Shifting paradigms and novel players in Cys-based redox regulation and ROS signaling in plants - and where to go next

Introduction

Cys-based redox regulation has claimed a central place in the control of metabolic, developmental and acclimatory processes of plants in recent years. Drivers in understanding were advancements in technologies and novel approaches. In this context, photosynthetic cells serve both as a biological system with outstanding importance for life providing carbon and energy on the one hand, and as a blueprint to scrutinize principles and dynamics of reactive oxygen species and redox regulation on the other hand. Since evolution has been utilizing the redox regulatory toolbox in different contexts in different organisms depending on their specific lifestyles we expect particular sophistication for a photoautotrophic, sessile organism. At the same time, we expect the toolbox to be employed under the same fundamental biophysical and biochemical constraints across biology. In this review, we discuss recent advancements in understanding the dynamics and the scope of redox regulation in plants. The article does not aim to present a comprehensive coverage of redox systems and target processes. For the sake of focus we deliberately chose to leave out important aspects, which have also experienced extensive progress, such as protein nitrosylation, sulfenylation, persulfidation, and H₂S signaling. Rather, we highlight a selection of recent major changes in thinking and concepts in context of Cys-based redox- and reactive oxygen species (ROS) signaling.

Paradigm shift 1: ROS production takes place under any condition

Superoxide (O₂⁻), hydrogen peroxide (H₂O₂) and other ROS are formed both as byproducts of cellular metabolism and through specifically evolved generator systems. Several of the underlying cellular processes and chemical mechanisms are known and have been extensively covered in reviews (Apel and Hirt 2004; Waszczak et al. 2018). For more than a quarter of a century, the relevant literature stresses that ROS should no longer be considered solely as damaging compounds but as signaling messengers, e.g., involved in controlling local and global acclimation responses or in triggering cell death programs. Before that paradigm shift, ROS were merely considered markers of severe stress and dysfunction.

Currently we are witness to a second major paradigm shift. H₂O₂ and other ROS are now regarded as essential under any physiological condition by functioning as vital, and readily available, electron sinks to properly adjust the redox state of cellular Cys-based redox systems.

ROS are products of enzymatic reactions of metabolism, electron transport chains and specific generator systems. L-2-hydroxyacid oxidases, which include glycolate oxidase in photorespiration, and glucose oxidase are examples of enzymes, which release stoichiometric H₂O₂ amounts in normal metabolism, usually in peroxisomes (Foyer and Noctor 2003; Pan et al. 2020). Some oxidases like polyamine oxidase (PAOs) reside in the apoplast and contribute to biotic and abiotic stress acclimation (Pottosin et al. 2014) (see Paradigm shift 9).

Photosynthetic electron transport (PET) and respiratory electron transport (RET) generate O₂⁻ which readily dismutates to O₂ and H₂O₂. Interestingly, evolution of PET in angiosperms is based on ferredoxin as the terminal electron hub, which eases ROS production, while cyanobacteria, algae and plant lineages other than angiosperms rely on a group of Fe- and flavin-dependent electron transmitters that scarcely produce ROS in the Mehler reaction but fully reduce O₂ to H₂O. Thus, flavodiiron proteins transfer electrons from PET to O₂ and produce H₂O without intermittent release of O₂⁻ or H₂O₂ (Santana-Sanchez et al. 2019). Flavodoxin-expressing tobacco produces less chloroplast-derived ROS, displays higher stress tolerance and delayed senescence (Mayta et al. 2018). The authors discuss three major mechanisms for this positive flavodoxin effect in transgenic angiosperms, namely (i) their function as efficient electron sink particularly when the active ferredoxin pool declines during senescence or under stress, (ii) the stimulation of the antioxidant systems and (iii) maintenance of a balanced reduced state of the cell.

PET-dependent ROS production comes with a negative tradeoff under severe stress due to its damaging potential but is an important regulator under conditions of low and medium stress, and possibly particularly relevant under fluctuating environmental conditions. An analogous argument can be made for RET in the mitochondria, where O₂⁻ is readily formed at specific sites at the complexes I and III, but not others (Huang et al. 2016; Murphy 2009). The fact that mitochondrial ROS production has not been ‘fixed’ during evolution could either mean that (i) avoiding it is biochemically impossible or (ii) that ROS production has adopted an important physiological role and its absence would come with a disadvantage that is selected against.

The rate and function of Mehler reaction in the PET of angiosperms have long been a matter of debate but, more recently, evidence in favor of a suppressed rate of the Mehler reaction with a role in electron transport regulation appears to prevail (Heber 2002). O₂ photoreduction is lower in angiosperms than in gymnosperms (Shirao et al. 2013). The low rate of O₂ reduction is consistent with results in French bean from Driever and Baker (2011) who interpreted the low rate as indication for function in regulation, rather than for function as a major alternative electron sink. While the paradigm of ROS generation is still in the process of shifting, the emerging new picture regards ROS production by PET and RET as required for normal physiology, selected for and maintained by evolution, and actively regulated.

Paradigm shift 2: photosynthesis makes the chloroplast stroma a unique ‘redox battle ground’

Ever since the discovery of the thioredoxins (TRXs) in the chloroplasts (Wolosiuk and Buchanan 1977), they have retained their significance as a biological hotspot in redox regulation. Their importance is reflected by the number and diversity of TRXs and TRX-like proteins (Geigenberger et al. 2017), which greatly exceeds the number in non-plant organisms (see Paradigm shift 3).

Three key characteristics of photosynthesis make stromal Cys- and ROS-based redox regulation unique: (i) the option for direct delivery of electrons from the photosynthetic light reactions to the Cys-based redox machinery via ferredoxin-thioredoxin-reductase (FTR), (ii) the very high rates of electron flux that photosynthesis can adopt along with inherent ROS-production (see Paradigm shift 1) and (iii) the dependence of PET-derived ROS production on external illumination. As a result, there is a uniquely direct and dominant connection between the external environment (the sun/light) and the stromal Cys-redox machinery. By contrast, internal control is possible over the electron fluxes that enter or leave the Cys-redox machinery in heterotrophic cells and cell compartments of plant cells other than the chloroplast, e.g., at the level of carbon metabolite supply through membrane transport into the organism, the cell or the cell compartment. Illumination-coupled electron fluxes through the stromal Cys-redox machinery make fluctuations during the illumination phase unpredictable. Additionally, changes in light-driven electron flux rates and changes in redox potential of the involved redox couples can be drastic and potentially difficult to actively over-write by non-light-driven, endogenous mechanisms of reduction or oxidation (i.e., NADPH and ROS not derived from illumination-driven processes) (Figure 1). Hence, regulation apart from adjusting photochemical and non-photochemical quenching mechanisms must occur downstream of illumination, and both reductant and oxidant (i.e., ROS, see Paradigm shift 1) fluxes to drive Cys-based redox switching need to be supplied on demand by PET. These considerations have remained at the level of plausibility arguments for years, e.g., rates of ROS production could be estimated and correlated with particularly high levels of ascorbate suggestive of a particularly high need for ROS scavenging (Foyer and Noctor 2011). Recently, however, in vivo biosensing techniques to measure the in situ redox dynamics of the glutathione pool as well as estimations of stromal H₂O₂ dynamics have enabled a more direct picture. Interestingly, the onset of illumination in Physcomitrella patens consistently led to an initial oxidation of the glutathione pool during the first seconds of the light phase as measured with the Grx1-roGFP2 biosensor (Müller-Schüssele et al. 2020). Subsequently, gradual reduction of the glutathione pool to values below the redox potential in dark-adapted plants occurred. In the absence of plastidic glutathione reductase (GR), the initial oxidation of the probe was maintained, and the subsequent reduction was completely suppressed. Pronounced transient glutathione oxidation at dark-to-light and light-to-dark transitions was also observed in Arabidopsis chloroplasts albeit at a time scale of minutes to hours (Haber and Rosenwasser 2020). This dynamic hysteresis of the glutathione redox potential (E_GSH) appears to reflect the permanent ‘battle’ of reducing and oxidizing powers and the dynamics in the directionality of electron flux through the redox systems in the stroma (Figure 1). In extreme cases, imbalances created by the lack of key enzymes for mediating efficient electron transfer can lead to death as shown for Arabidopsis thaliana mutants lacking plastidic GR (Marty et al. 2019). Using a novel, rationally designed fluorescent protein sensor construct with single excitation and dual blue-green emission named FROG/B, Sugiura et al. (2020) interestingly found no transient oxidation during the illumination transitions but observed a gradual light-dependent reduction in both photosynthetically active vegetative cells and heterocysts of Anabaena sp. PCC 7120. Although this response was clearly dependent on PET, the response in heterocysts may actually suggest that it is more likely to reflect metabolic fluxes and increased supply of reducing equivalents rather than primary inputs of electrons from the PET. Interestingly, re-oxidation in the dark took several hours, which probably also reflects metabolic shifts and gradual diminishing electron supply from these sources. In higher plants, high light or application of PET-inhibitors cause increased production of H₂O₂ in chloroplasts, from where it is at least in part transferred to the cytosol and may act as a signal to the cytosol and the nucleus (Figure 1) (Exposito-Rodriguez et al. 2017; Ugalde et al. 2020a). These recent observations indicate how quantitative time-lapse imaging may provide new insights into chloroplast redox physiology and enable exploration of the battle of different reductive and oxidative forces in the future.

