|Home | About | Journals | Submit | Contact Us | Français|
For nearly 100 years, pediatricians have regularly employed oxygen to treat neonatal and childhood diseases. Over this time, it has become clear that oxygen is toxic and that overzealous use can lead to significant morbidity. As we have learned more about the appropriate clinical indications for oxygen therapy, studies at the bench have begun to elucidate the molecular mechanisms by which cells respond to hyperoxia. In this review, we discuss transcription factors whose activity is regulated by oxygen, including Nrf2, AP-1, p53, NF-κB, STAT and CEBP. Special attention is paid to the mechanisms by which hyperoxia affects these transcription factors in the lung. Finally, we identify downstream targets of these transcription factors, with a focus on heme oxygenase-1. A better understanding of how oxygen affects various signaling pathways could lead to interventions aimed at preventing hyperoxic injury.
Oxygen therapy has a long and tortuous history in Neonatology. The pendulum has swung from a liberal use of supplemental oxygen in the early 20th century, to limited application in the 1950’s based on the association with retinopathy of prematurity. Today, clinical studies are focused on addressing which neonatal pathologic states require treatment with oxygen, and what level of oxygen administration is safe. In concert with these clinical studies, much work has been done at the bench to ascertain how oxygen affects gene expression. This is of particular relevance in neonates because changes in gene expression at critical times in development can have long lasting effects and subsequent consequences on lung structure and function. This review will address lessons learned and new insights as to the effects of hyperoxia on pulmonary gene expression.
Responses to atmospheric oxygen have evolved in eukaryotes over the last 1.5 billion years (1). The ability of organisms to reduce oxygen to water critically altered cellular metabolism and energy production, but also resulted in the formation of toxic reactive oxygen species via the mitochondrial respiratory chain. These radicals are electron donors, which can damage DNA, RNA, protein and lipids. They can also propagate deleterious reactions throughout cells and tissues resulting in death and apoptosis. In addition, these ROS can alter gene expression by modulating transcription factor activation, which then impact downstream targets. In oxygen breathing animals, only three tissues — the cornea, the skin, and the respiratory tract epithelium - are exposed to 21% oxygen, equivalent to a partial pressure of about 160 mm Hg at sea level. The remaining tissues are exposed to much lower oxygen tensions. The affinity of hemoglobin for oxygen maintains the partial pressure of oxygen in the mitochondria below 0.5 mm Hg, limiting the production of reactive oxygen species and effectively protecting the body from oxygen toxicity (2). Before the advent of the medical use of oxygen, humans were rarely exposed to oxygen tensions that were greater than those in their ambient environment. Thus, it stands to reason that evolution may not have dictated a well-developed response to acute increases in oxygen tension. The notable exception is the transition at birth from the womb to the outside world where we were rapidly shifted from a relative hypoxic environment to relative hyperoxia. Additionally, the lung epithelium is constantly exposed to “relative hyperoxia” compared to other tissues and is further stressed by oxygen therapy.
From the time of its discovery in the 1770’s, oxygen has held promise as an elixir for multiple human ailments. Within 10 years of its discovery, Anton Lavoisier applied oxygen to newborn infants requiring resuscitation (3). By the early 1900’s, physicians were administering oxygen to treat cyanosis in premature infants (4). Shortly thereafter, oxygen therapy became widespread in neonatal units, with therapeutic indications ranging from respiratory distress to periodic breathing. However, by the early 1950s, published reports linking oxygen to the pathogenesis of retinopathy of prematurity began to appear, and the use of oxygen was quickly curtailed (5). Nevertheless, physicians were reminded that oxygen was a powerful and life-saving therapy when increased mortality from hyaline membrane disease (6) and the resurgence of cerebral palsy (7) were observed. This demonstrated that both too much and too little oxygen were problematic. Vigorous debates about the appropriate use of oxygen during newborn resuscitation (8) and the proper pulse oximetry saturation goals for premature infants (9) currently rage on. At this time, six multi-center randomized controlled trials are attempting to define optimal oxygen therapy goals for preterm babies (9).
Studies at the bench pair nicely with these clinical trials. Investigators have used multiple in vivo and in vitro models to determine how oxygen affects gene expression and subsequent lung structure and function. Hyperoxia results in alveolar and endothelial cell destruction, fluid leak into the air space, respiratory failure and mortality (10). The lungs of animals exposed to hyperoxia show increased mean linear intercepts, influx of macrophages, extracellular matrix turnover and fibrin deposition (11). During hyperoxia, reactive oxygen species (ROS) are produced both by the electron transport chain in the mitochondria and by the membrane-bound NADPH oxidase (12-15). ROS cause DNA strand breaks and other chromosomal aberrations (16, 17), which stimulate the expression of genes involved in inhibiting cell cycle progression (18). There is clear evidence in animal models that exposure to hyperoxia results in lung morphology similar to that of bronchopulmonary dysplasia (11, 19). These studies serve as important correlates to the ongoing trials involving oxygen therapy for premature infants.
