Increasing evidence links early exposure to oxidative stress with potentially lifelong consequences (1). However, the role of reactive oxygen species (ROS) in biologic systems is entirely dependent on context: location, neighbors, and timing (2). ROS are oxygen ions [singlet oxygen, superoxide (O2·−)] or oxygen-containing radicals [hydroxyl, OH·−]. ROS and their reaction products [e.g. hydrogen peroxide (H2O2)] are increasingly recognized as signaling intermediates in their own right that can contribute to adaptive or maladaptive molecular responses (3). This review will focus on—1) origins of ROS: environment, cells, and cellular components; 2) molecular targets: classic and novel macromolecular targets and associated toxicity in infants and children; 3) antioxidant defenses: developmental regulation and vulnerabilities; and 4) antioxidant therapies: enzymatic and nonenzymatic approaches.

Origins of ROS

Oxygen has a unique molecular structure and is abundant within cells. It readily accepts free electrons generated by normal oxidative metabolism within the cell, producing ROS, such as O2·− and hydroxyl radical (HO·), as well as the oxidant H2O2. Processes causing uncoupling of electron transport can enhance the production of ROS, with mitochondria being a major source (4). However, other cellular components, such as endoplasmic reticulum-bound enzymes, cytoplasmic enzyme systems, and the surface of the plasma membrane, also contribute (5,6). Activity of multiple enzyme systems, such as the cytochrome P450 monoxygenase system, xanthine oxidoreductase, nitric oxide synthases, and several others involved in the inflammatory process (cyclooxygenase and lipoxygenase), can also increase the generation of ROS. Cellular production of O2·− and H2O2 can facilitate the formation of the more toxic and reactive HO· in the presence of reduced transition metals such as iron. Importantly, O2·− reacts rapidly with nitric oxide to form peroxynitrite (ONOO), a strong nitrating and oxidizing compound (7). Such highly reactive species, such as HO· or ONOO, can react with membrane lipids to cause more complex radicals by initiating lipid peroxidation.

In addition to ROS generated as a “byproduct” of cellular respiration, endogenous production of O2·− also arises from NADPH oxidases (NOX 1–3; typically at low levels in smooth muscle and vascular endothelium), dual oxidases 1 and 2, and NOX 4 (epithelial cells) (8). ROS are also important in the regulation of nitric oxide bioavailability, dramatically influencing airway and vascular reactivity (2). The burden of ROS can be further amplified by the presence of “free” metals, such as iron, copper, and manganese, which can be released from metalloprotein complexes. Although free iron (unbound to ferritin or heme, for example) has been documented in the circulating plasma of preterm newborns, the detrimental effects of free iron or other metals have not been definitively established in newborns (9,10). In contrast, indirect evidence has linked the presence of free iron with increased protein carbonyl formation in patients treated with high concentrations of supplemental oxygen. Failure to adequately sequester or store iron could be a developmental liability in premature infants with relative deficiencies in iron carriers, such as transferrin (11).

Molecular Targets: How ROS Damage Cells and Organs

A delicate balance exists between ROS production and the antioxidant defenses that protect cells in vivo. This balance may become disturbed under conditions of hyperoxia, inflammation, or ischemia-reperfusion (excessive generation of ROS) or in the presence of limited or impaired antioxidant defenses. Multiple pathways involved in ROS-induced cell death have been proposed. ROS can cause direct injury to proteins, lipids, and nucleic acids, leading to cell death. Some of these pathways are illustrated in Fig. 1. For example, protein oxidation and nitrosylation (carbonyl, nitration, and nitrotyrosine formation) can impair a wide variety of enzymatic processes and growth factors that can result in marked cellular dysfunction (12). Lipid peroxidation has been linked to cell death through effects on cellular phospholipids (major cell membrane components) through activation of sphingomyelinase and release of ceramide, which activates apoptosis (13). Nucleic acid oxidation has been linked with physiologic and premature aging as well as DNA strand breaks, leading to necrosis and/or maladaptive apoptosis (14). The magnitude of these changes and the cell's ability to repair this damage determines whether the effects are adaptive or maladaptive.

