Physiological Reviews

Role of Oxidative Modifications in Atherosclerosis

Roland Stocker, John F. Keaney Jr.


This review focuses on the role of oxidative processes in atherosclerosis and its resultant cardiovascular events. There is now a consensus that atherosclerosis represents a state of heightened oxidative stress characterized by lipid and protein oxidation in the vascular wall. The oxidative modification hypothesis of atherosclerosis predicts that low-density lipoprotein (LDL) oxidation is an early event in atherosclerosis and that oxidized LDL contributes to atherogenesis. In support of this hypothesis, oxidized LDL can support foam cell formation in vitro, the lipid in human lesions is substantially oxidized, there is evidence for the presence of oxidized LDL in vivo, oxidized LDL has a number of potentially proatherogenic activities, and several structurally unrelated antioxidants inhibit atherosclerosis in animals. An emerging consensus also underscores the importance in vascular disease of oxidative events in addition to LDL oxidation. These include the production of reactive oxygen and nitrogen species by vascular cells, as well as oxidative modifications contributing to important clinical manifestations of coronary artery disease such as endothelial dysfunction and plaque disruption. Despite these abundant data however, fundamental problems remain with implicating oxidative modification as a (requisite) pathophysiologically important cause for atherosclerosis. These include the poor performance of antioxidant strategies in limiting either atherosclerosis or cardiovascular events from atherosclerosis, and observations in animals that suggest dissociation between atherosclerosis and lipoprotein oxidation. Indeed, it remains to be established that oxidative events are a cause rather than an injurious response to atherogenesis. In this context, inflammation needs to be considered as a primary process of atherosclerosis, and oxidative stress as a secondary event. To address this issue, we have proposed an “oxidative response to inflammation” model as a means of reconciling the response-to-injury and oxidative modification hypotheses of atherosclerosis.


A. Atherosclerosis and Its Relationship to Coronary Artery Disease

Atherosclerosis is the major source of morbidity and mortality in the developed world. The magnitude of this problem is profound, as atherosclerosis claims more lives than all types of cancer combined and the economic costs are considerable. Although currently a problem of the developed world, the World Health Organization predicts that global economic prosperity could lead to an epidemic of atherosclerosis as developing countries acquire Western habits.

Atherosclerosis is characterized by the accumulation of cholesterol deposits in macrophages in large- and medium-sized arteries. This deposition leads to a proliferation of certain cell types within the arterial wall that gradually impinge on the vessel lumen and impede blood flow. This process may be quite insidious lasting for decades until an atherosclerotic lesion, through physical forces from blood flow, becomes disrupted and deep arterial wall components are exposed to flowing blood, leading to thrombosis and compromised oxygen supply to target organs such as the heart and brain. The loss of heart and brain function as a result of reduced blood flow is termed heart attack and stroke, respectively, and these two clinical manifestations of atherosclerosis are often referred to as coronary artery disease and cerebrovascular disease. To simplify any discussion of the underlying pathology in these clinical syndromes, coronary artery disease and cerebrovascular disease are commonly referred to by the collective term cardiovascular disease. With respect to the underlying pathology of atherosclerosis, there are a number of environmental and genetic “cardiovascular risk” factors that warrant consideration.

1. Epidemiology and risk factors

Over the last 40 years, a number of clinical and laboratory variables have proven predictive of the incidence of cardiovascular disease and thus qualify as cardiovascular disease risk factors. In the following, we will restrict our discussion to those traits that are strongly and consistently associated with cardiovascular disease in a manner that is independent of other traits.


Although it is not subject to modification, age is among the most important risk factors for predicting incident cardiovascular disease. This concept is, perhaps, best illustrated if one considers the risk of developing cardiovascular disease over a 10-year period. Based on experience in the United States, the average risk of developing cardiovascular disease for a 30- to 34-year-old male is ∼3% (1040). This number rises some sevenfold to 21% for a comparable individual aged 60–64 yr. Prediction of coronary heart disease uses risk factor categories (1040). The exact magnitude of age-related risk compared with other cardiovascular disease risk factors is illustrated by work from the Framingham Heart Study that has resulted in a 14-point scoring system to predict incident 10-yr cardiovascular disease. In this system, increasing risk is characterized by a higher score, and up to 7 points can be attributed to age alone. Thus age is an overriding risk factor for incident cardiovascular disease.


Numerous observational studies have indicated that males exhibit excess risk for cardiovascular disease compared with age-matched women (40). There has been considerable speculation that estrogens offer a “protective” effect to women, as cardiovascular disease accelerates in women after menopause. However, this speculation has been difficult to substantiate, as the treatment with estrogen has not reduced the incidence of cardiovascular disease of postmenopausal women (419). Alternatively, some of this apparent protection could be due to the fact that women exhibit relatively higher concentrations of high-density lipoprotein (HDL) cholesterol than do age-matched men. Nevertheless, incident cardiovascular disease is less common in premenopausal women than their age-matched male counterparts.


There is now a growing appreciation that obesity, defined as an excess body weight with an abnormal high preponderance of body fat, is a condition that increases the incident risk of cardiovascular disease. The exact mechanism(s) to explain this phenomenon, however, are controversial. A number of other risk factors for cardiovascular disease, such as hypertension, low HDL cholesterol, and diabetes mellitus, often coexist with obesity (1041). This relation between obesity and cardiovascular disease has become of considerable concern as the prevalence of obesity in the developed world is increasing at an alarming rate.


The notion that cigarette smoking is linked to heart disease dates back to a series of studies that unequivocally linked smoking and the incidence of myocardial infarction (208, 231, 355). More recently, the Surgeon General's report estimates that smoking increases atherosclerotic disease by ∼50% and doubles the incidence of coronary artery disease (962a). There is now considerable confidence that smoking is causally related to coronary artery disease, as smoking cessation is quite effective in lowering the future risk of the disease. In fact, the risk of heart attack in ex-smokers approaches that of nonsmokers in only 2 years (296).


Hypertension is defined as a systolic blood pressure in excess of 140 mmHg or a diastolic blood pressure above 90 mmHg (734). The current estimates indicate that the elderly are particularly predisposed to hypertension, with up to 75% of people over 75 years of age qualifying for this diagnosis (734). There appears to be an approximately linear relation between blood pressure elevation and the increased incidence of atherosclerotic vascular disease (573). The causal nature of this association is supported by numerous studies demonstrating that both heart attack and stroke are significantly reduced in hypertensive patients with the institution of antihypertensive therapy (380).


Approximately 17 million people in the United States, or 6.2% of the population, carry the diagnosis of diabetes mellitus (165). In patients with diabetes, the risk of coronary atherosclerosis is three- to fivefold greater than in nondiabetics despite controlling for other risk factors (60, 739). A number of other known risk factors for coronary disease such as hypertension and abnormal lipids are also more common in diabetics than the general population (60), but despite this association, no more than 25% of the excess coronary atherosclerosis risk from diabetes can be attributed to these known risk factors (670). Thus diabetes represents a major contributing factor to atherosclerosis.


The association between LDL cholesterol and atherosclerosis has been established based, in part, on an experiment of nature. Familial hypercholesterolemia is an autosomal dominant disorder that affects ∼1 in 500 persons from the general population. Heterozygotes for this disease manifest a two- to fivefold elevation in plasma LDL cholesterol that is due to a functional impairment of the LDL receptor, resulting in a defect in LDL clearance. Homozygotes for this disorder demonstrate a four- to sixfold elevation in plasma cholesterol that produces precocious atherosclerosis. In heterozygotes, 85% of individuals have experienced a myocardial infarction by the age of 60, and this age is reduced to 15 yr in patients homozygous for the disease (325). In the general population, the cardiovascular disease risk from increased LDL cholesterol is supported by observations that cholesterol-lowering therapy greatly diminishes the clinical manifestations of atherosclerosis, particularly since the advent of inhibitors of 3-hydroxy-3-methylglutaryl coenzyme A reductase (i.e., statins) that profoundly lower LDL cholesterol (326).

In contrast to the situation with LDL cholesterol, the relation between HDL cholesterol and atherosclerosis is an inverse one (320). The causal nature of this association is also supported by an experiment of nature, Tangier disease (682). This autosomal codominant condition is characterized by the essential absence of HDL cholesterol levels due to a defect in the ATP binding cassette transporter-1 (69, 584, 790) that impairs cholesterol efflux from cells (525). Tangier patients demonstrate a tissue cholesterol-loading syndrome, characterized by large tonsils, neuropathy, and premature coronary artery disease in some kindreds (682). Thus considerable evidence supports the inverse relation between coronary artery disease and serum levels of HDL cholesterol.

2. Morphological features of atherosclerosis


The arterial wall normally consists of three well-defined concentric layers that surround the arterial lumen, each of which has a distinctive composition of cells and extracellular matrix. The layer immediately adjacent to the lumen is called the intima, the middle layer is known as the media, and the outermost layer comprises the arterial adventitia. These three layers are demarcated by concentric layers of elastin, known as the internal elastic lamina that separates the intima from the media, and the external elastic lamina that separates the media from the adventitia.

A single contiguous layer of endothelial cells lines the luminal surface of arteries. These cells sit on a basement membrane of extracellular matrix and proteoglycans that is bordered by the internal elastic lamina. Although smooth muscle cells are occasionally found in the intima, endothelial cells are the principal cellular component of this anatomic layer and form a physical and functional barrier between flowing blood and the stroma of the arterial wall. Endothelial cells regulate a wide array of processes including thrombosis, vascular tone, and leukocyte trafficking among others.

Progressing outwards from the internal elastic lamina, the media consists principally of smooth muscle cells arranged in layers, the number of layers depending on the arterial size. An extracellular matrix consisting largely of elastic fibers and collagen with a lesser content of proteoglycan holds the smooth muscle cells together. An increasing content of elastin is typical of larger arteries that need to provide for considerable elastic recoil during diastole, the time period between ejections of blood from the heart.

The adventitia is the outermost layer of the artery and typically consists of a loose matrix of elastin, smooth muscle cells, fibroblasts, and collagen. Most of the neural input into blood vessels also traverses through the adventitia. At one time, the adventitia was considered inactive with respect to vascular homeostasis; however, recently it has become clear that the adventitia, through the production of reactive oxygen species (ROS), may play an important role in controlling vascular remodeling and nitric oxide (NO) bioactivity (766).


Atherosclerosis manifests itself histological as arterial lesions known as plaques that have been extensively characterized (878880) into six major types of lesions that reflect the early, developing, and mature stages of the disease (Fig. 1). In lesion-prone arterial sites, adaptive thickening of the intima is among the earliest histological changes. As macrophages accumulate lipid, type II lesions form as nodular areas of lipid deposition that are also known as “fatty streaks,” and these represent lipid-filled macrophages (i.e., foam cells). Continued foam cell formation and macrophage necrosis can produce type III lesions that contain small extracellular pools of lipid. Types II and III lesions are readily apparent through the use of fat-soluble dyes that stain cholesterol esters accumulated in macrophages and the extracellular space. These early lesions are often evident by age 10 (877) and can occupy as much as one-third of the aortic surface by the third decade. Developing lesions represent the next two types of lesions and, as shown in Figure 1, are characterized by significant areas of extracellular lipid that represents the “core” of the atherosclerotic lesion. Type IV lesions are defined by a relatively thin tissue separation of the lipid core from the arterial lumen, whereas type V lesions exhibit fibrous thickening of this structure, also known as the lesion “cap.” These type IV and V lesions can be found initially in areas of the coronary arteries, abdominal aorta, and some aspects of the carotid arteries in the third to fourth decade of life.

FIG. 1.

Varying stages of atherosclerosis as outlined by Stary et al. (879). The progression of atherosclerosis is depicted from the earliest stages (top left) to the most advanced (bottom right) culminating in plaque rupture and associated thrombosis. [From Stary et al. (879).]

Mature type VI lesions exhibit architecture that is more complicated and characterized by calcified fibrous areas with visible ulceration. These types of lesions are often associated with symptoms or arterial embolization. It was once thought that end-organ damage and infarction were due to gradual advancement of these lesions, but we now know the processes involved in precipitating heart attack and stroke are considerably more complex. As a consequence, we direct our attention to the histology of an atherosclerotic plaque.


The light microscopic appearance of a prototypical atherosclerotic plaque is depicted in Figure 2. Plaques contain a central lipid core that is most often hypocellular and may even include crystals of cholesterol that have formed in the aftermath of necrotic foam cells. In this late stage of lesion development, residual foam cells may be difficult to see but have often left the core with an abundant quantity of tissue factor (1033), an important activator of the clotting cascade. This lipid core is separated from the arterial lumen by a fibrous cap and myeloproliferative tissue that consists of extracellular matrix and smooth muscle cells. The junction between the cap and the morphologically more normal area of artery is known as the “shoulder” region of the atherosclerotic plaque. This area is typically more cellular than other areas of the plaque and may contain a variable composition of smooth muscle cells, macrophages, and even T cells. The shoulder region is most prone to rupture and, as shown in Figure 2, may even contain evidence of previously healed fissures.

FIG. 2.

Light micrograph of a fibrofatty plaque in the coronary artery. The lumen is eccentric and separated from the fibrous cap (FC) by myeloproliferative (MP) tissue. Under the fibrous cap are calcified (C) areas of the plaque indicative of advanced lesions. The fibrous cap is separated from the more normal area of the artery by the “shoulder” region that, in this plaque, demonstrates evidence of a healed fissure (F). The adventitia (A) is the outermost area of the artery, and its border with the media is defined by the external elastic lamina (*). [From Gravanis MB. Histopathology of atherosclerosis. In: Atlas of Atherosclerosis: Risk Factors and Treatment, edited by Wilson PWF. Philadelphia, PA: Current Medicine, 2002.]


Early concepts of atherosclerosis involved progressive luminal narrowing until the blood flow was compromised to the point that organ metabolic needs could no longer be met, producing ischemia and infarction of the subtended tissue such as the heart or the brain. Over the last 15 years, this concept has changed dramatically to include the notion of plaque rupture as both a precipitant of clinical events but also a component of plaque progression in atherosclerosis.

Using perfusion fixation techniques, Davies and Thomas (188) published a seminal study demonstrating that acute myocardial infarction and crescendo angina, two cardinal manifestations of atherosclerosis, were associated with atherosclerotic plaque rupture and fissuring in the artery with compromised blood flow. These observations suggested that clinical events were the consequence of an abrupt, catastrophic change in plaque morphology rather than a gradual narrowing of the lumen. Evidence for plaque rupture can also be found in patients dying from noncardiac causes (186), suggesting that plaque rupture is part of atherosclerotic lesion progression rather than a unique feature of clinical events from atherosclerosis.

Given that plaque rupture is implicated as a precipitating event in the clinical manifestations of atherosclerosis, a considerable effort has been directed at understanding the events involved in this process. Mature atherosclerotic plaques can be categorized as either stable or vulnerable to rupture. Stable plaques tend to be characterized by a smaller lipid core, a thick fibrous cap, and shoulder regions with few inflammatory cells, whereas vulnerable plaques contain considerable lipid in their core, a thin fibrous cap, and a robust population of macrophages and T cells in their shoulder regions. These differences in morphology suggest that vulnerable plaques may be weaker structurally and more likely to rupture in response to the physical forces of flowing blood. This contention is supported by experimental data linking an increased content of macrophages in lesions to structural weakness (539).

In summary, atherosclerosis is a major source of morbidity and mortality in the developed world that is characterized by LDL deposition in the arterial wall, a process that is stimulated by environmental and genetic factors such as tobacco use, diabetes, and hypertension. This LDL deposition occurs primarily within macrophages and ultimately begets the formation of well-defined lesions in the arterial intima. Such lesions then develop and are prone to rupture and, as a consequence, can precipitate the clinical events such as heart attack and stroke. Given the public health implications of this disease, it is not surprising that considerable effort has been devoted to understanding the molecular mechanisms of atherosclerosis and the factors that predispose individuals to clinical events. Thus we now turn our attention to contemporary theories of atherogenesis.

B. Hypotheses of Atherogenesis

Over the past 150 years, there have been numerous efforts to explain the complex events associated with the development of atherosclerosis. In this endeavor, three distinct hypotheses have emerged that are currently under active investigation. These hypotheses of atherosclerosis are not mutually exclusive but rather emphasize different concepts as the necessary and sufficient events to support the development of atherosclerotic lesions. For the purposes of this review, we refer to these paradigms as follows: 1) the response-to-injury, 2) the response-to-retention, and 3) oxidative modification.

1. The response-to-injury hypothesis

Early hypotheses of atherosclerosis included the “incrustation” hypothesis of Rokitansky (778), suggesting that intimal thickening was due to arterial fibrin deposition, and the “lipid transudation” hypothesis of Virchow (995) according to which lipid complexes with mucopolysaccharides caused atherosclerosis. A common theme to these early hypotheses was their dependence on passive deposition rather than an active cellular component. Ross and Glomset (782) offered a radical departure from this thinking by proposing the “response-to-injury” hypothesis of atherosclerosis. In this hypothesis, the proposed initial step in atherogenesis is endothelial denudation leading to a number of compensatory responses that alter the normal vascular homeostatic properties. For example, injury enhances endothelial adhesiveness for leukocytes and platelets and alters the local vascular anticoagulant milieu to a procoagulant one. Recruited leukocytes and platelets then release cytokines, vasoactive agents, and growth factors that promote an inflammatory response that is characterized by migration of smooth muscle cells into the intima and their proliferation to form an intermediate lesion.

Another component of the inflammatory response is the recruitment of macrophages into the arterial wall. These macrophages take up deposited LDL lipid to form lipid-laden “foam cells,” the hallmark of an early atherosclerotic lesion. The process of lipid accumulation and foam cell formation perpetuates an inflammatory response that perpetuates macrophage and lymphocyte recruitment (454, 981). Continued inflammation allows for cellular necrosis, with a concomitant release of cytokines, growth factors, and proteolytic enzymes that sets the stage for autocatalytic expansion of the lesion to form a space-occupying collection in the intima not unlike an abscess that would form in other tissues. As the lesion enlarges it begins to encroach upon the lumen and, ultimately, blood flow is impaired.

This response-to-injury hypothesis was originally based on the notion of endothelium desquamation as a principal event initiating atherosclerosis (783, 784). More recently, it has become clear that endothelial desquamation is not common and that an intact endothelial cell layer covers developing atherosclerotic lesions. These facts, among others, promoted refinement of the initial hypothesis such that endothelial dysfunction is sufficient to initiate atherogenesis through increased endothelial permeability to atherogenic lipoproteins (781) (Fig. 3). This contention is not without its problems, however, as even normal artery segments exhibit rates of LDL entry that exceed the rate of LDL accumulation (124), suggesting that atherogenic lipoprotein entry into the arterial wall may not depend on endothelial dysfunction. In fact, the rate of LDL entry into the arterial wall is rather uniform, but the accumulation of atherogenic lipoproteins is concentrated in areas that are predisposed to future lesion development (820, 821). Such lesion-prone sites tend rather to demonstrate an enhanced retention of atherogenic apolipoprotein B-containing lipoproteins (238, 820, 821). Such observations have prompted alternative hypotheses for the initiation of atherosclerosis.

FIG. 3.

Response-to-injury hypothesis of atherosclerosis as proposed by Ross (781). In this hypothesis atherosclerosis begins with endothelial injury or dysfunction (A) that is characterized by enhanced endothelial permeability and low-density lipoprotein (LDL) deposition in the subendothelial space. This is followed by leukocyte adhesion and transmigration across the endothelium. In intermediate stages (B), atherosclerosis is characterized by foam cell formation and an inflammatory response including T-cell activation, the adherence and aggregation of platelets, and further entry of leukocytes into the arterial wall along with migration of smooth muscle cells into the intima. Finally, advanced atherosclerosis (C) is characterized by continued macrophage accumulation, fibrous cap formation, and necrosis in the core of the lesion. [From Ross (781), copyright 1999 Massachusetts Medical Society.]

2. The response-to-retention hypothesis

This hypothesis submits that the lipoprotein retention is the inciting event for atherosclerosis (Fig. 4). Within 2 h of injecting LDL into rabbits, arterial retention of LDL and its microaggregates can be observed (667). The underlying mechanisms involved in this process are just now coming to light. It is estimated that ∼85% of subendothelial lipoprotein delivery is the result of transcytosis, and this process is restricted to particles <70 nm in diameter (852). This size restriction is important as it suggests that lipoprotein lipase activity is needed for triacylgycerol-rich lipoproteins to reach the subendothelial space (1110). The retention of lipoproteins within the arterial wall, however, appears tightly linked to components of the extracellular matrix. Apolipoprotein B-100, the single protein associated with LDL, is retained within the arterial wall in close association with arterial proteoglycans (118, 1091). This interaction is mediated by specific residues (3359–3369) (71) that, when mutated, protect experimental animals against the development of atherosclerosis (71, 857). Apolipoprotein B-48, which appears to be equally atherogenic as apolipoprotein B-100 in mice (990), also binds avidly to proteoglycans, and this interaction is mediated by residues 84–94 of apolipoprotein B-48 (250). Thus these data support an important role for proteoglycan binding in the retention of apolipoprotein B-containing lipoproteins in the early stages of atherosclerosis.

FIG. 4.

Response-to-retention hypothesis of atherosclerosis. According to the original hypothesis (1036), mild to moderate hyperlipidemia causes lesion development only in specific sites within the arterial tree characterized by local synthesis of apolipoprotein B-retentive molecules such as biglycan and decorin. The cartoon shows the initial stages of arterial lipoprotein delivery, retention, and efflux (1–5). Accumulation (5) is thought to result from both apolipoprotein B-100 motifs that mediate proteoglycan binding and arterial factors such as secretory sphingomyelinase that facilitate lipoprotein aggregation (reviewed in Ref. 1037). The accumulation of apolipoprotein B-100-containing lipoproteins within the arterial wall is thought to further trigger a proinflammatory cascade (6–13). Lipoprotein oxidation (not shown) may or may not be part of these responses. Similar to LDL, apolipoprotein B-48-containing chylomicron remnants may also bind to proteoglycans via residues 84–94 of the apolipoprotein (250) and hence be retained within the arterial wall (not shown). [Modified from Proctor et al. (736).]

In addition to proteoglycan binding, lipolytic and lysosomal enzymes in the extracellular matrix also appear to play a role. For example, lipoprotein lipase enhances the adherence of LDL in vitro (1035), and this effect is independent of enzymatic activity (1034). Once retained within the arterial wall, LDL can form microaggregates (667, 931), perhaps through the action of secretory sphingomyelinase (1074), an enzyme that also generates ceramides that mediate apoptosis and mitogenesis (357, 455), as well as lysosomal enzymes such as cathepsin D and lysosomal acid lipase (350). Most importantly, aggregated LDL is avidly taken up by macrophages and smooth muscle cells (440) and thus can support foam cell formation (991). Thus many features of atherosclerosis can be attributed to enhanced retention of LDL within the arterial wall and its association with proteoglycans.

3. The oxidative modification hypothesis

The oxidative modification hypothesis, discussed in more detail in section iii, focuses on the concept that LDL in its native state is not atherogenic. However, LDL modified chemically is readily internalized by macrophages through a so-called “scavenger receptor” pathway (317). Exposure to vascular cells in medium that contains transition metals also results in modification of LDL such that it serves as a ligand for the scavenger receptor pathway (395). It is now clear that one mechanism whereby cells in vitro render LDL a substrate for the scavenger receptor pathway is via oxidation of LDL lipids and the resulting modification of apolipoprotein B-100 (886). These observations form the basis for the oxidative modification hypothesis of atherosclerosis (Fig. 5), in which LDL traverses the subendothelial space of lesion-prone arterial sites. During this process, LDL lipids are subject to oxidation and, as a consequence, apolipoprotein B-100 lysine groups are modified so that the net negative charge of the lipoprotein particle increases (347). This modification of apolipoprotein B-100 renders LDL susceptible to macrophage uptake via a number of scavenger receptor pathways producing cholesterol ester-laden foam cells (349). It is this accumulation of foam cells that forms the nidus of a developing atherosclerotic lesion.

FIG. 5.

Oxidative modification hypothesis of atherosclerosis. LDL becomes entrapped in the subendothelial space where it is subject to oxidative modification by resident vascular cells such as smooth muscle cells, endothelial cells, and macrophages. Oxidized LDL stimulates monocyte chemotaxis (A), prevents monocyte egress (B), and supports foam cell formation (C). Once formed, oxidized LDL also results in endothelial dysfunction and injury (D), and foam cells become necrotic due to the accumulation of oxidized LDL (E). (From Diaz M, Frei B, Vita JA, and Keaney JF Jr. Antioxidants and atherosclerotic heart disease. N Engl J Med 337: 408–416, 1997, copyright Massachusetts Medical Society.)

The process of LDL oxidation is associated with a number of other potentially proatherogenic events. For example, during the initial stages of in vitro LDL oxidation, modification of LDL lipids can occur in the absence of any changes to apolipoprotein B-100. Such modified LDL has been termed “minimally modified LDL” and shown in vitro to induce the synthesis of monocyte chemotactic protein-1 in both smooth muscle and endothelial cells (171, 754), resulting in the recruitment of inflammatory cells (659). This particular step appears critical as mice lacking the receptor for monocyte chemotactic protein-1 are resistant to atherosclerosis (72, 323). More heavily in vitro oxidized LDL, commonly termed “ox-LDL,” is chemotactic for monocytes (742) and T lymphocytes (611), perhaps as the result of lysophosphatidylcholine formed during oxidation (886). Oxidized LDL has also been shown to stimulate the proliferation of smooth muscle cells (890) and to be immunogenic by eliciting the production of autoantibodies (710, 795) and the formation of immune complexes that can also facilitate macrophage internalization of LDL (334, 492). The recruitment of inflammatory cells may result in the continued oxidation of LDL, setting the stage for catalytic expansion of the atherosclerotic lesion and the full-blown spectrum of atherosclerosis.

In summary, the aforementioned hypotheses of atherosclerosis have each attempted to explain the complex cellular events of atherosclerosis along a common theme. The response-to-injury hypothesis focuses on vascular injury as the inciting event in atherosclerosis with the entire spectrum of the disease representing an attempt to heal an ongoing vascular insult. In contrast, the response-to-retention hypothesis uses lipoprotein-matrix interactions as the critical event in early atherosclerosis, whereas the oxidative modification hypothesis requires oxidation of LDL lipids. Although each hypothesis points to its own critical initiating event, there are many common features among these competing hypotheses. For example, each involves a significant component of inflammation, a known feature of atherosclerosis (556). Each hypothesis also includes LDL as a central element, an important point as reduction in LDL cholesterol is among the most effective means of treating atherosclerosis. Perhaps the most unique, however, is the oxidative modification hypothesis as it alone proposes a particular importance of oxidative events and redox reactions in the genesis of vascular disease. The concept of oxidative events in atherosclerosis has changed considerably since the early days of the oxidative modification hypothesis, and it is this subject that will form the basis for the remainder of this review.


A. Oxidative Stress: A Definition

The notion of “oxidative stress” in biological systems goes back to the early period of research on oxygen activation with an initial focus on oxygen toxicity and X-irradiation (300). Much of the relevant early literature was reviewed in 1979 by Chance et al. (143) in an article on hydroperoxide metabolism in mammalian organs. In the following, the concept of oxidative stress was developed primarily by Sies (847849), with synonymous terms such as “oxidant stress” and “pro-oxidant stress,” or the related term “reductive stress” receiving comparatively less emphasis.

Sies described oxidative stress as a “disturbance in the pro-oxidant/antioxidant balance in favor of the former” (847). This original denotation has been modified since to the more refined definition of “imbalance between oxidants and antioxidants in favor of the oxidants, potentially leading to damage” (848). This more careful definition accounts for some important operational considerations. For example, an oxidative challenge or a loss of antioxidants alone does not constitute oxidative stress. However, if increased formation of oxidant(s) is accompanied by a loss of antioxidant(s) and/or accumulation of oxidized forms of the antioxidant(s), oxidative stress is approached. The refined definition also conceptually distinguishes oxidative stress from oxidative damage. Thus even a severe oxidative assault that is accompanied by a loss of antioxidants may not necessarily result in oxidative damage (848). For example, biological systems are characterized by adaptive responses that may compensate, and perhaps even overcompensate, the oxidative stress that may manifest itself as a situation of increased redox environment (898).

The refined definition of oxidative stress and its underlying redox chemistry involving reduction-oxidation reactions implies that any form of “tipping the balance” causes an “imbalance.” This has led to the concept of reductive stress to describe a situation where the balance is altered in favor of reductants (1027). Reductive stress can be intimately linked to oxidative stress. For example, an overproduction of reducing equivalents such as NAD(P)H may result in increased redox cycling of substances that can undergo repetitive rounds of oxidation/reduction, ultimately leading to the increased generation of superoxide anion radical (O2·) and secondary oxidants. This has been implicated in the formation of ROS by hypoxia-like metabolic imbalances (951).

With the increased appreciation of interplay between ROS and reactive nitrogen species (RNS), including their responses in cells, the term nitrosative stress has also been introduced (364). Nitrosative stress, defined as increase in S-nitrosated compounds associated with a decrease in intracellular thiols, may be associated with a number of biological responses, some of which are of particular interest to vascular physiology and pathophysiology (1044, 1112).

B. Oxidants and Markers of Oxidant Events

1. Free radicals or 1e-oxidants

A free radical can be defined as any species capable of independent existence that contains one or more unpaired electrons (353). In biological systems, a variety of radicals can be generated (Table 1) with their reactivity depending on their nature and the molecule(s) encountered. If two radicals meet, they can join their unpaired electrons to form a covalent bond in reactions that are often kinetically fast and that lead to nonradical products. An example relevant to the vessel wall is the very fast reaction of O2· with ·NO to form peroxynitrite (ONOO) (reaction 1): Math(1)

View this table:

Examples of free radicals in biological systems

Alternatively, a radical may add to a nonradical molecule or abstract a hydrogen atom from a C-H, O-H, or S-H bond of nonradical molecules. These types of radical reactions are common in biological systems where most molecules are nonradical species. The molecules potentially affected include low-molecular-weight compounds like antioxidants and cofactors of enzymes, lipids, proteins, nucleic acids, and sugars. In this case, a new radical is generated, and this can set up a chain reaction.

A typical example of such a chain reaction is the process of lipid peroxidation that may be initiated by, for example, a hydroxyl radical (·OH) abstracting a hydrogen atom from a fatty acid side chain (LH) containing carbon atoms with bisallylic hydrogens (Embedded Image) (reaction 2). The resulting, carbon-centered radical (L·) adds rapidly to O2 to generate a lipid peroxyl radical (LOO·) (reaction 3) that itself can propagate the chain by reacting with a neighboring lipid molecule to generate another L· and lipid hydroperoxide (LOOH) (reaction 4). In this fashion, many molecules of LOOH may be generated for each initiating radical. Math(2) Math(3) Math(4) Whereas the highly reactive ·OH abstracts H atoms almost without discrimination, less reactive radicals, such as LOO· preferentially abstract H atoms from molecules with weaker bonds, such as the chromanol O-H bond contained in α-tocopherol (α-TOH). In this case, the α-tocopheroxyl radical (α-TO·) is produced and, for LOO·, a molecule of LOOH (reaction 5): Math(5)

A radical may be an oxidizing agent, accepting a single electron from a nonradical, or a reducing agent, donating a single electron to a nonradical. As implied above and like other reactions, free radical reactions are governed by thermodynamic and kinetic principles. A thermodynamic parameter commonly applied in free radical chemistry is the reduction potential that determines the feasibility of a compound X to chemically reduce compound Y. Buettner (104) has compiled a useful list of biologically relevant standard reduction potentials that predict the direction of reactions. Accordingly, the ascorbate/ascorbyl radical system is, for example, capable of reducing the α-TO·, H+/α-TOH system (reaction 6) that has a more positive standard reduction potential. Math(6)

2. 2e-Oxidants

In addition to radicals, several nonradical oxidants are important when considering oxidative modifications in the vessel wall (Table 2). Arguably, the most abundant of these is hydrogen peroxide (H2O2) derived from the action of oxidases such as glucose oxidase on O2, or from the dismutation of O2· (reaction 7): Math(7)

View this table:

Examples of nonradical oxidants of potential relevance to oxidative stress in the vasculature

As is the case for radicals, the reactivity of the different nonradical species varies. Hydrogen peroxide is generally a weak oxidant, although it can directly oxidize thiol (-SH) groups, for example, at the active site of enzymes like glyceraldehyde-3-phosphate dehydrogenase. Hydrogen peroxide can also react with certain heme proteins (e.g., myoglobin and cytochrome c), and this can result in release of iron and/or the formation of ferryl heme plus amino acid radicals that can propagate oxidation reactions, e.g., via radical chemistry. However, transition metal-catalyzed “decomposition” with the resulting formation of ·OH is generally considered the basis of most oxidative damage resulting from H2O2. Like H2O2, ONOO is also a relatively weak oxidant at alkaline pH. However, its protonated form, peroxynitrous acid (ONOOH), is extremely reactive, even more so than hypochlorite/hypochlorous acid (OCl/HOCl). As the pKa value of ONOOH is ∼7.5 (488), in biological systems the formation of ONOO likely results in a powerfully oxidizing environment, comparable to that resulting from generation of ·OH.

Nonradical oxidants like ONOOH (947) and HOCl (371373) appear to react preferentially with proteins rather than lipids (998). Cysteine and methionine residues are the preferred targets, followed by tyrosine, tryptophan, and phenylalanine residues; lysine residues can also be oxidized and, due to their relative abundance, may account for a majority of the ONOOH or HOCl that reacts with proteins (714, 1046). This preference for a reaction with proteins differentiates nonradical from radical oxidants, the latter commonly initiating lipid peroxidation (see above).

In the case of methionine residues, methionine sulfoxide is produced as the predominant species. In comparison, thiol oxidation by HOCl gives rise to a complex array of products, including disulfides (RSSR), sulfenic (RSOH), sulfinic (RSO2H), and sulfonic acids (RSO3H), as well as intra- and intermolecular sulfenamides (RSN-), sulfinamides [RS(O)N-], and sulfonamides [RS(O2)N-] (279, 750, 1047). The reaction with aromatic amino acid residues is studied best for tyrosine, with 3-chloro- and 3-nitrotyrosine being chemically characteristic oxidation products of HOCl and ONOOH, respectively (46, 377), although aldehydes can also be formed (see sect. iiC4). An example of formation of 3-nitrotyrosine is the specific nitration of tyrosine residue 430 of bovine prostacyclin synthase by ONOO that results in a decrease in enzymatic activity (814).

The local environment also governs the likelihood with which 2e-oxidants react with amino acids in protein side chains. For example, vicinal thiol and methionine residues are particularly sensitive to oxidation, as exemplified by H2O2-mediated oxidation of the carboxy-terminal vicinal methionine residues in calmodulin (1083). Similarly, a local environment that promotes ionization of the thiol (Cys-SH) group a cysteine residue, even at neutral pH, to the thiolate anion (Cys-S) enhances its susceptibility to oxidation. This may be achieved via neighboring amino acids to lower the pKa value of cysteine residues, or via complex formation with metal ions. Examples of the latter relevant to oxidative stress in the vessel wall are the zinc thiolate (ZnS4) cluster of endothelial nitric oxide synthase (eNOS) (757) and the cysteine-switch domain of prometalloproteinases (984) that are sensitive to ONOOH (1112) and HOCl, respectively (278).

Two-electron oxidants like HOCl and ONOOH avidly react with certain prosthetic groups, such as heme and iron sulfur centers of proteins (8), often in preference to reaction with amino acid side chains, as in the case of ONOO-induced inactivation of inducible NOS in the presence of calmodulin (418). As mentioned, lipids are poor targets compared with proteins, although chlorinated species have been characterized as products upon reaction with HOCl (978, 1049).

