Reversible oxidation of thiol proteins is an important cell signalling mechanism. In many cases, this involves generation or exposure of the cells to H2O2, and oxidation of proteins that are not particularly H2O2-reactive. There is a conundrum as to how these proteins are oxidised when other highly reactive proteins such as peroxiredoxins are present. This article discusses potential mechanisms, focussing on recent evidence for oxidation being localised within the cell, redox relays involving peroxiredoxins operating in some signalling pathways, and mechanisms for facilitated or directed oxidation of specific targets. These findings help define conditions that enable redox signalling but there is still much to learn regarding mechanisms.
Hydrogen peroxide is produced as a by-product of mitochondrial respiration and an end product of many metabolic reactions, and (either directly or via superoxide) by membrane-associated NADPH oxidases (NOXs) in response to external signals such as growth factors or hormones. Cells may also be exposed to exogenous H2O2, for example from radiation or redox-active xenobiotics. Although toxic at high concentrations, H2O2 and other reactive oxygen species can also elicit regulated cell responses by a process termed redox signalling. Of the various mechanisms identified for initiating or transmitting redox signals, the best established involves the reversible oxidation of thiol proteins [1–5]. Cellular responses to endogenous and exogenous H2O2 have been studied for many years, and although H2O2 is not the only species capable of thiol oxidation, it is the most widely implicated oxidant. Put simply, the underlying basis for redox regulation is the targeting of particular thiol protein(s) by an oxidant (in the present case H2O2) to cause a change of function. This may be activation or inhibition of enzymatic activity, or a change in conformation or binding characteristics. Downstream consequences include activation or repression of transcription factors and gene expression, and modulation of phosphorylation pathways, energy metabolism, ion transport and cytoskeletal structure, as extensively reviewed by others [4,6,7]. Numerous cell processes have been shown to be redox regulated. These include growth and differentiation, activation of protective stress responses such as expression of chaperones and phase II enzymes, and the initiation of programmed cell death.
A key requirement for cell signalling is specificity, and the focus of this article is on how oxidation of specific protein(s) in a particular pathway is achieved, when there are numerous potential targets in a cell. Additionally, different stimuli need to elicit their own specific responses. If H2O2 itself is the signal and the outcome is a general stress response, its source may not matter. However, if H2O2 is a secondary product of receptor activation, then the pathway targeted by H2O2 should depend on the nature of the activator.
Understanding how specificity is achieved has been a major challenge. There are many examples of thiol proteins in regulatory pathways that have been shown to be selectively oxidised during receptor activation or mild oxidative stress [8–11]. Yet in isolation, these proteins are not particularly reactive with H2O2 and there are few cases where the initial target for H2O2 and the specific reactions involved in transmitting the signal are well understood. There are also uncertainties about how redox signalling pathways operate in cells rich in antioxidant defences and whether they can distinguish H2O2 generated by different sources such as NOX activation, mitochondria, or as a metabolic end product in peroxisomes or the endoplasmic reticulum. The conundrum of rationalising the chemistry and cellular effects of reactive oxygen species has been addressed by us [2,12,13] and others [14–16], and potential mechanisms for providing specific oxidation have been proposed. There have been advances In the 12 years since we originally discussed these issues, and after first summarising key properties of H2O2 and thiols relevant to their reactions in cells, this article assesses how newer information supports, or otherwise, various proposed models. Its focus is predominantly on mammalian systems. However, it is important to appreciate that redox signalling mechanisms are widely employed in plants and there have been significant recent advances in elucidating these mechanisms, as for example reviewed in .
H2O2 and thiol reactivity
H2O2 is a strong oxidant but it reacts with relatively few biological molecules. Transition metal centres, and sulfur and seleno compounds, are its most favoured targets. For thiols, the initial product is a sulfenic acid, which is generally a transient species that reacts with another thiol to form a disulfide. Further oxidation of the sulfenic acid progressively gives the sulfinic and sulfonic acids. As H2O2 reacts much faster with the thiolate anion than the protonated thiol, at physiological pH thiols with a low pKa are more reactive. GSH and most thiols have a pKa above 8, but a minority of protein thiols have a low pKa and are ionised. However, this alone is insufficient to make them favoured cellular targets. Reactivity is assessed based on rate constant, which for H2O2 reacting with the thiolate of cysteine and most thiol proteins is ∼20 M−1 s−1, or in a few instances up to 10-fold higher . In contrast, peroxiredoxins and glutathione peroxidases are abundant throughout cells and have rate constants in the 106–108 M−1 s−1 range. So a major conundrum is how other proteins can be oxidised in the presence of these highly reactive targets.