Figure 1
figure 1

ROS generation and detoxification in alveolar epithelium. Molecular oxygen first contacts the alveolus through the layer of surfactant phospholipids contained in the epithelial lining fluid, which is rich in glutathione and can, scavenge ROS. Under oxidative stress, ROS may be generated at the epithelial layer by DUOX and NOX4 that generate H2O2. SOD3 is poised to detoxify extracellular O2·−, although extracellular expression may be relatively deficient in newborns. Intracellularly, SOD2 in mitochondria detoxify O2·− generated during normal cellular respiration. The intracellular H2O2 burden is detoxified by peroxisome-bound catalase. Alveolar epithelium and other tissues may enhance generation of ROS in endothelial cells via NOX2,4, which can terminate NO-mediated reactions.

ROS at the proper locations and concentrations can also function as “2nd messengers” and activate multiple signal transduction pathways within the cell, facilitating the actions of growth factors, cytokines, and calcium signaling. ROS can activate c-Jun N-terminal kinase (possibly through production of lipid peroxide intermediates), a crucial mitogen-activated protein kinase, which then phosphorylates and releases two Bcl-2–related proteins that are normally sequestered within the cell (15,16). The release of these key proteins can directly activate Bax by causing dissociation from its cytoplasmic anchor. Bax is then free to translocate to the mitochondria, where it undergoes oligomerization and initiates the release of cytochrome c and other pro-death mediators into the cytosol.

Relatively high levels of O2·− are generated by NOX in phagocytes, such as neutrophils and macrophages (1000-fold higher than nonphagocytic cells), an essential process in bacterial killing. Blocking neutrophil influx in hyperoxia-exposed newborn rats mitigates oxidative DNA damage, HO· formation, and O2·− accumulation, while enhancing alveolar development (14,1719). In experimental lung injury models, genetic ablation of NOX routinely reduces pulmonary ROS accumulation, but is not necessarily protective, because accompanying inflammation is worse in the NOX null mice (20,21). Loss of nonphagocytic NOX function may also impair physiologic signaling (22). The relevance of ROS generated by NOX depends on the organ system, with inhibition preventing radiation-induced oxidative stress in rat brain microvascular endothelium (23). In general, adult models of oxidative stress to the CNS show decreased injury with inhibition or genetic ablation of NOX activity (24). In contrast, newborn mice exposed to hypoxia-ischemia sustain worse injury with inhibition of NOX (e.g. NOX2 null mice), implying that endogenous O2·− signaling may have critical adaptive roles in different organ systems (25).

The brain and the lung have been most intensively studied as target organ systems prone to damage by ROS. In premature and full-term newborns, the control of cerebral perfusion is less tightly regulated, increasing vulnerability to reperfusion-type injury and oxidative stress. For example, microglial activation is believed to cause accumulation of markers of oxidation (e.g. nitrotyrosine and protein carbonyls) in oligodendrocytes, leading to the development of periventricular leukomalacia (26). Activation may also have secondary effects through neuronal excitotoxicity via effects on calcium flux. Most experimental studies have focused on ischemia-reperfusion models, with pretreatment with antioxidants or free radical scavengers typically reducing apoptosis (e.g. less DNA fragmentation and caspase expression) and ameliorating histologic evidence of brain injury (27,28). Recently, studies of hyperoxia in newborn rats have also implicated ROS in causing neuronal cell death. In vitro exposure to high-oxygen atmospheres induces apoptosis in oligodendroglial cells in a developmentally dependent pattern, which is prevented by inhibition of lipoxygenase, with decreased expression of myelin basic protein in vivo in hyperoxia-exposed rat pups (29).

The developing retina is particularly prone to ROS-mediated damage that contributes to retinopathy of prematurity in preterm infants (30). Vascular growth into the developing posterior retina is normally driven by redox-sensitive pathways that up-regulate VEGF. After birth, the marked increase in systemic oxygen tensions in the preterm newborn suppresses VEGF production. This occurs in conjunction with impairment in autoregulation of retinal blood flow as well as a relative deficiency of antioxidants in the immature retina (31). In the absence of VEGF (and other factors), angiogenic budding stops, and apoptosis of developing vessels occurs secondary to the formation of reactive oxygen and nitrogen species (3234). Endogenous generation of ROS through NOX may be critical to this pathway, because its pharmacologic inhibition prevents retinopathy of prematurity in a newborn rat model (35). The second phase of retinopathy of prematurity occurs after birth, when the avascular retina continues to grow, overreaching its blood supply. This results in local tissue hypoxia, increased VEGF release, and an abnormal neovascular response. Once again, this process involves the formation of ROS and may be amenable to reductions in exposure to oxygen as well as treatment with antioxidants (36,37).