A major biological target for ONOO is carbon dioxide (CO2) (reviewed in Ref. 869). The reaction of ONOO with carbon dioxide is complex and produces metastable species that promote nitration (i.e., addition of a nitro group -NO2), nitrosation (addition of a nitrogen monoxide group -NO), and oxidation reaction. This probably occurs via the intermediate nitrosoperoxycarbonate (ONOOCO2) that homolyzes to form a pair of caged radicals (CO3· ·NO2) that can then diffuse apart to become free radicals. Alternatively, the radicals may combine to nitrocarbonate (O2NOCO2) that then decomposes to nitrite and CO2. Therefore, in the presence of carbon dioxide, the redox chemistry of ONOO is mediated largely by ·NO2 and CO3·; in the absence of carbon dioxide it is mediated by ·NO2 and ·OH (487). Nitrogen dioxide (·NO2) is a strongly oxidizing radical that can initiate a variety of oxidative pathways (486). For example, ·NO2 initiates and promotes lipid peroxidation including that of LDL (115), and it gives rise to nitrotyrosine formation (327, 486). However, ·NO2 is not only derived from ONOO, as other sources, including enzymatic ones, have been identified (see sect. iiC4). Therefore, systems that simultaneously generate ·NO and O2· have the potential to generate ONOO and a complex mixture of oxidants including the radical oxidant ·NO2 that can readily oxidize a number of targets in the vessel wall, including LDL.

C. Sources of Oxidants and Markers of Oxidative Events

Several lines of evidence implicate ROS and RNS in atherogenesis. The advent of gene targeting to predictably modify genes has generated valuable animal models and allowed a genetic approach to gain information on which oxidants affect disease. The development of sensitive analytical methods also enables researchers to obtain direct chemical evidence for specific reaction pathways that promote oxidative events and lesion formation in vivo, and to relate the accumulation of specific oxidized lipid and protein to different stages of atherosclerosis. Understanding the processes that lead to the different oxidative events is important in developing strategies to effectively inhibit such processes and hence potentially atherosclerosis. Within the vessel wall, the different oxidants can originate principally from cellular and extracellular sources, and from enzymatic and nonenzymatic paths that are reviewed briefly in the following sections. The primary consideration given here will be on oxidants of plausible physiological relevance, noting those with the strongest support.

1. NAD(P)H oxidases

It has long been known that phagocytes, including neutrophils, monocytes, and macrophages, contain a plasma membrane-bound, multicomponent oxidase that utilizes electrons derived from NADPH to reduce molecular oxygen to O2· in a reaction that is insensitive to the mitochondrial poison KCN. While O2· is principally a reducing agent, it can give rise to secondary products that include strong oxidants.

The phagocyte NADPH oxidase (Nox) consists of four major units: p22phox (CYBA gene product) and gp91phox (CYBB gene product), the two subunits of the membrane-spanning cytochrome b558, and two cytosolic components, p47phox (NOXO2 gene product) and p67phox (NOXA2 gene product) (25). Other proteins associate or copurify with the oxidase and/or participate in its assembly/activation or inactivation. Phagocyte NADPH oxidase needs to be activated to produce O2·, a process that begins with the phosphorylation of p47phox leading to its association with cytochrome b558(25). Translocation of the G protein Rac to cytochrome b558 is also involved (706) but independent of both p47phox and p67phox translocation (397). Rac1 is the predominant protein expressed in monocytes, in contrast to Rac2 in neutrophils (137). Once activated appropriately, phagocytic cells produce large amounts of O2· over relatively short periods that are involved in host defense.

In analogy to phagocytes, adventitial fibroblasts, vascular smooth muscle cells, and endothelial cells contain a membrane-associated NAD(P)H oxidase that utilizes NADH or NADPH as the electron donor to generate O2· via 1e-reduction of molecular oxygen (reaction 8) (for reviews, see Refs. 333, 766). Math(8)

There is evidence that NAD(P)H oxidase activity represents a major source of ROS in the vasculature (631, 692, 752), and there is some controversy regarding the orientation and cellular location of these enzymes. Whereas in phagocytes electrons are transferred across the membrane to extracellular oxygen, in vascular smooth muscle cells, O2· and H2O2 appear to be produced predominantly inside the cells, and addition of NAD(P)H to the cells augments O2· generation (333). For endothelial cells, there is evidence for extracellular release of O2· as indicated using cell-impermeable “trapping” agents (972).

There are also differences in the protein components of the NAD(P)H oxidases in different vascular cells. Fibroblast and endothelial cells resemble phagocytes in that they contain mRNAs for gp91phox, p22phox, p47phox, and p67phox, whereas vascular smooth muscle cells appear to lack gp91phox and gp67phox (333, 766). Instead, these cells contain homologs of gp91phox, the expression of which increases cellular O2· production (918). A number of homologs including Nox1, -4, and -5, have been identified in vascular cells (147, 864), with endothelial cells expressing very low levels of Nox1, intermediate levels of Nox2, and abundant Nox4 mRNA (6, 864). In contrast, vascular smooth muscle cells express predominantly Nox4 and to a lesser extent Nox1 with negligible amounts of Nox2 (864). There also appear to be distinctions in Nox expression based on the particular type of blood vessel, as Nox2 is relatively more abundant in human aortic smooth muscle cells derived from resistance compared with conductance arteries (955). In addition, there is increasing evidence that Nox1 and Nox4 are located in different cellular compartments (398).

In contrast to phagocytes, endothelial cells, vascular smooth muscle cells, and fibroblasts exhibit low basal activity for O2· generation (691), suggestive of a constitutively active oxidase. Consistent with this, Nox2 in endothelial cells appears to be present as a preassembled, intracellular complex including the p22phox, p47phox, and p67phox subunits (547). Similar to the situation in phagocytes, however, NAD(P)H oxidases in vascular cells can also be activated by stimuli such as angiotensin II, thrombin, platelet-derived growth factor, tumor necrosis factor-α, interleukin-1, and for endothelial cells, mechanical forces (including shear stress) and vascular endothelial growth factor (972). Activation appears to involve the association of additional proteins with the oxidase components (972). In the case of angiotensin II-induced production of ROS by vascular smooth muscle cells, the early phase (30 s) is protein kinase C dependent, whereas the prolonged phase (30 min) is dependent on Rac, Src, and phosphatidylinositol 3-kinase (828). In endothelial cells, angiotensin II rapidly induces ROS production through serine phosphorylation of p47phox, and its enhanced binding to p22phox (548). More recently, homologs of p47phox and p67phox have been identified and proposed to participate in the activation of the Nox family oxidases (929). However, even under stimulated conditions, the amounts of O2· produced by vascular cells are only a fraction of those generated by activated phagocytes and are commonly considered to represent second messengers for a number of key regulatory proteins and cellular responses, rather than toxic species that cause oxidative damage. For example, increased NAD(P)H oxidase activity and p47phox expression is involved in hypercholesterolemia-induced leukocyte-endothelial cell adhesion (902), and increased Nox4 expression during chronic hypertension has been associated with enhanced cerebral vasodilation in vivo (702).

2. Xanthine oxidase

Xanthine oxidase is an iron sulfur molybdenum flavoprotein with multiple functions and present in high concentrations in endothelial cells of capillaries and sinusoids (447). It exists in two forms, xanthine dehydrogenase and xanthine oxidase, of which the former is predominant. Upon oxidation of xanthine or hypoxanthine to uric acid, the dehydrogenase produces NADH, whereas the oxidase generates O2·. However, the dehydrogenase can be converted into the oxidase by nonreversible proteolytic attack or by reversible oxidation of thiol groups. Another signal identified within the endothelium for xanthine oxidase formation is oscillatory shear stress, a condition that is often found at sites prone to atherosclerotic lesion formation (612). The role of xanthine oxidase in the cellular production of O2· has been studied most intensively in the setting of ischemia reperfusion (609).

3. NOS

NOS are a family of enzymes that catalyze the oxidation of l-arginine to l-citrulline and the potent vasodilator ·NO. In the context of the vasculature and atherosclerosis, eNOS and inducible NOS are most relevant. Active eNOS is a homodimer with each monomer consisting of a reductase (containing the binding sites for NADPH, FAD, and FMN) and an oxygenase domain (containing Zn, tetrahydrobiopterin, heme, and l-arginine) that are linked by a hinge region to which calmodulin binds. Reviewing the structure and functions of NOS is beyond the scope of the present work, and the interested reader is referred to several comprehensive and excellent recent reviews (11, 155, 248, 779, 836).

In recent years, it has become apparent that under specific circumstances, eNOS may become “uncoupled” and thereby an important source of ROS. In the absence of sufficient cofactors such as tetrahydrobiopterin for enzyme catalysis, the enzyme may reduce molecular oxygen rather than transfer electrons to l-arginine, thereby generating O2· (987, 1071). The O2· is likely generated by the oxygenase domain of the enzyme through dissociation of a ferrous-dioxygen complex that is normally stabilized by tetrahydrobiopterin (42, 322, 987).

·NO may react with metal complexes, molecular oxygen (i.e., autoxidation) or O2· to form RNS, whereas direct reaction of ·NO with thiols is too slow to be of biological significance (746, 1045). There are three major types of ·NO reactions with metals: the direct reaction of ·NO with the metal center, and ·NO redox reaction with dioxygen metal complexes and high valent oxocomplexes. For example, reaction of ·NO with iron sulfur (Fe4S4) clusters forms an iron nitrosyl complex. Different metal nitrosyl complexes vary in stability, and this dictates their biological relevance (1045). A facile reaction is that of ·NO with proteins containing heme. Biologicallly relevant reactions include those of ·NO with the heme proteins guanylyl cyclase, NOS, and hemoglobin. For example, binding of ·NO to heme of NOS results in reversible inhibition of the enzyme activity (335), whereas reaction of ·NO with the metallo-oxo complex of hemoglobin results in formation of methemoglobin and nitrate.

·NO is unstable in the presence of molecular oxygen. In aqueous solutions it gives rise to dinitrogen trioxide (N2O3) that undergoes hydrolysis to nitrite, whereas in hydrophobic environments, ·NO autoxidation forms ·NO2 and N2O3 (1045). Dinitrogen trioxide is a relatively mild 2e-oxidant, and its primary reaction is nitrosation such as nitrosation of thiols to S-nitrosothiols (RSNO). In addition to nitrosation caused by N2O3, nitrosothiols may be formed via intermediate 1e-oxidation of thiols, possibly mediated by ·NO2, and the subsequent reaction of thiyl radicals with ·NO (456). Thus nitrosothiol formation in vivo depends not only on the availability of ·NO and molecular oxygen but also on the degree of oxidative stress by affecting the steady-state concentration of thiyl radicals (456).

As indicated in section iiB1, ·NO reacts rapidly with O2· to generate ONOO (reaction 1) (44). In fact, this reaction proceeds at near diffusion-controlled rates (k = 1.9 × 1010 M−1 · s−1) (488) and dictates that ONOO formation is kinetically favored over both ·NO autoxidation (k = 2 × 106 M−1 · s−1) and spontaneous O2· dismutation (k = 5 × 105 M−1 · s−1). Moreover, the reaction of ·NO with O2· is more rapid than either its reaction with enzyme-bound heme (k = 102-106 M−1 · s−1) or the reaction of O2· with superoxide dismutase (SOD) (k = 2 × 109 M−1 · s−1) (745).

An emerging paradigm in vascular disease involves the balance between signals mediated by ROS and ·NO. In this context, LDL oxidation and lipid peroxidation is deemphasized in favor of the phenotypic implications of ROS-mediated signals (see sect. iiE).

4. Myeloperoxidase

Myeloperoxidase is a heme-containing enzyme that catalyzes the conversion of Cl to the 2e-oxidant HOCl as the major reaction (reaction 9): Math(9)

Myeloperoxidase is the only human enzyme known to generate HOCl, and chlorinated biomolecules are therefore considered specific markers of oxidation reactions catalyzed by the enzyme (384).

The myeloperoxidase/H2O2/Cl system can also give rise to 3-chlorotyrosine (377), chlorohydrins such as those of cholesterol and fatty acids (386, 1049), α-chloro fatty acid aldehydes (950), and free amino acid or protein-bound tyrosyl radicals (385). Tyrosyl radicals themselves may participate in secondary oxidation reactions, including the oxidation of LDL (806). In addition, the myeloperoxidase/H2O2/Cl system or HOCl convert l-tyrosine into p-hydroxyphenylacetaldehyde (374) that itself can react with the amino head group of phospholipids (378, 391), and the ε-amino groups of protein lysine residues to generate hydroxy-amino acids (376). These hydroxy-amino acids themselves may then be converted by the myeloperoxidase/H2O2/Cl to highly reactive α-hydroxy and α,β-unsaturated aldehydes such as glyceraldehyde, 2-hydroxypropanal, and acrolein (17, 376). Furthermore, the myeloperoxidase/H2O2/Cl system and HOCl convert α-amino acids to their corresponding aldehydes (379), and l-serine to the chemically well-characterized advanced glycation end product, Nε-(carboxymethyl)lysine (18). Thereby, and similar to the process of lipid peroxidation, the myeloperoxidase/H2O2/Cl system and HOCl can generate a series of secondary oxidation products that have the capacity to give rise to oxidized biomolecules including oxidized LDL capable of converting macrophages into foam cells (466).

The myeloperoxidase/H2O2/Cl system and HOCl also oxidize nitrite to the nonradical oxidant nitryl chloride (NO2Cl) and the radical ·NO2, both of which promote nitration and can convert tyrosine into 3-nitrotyrosine (226, 227, 979). Recent studies provide evidence that myeloperoxidase, in fact, plays a major role in the generation of nitrating species in vivo and that formation of 3-nitrotyrosine is strictly dependent on the availability of ·NO2 (486). For example, in a mouse model of inflammation, 3-nitrotyrosine levels increase in wild-type mice but not animals that lack functional myeloperoxidase (295), demonstrating that myeloperoxidase functions to generate RNS in vivo when nitrite/nitrate is available.

5. Lipoxygenases

Lipoxygenases are iron-containing dioxygenases that catalyze the stereospecific insertion of molecular oxygen into polyunsaturated fatty acids to give rise to a complex family of biologically active lipids, including prostaglandins, thromboxanes, and leukotrienes (reviewed in Refs. 780, 811). Prostaglandins and thromboxanes (or prostanoids) comprise metabolites from arachidonate and similar fatty acids. Prostaglandins contain a cyclopentane ring and are derived from the action of cyclooxygenase I (also referred to as prostaglandin G/H synthase I) that is expressed constitutively in almost all tissues and that acts on arachidonate. The enzyme requires activation by a “seeding” peroxide to oxidize the heme-iron at the active site to an oxo-heme species and a protein-derived tyrosyl radical (compound I). The tyrosyl radical is thought to abstract a bisallylic hydrogen atom from arachidonate to initiate formation of prostaglandin G2 that is then reduced to prostaglandin H2 by the peroxidase activity of cyclooxygenase. Both prostaglandin G2 and H2 are rapidly transformed into other prostaglandins, such as thromboxane A2 and prostacyclin, several of which are involved in the regulation of vascular tone and homeostasis. For example, vascular endothelium produces prostacyclin that dilates blood vessels and is a powerful inhibitor of platelet aggregation, similar to ·NO. In contrast, thromboxane A2 exhibits opposing biologic activities of vasoconstriction and platelet aggregation.

The non-heme iron-containing lipoxygenases oxidize certain fatty acids at specific positions of the carbon chain to their corresponding hydroperoxides that are the precursors of leukotrienes. Leukotrienes represent a family of chemicals that have potent biological activities and that differ from prostanoids in that they have a conjugated triene structure and no cyclopentane ring. These mediators may contribute to inflammatory reactions and, in the case of slow-reacting substance A, to an increase in vascular permeability. Like cyclooxygenase, lipoxygenases require low levels of “seeding peroxides” to oxidize inactive Fe2+ to active Fe3+ enzyme and are likely affected by the “peroxide tone” of cells.

In addition to the oxidation of fatty acids, 12- and 15-lipoxygenase have been reported to also oxygenate complex, esterified fatty acids, such as those in cholesterol esters and phospholipids (50, 87, 652). In vitro, 15-lipoxygenase can oxidize LDL (51), and this is achieved by a combination of direct, enzymatic and indirect, nonenzymatic oxidation reactions (396, 965, 1076). Formation of nonenzymatic oxidation products has been linked to the ability of cyclooxygenase and lipoxygenases to generate oxidants in addition to the specifically oxidized fatty acids. It is well known that cyclooxygenase also engages in the oxidation of other substrates (e.g., 13-cis-retinoic acid) via 1e-reactions (797), implying the generation of radicals as by-products during normal enzymatic reaction. Similarly, LOO· are generated and released during the catalytic action of lipoxygenases (142) and may participate in subsequent nonenzymatic lipid peroxidation (665, 964, 965). In addition, thromboxane synthase cleaves prostaglandin H2 to 12-hydroxy-5,8,10-heptadecanoic acid and produces malonyldialdehyde.

6. Mitochondrial respiration

As a by-product of electron flow through the mitochondrial electron transport chain, 1–2% of the molecular oxygen consumed may be converted to O2· (143), raising the possibility that this represents a major intracellular source of ROS. The steady-state levels of mitochondrial O2· depend on the activity of Mn-containing SOD (Mn-SOD), an enzyme located in the mitochondrial matrix that converts O2· to H2O2 and molecular oxygen. In addition to the electron transport chain reactions of the inner mitochondrial membrane, monoamine oxidase in the outer mitochondrial membrane may represent another source of H2O2. Within the mitochondria, O2· production occurs primarily at complex I (NADH dehydrogenase) and complex III (ubiquinone-cytochrome bc1).

A number of cellular responses have been linked to mitochondrial O2· generation including growth arrest, apoptosis, and necrosis (551, 741). With respect to vascular disease, complications from diabetes have been linked to mitochondrial O2· generation. For example, overproduction of O2· has been associated with changes in glycosylation (212) and protein kinase C activation (671) in vascular cells. Overexpression of Mn-SOD appears to ameliorate many of these deleterious consequences of hyperglycemia (212, 671). These data are in keeping with known genetic alterations in mitochondrial energy metabolism that involve a predisposition to glucose intolerance and diabetes (597, 851). A recent study has also linked mitochondrial production of ROS to the early atherosclerotic lesion development (30), and emerging evidence indicates that many cardiovascular syndromes are associated with some evidence for mitochondrial dysfunction, although a causal role has yet to be established in vivo (for review, see Ref. 755).

7. Transition metals

Free transition metals like iron and copper are strong catalysts for oxidation reactions in the presence of hydroperoxides, such as LOOH proposed to be “seeded” in LDL by 15-lipoxygenase (708); they catalyze homolytic cleavage of LOOH to lipid alkoxyl radicals that can initiate lipid peroxidation and other oxidation reactions. However, the concentration of free transition metals in vivo appears to be very low (e.g., see Ref. 749), and there is little convincing evidence that they are related to atherosclerosis. Indeed, an autopsy study on patients with hemochromatosis, a genetic disorder that results in elevated plasma and tissue levels of iron, showed that they have less coronary artery disease than age- and sex-matched controls (624).

In addition to free iron, biological forms of iron such as heme, hemoglobin, and myoglobin all have the potential to catalyze oxidative reactions in the vessel wall. In vitro, these iron-containing molecules are able to oxidize isolated LDL (625). In contrast to free transition metals, heme also binds to and oxidizes LDL in diluted serum (119), raising the possibility of heme, e.g., derived from hemolysis, being an in vivo oxidant for LDL. In support of this notion, intravascular hemolysis increases atherogenicity of diet-induced hypercholesterolemia in the rabbit (246).

8. Other oxidants

Lipid peroxyl radicals are regarded as the major chain carrier of nonenzymatic, free radical-mediated lipid peroxidation (112) and the ultimate precursor for LOOH. The presence of LOOH in human and animal atherosclerotic lesions (see sect. iiiB2) therefore implies the formation of LOO· as intermediate species. Peroxyl radical-mediated peroxidation of linoleate, the major lipoprotein lipid that oxidizes readily, theoretically yields four different regioisomers of LOOH that do not accumulate with equal abundance (725) (Fig. 6). In pure lipids the thermodynamically favored 9- and 13-trans,trans-isomers predominate, whereas LDL and other lipoproteins that contain α-TOH predominantly yield the 9-cis,trans- and 13-trans,cis-regioisomers of cholesteryl linoleate hydroperoxides (476, 964). This is explained by α-TOH reacting with LOO· before it acts as a chain carrier and oxidizes another lipid. In human atherosclerotic lesions, most LO(O)H are present as 9-cis,trans- and 13-trans,cis-regioisomers, particularly during the early developmental stages of the disease (967). This suggests that lipoprotein lipid peroxidation in the arterial wall most likely takes place in the presence of α-TOH. As a consequence, α-TO· rather than LOO· is likely the major radical oxidant that “carries” the chain reaction of lipid peroxidation (see sect. iiiC4).

FIG. 6.

Formation of specific regioisomers of lipid hydroperoxides during nonenzymatic lipoprotein oxidation in the presence and absence of α-tocopherol (α-TOH). Lipids containing bisallylic hydrogen atoms such as cholesteryl linoleate are present in the cis,cis (Z,Z) configuration. Upon hydrogen abstraction, β-fragmentation competes with oxygen addition. In the presence of suitable hydrogen donors such as α-TOH, formation of the kinetically favored trans,cis (Z,E) regioismer predominates. In the absence of hydrogen donors (α-TOH), the thermodynamically favored trans,trans (E,E) regioisomer is formed in preference. [Adapted from Upston et al. (963).]

While different types of LOOH can be analyzed with high sensitivity and specificity by HPLC with postcolumn chemiluminescence detection (805, 846, 1075), they are metabolized readily in biological systems, complicating their use as markers of in vivo lipid peroxidation. LOOH are metabolized principally through peroxidase-mediated reduction to the corresponding alcohols. In the extracellular space, where lipoprotein lipid oxidation is believed to take place yet suitable peroxidases are not present, reduction of LOOH can be achieved by methionine residues of apolipoproteins (292, 293, 596). Therefore, the accumulation of LOOH plus corresponding lipid alcohol [i.e., LO(O)H] may represent a suitable index of the extent of LDL oxidation in the vessel wall (1055).

Hydroxyl radicals may also be generated by reaction of metalloproteins with peroxides (13) and perhaps by the decomposition of ONOO (44, 868). Because of their extremely high reactivity, ·OH are commonly quantified indirectly with biomarkers such as o-tyrosine and m-tyrosine, and protein-bound 3,4-dihydroxyphenylalanine, which are produced from hydroxylation of phenylalanine and tyrosine, respectively (534).

D. Antioxidant Defenses

Oxidative modifications within the arterial wall that may initiate and/or contribute to atherogenesis likely occur when the balance between oxidants and antioxidants shifts in favor of the former. Therefore, it is important to consider sources of oxidants in the context of available antioxidants. Many substances may prevent, or significantly delay, the oxidation of other substrates. However, an antioxidant is defined as a substance being effective against oxidative damage when present in much smaller quantity than the substance that it protects (353). With regard to atherosclerosis, vascular antioxidants need to protect against 1e- (radical) and 2e-oxidants, both within and outside cells. In the context of the oxidative modification hypothesis, antioxidant protection of LDL in the extracellular space deserves focus, as oxidized LDL has many potential proatherogenic activities (57), and the cellular accumulation of oxidized LDL is considered a hallmark of atherosclerosis (882). In addition, cellular antioxidants are likely important in the context of the presence of heightened oxidative stress within the vessel wall and the known effect of oxidative events on key cellular activities such as ·NO-related bioactivities (see sect. ivC1). The following briefly describes the antioxidant defenses in arterial wall cells and lipoproteins that counteract oxidative modifications. The topic has been reviewed previously (895, 900), and for a more comprehensive, general review on antioxidant defenses, the reader is referred to an excellent monograph (353).

1. Enzymatic antioxidants

The classic antioxidant enzymes are largely cell-associated proteins whose function is to maintain a reducing tone within cells (353) (Table 3); they may also be involved in the maintenance of extracellular antioxidants (420). Enzymatic antioxidants principally include SOD, catalase, glutathione peroxidases, glutathione reductase and transferases, thiol-dissulfide oxidoreductases, and peroxiredoxins. Many of these enzymatic antioxidants are present in normal arteries, most likely within vascular wall cells as extracellular fluid is largely devoid of enzymatic antioxidants (reviewed in Refs. 352, 897).

View this table:

Cellular antioxidants


SODs catalyze the univalent reduction and oxidation of O2· to H2O2 and molecular oxygen (reaction 10): Math(10)

There are three forms of SOD in mammalian systems: the copper-zinc (Cu,Zn-SOD), Mn-SOD, and extracellular SOD (EC-SOD). The copper-zinc enzyme is present in virtually all cells, where most of it is located in the cytosol, with some activity in lysosomes, peroxisomes, nucleus, and the space between inner and outer mitochondrial membrane. The copper ion functions in the dismutation reaction by undergoing alternate oxidation and reduction, and this forms the basis for the ability of copper-chelating agents, such as diethyldithiocarbamate, to inhibit SOD (383). The rate of the dismutation reaction catalyzed by Cu,Zn-SOD is essentially independent of pH, in contrast to the noncatalyzed dismutation of O2· and that catalyzed by Mn-SOD (which decreases at alkaline pH). Manganese-containing SOD is largely located in the mitochondria, is cyanide insensitive, and contributes ∼10% of total cellular activity.

As indicated by its name, EC-SOD (587, 588) is an extracellular form of the enzyme. It also contains copper-zinc and is a notable exception in that significant amounts of this antioxidant enzyme are present in the normal arterial wall outside cells (907). The enzyme is bound to heparan sulfate proteoglycans in the glycocalyx of various cells, including endothelial cells, and in the connective tissue matrix. In human aorta, the enzyme is localized to the connective tissue matrix and is produced by smooth muscle cells (686). Recent evidence suggests that the enzyme exists in two forms, each with a unique disulfide pattern, and with only one of the two forms being active (718). The synthesis of EC-SOD by smooth muscle cells is modulated by cytokines, growth factors, vasoactive factors, and oxidants (906, 908), suggesting that changes in enzyme activity may occur in vascular diseases. Consistent with this, SOD activity is decreased by homocysteine (675) and increased by angiotensin II and hypertension (281). Indeed, a key function of Cu,Zn-SOD is thought to be protection against O2·-induced inactivation of endothelial cell-derived ·NO (3). In direct support of this contention, overexpression of EC-SOD improves endothelial function in a rat model of hypertension (245). In the presence of bicarbonate, Cu,Zn-SOD also has peroxidase activity towards H2O2 (399, 798) that results in the oxidation of cosubstrates and may inactivate the enzyme distinct from that seen with H2O2 at high pH (402). Reducing substances like urate can prevent inactivation of the enzyme (399). An additional feature of EC-SOD is its ability to bind to and protect type I collagen from oxidative fragmentation (717).


Two types of enzyme usually metabolize the H2O2 resulting from the dismutation of O2· or generated by oxidase enzymes including xanthine oxidase. Catalase directly decomposes H2O2 to water and molecular oxygen (reaction 11), whereas peroxidases eliminate H2O2 by using it to oxidize another substrate (reaction 12): Math(11) Math(12)


Glutathione peroxidases cooperate with catalase in the removal of H2O2 in vivo by using reduced glutathione (GSH) to reduce H2O2 to H2O and oxidized glutathione (GSSG) (reaction 13): Math(13)

Glutathione peroxidases are widely distributed in animal tissues and can act on peroxides other than H2O2. Most importantly with regard to oxidative modification in the artery wall, they can catalyze GSH-dependent reduction of LOOH to the corresponding alcohol (LOH) (reaction 14): Math(14)

This conversion represents an antioxidant defense as the potential to form pro-oxidants from the degradation of hydroperoxides is negated.

Glutathione peroxidases are tetrameric enzymes that contain and require selenium for activity. Selenium actively participates in the catalytic reaction (249), and this is often the basis for antioxidant protection offered by supplemental selenium. Selenium-containing peroxidases comprise a family of enzymes of at least four types. The “classic” glutathione peroxidase (cGPx) acts on H2O2 and hydroperoxides of fatty acids and cholesterol, but not esterified lipids such as those present in lipoproteins. In the mouse, heterozygous deficiency of the enzyme is associated with endothelial dysfunction and an increase in circulating and aortic concentrations of F2-isoprostanes (252), i.e., secondary oxidation products formed during the nonenzymatic oxidation of arachidonic acid (640, 1084) that are also used as a marker of oxidative stress (526, 642, 712, 772).

Phospholipid hydroperoxide glutathione peroxidase (PHGPx) is the only enzyme known to reduce complex lipid hydroperoxides in lipoproteins (576, 804). In addition to GSH, PHGPx can utilize a number of different low-molecular-weight thiols as reducing agents. This isoform is membrane bound (944) and, like the classic glutathione peroxidase, not found in extracellular fluids. The tetrameric gastrointestinal glutathione peroxidase found in the cells lining the gastrointestinal tract is most likely involved in the detoxification of dietary peroxides (1043). Human plasma contains low concentrations of a glycosylated peroxidase, now called extracellular peroxidase, that acts on H2O2 and phospholipid hydroperoxides (928). It is less clear whether, and if so to what extent, this isoform acts as a glutathione peroxidase, as plasma levels of GSH are low compared with the Michaelis constant (Km) of the enzyme. In addition to extracellular peroxidase, human and rat plasma contain selenoprotein P (108) that reduces phospholipid hydroperoxides (but not H2O2) (791) similar to PHGPx.

Glutathione peroxidases operate in concert with glutathione reductase that catalyzes the reduction of GSSG at the expense of NADPH (reaction 15): Math(15)

NADPH is provided primarily by the oxidative pentose phosphate pathway (143) that is initiated by glucose-6-phosphate dehydrogenase and that maintains the cellular redox couple NADPH/NADP+ in balance. In this context, glucose-6-phosphate dehydrogenase can be considered an antioxidant. For example, overexpression of glucose-6-phosphate dehydrogenase in endothelial cells decreases oxidative stress and increases the bioavailability of ·NO (540).

In addition to its central role in glutathione peroxidase activity, GSH is involved in several other antioxidant pathways, including direct scavenging of oxidants, ascorbate metabolism, the maintenance of protein sulfhydryl groups, and the detoxification of xenobiotics via the action of glutathione S-transferases (reviewed in Refs. 367, 910). Some glutathione transferases show peroxidase-like activity, including against phospholipid hydroperoxides (421), suggesting that these enzymes may contribute to the cellular antioxidant defense.


The ratio of GSH to GSSG regulates the activity of protein disulfide isomerases, enzymes important in the correct folding of proteins during their synthesis. Protein disulfide isomerases are also found on the surface of cells where they have been proposed to be involved in the control of the redox state of existing exofacial protein thiols or reactive disulfide bonds (450), and transnitrosation reactions (i.e., the movement of the nitrogen monoxide group -NO from one molecule to another), and the “ordered movement” of ·NO across the plasma membrane (756, 1098). Protein disulfide isomerases belong to a group of enzymes called thiol-disulfide oxidoreductases that also includes thioredoxin and glutaredoxin. While thioredoxin acts in conjunction with thioredoxin reductase and NADPH, glutaredoxin can be reduced directly by GSH (408). As well as acting as thioltransferases on proteins such as ribonucleotide reductase and methionine sulfoxide reductase (408), these oxidoreductases participate in the antioxidant defense in many different ways and may play a major role in cellular redox signaling (409). For example, protein disulfide isomerases also have dehydroascorbate reductase activity (1026) and therefore are involved in the maintenance of cellular ascorbate, as shown for glutaredoxin (703) and thioredoxin reductase (604, 605). In addition, protein disulfide isomerases, thioredoxin, and glutaredoxin all reduce lipid hydroperoxides (65), mitochondrial glutaredoxin is involved in dethiolation reactions (451), and thioredoxin can reactivate oxidized glyceraldehyde-3-phosphate dehydrogenase. Interestingly, the proteins are produced and at least some of them may be secreted by various cells (655, 786), raising the possibility that they also contribute to the antioxidant defense in extracellular fluid, although it is unknown whether they can mediate the reduction of lipoprotein lipid hydroperoxides.


Peroxiredoxins are a family of antioxidant enzymes (482) that comprises several members located in the cytosol, mitochondria, peroxisomes, and plasma membrane of cells, as well as in plasma (reviewed in Ref. 280). The enzymes commonly exhibit peroxidase activity that is dependent on reduced thioredoxin and/or GSH. Peroxiredoxins exist as homodimers and all contain an amino-terminal, reactive cysteine residue that is converted to sulfenic acid via reaction with H2O2 (140). The sulfenic acid is a reaction intermediate and either reacts with an accessible thiol such as that of the other subunit to form an intermolecular disulfide that can then become reduced by thioredoxin to reestablish active peroxiredoxin. Alternatively, the sulfenic acid may become oxidized further to sulfinic acid so that it is no longer reduced by thioredoxin (1082), but instead by a recently discovered, sestrin-dependent reduction pathway (103, 1062). There is increasing evidence that peroxiredoxins play an important role as antioxidants in vivo. For example, mice with a targeted mutation of peroxiredoxin 6 develop normally but are susceptible to oxidative stress (1013). In addition to their role as a peroxidase, individual members of the peroxiredoxin family appear to serve a viariety of functions associated with different biological processes such as cell proliferation, differentiation and gene expression (280).


Despite the overall lack of the major antioxidant enzymes in extracellular fluid, hydroperoxides of lipoprotein phospholipids and cholesterylesters may be reduced to the corresponding alcohols. This reduction appears to be mediated via a nonenzymatic reaction (803) that nonetheless requires apolipoproteins. Recent evidence suggests that methionine residues in apolipoproteins A-I, A-II (292, 595) and, to a lesser extent, apolipoprotein B-100 (generically referred to as MetApo) (596), reduce lipoprotein lipid hydroperoxides. In exchange for hydroperoxide reduction, the methionine residues are oxidized to methionine sulfoxide (MetO) (292) (reaction 16): Math(16)

The physiological relevance of this lipid hydroperoxide reduction has not been demonstrated. In vitro, the above reaction is slow, although it is accelerated by unknown factor(s) in liver (154), and a slow reduction could still be important under circumstances such as in an atherosclerotic vessel where the average residence time of lipoproteins is long. Peptide methionine sulfoxide reductase acts on methionine sulfoxides in lipid-free and lipid-associated apolipoprotein A-I, suggestive of enzymatic repair of the oxidized protein (850), although it remains unclear whether the cellular sulfoxide reductase can act on extracellular apolipoprotein(s).

2. Metal sequestration

Transition metals, specifically iron and copper, are essential for the synthesis of a very large range of proteins, including enzymes, like eNOS, that play a central role in the normal function of blood vessels. However, these metals can undergo 1e-transfer reactions that result in autoxidation reactions or the decomposition of peroxides to peroxyl, alkoxyl, and hydroxyl radicals. For example, transition metals can induce oxidative damage to lipoproteins, and copper is commonly used as the in vitro oxidant for LDL (234). Thus binding of adventitious transition metals to inactive chelates in the vascular wall may represent an antioxidant defense. Several proteins, such as ferritin, transferrin, haptoglobin, hemopexin, and ceruloplasmin, specifically bind biological iron and copper complexes and are considered to be part of the body's antioxidant defense system (352).


Ferritin, the principal cellular iron-binding protein, comprises 24 subunits of two types (H- and L-chains). The H-chains contain the metal-binding site and can oxidize Fe2+ to Fe3+, a process required for the intact protein to take up iron. The synthesis of ferritin is regulated tightly and in concert with that of transferrin receptors, via binding of iron-regulatory proteins to iron-responsive elements (481). When iron is low, binding of the regulatory proteins increases translation of transferrin receptor mRNA, while it inhibits that of ferritin mRNA. Conversely, when cellular iron is high, transferrin receptor mRNA is degraded rapidly. Synthesis of ferritin H-chains is also upregulated following activation of cellular heme oxygenase-1 (993) that produces Fe2+, carbon monoxide, and biliverdin from the oxidative degradation of heme.