Organs, tissues and cells have evolved systems to rapidly respond to changes in their microenvironment. A stimulus, which causes a perturbation, must be detected and translated into a response, which then facilitates a return to the steady state (Figure 1). Receptors, signaling pathways, transcription factors and downstream changes in proteins and metabolic function have evolved for this purpose. Only a few transcription factors that specifically alter gene expression in response to increased oxygen tension have been identified, as well as some direct down-stream targets (see Table 1). These will be discussed below.
The detoxification of ROS and electrophiles is important to prevent cellular injury. The transcription factor nuclear factor, erythroid 2 related factor 2 (Nrf2) regulates the inducible expression of a group of detoxification enzymes, such as glutathione S-transferase and NAD(P)H:quinone oxidoreductase, via antioxidant response elements (ARE). Under normal circumstances, Nrf2 is retained in the cytoplasm by a repressor protein Kelch-like ECH-associated protein 1 (Keap1). Exposure to xenobiotics and oxidants leads to the dissociation of Nrf2 from Keap1, which allows the free Nrf2 to translocate to the nucleus where it heterodimerizes with c-Jun, an activator protein-1 (AP-1) family protein (20). The consensus binding sequence of Nrf2 shows high similarity to the ARE/electrophile-responsive element (EpRE) sequence previously identified (21-23). Nrf2 can also heterodimerize with small Maf proteins to regulate ARE-mediated gene expression (24). These Maf proteins are so named because of their structural similarity to the founding member, the oncoprotein v-Maf. They include a characteristic basic region linked to a leucine zipper (b-Zip) domain which mediate DNA binding and subunit dimerization respectively (25).
Lung Nrf2 responds to hyperoxia (26). Linkage analysis identified Nrf2 as an important mediator of protection against lung hyperoxic injury (27) and mice deficient in Nrf2 exhibit aggravated lung injury and a lack of upregulation of ARE-mediated phase 2 detoxifying and antioxidant enzymes (28). Further gene array analysis of wild type versus Nrf2 deficient mice revealed discordance in multiple genes, thus identifying potential downstream targets of this important transcription factor (29). In fact, a single nucleotide polymorphism found in the Nrf2 promoter increases the risk of acute lung injury in human subjects (30). This evidence provides an important translational correlate and may lead to the development of therapeutic strategies.
Activator protein-1 was first identified as a transcriptional factor that binds to an essential cis-element of the human metallothionein II gene (31). It is composed of fos and jun protein dimers which bind via hydrophobic interactions of their leucine-zipper regions (32). The jun/jun and jun/fos dimers form the AP-1 complex. This transcription factor controls genes involved in cellular proliferation and death in response to various stimuli including hyperoxia. The consensus AP-1-binding site is embedded in the ARE where fos and jun proteins may heterodimerize to Nrf2 in the presence of electrophiles and oxidants as discussed above (33). Blocking AP-1 activation enhances hyperoxia-induced cell death in murine lung epithelial cells (34, 35). One specific target of hyperoxia-induced JNK1/AP-1 activation in A549 cells is the IL-8 promoter (36). This could modulate inflammatory responses with hyperoxic exposure. It is interesting to note that neonatal mice exposed to hyperoxia show no increase in lung AP-1 consensus sequence binding (37) in contrast to their adult counterparts (37, 38). However, in the brain, increased AP-1 consensus sequence binding occurs in the forebrain and hippocampus of both adult and younger rats exposed to hyperoxia (39, 40). These data suggest both maturational differences and tissue specificity of AP-1 activation.
The transcription factor p53 regulates the expression of a large number of target genes including those related to cell cycle arrest, cell death and DNA repair (41). Since its discovery in 1979, p53 has been identified as a tumor suppressor and its role in human cancer has become clearer (42). Under basal conditions, p53 resides in the cytoplasm and is subjected to ubiquitin-mediated proteolysis. However, in response to stimuli such as DNA damage, p53 is phosphorylated, stabilized and enters the nucleus (41). Under conditions of cellular stress, activated p53 initiates growth arrest and induces proapoptotic gene expression (42). Hyperoxia increases p53 gene transcription, protein levels and activity (43-46). In preterm baboons, exposure to hyperoxia results in increased p53 protein levels in airway epithelium (47, 48). However, in p53-/- mice exposed to hyperoxia, lung injury and lethality did not differ from similarly exposed wild-type animals (44, 49). These data indicate that the exact role of p53 in modulating the cellular response to hyperoxia remains to be elucidated.