Similar mechanisms may be at play in ROS-induced damage to the immature postnatal, developing pulmonary system, where both epithelial and endothelial cells may be damaged. Pulmonary epithelial DNA oxidation (14), accumulation of HO· (19), lipid peroxidation (38), and protein oxidation (39) in whole lung have all been demonstrated in experimental models of bronchopulmonary dysplasia (BPD). In human BPD, studies do strongly support a role for ROS-mediated damage. Plasma 3-nitrotyrosine, a footprint of ONOO formation, and protein carbonyls, a marker of protein oxidation, are elevated in premature newborns at highest risk of developing BPD (40). ROS may inactivate antioxidant enzymes, with oxidized or nitrated proteins critical to lung function having been identified (41). The weight of the evidence implicates ROS in the development of impaired lung development in BPD (9,42).

Because exposure to more moderate oxygen concentrations are associated with modern-day BPD, the disruption to mesenchymal-epithelial-endothelial signaling, rather than acute cell necrosis or apoptosis induced by ROS, may be more critical. The “new” BPD is characterized by milder exposure to oxidative stress and mechanical injury, but at an earlier stage of pulmonary development, which ultimately causes alveolar hypoplasia (43). The inactivation of NO signaling, which is required for normal alveolar development, is one likely candidate pathway, either through direct inactivation (ONOO formation) or through indirect effects on endogenous NO production (44,45). In some (46,47), but not all, (48,49) experimental animal models of BPD, treatment with inhaled NO significantly protects alveolar development. Clinical trials of inhaled NO to prevent BPD have also had mixed results (5053), possibly due to differences in patient selection and treatment strategies as well as inactivation of NO by ROS. Because cGMP is a major target of NO action, type V phosphodiesterase inhibition would be expected to also enhance alveolar development, which indeed has been shown in a newborn rat model of BPD (54). Further support for this concept is suggested by strategies aimed at interfering with O2–mediated inactivation of NO using exogenous recombinant human CuZnSOD (rhSOD). In newborn lambs with pulmonary hypertension, treatment with inhaled NO in conjunction with rhSOD results in enhanced NO signaling, significant improvements in oxygenation, and marked reductions in oxidation in the lung (55). Clinical trials of inhaled NO in conjunction with rhSOD administration for the treatment of pulmonary hypertension in newborn infants are currently being planned.

Antioxidant Defenses: Developmental Regulation and Vulnerabilities

The vulnerability of target molecules, cellular compartments, and organ systems to ROS-mediated pathways depends on the local redox milieu, which, in turn, depends on developmental regulation of antioxidants. Several foci of ROS generation are spatially adjacent to opposing antioxidant enzymatic systems within subcellular compartments, as illustrated in Fig. 2. Tight temporo-spatial control of antioxidant expression has been linked to normal control of apoptosis in physiologic development. For example, opposition to oxidative signaling by use of antioxidant mimetics has been shown to prevent the physiologic apoptosis required for normal vertebrate limb development (56).

Figure 2
figure 2

ROS, macromolecular damage, altered signaling. ROS damage DNA through strand breaks and base oxidation that, if unrepaired, induces apoptosis or oncosis. Protein oxidation and nitration damage antioxidant enzymes, surfactant proteins, and anti-inflammatory pathways that can further propagate maladaptive inflammation. Lipid peroxidation products generate pro-inflammatory prostanoids, and can generate further radical formation through lipid chain reactions, possibly releasing damaging enzymes packaged in cellular organelles. Direct effects of ROS on signaling pathways include redox-sensitive transcription factors—e.g. HIF, Nrf-2, and NF-κB—as well as indirect effects through inactivation of NO-based signaling.

Gestational dependence of enzymatic antioxidants has been recognized for decades, as has the developmental regulation of oxygen-dependent signaling (previously reviewed in this series by Maltepe and Saugstad) (57,58). The cellular, subcellular, and tissue-specific expression of antioxidant enzymes, such as superoxide dismutases (SOD), catalase, glutathione peroxidases, and peroxiredoxin largely determine the relative vulnerabilities of tissues and cells to ROS-mediated injury (59,60). However, oxidative stress may actually regulate antioxidant capacity, with newborn rats demonstrating up-regulation of glutathione peroxidase catalase, and CuZn (cytosolic) SOD expression and activity in response to hyperoxia (61).