In the extracellular space, iron is bound to transferrin that transports the metal to the various tissues. Importantly, under normal conditions, transferrin is loaded to only ∼20–30% so that there is substantial iron binding capacity remaining and “free iron” is essentially nondetectable. In addition, human plasma contains haptoglobin and hemopexin that can prevent the pro-oxidant activity of hemoglobin (344) and heme (36), respectively. For example, hemopexin can inhibit in vitro LDL oxidation induced by heme (29) and hemoglobin (626).

Albumin transports dietary copper to the liver where it is incorporated into ceruloplasmin for release into circulation and transport to various tissues. Ceruloplasmin has ferroxidase activity that is required for iron incorporation into ferritin (see above). The protein can also catalyze oxidation of a wide range of phenols and, surprisingly, was reported to induce in vitro (223) and to faciliate cell-mediated, metal-dependent LDL oxidation (648). As a result of this, it has been speculated (648) that ceruloplasmin may participate in LDL oxidation in the arterial wall. It should be noted, however, that many more proteins bind metals nonspecifically. For example, albumin has several metal binding sites and at a concentration similar to that in the vascular wall (860), is able to inhibit in vitro lipoprotein oxidation induced by transition metals (982) and ceruloplasmin (223). Similarly, effective antioxidation of lipoproteins in the presence of Cu2+ is mediated by the high-molecular-weight fraction of human suction blister fluid (178).


Heme oxygenases catalyze the breakdown of heme to carbon monoxide, iron, and biliverdin, the latter of which is then used to synthesize bilirubin via the action of NADPH-dependent biliverdin reductase (575). Of the three isoforms, heme oxygenases-2 and -3 are expressed constitutively, whereas heme oxygenase-1 is induced in response to various stimuli, including heme, angiotensin II, ·NO, inflammatory cytokines, ultraviolet irradiation, and heat shock. Iron release from the active enzyme results in increased synthesis of ferritin, and this has been proposed to restrict the participation of cellular iron in oxidation reactions (993). The overall effect of heme oxygenase may be seen as that of removing the potential pro-oxidants heme and iron, while at the same time producing the antioxidants biliverdin and bilirubin (see below). Induction of heme oxygenase (previously known as heat shock protein 32) represents a general response of cells to oxidative stress (20) and has been proposed to represent an antioxidant defense (891). Indeed, cells deficient in heme oxygenase-1 are hypersensitive to oxidants like H2O2 (726), and experiments modulating cellular enzyme activity demonstrate that heme oxygenase-1 can mediate protection against oxygen toxicity, although protection is seen only over a narrow range of enzyme expression (197). Multiple mechanisms appear to be involved in this protection, including the formation of bilirubin and possible upregulation of catalase and glutathione content (411). Deficiency in heme oxygenase-2 results in increased sensitivity to hyperoxia and increased markers of oxidative injury (196). Recently, biliverdin reductase has been proposed as a cellular protective agent via regeneration of bilirubin (32).

By comparison to radical oxidants, relatively little is known about antioxidants effective against 2e-oxidants such as HOCl, against which protein thiols provide the primary defense (414) so that albumin in the vascular wall may be protective. However, in the early stages of atherosclerosis, when the albumin concentration is low (860) yet HOCl is likely to be produced (369), other protein thiols may be targets for HOCl. Thus this type of antioxidation could become limited or negated by loss of function to essential proteins such as those found in extracellular space (Table 4) and cell plasma membranes.

View this table:

Extracellular antioxidants

3. Nonproteinaceous antioxidants

As well as protein antioxidants, there are several low-molecular-weight compounds in addition to GSH that are thought to contribute to the antioxidant defense systems. For example, there is a large body of literature on the role of dietary antioxidants in diseases including atherosclerosis (266). In addition, there are several compounds synthesized in our body that can act as antioxidants. In the following, important low-molecular-weight antioxidants will be divided into water- and lipid-soluble compounds.


Biological systems contain a large number of different redox-active compounds. These include the endogenous compounds uric acid and bilirubin, as well as diet-derived compounds like vitamin C and a variety of flavanoids and polyphenols. The latter include compounds for which beneficial effects, such as improvement of vascular function (see, e.g., Ref. 214), have been reported. However, flavonoids and polyphenols are poorly absorbed, extensively metabolized, and possess only modest antioxidant activity in vitro so that they are not likely to make a significant contribution to the overall antioxidant defenses in humans. The following discussion is therefore limited to vitamin C, uric acid, and bilirubin.

I) Ascorbate (vitamin C).

Most animals can synthesize ascorbate from glucose, but humans and other primates lost gulonolactone oxidase required for the terminal biosynthetic step, and so depend on dietary intake of the vitamin. Vitamin C is ubiquitous in biological fluids and, at physiological pH, present as the mono-hydro conjugate base ascorbate. Ascorbate is labile due largely to its redox potential of E1/2 of approximately +282 mV at pH 7.0 (104) and generally considered to function as a reducing agent capable of enhancing enzymatic activity through maintenance of the iron center in the active, ferrous state. Ascorbate is a cofactor for several enzymes engaged in hydroxylation reactions. Relevant to the vascular wall, ascorbate is required for the biosynthesis of collagen via proline and lysine hydroxylases. Collagen formed in the absence of ascorbate is insufficiently hydroxylated and does not form proper fibers, resulting in fragile blood vessels.

This inherent metal-reducing activity is also the basis for the pro-oxidant activity of ascorbate in vitro, attributed to the effective redox cycling of iron and the production of hydroxyl radicals through Fenton chemistry (351). There is compelling evidence, however, that this pro-oxidant activity is not observed in vivo where ascorbate generally protects biological molecules from oxidation (129) and even after iron supplementation (55). Indeed, ascorbate is a quantitatively important antioxidant (1023); qualitatively, it also provides the first line of defense against oxidative damage in human plasma (267, 269, 835), where it is considered the most effective water-soluble antioxidant (265, 270) (Table 4).

The potential protective activities of vitamin C related to atherosclerosis have been reviewed previously (135). Ascorbate is generally regarded as a primary, first line protective agent that repairs or nullifies free radicals by donating a single electron followed by a proton to yield a chemically reduced, nonradical product and ascorbyl radical that readily dismutates to ascorbate and dehydroascorbic acid. Both ascorbyl radical (9, 104) and dehydroascorbic acid (604, 620, 657, 703) can be reduced by glutathione-dependent enzymatic systems, effectively recycling the pool of bioavailable vitamin C.

Ascorbate at physiological concentrations inhibits LDL lipid peroxidation initiated by vascular cells (591), activated neutrophils (896), and cell-free systems (802). This is achieved by ascorbate scavenging aqueous oxidants and acting as a coantioxidant for LDL's α-TOH (270, 1058) (see sect. iiiE). For example, ascorbate efficiently scavenges aqueous peroxyl (33, 34) and protein radicals (202, 1052, 1060) that can initiate LDL lipid peroxidation. Ascorbate also stongly attenuates, though does not completey prevent, LDL and plasma protein oxidation induced by the 2e-oxidant HOCl (134, 365, 373), indicating that the vitamin could contribute to the antioxidant defense against HOCl. However, the antioxidant protection offered by ascorbate against ONOO is limited. Thus a recent study calculated that under conditions relevant for the extracellular space (i.e., pH 7.3, 100 μM ascorbate, 1 mM carbon dioxide), the vitamin would be expected to scavenge only ∼3.5% of the ONOO (10 μM); at 10 mM (i.e., reflecting intracellular concentrations), ascorbate would scavenge ∼67% of ONOO (512). Together, these studies show that ascorbate is highly effective in preventing oxidation reactions induced by most 1e-oxidants and that it also strongly protects, but does not completely prevent, oxidative damage induced by 2e-oxidants, particularly ONOO.

The ability of ascorbate to reduce α-TO· to α-TOH, recognized first by Packer et al. (688), is now well established as a mechanism that spares vitamin E from oxidation in micelles (62, 1059) and isolated biological membranes (462, 606). Ascorbate completely inhibits LDL lipid oxidation initiated by aqueous and lipophilic peroxyl radicals (78, 79, 802), horseradish peroxidase (1058), and lipoxygenase (965) in vitro. At least in the case of horseradish peroxidase, addition of ascorbate results in immediate disappearance of α-TO· and cessation of lipid peroxidation, with concomitant formation of ascorbyl radical (1058). Despite these data, however, the importance of this interaction for in vivo maintenance of vitamin E and inhibition of lipid oxidation remains to be established (113, 416).

Another example of synergistic activity of ascorbate is the maintenance of cellular glutathione via reduction of glutathione thiyl radical (228). Reduction of glutathione thiyl radical by ascorbate is thermodynamically favored and kinetically fast (ΔG ∼60 kJ/mol, k ∼6 × 108 M−1 · s−1) (930, 1014) so that ascorbate is a preferred radical sink in cells (914) maintaining GSH in the antioxidant active, reduced form.

In addition to oxidant scavenging, ascorbate has a number of potentially protective activities related to atherosclerosis and cardiovascular disease. Patients with unstable coronary syndromes (1001) or hypertension have low plasma concentrations of ascorbate, and vitamin C supplements have the potential to attenuate defective endothelium-dependent vasodilation (see sect. ivC1). Multiple mechanisms may operate (603), including the maintenance of eNOS activity, attenuation of cellular oxidative stress, inhibition of LDL oxidation, and preservation of compounds that exhibit bioactivity similar to authentic ·NO, such as S-nitrosothiols. Ascorbate can also compete with O2· for reaction with ·NO, although effective competition is observed only at supraphysiological concentrations of vitamin C (≥10 mM) (444).

Ascorbate may also maintain ·NO synthesis by eNOS that requires cofactors such as NADPH, flavin adenine dinucleotide, flavin mononucleotide, and tetrahydrobiopterin. A deficiency in some of these cofactors can change the enzyme into an oxidase that produces O2· in addition to, or in place of, ·NO (987). Addition of ascorbate to aortic endothelial cells in vitro enhances eNOS activity, and this is due to an increase in cellular tetrahydrobiopterin (415). Ascorbate chemically stabilizes tetrahydrobiopterin, likely due to reduction of the 1e-oxidation product trihydrobiopterin radical (711) rather than dihydrobiopterin (989). Therefore, ascorbate effectively increases the half-life and concentration of tetrahydrobiopterin in cells (392) and decreases O2· production by NOS (986). In contrast to ascorbate, GSH does not appear to be important in the preservation of cellular tetrahydrobiopterin (415). Chemical stabilization also attenuates autoxidation of tetrahydrobiopterin, a process that itself can cause inactivation of ·NO (607). Furthermore, ascorbate may preserve ·NO bioavailability indirectly by sparing cellular thiols, such as GSH, that have the potential to stabilize ·NO via formation of S-nitrosothiols (431).

The preservation of ·NO formation and bioavailability has potential downstream effects on vascular homeostasis independent of vessel relaxation. These include inhibition of leukocyte adhesion to endothelial cell, smooth muscle cell proliferation, and platelet aggregation (135, 504). Supplemental vitamin C protects leukocytes from smoke-induced adhesion to micro- and macrovascular endothelium in hamsters (536) and humans (1065). The underlying mechanism for the inhibitory effect remains unclear, although the activation of important monocyte adhesion proteins, such as CD11b, is under redox regulation and hence a potential site of action of ascorbate (67). Other factors that may be involved in the regulation of leukocyte adhesion are platelet activating factor-like lipids formed during nonenzymatic phospholipid oxidation (382, 538) and F2-isoprostanes (629), raising the possibility of ascorbate modulating cell adhesion via regulation of lipid oxidation.

II) Uric acid.

Uric acid is produced by the oxidation of hypoxanthine and xanthine catalyzed by xanthine oxidase and dehydrogenase (see sect. iiC2). At physiological pH almost all uric acid is present as urate monoanion. While in most species urate is effectively metabolized by urate oxidase, humans and other primates lack this enzyme so that urate accumulates to high concentrations in the blood (200–400 μM). Inside cells urate is present at much lower concentration. Urate can directly scavenge singlet oxygen, ·OH and peroxyl radicals (16), certain oxidants produced by enzymes (1058) and CO3· and ·NO2 derived from ONOO (867). These findings led Ames et al. (16) to propose that the biological function of urate may be that of an antioxidant. Indeed, scavenging of ONOO by urate may protect proteins against nitration. Urate also binds transition metals (185) that may be relevant for LDL lipid peroxidation in the vascular wall (see, e.g., Ref. 527). Reaction of urate with 1e-oxidants leads to the formation of urate radical that can be reduced by ascorbate. However, unlike ascorbate, urate is not able to reduce α-TO· (see above).

III) Bilirubin.

Bilirubin is the end product of heme degradation in humans, with ∼275 mg or 500 mmol being produced each day. Bilirubin is a strong reducing agent and a potential physiological antioxidant (901). In extracellular fluids, the pigment is present at ∼15 μM and predominantly bound to albumin, rendering the otherwise highly lipophilic pigment water soluble. Free and albumin-bound bilirubin is able to reduce α-TO· and to inhibit plasma and LDL lipid peroxidation (663), although likely as a secondary line of antioxidant defense behind ascorbate (269). Bilirubin can also protect proteins against oxidation in vitro (662), and when bound to albumin the pigment can protect cells against oxidative damage (1067, 1068). Within cells, bilirubin appears to be present primarily within membranes and at submicromolar concentrations; higher concentrations are cytotoxic. To date, few studies have attempted to provide direct evidence for bilirubin acting as an important antioxidant in vivo, although in rats hyperbilirubinemia attenuates oxidative injury in response to hyperoxia (195).


In addition to aqueous antioxidants, lipid-soluble antioxidants also play an important role in preventing oxidative damage in biological tissues. Tocopherols, that make up vitamin E, and ubiquinols are the major lipid-soluble antioxidants in lipoproteins and cells (93, 233). Indeed, lipoproteins are the primary vehicles for the transfer of vitamin E from the liver to peripheral tissues (467, 957). Lipoproteins also carry carotenoids that have received attention as antioxidants in the context of cardiovascular disease. However, compared with vitamin E and ubiquinols, carotenoids are generally poor antioxidants (111), except for singlet oxygen, an oxidant of presently unknown relevance to atherosclerosis. Also, the nature of the relationship between cardiovascular disease and all, or specific, carotenoids remains controversial (494, 829) so that the following section will be limited to vitamin E and ubiquinol.

I) Vitamin E.

Vitamin E is a nutritional term that refers to a fat-soluble, dietary “factor” essential to permit normal reproduction of rats. Eight different, naturally occurring substances have vitamin E activity in animals: d-α-, d-β-, d-γ-, and d-δ-tocopherols and d-α-, d-β-, d-γ-, and d-δ-tocotrienols. The tocopherols have three asymmetric carbon atoms, giving rise to eight optical isomers, of which RRR-α-tocopherol (formally called d-α-tocopherol) is the most active form of vitamin E. The terms α-TOH and vitamin E are commonly, though incorrectly, used interchangeably. Synthetic “vitamin E,” also called all-rac-α-tocopherol, contains ∼12.5% RRR-α-tocopherol together with the seven other α-isomers that are biologically less active.

Being a fat-soluble molecule, α-TOH tends to localize in membranes and lipoproteins. Indeed, α-TOH is quantitatively and qualitatively the major antioxidant in extracts prepared from LDL (234) and central to the control of radical-induced lipid peroxidation (see sect. iiiC4). As indicated in section iiB1, tocopherols and tocotrienols are excellent scavengers of LOO·, and this is generally thought to be the major antioxidant action of the vitamin (reviewed in Ref. 112). In the case of α-TOH, the rate constant for the reaction with LOO· (reaction 5) is about four orders of magnitude faster than that for the reaction of LOO· with LH (reaction 4) (110). The resulting α-TO· · is relatively nonreactive and, at least in some systems such as micelles, has a long half-life (t1/2 ∼5 min, Ref. 62). However, α-TO· can react further with LOO· or another radical species (R·) to yield nonradical products (NRP in reaction 17): Math(17a) Math(17b)

In the case of LOO· (reaction 17a), 8a-hydroperoxy tocopherones are formed that hydrolyze to tocopheryl quinone epoxides, whereas in the case of R· (reaction 17b), 8a-substituted tocopherone adducts are formed that hydrolyze to tocopheryl quinone (560) (Fig. 7). In addition to scavenging peroxyl radicals, α-TOH can also react with singlet oxygen and the 2e-oxidants HOCl and ONOO. In the latter case, α-tocopheryl quinone is formed via tocopheroxylium cation (Fig. 7).

FIG. 7.

Scheme of major pathways of α-TOH oxidation. Two-electron oxidants such as HOCl and ONOO oxidize α-TOH to the intermediate tocopheroxylium cation (α-TO+), which hydrolyzes to α-tocopherylquinone (α-TQ). Radical oxidants (R·) generate the α-TO· that can further scavenge radicals, to produce 8a-substituted tocopherone adducts, or scavenge LOO·, to produce 8a-hydroperoxy-epoxytocopherones. These hydrolyze to α-TQ and α-tocopheryl quinone epoxides (TQEs), respectively. [Adapted from Terentis et al. (939).]

It follows from the above that one molecule of α-TOH can, in principle, scavenge two molecules of radicals such as LOO·. The extent to which this takes place in vivo is largely unknown. Alternatively, α-TO· produced as a result of initial radical scavenging (reaction 5) may be reduced. As mentioned earlier, ascorbate rapidly (k18 ∼1.5 × 106 M−1 · s−1) reduces α-TO· back to α-TOH.

Several other biological reducing agents, such as ubiquinols (647), bilirubin (663), α-tocopherylhydroquinone (666), caffeic acid, 2-hydroxyestradiol, and epinephrine also reduce α-TO· to α-TOH (1059).

Like ascorbate, α-TOH can reduce transition metals including iron and copper, and this may result in pro-oxidant effects of the vitamin (497). Perhaps more importantly, α-TO· has a finite reactivity with lipids and can abstract hydrogen from LH (reaction 18): Math(18)

While the rate constant for reaction 18 is very low (k19 ∼5 × 10−2 M−1 · s−1) (81), we will see in section iiiC4 that this reaction can become relevant for in vitro LDL oxidation.

In addition to α-TOH, γ-tocopherol is present in lipoproteins but is less antioxidant active than α-TOH and, hence, less able to control radical-induced LDL lipid peroxidation (1051). However, γ-tocopherol may be important in the detoxification of ·NO2 (153, 166), although the physiological significance of this remains unknown. Interestingly, γ-tocopherol, but not α-TOH, has recently been reported to inhibit both carrageenan-induced inflammation in rats and the associated increase in lipid peroxidation and proinflammatory eicosanoids (449). This could be relevant given the importance of inflammation in atherogenesis (see sect. vC).

II) Coenzyme Q10.

Coenzyme Q10 belongs to a family of compounds known as ubiquinones that are fat-soluble molecules synthesized by all animals including humans and that contain a benzoquinone structure with anywhere from 1 to 12 isoprene units. Ubidecaquinone or coenzyme Q10 has 10 isoprene units and is the predominant form of ubiquinone in humans. Coenzyme Q10 is found in virtually all cell membranes, as well as lipoproteins (233). Its ability to accept and donate electrons is critical to its physiological functions, and coenzyme Q10 can exist in three oxidation states: the fully reduced ubiquinol-10 form (CoQ10H2), the ubisemiquinone radical intermediate (CoQ10H·), and the fully oxidized ubiquinone-10 form (CoQ10). Coenzyme Q plays an important part as an electron and proton transfer agent in mitochondrial ATP production and in the maintenance of optimal pH of lysosomes (168).

In its reduced form, ubiquinol is an effective fat-soluble antioxidant (268, 461). The presence of a significant amount of ubiquinol in cell membranes, along with enzymes that are capable of reducing ubiquinone back to ubiquinol, supports the idea that the latter is an important cellular antioxidant (233). Indeed, CoQ10H2 can inhibit protein and lipid oxidation in cell membranes (254). It also provides the first line of lipid-soluble antioxidant defense against human LDL lipid peroxidation (896), although it is present in smaller quantity than α-TOH. Ubiquinol-10 reduces α-TO· (reaction 19) (646) and its ability to interact with α-TOH (433) probably accounts for its strong inhibition of lipid peroxidation. Math(19) Math(20) Math(21)

CoQ10H· can react with O2 to produce CoQ10, H+ and O2· (reaction 20) (674). This reaction increases the formation of O2· and hence the possibility for oxidative stress in the aqueous phase (433), although it provides overall antioxidant protection to lipids as O2· is inefficient in initiating lipid peroxidation. In addition, CoQ10H· may also reduce α-TO· back to α-TOH (905) (reaction 21), resulting in the formation of CoQ10 that does not react with O2. In addition to its interaction with vitamin E, there is some evidence that coenzyme Q-dependent electron transport across the cell membrane can be used to regenerate ascorbate from the ascorbyl radical (22).

E. Redox Reactions in Cell Signaling

There is now a wide appreciation that the concept of oxidative stress in the vasculature has become more complex in recent years. Although the above discussion on ROS focused principally on their capacity to affect oxidative damage, we now know these species have specific roles in the modulation of cellular events. Importantly, in the case of “redox signaling” this is achieved by discrete, localized redox circuitry rather than generalized “oxidative stress” (309). This realization represents a significant departure from traditional views that ROS are simply a by-product of normal oxidative metabolism or a tool through which phagocytes accomplish antimicrobial action. With this paradigm shift comes the challenge of understanding how ROS production is regulated and localized within cells in both normal and pathological circumstances. Current evidence would support a role for ROS as a generalized “injury” response in tissues. This contention is supported by observations that ROS may underlie growth promotion and the response to many repair stimuli such as platelet-derived and epidermal growth factors. Moreover, this role for ROS in the injury response is in keeping with their role in antimicrobial action and cell killing, two functions that often precede tissue repair. Only further investigation will determine if this paradigm for ROS in the vasculature proves true and to what extent these species mediate vascular pathology (see sect. v).

1. Targets of ROS in signal transduction

Intracellular production of ROS is elicited in response to a host of stimuli (Table 5). To differentiate a nonspecific response from signal transduction, several groups overexpressed antioxidant enzymes to scavenge stimulus-induced ROS. Addition of catalase to rat vascular smooth muscle cells scavenges ROS and limits protein tyrosine phosphorylation and cell growth in response to platelet-derived growth factor (923). In addition, tyrosine phosphorylation due to epidermal growth factor or angiotensin II is inhibited by intracellular catalase overexpression (26, 971). These approaches support a role in cell signaling; however, little insight is provided concerning the specific targets of ROS.

View this table:

Generation of reactive oxygen species in vascular cells induced by different stimuli

For ROS to mediate signaling, one would assume that some specific protein modification would be involved. Available evidence supports a role for protein thiol groups in this process as they represent a well-known target for reactive species (Table 6). Certain cysteine residues, particularly those surrounded by basic amino acids, exhibit pKa values in the range of ∼5.0 such that they are not protonated at physiological pH (564, 1105), making them particularly attractive for redox reaction signaling. This situation is illustrated by the bacterial transcription factor OxyR. OxyR positively regulates several target genes involved in the response to ROS, and this activity depends on a critical cysteine residue (Cys-199) (509). This residue is subject to distinct redox modifications (S-OH, S-S, and S-NO), each of which produces specific functional responses (483), lending credence to the notion that redox-mediated thiol modification is one mechanism for signal transduction. Recently, a number of proteins have been identified as targets for ROS-mediated signal transduction.

View this table:

Selected reactive oxygen species and their known biologic targets


Thioredoxin is ubiquitously expressed in endothelial cells and medial smooth muscle cells. It is a major cytosolic protein thiol reductant and appears to be a target for ROS with implications for cell signaling. In its reduced state, thioredoxin forms a complex with apoptosis signal-related kinase-1 in a manner that inhibits kinase activity (792). This interaction between the two proteins is regulated by the redox status of thioredoxin. Oxidation of the sulfhydryl groups of Cys-32 and Cys-35 of thioredoxin by reagent H2O2 or cytokine-stimulated ROS production releases the protein from the kinase leading to kinase activation and the stimulation of downstream targets such as c-Jun amino-terminal kinase and p38 kinase (324, 792). In addition, Cys-62 and Cys-69 represent a second dithiol/disulfide motif implicated in the transient redox regulation of thioredoxin activity (1021), and S-nitrosation of Cys-69 has also been linked to the antiapoptotic activity of thioredoxin (1077).


Aconitase is present in both a mitochondrial and cytosolic form and is involved in the citric acid cycle that catalyzes the conversion of citrate to isocitrate. The enzyme contains a cubane moiety [Fe4S4]2+ consisting of three cysteine-ligated iron residues and the fourth available for interaction with substrate during the catalytic cycle (49). This iron residue is also available for reaction with ROS and RNS, and its loss has been linked to formation of a Fe3S4 cluster and enzyme inactivation (288, 477). Aconitase is inactivated rapidly by O2·, and this has led to the suggestion that inactivation of aconitase may be useful as a sensitive measure for intracellular O2· (287), although H2O2, ·NO, and ONOO also inactivate the enzyme. In intact cells, reactivation of the enzyme is achieved rapidly by iron sulfur cluster reduction and Fe2+ insertion, suggesting that iron sulfur center recycling could serve adapative functions related to cellular redox signaling (286).

Cytosolic aconitase is identical to the iron-regulatory protein 1 that regulates intracellular iron by controlling biosynthesis of the transferin receptor and ferritin via binding to the iron-response element in the 3′-untranslated region of the mRNA for these proteins (224). Loss of iron from the Fe4S4 cluster (and loss of aconitase activity) induces binding of iron-regulatory protein to the iron response element that represses ferritin translation and stabilizes transferrin receptor message. This pathway is subject to redox regulation as cells or intact organs exposed to H2O2 exhibit reduced ferritin synthesis and upregulation of transferrin receptor mRNA (644, 701). The mechanism for this effect is complex, as H2O2 releases Fe from iron-regulatory protein, but such H2O2-treated protein is not able to bind to iron-response elements (700), suggesting some other signaling mechanism for changes in iron status mediated by iron-regulatory protein. This contention is supported by observations that H2O2 treatment of cell lysates does not stimulate binding of iron-regulatory protein to response elements (75). Thus the specific role of ROS in regulating aconitase and, as a consequence, iron-regulatory protein activity remains to be determined.


Guanylyl cyclases are enzymes responsible for the synthesis of the second messenger cGMP that mediates many physiological functions including vasorelaxation and inhibition of platelet aggregation and smooth muscle cell proliferation (427). Soluble guanylyl cyclase is a heterodimeric enzyme consisting of an α- and β-subunit and containing heme as the single prosthetic group that is crucial for binding of and activation by ·NO. In addition to this activation pathway, there is evidence for modulation of the activity of soluble guanylyl cyclase by translocation, cations, adenine nucleotides, allosteric changes, and oxidative modification (272, 705, 789, 1096). With regard to regulation by oxidative processes, early studies showed that the redox state of heme and protein thiols is important (339). The enzyme is reversibly inactivated by the quinoxalin derivative 1H-[1,2,4]oxadiazole[4,3-a]-quinoxalin-1-one (ODQ) via oxidation of heme iron (1106). Guanylyl cyclase has also been reported to undergo reversible inactivation via formation of mixed disulfide bonds (84), while other reports provide evidence for peroxides activating the isolated enzyme (reviewed in Ref. 1009), and fatty acid hydroperoxides the enzyme in intact platelets (102). To reconcile these findings, it has been proposed that reversible oxidation of a single cysteine residue, possibly via glutathiolation, activates whereas oxidation of additional, essential thiol groups inactivates the enzyme (1069).

The in vivo relevance of modulation of soluble guanylyl cyclase activity via redox processes other than ·NO binding remains to be established. However, blood vessels exposed to H2O2 exhibit relaxation that is blocked by methylene blue, an inhibitor of soluble guanylyl cyclase (109). The activity of H2O2 to increase smooth muscle cell cGMP is dependent on catalase and is temporally related to the formation of compound I. The specific mechanism whereby compound I may activate guanylyl cyclase is not yet clear but may involve some change in the oxidation state of heme iron (148). Emerging evidence that H2O2 also acts as an endothelium-derived hyperpolarizing factor suggests that the mechanism outlined above is relevant for the control of vascular tone (598).


Ras GTPases are a family of ∼21-kDa proteins that serve as molecular switches that control many aspects of cellular function including proliferation, motility, differentiation, and death. There is a reversible interaction between RNS and Ras that results in its nitrosylation at Cys-118 and activation via GDP/GTP exchange (517). Among the known Ras effectors, phosphatidylinositol 3′-kinase appears particularly important for propagating the RNS signal (200), and this RNS response is dependent on Ras-mediated Raf-1 activation (199). The modification of Ras at Cys-118 has also been implicated in ROS-mediated Ras activation, particularly in cells exposed to either advanced glycation end products (518) or angiotensin II (4a). Thus Ras appears to be a target of ROS/RNS, and further study will be required to identify the exact role of posttranslational oxidative Ras modification in cell signaling pathways.


Nonreceptor tyrosine kinases, particularly the Src-family kinases, are frequently implicated as targets of ROS. Treatment of T cells with either H2O2 or thiol-modulating agents such as diamide results in the tyrosine phosphorylation of multiple proteins (656), including Src, Lck, and Fyn. It is now widely recognized that Src-family tyrosine kinases are activated by oxidative events (1, 203, 656, 971), and this process is involved in the activation of many downstream kinases such as the mitogen-activated protein kinase (MAPK) family, Akt, protein kinase C, and the epidermal growth factor receptor kinase (for review, see Ref. 332). In addition, ROS also activate other tyrosine kinases, such as c-Abl, pyk2, Jak2, and others (510, 853, 927). This activation is likely indirect, perhaps through inhibition of tyrosine phosphatases (see below). A comprehensive review of all tyrosine kinase pathways sensitive to ROS-mediated activation is beyond the scope of this review; however, the interested reader is directed to recent excellent reviews on the subject (332, 767).


Protein tyrosine phosphatases are important for the signaling mediated by cellular tyrosine kinases, because these kinases are under a tonic inhibition by the phosphatases. Protein tyrosine phosphatases share an active site motif consisting of an invariant cysteine and an arginine separated by five residues (I/V-C-X-X-G-X-X-R-S/T, where X is any amino acid) (35). This sequence motif confers a low pKa environment that facilitates the cysteine residue to function as a nucleophile but also renders it susceptible to oxidation. Consistent with this fact, treatment with H2O2 leads to inactivation of many phosphatases via cysteine oxidation to a sulfenic acid, both with the isolated protein (136, 198, 532) and in cells in response to epidermal growth factor (532). Recent data suggest the sulfenic acid moiety in protein tyrosine phosphatase 1B is converted readily into a sulfenyl-amide species, in which the sulfur atom of the catalytic cysteine is covalently linked to the main chain nitrogen of an adjacent residue (793). This is accompanied by large conformational changes in the catalytic site that inhibit substrate binding and that protect the active-site cysteine residue from irreversible oxidation to sulfonic acid and permit redox regulation of the enzyme by promoting its reversible reduction by thiols.

These studies suggest that at least part of H2O2 signaling is mediated through the inactivation of protein tyrosine phosphatases. Indeed, in cultured cells, stimulation with platelet-derived growth factor leads to reversible inactivation of the SH2 domain-containing protein tyrosine phosphatase (SHP-2) that requires its association with the platelet-derived growth factor receptor (621). Moreover, Rac-dependent Rho GTPase inactivation now appears to be due to NADPH oxidase-mediated inactivation of a low-molecular-weight protein tyrosine phosphatase (669). In addition to inactivation, H2O2 may also alter the cellular content of protein tyrosine phosphatases. In particular, treatment of HeLa cells with H2O2 is characterized by rapid degradation of the Cdc25C (807), a phosphatase involved in the cell cycle progression and checkpoint control. This effect of H2O2 is due to cysteine disulfide formation in Cdc25C, as it can be recapitulated by mutation of either cysteine residues at positions 377 and 330 in the protein. Thus H2O2 may impair protein tyrosine phosphatase activity through either enzymatic inactivation or protein degradation.

Although cysteine oxidation plausibly explains protein modification by H2O2, one must also consider potential cellular means by which this protein modification is controlled. Oxidation of cysteine beyond sulfenic acid is generally considered to be irreversible (156), a situation poorly suited for temporal control of any signaling event. However, cells appear to have access to several options to deal with this situation. As indicated above, one solution to this problem is formation of an intramolecular sulfenyl-amide bond (793). In addition, a recent study showed that at least in the case of peroxiredoxin, oxidation of cysteine to its sulfinic acid is a reversible process via a sestrin-dependent pathway (103, 1062). Furthermore, glutathiolation via a redox-sensitive mechanism offers an alternative solution (39). Indeed, the process of glutathiolation is now well recognized and has been implicated in the posttranslational modification of many proteins in response to ROS/RNS (for review, see Ref. 490). Reversibility of the oxidation-mediated protein modification can be achieved via the action of glutaredoxin (39). It remains to be seen if the “regeneration” of reduced protein cysteines is also a regulated process. In contrast to sulfenyl-amide bonds and glutathiolation, formation of intra- and intermolecular sulfinamide bonds, as can be observed upon exposure of proteins to HOCl, does not appear to be a reversible modification (750).

Although inactivation of protein tyrosine phosphatases by H2O2 may, in part, explain redox-sensitive signaling, many questions remain. For example, little is known about the specific mechanisms involved in H2O2-mediated protein tyrosine phosphatase inactivation. Although reagent H2O2 inactivates multiple forms of protein tyrosine phosphatases, engagement of platelet-derived growth factor with its receptor specifically inactivates SHP-2 (621), and the basis for this selectivity is not clear. As is the case for other signaling microdomains, the source(s) of ROS stimulated by ligand engagement might be expected to bear some spatial or vectorial arrangement to their intended targets. Although many features of H2O2 as a signaling molecule parallel that of ·NO, we have little knowledge at present of how H2O2 may function as a paracrine mediator.


Phase 2 genes encode for proteins including glutathione reductase and heme oxygenase-1 that protect against the damage of electrophiles and ROS, and many of these genes are regulated by upstream antioxidant response elements that are targets of the leucine zipper transcription factor Nrf2. Under basal conditions, a homodimer of Keap1, a recently identified, cysteine-rich protein associated with the actin cytoskeleton, binds very tightly to Nrf2 so that the transcription factor is anchored in the cytoplasm, thereby repressing the activity of Nrf2 by targeting it for ubiquitination and proteasome degradation (442, 888). Inducers of phase 2 enzymes disrupt the Keap1-Nrf2 complex. This allows Nrf2 to translocate to the nucleus, where it binds (in heterodimeric combinations with other basic leucine zipper proteins) to antioxidant response elements of phase 2 genes and accelerates their transcription. Inducers of phase 2 genes belong to nine structurally highly diverse classes of chemicals that share only a few common properties (206), including the capacity to modify cysteine residues.

The mechanism underlying this induction has been clarified recently (1007). Inducers of the phase 2 response interact with specific thiol groups of Keap1 eventually causing the formation of an intermolecular disulfide bond, most likely involving Cys-273 from one and Cys-288 from the other Keap1 molecule. This results in conformational change that renders Keap1 unable to bind to Nrf2 so that it can translocate into the nucleus to enhance phase 2 gene transcription. The Keap1-Nrf2 system may be unique in that it depends initially on different types of chemical modification of cysteine thiols and that the modifying agents then appear to be displaced by intermolecular sulfhydryl disulfide interchange to lead to a covalent disulfide dimer of Keap1 (1007).

I) Nuclear factor κB.

