When oxidative stress occurs, cells function to counteract the resulting oxidant effects and to restore the redox balance. All organisms have adaptive responses to oxidative stress, with antioxidant defense enzymes being induced by changes in the levels of H₂O₂ or O₂•^-, leading to the activation or silencing of genes encoding defensive enzymes, transcription factors and structural proteins [5]. ROS have also been proposed to function as second messengers independent of oxidative stress and to signal such cellular fates as cell proliferation, necrosis, and apoptosis. Although various observations have led to the suggestion that cells have the means to sense ROS and to induce specific responses [1], the underlying mechanisms are still not fully understood. Regulators of oxidative stress responses are currently best characterized in bacteria [6], but progress is now also being made towards understanding such regulators in higher eukaryotes [5,7].

The transcriptional network that responds to ROS in eukaryotes is only now being unraveled but the prokaryotic system is quite well understood. Seventeen years ago, it was shown by two-dimensional gel electrophoresis that the expression of approximately 30 proteins was induced in bacteria by H ₂O₂. Of these 30 proteins, 12 were maximally induced within 10 minutes and the other 18 between 10 and 30 minutes [8]. Subsequent work led to the discovery of the OxyR regulatory protein, which was shown to regulate the expression of 9 of the 12 rapidly induced proteins. The tetrameric OxyR protein is a member of the LysR family of transcription activators and exists in two forms, reduced and oxidized; the latter is the only form able to activate transcription. Further studies led to the identification of a number of OxyR-activated genes [9]. Similarly, the superoxide response or SoxRS regulatory proteins were found to regulate expression of O₂•^--responsive proteins in bacteria [10]. Regulation of the soxRS regulon occurs by conversion of SoxR protein to an active form, which enhances soxS transcription. The enhanced levels of SoxS, in turn, activate expression of the regulon [10]. In addition to SoxR and OxyR, several other transcriptional regulators modulate the expression of antioxidant genes in bacteria, indicative of the complexity and connectivity of overlapping regulatory networks [6].

There are no apparent homologs of OxyR, SoxR, or SoxS in eukaryotes, but a number of other transcription factors have been found to play a role in regulating eukaryotic antioxidant genes. In yeast, transcriptional regulators of antioxidant genes include ACE1, MAC1, YAP1, YAP2, HAP1, and HAP2/3/4 [11]. In higher eukaryotes, oxidative stress responses are more complex and are modulated by several different regulators. In mammalian systems, nuclear factor κB (NFκB) and activator protein-1 (AP-1) are involved in regulating the oxidative stress response. The antioxidant responsive element (ARE) is present in the promoter region of genes encoding mammalian glutathione S-transferase (GST), metallothionein-I and manganese superoxide dismutase (MnSOD), and is responsible for the induction of these genes in response to oxidants [12]. The promoters of antioxidant genes in higher plants also contain the ARE and the NFκB and AP-1 binding sites [1]. The role of these factors is not unique to activation of antioxidant genes, however: NFκB, in particular, is known to play central roles in regulating cellular responses to stresses other than oxidants, as well as regulating normal growth and metabolism.

There is substantial evidence to suggest that a variety of biotic and abiotic stresses induce H₂O₂, which serves as a common factor in regulating various signaling pathways [1,13]. Similar stresses also activate mitogen-activated protein (MAP) kinases, with kinetics that either precede or parallel H₂O₂ production, suggesting that MAP kinases may be among the many converging points in the defense-signaling network [14]. In addition, exogenous application of several plant hormones and toxins has been shown to induce synthesis of O₂•^- and H₂O₂, leading to differential induction of some antioxidant genes and isozymes [1,7,12,15]. Thus, from a utilitarian viewpoint, the identification of all genes and proteins regulated by H₂O₂ is an important step toward treatments that might confer tolerance to multiple, but interrelated, stresses. In addition to induction or repression of antioxidant defense genes, ROS are known to similarly affect expression of a variety of other genes involved in different signaling pathways in microbes [6], yeast [16], plants [17], and animals [2].

Efforts to identify ROS-responsive genes on a global scale were limited until the advent of microarray-based gene-expression analysis [18]: DNA microarrays are now being used to comprehensively examine gene expression networks during oxidative stress. Reports on the stress responses of Escherichia coli [19], yeast [16,20], and higher plants [21] have now provided significant progress in surveying gene expression in response to H₂O₂.