Figure 1:

Schematic illustration of antagonistic reducing and oxidizing powers acting on proteins with thiol moieties and low-molecular weight thiol metabolites.

In chloroplasts, the redox status of protein Cys and metabolites is dominated by inputs of electrons and H₂O₂ from photosynthesis. Stromal metabolism has only a minor impact during active photosynthesis. Compared with other compartments, photosynthetic electron transport takes the dominating role for supply of reducing capacity in chloroplasts. This makes the stroma the major ‘redox battle ground’ with respect to electron flux through the Cys-based redox systems in the cell. Thus, sophisticated means are essential to maintain balance between the processes involved, such that appropriate Cys-based redox switching and signaling remains possible. Schematic Cys-based redox switch illustrates both proteins as well as glutathione. While Cys-based redox dynamics in the stroma are separated from those in the cytosol, export of NAD(P)H-based redox equivalents, e.g., via the triose-phosphate and the malate-oxaloacetate shuttles (dashed grey arrow) and release of surplus H₂O₂ (solid grey arrow) from the chloroplasts also impact on redox metabolism in cell compartments beyond the chloroplast.

Paradigm shift 3: the specificity of TRX arises from the microenvironment and microcompartmentation around the catalytic thiol

Redox regulation relies on conditional reduction and oxidation of process regulators, i.e., proteins that contain cysteine thiols that are targets of the redox regulation machineries including redox transmitters (Dietz 2008). Redox transmitters operate the Cys-based switches in the redox regulatory network. Angiosperms fascinate redox researchers by their unique gene family complexity of redox transmitters, in particular TRXs, certain glutaredoxins (GRXs) and protein disulfide isomerases (PDIs) (Chibani et al. 2009; Geigenberger et al. 2017). The best studied example of TRX complexity is the chloroplast with 20 TRXs and TRX-like proteins in A. thaliana. As outlined by Geigenberger et al. (2017) the simultaneous presence of FTR and NADPH-dependent thioredoxin reductase c with a fused TRX domain (NTRC) enables the coordination of photosynthetic light reactions and associated anabolic pathways and the complementary catabolic or dark reactions. However, this qualitative assignment does not explain the highly evolved and diversified setup of the stromal TRX network.

To give an example, chloroplast TRX-y1 was initially associated with the regeneration of peroxiredoxin Q as part of the antioxidant defense (Collin et al. 2004). More recently, biochemical analyses linked TRX-y1 with other chloroplast processes like starch synthesis, chlorophyll biosynthesis, oxidative pentose phosphate cycle and redox regulation of β-carbonic anhydrase (βCA) (Dreyer et al. 2020; Nee et al. 2009; Valerio et al. 2011; Yoshida and Hisabori 2016). In general, nearly each study comparing different chloroplast TRXs for their efficiency to reduce or regulate target proteins reported certain redundancy among different TRX isoforms. In other cases, studies on specific donor-acceptor pairs often suffer from the drawback that only few TRXs were tested (for overview see Geigenberger et al. (2017)), possible effectors were missing and tested concentrations were low in comparison to in vivo conditions.

At present, we encounter a paradigm shift in understanding the redox network by incorporating information on structural properties and electrostatic interfaces of TRXs (Berndt et al. 2015; Gellert et al. 2019). The authors of the former study compared and clustered 35 TRX-fold domains in 33 proteins. The results reveal electrostatic features on the protein surface rather than structural properties as determinants of interaction preference of the thiol-disulfide exchange proteins with target proteins. Complementary interfaces have been suggested in earlier work, e.g., based on redox-independent binding of target proteins to TRX (Balmer et al. 2004).

Despite the distinct interface properties, promiscuity often remains, as indicated in in vitro work and in genetic studies with single or multiple deletions of specific TRX without apparent phenotypic variation. Many studies have addressed this issue, as illustrated by a recent example: the βCA1 is reduced and activated by chloroplast TRXs in vitro and in the absence of the respective other TRXs in the efficiency order TRX-y1>TRX-y2>TRX-f1>TRX-f2=TRX-m1=NTRC=TRX-m4>TRX-m2=TRX-m3=TRX-x>TRX-like CDSP32 (Dreyer et al. 2020). Only CDSP32 was ineffective as a reductant to βCA1 relative to the control, while 10 other TRXs were able to donate electrons to βCA1. The electrostatic surfaces of these TRXs are shown in Figure 2 and reveal neutral or positive surfaces in the vicinity of the catalytic thiols of those TRXs, which efficiently reduce βCA1. Similar promiscuity is known from many TRX / target protein pairs (Geigenberger et al. 2017). Based on such results, one may argue why plant compartments contain these large sets of TRXs. Two interpretations are at hand; (i) small kinetic differences in electron transfer efficiency as determined in vitro may scale into significant differences in an in vivo situation where the different TRXs compete for targets and/or (ii) interactions with additional factors or the build-up of regulatory units accentuate these differences in vivo. There is the further, more trivial possibility of conditional and tissue-specific expression of specific isoforms, but even if that is the case for some TRXs, this leaves a sufficiently large number of TRXs for the basic problem to persist.

Figure 2:

Electrostatic surfaces of the different chloroplastic TRXs and their targets βCA1 and FBPase.

The TRXs are ordered from best reductant (TRX-y1) to inefficient reductant (CDSP). Neutral and positive potentials ease reduction of β-carbonic anyhydrase 1 (βCA1) and positive potentials reduction of fructose-1,6-bisphosphatase (FBPase). The predicted protein structures were obtained from swissmodel.expasy.org, except of the βCA1 structure. The dimeric βCA1 structure derives from the octameric structure of βCA from Pisum sativum (pdb: 1ekj). The structures were analyzed using PyMOL 2.4 together with the ABPS/pdb2pqr plugin (Baker et al. 2001; Dolinsky et al. 2004, 2007; Schrödinger 2015). The regions of the redox-sensitive cysteines are within the black circles. Blue: positive, red: negative potential.

Three examples of a persistent binding of TRX to target complexes are the discovery of TRX-z as part of the chloroplast transcriptional machinery (Pfalz et al. 2006), the binding of Escherichia coli TRX to the TRX-binding domain of bacteriophage T7 DNA polymerase increasing the processivity of the polymerase more than 100-fold (Ghosh et al. 2008) and the identification of human TRX as RNA-binding protein (Castello et al. 2016). As such, option (ii) has the potential for another paradigm shift (Figure 3) and is analogous to recent observations of conditional association of regulatory supramolecular assemblies. Examples concern the association of scaffolding proteins, signaling proteins (e.g., RBOHs), transporter proteins (e.g., Ca²⁺ channels) and specificity factors in membranes or the reversible formation of regulatory complexes with chloroplast protein 12 (CP12) together with ribulose-5-phosphate kinase and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (Fuglsang et al. 2007; Han et al. 2019; Oesterhelt et al. 2007; Wedel et al. 1997).

Figure 3:

Two scenarios for thioredoxin (TRX) function in redox signaling.

(A) Reduced TRX binds to target protein in a short-lived enzymatic reaction. Specificity is determined by small complementary interfaces (electrostatic properties, red and blue) and therefore not exclusive. (B) Extended complementary interfaces of TRXs and additional factors (green) and the target protein facilitate the assembly of regulatory modules. This process increases the specificity at the cost of decreased turnover.

The underlying hypothesis is that the various TRXs do not exclusively function as disulfide reductases by short-lived bimolecular interactions with target proteins but, instead, build up regulatory modules and this involves additional facilitators and components. Demonstration of TRX-f1 binding to 2-Cys peroxiredoxin (2-CysPRX) in overlay assays indicates persistent association between both proteins (Liebthal et al. 2020). A significant overlap in proteomic studies was revealed between the set of polypeptides binding to 2-CysPRX (Cerveau et al. 2016; Liebthal et al. 2020) and the polypeptide set recognized as TRX partners (Schürmann and Buchanan 2008). This observation may indicate the existence of ternary or higher order complexes of target proteins, TRX, 2-CysPRX and target proteins.

Complex formation between components of the redox regulatory network may increase the specificity of regulation in time and space and may explain the requirement for more extended families of involved proteins. The formation of such regulatory assemblies may be suppressed under unstressed conditions by distinct compartmentation. N-myristoylation or S-palmitoylation associates certain TRX-h isoforms to cell membranes (Traverso et al. 2013). Upon cleavage of the fatty acyl residue, TRX-h may relocate to other compartments. Serrato and Cejudo (2003) reported TRX-h accumulation in the nucleus of wheat seeds under oxidative stress. Such a dynamic formation of regulatory units should be taken into the focus of research, bringing together redox biochemistry and functional cell biology approaches. There is a need to search for Cys-redox microdomains that can be stabilized by structural arrangements shielding them from other potential interaction partners and allowing heterogeneity and specificity of redox regulation.