The nuclear factor kappa B (NF-κB) family is composed of highly conserved dimeric proteins, which activate genes that regulate apoptosis, inflammation and oxidative stress (50-52). This factor regulates gene expression and was first described by Baltimore and Sen (53). In quiescent cells, NF-κB dimers remain sequestered in the cytoplasm bound to a member of the IκB family of inhibitory proteins (51). IκBα is the prototypical member of this family and the most well studied. With inflammatory or oxidant stress, IκBα is phosphorylated, resulting in dissociation and unmasking of the nuclear localization sequence of NF-κB (52). Following inflammatory stimuli, such as TNF-α activation, IκBα is phosphorylated on serine 32/36 and degraded through the proteosomal pathway (52). In addition to this canonical pathway, an atypical pathway of NF-κB activation results from specific phosphorylation of IκBα on tyrosine 42 (54). This occurs after stimulation with pervanadate, nerve growth factor, hydrogen peroxide and ischemia-reperfusion (54-56) and, as most recently demonstrated, with hyperoxia (57). This latter pathway represents an intriguing molecular target for modulating the pulmonary response to hyperoxia.
Hyperoxia-induced NF-κB activation appears to be stimulus and cell type specific. It is important to note that NF-κB nuclear translocation and DNA binding can either enhance or suppress target gene expression. The subunit composition of the NF-κB dimer likely confers specificity to the expression of target genes following activation (58). The most abundant NF-κB protein is the p65-p50 dimer (59). The p65 subunit contains a transactivation domain that interacts with other transcription proteins to increase gene expression (60). The p50 subunit lacks this transactivation domain, and can repress transcription when bound to DNA as a p50-p50 homodimer (60, 61). Furthermore, the ability of NF-κB to alter gene expression is affected by post-translational modifications including phosphorylation and acetylation (60).
Nuclear translocation of NF-κB was shown in A549 lung adenocarcinoma cells exposed to hyperoxia-induced but this activation did not protect against cell death (62). Also, in adult mice exposed to hyperoxia, NF-κB activated pro-inflammatory markers in pulmonary lymphocytes (63). Furthermore, in fetal mouse lung explants, hyperoxia-induced NF-κB activation was associated with increased apoptosis which was reversed by blocking NF-κB activation (64). In contrast, inhibition of hyperoxia-induced NF-κB activation accelerated nonapoptotic cell death in primary and transformed lung epithelial cells, resulting in decreased levels of MnSOD (65). Additionally, A549 cells pre-treated with hyperoxia showed less apoptosis following exposure to hydrogen peroxide, an effect reversed by inhibiting NF-κB activation (66). In other examples, NF-κB was not activated in response to hyperoxic exposure (67, 68), suggesting that this signaling pathway is cell specific. The lung contains over forty different cell types (69), and the response to hyperoxia is cell type specific. Endothelial cells are very sensitive to oxygen toxicity, while type II epithelial cells are resistant and proliferate in the recovery phase (70). Furthermore, in the developing lung, exposure to hyperoxia prevents the normal differentiation of type II cells to type I cells in the developing lung (71). Further studies are necessary to fully dissect the specificity and complexity of hyperoxia-induced NF-κB activation. Nevertheless, these findings suggest that interventions to either inhibit or enhance NF-κB activation in hyperoxia could be of therapeutic benefit.
Various clinical interventions can affect NF-κB activation. Adrenalectomized adult mice exposed to hyperoxia had less lung injury and had improved survival due to increased NF-κB activation (72). Glucocorticoids are known to inhibit NF-κB activation (73-76). Thus, hyperoxia-induced NF-κB activation, when not limited by endogenous glucocorticoids, protects the adult lung from oxygen toxicity (72). Interestingly, following glucocorticoid therapy for BPD, cells obtained from tracheobronchial lavage fluid of premature neonates showed inhibition of NF-κB activation (77). Nitric oxide, which may prevent BPD in some infants(78), also inhibits NF-κB activity (79). The clinical implications of these findings remain to be explored in humans.