Diffusion and sites of H2O2 production
Redox signalling frequently occurs at a localised site within the cell, so the source of oxidant and its ability to diffuse are important considerations. H2O2 can pass through membranes, albeit aided by aquaporins (AQPs) , so a membrane barrier alone is insufficient to localise its actions. The distance it can diffuse from a site of production depends on what reactive targets are present in its path. High concentrations of reactive species such as peroxiredoxins will consume H2O2 locally . However, less reactive targets are not able to restrict H2O2 from diffusing beyond the dimensions of a cell. To illustrate this point, a peroxiredoxin could consume 100 µM H2O2 in a fraction of a second and thus restrict its movement whereas a typical low pKa thiol protein would take many minutes and allow it to diffuse away. This implies that even when production is localised, it does not necessarily follow that H2O2 will react at that site. Thus, there is a need to understand how less reactive regulatory thiol proteins undergo site-localised oxidation.
NOXs, along with the mitochondrial electron transport chain, are an important source of H2O2 and NOX activation is strongly implicated in redox signalling [19,20]. The NOX family are membrane-associated multiprotein complexes that catalyse the transfer of electrons from NADPH to oxygen. The product is superoxide which then dismutase to H2O2, or in some cases (as with NOX4) H2O2 is generated directly . NADPH is oxidised in the cytoplasm and superoxide (or H2O2) released on the external side of the membrane. This may be to the outside of the cell, possibly in a localised area, or into an internal organelle such as an endosome . The oxidant must then act externally or with targets in the organelle, or it must traverse the membrane to act intracellularly. The extent to which NOX-derived oxidants cross membranes or act locally is not fully understood.
Substantial amounts of H2O2 are also produced in peroxisomes , for example in the β-oxidation of fatty acids, and in the endoplasmic reticulum during oxidative folding of exported proteins [23,24]. For example production of antibodies by plasma cells has been estimated to generate tens of micromolar H2O2 per minute from disulfide bond formation . Such levels may well exceed the H2O2 generated from NOX activation. Peroxisomes are rich in catalase and the endoplasmic reticulum in peroxiredoxin 4 and glutathione peroxidases 7 and 8, so they have a high capacity to consume H2O2 . However, some appear able to exit the organelles , and another conundrum is whether cell signalling mechanisms distinguish this H2O2 from other sources.
Proposed mechanisms for specificity in cell signalling
Various mechanisms have been proposed for thiol-based redox signalling that accommodate the chemical properties of H2O2. These mechanisms are not mutually exclusive, they may act in combination, and the same mechanism need not apply for all redox signals. Also, in extrapolating chemical data, the complexity of the cellular environment needs to be appreciated. H2O2 chemistry has largely been studied in homogeneous solution, whereas even without considering their numerous membranous structures and organelles, cells are clearly not homogeneous and H2O2 production and targets not necessarily evenly distributed. While the underlying chemistry of the reacting species will still hold, it will be modulated by cellular conditions including concentration gradients, intermolecular interactions, and compartmentalisation. This may make some reactions more favourable in cells than they are in free solution.
Figure 1 gives a (not necessarily complete) list of mechanisms that have been proposed for transmitting a thiol redox signal. (A) is simply direct oxidation of a specific target and presumably requires a highly reactive thiol and/or a minimum of competitors. (B) is a relay in which a highly reactive sensor protein transmits oxidising equivalents from H2O2 to another target. (C) is also a sensor mechanism in which the sensor is recycled by a redoxin (such as thioredoxin) that regulates the redox state of other target proteins. In (D), oxidation of a sensor disrupts its interaction with a target protein, thereby affecting its localisation or activity. (E) requires inactivation of highly reactive proteins (such as peroxiredoxins) to enable the oxidation of less favoured targets (the floodgate model ). In (F) an association with the H2O2 source enables facilitated oxidation of the target without free diffusion and scavenging by more reactive competitors.