Paradigm shift 4: thiol peroxidases link ROS to functional redox signaling

ROS-sensitive cell membrane-permeable reversible fluorescent dyes (Prasad et al. 2019) and genetically encoded probes like roGFP2-Orp1 and HyPer (Bilan and Belousov 2018; Meyer and Dick 2010; Schwarzländer et al. 2016) provide experimental access to changes in ROS dynamics. They give reliable and important insight into transient changes in ROS fluxes, e.g., induced by wounding and addition of elicitors of pathogens (Nietzel et al. 2019; Prasad et al. 2019). Absolute ROS concentrations cannot be determined because the sensors monitor the balance between oxidation and re-reduction. The full potential of genetically encoded sensors to detect local redox changes is limited by the exact kinetic properties of the probes, the match of their sensitivity to the physiological changes and their response kinetics. Hence, a given biological question requires careful selection of the most suitable biosensor and in many cases such a sensor is still not available. Some limitations with respect to sensitivity and kinetics may be overcome through placing the sensor in close proximity to the process of interest (e.g., a specific Cys-based redox switch), subcellular targeting, fusion with specific proteins and/or immobilization of probes to membranes or membrane subdomains by membrane anchoring. Further, sensitivity and responsive kinetics can be altered through mutational approaches of established sensor proteins (Aller et al. 2013; Steinbeck et al. 2020).

In mathematical terms, a single measuring parameter cannot solve an equation with several variables, namely the H₂O₂ concentration and the re-reduction rate of oxidized roGFP2-Orp1 by various redox transmitters like TRXs and GRXs [discussed at depth in Schwarzländer et al. (2016)]. Nietzel et al. (2019) showed different reduction efficiencies of oxidized roGFP2-Orp1 by (i) TRX-o1 coupled to NADPH-dependent thioredoxin reductase B (NTRB) and NADPH, (ii) glutathione, GR and NADPH and (iii) GRX-C1, glutathione, GR and NADPH. In fact, this kinetic probe cannot be calibrated in vitro for its response in vivo due to the impossibility of fully reconstituting the system in vitro.

The redox adjustment of this sensor is a good example of the fact that the redox state of proteins of interest always needs to be considered in the context of oxidation and reduction rates. The oxidation state of the protein Cys increases upon stimulated ROS-dependent electron drainage or upon decreased electron provision by the reductive pathway(s). Cys-based redox switches function as kinetic sensors. Oxidation of sensitive thiols by O₂ should also be considered. However, reliable chemical reaction constants still need to be determined. Available data suggest that low concentrations of H₂O₂ (30 nM as discussed below) and O₂ [in equilibrium with air of 21% v/v (Bagiyan 2003)] may equally contribute to thiolate oxidation [reviewed in Vogelsang and Dietz (2020)].

In the case of the photosynthesizing chloroplast, PET continuously generates ROS in the light as outlined above. The electron flow from the PET into the reductive branch of the redox network and the continuous drainage of electrons by the oxidizing branch determine the redox state and thus activity of the redox target proteins. Enzymes of the Calvin–Benson-Cycle are the most prominent target proteins, which are tuned dependent on the environmental conditions for photosynthesis. The chloroplast 2-CysPRX and PRXQ are kept in a partly oxidized state by reacting with H₂O₂. O₂ may contribute to the slow oxidation of thiols, e.g., in TRXs [reviewed in Vogelsang and Dietz (2020)]. The PRXs participate in oxidizing TRXs which in turn oxidize the target protein, e.g., upon darkening the cells or a drop in light intensity (Ojeda et al. 2018; Telman et al. 2020; Vaseghi et al. 2018; Yoshida et al. 2018).

The rapid responsiveness of the chloroplast redox system is linked to the balance of continuously active oxidizing and reducing pathways. This mechanism “tensions” the regulatory “spring”. Importantly, the relative sensitivity of redox target proteins for reduction and oxidation is strongly affected by the presence of metabolites (Knuesting and Scheibe 2018), e.g., TRX-f1 efficiently activates chloroplast fructose-1,6-bisphosphatase (FBPase) in the presence of its substrate fructose-1,6,-bisphosphate (Wolosiuk et al. 1980). This interrelation with metabolic regulators adds an additional level of fine-tuning to redox regulation. ROS at low concentrations are likely to exist in most cell compartments. To estimate the electron fluxes through the redox regulatory network, it would be important to know the rates of ROS generation in a particular cell compartment and ROS influx from neighboring compartments rather than the ROS concentration. Experimental approaches to this challenging question are unavailable at present. Case studies suggest that chloroplast 2-CysPRX, PRXQ and possibly other thiol peroxidases are essential, allocating an important house-keeping function to H₂O₂ and suggesting that H₂O₂ needs to be actively generated and be available at all times.

Paradigm shift 5: synergism, antagonism and reversibility are part of the flexibility in the redox regulatory network

Angiosperms express a particularly elaborated Cys-based redox regulatory system. Chibani et al. (2009) compiled the gene families encoding TRXs and TRX reductases of five angiosperms and compared them with P. patens, a moss, two algae and two cyanobacteria. The TRX number encoded in the genome increases from four in Synechocystis sp. PCC 6803 to 43 in poplar (Chibani et al. 2009). Recent work also illustrates the functional diversification of TRXs. The biochemical in vitro and the genetic in vivo data suggest that certain TRXs are more likely to function in the reducing pathway, while others, like TRX-like2 isoforms and Lilium-TRX (atypical chloroplast TRX: ACHT), participate in the oxidizing pathway (Dangoor et al. 2009; Yoshida et al. 2018).

The discussion of possible dithiol-disulfide transfer reactions often centers on redox potentials based on thermodynamic considerations. A redox pair A with more negative midpoint potential donates electrons to a redox pair B with less negative potential and in equilibrium, A will be more oxidized and B more reduced (Segel 1975). However, this assumption needs refinement if we consider fluxes and redox transmitters like TRXs. In this case, the decisive thermodynamic boundary condition is the thermodynamic difference between the redox input elements, e.g., NADPH/NTR, and the final redox target protein, while the midpoint redox potential of the redox transmitter is less critical. Rather the electrostatic interface (Berndt et al. 2015; Gellert et al. 2019) and efficiency of transfer, possibly facilitated in a regulatory complex, will determine the kinetics of target protein reduction or oxidation.

The complexity of the network with multiple electron transfer pathways questions the expectation that genetic approaches with single or multiple deletions, or overexpression of redox network components can provide full understanding. To obtain a detailed understanding of the network performance, there is a need to determine kinetic constants for more reactions, to reconstitute subnetworks in vitro, to challenge their redox response and to simulate the network performance by mathematical modeling.

Gerken et al. (2020) developed a mathematical model of a chloroplast subnetwork in order to simulate the electron flow via NADPH/NTRC or FTR/TRX to 2-CysPRX, FBPase for activation and deactivation and H₂O₂. The H₂O₂ concentration-dependent simulation of the system allowed for estimating experimentally inaccessible variables. The resting H₂O₂ concentration in the stroma was estimated to be about 30 nM in the chloroplast. Electron flow into metabolism exceeded the electron flow into redox regulation of FBPase about 7000-fold.

The model enabled assessment of the synergism between the NTRC- and TRX-f1-dependent reduction of 2-CysPRX, with the former being 5.3-times more efficient than the latter (Gerken et al. 2020). The model also illustrated how the NTRC systems and the TRX systems may act in synergy, e.g., if high photosynthetically active radiation reduces both the NADPH and the ferredoxin system. But antagonistic effects are realized in darkness or upon lowering the light intensity (Figure 4). Our understanding of redox network complexity is only fragmentary and the inclusion of the less explored TRXs like TRX-like and ACHT will further enhance the opportunities for flexible adjustment in dependence on environmental conditions controlling photosynthesis.

Figure 4:

Synergism and antagonism in redox regulatory electron flow.

(A) In vitro regulation of fructose-1,6-bisphosphatase (FBPase). In an enzyme assay, low concentrations of dithiothreitol in conjunction with TRX-f activate FBPase in the presence of its substrate FBP. Addition of excess amounts of oxidized 2-CysPRX rapidly inactivates FBPase in vitro in the presence of TRX-f (Vaseghi et al. 2018). (B) Redox regulatory network of the chloroplast comprising the electron flow from photosynthetic electron transport chain (PET) to NADPH and metabolism, NADPH-dependent thioredoxin (TRX) reductase C (NTRC) and 2-CysPRX and as second branch ferredoxin (FD)-dependent TRX reductase (FTR), TRX-f and FBPase. In the light there is sufficient electron pressure to reduce FBPase despite continuous electron drainage by PRXs. Upon darkening, 2-CysPRX rapidly drains electrons from TRX-f, which in turn oxidizes the regulatory Cys of FBPase or other targets. Other TRXs such as TRX-like and Lilium-TRX contributes to the regulatory electron flow. The central part of this network was mathematically simulated by Gerken et al. (2020) using the available and estimated velocity constants v1-v10. Additions not being part of the simulation are PRXQ, TRX-like and other targets.