Because enzymatic antioxidants are gestationally regulated, premature newborns would be expected to have decreased expression relative to full-term newborns, and this has been demonstrated in most animal models (58,59,62,63). Susceptibility of the premature infant to ROS-mediated damage also depends on the expression and activity of many of these antioxidant enzymes, with heme oxygenase-1 and thioredoxin mRNA expression levels increased immediately after birth in both preterm and term newborns (64,65). Hypoxia may generate ROS that directly or indirectly stimulate hypoxia-inducible factors important in lung or brain development (66,67). The contribution of these induced responses may vary depending on the nature and magnitude of the oxidative stress. Whether this gestation-dependent expression/induction of antioxidants is adequate and actually involved in human diseases is not completely clear. Some antioxidant levels—glutathione, ascorbate, and urate—have been analyzed in tracheal aspirates and found to be poorly predictive of the risk of developing BPD (68). However, other antioxidants (e.g. SOD) may be more important in view of what has been found in animal model systems as well as clinical specimens obtained from newborn infants (69).

Although nonenzymatic antioxidants are also depleted in conditions characterized by ROS-mediated stress, the interpretation of these measurements is quite complex (70). For example, “low” vitamin E levels in premature newborns were interpreted as representing a deficient state; whether or not this is the case actually depends on the specific target cell or organ and on the actual ROS milieu (71). In contrast, glutathione may be deficient in premature infants because of excessive oxidization by ROS coupled with reduced glutathione reductase reaction with the electron acceptor NADPH (72). Finally, melatonin acts as an antioxidant in the retina and brain (73), and its cyclic production is disrupted in premature infants (74), possibly increasing the risk of ROS-mediated damage.

Antioxidant Therapies: Enzymatic and Nonenzymatic Approaches

Although the use of supplemental antioxidants represents a logical strategy to prevent or ameliorate lung injury from excess generation of ROS, caution must be exercised because ROS are critically important second messengers in various cell signaling pathways that control normal cellular functions. In addition, intracellular generation of ROS is important in bacterial killing by alveolar macrophages and neutrophils, and antioxidants may interfere with this process and contribute to worsening tissue injury.

Multiple cell culture models have suggested that overexpression of antioxidants prevents ROS-induced injury. Ilizarov et al. (75) generated stable cell lines overexpressing MnSOD and/or catalase (1.5- to 2-fold increase in activity) and then exposed them to 95% O2 for 10 d. Significantly, more cells overexpressing MnSOD were viable (40%) compared with cells overexpressing catalase alone or control cells (10%). Overexpression of catalase with MnSOD had a small additional benefit, suggesting that scavenging O2·− is the important rate-limiting step. Overexpression of either MnSOD or CuZnSOD also reversed the growth inhibitory effects of hyperoxia, with optimal protection from hyperoxic injury occurring with 1.5- to 3-fold increases in activity (76). Prevention of mitochondrial oxidation seemed to be a critical factor, because markers of mitochondrial function and cell survival correlated directly with the extent of mitochondrial localization of antioxidant activity and not overall activity within the cell (77). Overexpression of SOD not only reduced ROS production, but also mitigated the activation of the JNK/AP-1 pathway (78). Activation of this MAPK signal transduction pathway has been implicated in the pathogenesis of ROS-induced mitochondrial injury and apoptotic cell death. Finally, bacterial infection and associated inflammation have been shown to significantly increase ROS production. Exposure of lung epithelial cells (both airway and alveolar), monocytes, and macrophages to hyperoxia for as little as 24 h is associated with significant increases in bacterial adherence and IL-8 production as well as impaired phagocytosis and bacterial clearance, with overexpression of SOD having significant beneficial effects (7981). Because nosocomial infection is a predictor of BPD (82), antioxidant therapy could also be protective through this mechanism.

Other data demonstrating the efficacy of SOD in preventing hyperoxia-induced lung injury come from studies of genetically engineered mice. Transgenic mice lacking MnSOD die within the first 10 d of life in room air, whereas mice lacking CuZn or EC-SOD have reduced survival and more lung injury in response to ROS, but a normal lifespan (8385). In contrast, transgenic mice overexpressing MnSOD in alveolar type II cells are able to survive longer with significantly less lung injury in hyperoxia compared with wild-type controls (86). In addition, newborn EC-SOD transgenic (SP-C promoter driven) mice exposed to hyperoxia showed significantly less pulmonary neutrophil influx and reduced glutathione, with preservation of alveolar development compared with wild-type littermates (60). Transgenic mice had significantly less pulmonary neutrophil influx and oxidized glutathione at 7 d, preservation of alveolar surface and volume density, and preserved differentiation of type I alveolar epithelium, compared with wild-type littermates. Taken together, these data indicate that SOD is critically important in preventing hyperoxia-induced lung injury and preserving normal alveolar architecture.

