A look back at the early literature on reactive oxygen species (ROS) gives the impression that these small inorganic molecules had a singular defined role, that of host defence in mammalian systems. However, it is now known that their roles also include a major part in cell signalling, in a broad range of organisms from mammals to plants. Similarly, a look back at papers on the proteins now thought to be involved in the perception of hydrogen peroxide (H2O2) will show that they too had defined functions assigned to them, completely independent to H2O2 signalling. These proteins have disparate roles, in ethylene perception or even major metabolic pathways such as glycolysis. However, the chemistry of H2O2 sensing dictates that the proteins have a commonality, with active thiol groups being potential ROS targets. The challenge now is to determine the full range of proteins which may partake in the role of H2O2 perception, and to determine the mechanisms by which the signal is transmitted to the next players in the signal transduction pathways.
Despite the fact that, for many years, ROS have been considered to be damaging molecules used primarily for host defence, particularly in mammalian systems, it is recognized that ROS, especially hydrogen pexoxide, have a major role in cellular signalling pathways, across a wide range of organisms, including plants (reviewed by Dröge, 2002; Neill et al., 2002; Foyer and Noctor, 2005b). ROS can be generated in plants via the leakage of electrons from mitochondria or photosynthesis, or can be synthesized by a variety of dedicated enzymes, such as peroxidases and NADPH oxidases (Neill et al., 2002). In fact, the production of H2O2 by chloroplasts was reported over a quarter of a century ago (Kaiser, 1979; Charles and Halliwell, 1980), although it is only now that the full impact of such activity is coming to light. Once produced, ROS can partake in signalling, although these events will be modulated by the complement of antioxidants in, or even around, the cell. Some of the key events in plants which are controlled, perhaps in part, by ROS include stomatal closure (Pei et al., 2000), root growth (Foreman et al., 2003), and programmed cell death (Desikan et al., 1998), including the hypersensitive response.
To establish that ROS do in fact have a role in signal transduction in plants, the impact of the addition of ROS to cells on the activity of signalling proteins has been investigated. The overall effect of ROS on the intracellular redox state in a cell may mean that a myriad of proteins can be affected, and this has been the subject of recent reviews (Buchanan and Balmer, 2005; Foyer and Noctor, 2005b). The activity of kinases has been a particular focus. Exogenously added H2O2 has been shown to lead to the activation of mitogen-activated protein kinases (MAP kinases; Desikan et al., 1999), in particular the Arabidopsis AtMPK3 and AtMPK6 (Kovtun et al., 2000; Desikan et al., 2001b). Another kinase found to be involved in H2O2 signalling is one which has been suitably named oxidative signal-inducible 1 (OXI1), and interestingly this protein has been shown to be up-stream of the MAP kinases AtMPK3 and AtMPK6 (Rentel et al., 2004). In addition, the MAPKKKs ANP1 and OMTK1 have also been shown to be activated by H2O2 (Kovtun et al., 2000; Nakagami et al., 2004). Thus, several types of protein kinases have been shown to be activated in the presence of H2O2. However, whether these kinases act in isolation or via the formation of macromolecular complexes is not known.
In addition to protien kinases, tyrosine phosphatases have been shown to be inactivated by hydrogen peroxide in mammals (Cho et al., 2004), and phosphatases may be involved in H2O2 signalling in plants too. The Arabidopsis protein phosphatase 2C enzymes ABI1 and ABI2 and the protein tyrosine phosphatase AtPTP1 have been suggested to play such a role (Meinhard and Grill, 2001; Gupta and Luan, 2003). Interestingly, AtPTP1 regulates the activity of MAPKs, suggesting a tight link between H2O2, kinases and phosphatases.
Collectively, these observations provide quite compelling evidence of a specific signalling role for H2O2. However, there is little or no evidence to suggest that many of these proteins are the direct targets for oxidative modification by H2O2, and therefore the question of the identity of the H2O2 sensors up-stream of these proteins in a signalling cascade needs to be addressed.