Paradigm shift 6: refined subcellular localization outlines novel roles of plant glutathione peroxidase-like proteins

Removal of H₂O₂ and fatty acid hydroperoxides in mammalian cells has long been attributed at least in part to reduced glutathione (GSH) and a peroxidatic function of glutathione peroxidases (Gpxs) (Mills 1957). Later research showed that this peroxidatic function strictly depends on supply with selenium because the respective Gpxs contain a selenocysteine (SeC) in their catalytic side (Flohe et al. 1973; Rotruck et al. 1973). While four out of eight isoforms in the human Gpx family are SeC enzymes, the other four enzymes contain Cys instead of SeC (Brigelius-Flohe and Maiorino 2013). Like other non-mammalian model species such as Saccharomyces cerevisiae or Trypanosoma brucei, plants contain several members of the Gpx protein family, however, none of these proteins contain SeC but rather redox-active Cys in the active site (Figure 5) (Inoue et al. 1999; Melchers et al. 2008). These non-SeC enzymes, including their plant orthologues, have been functionally identified as TRX-dependent peroxidases (Iqbal et al. 2006; Lubos et al. 2011; Navrot et al. 2006). Based on substrate preferences and the respective electron donor, these non-SeC enzymes have also been referred to as phospholipid hydroperoxide GPXs, TRX peroxidases or GPX-type enzymes (Herbette et al. 2002; Jung et al. 2002; Maiorino et al. 2015; Schlecker et al. 2005). To avoid misleading assumptions about their catalytic function and the respective electron donor while still emphasizing the structural homology, the name GPX-like (GPXL) has been proposed for the eight A. thaliana proteins (Attacha et al. 2017). Although the exact substrate specificity of A. thaliana GPXLs remains to be elucidated, detailed analysis of the subcellular localization of A. thaliana GPXLs recently paved the way to further discoveries in Cys-based redox biology of plants. In contrast to bioinformatic predictions for mitochondrial localization, GPXL3 was found to be a membrane-anchored luminal protein in the secretory pathway, and GPXL4 and GPXL5 are both anchored to the plasma membrane with an N-terminal myristoyl-residue (Attacha et al. 2017). Putative physiological implications are outlined in Paradigm shifts 7 and 8.

Figure 5:

Models depicting the catalytic cycle of different members of the glutathione peroxidase family.

Hydrogen peroxide and organic hydroperoxides (ROOH) are degraded in part by different types of peroxidases of the entire glutathione peroxidase (Gpx) superfamily, which is represented in all kingdoms. While in mammalian cells, classical selenocysteine (SeC) Gpx receives electrons from GSH (panel A), plant GPXLs are not reduced by GSH but use thioredoxin (TRX) as electron donors instead (panel B). During the catalytic cycle, two of the three cysteines (C) in the active site of GPXLs form a disulfide bridge that is subsequently reduced by TRX.

True GPX activity with GSH as electron donor for reduction of H₂O₂ or lipid hydroperoxides in plants is performed by some glutathione S-transferases (GSTs) (Dixon et al. 2009; Sylvestre-Gonon et al. 2019). Specifically, members of the GST subfamilies phi (F), tau (U) and theta (T) accept peroxides and reduce them to the respective alcohols with two GSH molecules as electron donors (Edwards et al. 2000). Several members of the GSTU and GSTF subfamilies are strongly induced under lipid stress responses, which suggest a role in removal of toxic intermediates resulting from lipid peroxidation (Mueller et al. 2008). Pronounced catalytic activities of several GSTF isoforms with organic peroxides as substrates have been shown (Cummins et al. 2013; Pégeot et al. 2017). It should be noted though, that this activity might be several orders of magnitude lower than for some other thiol peroxidases (Pégeot et al. 2017). Recent data suggest that the cytosolic GSTU7 is involved in mitigation of oxidative stress resulting from methyl viologen-induced O₂⁻ formation in plastids (Ugalde et al. 2020b). Interestingly, GSTU7 also has a pronounced effect on root growth but the detailed mechanism of its function still needs to be elucidated.

Paradigm shift 7: ferroptosis in plant redox research cannot just follow the mammalian blueprint

Exposure of plants to abiotic and biotic stresses generally triggers the activity of RBOHs in the plasma membrane to generate O₂⁻ in the apoplast (Baxter et al. 2014; Mittler 2017). O₂⁻ is then converted to H₂O₂, which may enter the cytosol through aquaporins (plasma membrane intrinsic proteins (PIPs)) (Rodrigues et al. 2017) to trigger signaling cascades. In addition, H₂O₂ may participate in a Fenton reaction with ferrous iron (Fe²⁺) leading to formation of hydroxyl radicals (OH^•). Lack of appropriate immunity in mutants compromised in H₂O₂ production or deficient in iron suggests that the Fenton reaction acts as a key step in plant immune responses similar to iron-dependent cell death processes in mammalian cells (Herlihy et al. 2020).

Ferroptosis is a regulated form of cell death driven by peroxidation of phospholipids (Stockwell et al. 2020). Hydroxyl radicals resulting from the Fenton reaction can oxidize membrane phospholipid-associated polyunsaturated fatty acids (PL-PUFAs) to form PL-hydroperoxides (PL-OOH) (Figure 6). If not removed immediately, PL-OOH can undergo fragmentation to form highly toxic lipid radicals and other reactive lipid breakdown products (Schneider et al. 2008). Non-enzymatic systems preventing uncontrolled accumulation of PL-OOH include chelators of iron and lipophilic radical-trapping antioxidants, especially ɑ-and ɣ-tocopherol (Kanwischer et al. 2005; Stahl et al. 2019; Stockwell and Jiang 2020). In contrast to the primary function of tocopherols at the plasma membrane in mammalian cells, their protective function in plants, however, is primarily exerted at the chloroplast membrane (Maeda and DellaPenna 2007). The key factor in preventing ferroptosis in mammalian cells is the SeC-containing glutathione peroxidase 4 (Gpx4), which reduces PL-OOH to the corresponding alcohol with electrons provided by GSH (Yang et al. 2014).

Figure 6:

Model for iron- and H₂O₂-dependent lipid peroxidation resulting in ferroptosis in plants.

Peroxidation of phospholipid polyunsaturated fatty acids (PL-PUFA) leads to PL-PUFA hydroperoxides (PL-PUFA-OOH) which, if not efficiently removed, trigger ferroptosis. Accumulation of PL-PUFA-OOH is presumably prevented by glutathione peroxidase-like enzymes GPXL4 and 5, which are both attached to the plasma membrane with a myristoyl residue. Electrons for peroxide reduction are likely to be provided by NADPH through the TRX/NTR system.

Ferroptosis-like cell death was recently also described for heat-shock induced death of root hairs in A. thaliana (Distefano et al. 2017) and as part of the hypersensitive response during interactions between rice and the rice blast fungus Magnaporthe oryzae (Dangol et al. 2019). A 55° C heat-shock caused morphological changes in mitochondria that mimic those observed in cancer cells undergoing ferroptosis (Distefano et al. 2017). In both heat-shocked root hairs and the hypersensitive response, models for interpretation of the results show dependence of ferroptosis on iron and glutathione and include one or more Gpx4 orthologues. Inhibition of these peroxidases or depletion of GSH is suggested to trigger ferroptosis (Conlon and Dixon 2017; Dangol et al. 2019). It is important to consider, however, that, in contrast to HsGpx4, the non-SeC plant GPXLs are not reduced by GSH but rather employ TRXs as electron donor (see Paradigm shift 6). This implies that in plants, ferroptosis depends on TRXs and NTRs rather than electrons provided by GSH (Figure 6). Nevertheless, heat-induced death of root hairs was suppressed by external supply of GSH and vice versa increased by pharmacological inhibition of GSH biosynthesis (Distefano et al. 2017). One possible interpretation for this observation is the functional redundancy between the cytosolic glutathione redox system and the cytosolic TRX redox system (Marty et al. 2009; Reichheld et al. 2007). Alternatively, it can also be envisaged that GSTs may contribute to the removal of PL-OOH at the expense of GSH. With the outlined inconsistencies in current models, it is obvious that the putative roles of GPXLs as well as GSH and GSTs in ferroptosis in plants await further investigation.