Within the vessel wall, the transcription factor nuclear factor κB (NF-κB) plays a critical role in the regulation of inflammatory and immune response genes relevant to atherosclerosis. Normally, NF-κB is restricted to the cytoplasm through an interaction with its inhibitor, I-κB. Activation of NF-κB is accomplished through phosphorylation and ubiquitination of I-κB (52), leading to translocation of NF-κB to the nucleus and permissive phosphorylation of Ser-276 on the p65 subunit (302). DNA binding of NF-κB appears to require a reducing environment in the nucleus (600), possibly provided by nuclear translocation of thioredoxin (401) and redox factor-1 (1070). An oxidizing stimulus in the cytoplasm appears to be associated with I-κB degradation (549), and it has been proposed that H2O2 activates NF-κB in certain cell lines via this mechanism (816). However, the contribution of redox regulation in NF-κB activation is still subject to intense debate. Troubling aspects include many conflicting reports, reliance on indirect support such as the use of nonspecific agents like the copper chelator pyrrolidine dithiocarbamate as “antioxidants,” and the lack of specific targets for redox-mediated posttranslational modification (reviewed in Ref. 549). Also, modulation of intracellular production of O2· via Rac/NADPH oxidase does not mediate NF-κB signaling, but rather lowers the magnitude of its activation (366). However, I-κB contains redox-sensitive methione residues such as Met-45, the oxidation of which can be achieved by 2e-oxidants such as taurine chloramines, and this leads to inhibition of NF-κB activation (464). Also, H2O2 promotes phosphorylation of human ribonuclear protein-U (903) that regulates specific SCF (Skp-1/Cul/F box) family ubiquitin ligase involved in I-κB degradation (189). These observations lend plausibility to the notion that ubiquitination of I-κB may be a redox-sensitive step for NF-κB activation (549).

The notion that NF-κB plays a role in atherosclerosis has received much attention, as activation of this transcription factor in vascular cells such as endothelial cells, smooth muscle cells, and macrophages leads to or enhances the expression of several genes, including those encoding for certain cytokines, leukocyte adhesion molecules, matrix metalloproteinases, cyclooxygenase-2, and inducible NOS (163, 975). Thus activation of NF-κB can promote the recruitment and activation of inflammatory cells in the vessel wall. Indeed, activated NF-κB has been identified in endothelial cells, smooth muscle cells, and macrophages in human atherosclerotic lesions (74, 83), where it may contribute to the dysregulation of vascular smooth muscle cell function (74, 654). Also, NF-κB activation is seen in atherosclerosis-prone mice in response to a high-fat, high-cholesterol diet (552). What is less clear is how precisely NF-κB is activated during atherogenesis, although induction of inflammatory genes and activation of NF-κB-like transcription factors cosegregate with aortic atherosclerotic lesion formation in mice (553). In vitro several pathological stimuli, including oxidized LDL (552) and microorganism components such as Chlamydia pneumonia (495), are able to activate the transcription factor.

II) AP-1.

Another transcription factor thought to be redox-sensitive is AP-1, a member of the basic leucine zipper family of proteins that plays an important role in the regulation of growth, differentiation, and stress adaptation of cells. AP-1 is a heterodimer consisting of c-Fos and Jun, typically c-Jun. The activation of AP-1 occurs at multiple levels, including changes in Fos and Jun mRNA, effects on Fos and Jun protein turnover, posttranslational modification of Fos and Jun, and interaction with other transcription factors (reviewed in Ref. 837). With respect to redox regulation of AP-1 activity, the first and third of these mechanisms appear most relevant and involve activation of the MAPK cascade (144). Whereas growth factor-induced AP-1 activation typically involves the extracellular signal-regulated kinase subgroup of MAPK, oxidative stress and cytotoxic stress principally mediate AP-1 activation through the c-Jun amino-terminal and p38 kinase pathways (144). In particular, c-Jun amino-terminal kinase is activated by H2O2 leading to both c-Jun phosphorylation with concomitant increased AP-1 transactivation (146) and activation of ATF2, which enhances transcription of c-Jun (342). The two major pathways of H2O2-induced AP-1 activation involve MAPK stimulation through either apoptosis signal-regulating kinase 1 (ASK1) (952) or inhibition of MAPK phosphatases (139) (see above). Thus, in some cases, transcription factor activation is a downstream target of redox-sensitive signal transduction.

A number of other transcription factors are redox sensitive and all contain critical cysteine residues that dictate, in large part, their DNA binding (reviewed in Ref. 145). Thus far, a comprehensive mechanism that explains the redox sensitivity of these transcription factors is not available. Although in many cases a chemically reduced state of the respective cysteine residue facilitates DNA binding by these transcription factors, the specific nature of their contribution to redox-sensitive signaling is not yet clear. Moreover, we are only beginning to develop paradigms on how specific redox-related signals in the cytosol may translate into DNA binding of transcription factors. Thus considerable work is needed to link specific redox-related events into a coordinated genetic program.

2. Targets for RNS in cellular signaling

It is now known that a number of RNS participate in cell signaling within the vasculature (Table 7) as well as many other tissues (for review, see Ref. 432). The prototypical RNS involved in cell signaling is ·NO produced by NOS. Once synthesized, ·NO may interact with a variety of species yielding its oxidation or reduction, and resulting in the production of a spectrum of RNS (Tables 1, 2, and 7). Let us now turn our attention to the targets of these RNS that are relevant to the vasculature.

View this table:

Selected reactive nitrogen species implicated in cell signaling

To consider targets of RNS, let us focus initially on ·NO reactions with proteins. Initially, it was thought that all the biologic activity of ·NO was readily attributable to its diffusion and reaction with a single target, the soluble isoform of guanylyl cyclase (638). Contemporary understanding of ·NO biology lies in stark contrast to this concept. We now know that ·NO has a number of relevant biologic targets that are dictated, in part, by both its site of synthesis, its relative concentration, and the availability of coreactants. At physiologically relevant low nanomolar concentrations, the predominant reactions of ·NO are with heme and heme-copper centers. The reaction with O2· will not be considered here, despite the fact that kinetic considerations alone indicate it to be a favored reaction, and its product, ONOO, has properties distinct from those of ·NO (see sect. iiC3).

Perhaps the best-characterized heme target for ·NO is the soluble isoform of guanylyl cyclase. This heme protein binds ·NO with a bimolecular rate constant of ∼108 M−1 · s−1 (31) permitting its effective competition with other ·NO targets and facilitating the well-characterized increase in cGMP in response to ·NO (428). Binding of ·NO to the heme results in an initial six-coordinated NO-Fe2+-histidine complex. Subsequent breakage of the histidine-to-iron bond leads to formation of a five-coordinated nitrosyl-heme complex that initiates a conformational change resulting in activation of the enzyme (reviewed in Ref. 272). The resulting increase in cGMP is largely responsible for vasodilation and the inhibition of platelet aggregation and proliferation of smooth muscle cells (561, 638). Another important target for ·NO at physiological concentrations is the heme-copper protein cytochrome oxidase (158). Binding of ·NO inhibits the oxidase, and this is associated with an improvement in the efficiency of energy metabolism in the mitochondria. As a consequence, endogenously produced ·NO from the endothelium has considerable influence over tissue oxygen consumption (839). Compared with ·NO, ONOO is a poor agonist for soluble guanylyl cyclase (936), and it irreversibly inhibits mitochondrial respiration through the release of iron from iron sulfur centers (747).

When considering the role of ·NO in signal transduction, it is important to recognize that multiple isoforms and splice variants of NOS exist in discrete cellular compartments permitting the enzymes to fulfill distinct functions. For example, both skeletal and cardiac muscle cells express all three different isoforms of NOS in several different cellular compartments. Localization of eNOS to the plasmalemma facilitates regulation of β-adrenergic-mediated force production (37), whereas NOS-derived ·NO controls respiration at the level of cytochrome c oxidase (see above). The presence of NOS in the sarcoplasmic reticulum affords control of calcium homeostasis that, paradoxically, enhances cardiac contractility (37). Finally, the specific functions of constitutive and inducible isoforms of NOS that exist in the cytosol remain unknown. Thus emerging data indicate the site of ·NO generation has important implications for ·NO-mediated signal transduction, consistent with the growing appreciation that signaling takes place within the confines of subcellular compartments that are integral to both the specificity and phenotypic response of such signals.

The concentration of ·NO produced within a cell also has significant implications for the ultimate signals produced. For example, under certain pathological conditions such as inflammation, upregulation of inducible NOS affords production of ·NO at low micromolar concentrations. In this concentration range, ·NO competes effectively with SOD for O2·, facilitating formation of ONOO and other RNS. This increase in tissue RNS, a condition termed nitrosative stress (364), leads to the modification of cellular targets such as thiols, proteins, and lipids, many of which have implications for cellular signaling.


The production of RNS within cells has important implications for cellular thiol status. Within cells, the most abundant low-molecular-weight thiol is GSH with concentrations in the range of 1–5 mM (89). This abundance of GSH facilitates the formation of its nitrated (GSNO2) and nitrosated (GSNO) forms that may serve, in part, as tissue “buffers” of ·NO (1045). The formation of nitrosated glutathione has been observed in vivo (294), although the exact concentrations that occur within cells remain a matter of question. Recent data suggest that S-nitroso species are present in tissues at concentrations of ∼40 nM (774) comparable to that of ·NO itself (578). These GSH adducts possess biologic activity that is comparable to ·NO, including vasodilation and inhibition of platelet aggregation (430, 431, 618). The reaction of ·NO with GSH is by no means exclusive. Indeed, low-molecular-weight thiol adducts of ·NO have been demonstrated with a variety of species such as cysteine, homocysteine, and synthetic species including N-acetyl cysteine and penicillamine (for review, see Ref. 405). The abundance of GSH over other low-molecular-weight thiols suggests its role in ·NO biology is likely more profound than for the other species. Despite this, however, important questions remain. For example, we know precious little about the relative contribution of ·NO adducts of GSH with regard to the biologic activity of ·NO, and to what extent they contribute to posttranslational protein modification (see below).


Given that thiols represent a prominent biologic target for RNS (Table 7), it is perhaps not surprising that protein cysteine residues are also subject to modification mediated by RNS. This concept was initially brought to light through the demonstration that albumin and tissue-type plasminogen activator may form S-nitrosated species (871, 874, 875) with biologic activity reminiscent of authentic ·NO (472). Evidence that S-nitrosation of protein cysteine residues is a motif for ·NO-related signaling is now pervasive in nature. For example, the function of L-type calcium channels in the heart is subject to modulation through S-nitrosation (122), and the function of the ryanodine receptor is readily modified by receptor S-nitrosation at Cys-3639 (922). S-Nitrosation of thioredoxin is thought to activate apoptosis signal-regulating kinase 1 (920). Indeed, ·NO synthesis by many cells results in S-nitrosation of numerous proteins (328), although the particular functional consequences of these phenomena remain to be determined. This widespread utilization of S-nitrosation throughout nature has prompted considerable speculation that it represents a conserved method of signaling throughout biologic systems (217).

Initially it was thought that any protein thiol species would be subject to modification by RNS; however, just as some cysteine residues have a predilection for oxidative modification (see above), it appears the same holds true for modification via RNS (872). Such implied selectivity supports the notion that this phenomenon may represent another means for allosteric control of protein function (634). The specific mechanism(s) that control protein nitrosation are not yet clear, but using the paradigm of protein phosphorylation, some investigators have proposed consensus motif(s), such as adjacent aspartate or glutamate residues, for this process (872, 873). Conceptually, this proposal is attractive as protein phosphorylation is controlled, in part, by consensus motifs that dictate targets for kinases and phosphatases. With regard to the latter, one might expect enzymatic mechanism(s) for the regeneration of cysteine residues from their S-nitrosated forms.


The chemical modification of protein thiols in the setting of ROS and RNS does not occur in isolation. Considering its relative abundance, it is not surprising that GSH functions as a prominent coreactant for protein thiol modification in the face of ROS and RNS. Considerable evidence, dating back some 20 years, indicates that GSH is involved in the dynamic regulation of protein function through the reversible formation of mixed disulfides involving protein cysteinyl residues (303). Early data demonstrating protein S-glutathiolation were largely derived from the exposure of cells to toxic levels of oxidants or activation of an inflammatory response from neutrophils (141), macrophages (961), and monocytes (764). Such studies readily demonstrated a number of S-glutathiolated proteins that spanned a broad range of cellular functions including metabolism (glyceraldehyde-3-phosphate dehydrogenase and creatine kinase) (164, 764), the cytoskeleton (actin) (141), and antioxidant protection (SOD and glutathione transferase) (180, 812) to name a few. For a more complete list of S-glutathiolated proteins, readers are directed to an excellent review (490).

With regard to cell signaling in the context of ROS and RNS, the potential formation of S-glutathiolated protein adducts has several attractive features. First, S-glutathiolation of protein cysteine residues protects against higher oxidation states of the protein thiol, thereby preserving the reversibility of this type of modification (39). Second, reduced protein thiols can be regenerated from their S-glutathiolated forms enzymatically through the action of protein disulfide isomerase, mitochondrial glutaredoxin, or thioredoxin (451, 458). For example, the production of ROS in response to epidermal growth factor is temporally related to glutathiolation of protein tyrosine phosphatase 1B that is reversed by glutaredoxin (38, 39). The speculation that S-glutathiolation protects proteins from irreversible oxidation and loss of function is supported by data in yeast (331) and cardiac tissue (219) where glutathiolation of glyceraldehyde-3-phosphate dehydrogenase preserves its function under conditions of high oxidative stress. Protein S-glutathiolation has also been implicated in the control of ubiquitination (445), the binding of transcription factor c-Jun to DNA (491), and sarcoplasmic Ca2+-ATPase activity (994).

3. Summary

From the preceding paragraphs, it should be clear that ROS and RNS are routine products found in biological tissues. Although once thought to exist principally in the setting of inflammation, it is now clear that they may also mediate a number of physiological responses that are pervasive in nature. Within the vasculature specifically, ·NO is a major mediator of normal vascular homeostasis manifest as the maintenance of vascular tone, blood fluidity, and leukocyte entry into tissues. Roles for ROS are best established for H2O2 that has been implicated in the vascular injury response and, perhaps, various forms of hypertrophy for both smooth muscle cells and cardiac myocytes. In later sections we will touch on the notion that production of ROS and RNS is often increased in the setting of vascular pathology and that increases in both oxidative and nitrative stress may lead to significant phenotypic changes of vascular and inflammatory cells that contribute importantly to human disease.

F. ROS and Cell Proliferation

The accumulation and hypertrophy of smooth muscle cells are characteristic of atherosclerotic, restenotic, and hypertensive vascular diseases. Many pathological vascular conditions also involve to some degree the proliferation of endothelial cells and fibroblasts. Overall, the net balance between proliferation and apoptosis/necrosis determines the extent of cell growth. A role of ROS in the control of growth of vascular cells is supported best for H2O2 mediating a proliferative phenotype in vascular smooth muscle cells (24, 758, 761, 1097). In fact, H2O2 may be required for smooth muscle cell survival, as overexpression of catalase inhibits proliferation and can induce apopotosis (100). Low concentrations (i.e., <10 μM) of added H2O2, but not O2·, are also mitogenic to or enhance the survival of other cells (107) including endothelial cells (904). ROS derived initially from mitochondria and then NAD(P)H oxidase have also been suggested to be responsible in the autonomous proliferative response of endothelial cells to hypoxia (810). In contrast to these low concentrations, H2O2 at >50 μM, corresponding to intracellular concentrations of >1 μM (76), can cause cultured cells to undergo growth arrest, apoptosis, or necrosis (550).

The molecular mechanism(s) by which low concentrations of H2O2 stimulate growth of vascular cells remains poorly understood. Unlike the situation in bacteria with the transcription factor OxyR (1107) a similarly acting “sensor” for H2O2 has not been identified in mammalian cells. In fact, human cells respond to H2O2 in a complex manner, with a multitude of genes affected. For example, cells with ∼10-fold increase in cellular H2O2 concentration due to overexpression of Nox1 show increased expression of some 200 genes, whereas overexpression of Nox1 plus catalase reverts ∼70% of the altered genes to normal (21). Many of the genetic changes seen are in proteins related to cell cycle, signal transduction, and transcription rather than antioxidant defense, indicating that the response triggered by H2O2 is specific and different from the stress response seen in bacteria. For example, in vascular smooth muscle cells in vitro, cell growth-enhancing, low concentrations of H2O2 activate multiple intracellular proteins and enzymes, including the epidermal growth factor receptor, c-Src, p38 MAPK, Ras, and Akt/protein kinase B (332). In contrast, intermediate concentrations of added H2O2 (100 μM) inhibit serum-induced progression through the cell cycle via a decrease in cyclin A expression and cyclin-dependent kinase 2 activity, as well as upregulation of p21 and p53, whereas higher concentrations of H2O2 (≥500 μM) cause apoptosis of vascular smooth muscle cells (201).

It is known, however, that in addition to the potential targets described above, many growth factor-mediated responses also involve the generation of ROS as a component of their signaling, although at present little detail is known of the molecular mechanism(s) underlying these responses. For example, growth factors including platelet-derived growth factor (923) and epidermal growth factor (26) trigger H2O2 production. Similarly, vascular endothelial growth factor-mediated proliferation and migration of endothelial cells appear to depend on ROS production (2, 160, 972).

Growth factor receptors may be activated through both ligand-dependent and -independent means, although these two processes are distinct (146, 765). This distinction has functional implications, since ligand-induced epidermal growth factor receptor activation is regulated, in part, through receptor internalization and degradation, whereas H2O2-mediated receptor transactivation is not (765). The latter fact has been cited as one potential mechanism for the relation between ROS and unregulated cell growth (765).

The specific extent to which ROS contribute to vascular cell growth remains to be determined. Teleologically, one could envision cell growth mediated by ROS as part of a spectrum that encompasses the tissue injury response. Indeed, the response to cellular injury is characterized by the recruitment of inflammatory cells such as neutrophils and macrophages. These cells are needed to “clean” the injured area of necrotic debris through phagocytosis, a process associated with the production of ROS. This burst of ROS also appears critical for the clearance of neutrophils and macrophages from damaged tissue (1101), a prelude to the resolution of inflammation and formation of scar tissue. Thus it is not surprising that ROS, an important component of inflammation, would be involved in the regrowth of tissue after damage.

G. Redox Reactions in Cell Death

There is now a growing body of evidence that ROS play a role in both cellular necrosis and apoptosis. It has long been known that ROS in sufficient high doses will lead to cell killing, as would be expected with any compound. More recently, however, it has become clear that ROS generated within cells play an important role in cellular fate. In particular, the role of ROS in apoptosis, or programmed cell death, has been examined in great detail.

Early hints that ROS may be involved in apoptosis were derived from apoptosis induced by the cytokine tumor necrosis factor-α. In cytokine-exposed cells, mitochondria undergo prominent morphological changes that coincide with cell death (601), and reducing ambient oxygen levels affords some protection (660). The involvement of mitochondrial ROS in the response to this cytokine is supported by observations that inhibition of mitochondrial respiration at complex I limits apoptosis induced by tumor necrosis factor-α (318). Apoptosis from tumor necrosis factor-α may also be inhibited by Mn-SOD, further implicating the involvement of ROS in this process (247).

The role of mitochondrial ROS extends to a form of programmed cell death known as anoikis that results from the loss of cell attachment to extracellular matrix (273). This process is associated with a significant increase in intracellular ROS that appears derived from mitochondria (544). Mitochondrial production of ROS results in the activation of components of the apoptotic pathways, such as caspases and c-Jun amino-terminal kinase (JNK), and inhibition of mitochondrial ROS generation ameliorates anoikis (544).

Although the data outlined above point to mitochondrial ROS as critical for apoptosis, other means of programmed cell death require ROS from sources outside the mitochondrion. Neutrophils are recruited to sites of inflammation, and their clearance from these sites requires apoptosis. Neutrophil apoptosis triggered by phagocytosis of complement- or immunoglobulin-opsonized particles is dependent on ROS generation, as cells deficient in active NADPH oxidase do not undergo apoptosis in response to phagocytosis (167). It appears that ROS produced during phagocytosis cause activation of caspase-8 and subsequent induction of the extrinsic apoptotic pathway (1101) via a mechanism that remains to be determined.

Damage to DNA has long been known to be important for carcinogenesis, and recent data implicate DNA damage in the process of atherosclerosis (30). Cellular ROS are now implicated in the apoptotic response to DNA damage. For example, DNA damage is associated with rapid activation of p53 that subsequently promotes cellular apoptosis, in part, via ROS (724). Consistent with this observation, p53 overexpression results in increased cellular ROS generation and apoptosis (453). In this regard, a transcription factor known as REDD1 has been identified as one important downstream target of p53 (230). The upregulation of REDD1 in response to DNA damage is associated with increased production of and an increased sensitivity to ROS, by presently unknown mechanism(s).

H. Redox Reactions in Platelet Function

1. Platelets and cardiovascular disease

The critical role for platelets in the acute manifestations of atherosclerosis is supported by data that platelet inhibition yields a reduction in myocardial infarction and stroke (161). To prevent inappropriate thrombus formation and vascular events, platelet function is strictly regulated in the vasculature. In particular, a number of autocrine and paracrine factors are secreted by platelets and the adjacent endothelium to prevent platelet adhesion and aggregation. Many of these control mechanisms are under redox regulation and, in many cases, involve ROS and RNS.

2. ROS and platelet aggregation

Platelet aggregation is associated with considerable oxidative stress manifest as a burst of oxygen consumption (92) and an increase in GSSG (106). This increase in GSSG arises, in part, from ROS such as O2· (260, 585) and H2O2 (912) that are produced during the aggregation. Implicated sources for these ROS produced during aggregation include cyclooxygenase, lipoxygenase (856), and even NADPH oxidase (503, 825). Platelets may also be exposed to ROS that arise from the vascular wall, particularly in the setting of atherosclerosis or its established risk factors such as hypercholesterolemia (473, 680) and diabetes (940). This association between platelet activation and ROS production has driven considerable investigation into the potential role(s) of these species in regulating platelet function.

Early investigation of ROS and platelet aggregation focused on O2·. Platelets exposed to ROS from xanthine/xanthine oxidase undergo aggregation and serotonin release (356). Chronic exposure of platelets to low fluxes of O2· also appears to sensitize the platelet response to thrombin (356), suggesting there is a synergy between receptor-mediated aggregation and ambient O2·. Consistent with these observations, SOD inhibits thrombin-stimulated platelet adhesion and aggregation (796). The mechanisms underlying these observations are not entirely clear but may reflect a stimulatory effect of O2· on platelet thromboxane production that, in turn, enhances aggregation (636).

The effect of H2O2 on platelet aggregation has been mixed. The presence of H2O2 at low micromolar concentrations inhibits the aggregation (912), in part, by raising intracellular cGMP (15). Although not studied in detail in platelets, this effect of H2O2 on intracellular cGMP may relate to findings in vascular smooth muscle in which H2O2-mediated formation of catalase compound I activates guanylyl cyclase (109). Higher concentrations of H2O2, in the millimolar range, appear to stimulate platelet aggregation (776). This observation may be physiologically relevant to certain platelet agonists. For example, collagen appears to stimulate considerable platelet H2O2 production that may be involved in platelet arachidonic acid mobilization and thromboxane production (719). In fact, H2O2 has also been implicated in platelet “priming” that facilitates tyrosine phosphorylation of the fibrinogen receptor that is ultimately required for aggregation (436). This latter contention is supported by data that ultraviolet irradiation enhances platelet aggregation through the upregulation of fibrinogen binding sites as a direct result of activation of protein kinase C mediated by ROS (983). Thus the role of ROS in modulating platelet function is dependent on the particular type and concentration of ROS involved.

Part of this ambiguity concerning ROS and platelet function may result from an incomplete understanding of platelet aggregation and the factors that influence this process. It may turn out that investigation into the autocrine aspects of platelet aggregation has been misplaced. Most early reports focused on the products of platelet aggregation, such as O2·, and how these products may alter the aggregation response. However, in vitro, platelet aggregation is an “all-or-none” type of response in which all the platelets are activated simultaneously. In the setting of an all-or-none event, it is unlikely that factors released from aggregating platelets will have material impact on the same population of platelets that have been committed to aggregation. Rather, it is much more likely that autocrine factors elaborated from activated platelets will have a more profound impact on subsequent populations of platelets that approach a site of arterial injury. This latter process is known as “platelet recruitment,” and its recognition has shed important light onto the autocrine behavior of platelet aggregation.

If one studies both platelet aggregation and the formation of ROS in real time, the release of ROS and RNS occurs near the end of the aggregation process (260). This observation reinforces the notion that reactive species from platelets may have a more profound impact on platelet recruitment rather than aggregation. The relation between ROS and platelet recruitment has recently been examined. In washed platelets, the release of O2· in response to collagen appears to be due to NAD(P)H oxidase, and inhibition of this signal is associated with a marked reduction in platelet recruitment to a growing thrombus (503). Other agonists are also associated with platelet O2· production. For example, ADP-stimulated platelets exhibit O2· production that may, in part, be due to eNOS, as a large part of this signal is lost in platelets from mice lacking the NOS-3 gene (261). In the context of atherosclerosis and its associated increase in ambient O2·, one would expect platelet recruitment to also be enhanced by the vascular flux of O2·.

3. RNS and platelet aggregation

With regard to a role of RNS, the effect of ·NO on platelet aggregation is the best characterized. This information is derived from long-standing knowledge that an endothelium-derived product inhibits the adhesion and aggregation of platelets to the endothelial surface (23). Once the identity of endothelium-derived relaxing factor was established as ·NO, it was rather straightforward to demonstrate that endothelial-derived ·NO limits the adhesion and aggregation of platelets (748). These observations were completely consistent with previous data indicating that ·NO donors inhibited platelet aggregation in a cGMP-dependent manner (618). A large body of work has demonstrated that authentic ·NO activates the platelet-soluble isoform of guanylyl cyclase, increasing cGMP and resulting in reduced intracellular calcium (619) and inactivation of the thromboxane receptor (1011). Both of these effects act to inhibit the aggregation and adhesion of platelets.

In addition to ·NO, ONOO has been investigated with regard to its effects on platelets. In fact, the action of ONOO is complex. In washed platelets, ONOO induces aggregation and even reverses any inhibitory effects of ·NO donors (637). In contrast, the presence of plasma transforms ONOO into an inhibitor of platelet aggregation (637). These divergent results are likely due to ONOO reaction with plasma constituents. In the presence of proteins and low-molecular-weight thiols, ONOO is a potent nitrosating agent that can form S-nitrosothiols (45) that, as discussed above, possess biologic activity similar to authentic ·NO (430).

Up to this point, we have restricted our discussion to RNS that are external to the platelet. However, platelets themselves are capable of producing RNS as they contain isoforms of NOS (615, 801). During the aggregation response, platelets produce ·NO (579) that has modest autocrine implications, as inhibition of platelet ·NO synthesis by treatment with NG-nitroarginine methyl ester enhances aggregation only by ∼10–15% (262). In contrast, the recruitment of subsequent platelets is markedly enhanced if aggregation-induced ·NO production by platelets is inhibited (262). These observations may be clinically relevant as the bioactivity of platelet-derived ·NO is also altered in the setting of atherosclerosis and its risk factors. For example, patients with acute coronary syndromes have impaired platelet ·NO production compared with individuals without active atherosclerosis (264). In terms of atherosclerosis risk factors, impaired platelet ·NO bioactivity has been described in association with essential hypertension (116) and smoking (425). Thus considerable data indicate that platelet ·NO production is involved in the regulation of platelet aggregation and, more importantly, platelet recruitment to a growing thrombus.

One implication of the data demonstrating platelet-derived ·NO production involves its reaction with O2·. In particular, data demonstrating that O2· enhances platelet aggregation and recruitment must be interpreted with some caution. Because platelets produce ·NO during aggregation, and platelet-derived ·NO inhibits platelet recruitment (262), it is possible that the major effects of O2· on platelets is related to its ability to “quench” the bioactivity of ·NO (44). Indeed, treatment of aggregating platelets with SOD enhances platelet-derived ·NO, and inhibition of ·NO production boosts aggregation-induced O2· production (259, 260). These observations underscore the notion that ·NO bioactivity is tightly linked to ambient levels of O2·.


A. Original Hypothesis

In 1989, Steinberg et al. (882) proposed the original oxidative modification hypothesis of atherosclerosis, based on the hypothesis that oxidation represents a biologic modification analogous to the chemical modifications discovered by Brown and Goldstein that give rise to a foam cell (see sect. iiiB1). Accordingly, oxidized LDL contributes to atherogenesis by 1) aiding the recruitment of circulating monocytes into the intimal space, 2) inhibiting the ability of resident macrophages to leave the intima, 3) enhancing the rate of uptake of the lipoprotein leading to foam cell formation, and 4) being cytotoxic, leading to loss of endothelial integrity (743). The latter property was proposed to provide the link between fatty streak formation due to lipid infiltration and the progression of the fatty streak to more advanced lesions according to the response-to-injury hypothesis of atherogenesis (882). As such, a key feature of the oxidative modification hypothesis is that it no longer presupposed the loss of endothelial cells as an initiating event in atherogenesis.

B. Evidence in Support of the LDL Oxidation Hypothesis

1. LDL does not support foam cell formation

Although macrophages were identified as the predominant cell type that gives rise to foam cells, early atherosclerosis research was hampered by observations that LDL particles did not appear to be atherogenic in vitro as macrophages incubated with LDL failed to internalize excess lipoprotein-cholesterol due to downregulation of the LDL receptor (316). Thus foam cell formation did not appear to be mediated by the LDL receptor, a fact consistent with the occurrence of atherosclerosis in patients with familial hypercholesterolemia, who lack functional LDL receptors. In a search for alternative receptors, Goldstein et al. (317) observed that acetylation of LDL leads to extensive macrophage cholesterol uptake and foam cell formation. This phenomenon is mediated by a saturable, specific receptor later termed the “acetyl-LDL receptor” (317). It is now known that this receptor is one of many so-called “scavenger receptors” present on macrophages and other cell types (501).

The original description of the acetyl-LDL receptor did acknowledge that there was no known means of LDL acetylation in vivo (317). However, Brown and Goldstein did speculate that other, as yet unknown, modifications of LDL could facilitate its recognition by this acetyl-LDL receptor. Approximately two years later, Henriksen et al. (395) found that LDL incubated with endothelial cells did serve as a ligand for macrophage foam cell formation. Considerable work has since established that several receptors on macrophages and other cells function as “scavenger” receptors (500). The original acetyl-LDL receptor has been identified and exists in two forms known as scavenger receptors A1 and A2 (493). Other receptors such as CD68, CD36, SR-B1, and LOX-1 are also known to possess properties similar to the classic scavenger receptors (500).

2. Oxidative modifications in atherosclerotic lesions

In studies dating back half a century, it was recognized that oxidation represented a chemical modification occurring in atherosclerosis. However, this recognition of oxidative modifications was limited originally to that of lipids, and it was not linked to the formation of foam cells until 1984, when it was realized that conversion of LDL into a high-uptake form via incubation with endothelial cells was associated with oxidation of the lipoprotein (886).


Using direct chemical tools for analysis, early studies clearly established the presence of oxidized lipids in human atherosclerotic lesions (95, 128, 304, 307, 361). Approximately half a century ago, Glavind et al. (307) reported that the content of peroxides in a chloroform extract prepared from human diseased aorta strongly correlated with the degree of atherosclerosis, whereas normal arteries did not contain lipid peroxides. As a result, these authors suggested lipid peroxidation was either secondary to the deposition of lipids or played an active role in the pathogenesis of atherosclerosis.

I) Fatty acid oxidation products.

Several different types of fatty acid oxidation products have been reported to be present in human atherosclerotic lesions. Of these, hydroxy products of linoleic acid or hydroxyoctadecaenoic acids (HODEs) were the first to be detected (94, 360, 938). These, together with hydroxyeicosatetraenoic acids (HETEs or hydroxy fatty acid oxidation products of arachidonic acid), are the most abundant type of oxidized lipid in atherosclerotic lesions (128, 915, 1006). Most of these products are present as cholesterol esters (507, 915). This is not surprising given that cholesteryllinoleate and cholesterylarachidonate are the major, readily oxidizable lipid in LDL (234), and LDL is the major source of lipids that accumulate during atherosclerosis. As judged by stereospecific analysis, most esterified and nonesterified HODEs are derived from nonenzymatic oxidation reactions (251, 966, 1006). This is consistent with the observation that during lipoxygenase-mediated LDL oxidation, enzymatic oxidation reactions are quickly superceded and dominated by nonenzymatic oxidation reactions (396). In addition, most of the esterified HODEs in human atherosclerotic lesions are generated in the presence of α-TOH, as judged by their regioisomeric distribution (966) (Fig. 6).

Human atherosclerotic lesions also contain oxo-octadecaenoic acids (507, 916, 1006) and F2-isoprostanes (308, 731), the latter representing secondary radical oxidation products of arachidonic acid (642). However, these lipid oxidation products generally occur at concentrations 20- to 40-fold lower than those of HODEs and HETEs (580, 1006).

II) Oxysterol.

Like fatty acids, cholesterol can undergo enzymatic and nonenzymatic oxidation to a range of oxysterols, and a number of studies reported the presence of oxysterols in organic extracts of human aortas (reviewed in Ref. 97). The product of mitochondrial 27-hydroxylase, i.e., 27-hydroxycholesterol (previously also referred to as 26-hydroxycholesterol), is the most abundant oxysterol in atherosclerotic lesions (861) and human macrophage foam cells (602). Its concentration increases in parallel with that of cholesterol and increasing severity of atherosclerosis (127, 861, 966), suggesting that 27-hydroxycholesterol may be produced by cells in atherosclerotic lesions in response to cholesterol accumulation (97). After 27-hydroxycholesterol, 7-ketocholesterol is the next most abundant oxysterol in advanced human atherosclerotic lesions, followed by 7β- and 7α-hydroxycholesterol. The latter three oxysterols are generally considered to be nonenzymatic oxidation products derived from dietary sources or generated in vivo. Other oxysterols present in human lesions include 7β-hydroperoxycholesterol that may represent a major cytotoxin (151). Oxysterols appear to concentrate in foam cells (602), although with the exception of 27-hydroxycholesterol, their tissue concentrations do not generally parallel disease severity (127). Oxysterols are predominantly present in esterified form (98), as mono- and diesters, for presently largely unknown reasons. However, despite their presence, there is no direct evidence in humans that oxysterols contribute to atherosclerosis (97).


In addition to lipid oxidation, there is also good evidence for protein oxidation in human atherosclerotic lesions (for reviews, see Refs. 276, 384). Such evidence comes principally from studies employing either immunohistochemical methods based on antibodies that recognize protein-bound oxidation products, or analytical methods such as high-performance liquid chromatography or gas chromatography/mass spectrometry. The initial documentation for the presence of oxidized proteins including LDL's apolipoprotein B-100 in human lesions comes from immunohistochemical studies (see sect. iiiB3). While this approach provides important information on the location of oxidized proteins, its disadvantages include the potential interference from structurally related molecules and its semi-quantitative nature. On the other hand, the analytical methods used offer quantitative information and high specificity, although they generally do not provide information on the location of the oxidized proteins.

In general, the types of protein oxidation observed in atherosclerotic lesions are similar to those observed during in vitro oxidation of LDL. As mentioned earlier, oxidative damage to proteins may result from electrophilic (2e-) and radical (1e-) reactions, e.g., initiated by electron leakage, metal-ion-dependent reactions, autoxidation of lipids and sugars, and breakdown products of lipid oxidation. The products include several reactive species such as hydroperoxides and protein-bound reductants (notably dopa), as well as compounds that retain the initial oxidant as a covalent adduct (190). The latter are commonly formed as a result of proteins undergoing electrophilic addition and substitution reactions, or radical-radical recombination reactions that give rise to specific markers of oxidative damage, indicative of the oxidant(s) involved (see sect. iiB).

Compared with normal arteries, human carotid endarterectomy samples contain higher concentrations of several amino acid oxidation products (i.e., dopa, o-tyrosine, m-tyrosine, hydroxyleucine, and hydroxyvaline) indicative of ·OH-mediated protein oxidation (276). In addition, such lesions contain elevated levels of o,o′-dityrosine suggestive of the involvement of HOCl-mediated reactions (276). The latter notion supports an earlier finding that atherosclerotic lesions contain increased concentrations of chlorotyrosine, thought to represent a characteristic marker for myeloperoxidase-mediated oxidation reactions (377). However, limited information is available regarding the relationship between the accumulation of markers of oxidized proteins and severity of atherosclerosis. Upston et al. (966) reported the absence of a clear disease stage-dependent accumulation of several amino acid oxidation products, except dityrosine, although the study did not exclude the possibility that specific tyrosine modifications may be early disease events (966).