Possible mechanisms for transmission of a redox signal initiated by H2O2.
Technological advance — fluorescent probes and localised imaging of intracellular H2O2
Genetically encoded green fluorescent protein (GFP) derivatives have proved to be invaluable tools for the intracellular detection of H2O2. With the specificity achieved by coupling an engineered GFP to an H2O2-sensitive thiol protein (oxy R, Orp1 or more recently the more sensitive yeast peroxiredoxin, TSA2 [27–31]), they have provided reliable imaging of where H2O2 is produced, the dynamics of production, and estimates of intracellular levels. More recently, insight about localisation has been obtained by incorporating appropriate targeting sequences to direct the probes to different cell compartments, including mitochondria, endoplasmic reticulum, nucleus, and cytoplasm [27,32–34]. Using these probes, it has been possible to show that H2O2 production is localised following growth factor stimulation, and that it reacts close to the site of generation with only limited diffusion [32,35]. Another technological advance has been to express d-amino acid oxidase, as a regulated generator of H2O2 , at different cellular sites and follow the fate of the H2O2 with targeted fluorescent probes . This approach has substantiated the conclusion that intracellular diffusion is limited, and observations that thioredoxin reductase inhibition but not GSH depletion enabled probe oxidation to occur throughout the cell imply that localised consumption involves thioredoxin-dependent reactions . These data support kinetic arguments that site-localisation requires a highly reactive target, and point to peroxiredoxins (which undergo thioredoxin-dependent recycling) as likely candidates. They are less supportive of a mechanism (Figure 1E) in which inhibiting peroxiredoxin activity enables H2O2 to act locally with a less reactive target. Localised oxidation following inactivation of a peroxiredoxin has been reported [39,40] and it is possible that this may require some form of facilitated reaction (F) may also be involved to restrain the H2O2.
Advances in thiol proteomics
Rapid advances are being made in mass spectrometric methods for detecting thiol oxidation at specific cysteine residues and quantifying the sensitivity of different proteins to oxidation [41–43]. These include the development of reagents that enable rapid derivatisation without artefactual oxidation, derivatisation strategies that distinguish different oxidation states, and quantitative proteomic analyses that pick up low-expressing proteins. Each of these has been a challenge. There is still a way to go towards detecting low-expressing proteins and identifying those that undergo oxidation during a particular cell signalling event. However, with reports of methodologies that can quantify the redox state of some 70 000 protein cysteine residues , a route towards achieving this appears to be in sight.
As more quantitative information is being acquired about the kinetics of thiol reactions and the enzymology of thiol proteins, mathematical approaches are being applied to modelling the organisation of redox networks and assessing what reactions are likely in cellular environments . For example, modelling has been used to identify conditions that would favour localised redox relays , to assess how extracellular H2O2 exposure relates to intracellular peroxiredoxin , and show how the makeup of different cell types can influence their response to H2O2 .
Identification of redox relays
To overcome the low reactivity of H2O2 with thiol proteins involved in signalling pathways, redox relay mechanisms (Figure 1B) are being actively explored [2,4,16,48]. Reactive sensor proteins are proposed to act as the initial target for H2O2 then transfer oxidising equivalents to less reactive signalling proteins. This mechanism was first identified in Saccharomyces cerevisiae in 2002 by Delaunay et al.  for H2O2-mediated activation of the transcription factor, Yap1, via the transfer of oxidising equivalents from the peroxiredoxin, Orp1. This was soon followed by findings of a related mechanism for activation of the stress-activated protein kinase Sty1 in Schizosaccharomyces pombe . A relay mechanism is appealing as it accommodates the high peroxide reactivity of the peroxiredoxins, but surprisingly for more than a decade few other relays came to light. However, examples of transfer of oxidising equivalents via peroxiredoxin sensors in mammalian cells are now being identified. These include H2O2-mediated oxidation of apoptosis signal-regulating kinase 1 (ASK1) by peroxiredoxin 1 , oxidation of STAT3 via peroxiredoxins 1 and/or 2 , regulation of FOXO3 signalling by peroxiredoxin 1 , and regulation of the nuclear transfer of HMBG1 by peroxiredoxins 1 and 2 . Peroxiredoxin-mediated oxidation of other cellular thiol proteins has been demonstrated by Stocker et al. , who observed that fewer more thiol proteins formed disulfides in cells treated with H2O2 when peroxiredoxins 1 and 2 were knocked out and postulated that relay mechanisms are widespread. Identification of peroxiredoxin binding partners and oxidant dependent mixed disulfides with other thiol proteins [53,56,57] may also be indicative of redox relays.