In A. thaliana, GPXL4 and GPXL5 might be particularly good candidates for a functional role in ferroptosis because of their localization at the plasma membrane (Attacha et al. 2017). Increased amounts of the lipid peroxidation marker malondialdehyde in shoots of gpxl4 and roots of gpxl5 mutants exposed to osmotic stress (Bela et al. 2018) tentatively support their presumed role in efficient removal of PL-OOH. It is, however, not clear, from where GPXL4 and GPXL5 retrieve electrons for substrate reduction. Nominally, this could be any cytosolic TRX together with NTRA/B. Interestingly two subgroups of cytosolic h-type TRXs show N-terminal lipidation, which may enable their positioning at membranes (Traverso et al. 2013). While isoforms of subgroup 3 are both myristoylated and palmitoylated and appear at the plasma membrane, isoforms of subgroup 2, which are only myristoylated, were reported to reside on endomembranes when expressed in onion (Allium cepa) cells. Of particular interest might be TRX-h2, which belongs to subgroup 2 and is myristoylated in the same way as GPXL4 and 5. It seems reasonable to speculate that the protein may also localize to the plasma membrane in A. thaliana where it indeed has been found in proteome studies (Benschop et al. 2007; Elmore et al. 2012). This might bring TRX-h2 and other TRX isoforms in close proximity to GPXL4 and GPXL5 and thus potentially create a complete membrane–bound degradation system for PL-OOH at the site of their formation, which may increase the efficiency of the removal of these deleterious metabolites to suppress ferroptosis.

In contrast to other organisms, tocopherols in plants are exclusively located in plastids where they are being synthesized. Thus, tocopherols do not contribute to detoxification and scavenging at the plasma membrane, which further emphasizes the lead role of enzymatic detoxification of PL-OOH.

Paradigm shift 8: oxidative protein folding in the ER and implications for phytohormone signaling

Formation of structural disulfides is essential for many ER-resident proteins and for proteins passing the ER on the way to their final destination [reviewed in Meyer et al. (2019)]. De novo disulfide formation by ER thiol oxidases (EROs) is followed by transfer of disulfides to nascent proteins through a cascade of thiol-disulfide exchange reactions involving several disulfides in ERO and PDIs. Despite largely homologous sequences, the two A. thaliana isoforms ERO1 and ERO2 are present in distinct redox states indicating that they might be regulated differently and possess different activities. Nevertheless, both proteins are required for oxidative protein folding (Fan et al. 2019).

PDIs of the PDI-L subfamilies are characterized by their consensus domain structure a-b-b′-a′ in which a and a′ represent active TRX motifs while b and b′ are inactive TRX domains (Selles et al. 2011). In addition, several other PDIs with 1–3 TRX domains are present. The large number of PDIs and their structural diversity may reflect substrate specificities and specific functions. In addition, PDIs may work cooperatively to oxidize specific target proteins (Matsusaki et al. 2016).

The net transfer of two electrons leads to the formation of stoichiometric amounts of H₂O₂ as a potentially toxic by-product that demands appropriate detoxification mechanisms. In human cells, H₂O₂ produced by EROs has been shown to be detoxified by Gpx8 and at higher concentrations also by a luminal Prx (Ramming et al. 2014; Tavender and Bulleid 2010; Zito et al. 2010). If Gpx8 is missing, H₂O₂ may leak into the cytosol as a signal or to exhibit toxic effects (Appenzeller-Herzog et al. 2016; Ramming et al. 2014). In plants, the ER does not contain a luminal PRX (Dietz et al. 2006), but the recent identification of GPXL3 as a luminal protein that is anchored to the ER and Golgi membranes (Attacha et al. 2017; Nikolovski et al. 2012) indicates that H₂O₂ generated during oxidative protein folding may be degraded locally (Figure 7). In the absence of a functional NTR/TRX system, which is presumed to reduce cytosolic and mitochondrial GPXLs (see Paradigm shift 6), PDIs with functional TRX domains are the most likely candidates for reduction of GPXL3. PDI oxidized in this step may then again contribute the disulfide to the oxidation of thiols on nascent proteins entering the ER (Figure 7). The coupling of two pathways for oxidation of PDI via ERO and GPXL3 would double the efficiency of the oxidative protein folding machinery counted as disulfides formed per molecule of used O₂ compared to oxidation by ERO on its own.

Figure 7:

Pathways for oxidative protein folding in the endoplasmic reticulum.

After entering the ER, free thiols in nascent proteins get oxidized by protein disulfide isomerases (PDIs). PDIs are subsequently re-oxidized by ER thiol oxidase (ERO), which transfers the electrons further to O₂. H₂O₂ produced during this process is further reduced to H₂O by GPXL3, which in turn might also be reduced by PDIs. In this way, four electrons originating from the formation of two disulfide bridges on nascent proteins can be transferred to O₂. Targets of disulfide formation include, amongst others, secreted rapid alkalinization factor (RALF), the brassinosteroid receptor (BRI1) and the ethylene receptor (ETR1). Furthermore, ERO itself may be a target to modulate its activity. GSH entering the ER along its concentration gradient through the translocon SEC61 may resolve some disulfides, particularly thermodynamically less stable, incorrect disulfides. The resulting glutathionylated intermediate becomes a substrate for luminal GRXs that deglutathionylate the respective Cys and thus feed the partially folded protein back to PDIs. For the GSSG formed in this step the fate is not clear. It could either be exported to the cytosol for reduction or, alternatively, also oxidize PDIs (not shown). It should also be noted that depicted reactions formally are equilibrium reactions and may well run backwards if changes in reductive and oxidative power occur.

In yeast, under conditions of stress, GSH may enter the ER along its concentration gradient through the translocon Sec61 (Ponsero et al. 2017). Whether the same mechanism applies to plants and whether additional pathways for transport of GSH into the ER exist is not clear at this point. Once inside the ER, GSH may attack disulfides formed in nascent peptides and thus provide a force acting against disulfide formation. In particular, non-native and thermodynamically less stable protein disulfides are prone to non-catalyzed, thermodynamically driven attack of GSH. Glutathionylated peptides may re-enter the folding pathway after removal of the glutathione moiety. Deglutathionylation of proteins can be mediated by dithiol GRXs (Trnka et al. 2020). In the A. thaliana ER proteome, two dithiol GRXs have been identified (Nikolovski et al. 2012) but still await functional characterization. Whether the resulting glutathione disulfide (GSSG) is exported from the ER or whether it also contributes to oxidation of PDIs, or even to direct target protein oxidation mediated by GRXs, is not yet known. Thermodynamically, the GRX-mediated reactions are bound to run in reverse if luminal GSSG concentration increases too much. In the presence of GRXs, GSSG could thus also contribute to disulfide formation on nascent proteins. With these considerations, it becomes clear how closely oxidative protein folding in the ER is associated with other physiological processes in the cell. If GSH can freely enter the ER along a concentration gradient, the folding machinery will constantly need to work against the influx of reductants setting up another ‘redox battle ground’ in the cell where major oxidizing and reducing fluxes collide (see Paradigm shift 2). If the fine-tuned system is overloaded with GSH, the folding machinery may eventually be exhausted. Plants with elevated GSH have been generated through overexpression of GSH biosynthetic enzymes but frequently displayed phenotypic aberrations like early senescence and lower biomass [for review see Noctor et al. (2011)]. A higher GSH concentration and permanently increased flux through the biosynthetic pathway may impose continuous reductive stress on the oxidative folding machinery in the ER, which may contribute to the observed phenotypes. Deleterious effects of increased amounts of low-molecular weight thiols are particularly evident in gsh2 mutants that hyperaccumulate the GSH precursor γ-EC and in consequence develop a severe ER phenotype (Au et al. 2012).

Strict dependence of ERO on supply with O₂ renders oxidative protein folding in the ER sensitive to hypoxia with diminished O₂ supply. With accumulating unfolded proteins under hypoxia the ER might thus act as low oxygen sensor as has been proposed for mammalian cells (Wouters and Koritzinsky 2008). Efficient O₂ sensing depends on the respective K_m values of ERO for O₂ (Schmidt et al. 2018). While these values are not yet known for the plant EROs, one can only infer the possibility from K_m values of 4–10 µM for yeast ERO (Gross et al. 2006; Schmidt et al. 2018).

Loss of ERO function, either through lack of O₂ or mutation, is expected to cause pleiotropic effects on multiple downstream targets that rely on disulfide formation. Besides storage proteins (Onda et al. 2009), these targets include Cys-rich peptides (CRPs), the ethylene receptor ETR1 and leucine-rich repeat (LRR) receptor-like kinases (RLKs). CRPs constitute a large protein family with more than 100 members and are involved in multiple developmental and physiological processes (Marshall et al. 2011; Murphy and De Smet 2014). A well characterized representative of this protein family is the rapid alkalinization factor 1 (RALF1) which, after formation of two internal disulfides and secretion to the apoplast, interacts with the RLK feronia (FER) (Haruta et al. 2014; Pearce et al. 2001). After binding of RALF, the FER kinase domain is activated and phosphorylates the H⁺-ATPase AHA2. Multiple further interactions for RALF-FER with other hormone signaling pathways have been reported, including abscisic acid (Chen et al. 2016), auxin (Barbez et al. 2017), jasmonic acid (Guo et al. 2018), brassinosteroids (Deslauriers and Larsen 2010) and ethylene (Mao et al. 2015). Ethylene- and brassinosteroid signaling stand out, because the required receptors contain disulfide bridges that need to be established in the ER. The ethylene receptor ETR1 is active as a homodimer in which the subunits are linked by two disulfides (Schaller et al. 1995). The brassinosteroid receptor BRI1 contains five disulfides in its ectodomain to link consecutive LRR-domains plus further two disulfides to stabilize the N- and C-terminal caps (Hothorn et al. 2011). Mutation of a single Cys is sufficient to render BRI1 non-functional. In the corresponding bri1-5 mutant line the protein is retained in the ER and degraded, illustrating how essential correct oxidative protein folding is for proper receptor maturation and localization (Hong et al. 2008).