The transcriptional profile of E. coli cells exposed to H₂O₂ was examined with a DNA microarray composed of 4,169 E. coli open reading frames [19]. Gene expression was measured in isogenic wild-type and oxyR-deletion mutants (ΔoxyR) to confirm that the H₂O₂-response regulator OxyR activates most of the H₂O₂-inducible genes. There was a very rapid and strong induction of a set of OxyR-regulated genes in the wild-type but not in the ΔoxyR strain, providing an internal validation of the experiment and confirmation of the induction of the oxidative stress genes identified 17 years ago by other means [8]. In addition, several new H₂O₂-inducible genes were identified: some are members of the OxyR regulon and some are induced by an OxyR-independent mechanism. These findings indicate that other H₂O₂ sensors and regulators are present in E. coli [19]. Several genes that are known to be repressed by OxyR were found to be significantly expressed in the ΔoxyR mutant. Overall, the mRNA of 140 genes in the wild-type and 167 genes in the ΔoxyR strain were significantly induced after H₂O₂ treatment. It was also found that the superoxide response transcription factor gene soxS was induced by H₂O₂, indicating an overlap with other regulatory pathways. Also highly induced by H₂O₂ in both wild-type and ΔoxyR cells were two genes known to be members of the SoxRS regulon, Fpr and sodA (encoding NADH-ferredoxin oxidoreductase and manganese-superoxide dismutase, respectively). The microarray data revealed an overlap between the oxidative stress and heat-shock and 'SOS' DNA-damage responses [19]. Thus, the results from E. coli microarrays clearly indicate that the activities of transcription factors in addition to OxyR and SoxRS are likely to be modulated by oxidative stress.

In addition to bacteria, the transcriptional profile in response to oxidative stress has also been characterized in eukaryotes. In one broad-ranging study, the expression profile was examined in Saccharomyces cerevisiae cells exposed to H₂O₂, in addition to other stresses, and the global set of genes induced or repressed by each environmental signal was identified [16,20]. The results indicated that about two-thirds of the genome is involved in the response to environmental changes and the response to oxidative stress involves about one-third of the yeast genome, with the maximal effects on gene expression occurring slightly later relative to other stresses examined during similar time courses. Most of the transcriptome returns to pre-stress levels within 2 hours of exposure to H₂O₂ [20]. Genes that are repressed for approximately 1 hour after exposure to H₂O₂ are only transiently repressed in other stress time courses. Thus, genes encoding the translation apparatus and its regulators are remarkably coordinated in the responses to each environmental change, although the dynamics of each response are different. The expression programs following H₂O₂ or O₂•^- treatment were essentially identical, despite the fact that different ROS are involved. There was strong induction of genes known to be involved in detoxification of both H₂O₂ and O₂•^-, such as catalase, superoxide dismutase, and glutathione peroxidase, as well as genes involved in oxidative and reductive reactions (for example, thioredoxin, glutathione reductase, and glutaredoxin). The genes most strongly induced in response to H₂O₂ and O₂•^- were dependent on the transcription factor Yap1 for their induction. Genes that are moderately induced by ROS and other signals are regulated by different transcription factors, depending on the conditions, and their response may be governed by different upstream signaling pathways [20].

Recently, it has also been demonstrated that H₂O₂ activates the Sty1 (stress-activated MAP kinase) pathway in Schizosaccharomyces pombe in a dose-dependent manner, via two sensing mechanisms [22]. At low H₂O₂ levels, Sty1 is regulated by a two-component signaling pathway that feeds into either of the two - Wak1 or Win1 - stress-activated MAP kinase kinase kinases upstream of Sty1. In contrast, at high H₂O₂ levels, Sty1 activation is controlled mainly by an independent two-component mechanism, requiring the function of both Wak1 and Win1. In addition, the individual bZip transcription factors Pap1 and Atf1 were found to function within a limited range of H₂O₂ concentrations: Pap1 activates target genes at low H₂O₂ concentration, whereas Atf1 controls transcriptional responses to high H₂O₂, with some minor overlap. Some apparent cross-talk among Sty1, Atf1, and Pap1 has been detected [23]. Thus, S. pombe deploys a combination of stress-responsive regulatory proteins to gauge and trigger the appropriate transcriptional response to increasing H₂O₂ concentrations [22]. This yeast mounts two separate responses to oxidative stress: an adaptive response to low-level H₂O₂ exposure that protects it from subsequent exposures to higher H₂O₂ levels, and an acute response that allows the cell to survive a sudden, potentially lethal dose of H₂O₂.