3. Oxidized LDL is present in atherosclerotic lesions

If oxidized LDL is a required feature of atherosclerosis, it stands to reason that one should be able to detect this form of LDL in atherosclerotic lesions. The first approach to this issue utilized antibodies that recognize epitopes on oxidized LDL that are not present in its native, nonoxidized form. The oxidation of polyunsaturated fatty acids can lead to the formation of aldehydes that modify lysine residues in apolipoprotein B-100 (see Ref. 234 and sect. iiiD). Adducts of lysine residues with malondialdehyde and 4-hydroxynonenal have been characterized extensively and antibodies raised against them. These antibodies avidly stain atherosclerotic lesions in LDL receptor-deficient rabbits (82, 348), apolipoprotein E-deficient mice (694), and humans (695, 1089) with no demonstrable staining in normal arteries. As the oxidative modification hypothesis would predict, these epitopes largely colocalize with macrophages (348), although one might also argue they are not specific for LDL and could represent modification of other proteins in the atherosclerotic lesion. Inconsistent with this possibility, however, is a study showing that LDL isolated from atherosclerotic lesions possesses properties that resemble those of oxidized LDL formed in vitro (1089). Furthermore, LDL isolated from human atherosclerotic lesions contains elevated levels of chlorotyrosine (534) and o,o′-dityrosine (377), as assessed by gas chromatrography-mass spectrometry. These findings demonstrate that lesion LDL is oxidatively modified, and they suggest that HOCl is a likely oxidant participating in these modification reactions. Indeed, there is antibody-based evidence that apolipoprotein B-100-containing lipoproteins of human atherosclerotic lesions are modified by HOCl (368), and lesion lipoproteins contain high ratios of 3,5-dichlorotyrosine to 3-chlorotyrosine (966), indicative of local HOCl production, as would be expected if active myeloperoxidase binds to LDL (133).

4. Circulating oxidized LDL

LDL circulates in plasma, and a portion traverses the subendothelial space to arrive back in the circulation (820, 821). The plasma content of antioxidants provides potent protection against oxidation (269) and, as a consequence, the putative site of LDL oxidation is the subendothelial space. One might expect the transit of LDL across this space to yield small amounts of circulating LDL that is oxidized. Indeed, chemical analysis of circulating LDL has been reported to yield a minor fraction, termed LDL− that exhibits an enhanced content of oxidized lipid (830). Consistent with these findings, human plasma contains immunoreactivity towards epitopes generated from oxidized LDL (409b, 693). However, the existence of oxidized LDL in the circulation remains controversial on the basis of potential artifacts that may occur during the ex vivo handling of plasma and isolation of LDL.

Although the aforementioned data do not address a causal relation between oxidized LDL and atherosclerosis, several studies have shown that circulating levels of oxidized LDL epitopes can be used to distinguish between patients with and without clinically evident atherosclerosis (409a–409d). Using immunologic methods that detect oxidized phosphatidylcholine [including 1-palmitoyl-2-(9-oxononanoyl)phosphatidylcholine] and their protein adducts (441), but not native, acetylated, or malondialdehdye-treated LDL, Ehara and co-workers (221, 222) also reported that acute coronary syndromes are characterized by increased circulating levels of oxidized LDL. Together, these data indicate that relatively small amounts of LDL containing different types of oxidation-specific epitopes can be detected in the blood and may reflect atherosclerosis and its different manifestations. What is less clear at present is where these epitopes originate from and which, if any, of the different oxidation-specific epitopes directly relate to and/or are important for disease burden (959).

5. Autoantibodies to oxidized LDL

If oxidative modifications of apolipoprotein B-100 resulted in the formation of neoepitopes, one might expect them to be immunogenic, and this expectation has proven true (697). Oxidized LDL is immunogenic. A large number of epitopes within the apolipoprotein B-100 component of oxidized LDL have been identified that provoke an immune response (257), and autoantibodies against malondialdehyde-modified lysine residues have been demonstrated in the serum of both rabbits and humans (695). Some studies have reported that the titer of these autoantibodies is associated with the burden of and may predict progression of atherosclerosis (574, 795) and myocardial infarction (232, 1066). Higher titers of autoantibodies have also been associated with coronary artery disease (105), peripheral atherosclerosis (56), and higher risk for restenosis following balloon angioplasty (299). In addition, there is support for a role for antioxidized autoantibodies in animal atherosclerosis. For example, circulating titers of autoantibodies to oxidized cardiolipin correlate with the extent of atherosclerosis in apolipoprotein E-deficient mice (732). In LDL receptor-deficient mice fed an atherogenic diet, serum titers of autoantibodies increase over time and correlate with both the extent of blood cholesterol and lesion size (696) and the aortic content of oxidized LDL. Autoantibodies against oxidized phospholipid epitopes, present either as lipids or as lipid-protein adducts, may directly affect atherosclerosis by inhibiting macrophage uptake of oxidized LDL (412). In support of this contention, immunization with apolipoprotein B-100 peptides against which high levels of antibodies are present in healthy humans, reduces atherosclerosis in apolipoprotein E-deficient mice (258). It is worth noting here, however, that the above-mentioned correlations between atherosclerosis and cardiovascular disease burden on the one hand and circulating levels of oxidized LDL epitopes and autoantibodies to oxidized LDL on the other hand are not observed consistently (see sect. iiiF1c) and also not immediately consistent with biochemically assessed oxidant markers and the extent of atherosclerosis (see sect. iiiC).

6. Potential proatherogenic activities of oxidized LDL

In addition to its activity to support foam cell formation, oxidized LDL also has a number of other potential proatherogenic properties (Table 8). Endothelial cell activation is among the early events of atherosclerosis, and products of oxidized LDL facilitate this process (480, 511). Once oxidized, LDL increases its substrate suitability for sphingomyelinase, an enzyme that promotes LDL aggregation thereby enhancing its macrophage uptake (813). The initial oxidation of LDL and formation of what is often referred to as “minimally modified” LDL stimulates adjacent endothelial cells and smooth muscle cells to synthesize and secrete monocyte chemotactic protein-1 (754) that is thought to facilitate the recruitment of monocytes into the arterial wall (659, 1087). This activity of minimally modified LDL is contained in its lipid fraction (1018). The importance of monocyte chemotactic protein-1 for atherosclerosis is highlighted by the observation that mice deficient for the protein (323) or its receptor (72) show attenuated disease development.

View this table:

Potential proatherogenic activities of oxidized LDL

The formation of oxidized LDL in the subendothelial space can also facilitate atherogenic progression. There are dual activities of oxidized LDL that could enhance the arterial content of inflammatory cells. For example, and as pointed out earlier, oxidized LDL is chemotactic for monocytes (742) and T lymphocytes (611) by virtue of its content of lysophosphatidylcholine, formed during oxidation (886). In addition, oxidized LDL is chemostatic for macrophages (742). Thus the enhanced entry and impaired egress of inflammatory cells would be expected to stimulate arterial inflammation, a process strongly implicated in atherosclerosis (556). Finally, oxidized LDL limits the biologic activity of endothelium-derived ·NO (505), and the loss of ·NO bioactivity is associated with increased inflammatory cell entry into the arterial wall (504).

The gene expression pattern in the arterial wall is also subject to influence by modified forms of LDL. For example, minimally modified LDL as well as oxidized LDL can induce macrophages to express scavenger receptor, thereby enhancing foam cell formation and lipoprotein uptake (622). A number of genes associated with inflammation are also upregulated by oxidized LDL such as monocyte chemotactic protein-1, serum amyloid A, ceruloplasmin, and heme oxygenase-1 (552). In addition, a recent discovery is the stimulating effect of oxidized LDL on macrophage expression of peroxisome proliferator activated receptor-γ. This has been shown to alter scavenger receptor (CD36) expression and the expression of proinflammatory genes (244). Thus oxidized LDL has a number of biologic activities that could, in theory, contribute to the process of atherosclerotic lesion formation.

7. Scavenger receptors and atherosclerosis

The oxidative modification hypothesis holds that LDL recognition by the scavenger receptor is a prerequisite for foam cell formation. This notion is supported by considerable experimental evidence. Mice lacking the scavenger receptor-A gene demonstrate a defect in the binding and degradation of modified LDL (500). In mice lacking the apolipoprotein E gene, deletion of the scavenger receptor A confers resistance to atherosclerosis (925). Similarly, macrophages derived from mice lacking the CD36 scavenger receptor demonstrate reduced uptake of modified LDL (676) and reduced atherosclerosis when crossed with the apolipoprotein E-deficient mouse (243). These data are consistent with a critical role of scavenger receptors in early atherosclerosis.

8. Antioxidant studies

If one accepts the notion that LDL oxidation is an essential feature of atherosclerosis, then inhibiting LDL oxidation should limit atherosclerosis. This concept has been tested primarily in animals with a number of structurally distinct antioxidant compounds that can inhibit LDL oxidation in in vitro assays. In the following paragraphs, we will briefly touch upon this body of evidence with particular attention to the specific antioxidant compounds that have been tested.

Arguably the strongest evidence in support of antioxidant compounds providing protection against atherosclerosis comes from studies with probucol (594). Probucol is a synthetic cholesterol-lowering drug that also possesses antioxidant activity (590, 709). The lipophilic drug associates with and effectively protects LDL against in vitro oxidation induced by copper ions (489, 709), although it is a sterically hindered phenol, and its peroxyl radical scavenging activity is only ∼16% of that of α-TOH (737).

Early studies with probucol demonstrated inhibition of atherosclerosis in rabbits (502, 937) and monkeys (1050). Kita et al. (489) treated Watanabe heritable hyperlipidemic (WHHL) rabbits with probucol and observed an 87% reduction in lesion area and LDL resistance to oxidation compared with those animals not treated with probucol. These findings prompted the authors to conclude the reduction of atherosclerosis was due to the antioxidant effect of probucol; however, probucol also produced a 17% reduction in serum cholesterol. Carew et al. (125) controlled for the cholesterol-lowering effect of probucol with lovastatin, an inhibitor of 3-hydroxy-3-methylglutaryl coenzyme A reductase. Similar reductions in total cholesterol were observed with lovastatin and probucol; however, the latter provided an additional 48% reduction in atherosclerosis (125), suggesting that probucol inhibited atherosclerosis due to its antioxidant activity. This contention was supported in a study with modified WHHL rabbits treated with a structural analog of probucol devoid of cholesterol-lowering properties (583). Both probucol and the analog inhibited LDL oxidation and reduced atherosclerosis, suggesting an antioxidant-mediated mechanism for reduced lesion formation. Subsequent studies in primates (800), cholesterol-fed rabbits (834), and hamsters (704) have also demonstrated a reduction in atherosclerosis with probucol. The situation appears more complex in murine models of atherosclerosis, where probucol promotes atherosclerosis in the aortic root (61, 630, 1104) but inhibits disease formation at more distal sites (1054).

There have been two studies testing the antiatherosclerotic activity of probucol in humans. The Probucol Quantitative Regression Study (PQRST) reported probucol to be ineffective in attenuating lumen loss in the femoral arteries in hypercholesterolemic subjects over 3 years, as assessed by quantitative angiography (1010). Importantly however, this method does not directly assess disease burden (see sect. ivA). In contrast, the Fukoaka Atherosclerosis Trial (FAST) observed probucol to significantly decrease atherosclerosis progression in the carotid artery of hypercholesterolemic patients, as assessed by the intima-to-media thickness determined by B-mode ultrasound (808). In humans, probucol (934, 935) and a probucol analog with one of its two phenol moieties present as succinate ester (935), also protect against restenosis after percutaneous coronary intervention.

A number of other antioxidants have also been tested for their ability to inhibit atherosclerosis in animal models of the disease. N,N′-diphenyl-phenylenediamine (DPPD) is an aniline compound that attenuates atherosclerosis in the aorta of cholesterol-fed rabbits (865), and similar findings have been reported in a murine model of atherosclerosis (932). In addition, 2,3-dihydro-5-hydroxy-2,2-dipentyl-4,6-di-tert-butylbenzofuran (BO-653), a synthetic antioxidant with structural components of vitamin E (673), inhibits atherosclerosis in both rabbit and murine models (174). Similarly, 3,3′,5,5′-tetrabutyl-1,1′-biphenyl- 4,4′-diol, a lipophilic bisphenol, inhibits atherosclerosis in mice deficient in apolipoprotein E and the LDL receptor (1056). Furthermore, supplementation of the diet with butylated hydroxytoluene reduces atherosclerotic lesions in cholesterol-fed rabbits (64, 271). A common feature of all of these compounds is that they can inhibit LDL oxidation in vitro. Thus a number of lipid-soluble, synthetic antioxidants have been used to demonstrate an association between a reduction in atherosclerosis and ex vivo inhibition of LDL oxidation. The effect of vitamin E on atherosclerosis in animals and cardiovascular disease in humans is discussed in section iiiF.

C. LDL Oxidation

As we have learned, there is now convincing evidence that oxidatively modified LDL exists in atherosclerotic lesions, yet it remains largely unknown where precisely within the vessel wall, how, and to what extent, LDL becomes oxidized during atherogenesis. In addition to intrinsic properties of the lipoprotein, factors that prolong the life/residence time of LDL may also be conducive to oxidation. For example, proteoglycans not only “trap” LDL in the extracellular matrix of the intimal space, but also change the conformation of apolipoprotein B-100 (120), and such proteoglycan-exposed LDL shows greater sensitivity than native LDL to subsequent in vitro oxidative modification by Cu2+ (121, 422) but not other oxidants (967).

It is generally upheld (882) that LDL oxidation must occur in the arterial wall rather than the circulation, as lipoprotein lipids in plasma are well protected from oxidation due to the robust antioxidant defenses (897). It is noteworthy that LDL itself contains and in fact is the major transport vehicle for most of the plasma α-TOH. Furthermore, oxidized lipoproteins that may exist or form in plasma are diluted rapidly by either hepatic clearance (976) or accumulation and subsequent degradation in the arterial wall (460). Consistent with this, the plasma concentrations of oxidized LDL, as assessed by immunologic techniques, and oxidized lipids, determined by analytical techniques, in healthy humans are extremely low (895). For example, LOOH are present in the low nanomolar range, and there is no evidence that they are specifically associated with LDL (80). The situation is similar in patients with severe cardiovascular disease. Plasma of these subjects shows signs of mild oxidative stress, not damage, as indicated by a slightly elevated proportion of ubiquinone-10 to total coenzyme Q (157, 514), and the decreased concentrations of ascorbate (1001). Apart from these minor differences, concentrations of antioxidants and oxidant markers do not correlate with the extent of atherosclerosis (1001).

In sharp contrast to the plasma situation, and as we have seen, atherosclerotic lesions contain substantial amounts of oxidized lipid. Despite this however, homogenates of advanced human plaque also contain ascorbate, uric acid, and α-TOH in quantities comparable to human plasma (see sect. iiiF), sufficient to efficiently block LDL oxidation in vitro (269).

1. Putative oxidants involved

Much research has concentrated on modeling LDL oxidation mediated by putative physiological oxidizing species, including cells, in the presence of various antioxidants. In vitro, LDL oxidation is initiated by a variety of oxidants including 1e-oxidants (i.e., radical) and 2e-oxidants (i.e., nonradical). Many studies have attempted to determine the oxidants contributing to oxidative processes in atherosclerotic lesions. In most cases, the approach chosen has been indirect. This is not surprising, given the labile nature of most of the biologic oxidants potentially involved. In the case of human atherosclerotic lesions, a common biochemical approach has been to first chemically define the “footprint” of products of a particular oxidant with target molecules (see sect. iiB) and to then analyze vessels and materials derived from them for the presence of these footprints. With the advent of murine models of atherosclerosis, molecular approaches such as the use of animals deficient in a specific oxidant-forming enzyme have been utilized increasingly.

Together, these studies have provided useful insight into the origin and nature of the oxidative modifications taking place during atherogenesis. For example, there is now convincing evidence for the participation of myeloperoxidase and oxidants derived from this enzyme. Other oxidants, both enzymatic and nonenzymatic, have also been suggested to be involved. However, the approaches taken all have their limitations. For example, not in all cases has it been possible to localize the oxidative modification to LDL or even lipoproteins. Also, while the analysis of human atherosclerotic material provides direct information on the human disease, in most cases such information represents a snapshot of a single or few time points of what represents a disease developing over decades. In contrast, animal studies provide the possibility to follow changes over disease progression, or regression, yet suffer from the limitation that there is no single animal model in which atherosclerosis truly reflects the human pathogenesis. This may be reflected, for example, in differences in the activity of a particular oxidant-producing enzyme (e.g., myeloperoxidase) in human versus animal atherosclerotic lesions so that manipulation of such enzyme in the animal model does not necessarily provide useful information about human atherosclerosis. A further limitation of animal models is that interventions are commonly carried out before substantial disease has developed, whereas in humans, interventions mostly engage subjects with established atherosclerosis. Keeping these limitations in mind, let us now review the evidence for and against the involvment of different ROS and RNS in oxidative modifications in atherosclerosis.


Since the early discovery by Henriksen et al. (395), vascular cells including endothelial and smooth muscle cells and macrophages have attracted much attention as sources of oxidants, as they all can oxidize LDL to a form recognized by the acetyl LDL receptor. As described in section iiC, vascular cells contain NAD(P)H oxidases capable of generating O2·, and recent data suggest that Nox isoforms may have a role in atherosclerosis. For example, early atherosclerosis in Watanabe hyperlipidemic rabbits (that lack an LDL receptor) is associated with excess vascular production of O2· due to NAD(P)H oxidase activity (1015). Consistent with this observation, diet-induced atherosclerosis in primates is associated with excess vascular O2· and upregulation of p22phox (cytochrome b-245 alpha polypeptide, CYBA) and p47phox (neutrophil cytosolic factor 1, NCF1), two features that abate with regression of atherosclerosis through dietary cholesterol lowering (363). However, a causal role for NAD(P)H oxidase in atherosclerosis is controversial. Mice lacking gp91phox (Nox2) have no demonstrable difference in either the apolipoprotein E −/− or diet-induced models of atherosclerosis (484). Two studies have examined the effect of p47phox on atherosclerosis in the apolipoprotein E −/− model (41, 413). Both observed no difference in atherosclerosis at the level of the aortic sinus as a function of p47phox status; however, one study examined atherosclerosis in the descending aorta and found that p47phox−/− animals had considerably less atherosclerosis than their p47phox +/+ counterparts (41).

Recent studies in human tissue suggest that Nox isoforms are involved in atherosclerosis. Vascular production of O2· increases as a function of risk factors for disease (346), suggestive of NAD(P)H oxidase playing a role in the disease. Atherosclerotic lesions contain abundant p22phox and gp91phox in the vicinity of macrophages that correlated with the severity of atherosclerosis (463, 864). In contrast, Nox4 was found exclusively in nonphagocytic cells, and it did not vary with atherosclerotic lesions. Analysis of tissue in patients undergoing bypass surgery revealed that diabetes is characterized by increased expression of p22phox, p47phox, and p67phox (neutrophil cytosolic factor 2, NCF2), compared with nondiabetics (345). Given that diabetics have disproportionably more cardiovascular events than nondiabetics, these data suggest that NADPH oxidase activity has a role in the clinical expression of atherosclerosis. However, the specific contribution of each Nox isoform to atherosclerosis and the clinical expression of cardiovascular events remain to be determined.

In addition to NAD(P)H oxidases, xanthine oxidase has been implicated in endothelial cell O2· production in experimental models of hypercholesterolemia. Cholesterol-fed rabbits exhibit an increase in endothelial O2· production that is inhibited by oxypurinol (680), a xanthine oxidase inhibitor. This increased O2· flux from xanthine oxidase contributes to impaired endothelial ·NO bioactivity in the setting of hypercholesterolemia (123, 680), heavy smoking (343), and coronary artery disease (866), although the regulation of endothelial xanthine oxidase is not entirely clear. Some evidence suggests that endothelial xanthine oxidase activity is increased in response to interferon-γ (216) or the binding of neutrophils to the endothelial surface (1008). Other data suggest that “endothelial xanthine oxidase” is actually located extracellularly and derived from a circulating pool of enzyme that is increased in the setting of hypercholesterolemia (1031). More recently, evidence has been presented suggesting that NAD(P)H oxidase maintains endothelial cell xanthine oxidase and that xanthine oxidase is responsible for increased ROS production in response to oscillatory shear stress (612). Oscillatory shear stress occurs at sites of the circulation that are vulnerable to atherosclerosis, although a direct role for xanthine oxidase in atherosclerosis has not been investigated to date.

It is also possible that eNOS contributes to cellular O2· in the setting of vascular disease, because hypercholesterolemia (735), atherosclerosis (524), diabetes (400), and hypertension (520) have all been associated with eNOS uncoupling. Moreover, animals that serve as models of atherosclerosis and that have been treated with tetrahydrobiopterin or its precursor exhibit a decrease in vascular O2· and enhancement of ·NO bioactivity (346, 524). Tetrahydrobiopterin also improves endothelium-dependent vasodilation in chronic smokers (387). Thus considerable evidence exists for eNOS-derived O2· production, although the precise mechanism(s) responsible for this phenomenon and how it relates to atherosclerosis remains unknown.

As indicated above, the specific role of O2· in oxidative modification during atherosclerosis, particularly that related to LDL oxidation, remains unclear. Consistent with its relative chemical reactivity, O2· is not able to oxidize LDL, in contrast to its protonated form, the hydroperoxyl radical (HO2·) (48). However, at physiological pH only a fraction of O2· produced is present as HO2·. This suggests that O2·-generating systems, including NAD(P)H oxidases, are by themselves not efficient in oxidizing LDL (48), a notion confirmed experimentally for macrophages (289, 448), although superoxide has been proposed as a mediator of LDL oxidation by endothelial cells (883). However, O2· may act as the precursor for chemically more reactive oxidants, such as ONOO and H2O2 and oxidants derived from them, and these may participate in LDL oxidation in the vessel wall.


Using a sensitive and specific method involving gas chromatography and mass spectrometry, Leeuwenburgh et al. (533) reported LDL from aortic atherosclerotic intima to contain 90-fold higher levels of protein-standardized 3-nitrotyrosine than plasma LDL. This is consistent with an earlier study reporting increased tyrosine nitration in atherosclerotic lesions based on an immunological approach (47) and implicates RNS in the oxidation of LDL and other targets in the vessel wall. In addition to eNOS, inducible NOS is expressed in atherosclerotic lesions of humans (114) and rabbits (567), providing potential sources for RNS.

Although commonly assumed, the presence of 3-nitrotyrosine in human lesions does however not prove the involvement of ONOO. The reason for this is that 3-nitrotyrosine is also formed from the reaction of ·NO with HOCl (1048) or tyrosine phenoxyl radicals (341). In biological samples, ONOO and myeloperoxidase leave similar footprints with regard to protein nitration (479). Indeed, the myeloperoxidase/H2O2/nitrite system modifies LDL's apolipoprotein B-100 and lipids, probably via formation of ·NO2 (115, 723), and such modified LDL can convert macrophages into foam cells (723). In addition, in vitro reaction of nitrite with HOCl gives rise to nitryl chloride (NO2Cl) that can chlorinate and nitrate protein tyrosine residues (226), as well as oxidize LDL lipids (699), although the biological significance of this is not clear.


The role of myeloperoxidase and myeloperoxidase-derived oxidants in atherosclerosis has been reviewed recently (384, 720). Heinecke and co-workers (184) were the first to report the presence of active myeloperoxidase in human atherosclerotic lesions. The involvement of myeloperoxidase activity in oxidation reactions taking place during atherogenesis in humans is also supported by the detection of footprints of myeloperoxidase/HOCl, including HOCl-modified proteins (368), 3-chlorotyrosine (377), p-hydroxyphenylacetaldehyde (375), and p-hydroxyphenylacetaldehyde-ethanolamine (391). The potentially important finding of myeloperoxidase activity has been confirmed for lesions in humans (368, 581, 917), Watanabe heritable hyperlipidemic rabbits (85), and rabbits fed a high-cholesterol diet (582). In contrast, active myeloperoxidase appears to be absent in atherosclerotic lesions in LDL receptor −/− and in apolipoprotein E −/− mice (90), and therefore is not likely to contribute to atherogenesis in these animal models. This highlights one of the limitations of using animal models to study the pathogenesis of the human disease. It suggests that myeloperoxidase-deficient mice are not suitable to evaluate the potential role of this protein in human atherogenesis. Mice overexpressing myeloperoxidase may be a more useful tool for this purpose.

Hypochlorite-modified LDL is potentially proatherogenic (132). In vitro studies have shown that HOCl-modified LDL can stimulate macrophage foam cell formation (371) via binding to class B scavenger receptors (589), increases leukocyte adherence and migration into blood vessels (555), increases ROS production by leukocytes (499), and has chemotactic activity for neutrophils (but not monocytes) (1061). Myeloperoxidase may oxidize LDL via reactive amino acid intermediates, such as the tyrosyl radical (806), p-hydroxyphenylacetaldehyde (391), or α-hydroxy aldehydes (466). In addition, myeloperoxidase-derived HOCl may convert LDL into a high-uptake form of the lipoprotein via reaction with apolipoprotein B-100, with little oxidation of the lipids (371, 373). HOCl reacts readily with the Nε-amino groups of lysine residues, resulting in the formation of N-chloramines and increased negative charge of the lipoprotein particle (373). A small proportion of these chloramines break down to aldehydes (373, 1078), which may contribute to the cross-linking and aggregation of HOCl-exposed LDL. Other amino acid residues, such as cysteine and methionine, are also modified (373, 1081). Chloramines can break down to generate alkoxyl radicals that induce lipid peroxidation (370), although the overall extent of lipid oxidation, relative to protein oxidation, is small.


Several lines of evidence support a role for lipoxygenase in atherosclerosis. 15-Lipoxygenase and 5-lipoxygenase are expressed in atherosclerotic lesions of humans (1090) and apolipoprotein E −/− mice, respectively (613). Also, disruption of the 12/15-lipoxygenase gene or decreased expression of 5-lipoxygenase diminishes disease in apolipoprotein E −/− and in LDL receptor −/− mice (176, 298, 613). Furthermore, inhibiting 15-lipoxygenase lowers lesion formation in rabbits fed a high-fat and high-cholesterol diet (824).

There is indirect evidence in support of oxidative events in general and LDL oxidation in particular contributing to the observed link between lipoxygenase and atherosclerosis. Thus 15-lipoxygenase and epitopes of oxidized LDL colocalize in human lesions (1090), and the expression of 15-lipoxygenase in rabbit arteries results in the appearance of oxidized lipid-protein adducts (1088). Also, the plasma and urinary concentrations of isoprostanes correlate positively with 12/15-lipoxygenase activity and disease extent in apolipoprotein E −/− mice (175). In addition, oxidation products specific for 15-lipoxygenase have been reported to be present in lesions (251, 508), and the isolated enzyme (51), cells containing 12/15-lipoxygenase (1108), or overexpressing 15-lipoxygenase can oxidize LDL in vitro (54, 237), although the underlying mechanism for this remains unclear.

Emerging evidence has heightened the interest in the contribution of lipoxygenase to atherosclerosis. Analysis of atherosclerosis-prone and atherosclerosis-resistant mice identified a region on chormosome 6 that conferred resistance to atherosclerosis despite elevated levels of lipids (614). Further analysis of this locus determined that 5-lipoxygenase was one putative gene on chromasome 6 that conferred susceptibility to atherosclerosis (613). This suspicion was confirmed through the use of mice lacking one copy of the 5-lipoxygenase gene that, when bred with LDL receptor −/− mice, demonstrated a dramatic decrease in atherosclerosis (613). This observation has now been extended to humans as variant 5-lipoxygenase alleles segregate with evidence of atherosclerosis by carotid imtimal-to-medial thickness measurements on ultrasound (218). Thus 5-lipoxygenase appears to be one lipoxygenase isoform that is particularly germane to the development of atherosclerosis in both experimental animals and humans.


It has been known for a long time that the oxidative modification of LDL by vascular cells is absolutely dependent on the presence of low concentrations of transition metals, such as copper and iron, in the medium (reviewed in Ref. 290). For example, it can be completely prevented by the inclusion of metal chelators such as ethylenediamine tetraacetic acid or relatively small amounts of serum (that contains metal-binding proteins). This is also true for LDL oxidation induced by monocytes in which NADPH oxidase activity is stimulated by opsonized zymosan (138). In this case, iron contaminating the commercial preparations of zymosan has been shown to be required for LDL oxidation (1072). However, both monocytes (150) and neutrophils (130, 896) can oxidize LDL in metal-independent ways. Where transition metals are involved, the role of the cells appears to be to maintain the metal in the reduced form, thereby allowing it to redox cycle and to react with traces of LOOH (290, 291) (reaction 22) Math(22)

In fact, incubating LDL in serum-free medium in the presence of copper or iron ions mimics the process of cell-mediated oxidation of the lipoprotein (see sect. iiiC2). This has raised interest in the possibility that transition metals may participate in oxidative modification reactions in atherosclerotic lesions. Indeed, iron deposits are present in human lesions (1092). As pointed out earlier however, in vivo, transition metals are usually present as complexes to carrier proteins that prevent participation in inadvertent redox reactions, although it is possible that hemin (29) or hemoglobin derived from ruptured erythrocytes (690) may contribute to oxidation reactions. Macrophages that have previously phagocytosed iron-rich materials such as erythrocytes can exocytose redox-active iron that is capable of oxidizing LDL (1093). The possible contribution of transition metals to LDL oxidation during atherogenesis is supported by the presence of free transition metals in advanced human lesions (515, 859, 870). However, increasing or decreasing plasma and tissue stores of iron does not affect the formation of atherosclerotic lesions in cholesterol-fed rabbits (179). Also, in apolipoprotein E −/− mice, iron overload decreases lesion size despite detectable increases in hepatic concentrations of markers of oxidative events in the liver (485). Similarly, dietary supplementation with copper decreases atherosclerosis in cholesterol-fed rabbits (516), while copper deficiency increases atherosclerosis in C57B mice (354). Together, these results indicate that iron and copper are not likely important catalysts for oxidation events leading to atherosclerosis and that studies using free transition metals such as Cu(II) to oxidize LDL in vitro may not be meaningful biologically.


As indicated in section iiiB2b, there is some evidence for ·OH contributing to oxidative reactions in atherosclerosis based on the finding that certain hydroxylated amino acids are increased in human endarterectomy specimens (276) and in the artery wall of monkeys in early diabetic vascular disease (716). There is also a recent suggestion for a role of ozone in atherosclerosis based on the presence of cholesterol ozonolysis products in human carotid endarterectomy specimens (1028). Activation of leukocytes in arteries showing advanced atherosclerotic disease has been proposed to result in the generation of ozone, as assessed by the conversion of indigo carmine to isatin sulfonic acid, and that of cholesterol to specific cholesterol oxidation products that cause cytotoxicity and conversion of LDL into a high-uptake lipoprotein (1028). However, it remains controversial whether activated leukocytes produce ozone, and a recent study has shown that conversion of indigo carmine to isatin sulfonic acid can be mediated by ROS other than ozone (478). It will be important therefore to verify the specificity of the cholesterol oxidation products detected (1028) to substantiate a participitation of ozone in oxidative events in atherosclerosis.

What is clear is that human atherosclerotic lesions contain relatively large amounts of hydroxylated fatty acids associated with lipoprotein cholesterylesters and derived via nonenzymatic reactions. As mentioned in section iiiC2a, hydroxylated cholesterylesters are the major class of oxidized lipid present in advanced atherosclerotic plaque (672, 915, 966) and more abundant than isoprostanes and hydroxides of nonesterified fatty acids (1006). This implies that their precursors, i.e., lipoprotein-associated, esterified LOOH, LO·, and LOO· are generated as lesions develop. As both LO· and LOO· are scavenged rapidly by α-TOH, it follows further that α-TO· must also be formed in human lesions.

2. High-uptake or oxidized LDL

It is important to emphasize that although commonly used (as in sect. iiiC6), the term oxidized LDL does not define a characterized molecular species. In fact, oxidized LDL represents a heterogeneous population of modified forms of LDL that differ greatly in their chemical composition and functional properties (881). The differences in the chemical composition are due to individual differences in the lipid and antioxidant content of LDL, as well as the choice of the oxidant and the incubation conditions used to prepare oxidized LDL. Copper ions are commonly used as the oxidant and incubations carried out for prolonged periods of time. This causes drastic alterations to the lipoprotein particle that, as a result, gains alternative functions (Table 8). For example, as pointed out earlier, oxidized LDL is no longer able to effectively interact with the receptor for native LDL and instead is recognized by scavenger receptors, and upon exposure to macrophages induces cellular cholesterol accumulation. In fact, gain of this alternative function, also referred to as formation of “high-uptake” LDL, was used as the earliest definition of oxidized LDL (395, 885, 886). It is worth noting, however, that even such a functional definition does not characterize oxidized LDL unambiguously, as its relation with the various chemical modifications remains largely unknown.

Much of our knowledge on how native LDL is converted to oxidized LDL originates from in vitro studies exposing LDL to copper ions or to cells cultured in transition metal-containing medium (349, 707, 882, 885, 886). Under these experimental conditions, formation of oxidized LDL is generally a fast process, characterized by a sequence of events starting with the complete loss of LDL's endogenous antioxidants including CoQ10H2 and α-TOH, followed by the conversion of a majority of the polyunsaturated fatty acids to their corresponding hydroperoxides. These primary lipid oxidation products then fragment to secondary lipid oxidation products such as malonyldialdehyde or 4-hydroxynonenal that then react with the Nε-amino group of lysine residues of apolipoprotein B-100 such that the particle's electrophoretic mobility increases and the lipopoprotein becomes “high uptake” (reviewed in Ref. 234). When LDL is oxidized rapidly by transition metals, no measurable modification into high uptake forms occurs during the period when LDL still contains α-TOH (234, 235). α-Tocopherol strongly inhibits the degradation of LOOH to secondary lipid oxidation products such as 4-hydroxynonenal (see, e.g., Ref. 815).

It follows from the above discussion that any condition that gives rise to LOOH has the potential to generate high-uptake LDL if transition metals are available. In general, substantial LDL lipid peroxidation can be achieved by sufficiently reactive 1e-oxidants, although it is less clear whether transition metals are available in biological tissue like arteries. However, in vitro high-uptake LDL can also be generated in the absence of transition metals, as suggested by experiments carried out in the presence of the metal chelator diethylenetriamine pentaacetic acid. For example, Graham et al. (330) observed that reagent ONOO converts LDL to a form recognized by the macrophage scavenger receptor. Similar to the situation with copper ions, the modifications reported for ONOO involve radical reactions, as shown directly by the formation of α-TO· and lipid peroxidation. In addition, ONOO also induces the loss of Nε-amino groups of lysine residues, and this process is associated with an increase in the relative electrophoretic mobility of the lipoprotein particle (330). Interestingly, α-TOH is not able to prevent ONOO-induced loss of Nε-amino groups of lysine residues (947), suggesting that this modification occurs via pathway(s) separate from lipid peroxidation.

In addition to ONOO, the myeloperoxidase/H2O2/nitrite system causes nitration of LDL protein tyrosine residues, lipid peroxidation and binding of the lipoprotein to macrophages that results in foam cell formation (723). This provides a possible metal-independent pathway by which mononuclear phagocytes generate oxidized LDL in the vessel wall, although the particular nature of the alterations induced by the oxidizing conditions and responsible for the altered function remains unclear. What is known is that increased binding and degradation of LDL oxidized by the myeloperoxidase/H2O2/nitrite system is observed before measurable modification of Nε-amino groups of lysine residues takes place, and it is independent of lipoprotein aggregation and the scavenger receptor class A type I (723). It is not known at present how formation of high uptake LDL by this oxidative path relates to the consumption of LDL's α-TOH.