Of particular interest to pediatricians are the maturational differences found in NF-κB activation. Multiple models have shown increased NF-κB activation in neonates compared to adults following exposure to inflammatory and oxidant stimuli (80-82). In rat fetal alveolar type II cells, NF-κB translocates to the nucleus and binds DNA after hyperoxic exposure (83). This binding peaks soon after birth and gradually decreases postnatally, suggesting that NF-κB regulates genes involved in the transition from the relative hypoxic environment seen in utero (83). This activation may have important downstream effects as shown in hyperoxia exposed fetal lung fibroblasts where NF-κB activation prevented apoptosis through the suppression of pro-apoptotic genes (57). In contrast, this hyperoxic activation of NF-κB was not seen in adult lung fibroblasts (57). In the only published study evaluating hyperoxia-induced NF-κB activation in a neonatal in vivo model, Yang and colleagues showed that hyperoxia-induced NF-κB occurred in the lungs of neonatal but not adult mice (82). This activation was associated with the relative tolerance to hyperoxic injury in the neonatal animals when compared with adults, and this tolerance was reversed when hyperoxia-induced NF-κB activation was inhibited (82). In contrast, clinical studies show that enhanced NF-κB activation is linked to respiratory distress syndrome and an increased risk of developing BPD in preterm infants (84-86). Thus, it is not yet clear whether inhibition of lung NF-κB is beneficial or harmful in human neonates.
The hyperoxic activation of NF-κB has also been investigated in tissues other than the lung. Using a bioluminescent NF-κB reporter mouse line, Dohlen and colleagues showed increased NF-κB activity in the brain after resuscitation with 100% O2 (87). In other studies, hyperoxia without preceding ischemia decreased NF-κB activation in the basal forebrain, with a more pronounced effect in aged versus young mice (88).
It is clear that the NF-κB mediated response to oxygen is influenced by maturation. Whether these changes are beneficial or detrimental remains to be seen. Understanding the maturational differences in hyperoxia-induced NF-κB activation could help guide interventions aimed to modulate this response in neonates.
Another important transcription factor involved in hyperoxic gene regulation is the signal transducers and activators of transcription protein (STAT). This family of proteins is activated by various cell-surface receptors in response to ligands, including cytokines, growth factors, and peptides (89). Hyperoxic lung injury is attenuated in mice constitutively expressing Stat3 in respiratory epithelial cells (90). Conversely, mice with disruption of Stat3 in respiratory epithelial cells demonstrate exaggerated hyperoxic lung injury and increased expression of pro-inflammatory cytokines including IL-6 (91).
The ccat/enhancer binding protein (C/EBP) family of proteins are basic leucine zipper (bZIP) transcription factors that respond to extracellular signals to regulate cell proliferation, differentiation and tissue development (92). C/EBPβand C/EBPconsensus sequence binding was increased in the lungs of young and aged mice exposed to hyperoxia (38). In the mouse exposed to hyperoxia, there is down-regulation of the protective Clara cell secretory protein (CCSP) due to enhanced C/EBPβnuclear translocation and binding to the CCSP promoter (93). These studies are particularly relevant because is required for lung maturation (94).
Acute and chronic exposure to hyperoxia may result in activation of a variety of other transcription factors including cmyc, fos related antigen (Fra)-1, junB, c-fos as well as nerve growth factor (NGF)1-A and —B (95). Furthermore, in the neonatal lung, hyperoxia can cause down-regulation of sox-7 and sox-18 (95). The relevance of these signaling events is not fully clear.
Because transcription factors that are regulated in hyperoxia control a multitude of genes, it would be difficult to list all of these genes. For example, the activation of NF-κB can regulate the expression of over 100 genes. Nevertheless, only a small fraction of NF-κB responsive genes are activated in hyperoxia. Some of the genes regulated by Nrf-2 and NF-κB will be highlighted below.
Nrf2 binds to the ARE, driving the expression of genes including antioxidants such as glutathione peroxidase, catalase, superoxide dismutase, thiol metabolism-associated detoxifying enzymes such as glutathione-s-transferase and stress-response genes such as heme oxygenase (HO-1), amongst others (25-28). These genes are all highly responsive to hyperoxia. We will focus on HO-1 as an example of an Nrf-2 regulated gene regulated in hyperoxia.