The antioxidant vitamins, ascorbic acid (vitamin C) and α-tocopherol (vitamin E), are known to inhibit ROS-induced lipid peroxidation. Berger et al. (87) studied the administration of high-dose antioxidant vitamins in a premature baboon model of BPD compared with standard antioxidant vitamin supplementation. Although higher doses significantly raised vitamin C and E concentrations in plasma and the lung, no protective effects could be demonstrated. These studies question whether raising antioxidant vitamin concentrations alone will be effective in preventing ROS-induced injury in high-risk preterm infants.

Supplementation of vitamins A, C, and E has also been studied in preterm infants in an attempt to prevent with ROS-induced injury. Concentrations of vitamin A (i.e. retinol) may be deficient in very low birth weight infants, presumably from increased absorption of parenteral vitamin A into the i.v. tubing or from higher nutritional requirements (88). A multicenter trial of high-dose vitamin A supplementation in premature infants found a small (7%), but statistically significant reduction in the incidence of BPD (89). However, follow-up of treated infants did not demonstrate any long-term benefits of vitamin A in reducing chronic respiratory morbidity (90). Vitamin C has both oxidant and antioxidant activities and is thought to contribute to the regeneration of membrane-bound α-tocopherol (91). Although preterm infants may be relatively deficient in vitamin E, randomized controlled trials have consistently failed to demonstrate a significant benefit of α-tocopherol in preterm infants (92,93). Of concern, is that pharmacologic concentrations of vitamin E were associated with an increased risk of sepsis and necrotizing enterocolitis, precluding the routine use of these doses in high-risk preterm infants (94).

n-Acetylcysteine (NAC) is a source of the essential amino acid l-cysteine and a precursor of the antioxidant glutathione. A multicenter, double-blind trial of NAC was conducted in 391 ventilated, extremely low birth weight infants. Infants were randomized by 36 h of age to receive 16–32 mg/kg/d of NAC or placebo i.v. for 6 d (95). The study showed no reduction in survival or the incidence or severity of BPD at 36 wk corrected age or improved pulmonary function when the infants were studied at term (96).

A multicenter, randomized trial of prophylactic rhSOD has been performed to determine whether intratracheal treatment significantly reduced the incidence of BPD and improved pulmonary outcome at 1-y corrected age (97). Three hundred two intubated, premature infants (600–1200 g at birth) received either intratracheal rhSOD (5 mg/kg) or placebo every 48 h (as long as intubation was required) up to 1 mo of age. Although there were no differences in the incidence of death or BPD, 37% of placebo-treated infants had repeated episodes of wheezing or other respiratory illness severe enough to require treatment with asthma medications (e.g. bronchodilators and corticosteroids) compared with 24% of rhSOD treated infants at 1-y corrected age. In the highest risk infants, <27 wk gestation, 42% treated with placebo received asthma medications compared with 19% of rhSOD-treated infants. rhSOD was also associated with a 55% decrease in emergency department visits and a 44% decrease in subsequent hospitalizations. This study demonstrates that rhSOD may reduce ROS-induced pulmonary injury, although this may not be readily apparent when only evaluating early outcomes based on current BPD definitions. Future studies using long-term outcome variables may be needed to more definitively determine whether treatments to scavenge ROS are effective.


Oxidative stress, particularly, in the preterm newborn, arises in multiple organ systems and subcellular compartments. This occurs due to inadequate detoxifying mechanisms such as inducible antioxidant enzymes, glutathione stores, and nutritional antioxidants. Oxidative molecular damage to DNA can arrest appropriately timed proliferation and differentiation and damage to lipids in cell membranes, and key regulatory enzymes can provoke maladaptive inflammatory responses that can amplify the initial injuries. More subtle effects on ROS-mediated signaling and depletion of NO available for endogenous proangiogenic signaling can further contribute to disrupted organ development, including excitotoxic neuronal damage. Although these aspects have suggested the rationale for antioxidant therapy, its uses in the prevention of BPD, ROP, or brain injury in preterm newborns has not yet yielded unequivocal success. Further studies aimed at superior targeting to improve the therapeutic index of antioxidants will be necessary.