Rather than directly transferring oxidising equivalents to a target, a sensor could propagate a signal if oxidation altered its interaction with a signalling protein so as to affect its activity or location (Figure 1C). An example is the interaction of ASK1 with thioredoxin 1. Reduced thioredoxin binds ASK1 and inhibits its activity, but when the thioredoxin becomes oxidised the ASK1 dissociates and converts to its active form [58,59].
Evidence is accumulating to support redox signalling mechanisms involving oxidant sensors and redox relays . However, at the present time, the number of redox relays that have been truly validated from signal recognition to downstream effect is still low. A possible reason why characterisation has proved difficult is that other relays may require multiple components, as has been found with signalling via YAP1 . This relay requires a scaffold protein, which not only brings sensor and target together, but also changes their kinetic characteristics so as to favour oxidant transfer between them. It may be that scaffold proteins, or hubs are more generally required to bring redox signalling partners together.
Regulation of redox signalling by thiol reduction and exchange
Redox signalling pathways are mostly transient and reversible so they should be regulated by reductants as well as oxidants. The main mechanisms for disulfide reduction involve thioredoxin and GSH/glutaredoxin, each linked via its reductase to NADPH [61,62]. The activity of the relevant reducing system can regulate the oxidation state of its target protein, and preferential association or reactivity with different targets provides an opportunity for selectivity . These reductants, by recycling oxidised peroxiredoxins and glutathione peroxidases, provide a variation of the relay mechanism (Figure 1D), in which oxidising equivalents are transmitted from the sensor to its reductant and on to another protein . This mechanism then enables H2O2 to act via a sensor to regulate proteins that are reduced by thioredoxin or the glutaredoxin system, either by oxidising the reductant and making it less available for recycling client proteins, or by regulating their redox state by thiol/disulfide exchange. An example of this mechanism is H2O2-mediated activation of the AP1-like transcription factor pap1 in S. pombe, which involves the transfer of oxidising equivalents from a peroxiredoxin sensor via the thioredoxin-like protein Txl1 . This type of mechanism may be more widespread, especially as other proteins including protein disulfide isomerases [65,66] and cyclophilins  can reduce peroxiredoxins.
A variation on the peroxiredoxin/thioredoxin relay mechanism has been proposed , based on observations with S. pombe that peroxiredoxin hyperoxidation facilitates the repair of oxidised proteins after H2O2 exposure. In the classical floodgate model (Figure 1E), hyperoxidation of peroxiredoxins is proposed to allow H2O2 to oxidise less reactive thiol proteins. In this alternative model, hyperoxidation removes the peroxiredoxin from its reversible redox cycle, thus making thioredoxin available to reduce other proteins. Whether this mechanism applies to other systems is yet to be established.
2O 2 via AQPs
The AQPs have long been recognised as membrane-traversing proteins that are important for cellular uptake and transfer of water. More recently, members of this family, particularly AQPs 3, 8, 9, and 11, have been shown to be peroxiporins that transport H2O2 and facilitate its diffusion into and out of cells and between cell compartments . AQP expression can be regulated, and selective expression or membrane localisation could be a mechanism for directing H2O2 and promoting specificity in cell signalling. Peroxiporin transport of NOX-derived H2O2 has been implicated in growth factor signalling  and an interaction between NOX2 and AQP3 has been observed . This raises the possibility of peroxiporins channelling H2O2 into cells at the site of generation and providing directed oxidation of signalling targets (Figure 1F). While at present largely speculative, this mechanism warrants further investigation.