Thiol modification is key
For proteins to perceive the presence of ROS, such as H2O2, and to act as signal intermediaries, there needs to be a specific recognition of H2O2 by the protein, or a direct chemical interaction which leads to signal propagation. The former is unlikely, owing to the small size of H2O2, but the latter is likely as the oxidizing nature of H2O2 will allow it to modify thiol residues in proteins directly, as has been suggested (Cooper et al., 2002; Vranová et al., 2002; Foyer and Noctor, 2005a), although other amino acids may be able to be oxidized too, such as Tyr, Trp, and His (Dröge, 2002). Because numerous cellular and extracellular proteins will be potential targets of this type of oxidative modification, the ROS sensors must have some specific characteristic that enables them to propagate this signal.
Oxidation of methionine may yield MetSO, and this can be reduced back to Met by protein MetSO reductases (PMSR) (Sadanandom et al., 2000). There are two types of this enzyme, type A and type B, the former uses thioredoxins (TRX) as a reductant (Romero et al. 2004).
Using cysteine as a H2O2 target, several results are feasible (Fig. 1). If there are two cysteine residues involved, the outcome may be the formation of a disulphide bridge (S-S). However, if there is one cysteine, the -SH group may be oxidized to varying degrees. The thiol group may be oxidized to sulphenic acid (-SOH), but this can be further oxidized to sulphinic acid (-SO2H) or sulphonic acid (-SO3H). Of course, any oxidation of the thiol is dependent on its mid-point redox potential and its availability to the oxidant, and therefore only a low proportion of the -SH groups within any protein will be able to be modified in these ways. In fact, there is likely to be a redox hierarchy, as thiols in different proteins will have different mid-point potentials, the proteins will be differentially controlled by fluctuations in the intracellular redox state, some being regulated earlier, or later than others as the redox state becomes more oxidized, for example. In this way, a fine subtle control over cellular activity can be achieved by oxidants such as H2O2.
It has been shown that within the tyrosine phosphatase of mammalian cells, an enzyme known to be inhibited by H2O2, the -SOH group is relatively unstable, and a cyclization within the peptide structure occurs which leads to the formation of a sulphenyl-amide intermediate, which can be decyclized to re-generate the original thiol group (Salmeen et al., 2003; Van Montford et al., 2003). Mechanistically, this rapid and reversible modification of a protein that can impact on phosphorylation cascades appears to be an ideal candidate for a signalling event.
These oxidation states of the -SH group within cysteine may be restored by re-reduction too. TRXs and glutaredoxins can act as protein disulphide reductases (Schürmann and Jacquot, 2000; Lemaire, 2004) as well as re-oxidizing -SOH groups (Rouhier et al., 2001; Collin et al., 2004), while sulphinic acid groups can be reduced back to the sulphenic acid group by sulphiredoxins, first found in yeast as an ATP-dependent enzyme (Biteau et al., 2003). The sulphenic acid group created can, of course, then be reduced further by TRX or glutaredoxins to regenerate the thiol, -SH.
It is clear therefore that reducing mechanisms, such as the use of thioredoxins is also key to these events, as reversal of the oxidized signal is vital if cell signalling is to function properly. For example, thioredoxins are able to reduce oxidized 2-cys peroxiredixins, which themselves have a function in removing peroxides, and it has been suggested that 2-cys peroxiredixins may act as redox sensors in chloroplasts (Konig et al., 2002, 2003, and discussed further in this special issue). Work has also been carried out to determine a fuller range of proteins able to interact with thioredoxins and, using peanut seed extracts, at least 20 such proteins have been identified (Yano et al., 2001).
In summary, there are redox groups within proteins that can potentially toggle between oxidation and reduction states in a rapid and ROS dose-dependent manner, and in doing so the structures of the proteins will be altered and such proteins may partake in H2O2-mediated signalling. There are also other proteins that can re-reduce the oxidized proteins, allowing them to be regenerated for further rounds of signalling. However, the proteins that are involved in such re-reduction processes can themselves exist in different redox states and they too may be able to be involved in signalling, in events yet to be unravelled.