A common feature of the above-described pathways that lead to oxidized LDL is the occurrence of substantial lipid peroxidation, characteristic of the participation of 1e-oxidation reactions. This contrasts with the situation observed with the 2e-oxidant HOCl that also converts LDL into a high uptake form for macrophages (371). As mentioned earlier, oxidation of LDL by HOCl is characterized by the immediate and preferential oxidation of amino acid residues of apolipoprotein B-100 (with lysine residues representing the single major target) in the absence of substantial consumption of α-TOH and occurrence of lipid peroxidation. Similar to the situation with ONOO, α-TOH does not inhibit HOCl-induced oxidation of LDL's lysine residues (372).

In summary, formation of high-uptake LDL via oxidative processes can occur in two ways, one mediated by radical oxidants and the other mediated largely by 2e-oxidation reactions (Fig. 8). The two pathways differ from each other in several ways. For example, only the pathway mediated by radicals shows a requirement for extensive lipid peroxidation, and the 2e-oxidant pathway is not impacted on by lipid-soluble antioxidants such as α-TOH.

FIG. 8.

Conversion of native LDL into high uptake LDL via oxidative processes mediated by 1e-oxidants or radicals (left) and by 2e-oxidants (right). Radical-induced oxidative modification of LDL proceeds largely via distinct phases. Primary lipid peroxidation (a) refers to the period during which endogenous antioxidants such as α-TOH are being consumed and LOOH accumulate. Following depletion of endogenous antioxidants, secondary lipid oxidation proceeds (b), characterized by the breakdown of LOOH to reactive aldehydes and other products, and the general loss of polyunsaturated fatty acids (PUFA). As a result, minimally modified LDL (mm-LDL) may be formed. ApoB-100, apolipoprotein B-100.

3. Minimally modified LDL

Minimally modified LDL is a term introduced to describe LDL less drastically oxidized than LDL after its prolonged exposure to high concentrations of copper ions. Minimally modified LDL is a functional term referring to LDL that is 1) still recognized by the receptor for native LDL, 2) not recognized by macrophage scavenger receptors, and 3) capable of stimulating the release from cultured endothelial cells of monocyte chemoattractant protein-1 and macrophage colony-stimulating factor (58, 754). These biological activities of minimally modified LDL are distinct from those of native LDL and “oxidized LDL” (58, 171, 659, 754). Information on the chemical moieties responsible is available from studies showing that the induction of monocyte adhesion to endothelial cells seen with minimally modified LDL can be mimicked by oxidized phospholipids isolated from 1-palmitoyl-2-arachidonoyl-sn-glycero-3-phosphocholine that has undergone autoxidation (1020). Three biologically active oxidized arachidonic acid-containing phospholipids have been identified, namely, 1-palmitoyl-2-(5-oxovaleryl)-sn-glycero-3-phosphocholine, 1-palmitoyl-2-glutaryl-sn-glycero-3-phosphocholine (1018), and 1-palmitoyl-2-(5,6-epoxyisoprostane E2)-sn-glycero-3-phosphocholine (1020).

4. Role of vitamin E in LDL oxidation

As we have seen, there is good evidence for implicating oxidized LDL in atherosclerosis, although the identities of the oxidants that modify LDL in blood vessels remain speculative. Consequently, research on LDL oxidation has been disparate, employing a myriad of largely nonphysiological conditions in vitro. A number of early and influential studies focused on LDL oxidized by a large molar excess of cupric (Cu2+) ions over several hours (234, 236). As mentioned, such conditions are strongly oxidizing and result in a drastically oxidized lipoprotein capable of generating foam cells in vitro (566). During formation of such oxidized LDL, all α-TOH in the lipoprotein is depleted (236). This is in stark contrast to LDL retrieved from atherosclerotic tissue that contains relatively normal levels of α-TOH (672, 939, 967) but can also generate foam cells (884).


LDL contains 6–12 molecules of α-TOH per particle (234, 896) as its major redox-active constituent (Table 9), so it is important to discern the role of α-TOH in the process of LDL oxidation, particularly that mediated by radical oxidants. If LDL particles are exposed to a strong oxidant or encounter radicals with high frequency, α-TOH is consumed rapidly and there is little concomitant formation of LOOH (78, 81, 234, 236, 433). Under these conditions, α-TOH ostensibly performs an antioxidant role. In sharp contrast, if LDL particles encounter mild oxidants or radicals with low frequency, consumption of α-TOH is much slower and associated with a comparatively large accumulation of LOOH (78, 81, 433, 802, 896). It is important to consider that depletion of LDL's α-TOH is required for accumulation of secondary lipid oxidation products such as aldehydes (235) and F2-isoprostanes (570), although some transformation of LOOH into hydroxyalkenals takes place in the presence of the vitamin (815). It is these secondary lipid oxidation products that are largely responsible for the atherogenic properties of in vitro oxidized LDL (412, 721, 722, 885). Thus under strongly oxidizing conditions, the antioxidant function of α-TOH is in inhibiting an initial step in the overall conversion of native to high uptake LDL.

View this table:

Chemical composition of LDL

However, there is little evidence that strongly oxidizing conditions persist in vivo in atherosclerotic vessels. It is also important to recognize that in vivo LDL is surrounded by a myriad of other molecules, some of which will also react with the oxidants present, thereby essentially decreasing the frequency with which the lipoprotein particles themselves encounter radicals. Therefore, conditions of mild oxidants and low “fluxes” of radical oxidants appear more likely relevant in vivo. Under such conditions, the extent and ability of LDL lipid to oxidize is controlled primarily by α-TOH, as shown unambiguously by in vitro experiments. Thus so long as α-TOH remains in the particle, the process of LDL lipid peroxidation under mild oxidative conditions follows a model of tocopherol-mediated peroxidation (77, 81, 169, 433, 443, 496, 893, 968, 1057, 1058). In tocopherol-mediated peroxidation, α-TOH does not act as a classic chain-breaking antioxidant and the fate of α-TO· determines whether vitamin E in LDL exhibits pro- or antioxidant activity.

Tocopherol-mediated peroxidation of LDL and the central role exerted by α-TOH in this model is described in Figure 9 (77, 78, 81, 433, 893, 968). Essentially, α-TO· replaces LOO· as the peroxidation chain-carrying species and unless the former radical is eliminated, a large proportion of lipid in LDL can oxidize without significant loss of vitamin E. The α-TO· is less reactive than the LOO· and as such tocopherol-mediated peroxidation represents a model of retarded lipid peroxidation relative to that caused by LOO· in the absence of vitamin E. However, as α-TOH is highly reactive towards various oxidants, the mere presence of vitamin E in LDL renders the lipoprotein susceptible to oxidation (78, 81, 497, 1051). This property is referred to as the phase-transfer activity of α-TOH. Thus, in tocopherol-mediated peroxidation, the greater the content of vitamin E in LDL, the more susceptible the lipoprotein will be to oxidation.

FIG. 9.

Tocopherol-mediated peroxidation of LDL. According to the model, α-TOH acts as a phase-transfer agent reacting with aqueous radical oxidants (R·) that results in formation of α-TO· and hence “import” of radicals from the aqueous into the lipid phase. The model predicts that there is only one radical (mostly α-TO·) contained in an oxidizing LDL particle. Suitable reductants or coantioxidants (XH) react with this α-TO· thereby regenerating α-TOH at the expense of formation of a coantioxidant-derived radical (X·) that “exports” the radical back into the aqueous phase (·Xaq) where it eventually gives rise to nonradical products (NRP). In the absence of coantioxidants, α-TO· is forced to react with a lipid molecule containing bisallylic hydrogen atoms (LH), thereby initiating a chain reaction, in which α-TO· is the chain-carrying radical (chain transfer activity of vitamin E). In this chain reaction many molecules of LOOH can be formed without consumption of α-TOH. Tocopherol-mediated peroxidation is also inhibited under conditions of high radical flux, i.e., when LDL particles encounter radical oxidants at high frequency. Under this condition (not shown), a second radical oxidant reacts with an oxidizing LDL particle before its α-TO· initiates lipid peroxidation via reaction with LH.

As mentioned above, the strength of the oxidizing conditions largely determines whether α-TOH exerts a pro- or antioxidant function in LDL. Thus, under mild oxidant flux, α-TOH may exert pro-oxidant activity (81, 664). In addition, the type of oxidant involved impacts the function of vitamin E. Whereas α-TOH may be an effective antioxidant for 1e-oxidants in lipid environments (112), it does not protect LDL against oxidative modification by 2e-oxidants such as HOCl (372) and ONOO (947). As both 1e- and 2e-oxidants are implicated in oxidative processes in atherosclerosis, it is difficult to predict the overall contribution of α-TOH to LDL oxidation in atherosclerosis.

D. Antioxidant Status in Atherosclerotic Lesions

Compared with our knowledge of lipid changes in atherosclerosis, little is known of the accompanying changes to vascular wall antioxidants. This is true even in studies of animal models of atherosclerosis. The data currently available are primarily restricted to studies of advanced vascular disease and components released from homogenates of large pieces of aortic tissue. Thus systematic studies on antioxidant changes at various disease stages and, in focal areas of the intima, are not readily available. In addition, the currently available information is not entirely consistent, for example, when data from human lesions are compared with those of lesioned material isolated from animals used as models of the human disease. Furthermore, it is worth noting that the interpretation of changes to antioxidant defenses is complicated because the same material is not also analyzed for the extent of presence of oxidized biomolecules. For example, an increase in an antioxidant defense is commonly interpreted as a biological response to increased oxidative stress without knowledge of whether this is associated with an overall decrease or increase in oxidative tissue damage, and despite the fact that oxidants do not necessarily affect the expression of antioxidant enzymes (909). With these limitations in mind, the following discussion serves as an indication of the gross changes in vascular wall antioxidants that may occur during disease progression. Recent studies addressing the stage-dependent changes in antioxidant content and quality in diseased vessels are also highlighted.

1. Proteinaceous antioxidants

There is evidence that the levels of some important proteinaceous antioxidants are altered in the diseased vascular wall (Table 10), although information on antioxidant enzyme expression in human atherosclerosis is limited (reviewed in Ref. 892).

View this table:

Contents of antioxidants in human plasma and homogenates of normal human arteries and advanced atherosclerotic lesions


The activity of EC-SOD, the major SOD isoenzyme in the arterial wall, was reported to be increased in highly cellular rabbit lesions, but decreased in advanced, connective tissue-rich human lesions (567). Overall, however, the activity of the different SOD isoenzymes does not appear to be altered drastically compared with normal arteries (Table 10) (reviewed in Ref. 892). In a recent study, 't Hoen et al. (943b) reported mRNA levels of antioxidant enzymes in the aortic arch of wild-type and apolipoprotein E −/− mice of different age fed a regular chow. The mRNA levels of Cu,Zn-SOD (SOD1), Mn-SOD (SOD2), and catalase were increased in apolipoprotein E −/− mice in the period preceding lesion formation in the aortic root (weeks 6–12), but were decreased at 34 wk, when lesions had developed (943b). As we have seen, a decrease in the activity of Cu,Zn-SOD (205) or EC-SOD (459) may lead to decreased bioavailability of ·NO and hence dysfunctional endothelium-dependent vasodilation and increased production of ONOO. Chronic inhibition of Cu,Zn-SOD in rats has also been reported to result in increased nonenzymatic lipid peroxidation (569), indicating the potential for a protective role of SODs in atherosclerosis.

In apparent contradiction to this notion however, overexpression of Cu,Zn-SOD in fat-fed C57BL/6 mice increases rather than decreases lesion formation (958). Also, after 1-mo atherogenic diet, aortic lesions in apolipoprotein E −/− mice are larger in wild-type than in EC-SOD-deficient mice, and there is no differences between the EC-SOD genotypes in the larger lesions seen after 3 mo on the diet or after 8 mo on normal chow (826). In addition, on a wild-type background, there are no effects produced by the absence or presence of EC-SOD on atherogenic diet-induced aortic root lesions (826). In these studies, the urinary excretion of F2-isoprostanes was related to the rates of atherogenesis but was not influenced by the EC-SOD genotype. Likewise, the EC-SOD status has no effect on the staining for oxidized LDL epitopes in aortic root sections (826). Together, these findings indicate that EC-SOD appears to have surprisingly little influence on atherogenesis in mice and that the role of SOD in intimal LDL oxidation and atherogenesis remains unknown.


Compared with normal internal mammary arteries, selenium-dependent glutathione peroxidase and glutathione reductase activities are decreased in human carotid atherosclerotic plaques, with no measurable changes to glutathione transferase activity (521). In contrast, selenium-independent glutathione peroxidase activity is increased in plaque (Table 10).

In lesions of rabbits fed a hyperlipidemic diet, total thiol compounds and selenium-dependent glutathione peroxidase activity has been reported to progressively rise from 10 to 60 days, whereas the activities of catalase, glutathione reductase, and glutathione transferase significantly decrease, and selenium-independent glutathione peroxidase activity is not detectable (194). At 80 days, there is a significant decrease in the vascular content of GSH that is associated with reduced activity of γ-glutamyl transpeptidase but not γ-glutamylcysteine synthase (522), both enzymes involved in the de novo synthesis of GSH. In parallel, thiobarbituric acid-reactive substances in the vessel wall increase about three times when expressed per gram tissue (194, 522). In apolipoprotein E −/− mice, mRNA levels of glutathione peroxidases in the aortic arch increase in the period preceding lesion formation in the aortic root, but they decrease at 34 wk, when lesions have developed, similar to the changes seen for SOD and catalase (943b). When Japanese quails are fed a cholesterol-enriched diet, comparable levels of glutathione-related enzymes are found between control and atherosclerotic arteries and between arteries derived from cholesterol-fed animals whether or not they develop macroscopic lesions (312). Thus the changes in glutathione-related antioxidant enzymes during atherosclerosis are inconsistent between species and between animal models of atherosclerosis. The latter study suggests that atherogenesis proceeds in the absence of gross changes to glutathione-dependent enzymatic antioxidants; however, the relevance to human atherosclerosis remains unknown.


Heme oxygenase-1 is induced by oxidative stress (992) and is expressed in cells of human atherosclerotic lesions (1012). Thus increased heme oxygenase-1 activity could represent a local antioxidant response (891), particularly if biliverdin reductase activity was present together with ferritin synthesis (28, 992). This scenario would result in removal of heme and, hence, the removal of a potential pro-oxidant, the generation of bilirubin, a coantioxidant for α-TOH (663), and the sequestration of iron (28).

There is increasing evidence that induction of heme oxygenase provides protection against atherosclerosis and related diseases. Indeed, upregulation of heme oxygenase-1 activity by heme, transfection, or gene transfer reduces atherosclerosis in mice deficient in LDL receptor or apolipoprotein E −/− mice and in porcine coronary arteries (213, 439, 457), whereas pharmacological inhibition of heme oxygenase enhances atherosclerosis in LDL receptor −/− rabbits (438). Induction of heme oxygenase-1 also protects against intimal hyperplasia in rat aorta following balloon injury (953), and expression of the enzyme can determine cardiac xenograft survival (862). In contrast to SOD, catalase, and glutathione peroxidases, the expression of heme oxygenase-1 remains increased over the course of disease development in apolipoprotein E −/− mice (943a).

Whether this apparent protective effect of heme oxygenase depends on an antioxidant activity remains unclear. What is clear is that increased heme oxygenase-1 protects against vascular constriction and inhibits the proliferation of vascular smooth muscle cells (213) via induction of apoptosis (563). Heme oxygenase-derived carbon monoxide has been implicated in this process (685), as well as other potentially protective activities such as endothelium-dependent relaxation (1099) and inhibition of apoptosis of endothelial cells (96, 863).


Developing lesions have decreased ratios of albumin and apolipoprotein A-1 to LDL (860, 1085). Relative decreases in albumin may promote the availability of transition metal ions via decreased metal binding, or lower the concentration of sacrificial thiols to scavenge 2e-oxidants. Decreased apoplipoprotein A-1 infers a relative decrease in the content of HDL in the vascular wall. This could also favor oxidative events as a number of antioxidant activities have been assigned to HDL (571). The molecular basis for an antioxidant action of paraoxonase (an aryl esterase) is obscure, yet in HDL it has been reported to exert antioxidant activity including inhibition of lipid peroxidation (571, 1017). Paraoxonase is present in interstitial fluid associated with HDL (572), and gene knockout studies demonstrate increased (although small) lesion formation with decreased paraoxonase (842). Nonetheless, direct interaction of HDL paraoxonase with LDL lipid oxidative events in developing lesions remains to be shown. A decrease in the relative concentration of intimal HDL would also decrease methionine residue availability for LOOH reduction and removal of potential pro-oxidants.

2. Nonproteinaceous antioxidants

Most of the atherosclerosis-associated changes to enzymatic antioxidants described above are likely to take place within vascular wall cells. While an altered intracellular redox environment may be atherogenic (679), evidence also suggests that changes in extracellular and nonenzymatic antioxidants play a role in disease initiation and/or progression (882). In particular, antioxidants that are associated with LDL and required to inhibit lipoprotein lipid peroxidation are thought to be deficient and/or ineffective. As plasma constituents enter normal and atherosclerotic vessels (860), all of the low-molecular-weight, nonproteinaceous antioxidants described in section iiD3 may be present in arterial walls. However, as this material is difficult to obtain, nonproteinaceous antioxidants have not been characterized systematically, and to date, their content has only been studied in tissue homogenates prepared from normal or diseased tissue (Table 10).


The transfer of aqueous nonproteinaceous antioxidants, such as ascorbate and urate, to the intima likely occurs via simple diffusion. Thus the concentration of these antioxidants in the extracellular space of the vascular wall may approximate that of the lumen. Indeed, the concentration of ascorbate in interstitial fluid (178) and lymph (633) is similar to that in plasma.

I) Vitamin C.

Normal arteries contain approximately one-third the concentration of ascorbate than normal human plasma (915) (Table 10). Although it is not clear whether ascorbate originates from extra- or intracellular compartments, it is reasonable to expect a portion of the ascorbate detected to be extracellular. An early study found the concentration of ascorbate in diseased human aorta to be comparable to that found in plasma (1038). This finding was validated recently by a study that showed elevated levels of ascorbate in advanced human atherosclerotic plaque compared with normal arteries (915). Furthermore, only small amounts of vitamin C were present as dehydroascorbic acid (the 2-electron oxidation product of vitamin C) in atherosclerotic plaque. Together, the limited information available to date suggests that there is no gross alteration or deficiency in the arterial vessel wall content in ascorbate with atherosclerosis. As indicated, these studies do not however rule out the possibility of temporal or local (e.g., intra- versus extracellular) alterations in ascorbate concentrations during atherosclerosis.

II) Urate.

Compared with human plasma, healthy arteries contain significantly less urate. In contrast, the concentrations of urate in advanced human lesions are comparable to those in plasma (Table 10) and in some lesions approach if not exceed the limit of its solubility in aqueous solution (915). Thus, like with ascorbate, currently available evidence suggests that urate levels in the vessel wall do not become decreased during atherosclerosis. It remains to be established whether uric acid provides antioxidant protection in vivo, and the role of uric acid in cardiovascular disease is still unclear, despite a long-standing association between hyperuricemia and atherosclerosis. Some studies have suggested a protective role of uric acid (43, 858). For example, reaction of urate with ONOO results in formation of a compound that can relax blood vessels, apparently via release of ·NO (858). However, some (241, 738, 759) but not all studies (170) have reported a direct association between uric acid concentrations and atherosclerosis, hypertension, and cardiovascular mortality. Of potential relevance, urate has been shown to stimulate the proliferation of vascular smooth muscle cells in vitro, and this has been proposed to be important in the setting of ischemia reperfusion where substantial increases in the vessel concentration of uric acid may occur (759). Also, and potentially interesting in the context of the suprasaturable concentrations of uric acid present in some human plaque samples (915), urate crystals derived from injured cells have been shown to act as a danger signal in inflammation (840) that may contribute to leukocyte recruitment to atherosclerotic lesions.

III) Bilirubin.

There is increasing evidence supporting an inverse association between cardiovascular disease and plasma levels of bilirubin (207, 410, 543, 822, 823, 1002). Whether this reflects a direct protection by the bile pigments, or simply a surrogate measure for increased heme oxygenase activity, remains to be established. As indicated earlier, a protective role has been attributed to the heme oxygenase product carbon monoxide (684). However, recent studies indicate that bilirubin may be protective by inhibiting the proliferation of vascular smooth muscle cells (563), by mechanism(s) yet to be established. Unfortunately, there is no direct information available at present on the concentration of bilirubin in diseased or healthy arteries.


In the healthy vessel the concentrations of lipophilic nonproteinaceous antioxidants resemble the concentration of lipoprotein lipid present in interstitial fluid (178). They are considerably lower than their respective plasma levels or the vessel concentration of albumin (860). This is not surprising, given that lipophilic antioxidants are associated with lipoproteins, the concentration of which is low in healthy vessels. By implication, lipid-soluble antioxidants detected in healthy arteries are probably located within cells (860). However, as atherosclerotic lesions develop, lipoproteins including LDL, with their full complement of antioxidants, transfer to the vessel wall from plasma. Thus it is reasonable to assume that, as lesions develop, an increasing proportion of the lipid-soluble antioxidants detected is localized extracellularly within lipoproteins and lipid depositions derived from them.

I) Vitamin E.

Concordant with the above data for ascorbate and urate, the levels of α-TOH in homogenates of early, intermediate, and advanced human atherosclerotic lesions are comparable to plasma levels of the vitamin (915, 939, 967) (Table 10). Similar results have been reported for lipoproteins isolated from human lesions representing different developmental stages (672, 939), and this is true whether α-TOH is expressed per free cholesterol molecule or per cholesteryllinoleate, the major readily oxidizable lipid in lesions. In apparent contrast, Carpenter et al. (126) suggested that α-TOH is depleted in macrophage-rich lesions, based on both comparatively low ratios of α-TOH to cholesterol and the presence of hydroxycholesterol (126). It is worth noting that the biological relevance of expressing α-TOH relative to cholesterol is not known, as the vitamin is generally considered to affect lipid peroxidation rather than oxysterol formation.

Consistent with the notion that vitamin E remains essentially intact during atherosclerosis, available data from gas chromatography-mass spectrometry analysis suggest that only a fraction of the vitamin is oxidized (939). α-Tocopherylquinone is the single major oxidation product, with maximal, albeit limited, α-TOH oxidation observed at the earliest stage of atherogenesis, at a time when the extent of fatty acid oxidation is minimal and overall less than that of the vitamin. As the severity of the disease increases, the ratio of oxidized lipid to oxidized α-TOH also increases (939). However, at all stages of atherosclerosis, the observed relative abundance of α-tocopherylquinone over 2,3- and 5,6-epoxy-α-tocopherylquinones resembles the situation when LDL is oxidized in vitro by 2e- rather than 1e-oxidants (939). These findings are consistent with and further support the notion that 2e-oxidants are primarily involved in the oxidative events taking place in the artery wall and involving vitamin E.

In the context of atherosclerosis, it may be important to note that vitamin E also has activities that may not be directly related to its involvement in lipid peroxidation (reviewed in Refs. 93, 475). Briefly, vitamin E can enhance the bioactivity of ·NO (471), inhibits smooth muscle proliferation (73), and limits platelet aggregation (259). One common mechanism to account for these effects of vitamin E is the inhibition of protein kinase C stimulation (73). In the setting of atherosclerosis, inhibition of protein kinase C by vitamin E would be expected to maintain normal vascular homeostasis and thus reduce the clinical incidence of cardiovascular disease.

II) Coenzyme Q10.

To date, little information is available on the vessel content of coenzyme Q and the potential effect of atherosclerosis on this. Suarna et al. (915) reported comparable levels of total coenzyme Q10 in homogenates of healthy human arteries and carotid endarterectomy specimens, indicating that compared with cellular sources, lipoprotein-derived coenzyme Q10 is insignificant in diseased vessels. This interpretation is consistent with the observation that in apolipoprotein E −/− mice fed an atherogenic diet the aortic content of coenzyme Q also remains largely unchanged as lesions develop (541). However, supplementing these animals with large amounts of coenzyme Q, without and with additional vitamin E, decreases atherosclerosis (948, 1053), although supplements at therapeutic concentrations do not inhibit atherosclerosis in LDL receptor-deficient rabbits (86), and even large amounts of coenzyme Q10 do not inhibit intimal hyperplasia in balloon-injured rabbits (152).

It is presently not clear whether the vascular effects seen with coenzyme Q supplementation are related to antioxidant activity. What is known is that oral coenzyme Q10 supplementation increases the concentration of CoQ10H2 in human LDL (632) and in the blood vessels of animals used as models of cardiovascular disease (152, 1053), and this is associated with a decrease in the aortic content of oxidized lipids (152, 1053). In addition to this antioxidant activity, supplemental coenzyme Q10 has been reported to decrease integrin expression by monocytes in humans (962), suggestive of anti-inflammatory activity. There is also some recent evidence that supplemental coenzyme Q10 may improve blood pressure (403) and endothelium-dependent vasodilation in type 2 diabetics (1022), although similar beneficial effects were not observed in hyperlipidemic patients (751).

E. Inhibition of LDL Oxidation

Most studies of LDL lipid peroxidation in vitro utilize lipoproteins removed from native environments, although it is now recognized that such environs play an important role in the antioxidant function of vitamin E. For example, a number of endogenous reducing agents, termed “coantioxidants,” can impede the pro-oxidant activity of α-TOH (79, 666, 949, 1059). As depicted in Figure 9, coantioxidants (denoted as XH) reduce the α-TO· and eliminate the radical character from the LDL particle, thereby inhibiting tocopherol-mediated peroxidation. Thus, for this model, a balance between the level of vitamin E and available coantioxidants determines whether LDL lipid peroxidation occurs. Vitamin C (ascorbic acid) and CoQ10H2 are examples of endogenous coantioxidants for α-TOH in LDL. Both are lost easily during isolation of the lipoprotein.

Despite the presence of apparently adequate levels of nonenzymatic antioxidants and coantioxidants, a significant proportion of cholesteryllinoleate in advanced human lesions is nonetheless oxidized (915). As discussed above, so long as ascorbate and CoQ10H2 are present, tocopherol-mediated peroxidation of lipoprotein lipid is effectively prevented in vitro. An explanation for the coexistence of large amounts of oxidized lipid and ascorbate in advanced atherosclerosis is presently elusive. However, the above does not reflect oxidized lipid and antioxidant levels at focal areas of lesions as whole tissue homogenates were used for analysis. Hence, it may be that oxidative damage occurs at sites remote from available antioxidants, for example, in the extracellular matrix where LDL may become trapped and fuses to large vesicles (255). This rationale does not likely extend to the coantioxidant CoQ10H2, which associates with lipoproteins. However, there are no data available on the amount of CoQ10H2 in lesions (915).

To address the above conundrum, that is, antioxidants and oxidized lipids apparently coexist in lesions, and the fact that the levels of reduced, lipophilic coantioxidants for α-TOH in atherosclerosis is presently unknown, animal intervention studies employing coantioxidants have been carried out. As indicated above, the physiological levels of natural coantioxidants, in particular, ascorbate and CoQ10H2, can be manipulated by dietary supplementation. Thus coantioxidation is amenable as a potential antiatherogenic strategy. Importantly, results from intervention studies show that aortic lipoprotein lipid peroxidation can be effectively inhibited by lipophilic coantioxidants (152, 948, 1053, 1055, 1056). This is particularly striking in situations where coantioxidants were tested that are kinetically inferior peroxyl radical scavengers compared with α-TOH (1055, 1056). Therefore, these results support the notion that inhibition of tocopherol-mediated peroxidation is a useful strategy to inhibit lipoprotein lipid oxidation in the vessel wall.

F. Problems With the Oxidative Modification Hypothesis

As reviewed in the preceding sections, a large body of evidence supports the view that oxidative modification of LDL contributes to atherogenesis, likely in a variety of different ways. Despite this however, and as we will see in the following, there is also literature evidence inconsistent with the oxidative modification hypothesis of atherosclerosis.

1. LDL oxidation versus atherosclerosis

Over the last decade we have learned much about the chemistry underlying alterations in the redox status of antioxidants, lipids, amino acids, and proteins as well as the potential roles of ROS and RNS in cellular signaling, yet there remain lingering uncertainties about where, how, and to what extent LDL becomes oxidized in vivo, and how these processes directly relate to atherogenesis.


A common argument used to suggest that LDL oxidation must occur in the intima rather than plasma is that, relative to plasma, intima has a low antioxidant capacity. However, as we have seen (sect. iiiD), the concentration of major extracellular antioxidants is surprisingly high in human atherosclerotic lesions. In fact, in the case of ascorbate, urate, and α-TOH, the plaque concentrations approach those of human plasma (Table 10) so that the arguments put forward against the occurrence of oxidation of LDL in the circulation also hold true for the intimal space. It appears pertinent, therefore, that studies on LDL oxidation consider the presence of such antioxidants in general. This may be particularly true for α-TOH, the major antioxidant associated with LDL.

As mentioned earlier, a vast number of studies have established that in vitro, radical oxidants like copper ions or cells cultured in transition metal ion-containing medium can convert native LDL into oxidized or high uptake LDL. Such conversion requires drastic alterations to the lipoprotein particle, including the complete depletion of endogenous α-TOH. This is in sharp contrast to the situation in vivo where, irrespective of the developmental stage of atherosclerosis, homogenates of diseased vessels or apolipoprotein B-100-containing lipoproteins isolated therefrom, are neither depleted of α-TOH nor is the vitamin substantially oxidized, and a majority of lipoprotein-associated oxidized lipids are formed in the presence of α-TOH (see sect. iiiB2a). This apparent discrepancy questions the general pathophysiological relevance of in vitro generated oxidized LDL and its potential proatherogenic activities (see sect. iiiB6) (Table 8). It also questions the pathophysiological relevance of minimally modified LDL (see sect. iiiC3). Thus minimally modified LDL is formed in vitro only after complete depletion of its antioxidants (58, 1019). Specifically, formation of oxidized arachidonic acid products with chemotactic activity represents secondary lipid oxidation, a process largely prevented as long as α-TOH is present. Therefore, depletion of the vitamin appears to be a prerequisite for formation of minimally modified LDL, and the extent of its oxidative modifications is substantial when compared with the in vivo situation, although it is “minimal” when compared with LDL exposed to redox-active copper concentrations for prolonged periods.

One way to overcome the above-described conundrum is to speculate that in the arterial wall LDL oxidation occurs in “microenvironments” where oxidants are produced excessively and/or where the antioxidant shield is no longer intact. The macrophage phagosome may represent such a microenvironment, although this would imply that oxidation of LDL takes place in the intracellular space, and it is not a prerequisite for lipoprotein uptake by macrophages, a notion itself inconsistent with the hypothesis that LDL oxidation causes foam cell formation. In vitro aggregation of LDL or its aggregation with proteoglycans does not appear to generate microenvironments from which coantioxidants like ascorbate are excluded, as the vitamin remains antioxidant active in such systems (967). An alternative way to overcome the conundrum is to speculate that in vivo LDL oxidation is mostly a consequence of reactions with 2e-oxidants rather than radicals. Indeed, there is ample and increasing evidence for the involvement of oxidants like HOCl and ONOO (see sect. iiiC1), although in this scenario too, one would expect ascorbate to be protective at least in part, and it remains unclear whether the extent of such modifications seen in vessels or LDL isolated from them are sufficient to explain foam cell formation. An attractive aspect of this interpretation is, however, that it could help explain the overall disappointing outcome of human interventions with antioxidant supplements (see sect. iiiF2), as the antioxidants tested to date are radical scavengers and, at least in the case of α-TOH, offer little protection against LDL oxidation induced by 2e-oxidants.


It is important to recognize that a large body of the support referred to in this review to substantiate the oxidative modification hypothesis of atherosclerosis provides indirect rather than direct evidence for a causative link between the two processes. This is perhaps not surprising given the difficulties in experimental attempts to distinguish LDL oxidation as a cause rather than consequence of atherosclerosis. For example, associations such as the relative extent of LDL oxidation in the vessel wall and disease burden at best only strengthen the oxidative modification hypothesis; they do not prove the hypothesis.

I) Correlation between lipid oxidation and atherosclerosis.

As the formation of oxidized LDL in the vessel wall is thought to be an early event in atherogenesis and to contribute to on-going pathogenesis, it appears reasonable to assume that oxidized lipids are formed and accumulate early in disease and that their concentrations increase as lesions develop and decrease as lesions regress. Indeed, it has been reported that compared with C57Bl/6 mice, aortic concentrations of F2-isoprostanes are increased in apolipoprotein E −/− mice that spontaneously develop atherosclerosis and that have been reported to respond to vitamin E supplements by decreasing lesion size and tissue F2-isoprostanes (733). Similarly, the same group reported that hepatic gene transfer of apolipoprotein E decreases atherosclerosis and tissue concentrations of F2-isoprostanes in LDL receptor −/− mice fed an atherogenic diet (933) and that plasma levels of F2-isoprostanes directly correlate with lesion area in mice deficient in apolipoprotein E and 12/15-lipoxygenase (175). A potential problem with these studies is, however, that tissue levels of F2-isoprostanes were not standardized for arachidonic acid (from which they are derived) so that the results reported do not distinguish differences in lipid load (i.e., disease) versus relative extent to which the lipids that have accumulated are oxidized. It is also noteworthy that others failed to observe that vitamin E supplements decrease arachidonic acid-standardized concentrations of aortic F2-isoprostanes in apolipoprotein E −/− mice (948). Furthermore, in human atherosclerotic lesions, F2-isoprostanes represent minor lipid oxidation products, and the functional relevance of F2-isoprostanes for atherosclerosis remains unclear (580, 1006).

Compared with F2-isoprostanes, HETEs and HODEs in general, and their respective cholesterylesters in particular, are much more abundant in human lesions. In fact, these oxidized lipids can account for up to ∼1% of the corresponding parent lipids, yet the relationship between these oxidized lipids and atherosclerosis remains unclear (see sect. iiiB2a). Early studies reported the ratio of HODEs to linoleic acid in the abdominal aorta of men with chronic ischemic heart disease to increase slightly from stage I to III (507). Carpenter et al. (127) observed generally higher absolute concentrations of HODE in diseased than normal arteries. However, these authors (127) did not observe a disease stage-dependent increase in HODEs. Similarly, Waddington et al. (1006) reported comparable levels of HODEs in advanced lesions varying from type V to type VI. In addition, these authors failed to confirm an earlier study (580) reporting that increased HETEs and HODEs are associated with symptomatic atherosclerotic plaque in human carotid arteries (1006). Perhaps the most compelling evidence against a causative role for LDL oxidation in atherosclerosis comes from a recent study (966) reporting parent molecule-standardized concentrations of esterified HODEs in human aortas to increase significantly only at late developmental stages of atherosclerosis. Similarly, antioxidant intervention studies have revealed situations where aortic concentrations of esterified HODEs do not correlate with disease burden (1054, 1055).

Overall, it therefore appears that the levels of oxidized lipids correlate more weakly with lesion development than would be predicted by the oxidative modification hypothesis, despite the fact that established lesions contain increased concentrations of fatty acid oxidation products compared with healthy arteries. In addition, there is evidence that in human lesions oxidized lipids associated with LDL accumulate substantially only late in disease development, after accumulation of unoxidized cholesterol and cholesterol esters (966). Furthermore, lipoprotein lipid oxidation can be dissociated from disease in animal models of atherosclerosis (899, 1005, 1054, 1055). While these studies do not support a cornerstone of the oxidative modification hypothesis of atherosclerosis, namely, that LDL oxidation in the artery wall is an early event in the genesis of the disease and that it contributes to subsequent lipid accumulation, the data do not directly address the causal or temporal relationship between intracellular oxidative stress and disease progression. However, available literature such as the lack of clear disease stage-dependent increase in total HODEs (i.e., that include lipoprotein and cellular oxidized lipid) suggests that at least cellular lipid oxidation may not correlate with disease progression.