H2O2 reacts with CO2/bicarbonate to produce peroxymonocarbonate () (Figure 2A) [71,72]. is a more reactive species than H2O2 that has received little attention in relation to redox signalling. reacts faster than H2O2 with methionine and GSH [71,73]. However, because it is formed in an equilibrium that lies well to the left, at physiological CO2/bicarbonate concentrations only a few percent of H2O2 is present as . This is sufficient to increase their overall oxidation rates by no more than about twofold. In contrast, some thiol proteins are much more reactive with and oxidation by H2O2 is orders of magnitude faster when CO2/bicarbonate is present. This was first observed with the protein tyrosine phosphatases PTP1B and SHP2 , and it was subsequently shown that the addition of CO2/bicarbonate enabled H2O2 to inactivate PTP1B in the presence of a peroxiredoxin/thioredoxin system (Figure 2B) . Most significantly, PTP1B oxidation associated with epidermal growth factor receptor activation was observed only with cells in bicarbonate-containing medium (Figure 2C). These findings suggest a potential mechanism for regulating PTP-dependent signalling involving both peroxiredoxin recycling and variations in CO2/bicarbonate (Figure 2).
Effect of bicarbonate on inactivation of PTP1B.
The high reactivity of observed with PTP1B appears to be protein selective. Hyperoxidation of peroxiredoxins is also dramatically faster in bicarbonate buffer [76,77], but oxidation of papain and albumin by H2O2 showed only modest enhancement [71,74]. However, more extensive investigation is required to determine the selectivity of peroxymonocarbonate-mediated oxidation and whether it provides a plausible explanation for how some thiol proteins that react slowly with H2O2 (when bicarbonate is typically not present) appear to be oxidant-sensitive during cell signalling.
There is still much to learn about how thiol proteins transmit redox signals. However, advances have been made in understanding cellular mechanisms that could enable less reactive proteins to become oxidised. Based on current evidence, it is unlikely that there is a single unifying redox signalling mechanism, and probable that synergy between mechanisms occurs in some situations. With H2O2 exposure in the signalling range, imaging evidence points to it acting locally, close to the site of generation or entry into the cell. This implies that there are high reacting targets in the vicinity, or a mechanism for facilitated oxidation of other targets near the site. Simply inactivating reactive targets such of peroxiredoxins would enable H2O2 to diffuse more widely and not be restrained to its generation site.
Relays from a reactive sensor such as a peroxiredoxin fit well with localised action and there is now good evidence for redox signals being transmitted by a relay mechanism in mammalian as well as yeast cells. However, there are still relatively few redox signalling pathways that have been fully characterised. Exact mechanisms of transfer of oxidising equivalents need to be elucidated and as does the extent to which scaffold proteins are involved. These are productive areas for future research.
There is suggestive evidence for signalling via direct H2O2 transfer to specific targets. AQPs facilitate membrane transfer of H2O2, and an observed association with NOX2 needs further study. A well-characterised example of direct transfer of H2O2 from NOX4 has been observed in elF2α signalling . In this case, a metallo- rather than thiol enzyme, protein phosphatase 1, undergoes targeted oxidation to activate the pathway. Recent findings with bicarbonate highlight the potential for facilitated thiol protein oxidation by peroxymonocarbonate. This appears to have a role in growth factor signalling and warrants further exploration. It should also be noted that although this article has focussed on H2O2, signalling can involve other reactive oxygen species, and thiol proteins are not the only potential targets.
A conundrum for redox signalling is how thiol proteins involved in regulatory pathways become oxidised when their innate reactivity with H2O2 is low. A single unifying mechanism appears unlikely, as examples of redox relays, facilitated transfer from the oxidant source, and activation through reaction with bicarbonate have all been identified.
Cellular imaging of H2O2 has shown limited diffusion from its site of production. This implies that it reacts with highly reactive targets such as peroxiredoxins in the vicinity and is thus supportive of peroxiredoxin-mediated relays.
Several examples of thiol protein oxidation by H2O2 via a peroxiredoxin have now been identified in mammalian as well as yeast cells, but a lot more needs to be learned about the prevalence of such redox relays and exact mechanisms.
The author declares that there are no competing interests associated with this manuscript.
The author's work related to this article has been supported by the Health Research Council of New Zealand and the New Zealand Marsden Fund.