Further complexity is added as thiols can also undergo modification by the addition of other groups, in the formation of what is referred to as mixed disulphides, or S-thiolation. One of the most common is the addition of glutathione, glutathionylation, a process which is also involved in the de-cyclization process of the sulphenyl-amide group of the mammalian tyrosine phosphatases (Salmeen et al., 2003). With a combination of in vitro and in vivo labelling techniques and two-dimensional gel electrophoresis in conjunction with mass spectroscopy, Dixon et al. (2005) identified 79 polypeptides that could be modified in this way from Arabidopsis. These included dehydroascorbate reductase, zeta-class glutathione transferase, nitrilase, alcohol dehydrogenase, and methionine synthase.
A further protein that has been shown to be able to be S-thiolated is glyceraldehyde 3-phosphate dehydrogenase (GAPDH) in Saccharomyces cerevisiae. However, this process is isoform specific, the Tdh3 isoform is modified, but the Tdh2 isoform is not, showing that it is not a general feature of this enzyme (Grant et al., 1999). Both isoforms were inhibited by H2O2, but only the Tdh3 activity recovered with time.
Nitric oxide and its influence on ROS sensing
Nitric oxide (NO) is often produced at the same time and in the same locations in plants as ROS and, like ROS, NO is involved in a plethora of responses and functions (reviewed by Neill et al., 2003). Of particular note here is that NO may also react with thiol groups on proteins in a process known as S-nitrosylation, to yield a -S-NO group. Studies have been carried out to determine which proteins in plants are modified in this way and it is found that many of the proteins found are also those which are potentially modified by H2O2 (Lindermayr et al., 2005). It is likely that NO and H2O2 may be covalently altering and, therefore, potentially controlling, the same complement of proteins in cells, and there may be a competition between H2O2 and NO at the level of thiol modification which may determine the exact signalling processes that ensue.
The role of histidine kinases in H2O2 perception
One of the most well-recognized cellular effects of ROS is the alteration of gene expression. Early studies targeted genes to assess whether the addition of ROS to cells increased, or decreased, the levels of gene expression but more holistic studies have used microarray analysis. Such studies have shown that the expression of approximately 3% of an organism's genome may be altered by ROS, with expression being increased and repressed (Gasch et al., 2000). In Arabidopsis, a transcriptomic study found approximately 170 genes for which the expression was increased more than 2-fold by application of exogenous H2O2, and approximately 65 genes having their expression reduced, although the whole genome was not analysed in this case (Desikan et al., 2001a). However, expression of these genes changed in response to different abiotic and biotic stresses, both independent of, and requiring H2O2, suggesting cross-talk with different pathways (Desikan et al., 2001a). Similar studies have been performed in tobacco (Vandanabeele et al., 2003), where some of the H2O2-regulated genes appeared to be those involved in other hormone and stress response pathways (Vandanabeele et al., 2003). Using an alternative approach by performing a microarray study on catalase-deficient Arabidopsis grown under high light exposure, the expression of 349 genes was found to be induced, whilst the expression of 88 genes was repressed, in response to H2O2 (Vanderauwera et al., 2005).
As well as showing that cells clearly can perceive H2O2, and that there must be a mechanism in place to transduce that signal to the nucleus, such transcriptome studies can also be used to highlight proteins that might be involved in H2O2 signalling and perception. A closer look at the transcriptome data from Arabidopsis shows that proteins involved in ethylene signalling, including histidine kinases, are induced by the addition of H2O2 (Desikan et al., 2001a). Furthermore, it has been reported that histidine kinase signalling is involved in the transcriptional response and tolerance of yeast to oxidative stress (Singh, 2000; Buck et al., 2001). A focused study on the histidine kinase receptor ETR1 from Arabidopsis showed that it was essential for H2O2 perception leading to stomatal closure (Desikan et al., 2005). The etr1-1 mutant, that contains a Cys65Tyr mutation, had reduced stomatal closure in response to H2O2, suggesting that the thiol of Cys65 was important for H2O2 signalling. Using mutants that lacked either histidine kinase activity, or the complete histidine kinase domain of ETR1 indicated that the kinase domain was not required for H2O2 signalling. In its other role as an ethylene receptor, it appears that the presence of the histidine kinase domain of ETR1 is required (Gamble et al., 2002), suggesting that the signalling through ETR1 invoked by H2O2 is different from that invoked by the presence of ethylene.