Paradigm shift 9: the apoplast as an emerging playing field for key redox signaling processes

Production of ROS at the cell surface in the form of an oxidative burst is recognized as critical for pathogen defense responses (Hammond-Kosack and Jones 1996), developmental processes (Foreman et al. 2003; Fujita et al. 2020; Orman-Ligeza et al. 2016), guard cell dynamics (Pei et al. 2000; Kwak et al. 2003) and long distance signaling (Fichman and Mittler 2020). Key proteins are respiratory burst oxidase homologues (RBOHs), which produce O₂⁻ by transferring electrons from cytosolic NADPH across the plasma membrane to O₂ at the extracellular membrane face (Figure 8). In addition to several RBOH isoforms [10 in A. thaliana, RBOHA-RBOHJ (Suzuki et al. 2011)], plants also contain a large number of class III peroxidases (POXs) in the apoplast some of which have been shown to contribute to the oxidative burst (Daudi et al. 2012). Furthermore, PAOs may contribute to apoplastic ROS production on the condition of sufficient substrate supply (Moschou et al. 2008). Quiescin sulfhydryl oxidases (QSOXs) are thiol oxidases that are likely to contribute to oxidative protein folding in the secretory pathway, especially in the Golgi (Aller and Meyer 2013). The appearance of QSOX in the plasma membrane proteome (Mitra et al. 2009) raises the question of whether and to what extent QSOX may also attack extracellular thiols similar to its function in fibroblasts (Ilani et al. 2013) and contribute to H₂O₂ generation in the apoplast. Under osmotic stress, ascorbate-mediated reduction of Fe²⁺ and subsequent re-oxidation to Fe³⁺ may contribute to apoplastic O₂⁻ formation (Martiniere et al. 2019). RBOHs are not constitutively active but need a trigger to be activated either through binding of Ca²⁺ to a cytosolic EF-hand or through phosphorylation by kinases, prominently of the Ca²⁺-dependent protein kinase (CDPK) family (Oda et al. 2010; Ogasawara et al. 2008; Suzuki et al. 2011). A steadily growing number of upstream stimuli to trigger RBOH activity has been identified and added to well-established stimuli such as bacterial elicitors being perceived by LRR-RLKs (Gomez-Gomez et al. 1999) or osmolytes (Martiniere et al. 2019). Only recently another layer of control over RBOH activity was identified in the form of small GTPases of the Rho-of-plant (ROP) family that control RBOHD under waterlogging or during osmotic signaling (Smokvarska et al. 2020; Sun et al. 2019). A hyperosmotic stimulus triggers ROP6 to form nanodomains in which two RBOH cluster and subsequently produce O₂⁻ (Smokvarska et al. 2020).

Figure 8:

ROS and redox signaling processes in the apoplast.

Production of H₂O₂ in the apoplast results from multiple sources. NADPH oxidases (RBOHs) produce superoxide (O₂⁻), which is converted to H₂O₂ although it is not known whether this conversion occurs enzymatically or non-enzymatically. Other sources of H₂O₂ may include different apoplastic class III peroxidases (POX), polyamine oxidases (PAO) and possibly quiescin sulfhydryl oxidases (QSOX), which have also been found in the extracellular proteome. H₂O₂ may exhibit some signal function in the apoplast. Apoplastic H₂O₂ is known to activate Ca²⁺ channels and may also affect cysteine-rich-repeat receptor kinases (CRKs). Direct activation of the receptor-like kinase hydrogen-peroxided-induced Ca²⁺ increase 1 (HPCA1) by H₂O₂ depends on four exposed extracellular cysteines (red dots). A functional receptor for H₂O₂ in the apoplast, however, would require the presence of an efficient reducing power that could act against a high background of H₂O₂ mediated oxidation. Such a system is as yet unknown. For intracellular signaling H₂O₂ is admitted to the cytosol via aquaporins (PIP). Efficient production of extracellular O₂⁻ by RBOHs involves multiple regulatory inputs including Ca²⁺-binding, phosphorylation and also interaction with the small Rho-GTPase ROP6 that may trigger the formation of RBOH nanoclusters in nanodomains of the membrane. Solid arrows: confirmed reactions and inputs; dashed arrows: hypothetical reactions and signaling pathways not understood in their molecular details.

In contrast to intracellular compartments, to date no apoplastic superoxide dismutase (SOD) for enzymatic conversion of RBOH-derived O₂⁻ to H₂O₂ has been functionally confirmed. Two candidate genes possibly coding for apoplastic SODs in A. thaliana have been suggested (Waszczak et al. 2018). In the absence of an enzymatic catalyst apoplastic dismutation of O₂⁻ occurs spontaneously, albeit at rates that are five orders of magnitude slower. The resulting RBOH-dependent production of H₂O₂ is sufficient to induce so-called ROS-induced ROS waves in which the initial oxidation triggers a Ca²⁺-release in neighboring cells, which in turn again activates RBOH. Such waves may relay essential information over larger distances probably using the vasculature for signal propagation (Fichman et al. 2019; Xiong et al. 2020; Zandalinas and Mittler 2018). Whilst convincing evidence exists for H₂O₂ acting as a signal within the apoplast [reviewed in Waszczak et al. (2018)], several of these signaling functions rely on re-entry of H₂O₂ into the cytosol through PIPs. The oxidation resulting from this entry of H₂O₂ can be monitored dynamically with genetically encoded probes (Nietzel et al. 2019; Rodrigues et al. 2017). Despite plenty of evidence for H₂O₂ being a critical factor, the molecular mechanisms involved in downstream signaling triggered or mediated by extracellularly-derived H₂O₂, as well as the principles that confer specificity, are still far from being understood.

In addition to the LRR-RLKs highlighted above, the large family of RLKs with more than 400 members also contains a subgroup of Cys-rich repeat (CRR) RLKs (CRKs). These proteins also have the classical domain structure with a single transmembrane domain, an intracellular kinase and an extracellular ectodomain for signal perception. The CRKs are predicted to form disulfide bridges similar to the disulfides in LRRs. In contrast to LRRs, CRKs have been proposed as putative targets for redox modification in the apoplast (Bourdais et al. 2015; Yadeta et al. 2017). It is not clear, however, whether these CRKs are primary components of H₂O₂ sensing or whether they are rather modulators of RBOH activity. Indeed, physical interaction of CRKs and RBOHs can be deduced from a membrane-linked interactome study (Jones et al. 2014). Interestingly, the same study also indicates direct interaction of CRK33 with GPXL5, which has not yet been investigated for its functional significance.

Most recently, the LRR-RK hydrogen-peroxide-induced Ca²⁺ increase 1 (HPCA1) has been reported as an apoplastic H₂O₂ sensor (Wu et al. 2020). HPCA1 consists of an intracellular kinase domain, a single transmembrane domain, a hydrogen–peroxide domain (HP) with four cysteines in two CxxC motifs, and the typical LRR ectodomain (Figure 8). HPCA1 is activated by extracellular H₂O₂ and mediates H₂O₂-induced activation of Ca²⁺-channels in guard cells as part of the signaling leading to stomatal closure. While firm genetic evidence for HPCA1 as part of the signaling leading to H₂O₂-induced intracellular Ca²⁺ elevations has been established, the proposed function as the primary H₂O₂ sensor raises several questions. The authors had to use 4 mM H₂O₂ to see an effect. Even if the apoplast is the compartment in which the highest steady-state concentrations of H₂O₂ can be maintained in plant cells (Waszczak et al. 2018), millimolar concentrations are very likely to be far beyond the physiologically meaningful range. It is also surprising that single Cys mutants of HPCA1 are retained in the ER, similar to the bri1-5 (see Paradigm shift 8). This observation pinpoints the need for disulfide formation by the respective Cys to attain an appropriate structure for presentation at the plasma membrane. Most notably, H₂O₂ perception in the apoplast by oxidation will require reduced Cys and any reversible response will require re-reduction. It presently remains unclear, however, what the reducing factor in the apoplast for maintaining reduced Cys may be, and how it may itself be reduced. GRXs have been found in the extracellular proteome (Nguyen-Kim et al. 2016) and could potentially contribute to reduction of disulfides. The GRX, however, would need GSH for efficient reduction, which only occurs at very low concentrations in the apoplast and adopts oxidizing redox potentials (Foyer and Noctor 2016), questioning its suitability as electron donor. As another possible candidate for reduction of HPCA1 a membrane-bound electron-transport system involved in reducing apoplastic ascorbate has been proposed (Foyer 2020) but awaits further characterization. In summary, recent research is raising the pressing question about the Cys-redox protein inventory and the machinery for mediating Cys-based redox switching in the plant extracellular matrix. It should also be noted that under acidic conditions in the apoplast, the formation of thiolates required for redox activity is largely repressed unless the respective thiols have exceptionally low pKa values. A major next step in exploring this new frontier in plant redox research will be to establish the source of reductive power as a prerequisite for active redox switching. Once this question is addressed there will be plenty of room for further discoveries and integrating apoplastic redox dynamics into the emerging frameworks of cellular redox regulation as well as whole plant signaling.

