human restenosis. However, efforts for developing improved animal models must continue. New animal models which are now under investigation may provide additional insight into new and important aspects of animal models that could predict the success of therapeutic interventions in animals and ultimately in humans.