II) Animal intervention studies with antioxidants.

As we have seen in section iiiB8, several synthetic antioxidants inhibit lesion progression in animal models of atherosclerosis. It is important to point out, however, that not all synthetic antioxidants offer protection against the disease in animals (reviewed in Ref. 894). This is particularly intriguing in situations where the antioxidants have been shown to decrease the extent of oxidation. For example, supplementation of a butter-based atherogenic diet with butylated hydroxytoluene and butylated hydroxyanisole (both phenolic lipid-soluble antioxidants that effectively inhibit LDL oxidation) do not prevent atherosclerosis in rabbits (1042). Also, compared with probucol, its structural analog bis(3,5-di-tert-butyl-4-hydroxy-phenylether)propane offers superior protection to LDL against in vitro oxidation, yet the analog is ineffective in inhibiting atherosclerosis in LDL receptor-deficient rabbits (275). Perhaps even more striking, another structural analog of probucol, 3,3′-5,5′-tetra-tert-butyl-4,4′-bisphenol, prevents lipoprotein lipid oxidation in the vessel wall of LDL receptor-deficient rabbits as effectively as probucol, yet unlike probucol, the analog does not protect against atherosclerosis (1055). Conversely, probucol inhibits atherosclerosis in the aortic arch and thoracic and abdominal aorta of apolipoprotein E −/− mice without inhibiting aortic lipoprotein oxidation (1054). These latter studies establish that at least in animals, the process of lipoprotein oxidation can be dissociated from atherosclerosis, a finding inconsistent with the oxidative modification hypothesis of atherosclerosis.

In addition to the synthetic lipid-soluble antioxidants referred to above, vitamin E has been used repeatedly for intervention studies in a variety of experimental animal models (Table 11). This is not surprising, considering it is the most abundant endogenous antioxidant associated with LDL, although its role in protecting LDL lipids against oxidation is more complex, as we have seen in section iiiC4. As with the human clinical trials (see sect. iiiF2), a majority of the studies report a null effect of vitamin E supplementation on lesion formation in animals on a normal diet. Eleven of the 44 studies carried out over the last 50 years show vitamin E to attenuate disease. In 4 of these 11 studies, vitamin E supplements lowered plasma lipids so that this rather than an antioxidant function may have been responsible for the outcome. Thus an antiatherogenic effect independent of lipid lowering has in fact been observed in only seven studies (70, 177, 730, 733, 740, 921, 948). Notably, a similar number of investigations (n = 5) have shown increased lesion formation with vitamin E supplements, particularly when given at high concentration (101, 311, 469, 643, 969). The overall marginal impact of α-TOH on lesion formation is perhaps not so surprising given the lack of evidence for a deficiency in this antioxidant (541). Similar to the situation with vitamin E, vitamin C supplements do not offer consistent benefit against atherosclerosis in animals (reviewed in Refs. 468, 568).

View this table:

Vitamin E supplements and atherosclerotic disease in animals

As indicated, interpretation of the early animal intervention studies is often complicated by a cholesterol-lowering effect of vitamin E (88, 730, 1030, 1042). Some, though not all, of the more recent studies of atherosclerosis in mouse models have reported vitamin E supplements to inhibit disease. An important point when comparing studies, however, is that many murine studies examine atherosclerosis only in the aortic root, a site that may be less sensitive to manipulation of oxidative stress than sites more distal in the aorta (41, 948). Also, the interpretation of animal studies that utilize nutrient vitamins is not entirely straightforward. Many of these studies lack standardization, as control diets among studies are not comparable. In particular, the tocopherol content of “standard” laboratory diets can vary considerably. This may have major implications on experimental findings as reported previously by others (537, 1039). Interestingly in this context, studies that examine the effect of vitamin E supplements in otherwise vitamin E-deficient animals appear to more frequently report inhibition of lesion formation (687, 704, 919, 1073) (Table 11). The relevance of this to human atherosclerosis is questionable, however, as deficiency of vitamin E is rare.

III) Studies with transgenic and knockout mice.

Several recent studies utilizing transgenic and knockout mice have provided support for the oxidative modification hypothesis of atherosclerosis. Perhaps most strikingly, mice deficient in scavenger receptors have been reported to show decreased atherosclerosis (243, 925) (see sect. iiiB7). As these receptors recognize oxidized LDL, it seems reasonable to attribute the observed protection to a decreased uptake of oxidatively modified LDL and hence a decrease in the progression of atherosclerosis. However, this mechanism has not been proven experimentally. Importantly, the extent of decrease in atherosclerosis observed was only modest, indicating that the scavenger receptors and uptake of oxidized LDL are not major contributors to atherosclerosis in these mice. Also, scavenger receptors have functions beyond the uptake of oxidized LDL (203b). For example, the scavenger receptor A also plays an important role in the adhesion of macrophages to extracellular matrix (799) and cells, thereby participating in “proinflammatory” action providing for the recruitment of mononuclear phagocytes to and their retention in ligand-rich tissues such as in atherosclerotic lesion (256). In addition, and similar to the situation with antioxidant intervention studies, not all results from atherosclerosis experiments are in agreement (203b). For example, in APOE3Leiden mice, a deficiency in the scavenger receptor A results in increased atherosclerosis (203a), and macrophage-specific overexpression of class A scavenger receptor decreases atherosclerosis in LDL receptor −/− mice (1032). Thus scavenger receptor class A is a multifunctional player in atherosclerosis, and the original observation by Suzuki et al. (925) does not provide direct evidence for a role of oxidized LDL in atherogenesis.

As discussed in section iiiC1d, there is evidence for a role of lipoxygenases in atherosclerosis. However, not all data support a causal role for 15-lipoxygenase-induced oxidative events and atherosclerosis. For example, compared with controls, macrophage-specific overexpression of 15-lipoxygenase was reported to decrease atherosclerosis (838), whereas the reported marginal presence of 15-lipoxygenase-specific, oxidized lipids in human lesions (251) was not confirmed in studies by others, not even in lesions representing early developmental stages of the disease (967). In addition, it should be pointed out that lipoxygenases could conceivably affect atherosclerosis via their multiple effects on inflammatory processes (218), and this could be unrelated to oxidative events. Therefore, present knowledge does not provide direct support for a contribution of lipoxygenases to oxidative events evoked by the oxidative modification hypothesis of atherosclerosis.


As discussed in section iiiB, there is a large body of evidence for the presence of oxidized lipid and protein in human atherosclerotic lesions. However, in most of these studies, these oxidative changes were not directly linked to lesion LDL and, where this was done, the reports are conflicting. For example, while one group reported LDL isolated from atherosclerotic lesions to possess properties that resemble those of LDL oxidized in vitro with copper ions (1089), other groups reported lesion LDL to be comparatively less substantially oxidized (672, 884). A complication with these latter studies is the argument that drastically oxidized LDL may be expected to be taken up rapidly by macrophages within the vessel wall and hence not available for “extraction” from lesion material. This argument is, however, not consistent with the findings of Ylä-Herttuala et al. (1089). It is also not immediately applicable to minimally modified LDL that is not recognized by scavenger receptors. There is evidence for the presence of the specific, biologically active oxidized phospholipids of minimally modified LDL in lesions of rabbits fed an atherogenic diet, as assessed by studies with the monoclonal antibody EO6 (1018). These oxidized lipids colocalize to foam cells (693) so that the simplest explanation is that they represent secondary lipid oxidation events taking place in cells rather than lesion lipoproteins. This interpretation is consistent with the preferred presence of other secondary lipid oxidation products, such as F2-isoprostanes (731) and malonyldialdehyde (1089) in vessel wall cells rather than the extracellular space. Irrespective of the precise origin however, these oxidized fatty acids have potential proinflammatory activities that may contribute to atherosclerosis.

The role of antioxidized LDL autoantibodies in the development of atherosclerosis also remains unclear (reviewed in Ref. 1086). In addition to supportive evidence (see sect. iiiB5), a large number of studies do not support the notion that the titer of autoantibodies to epitopes of oxidized LDL track with atherosclerosis (437, 954, 974), coronary heart disease (981a, 996), myocardial infarction (817), or restenosis (220). This and the fact that the overlap in autoantibody titer between control and patient groups is often considerable and shows wide standard deviations clearly limit the usefulness and clinical and prognostic value of measuring autoantibody titers (959).

2. Cardiovascular disease epidemiology and antioxidant intake

In comparing populations with similar serum cholesterol values, rates of cardiovascular death and disease vary, and this relates to several dietary and nondietary factors. A consistent finding of the study of these populations is that consumption of diets rich in antioxidants offers benefit for cardiovascular disease (301) and, furthermore, that plasma antioxidant levels reflect such benefit (770). Understandably, this has led to a large number of studies investigating the impact of dietary antioxidants in general and vitamin E in particular on cardiovascular disease. As with the animal studies, a benefit with vitamin E supplements was anticipated. However, the overall outcome, particularly the results of the large, randomized controlled studies, has been disappointing and arguably provides the strongest evidence against the oxidative modification hypothesis of atherosclerosis.


Large population prospective cohort studies have examined the relationship between self-reported intakes of antioxidants and rates of cardiovascular disease. In the 87,000 population Nurses' Health study (876) and the 40,000 population male Health Professionals' Follow-up Study (771), benefit was derived with 100 IU/day vitamin E supplement. However, low-dose or dietary vitamin E was ineffective. In contrast, a population study of 35,000 postmenopausal women showed significant reduction in cardiovascular risk with a modest increase in dietary but not with supplemental vitamin E (513). Thus, while some self-reported studies seemingly suggest a role for vitamin E in diminishing cardiovascular disease risk, it appears that greater benefit is attributable to small increases in dietary vitamin E rather than large doses of supplemental vitamin E.

Similar to the situation with vitamin E, the evidence linking cardiovascular disease and vitamin C is inconsistent, as reviewed recently (131). Several prospective cohort studies reported high dietary vitamin C intake or supplementation to associate with reduced risk of cardiovascular disease. It is worth noting, however, that such studies tend to favor trends within the group of survivors (i.e., a survivor bias) that potentially skew the relative risk assessment. Nevertheless, different studies reported an association (n = 8) or no significant association (n = 8) of dietary vitamin C with cardiovascular disease incidence (131).


Clinical studies of antioxidant protection against cardiovascular disease have focused most attention on vitamin E, based on scientific studies describing lipid oxidation as a key determinant of atherosclerosis and also epidemiological studies showing significant benefit despite small differences in plasma concentrations between populations of low and high dietary antioxidant intake (301).

The results of the major clinical trials to date designed to prevent cardiovascular disease at the primary level are summarized in Table 12. Two of these trials investigated the effect of supplemental vitamin E on the traditional end points of myocardial infarction, cardiovascular disease, and stroke. For these two trials, vitamin E, equivalent to 50 or 448 IU/day vitamin E, or a placebo was administered. The population numbers varied from 4,500 to 29,000 and all subjects had risk factors. Follow-up was for 3.6–6.1 years before the relative risk for cardiovascular events was reported. In essence, neither study showed benefit for cardiovascular risk with supplemental vitamin E. The Finnish Alpha-Tocopherol Beta Carotene Cancer Prevention (ATBC) study (943a) reported no effect of vitamin E on the incidence of fatal or nonfatal myocardial infarction. An increased risk of hemorrhagic stroke with vitamin E supplementation was observed in this study population but not in several secondary prevention studies (see below), and as such, the importance of this is unknown. The Collaborative Primary Prevention Project (PPP) similarly showed no effect on cardiovascular deaths by vitamin E, although this study was halted early as it confirmed a protective effect by aspirin seen in other trials (728).

View this table:

Clinical studies of vitamin E supplementation and cardiovascular disease

In addition to the traditional clinical end points (stroke, cardiovascular death, and myocardial infarction), carotid intima-to-media thickness has been used as a marker of on-going atherosclerotic disease. This marker is a useful surrogate for longitudinal studies, although it correlates only weakly with disease measured by angiography (5). The Antioxidant Supplementation in Atherosclerosis Prevention (ASAP) study examined the effect of vitamin E or vitamin C. After a follow-up of 3 years, the rate of progression of intima-to-media thickness was not decreased by supplementation with either antioxidant (Table 12) (794). The Vitamin E Atherosclerosis Prevention Study (VEAPS) examined the effect of 400 IU/day d,l-α-TOH on the progression of carotid artery intima-to-media thickness in men and women with LDL cholesterol ≥3.37 mM (130 mg/dl) and no clinical signs or symptoms of cardiovascular disease (404). Vitamin E supplements for 3 years did not reduce the progression of atherosclerosis compared with subjects randomized to placebo; rather, there was a borderline disease-promoting effect of α-TOH supplements (404). Similarly, in a small, crossover study involving young patients with homozygous familial hypercholesterolemia, intima-to-media thickness increased with vitamin E supplements (400 mg/day) for 2 years, but decreased when subjects received statin therapy (744). Thus vitamin E supplementation fails to slow (or inhibit) the progression of intima-to-media thickness in healthy men and women at low risk for cardiovascular disease (Table 12).


To date, six secondary prevention clinical trials have been conducted to investigate the effect of vitamin E supplements in patients with preexisting cardiovascular disease (Table 12). Of these, two studies appear to show favorable effects with vitamin E. The Cambridge Heart Antioxidant Study (CHAOS) reported a major reduction in the risk of nonfatal acute myocardial infarction with vitamin E supplements, although this result was somewhat counterbalanced by the finding that vitamin E caused a nonsignificant increase in fatal myocardial infarction (887). Similarly, in a small group of hemodialysis patients (SPACE study), vitamin E supplements resulted in a significant decrease in rates of acute myocardial infarction (68), raising the possibility that certain subjects, for example, those with renal failure and perhaps increased oxidative stress, may benefit from supplemental vitamin E.

The extent of benefit on cardiovascular outcome in the CHAOS and SPACE study is considerable in light of the relatively short duration of follow-up of 1.4 yr. It is also inconsistent with other studies. Thus supplementation with vitamin E was found to exert no benefit on cardiovascular events in the secondary prevention subgroup of the ATBC trial (762). In the large GISSI Prevenzione trial, the effect of supplemental vitamin E, n-3 polyunsaturated fatty acids, or both on cardiovascular death, nonfatal myocardial infarction and stroke was examined in 11,000 patients with a recent history of myocardial infarction (435). In this study, polyunsaturated fatty acid intake significantly decreased cardiovascular event risk, whereas vitamin E produced only a nonsignificant trend. The combination of vitamin E and polyunsaturated fatty acids was neither additive nor interactive. This result is difficult to resolve with scientific studies supporting indirect evidence for the oxidation of polyunsaturated fatty acids in LDL in atherosclerosis. In this case, an adverse effect of polyunsaturated fatty acid supplementation would be expected as the susceptibility of LDL to oxidation increased and the antioxidant activity of vitamin E would be expected to be protective.

In the Heart Outcomes Prevention Evaluation (HOPE) study, patients with cardiovascular disease or diabetes plus another risk factor were supplemented with vitamin E or ramipril (angiotensin-converting enzyme inhibitor) or a combination (1094, 1095). Vitamin E treatment did not affect any cardiovascular event, whereas ramipril significantly reduced the risk of acute myocardial infarct, cardiovascular death, and stroke. The negative result with vitamin E is compellingly at odds with that in the CHAOS study despite similar supplementation usages and longer duration time (4.5 vs. 1.4 yr). The differences between the HOPE and CHAOS studies may lie in baseline medication (aspirin and calcium antagonists) usage and the study of a country-specific (CHAOS) versus multinational (HOPE) population. However, future secondary prevention strategies for cardiovascular disease will place importance on demonstrating that any benefit from antioxidants, including vitamin E, is incremental to the commonly prescribed therapeutic agents.

A substudy (SECURE) of the HOPE trial investigated the rate of intima-to-media thickness progression in secondary prevention (Table 12). Similar to the above, vitamin E realized no benefit, whereas ramipril significantly and dose-dependently reduced risk (565). Thus, hitherto, studies using intima-to-media thickness as a surrogate marker for atherosclerosis have returned a null effect of vitamin E on cardiovascular event risk.

Consistent with the data reported above, a recent meta-analysis of randomized, controlled trials with vitamin E including a total of 81,788 patients concluded that vitamin E supplements do not reduce the risk of cardiovascular end points (1003).


The effect of vitamin E on cardiovascular health has also been examined in combination supplementation studies with other antioxidants in both primary and secondary prevention trials (Table 13). In the ASAP study, the combination of vitamins E plus C was also tested, and this significantly decreased the intima-to-media progression rates in men (794). In postmenopausal women with ≥15–75% stenosis at baseline (WAVE), progression of coronary atherosclerosis worsened nonsignificantly with 800 IU vitamin E plus 1 g vitamin C daily for 2.8 yr compared with vitamin placebo (1016). The ATBC clinical study used a combination of vitamin E and β-carotene in men as a secondary prevention strategy; however, no benefit on major coronary events was established (762). The large MRC/BHF Heart Protection Study (HPS) for secondary prevention also examined the benefit of antioxidant combination (vitamins E and C and β-carotene). Although the supplementation regimen increased blood antioxidant levels substantially, no significant reductions in the mortality from, or incidence of, any type of vascular disease, or other major outcome was found (380a). This was in contrast to the beneficial effect observed with simvastatin (380b).

View this table:

Clinical studies of antioxidant combination supplements and cardiovascular disease

Two further studies have investigated antioxidant combination supplements together with lipid-lowering therapy. In the secondary prevention HDL Atherosclerosis Treatment study (HATS), subjects were randomized to simvastatin (lipid-lowering) plus niacin treatment, or antioxidants (vitamin C, α-TOH, β-carotene, and selenium) or a combination of both. Compared with no treatment, only simvastatin/niacin significantly lowered stenosis progression rate and favorably altered plasma lipid profiles (99). Antioxidant supplementation alone had no significant effect on clinical end points but, notably, when used in combination with simvastatin/niacin, antioxidants negated the benefit of the latter on plasma lipid profile and stenosis progression. The MRC/BHF Heart Protection Study (HPS) also examined antioxidant supplements in combination with simvastatin. While this combination did not counteract the favorable effect of simvastatin on mortality and cardiovascular events (in contrast to the previous study), it also did not benefit any outcome including cardiovascular event risk (380a).

In a recent small study involving 40 patients, the effect of vitamins C and E in addition to pravastatin on progression of cardiac transplant-associated arteriosclerosis was studied, with the change in average intimal index measured by intravascular ultrasound as the primary end point (242). Compared with control, antioxidant supplements reduced plaque growth independent of any change in endothelial function. In addition to the limitation associated with the small sample size of this study, it is worth noting that the underlying pathogenesis of transplant-associated arteriosclerosis is a fibroproliferative response to chronic allogenic immune activation, which differs from that of atherosclerosis. Also, the fibroproliferative response leading to restenosis after angioplasty has been reported to be inhibited significantly by probucol, but not multivitamins (1,400 IU vitamin E, 1 g vitamin C, and 60,000 IU β-carotene per day for 6 mo) (934).

In summary, high dietary intake of vitamins E and C is associated with reduced risk of cardiovascular disease. However, trials in atherosclerosis-related cardiovascular disease indicate that supplements with vitamin E or vitamin C alone do not provide a general benefit, as can be demonstrated with, e.g., statins. This is also the case for the combination of vitamin E with other antioxidants or with lipid-lowering therapy, which do not consistently convey cardiovascular benefit, in either primary or secondary prevention. It is possible, however, that certain subpopulations of patients, such as those with endothelial dysfunction, benefit from supplements with vitamin C (see sect. ivC3).


The oxidative modification hypothesis (Fig. 6) centers largely around the early events in atherosclerosis and the formation of lesions. However, there are oxidative events other than LDL oxidation that are involved in atherosclerosis and are thought to contribute to the clinical manifestations of the disease. For example, ROS and RNS have been implicated in smooth muscle cell proliferation, metalloproteinase activation, and endothelial function. A complete understanding of how oxidative events are involved in atherosclerosis will require some discussion of why atherosclerosis results in clinical events such as myocardial infarction and stroke.

A. Concept of Disease Activity

Traditionally, atherosclerosis was thought of as simply a problem of lesion growth into the arterial lumen. According to this paradigm, continual lesion expansion would progressively narrow the arterial lumen and eventually limit blood flow, thereby causing tissue ischemia. Inevitably, it was thought that arterial occlusion would result and necrosis of the target organ would ensue. It has become clear in recent years that the scenario outlined above is not correct and that clinical manifestations of atherosclerosis are the consequence of abrupt lesion disruption and subsequent thrombosis. In addition, it is also clear these events often occur in the absence of critical luminal narrowing. The specific nature of these events is the subject to which we now direct our attention.

A vexing problem in the cardiovascular literature is the prediction of patients that will experience a clinical event. To this end, angiographic studies have determined that more severely stenosed arteries frequently progress to coronary occlusion (10, 229). However, owing to their greater frequency, the less severe stenoses are the predominant source of coronary occlusions (10). Thus myocardial infarction most commonly arises from stenoses that are relatively mildly (<50%) blocked. Thus clinical manifestations of atherosclerosis cannot be adequately predicted from angiographic images.

The observations outlined above beg the question of how less severe atherosclerotic lesions give rise to clinical events. The short answer to this question is that traditional concepts of atherosclerosis and encroachment of the arterial lumen were incorrect. This realization stems, in large part, from the observations of Glagov et al. (306). In a series of pathological specimens, Glagov et al. (306) found that the development of early atherosclerotic lesions was characterized by compensatory enlargement of the artery such that luminal cross-sectional area was preserved and the bulk of early lesions are outside the arterial lumen. This information has subsequently been confirmed using imaging techniques such as intravascular ultrasound. If one examines early lesions using this technique, it is readily apparent that large atheroma can coexist with arteries that have preserved luminal architecture. Conventional angiography only images the lumen, thereby underestimating the extent of atherosclerosis within the artery. This concept, along with the typical appearance of an atherosclerotic lesion, is contained in Figure 10. Thus a normal-appearing artery on angiography can harbor very bulky atherosclerotic lesions.

FIG. 10.

Conceptual view of appearance and development of atherosclerotic lesions. According to the so-called “Glagov” hypothesis (306), the early stages of atherosclerotic lesion development are characterized by compensatory enlargement of the artery, such that the luminal diameter is preserved. It is only in later stages of atherosclerosis that the lumen becomes smaller. The result of this effect is that considerable atherosclerosis can exist before any detection of luminal narrowing is observed on angiography. [Adapted from Glagov et al. (306).]

Although the concept of compensatory arterial enlargement explains the coexistence of advanced atherosclerosis and normal arterial lumen, it does not explain how such lesions actually lead to luminal occlusion. DeWood et al. (204) provided considerable insight into this issue by performing immediate coronary angiography on patients suffering from acute myocardial infarction. They found that most myocardial infarctions involved total occlusion of the coronary artery that was principally due to thrombus. Thrombosis of the vessel appears to result from rupture of the atherosclerotic plaque and the exposure of its thrombogenic components to the bloodstream. This understanding is derived from pathological studies and investigations with both angioscopy and intravascular ultrasound. Thus plaque disruption and consequent thrombosis underlie most clinical atherosclerotic events.

Based on the preceding paragraphs, one can appreciate that the development of atherosclerotic lesions simply sets the stage for clinical events such as myocardial infarction and stroke. Indeed, most atherosclerotic lesions are clinically silent and may exist for years without any meaningful sequelae. It is only when atherosclerotic plaques become “active” that clinical events ensue. As outlined above, the active component of atherosclerosis involves plaque rupture and a number of derivative events, including platelet adhesion, thrombosis, and vasospasm. We now consider these “active” components of atherosclerosis and how oxidative events contribute to this phase of atherosclerosis.

B. Plaque Disruption

To understand this phenomenon, it is important to have a working knowledge of plaque anatomy. The “classic” mature atherosclerotic lesion (Fig. 10) involves a central core of foam cells with extracellular cholesterol arranged in so-called cholesterol “clefts.” Typically, there is also a considerable amount of necrotic debris, and overlying the central core is a fibrous cap comprised of extracellular matrix, smooth muscle cells, and collagen. Lesion activation is initiated by rupture of the atherosclerotic plaque such that the plaque contents are exposed to the luminal surface of the artery. This plaque rupture tends to involve a structural failure of the fibrous cap.

1. The fibrous cap

Although rupture of the atherosclerotic plaque was proposed in the first half of this century (528), it is only recently that autopsy studies have consistently identified morphological features associated with this phenomenon. These features include a large necrotic core of lipid and cellular debris and a thin fibrous cap that is often eccentric (187, 188), prompting concern about mechanical stresses on the fibrous cap. In particular, the presence of a large, soft lipid core focuses available forces on the fibrous cap (768) at its junction with more normal architecture, the so-called “shoulder” region (Fig. 10). This region of the plaque, coincidentally, is most often the site of plaque rupture based on autopsy studies (188). Thus it is not surprising that determinants of fibrous cap structural integrity have become the focus of considerable investigation, in part, to provide insight into the molecular events predisposing plaques to rupture.

The extracellular matrix is a major component of the fibrous cap, and normally, matrix production and turnover are remarkably slow (831). In atherosclerosis, the environment of injury and inflammation leads to an enhanced synthetic activity of matrix components such as elastin, collagen, and proteoglycans (557). For example, inflammatory cells such as foam cells and monocyte-derived macrophages within the plaque produce an environment that is replete with the cytokines and growth factors that have important implications for matrix production. In particular, cytokines such as transforming growth factor-β may stimulate collagen synthesis, whereas others such as interferon-γ suppress it. Thus any propensity toward matrix degradation may have serious structural implications for the fibrous cap.


A wealth of information now indicates that matrix-degrading enzymes including serine proteases (tissue-type and urokinase-type) plasminogen activators and plasmin, the matrix metalloproteinases, and cysteine proteases can degrade the structural components of the fibrous cap. Table 14 lists selected matrix-degrading enzymes implicated in atherosclerosis along with their typical substrates. Considerable evidence indicates these matrix-degrading enzymes are active in atherosclerosis and that their production is enhanced in situations known to be relevant for atherosclerosis. For example, human lesions contain matrix metalloproteinases-3 and -1 (394), with the former demonstrating intense staining in the shoulder regions of atherosclerotic plaques (668). Cultured smooth muscle cells express little matrix metalloproteinase-1 activity required to cleave fibrillar collagen. However, the atherosclerotic plaque contains a milieu of cytokines and growth factors including tumor necrosis factor-α and interleukin-1β that induce smooth muscle cells to produce enzymes capable of degrading structural collagen and other matrix components (284). In addition, macrophages from arterial lesions in rabbits demonstrate production of matrix metalloproteinases-1 and -3, consistent with cell culture findings (285). Thus the cellular environment of an atherosclerotic plaque contains a number of features that would be expected to promote degradation of the fibrous cap.

View this table:

Selected matrix-degrading enzymes implicated in atherosclerosis

A similar argument also applies to the oxidative events of atherosclerosis. Atherosclerosis is associated with an increased production of ROS and RNS that have implications for matrix metalloproteinase activity derived from smooth muscle cells. In particular, H2O2 increases smooth muscle cell gelatinase activity and ONOO activates matrix metalloproteinase-9 to generate collagenase activity (753). In transformed cells, the sustained production of H2O2 is associated with the activation of matrix metalloproteinases-2 through a tyrosine kinase-dependent mechanism. The production of HOCl is also associated with metalloproteinase activity, both directly and indirectly. With regard to the former, HOCl can fragment components of the extracellular matrix (1063), and there is evidence for this occurring in atherosclerotic plaques (1064). Matrix metalloproteinases are converted from a proenzyme to the activated form through amino-terminal cleavage (1029), a process that is facilitated by HOCl (278, 1025). Indirect activation of matrix metalloproteinases is related to a class of proteins known as the tissue inhibitors of metalloproteinases (192). ROS such as HOCl have been shown to inhibit the ability of these proteins to function (832), thereby removing inhibition of matrix metalloproteinase activity. Consistent with the notion that the cytokine and oxidant milieu of the fibrous plaque is important in determining plaque integrity, there is an inverse relation between plaque macrophage content and its mechanical strength (539). Indeed, macrophage infiltration of the fibrous cap and its shoulder regions is a common finding in morphologically unstable lesions (635, 980).

Thus the elaboration of cytokines and production of ROS within the atherosclerotic plaque have important implications for its structural integrity. Unregulated oxidant production has the potential to promote the elaboration and activation of matrix degrading enzymes in the fibrous cap of the plaque. This activity is one link between oxidative stress and cardiovascular disease activity that is independent of LDL oxidation.


The major source of matrix production within the fibrous cap is the smooth muscle cell. It is possible, therefore, that depopulation of the fibrous cap may also contribute to plaque weakening and the propensity to rupture (239). For example, fibrous caps that have ruptured tend to have many more macrophages than stable plaques with many fewer smooth muscle cells (239). Under such circumstances, one might expect a relative paucity of matrix production and an abundance of matrix degradation due to this imbalance in the ratio of smooth muscle cells to macrophages. There are considerable data to suggest this situation may arise from specific cytokines produced within the plaque. For example, activated T lymphocytes are found within atherosclerotic plaques (359) and may represent 10–20% of the total cell population (454), particularly at sites prone to plaque rupture (980). Interferon-γ is a major product of lesional T cells (274) that is known to inhibit the proliferation of vascular smooth muscle cells (358) and sensitize these cells to apoptosis (53, 297). One could certainly envision a situation where inflammation in specific areas of the atherosclerotic plaque may promote smooth muscle cell apoptosis and confer a defect in the synthetic capacity of the atherosclerotic plaque. This scenario could render the plaque susceptible to rupture in response to hemodynamic forces in the vessel.

The local environment of ROS may also have implications for the population of smooth muscle cells in the fibrous cap. Although H2O2 stimulates the proliferation of smooth muscle cells (758), this response is highly dose dependent, as mentioned earlier. At low concentrations of H2O2, there is clearly a proliferative response (201). In fact, a number of smooth muscle cell mitogens actually require H2O2 production for a proliferative response (332, 923). However, higher concentrations of H2O2 are associated with smooth mucle cell apoptosis and necrosis (201), and the shoulder regions of atherosclerotic plaques appear to exhibit considerable levels of oxidant-generating enzymes and ROS (864). Similarly, atherosclerotic plaques exhibit evidence for HOCl generation and HOCl-mediated oxidation (368, 369), all the more important considering that HOCl induces growth arrest and apoptosis of vascular cells (997). Thus it is possible that the excess ambient oxidants contribute to the depopulation of the fibrous cap, thereby tilting the balance of matrix production and degradation in favor of the latter.

In summary, development of the atherosclerotic lesion involves the generation of a fibrous cap overlying a lipid core. Stable lesions are characterized by a fibrous cap that remains thick and replete with collagen-producing smooth muscle cells that lend structural stability to the plaque. Under these circumstances, shear forces and mechanical stresses are not met with any structural failure, and the lipid core remains isolated from the circulation. However, under certain circumstances of cytokines and oxidative stress, shoulder regions of the atherosclerotic plaque become replete with matrix-degrading activity, and the population of smooth muscle cells is diminished. As a consequence, the plaque may become relatively acellular with less interstitial collagen. This sets the stage for structural failure and plaque rupture that we now understand is a critical event in the precipitation of cardiovascular disease events. Evidence for the contribution of oxidative events in the precipitation of cardiovascular events comes from studies showing that in patients with coronary artery disease low red blood cell peroxidase 1 activity (66) or increased plasma levels of myeloperoxidase (91) are associated with increased risk of events.

C. Vasomotor Function

The precipitation of acute vascular events in atherosclerosis involves processes that go beyond plaque vulnerability and rupture. There is now a growing appreciation that local homeostatic processes in the arterial wall are also abnormal in those patients with frank atherosclerosis and risk factors for atherosclerosis. Among the more important components of vascular homeostasis is the endothelium, as it serves as the interface between the vascular wall and flowing blood. Through the release of autocrine and paracrine factors, the endothelium regulates a number of important processes such as vascular tone, platelet adhesion, and leukocyte transit into tissues and the vascular wall. The principal factors released by the endothelium that regulate vascular homeostasis on a moment-by-moment basis are prostacyclin, leukotrienes, and ·NO.

With regard to the latter, in 1980, Furchgott and Zawadzki (283) described an endothelium-derived factor responsible for arterial relaxation in response to acetylcholine that was later identified as ·NO (427). Endothelial production of ·NO is important in the regulation of vascular tone, arterial pressure, platelet adhesion, and leukocyte trafficking, as mice lacking eNOS exhibit spontaneous hypertension, defective vascular remodeling, enhanced vascular thrombosis, and leukocyte interactions (263, 417, 535, 788). The “classic” model of bioactivity of ·NO involves its binding to the heme group of guanylate cyclase in target cells (e.g., platelets, smooth muscle cells) to increase cellular cGMP and activate cGMP-dependent protein kinase, thereby affecting ·NO-mediated vasodilation and platelet inhibition (426). Under aerobic conditions, ·NO may also S-nitrosate protein cysteine thiols, and this mechanism has been implicated in ·NO-mediated modulation of ion channels, protein kinases, caspase enzymes, and transcription factors (328).

The endothelial isoform of NOS is a 135-kDa protein that consists of a carboxy-terminal reductase domain linked by a regulatory calmodulin-binding site to an amino-terminal oxygenase domain (Fig. 11, head-to-head apposition of eNOS). Normally, eNOS exists as a homodimer with the oxygenase domains of each protein linked together through a Zn-thiolate cluster (757). The spatial characteristics of eNOS are such that its catalytic action involves the flavin-mediated transport of electrons from NADPH in the carboxy-terminal reductase domain of one monomer to the amino-terminal heme of the other monomer where O2 is reduced and incorporated into the guanidino nitrogen of l-arginine to form l-citrulline and ·NO. The regulation of eNOS-mediated ·NO production is present at the transcriptional level as well as through posttranslational regulation that includes substrate and cofactor availability, enzyme acylation, and targeting to Golgi membrane and plasma membrane caveolae, protein-protein interactions, and phosphorylation (for reviews, see Refs. 282, 946).

FIG. 11.

Model of head-to-head apposition of nitric oxide synthases (NOS). Shown is the typical structure of a NOS dimer where the electron flow (arrows) traverses the reductase domain of one monomer to enter the oxygenase domain of the other. In the case of endothelial NOS, a Zn-thiolate cluster stabilizes the dimer (546, 757). [From Stuehr et al. (913).]

1. Endothelial dysfunction

Endothelial dysfunction is a poorly defined term that we submit refers to a loss of normal homeostatic functions (e.g., vasodilatation, platelet inhibition). This condition often occurs early in the course of atherosclerosis with one important manifestation being a reduction in the bioactivity of endothelium-derived ·NO. Although the loss of ·NO bioactivity is not the only manifestation of endothelial dysfunction, it is an independent predictor of future cardiovascular events in patients with atherosclerosis (314). There are many potential reasons for impaired ·NO bioactivity. These range from inadequate ·NO production to ·NO degradation or an inadequate response to ·NO and include eNOS uncoupling. There is evidence to support defects in all facets of ·NO production and metabolism in the setting of vascular disease, but oxidative events figure prominently in many studies of impaired ·NO bioactivity.

2. Oxidative events and endothelial dysfunction


In models of vascular disease such as cholesterol-fed rabbits, blood vessels produce substantial amounts of nitrogen oxides (i.e., oxidation products of ·NO) despite the impairment in ·NO-dependent vascular relaxation (628), suggesting that global ·NO production in vascular disease is not attenuated. Rather, ·NO appears to become inactivated before reaching its cellular target, and one mechanism of ·NO inactivation considered initially involved O2· (340, 429, 785). Mugge et al. (645) observed later that treatment of cholesterol-fed rabbits with SOD conjugated to polyethylene glycol increased vascular SOD activity and improved ·NO-mediated arterial relaxation (645). It is now known that hypertension, hypercholesterolemia, diabetes, and atherosclerosis are associated with an increase in the steady-state flux of O2· in the vascular wall and O2· reacts with ·NO at near diffusion-controlled rates (see sect. iiC). Because the resulting ONOO inefficiently activates the soluble isoform of guanylyl cyclase (936), its formation effectively decreases ·NO bioactivity in the vascular wall.