While this manuscript was under review, an independent study identified HPCA1 as a receptor for quinone. Based on this receptor function, the protein is now also called CAnnot Respond to DMBQ 1 (CARD1) (Laohavisit et al. 2020). Although the cysteine residues attributed to sensing of H₂O₂ did not appear in a genetic screen for quinone perception, there are leads pointing to their importance for quinone sensing as well. Whether perception of both quinones and H₂O₂ are simultaneously relevant in vivo and if so, whether both functions compete with, interfere with, enhance or modulate each other remains to be studied.

Paradigm shift 10: redox regulation of respiration occurs upstream of alternative oxidase in vivo

The alternative oxidase (AOX) at the inner mitochondrial membrane has been an icon of Cys-based redox regulation in plants beyond the chloroplast (Umbach and Siedow 1993; Umbach et al. 2006). AOX constitutes a shortcut for electron flow from the ubiquinone pool to bypass the complexes III and IV and proton pumping. As such it acts as a key mechanism of the remarkable flexibility of the RET in plants (Vanlerberghe 2013). AOX proteins are conserved across plants (McDonald 2008), indicating that ‘dumping’ excess electrons is critical for a sessile and photoautotrophic lifestyle and an important requirement to tolerate stress conditions under which metabolic redox fluxes change (Del-Saz et al. 2018). Consistently, the abundance of the A. thaliana AOX1a transcript is highly stress responsive (Van Aken et al. 2009). Specific AOX protein isoforms are further regulated in their amount in response to environmental stresses, such as pathogen attack, high light, heavy metal exposure, cold and drought [reviewed in Vanlerberghe (2013)]. Since changes in electron fluxes are coupled to metabolic fluxes, which frequently change too quickly for transcriptional or translational regulation to keep up, posttranslational regulation is particularly critical for AOX activity control.

To stimulate the activity of the AOX dimers an intermolecular disulfide between the two monomers that is exposed to the mitochondrial matrix needs to be reduced (Umbach and Siedow 1993). Poplar TRX-h2 was found to reside in the mitochondria and was the first TRX shown to be able to mediate the reduction and activation of AOX in vitro using purified soybean (Glycine max) mitochondria, even though how externally-added TRX-h2 protein accessed the matrix exposed AOX Cys in intact mitochondria remains unclear (Gelhaye et al. 2004). Mitochondrial localization was subsequently also reported for A. thaliana TRX-h2 (Meng et al. 2010). However, reactivity between A. thaliana AOXs and TRX-h2 may not necessarily carry significance in vivo, but rather indicate overlapping biochemical specificities of different TRXs (see Paradigm shift 3). Despite a growing body of work building on the assumption of mitochondrial matrix localization of A. thaliana TRX-h2 (da Fonseca-Pereira et al. 2020; Daloso et al. 2015; Del-Saz et al. 2018; Geigenberger et al. 2017), the existing evidence-base relying on transient, heterologous GFP localization (Meng et al. 2010) remains thin. Instead there are cumulative indications for a localization of TRX-h2 to cellular compartments other than the mitochondria, such as (i) the lack of any defined organelle targeting sequence, (ii) its absence from mitochondrial proteome datasets (Fuchs et al. 2020b; Senkler et al. 2017), including mitochondrial redox proteomes enriched for Cys-peptides (that reliably detect TRX-o1 and Trx-o2 (Nietzel et al. 2020)), (iii) plausibility based on A. thaliana TRX phylogeny (Geigenberger et al. 2017), (iv) the presence of an N-myristoylation motif for membrane anchoring (Boisson et al. 2003), and (v) the observation of association with the ER and the Golgi membranes (Traverso et al. 2013). In vitro AOX reduction can also be mediated by the most abundant TRX isoform in the mitochondrial matrix, TRX-o1 (Fuchs et al. 2020b; Gelhaye et al. 2004; Yoshida et al. 2013). TRX-o1 has been validated as a mitochondrial matrix protein in several independent studies and its function has been linked to a range of different phenotypes at the whole plant level [recently reviewed by Marti et al. (2020)]. Only after TRX-mediated reduction of AOX full activation mediated by α-keto acids, such as pyruvate or glycolate, can occur (Millar et al. 1993; Umbach and Siedow 1993, 1996). Sensitivity to different intermediates of the TCA cycle has recently also been found for the five A. thaliana AOX isoforms, suggesting that respiratory carbon status in the mitochondrial matrix plays an important role in regulating AOX activities (Selinski et al. 2018).

Despite the finding that reduction of the intermolecular disulfide of AOX dimers is required for full activity in isolated mitochondria, the key question of whether the two Cys thiols represent a redox switch that is indeed operated in vivo had remained a matter of debate (Del-Saz et al. 2018; Millenaar and Lambers 2003; Riemer et al. 2015). For redox regulation to occur the Cys thiol will need to be oxidized preferentially over other Cys in the mitochondrial matrix, requiring either (i) high reactivity directly with a reactive oxygen species oxidant, like H₂O₂, or (ii) secondary oxidation via a specific thiol redox relay, as mediated by an oxidized TRX or a thiol peroxidase. Immunoblotting analyses of whole tissue extracts from different species (using suitable blocking of thiols to prevent oxidation after tissue disruption) have indicated that AOX is not oxidized in vivo, even under stress conditions. By contrast, AOX proteins are strongly oxidized in isolated plant mitochondria. Hence, AOX cysteines may passively oxidize during the organelle isolation process when the normal electron flows from metabolism into the thiol redox system ceases to provide a sufficiently negative redox potential (Umbach and Siedow 1997). A recent study has tested the impact of the absence of TRX-o1 on AOX1a redox status in A. thaliana (Florez-Sarasa et al. 2019). Despite TRX-o1 being the most likely TRX to reduce AOX1a, AOX1a redox status was unaffected, suggesting that efficient backup exists from other components of the matrix Cys redox system to keep AOX reduced (Schwarzländer and Fuchs 2019). This observation strongly suggests that AOX cannot be easily oxidized in vivo. Interestingly, respiration rates were even increased in the trx-o1 background, suggesting an important role for TRX-o1 in throttling the rate of mitochondrial carbon metabolism upstream from the RET (Florez-Sarasa et al. 2019). While those observations were made in A. thaliana leaves under different illumination regimes, the respiratory rate in the trx-o1 background and in the genetic background of plants lacking NTRA/B or GR2, respectively, was consistently increased also in seeds during the early stages of germination (Nietzel et al. 2020). Those results taken together suggest that AOX normally remains reduced in vivo and redox regulation does not limit AOX capacity, which contrasts with in vitro systems that promote covalent linkage of AOX dimers by accidental disulfide formation. Instead, the control of respiratory flux occurs upstream at the level of central carbon metabolism, that has indeed been found to be regulated in its activity by Cys redox switches, some of which appear also operational under physiological conditions [recently reviewed in Nietzel et al. (2017); additional players identified in Nietzel et al. (2020)] (Figure 9). Consistently, photorespiration was also deregulated in the absence of TRX-o1, which was linked to redox-regulation of glycine decarboxylase (Reinholdt et al. 2019). In vivo data from four studies (Daloso et al. 2015; Florez-Sarasa et al. 2019; Nietzel et al. 2020; Reinholdt et al. 2019) point to an inhibitory role of mitochondrial TRX-activity on respiratory flux, i.e., the opposite of what would be expected if AOX was the decisive target. This concept of redox regulation of plant respiration at the level of carbon metabolism eventually links with AOX regulation; nevertheless, constitutive reduction of the AOX Cys allows direct regulation by the relative organic acid concentrations in the matrix (Selinski et al. 2017, 2018) and Cys-based redox regulation of the enzymes of matrix carbon metabolism is likely to be determined by organic acid accumulation (Daloso et al. 2015; Florez-Sarasa et al. 2019; Nietzel et al. 2020).

Figure 9:

Schematic model of redox regulation of plant mitochondrial respiration.