The Sacccharomyces cerevisiae mutant TM219, lacks a functional two-component system SLN1-SSK1, and therefore has enhanced susceptibility to H2O2, showing a greater inhibition of growth (Singh, 2000). Transformation of the TM219 mutant with full-length Arabidopsis ETR1 restored the yeast tolerance to H2O2. Transformation with only the first 128 amino acids of ETR1 had a similar effect, again indicating that the histidine kinase domain is not required for H2O2 signalling in ETR1. However, transformation of TM219 with a construct that contained the first 128 amino acids of ETR1, but lacking Cys65, did not restore H2O2 tolerance, confirming that the Cys65 residue of ETR1 is important for H2O2 signalling (Desikan et al., 2005).
The exact mechanisms of how ETR1 signalling occurs in both yeast and Arabidopsis has yet to be determined, but it appears that thiol modification of the protein may be the key, although whether this modification is a direct effect of H2O2, or mediated by another H2O2-sensing protein is not known.
Control of transcription factors by H2O2
As discussed above, one of the profound changes seen on addition of H2O2 is the control of gene expression, either being increased or decreased, depending on the gene investigated. Several transcription factors have been identified that are redox controlled. For example, in Escherichia coli the transcription factor OxyR is activated by cellular H2O2, causing disulphide bond formation between Cys199 and 208, occurring at a rate of 9.7 s−1 (Lee et al., 2004).
In S. cerevisiae the transcription factor Yap1 activates the expression of genes for antioxidants after the cells are exposed to oxidative stress. Regulation of Yap1 involves its nuclear accumulation, and its activation is by oxidation, with reduction mediated by thioredoxins. However, oxidation of the transcription factor is not directly by the hydroperoxide, but is mediated by glutathione peroxidase (GPx)-like enzyme Gpx3. On oxidation by H2O2, Cys36 on Gpx3 bridges to Cys598 on Yap1, forming a disulphide. This inter-polypeptide disulphide is then rearranged forming a disulphide in Yap1 itself, which activates the transcription factor. Therefore, in this case, the enzyme Gpx3 is acting as the H2O2 perception protein, and transduces the signal to the transcription factor (Delaunay et al., 2000, 2002).
Alterations in gene expression during systemic acquired resistance in plants are often modulated by the protein NPR1 (Non Expressor of Pathogenesis Related Genes 1) and TGA transcription factors. The latter are in the class of bZIP (basic leucine zipper) type transcription factors. In the non-activated state, NPR1 is in an oligomeric form, held by intermolecular disulphide bonds. On activation, NPR1 becomes reduced, forming the monomeric form, which then accumulates in the nucleus and alters gene expression (Mou et al., 2003). Mutations in Cys82 and Cys216 in NPR1 lead to constitutive activity, showing the importance of these cysteine residues in this control. Furthermore, reduction of cysteine residues in TGA1 and TGA4 allows them to interact with NPR1, and the latter stimulates TGA1 binding to DNA (Fobert and Despres, 2005). However, the exact proteins used in redox perception have yet to be identified, but it has been suggested that thioredoxins are likely to be involved (Fobert and Despres, 2005).
Identification of glyceraldehyde-3-phosphate dehydrogenase as a possible H2O2-sensing protein
The susceptibility of thiol groups within proteins to undergo oxidation suggests that they may also be open to attack from other forms of covalent modification, such as by the addition of an iodoacetamide group. Therefore, a competition could be set up between H2O2 and a tagged iodoacetamide that could be used to identify proteins. Such an approach was taken to identify H2O2-susceptible proteins in mammalian cells by Wu et al. (1998). They used a fluorescein tagged iodoacetamide, and estimated the molecular weights by using a western blot approach with anti-fluorescein antibodies. Further identification of the bands also used antibodies and they reported that one of the primary targets for H2O2 in mammalian cells was a tyrosine phosphatase (Wu et al., 1998).