The bimolecular combination of O2· and ·NO provides only a partial explanation for reduced ·NO bioactvity in the setting of vascular disease, as evident in later stages of atherosclerosis. For example, in older (>24 mo) Watanabe heritable hyperlipidemic rabbits that lack a functional LDL receptor and develop chronic manifestations of atherosclerosis, ·NO bioactivity is impaired and is not restored by increasing endothelial SOD levels (623). This finding implies that other sources of O2· outside the endothelium must be considered or, alternatively, that other means of impaired ·NO bioactivity predominate in later stages of the disease, and we now direct our attention to this issue.


In the presence of metal ions or ·NO, a flux of O2· can promote the formation of LOO· (569) that have implications for ·NO bioactivity. Like O2·, LOO· can combine with ·NO to form adducts (689, 787) and therefore may “quench” bioactive ·NO (677). Lipid peroxidation within the vascular wall may also lead to the formation of oxidized LDL that directly inactivates ·NO (149) and may reduce eNOS protein in endothelial cells (554). Indirect effects of lipid peroxidation are also potentially important. The transfer of oxidized phospholipids from oxidized LDL to the endothelial cell plasma membrane stimulates protein kinase C (681) and impairs G protein-coupled signal transduction leading to abnormal ·NO-mediated arterial relaxation in response to receptor-dependent ·NO agonists (505).

It is important to realize that 1e-species such as O2· and LOO· do not represent the sole source of reactive species that have implications for ·NO bioactivity and endothelial cell function. Indeed, there is now a growing appreciation that 2e-oxidants also play a role in the modification of endothelial function.


Although evidence outlined above plainly indicates that ONOO formation effectively limits ·NO bioactivity by quenching ·NO, other properties of ONOO appear to limit endothelial function as well. Peroxynitrite readily oxidizes tetrahydrobiopterin, thereby limiting the activity of eNOS and facilitating O2· production (524, 627) (see sect. iiC3). Atherosclerosis (985) and diabetes (843) are associated with reduced vascular levels of tetrahydrobiopterin, and ONOO-mediated oxidation of tetrahydrobiopterin has been proposed as a physiologically relevant mechanism of impaired ·NO bioactivity (400). Observations that uric acid, a scavenger of ONOO, improves endothelial ·NO bioactivity in atherosclerotic mice is consistent with this notion (524).

Another plausible pathway of eNOS uncoupling involves RNS-mediated oxidation of the Zn-thiolate center, resulting in the conversion of active eNOS dimer to inactive eNOS monomers. Thus exposure of endothelial cells to low concentrations of ONOO has been reported to result in increasing uncoupling of eNOS activity, characterized by decreased ·NO production and increased O2· formation without apparent loss of tetrahydrobiopterin (1112). Such uncoupling, mediated by ONOO itself or ·NO2 derived from it (see sect. iiB2), likely involves the oxidation of one (or several) of the four cysteine residues coordinated to the Zn-atom present in the eNOS dimer (1112). As pointed out in section iiC3, ·NO reversibly inhibits eNOS activity via binding to the heme moiety of the protein. However, exposure of the isolated protein or endothelial cells to ·NO also causes monomerization of eNOS that is associated with loss of enzyme activity and release of zinc and that is prevented by thioredoxin plus thioredoxin reductase (763). These findings have been rationalized by the binding of ·NO to eNOS proceeding in two steps: a first, reversible one due to binding of ·NO to heme, followed by a second, irreversible step, caused by nitrosation of a critical thiol residue and resulting in the displacement of zinc and tetrahydrobiopterin, and loss of enzyme activity (763). Because ·NO itself is not an efficient nitrosating species, its autoxidation products, particularly the 2e-oxidant N2O3, need to be considered as the likely mediator of the proposed formation of nitrosothiols (see sect. iiC3). Such a mechanism may explain the apparent O2·-independent impairment of ·NO bioactivity in later stages of atherosclerosis. Thus, as lipids accumulate in the vessel wall, the local environment becomes increasingly hydrophobic. This will increase both the local concentration of molecular oxygen (that promotes autoxidation of ·NO) and the extent to which ·NO-derived N2O3 engages in nitrosation reactions (by decreasing N2O3 hydrolysis) (1045). Thus ONOO and N2O3 have multiple biologic activities that could lead to impaired ·NO bioactivity by limiting ·NO production.


As discussed in section ii, myeloperoxidase exists in atherosclerotic lesions and vascular disease, and HOCl is the major product at physiological concentrations of chloride ions (362). Hypochlorous acid can oxidize a large variety of biological molecules, particularly proteins and amino acids (362), with the latter activity being particularly germane to endothelial function. For example, l-arginine reacts with HOCl to form chlorinated derivatives that impair ·NO production from cultured endothelial cells and rat arterial segments (1102). Thus there is evidence to indicate that myeloperoxidase-derived products could contribute to endothelial dysfunction in the setting of atherosclerosis.

The link between myeloperoxidase and ·NO bioactivity is not limited to HOCl. In particular, endothelial cells are known to transcytose myeloperoxidase leading to enzyme deposition in the subendothelial space and the facilitation of extracellular matrix protein tyrosine nitration (27). This may have important implications for ·NO bioactivity as myeloperoxidase can catalytically consume ·NO (4) even when the enzyme is localized to the subendothelial space (225). These findings are consistent with human studies that circulating myeloperoxidase levels are an important determinant of endothelial ·NO bioactivity in the setting of coronary artery disease (999). Thus myeloperoxidase accumulation in the arterial wall represents another mechanism for impaired ·NO bioactivity through the catalytic consumption of ·NO.


Among 2e-oxidants produced in vivo, H2O2 is perhaps the most abundant. It freely diffuses through cell membranes and may travel several cell diameters before reacting with targets such as thiols and heme. Diseased vessels produce increased levels of H2O2 (333), and activated neutrophils at normal circulating concentrations can produce 200–400 μM H2O2 over a 60-min period (562). Emerging evidence now also supports the notion that H2O2 may be involved in modulating ·NO bioactivity. Although H2O2 does not react with ·NO, it does produce arterial relaxation in an endothelium- and eNOS-dependent manner (945, 1100). Hydrogen peroxide treatment also promotes chronic increases in eNOS activity by upregulating transcription and enhancing mRNA stability (211) via a mechanism that involves activation of Ca2+/calmodulin kinase II and janus kinase 2 signaling pathways (117). Therefore, H2O2 may promote both acute and chronic increases in eNOS activity that could serve as a compensatory response to oxidative stress.

Paradoxically, although treatment of endothelial cells with H2O2 promotes eNOS activity above basal levels, ambient levels of H2O2 can inhibit agonist-stimulated ·NO bioactivity (446, 586). For example, in cerebral arterioles, H2O2 impairs ·NO-mediated arterial relaxation in response to acetylcholine or authentic ·NO, an effect that appears to involve O2· as it is reversed by SOD (1024). Mice with a defect in H2O2 detoxification due to cellular glutathione peroxidase deficiency have impaired endothelium-dependent vasodilator function (253).

3. Antioxidants and endothelial function


Given the overwhelming evidence linking oxidative stress to impaired ·NO bioactivity and endothelial function, it is not surprising that a number of investigators have utilized antioxidants as a strategy to improve endothelial cell phenotype. Many studies have reported a consistent beneficial effect of acute and chronic ascorbate administration on the bioactivity of endothelium-derived ·NO in human subjects (for review, see Ref. 215). Both intra-arterial infusion and oral supplementation of ascorbate results in improved endothelial-dependent vasodilation in human patients with vascular pathological conditions including atherosclerosis, diabetes, hypertension, and cigarette smoking (388, 542). The effect of ascorbate is both acute and is maintained over 30 days of supplementation (313). This is likely due to ascorbate increasing endothelial concentrations of tetrahydrobiopterin (392, 415) rather than scavenging of O2· (444). Indeed, clinical studies demonstrate that tetrahydrobiopterin supplementation improves endothelial function in human subjects with vascular disease (387, 911). Thus these data indicate that both ascorbate and tetrahydrobiopterin are limiting determinants of endothelial ·NO production.

Glutathione is a major determinant of the intracellular redox environment and is present in millimolar concentrations. Studies have shown that patients treated with l-oxo-4-thiazolidine carboxylate, an agent that selectively increases intracellular glutathione concentrations, exhibit improved ·NO bioactivity in the brachial artery (1000). Likewise, intra-arterial glutathione infusion improves endothelium-dependent arterial relaxation in response to acetylcholine (506). The specific mechanisms responsible for these observations are not yet clear, as glutathione alone has limited activity in boosting endothelial ·NO production in culture (381, 415). However, it should be noted that thiols are excellent scavengers of 2e-oxidants such as H2O2 and HOCl, two species that have been implicated in endothelial dysfunction (see above). Taken together, studies with ascorbate, tetrahydrbiopterin, and glutathione indicate that intracellular redox environment is an important determinant of endothelial ·NO bioactivity.


Vitamin E has also garnered considerable interest as an antioxidant with the potential to improve endothelial function. Similar to ascorbate, α-TOH can scavenge O2· (k = 5 × 103 M−1 · s−1), but the reaction rate is unlikely to support effective competition with ·NO for O2· in vivo. Nevertheless, increasing vascular vitamin E levels have been shown to improve endothelial ·NO bioactivity in experimental animal models of vascular disease (470, 889), although this effect is not linked to the inhibition of lipoprotein lipid oxidation (469). Alternatively, increasing tissue vitamin E levels may act by inhibiting protein kinase C-dependent promotion of O2· production (471).

Despite the consistent favorable effect of vitamin E on endothelial ·NO bioactivity in experimental animals, the situation in humans is contradictory. Vitamin E supplementation of hypercholesterolemic or coronary disease patients improves ·NO bioactivity in some studies (390, 661) but not others (305, 610, 855). The discrepant findings in animals and humans are not well explained but may reflect differences in disease stages at the time the intervention is applied. Studies in experimental animals are typically early in the disease process, whereas human studies involve patients with established vascular disease. Thus although early atherosclerotic disease processes may be subject to modification by vitamin E, later stages may not.

Probucol has been shown to inhibit atherosclerosis in multiple animal models (468, 594). In keeping with this vascular protective action of probucol, its use in animal models of vascular disease has demonstrated an improvement in ·NO bioactivity in atherosclerosis (434, 473, 854) and diabetes (940). This effect appears independent of its lipid-lowering properties and may be related to a reduction in the vascular O2· flux in cholesterol-fed rabbits (473). Limited studies within human subjects have provided mixed results, with one study demonstrating an additive effect with lipid lowering and another demonstrating no effect in forearm vessels (19, 610).

Thus far, we have focused on endothelial vasomotor function as a prototypical vascular process that is subject to modulation by oxidative events. Some of this emphasis is driven by observations that endothelial vasomotor function is predictive of future cardiovascular events in patients (314, 315, 809). The link between oxidative events and endothelial function as it pertains to the process of atherosclerosis is further strengthened by observations that patients with oxidative stress-induced endothelial dysfunction are particularly prone to future cardiovascular events (389). Nevertheless, there are other events in the vascular wall that contribute to atherosclerosis that are also known to be sensitive to oxidative processes. We now turn our attention to this issue.

D. Adhesion Molecules

Despite the long-standing appreciation that inflammatory cells are involved in atherosclerosis (783), the specific mechanisms whereby vascular inflammation is initiated remained elusive for many years. In 1991, Cybulsky and Gimbrone (172) made a seminal observation that early rabbit atherosclerosis was associated with induction of a leukocyte adhesion molecule on the endothelium that was homologous to human vascular cell adhesion molecule-1. We now know the recruitment of inflammatory cells is dependent on cellular adhesion molecules. These are a diverse group of surface proteins that are divided, among others, into selectins, integrins, and immunoglobulin superfamily members (Table 15). Vascular inflammation begins with leukocyte rolling on the endothelium that is facilitated by endothelial expression of P-selectin and its interaction with leukocyte P-selectin ligand-1 (59). Firm leukocyte adhesion requires interaction between leukocyte β1- and β2-integrins and endothelial immunoglobulin superfamily members such as vascular cell adhesion molecule-1 and intercellular adhesion molecule-1, respectively (1111). Once firm adhesion is established, leukocytes may then transmigrate across the endothelium along a chemotactic gradient such as that produced by monocyte chemotactic protein-1 (171).

View this table:

Cellular adhesion molecules implicated in vascular injury

Cellular adhesion molecules are an important component in atherosclerosis and the response to vascular injury. Histological studies demonstrate increased endothelial expression of vascular cell adhesion molecule-1 and intercellular adhesion molecule-1 in developing and established atherosclerotic lesions (545, 727). Early atherosclerotic events and the initiation of lesion formation appear particularly dependent on vascular cell adhesion molecule-1 (173). Mice lacking intercellular adhesion molecule-1 (653), P-selectin (209, 452, 653), or β2-integrins (653) are protected against the full spectrum of atherosclerotic lesion development. Human studies have demonstrated that plasma levels of intercellular adhesion molecule-1 and E-selectin correlate with the clinical manifestations of coronary atherosclerosis (424, 769). Thus adhesion molecules modulate the biologic response to vascular injury, with atherosclerosis and plaque activation representing two prominent examples of vascular injury.

The activity of adhesion molecules is regulated via distinct mechanisms. The immunoglobulin superfamily members, vascular cell adhesion molecule-1 and intercellular adhesion molecule-1, are predominantly regulated at the transcriptional level. In contrast, P-selectin activity is dependent on rapid mobilization to the surface from preexisting pools. Integrins are activated by G protein agonists (e.g., ADP), growth factors (e.g., platelet-derived growth factor), and cytokines (e.g., tumor necrosis factor-α), through a process known as “inside-out” signaling. Integrin activation is thought to reflect conformational change(s) that increase ligand affinity of β1- (182), β2- (773), and β3-integrins (162). Cross-linking of cell surface integrins by ligand occupancy triggers “outside-in” signals that induce cytoskeletal rearrangement, cell spreading, and, in leukocytes, respiratory burst (1080). The signaling pathways mediating integrin activation are incompletely defined but may involve altered intracellular pH, increased intracellular Ca2+, and/or activation of protein kinase C, phospholipase A2, phosphoinositol 3-kinase, and protein phosphatases (393, 1004).

Oxidative events have been implicated in the regulation of endothelial cellular adhesion molecules. Cytokine-induced endothelial cell vascular cell adhesion molecule-1 expression is inhibited by the thiol N-acetylcysteine (592), the metal-chelator pyrrolidine dithiocarbamate (592), or a glutathione peroxidase mimic (181). Intracellular O2· generation appears critical to cytokine-induced upregulation of vascular cell adhesion molecule-1, as it is inhibited by cellular overexpression of SOD (145). Similarly, a O2· signal has been implicated in the upregulation of intercellular adhesion molecule-1 in endothelial cell by oscillatory shear stress (423). Cytokine-induced vascular cell adhesion molecule-1 expression is also enhanced by oxidized LDL or its oxidized fatty acids (480). Even in the absence of cytokines, human arterial endothelial cells exposed to oxidized LDL alone express intercellular adhesion molecule-1, vascular cell adhesion molecule-1, and E-selectin (14). Thus considerable data indicate that cellular oxidative events modulate the expression of adhesion molecules on the endothelium.

The role of ROS in regulating cellular adhesion is not limited to the endothelium. Adhesion of neutrophils is associated with a ROS burst that is required for the activation of Src-family kinases that are important for “outside-in” signaling (1079). In fact, there appears to be considerable cross-talk between inflammatory cells and the endothelium with regard to adhesion molecule regulation. For example, a respiratory burst from the leukocyte oxidase is necessary for tumor necrosis factor-α-induced upregulation of intercellular adhesion molecule-1 on endothelial cells in vivo (240). When taken together, the paragraphs outlined above provide secure evidence that endothelial-leukocyte adhesion involves oxidative events within both cells involved.


In the sections above, we have reviewed the data linking oxidative events to the pathogenesis of atherosclerosis. Without question, there is a wealth of data demonstrating that atherosclerosis and its resultant cardiovascular events are associated with a number of oxidative events ranging from LDL oxidation to the production of intracellular ROS and RNS. Despite abundant data, however, a fundamental problem with implicating oxidative stress as pathophysiologically important for atherosclerosis is the poor performance of antioxidant strategies in limiting either atherosclerosis or cardiovascular events from atherosclerosis. There remains today no consensus that antioxidant supplementation of patients at risk for atherosclerosis has any effect on the disease process (see sect. iiiF2). These observations have called into question the relation between oxidative events and atherosclerosis and provide a real challenge to designing strategies targeted at the oxidative stress associated with atherosclerosis.

There are several potential explanations as to why antioxidant trials might have proven ineffective in the treatment of atherosclerosis and, as a consequence, reconcile available data. First and foremost, one must consider the possibility that oxidative events are not causally related to the process of atherosclerosis and clinical cardiovascular disease. For many, this represents an unattractive explanation given the wealth of associative data linking oxidative stress to atherosclerosis. Second, we may not have the requisite fundamental understanding of the relevant oxidants involved in atherosclerosis and thus have not tested the appropriate antioxidant treatment for atherosclerosis. In this regard, the distinction between 1e- and 2e-oxidants is of particular importance since most antioxidant trials involved species that are solely effective against 1e-oxidation reactions. Finally, it is possible that oxidative events principally represent an injurious response to the process of atherosclerosis and are not causative. In this instance, any strategy aimed at minimizing oxidation reactions would be akin to treating the symptoms of a disease rather than the disease itself. In the following paragraphs, we briefly consider each of these potential explanations.

A. Oxidative Events and Atherosclerosis Are Not Causally Linked

Despite the fact that this interpretation of available data is, perhaps, the darkest, there are data to support this contention. Human (915) and animal atherosclerotic lesions (541, 1055) are characterized by the presence of oxidized lipids thought to result from LDL oxidation. The latter, of course, has been proposed to cause foam cell formation, a critical part of the initiation of atherosclerotic lesion formation. Nevertheless, atherosclerotic lesion formation can be dissociated from the occurrence of lipid peroxidation in the arterial wall (see sect. iiiF1b), suggesting lipid peroxidation is not necessary for foam cell formation.

If one discounts lipid peroxidation as necessary for foam cell formation, an alternative means of macrophage cholesterol loading must be found. In this regard, the response-to-retention hypothesis of early atherogenesis is of particular interest. A number of studies indicate that arterial wall retention of lipoproteins has implications for atherogenesis (see sect. iB2). One interpretation of these and other studies (238, 819, 820) is that lesion-prone areas are characterized by enhanced retention of apolipoprotein B-containing lipoproteins, and this process is the inciting event for atherosclerosis rather than LDL oxidation. Entrapment of LDL in the arterial wall has important sequelae. For example, proteoglycan-bound LDL may be more susceptible to oxidation (422), and once oxidized, this species could obviously promote atherosclerosis as discussed above for oxidized LDL. Retained LDL is a substrate for sphingomyelinase (1074) that may generate ceramides that are known to promote apoptosis and mitogenesis (357, 455). Most importantly, however, and as pointed out earlier, aggregated LDL is avidly internalized by macrophages and smooth muscle cells (440), thereby supporting foam cell formation without the need for LDL oxidation (991). Such at least initially LDL oxidation-independent formation of foam cells could explain why in human aortic lesions nonoxidized lipids accumulate before oxidized lipids as disease progresses (966).

Considering the totality of evidence, one can make a cogent argument that LDL aggregation is sufficient for the formation of foam cells and the generation of atherosclerotic lesions. It is entirely possible this process will also produce some LDL oxidation; however, this would not be causally linked to atherosclerotic lesion formation. More investigation will be needed to determine the relative roles of LDL oxidation versus LDL retention to the early aspects of atherosclerotic lesion formation.

B. Incomplete Knowledge of Oxidants Involved in Atherosclerosis

The initial link between oxidative stress and atherosclerosis was derived from studies of LDL oxidation and its prevention with antioxidants such as probucol, vitamin E, and butylated hydroxy toluene. Most of these antioxidant species were chosen because they had considerable activity to inhibit LDL oxidation in vitro in response to 1e-oxidants such as copper ions or radical-generating systems. As a consequence, almost all evidence on antioxidant treatment and atherosclerosis involves these “traditional” antioxidants that are only active against 1e-oxidant species. If, however, the putative oxidative stress of atherosclerosis were actually due to 2e-oxidants, many of the antioxidant species outlined above would be ineffective scavengers and thus expected to fail. Whether, and if so to what extent, this line of argument can be extended to ascorbate requires further investigation, as the vitamin offers near-complete protection against HOCl and chloramines, partial protection against ONOO, and its efficacy against N2O3, particularly within a lipid environment, is not known.

As we have learned in section iiiC, there is considerable evidence that 2e-oxidants are involved in the process of atherosclerosis. With regard to LDL, 2e-oxidants like HOCl oxidize LDL by preferentially targeting apolipoprotein B-100 rather than lipids (372). Myeloperoxidase, the major source of HOCl, is present and active in human lesions (184, 368, 581), and leukocyte myeloperoxidase levels are positively associated with coronary artery disease (91, 1103). Lesions and lesion-derived LDL in humans and WHHL rabbits contain HOCl-modified epitopes (368, 369, 581, 582) and chemical markers for active myeloperoxidase and HOCl (277, 377, 391). Thus abundant data implicate the 2e-oxidant HOCl in the pathology of atherogenesis, and this component of oxidative stress would not be altered by classic antioxidants such as vitamin E that are active only against radicals. As discussed in section iiiC1c, this conclusion is not mitigated substantially by the observation that mice deficient in myeloperoxidase have somewhat increased atherosclerosis (90).

In addition to HOCl, H2O2 has become an important oxidant in atherosclerosis, in particular as a component of cell signaling (see sect. iiF) that mediates many maladaptive responses in the vasculature. Growth factors and cell proliferation have long been implicated in atherosclerosis (781), and these cellular responses are dependent, in large part, on intracellular generation of H2O2 (145, 923). As discussed above, reactions involving H2O2 are not subject to inhibition by classic antioxidants such as vitamin E. As a consequence, features of atherosclerosis due to unchecked H2O2 production would not be altered by most antioxidant interventions in both human and animal studies that have been published thus far.

This latter point raises a common problem with the pursuit of antioxidant strategies as a means to limit oxidative stress and ameliorate oxidant-mediated pathology. In particular, such strategies presuppose a detailed knowledge of the relevant oxidants involved in a disease process and the specific consequences of scavenging such oxidants. This problem is illustrated in Figure 12. As can be seen, the production of ROS has a number of potentially important consequences that have been implicated in the pathophysiology of vascular disease. Introducing specific antioxidant species has the potential to scavenge or metabolize some, but not all, of the relevant oxidizing species implicated in atherosclerosis. For example, the introduction of lipid-soluble antioxidants will limit lipid peroxidation but will have no effect on 2e-oxidants and thus no effect on protein modification by ONOO, cell signaling by H2O2, or HOCl-mediated oxidation reactions. Conversely, species that scavenge H2O2 will have a profound effect on HOCl-mediated oxidation but may also wipe out H2O2-mediated signaling, which may, under certain circumstances, be adaptive rather than maladaptive. This latter point is illustrated nicely with regard to ·NO production and the clinical syndrome of sepsis. Although excess ·NO production has been linked to the vascular leakage and hypotension of sepsis (for review, see Ref. 833), the use of general NOS inhibitors has met with uniform failure to improve outcome in this disease (159). Thus, whenever a physiological process goes awry in the setting of disease (e.g., ROS signaling), strategies that rely simply on scavenging the offending species must be employed with extreme caution lest physiological responses may well be interrupted and worsen the clinical situation.

FIG. 12.

Diversity of potential oxidants involved in atherosclerosis and inhibitors required counteracting the biological consequences of the different oxidants. As stated in the text, there is a plethora of sources for reactive oxygen species (ROS) in the vasculature with a number of secondary reactions that have been described. Any one “scavenger” strategy that tries to intercept the potentially damaging species is not likely to have a material effect on other potentially harmful reactions. As a consequence, identifying the signals and sources that lead to abnormal production of ROS and/or reactive nitrogen species (RNS) could prove more fruitful in limiting oxidative damage. MPO, myeloperoxidase; Mn+, metal ions. [Adapted from Munzel and Keaney (649).]

How can one resolve the aforementioned conundrum? A good starting point would be to develop a more complete understanding of the relevant oxidant sources and of what controls their production. With this information in hand, one could devise strategies to ameliorate the unchecked production of oxidants while preserving physiological signaling by ROS and RNS. The advantages of this strategy are illustrated by the experiences with chronic myelogenous leukemia. For many years, this disease was treated with cytotoxic chemotherapeutic agents that produced brief palliation, yet the disease remained incurable. Subsequently, research determined that chronic myelogenous leukemia was characterized by chromosomal rearrangement (the so-called Philadelphia chromosome) resulting in a unique fusion gene known as BCR-ABL that encoded a dysregulated tyrosine kinase (845). With this knowledge, a specific inhibitor of this kinase was devised, and it proved to be extremely effective in treating the disease (210). Thus, by understanding the nature of dysregulation, specific therapeutic approaches can be developed that target the disease process rather than physiological signaling. Only through further understanding the mechanisms of dysregulated oxidant production can we hope to achieve a similar result with atherosclerosis.

C. The Oxidative Response to Inflammation Hypothesis of Atherosclerosis

With the wealth of data indicating that LDL oxidation is a prominent feature of atherosclerosis, there has been considerable interest in linking atherosclerosis risk factors to the process of LDL oxidation and oxidative stress in general. Indeed, many studies indicate that oxidative stress is a feature of many atherosclerosis risk factors such as diabetes (319), hypertension (333), and smoking (641). The notion that oxidative stress is causally related to atherosclerosis was thought to be so compelling that some investigators have even proposed that all atherosclerosis risk factors act through the production of excess oxidative stress (Fig. 13 A) (12). With regard to lipid peroxidation, this link with oxidative stress was recently tested in the Framingham Study, the study that gave rise to the concept of atherosclerosis risk factors. This investigation found that lipid peroxidation was linked to some, but not all, risk factors for atherosclerosis (474). In fact, statistical modeling with established cardiovascular risk factors was only able to explain ∼15% of the observed evidence for lipid peroxidation. Thus lipid peroxidation is only modestly linked to atherosclerosis risk factors.

FIG. 13.

Conceptual views of oxidative stress as a cause (A) or consequence (B) of atherogenesis. In scheme B, one can see that scavenging ROS/RNS and preventing markers of oxidative stress will not have any material impact on the course of atherosclerosis.

Another process integral to atherosclerosis is inflammation, and this concept is now firmly established (for review, see Refs. 683, 715). Some 25 years ago, Ross (781) observed that atherogenic diets produced inflammatory cell adhesion to the arterial wall in a matter of days. As we have seen, this recruitment of inflammatory cells is dependent on cellular adhesion molecules and requires an interaction between leukocyte β1- and β2-integrins and endothelial adhesion molecules such as vascular cell adhesion molecule-1 and intercellular adhesion molecule-1, respectively (1111). Many studies have established that circulating markers of inflammation are predictive of both atherosclerosis and the clinical events associated with atherosclerosis (424, 556, 577, 729, 769).

With this information in mind, one needs to consider whether antioxidant strategies have been disappointing in atherosclerosis largely because oxidative events may be a consequence, rather than a cause, of the atherosclerotic process (Fig. 13B). In this scenario, the inflammatory process produces oxidative events as a by-product, but these events are not strictly required for the progression of atherosclerosis. As a consequence, antioxidant strategies that lead to scavenging of oxidants and limiting markers of oxidative stress would not be predicted to have a major impact on the disease process, as the link between inflammation and atherosclerosis would not be altered. Thus the scheme depicted in Figure 13B provides an explanation for both the occurrence of oxidative events with atherosclerosis and the disappointing results that have been achieved with antioxidant supplementation trials.

It is reasonable to ask if there is evidence to support such a contention. To address this issue, let us contrast the relation between atherosclerosis risk factors and the processes of oxidative stress and inflammation. With regard to the former, there is only a loose relationship as outlined above (474). Of the established risk factors, only smoking and diabetes are predictive of increased urinary levels of F2-isoprostanes (474). In comparison, the relation between atherosclerosis risk factors and inflammation is quite firm. Epidemiological studies using circulating inflammatory markers such as soluble intercellular adhesion molecule-1 (777) or C-reactive protein (558) have found that all established cardiovascular disease risk factors (i.e., smoking, hypertension, diabetes, hypercholesterolemia, age, and obesity) are predictive of circulating inflammation markers. In addition, there is now evidence that hypercholesterolemia promotes atherosclerosis lesion development by activating macrophage innate immunity signaling pathways (63), thereby directly linking elevated serum lipids to a pro-inflammatory signaling cascade. Taken together, these findings tend to support the notion that inflammation is a primary process in atherosclerosis, and oxidative stress is secondary. Additional studies will be needed to substantiate this notion or otherwise. In particular, it will be important to distinguish atherosclerosis, inflammation, and oxidative stress to extend the measures of oxidative stress to lipid oxidation markers other than urinary F2-isoprostanes, and to include markers of protein oxidation. Concerning the latter, a recent case-control study showed plasma 3-nitrotyrosine levels to be higher in patients with established coronary artery disease than controls, and preliminary results suggested this to be independent of inflammation (844).

The notion that oxidative stress is an important secondary consequence of inflammation is well established (658). The inflammatory process is modulated by the activities of several families of enzymes including cyclooxgenases, lipoxygenases, NADPH oxidases, nitric oxide synthases, and peroxidases that all possess the catalytic capacity to produce ROS and RNS. The roles of these reactive species in host defense and antimicrobial activity are well documented as impaired ROS or RNS production results in susceptibility to bacterial (25) or parasitic infection (651). The specific role that reactive species play in the inflammation of vascular disease, however, is not yet clear. We would submit that ROS- and RNS-mediated oxidative events are most important for modulating the healing response to inflammation. As a consequence, we propose the “oxidative response to inflammation” hypothesis as a means to reconcile available evidence concerning oxidative events and vascular disease. According to our proposal (Fig. 14), the injurious response to cardiovascular risk factors is manifest as an inflammatory response in the vascular wall subsequent to lipoprotein retention and vascular injury. This response is characterized by the production of ROS and RNS, not as a means to mediate atherosclerosis, but as a by-product of the inflammatory process. In this scheme, the production of ROS is, however, involved in promoting tissue reorganization and the regenerative response to injury.

FIG. 14.

Conceptual view for the oxidative response to inflammation hypothesis of atherosclerosis. In this scheme, the presence of cardiovascular risk factors promotes the extent of arterial retention of lipoproteins that leads to the vascular endothelium adopting an “injury” phenotpe. This event stimulates inflammatory cell recruitment into the arterial wall and, as a consequence, vascular inflammation. The inflammatory process is central to lesion development but also leads to the production of ROS/RNS. The latter may amplify vascular inflammation, and they cause the oxidation of biomolecules (i.e., oxidative stress) that is important for promoting tissue remodeling and has a modulatory influence on the characteristic of atherosclerotic lesions, but does not directly cause atherosclerosis.

The model depicted in Figure 14 lends itself to several predictions about atherosclerosis. For example, this model would predict that modification of atherosclerotic risk factors should be manifest as a reduction in vascular inflammation. In large part, this prediction has already been fulfilled as lipid-lowering (7), smoking cessation (698), weight loss (1109) and improved glucose control (775) have all been associated with a reduction in circulating markers of inflammation. Another prediction from the oxidative response to inflammation hypothesis is that any state of heightened inflammation should be associated with a greater degree of atherosclerosis. Considerable data support such a contention as many conditions known to activate the immune system have now been linked to the activity of atherosclerosis. For example, periodontal disease (639) and prior infection (608) have been associated with more advanced or active atherosclerosis. The converse is also true, in that agents known to increase circulating markers of inflammation would be expected to exacerbate atherosclerotic vascular disease. With regard to this latter point, hormone replacement therapy increases circulating C-reactive protein levels (559), and clinical trials thus far have demonstrated its promotion of atherosclerotic cardiovascular events (329). Thus the oxidative response to inflammation model of atherosclerosis is consistent with available clinical and epidemiological evidence on the link between inflammation and atherosclerosis.

The oxidative response to inflammation model is also consistent with available data on tissue injury, and this has implications for vascular remodeling. For example, the migration and proliferation of smooth muscle cells is a common response to vascular injury (818). The smooth muscle cell response to growth factors is now known to be dependent on the intracellular generation of H2O2 (26, 923). This putative role for ROS in the response to tissue injury is consistent with a recent report in which Nox2-deficient mice demonstrate an impaired proliferative response to vascular injury (146a). Teleologically, one can rationalize the relation between ROS and the response to tissue injury. If high levels of ROS and RNS are common sequelae of host invasion and tissue injury, and one considers that the concentration of these reactive species falls as a cubed function of the distance from tissue injury, it makes sense that at low levels (at the fringes of injured tissue) ROS and RNS serve as an initiating signal for tissue remodeling. A similar scenario exists for ·NO whereby low levels are associated with normal vascular homeostasis and high levels of ·NO accompany overwhelming tissue invasion and sepsis. The specific means, however, by which ROS and RNS modulate tissue remodeling will require considerable further study.

D. Conclusions

There is a wealth of data linking oxidative events to the pathogenesis of atherosclerosis and its resultant cardiovascular events. These include unambiguous evidence for the occurrence of oxidative modifications in atherosclerotic lesions, proatherogenic activities of oxidized LDL, successful outcome of some antioxidant intervention studies in animal models of atherosclerosis, the increased production of intracellular ROS and RNS during disease, and the role of oxidative processes in vasomotor function and plaque disruption. Despite this, the causative relationship between oxidative events and atherosclerosis in general and the pathophysiological importance of LDL oxidation in particular have been challenged by the overall poor performance of antioxidant strategies in limiting atherosclerosis and its cardiovascular events, the overall lack of clear disease stage dependency in the vessel wall contents of oxidized molecules and antioxidants, and by the reported dissociation of atherosclerosis and lipoprotein oxidation in the vessel wall of animals. As discussed, the poor outcome of antioxidant trials may be explained by the lack of appreciation of the importance of oxidative events caused by 2e-oxidants combined with the inappropriate focus on classic antioxidants effective only against 1e-oxidation reactions. It will be important in this context to provide experimental evidence that scavengers of 2e-oxidants can offer protection against disease in suitable animal models of atherosclerosis. Perhaps just as important, the relation between atherosclerosis and cellular redox processes such as the role of ROS and RNS in cell signaling require experimental testing employing targeted strategies. Finally, the real challenge will be to distinguish oxidative events as a cause rather than an injurious response to the process of atherosclerosis. In this context, it will perhaps be particularly important to experimentally examine whether oxidative modifications in the vessel wall occur as a process secondary to inflammation, as is implied by the oxidative response to inflammation hypothesis presented here.


We thank Drs. Proctor and Mamo for their original drawing (736) of Figure 5.

The work in R. Stocker's laboratory is supported by National Health & Medical Research Council of Australia (NH&MRC) Programme Grant 222722 and by grants from the Heart Foundation of Australia. R. Stocker also acknowledges the support he receives as a recipient of a NH&MRC Senior Principal Research Fellowship. The work in J. F. Keaney's laboratory is supported by National Institutes of Health Grants DK-55656, HL-60886, HL-67206, and HL-68758. J. F. Keaney is an Established Investigator of the American Heart Association.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 4a.
  6. 5.
  7. 6.
  8. 7.
  9. 8.
  10. 9.
  11. 10.
  12. 11.
  13. 12.
  14. 13.
  15. 14.
  16. 15.
  17. 16.
  18. 17.
  19. 18.
  20. 19.
  21. 20.
  22. 21.
  23. 22.
  24. 23.
  25. 24.
  26. 25.
  27. 26.
  28. 27.
  29. 28.
  30. 29.
  31. 30.
  32. 31.
  33. 32.