The oxidative-stress response has also recently been characterized in an organism more complex than yeast and bacteria. A recent large-scale cDNA microarray analysis of the Arabidopsis transcriptome during oxidative stress identified 175 non-redundant expressed sequence tags (ESTs) from a sample of 11,000 that are regulated by H₂O₂. Of these, 62 are repressed and 113 are induced; and RNA blots showed that some of the H₂O₂-regulated genes are also modulated by other signals known to involve oxidative stress [21]. Furthermore, a substantial number of these genes have predicted functions in defense responses, cell signaling, transcription, and cell rescue (from environmental insults and developmental arrest), underscoring the pleiotropic effects of H₂O₂ in the response of plants to stress. Overall, the microarray used was estimated to represent only about 30% of the Arabidopsis genome, depending on redundancy, and 1% to 2% of the genes represented in the array are affected by H₂O₂-imposed oxidative stress [21], which is comparable to the situation in yeast [20]. Of the 175 genes identified as H₂O₂-responsive, most have no obvious direct role in oxidative stress but may be linked to oxidative stress indirectly, as a consequence of other biotic and abiotic stresses, explaining their sensitivity to H₂O₂. Among the genes induced by H₂O₂ were genes encoding transcription factors, suggesting that they may mediate downstream H₂O₂ responses consistent with genomic studies in other species. Also, expression of the MAP kinases in Arabidopsis is induced by oxidative stress, as in other organisms, which in turn can mediate the induction of oxidative stress-responsive genes [24].

During the past ten years, the traditional view of ROS as mere indiscriminate reactive byproducts of cellular metabolism has undergone a metamorphosis. This change is primarily as a result of the discovery that ROS, and particularly H ₂O₂, may act as signal-transducing molecules, and that activation of intracellular transcription factors such as OxyR, SoxRS, NFκB, and AP-1 occur via interaction with ROS, leading to gene transcription (Figure 3). More recently, genome sequencing and expression profiling using DNA or oligonucleotide microarrays, and related technologies, have been used effectively in the study of global gene-expression patterns in response to different growth and environmental conditions to which organisms are exposed. Subsequent hierarchical clustering methods allow for the allocation of genes, co-regulated temporally or in response to a given signal, into specific expression groups, or regulons [19,25]. The numbers of genes that can be detected by these methods in response to H₂O₂ or any given environmental or developmental signal far exceed the limited number that could have been detected only a few years ago.

Given that transcription of genes into mRNA is governed by transcription factors, which bind to cis-regulatory regions of the DNA in the vicinity of their target genes, the question arises as to whether large co-regulated groups of genes share cis-regulatory elements that bind to common transcription factors. The data available with respect to oxidative stress seem to suggest this is indeed the case. Cis-acting elements within the promoters of ROS-activated genes are being defined as well as their cognate trans-acting factors [1]. A comparative analysis of promoter sequences of genes with similar expression profiles should provide a basis for unraveling common regulatory sequences and overlapping gene-expression networks modulating ROS-responsive genes. The antioxidant enzymes catalase and superoxide dismutase play key roles in modulating the levels of endogenous H₂O₂ and O₂•^-, which at specific concentrations act, in turn, to modulate the expression of other ROS-responsive genes.

Despite some limitations, mentioned in the papers cited [16, 19,20,21], it has become clear through the use of microarrays that there are far more genes responding to ROS than previously thought. We currently have a large and growing number of sequenced genomes and emerging technologies presenting us with enormous opportunities to advance biological science and our knowledge of how genomes perceive signals to respond to variable environments. Knowing the sequences of tens of thousands of ROS-responsive genes, however, only reminds us that we still do not know the many proteins they encode, nor the biochemical or biological function of the great majority of such proteins. How such proteins interact with ROS to drive the various physiological processes in aerobic organisms remains a great puzzle. For the future, the fundamental challenge will be to integrate the information now being obtained on gene-expression patterns with structural and functional parameters and interactions of the various proteins encoded by ROS-responsive regulons, and to view the cell in which they function holistically. The future indeed seems exciting for studies of oxidative stress responses.