Using the same basic approach, 5′-iodoacetamide fluorescein (IAF) could be seen to tag a number of proteins in cytosolic extracts from Arabidopsis cells (Hancock et al., 2005). However, pre-treatment with H2O2 severely reduced the binding of IAF to one particular protein at an approximate molecular weight of 40 kDa. Analysis of the sample using 2-dimensional gel electrophoresis showed that there were two main spots at approximately 40 kDa. Identification of such spots was undertaken via MALDI-TOF mass spectroscopy, and both were identified as the enzyme glyceraldehyde 3-phosphate dehydrogenase (GAPDH; Hancock et al., 2005). This is not an enzyme that would be expected to be involved in signalling events, being an enzyme integral within the glycolytic pathway. However, this enzyme has already been reported to be present in plant cells in locations other than that expected if it was only involved in glycolysis, for example, it has been identified using a proteomics approach from the cell walls of Arabidopsis (Chivasa et al., 2002), and found in the peribacteroid membrane of Lotus japonicus (Wienkoop and Saalbach, 2003). In mammalian cells it has been shown that not only does oxidation of a cysteine within GAPDH cause inactivation of its normal function, but it causes structural changes which allows the enzyme to form protein–protein interactions with the enzyme phospholipase D, activating the latter, and so propagating signalling events (Kim et al., 2003).
Further unexpected roles and activities of GAPDH include its involvement in age-induced apoptosis in neuronal cells (Ishitani et al., 1996) and its translocation within cells during oxidative stress. Such translocation has been reported to occur into the nucleus in animal cells (Dastoor and Dreyer, 2001) and to become associated with mitochondria in plants (Sweetlove et al., 2002).
For GAPDH to be involved in ROS signalling in Arabidopsis it might be assumed that a change in its activity will be seen in the presence of H2O2. In mammalian cells Cys149 of GAPDH has been shown to be oxidized by H2O2 (Brodie and Reed, 1987), and so similar effects would be expected in the plant enzyme. Addition of H2O2 to cytosolic extracts from Arabidopsis inhibited GAPDH activity in a concentration-dependent manner, although inhibition was not instantaneous, but rather had a time-dependent element, suggesting that turnover of the enzyme is required to allow oxidation to take place (Hancock et al., 2005).
However, as discussed above, for a protein to be involved in ROS signalling any oxidation that takes place ideally needs to be reversible. When GAPDH activity in Arabidopsis extracts was inhibited with low concentrations of H2O2, the activity could be restored with reduced glutathione, or the reductant dithiothreitol, suggesting that the inhibition is indeed reversible. However, little or no reversal was seen if the inhibition of the activity was with higher concentrations of H2O2 (>0.25 mM), suggesting that higher irreversible oxidation states of a thiol within the GAPDH enzyme are achieved (Hancock et al., 2005).
GAPDH from plants has also been reported to be inhibited by nitric oxide, and such inhibition could be reversed with reductants, in this case dithiothreitol (Lindermayr et al., 2005). With peptide mass spectrometry, the exact covalent modifications of the thiol groups within GAPDH that take place on H2O2 addition should be able to be determined, and it will be interesting to see if H2O2 and NO, both key signalling molecules, are targeting the same thiol groups in this enzyme.
Lastly, GAPDH has been shown in other systems to be S-thiolated, by the addition of glutathione to the thiol side group, for example in S. cerevisiae (Grant et al., 1999). Whether such modifications take place in plants has yet to be determined, but it is not unlikely.
Other proteins which may be involved in H2O2 sensing
The notion that histidine kinases might be involved in ROS signalling stemmed from microarray studies (Desikan et al., 2001a). Similarly, it is not impossible that other genes regulated by H2O2 would also be targeted by H2O2 at the protein level. For example, from the microarray studies, several transcription factors have been identified whose expression is regulated by H2O2 (Desikan et al., 2001a; Vandenabeele et al., 2003). It is possible that these transcription factors are also direct targets for redox modification by H2O2, as discussed for other systems above, and there is clearly more work to be done in this area.
Many proteins will be able to be modified covalently by ROS, but some may alter their structure because they contain a prosthetic group which can be reduced or oxidized. Here, cytochrome c is a candidate for ROS attack (Hancock et al., 2001, 2003). Cytochrome c can be oxidized by H2O2, but it can be reduced by superoxide anions, and therefore cytochrome c may exist in a reduced state, or an oxidized state, depending on which ROS is most prevalent in the cell. It has also been shown that the structure of the protein, and therefore its potential ability to signal during, for example, apoptosis, is determined by its oxidation state (Calver et al., 1997).