The HO-1 gene encodes for the rate-limiting enzyme in the degradation of heme and the formation of biliverdin, which is subsequently reduced to bilirubin by biliverdin reductase. In recent years, many roles have been identified for this protein and it has been clearly demonstrated that HO-1 is a generalized response to oxidative stress (96). The mouse HO-1 gene is 6.8 kb in length and organized into 4 introns and 5 exons. A promoter sequence is located 28 base pairs (bps) upstream of the transcription initiation site. There is a proximal enhancer (PE) directly upstream of the promoter and there are two more distal enhancers located at 4 kb (DE1) and 10 kb (DE2) upstream of the transcription initiation site. Each enhancer region contains multiple transcription factor binding sites including composite AP-1 and NF-E2 or CREB/ATF sites (see Figure 4) (97-99). Induction of HO-1 in oxidative stress is via Nrf2 and small Maf proteins binding to the ARE (100). Competitive binding between Nrf2 and BTB and CNC homology 1, basic leucine zipper transcription factor 1 (Bach1), at the ARE is important in heme-mediated regulation of HO-1 (101). Several investigators have documented hyperoxic induction of HO-1 in adult mice. However, in the neonatal rodent HO-1 induction is limited. In the neonatal mouse and rat, hyperoxic exposure did not result in a significant increase in HO-1 mRNA as it did in similarly exposed adult (102, 103). In another study, lung HO-1 mRNA only increased after ten days of hyperoxic exposure in neonatal mice (95) whereas this occurred within 24 hours in adult mice (104). There may be some teleological wisdom in not further elevating the levels of HO-1 when they are already quite high at birth and in the neonatal period, especially if this could lead to deleteriously high levels thus aggravating hyperoxic injury (105). We have also observed increased protein levels and DNA binding for Bach1, an inhibitor of HO-1 transcriptional activation, in neonates at baseline and after exposure to hyperoxia compared to adults (103). Typically, Bach1 is degraded in the presence of ROS (106). Enhanced Bach1 expression could ensure that there are sufficient levels for HO-1 gene inhibition in the neonate.
The IGF-binding protein (IGFBP)2 promoter has NF-κB consensus sequence binding sites, and both NF-κB consensus sequence binding and IGFBP2-promoter reporter activity increase in response to hyperoxia (107). This binding protein inhibits DNA synthesis and cellular entry into the S-phase, indicating a role for hyperoxia-induced NF-κB activation in modulating oxygen toxicity in the lung. Methylprednisolone treatment inhibits hyperoxia-induced NF-κB activation and downregulates ICAM-1 expression in human pulmonary artery endothelial cells (108), resulting in less neutrophil adhesion to the endothelium. As discussed above, adrenalectomized mice show attenuation of hyperoxic lung injury, and this is associated with preservation of NF-κB activation and induction of IL-6 (72). This cytokine is under the exclusive regulation of NF-κB with inflammation (109, 110). Whether IL-6 is exclusively regulated by NF-κB in response to hyperoxia is not known. Nevertheless, IL-6 is enhanced in the lungs of neonatal and adult mice in response to hyperoxia (111), although this phenomenon is not consistently observed in adult mice (63, 104). The amiloride-sensitive sodium channel, ENac, responsible for sodium and fluid absorption from the alveolar space (112), has an NF-κB binding site (113), and both NF-κB activation and ENaC gene expression increase with relative hyperoxia (114, 115). Furthermore, hyperoxia-induced ENaC expression is prevented with NF-κB blockade (115) in some reports but not others (116, 117).
Another important effect of hyperoxia is the modulation of genes involved in cell cycle regulation. Both acute and chronic exposure to hyperoxia results in upregulation of p21 (95). Of note, NF-κB is known to regulate the expression of p21 in some cells (118). This key inhibitor of cell cycle regulation and cellular proliferation is increased in both the neonatal (119) and adult (120) lung following exposure to hyperoxia. Expression of this protein in response to hyperoxia relies on either TGF-β signaling (121) or p53 activation (122, 123). Upregulation of p21 is protective against hyperoxic injury in both neonatal (124) and adult (125) mice. It is hypothesized that inhibition of cellular proliferation during periods of oxidative stress allows for additional time to repair damaged DNA (126) thus providing cytoprotection.
Hyperoxia regulates multiple transcription factors in the lung. These in turn regulate a variety of downstream targets including ARE regulated genes such as HO-1, antioxidant enzymes which are important in the detoxification of electrophiles, as well as genes involved in cell cycle regulation and the inflammatory response. The overall effect of hyperoxia in the lung depends on the maturational stage of the organism. The net effect of hyperoxic lung gene regulation may be both enhanced cytoprotection and worsened lung function. In the neonate where postnatal lung development is crucial to proper alveolar formation, hyperoxic gene regulation may have long-lasting effect on lung structure and function. A further understanding of how hyperoxia affects specific signaling pathways and subsequent gene expression could lead to interventions aimed at preventing hyperoxic injury.
Financial Support: This work was supported by the National Institutes of Health (RO-1 HL-58752 P.A.D.) and Pediatric Scientist Development Program (K12-HD00850, C.J.W.).