Clearly, there are several proteins that could potentially be oxidized by H2O2, and many of those, if not all of them, could also undergo S-nitrosylation, as the thiol groups which are susceptible to H2O2 attack will also be open for nitric oxide reactions. Fluorescent tagging with iodoacetamide in the presence and absence of H2O2 revealed not only GAPDH, but also other potential H2O2 targets. These included alcohol dehydrogenase, S-adenosyl methionine synthase 2, and glutamine synthetase (Hancock et al., 2005). Alcohol dehydrogenase has already been shown to be inhibited by H2O2 in other systems (Echave et al., 2003), and methionine adenosyltransferase from mammalian systems is reversibly inactivated by H2O2 through covalent modification of a cysteine residue (Sanchez-Gongora et al., 1997), a residue that is also the target for nitric oxide (Perez-Mato et al., 1999).
Glutamine synthetase from mammalian systems has been shown to be sensitive to oxidative modification, a process thought to be involved in neurogenerative disease (Butterfield et al., 1997), while in bacteria, oxidation of a histidine residue in glutamine synthetase is involved in the proteolytic breakdown of the enzyme (Levine, 1983), a process also reported in plants (Ortega et al., 1999).
However, the thiol tagging reported (Hancock et al., 2005) is only a limited study, with proteins only being identified within the separation range of the gels used. Further, more broad analysis of tagged proteins will no doubt reveal a plethora of other proteins that will be sensitive to ROS, and to H2O2 in particular.
Redox control of many other proteins has also been suggested, for example, ones involved in carbon storage and partitioning in plants. ADP-glucose pyrophosphorylase (AGPase) is the key regulatory enzyme of starch synthesis in the plastid and it has been reported that, in potato tubers, AGPase is subject to post-translational redox modification, and that the enzyme from pea leaf chloroplasts is activated by reduced thioredoxin f or m (Geigenberger et al., 2005).
Using a variety of techniques, either an informed study arising from transcriptomic analysis, or thiol tagging and targeted proteomics, a variety of proteins can be identified that are potentially involved in the direct reaction with H2O2, either when it enters a cell from outside, or is generated inside the cell itself (Fig. 2). Clearly, the location of the H2O2-sensitive protein is key to whether it is able to react and respond, and thus initiate a signal transduction pathway. Also key to whether a response is mounted is the redox environment in which the potential H2O2-sensing protein is located, as in a very reducing environment little H2O2 signalling might be able to proceed.
With a range of proteins able to respond to a single signal, probably each with a different outcome, it is clear that how such proteins operate and interact with ROS from different sources needs to be determined. The measurement of the presence of ROS along with the estimation of intracellular redox states, both spatially and temporally, will be vital to an understanding of ROS signalling.
Oxidation of thiol groups in proteins suggests that these side chains are in a redox state that is able to lose a proton readily, and that these groups are located within a protein such that they are open to attack. These thiols will also be able to be covalently modified by nitric oxide, and, as discussed above, several proteins have already been shown to be both oxidized and S-nitrosylated. It is likely that all S-nitrosylation sites are able to be oxidized by H2O2. ROS and nitric oxide react together chemically to produce peroxynitrite, and therefore the interplay between ROS and nitric oxide at both the chemical level and at the level of protein modification need to be understood, as these key signalling molecules obviously do not work in isolation.
Finally, it is intriguing that several of the proteins discussed above that are now thought to be involved in ROS signalling had defined roles already assigned to them, for example, the ethylene receptor ETR1, and the glycolytic enzyme GAPDH. It is now clear that such proteins can partake in unexpected functions, that of signalling the presence of ROS, in particular H2O2. It may be that the expression patterns, post-translational modifications, protein–protein interactions, the cellular locations of the proteins or the redox environments in which they exist, are involved in the mechanisms that allow such proteins to balance their roles in the cell. Clearly, a full understanding of how such proteins have multiple, often unexpected, roles in cells remains a challenge.
We would like to thank the Biotechnology and Biological Sciences Research Council (BBSRC) UK and the Leverhulme Trust for funding this research.