While the thiols of alternative oxidase (AOX) are efficiently maintained, even in the absence of thioredoxin-o1 (TRX-o1), carbon metabolism is de-regulated in the absence of TRX-o1 running at increased rates. AOX activity is unlikely to be redox-regulated, but the Cys thiols are maintained in a reduced state in vivo. Instead AOX is activated by organic acids, the concentration of which depends on redox regulation of carbon metabolism. TRX-o1 is the most abundant TRX in the A. thaliana mitochondrion, followed by TRX-o2 and receives electrons from matrix metabolism via NADPH and NADPH-dependent TRX reductases A and B (NTRA/B).

Paradigm shift 11: seed germination requires redox re-programming through global reduction and oxidation events

The germination of orthodox seeds is a particularly interesting biological situation, not only from a perspective of plant propagation and global nutrition, but also for redox regulation. Extended developmental and metabolic quiescence in the desiccated seed needs to be followed by a sharp, but well-controlled, transition to high activity in order to initiate germination. Overexpression of h-type TRX proteins in the starchy endosperm of cereal seeds (barley and wheat) accelerated germination as indicated by a more rapid increase of protein solubility, α-amylase activity and gibberellin content (Li et al. 2009; Wong et al. 2002). Vice versa, under-expression of specific h-type TRX isoforms prevented pre-harvest sprouting, indicating crosstalk between redox regulation, hormonal control of dormancy and germination, as well as metabolism across different seed tissues to execute germination.

Evidence has been accumulating for essential ROS-mediated oxidation as a critical factor for alleviating dormancy and driving germination. Non-dormant sunflower (Helianthus annuus) seeds appeared to generate more ROS within the first 24 h of imbibition than dormant seeds as indicated by fluorescein fluorescence in the embryo axis after 2′,7′-dichlorodihydrofluorescein diacetate (H₂DCF-DA) loading (Oracz et al. 2009). If dormancy was actively overcome by cyanide treatments, ROS production was accelerated reaching similar levels as early as 3 h after onset of imbibition. The stimulating effect of cyanide could be partially abolished by diphenyleneiodonium (DPI), indicating that RBOH activity (see Paradigm shift 9) may be responsible for the ROS burst promoting germination. A stimulating effect of H₂O₂ and a contribution of RBOHs in its generation are in line with reports on rice and barley seed germination (Bahin et al. 2011; Liu et al. 2007). An assessment of ROS generation in germinating pea (Pisum sativum) seeds indicated two bursts of extracellular ROS production. The first occurred within the first 30 min of imbibition, even though ambiguities in the detection did not allow interpretation with definitive confidence (Kranner et al. 2010b). For such a ROS burst to have effective regulatory impact on specific protein Cys targets, they must be in a state for oxidation to have a structural and functional impact, i.e., the Cys are typically required to be reduced. However, minimal metabolic electron flux and desiccated proteins and membranes in the quiescent seed allows for unregulated and unspecific redox modification of protein Cys as well as oxidation of glutathione to dominate (thermodynamic control; i.e., redox enzymes are present but redox potential is limiting), providing a non-ordered intracellular redox landscape as a starting point. Indeed, it has been recognized that progressive oxidation of GSH and protein Cys occurs during seed aging and that the Cys redox-landscape must be re-set at germination (Kranner et al. 2010a). Analytical measurements of glutathione redox status in wheat seeds suggested reduction in the first 4 h of imbibition followed by subsequent re-oxidation later in germination (Gerna et al. 2018). Redox-proteome analyses of Medicago truncatula seed extracts have further indicated TRX-mediated reduction of protein Cys as an early step in germination (Alkhalfioui et al. 2007). The precise time that the overall reduced intracellular protein Cys redox landscape is established for physiological redox regulation (kinetic control, i.e., the relative kinetics and specificities of redox enzymes set the redox state of specific Cys driven by redox potentials from oxidizing and reducing interactors that are not limiting) was not unambiguously delineated.

Recently, in vivo biosensing of glutathione redox potential in A. thaliana seeds identified the global reductive event as particularly rapid, occurring simultaneously with imbibition within the first minutes after addition of water and being completed within about 1 h (Nietzel et al. 2020). The rapid reduction is driven by restart of metabolic activity to provide electrons in the form of NADPH. Based on the speed of the reduction it is likely to be separated in time from the phase of (RBOH-associated) H₂O₂ production (Figure 10); alternatively, the phases of initial reduction and oxidation may overlap since metabolic NADPH provision is likely to be the prerequisite for both thiol reduction and a RBOH-mediated H₂O₂ burst. Hence, redox regulation of protein Cys occurs at a phase in early germination when the physiological conditions have not yet been set for hormonal control to be executed efficiently (e.g., by de novo gene expression). To investigate potential crosstalk, it will be of the highest interest to pinpoint the identity and the dynamics of the Cys-based redox switches that act as the most relevant regulators in situ and to dissect their mechanistic relationship to the hormonal control. Interestingly, a role for Cys redox regulation of delay of germination 1 (DOG1), a central regulator of seed dormancy, has recently been proposed (Nee et al. 2017). It will be intriguing to test whether the strategy of redox-based re-programming is unique to seed germination or whether more general underlying principles apply also in other systems that undergo transitions between quiescence and activity, such as germinating fungal spores or a mammalian egg cell at fertilization or rehydration of resurrection plants.

Figure 10:

Model of the dominating Cys redox dynamics in different stages of seed germination.

While the pronounced reductive transition at imbibition is driven by the thioredoxin and the glutathione systems based on NADPH-derived from metabolism, oxidation due to an RBOH-mediated H₂O₂ burst is also likely to rely on the availability of NADPH. The indicated changes apply for most exposed, redox-responsive Cys in cell compartments where thiols dominate after completed germination, e.g., in the cytosol and the mitochondrial matrix, but not in compartments in which active Cys oxidation is maintained, e.g., the ER.

Paradigm shift 12: plant mitochondrial respiration allows alleviation of thiol-based reductive stress

One of the most important insights from the use of the roGFP-based redox sensors has been the strict subcellular compartmentation of E_GSH (Meyer et al. 2007; Schwarzländer et al. 2016). As a result, the previous picture derived from glutathione measurements in whole cell extracts could be refined towards a cell biological understanding, where E_GSH is maintained at highly reducing potentials of about −320 mV in the cytosol of most organisms (equaling GSH:GSSG between 50,000:1 and 500,000:1 (Schwarzländer et al. 2016)), while that of the ER is much less reducing (about −210 mV). Such negative redox potentials in the cytosol reveal that the concentration of GSSG is in the picomolar range, equaling to only a few molecules of GSSG in the entire cytosol of a cell. This new picture has major implications on our understanding of redox regulation in the cytosol, and other compartments with similarly reducing E_GSH, such as the mitochondrial matrix. Importantly, there is little scope to rapidly shift the steady-state E_GSH towards even more negative values. The same is likely to be the case for most of the cytosolic protein Cys, even though the midpoint potentials of specific Cys can adopt more negative values, and specific Cys reactivity as well as kinetic control will allow exceptions from this rule (McConnell et al. 2019; Schneider et al. 2018). While any significant ‘overreduction’ of the glutathione pool by shifting GSSG to GSH cannot occur (and the same applies for a majority of protein Cys), such a shift is entirely feasible in compartments that contain a meaningful amount of oxidized Cys, like the ER (Birk et al. 2013; Meyer et al. 2007; Schwarzländer et al. 2008). The potential impact of thiol-based ‘reductive stress’ is hence particularly high in the compartments of the secretory system, but also the apoplast, the thylakoid lumen and the mitochondrial intermembrane space (Meyer et al. 2019). The other compartments are less vulnerable to dysfunction through Cys reduction and they are also poorly suited to safeguard neighboring compartments with oxidized Cys from reductive challenge. However, a protective role can be fulfilled by sinks, rather than Cys-based redox buffers, for superfluous thiol-based reductant. As an important example the ER has itself a sophisticated inventory to introduce and maintain disulfide bridges, with ERO using O₂ as a terminal electron acceptor (see Paradigm shift 8). The system is prone to limitations under high electron influx, however, leading to reduction-induced ER stress (Hwang and Qi 2018; Meyer et al. 2019).

While the genetic response program of the reduction-mediated unfolded protein response (UPR) has been researched at depth, indications are accumulating that direct help may also come from neighboring cell compartments. Such safeguarding mechanisms may include ER-membrane passage of external oxidant, such as H₂O₂ into the ER (Appenzeller-Herzog et al. 2016), as well as the active oxidation/detoxification of reducing molecules already in the cytosol, before ER-entry can occur (Figure 11). A relevant high-capacity electron sink may be the mitochondrial electron transport chain. Mitochondrial electron transport can serve as electron sink for oxidative protein folding as mediated by the sulfhydryl oxidase ERV1 system in the plant IMS (Meyer et al. 2019; Peleh et al. 2017). While the electron flux from IMS oxidative protein folding is small compared to fluxes that need to be dealt with under reductive challenge, the high capacity electron acceptor system that the plant RET provides is very well suited as a multipurpose oxidation platform.

Figure 11:

The role of the mitochondrial respiratory chain in alleviating thiol-based reductive stress.

Excess thiol-based reductant (blue ellipses), which may enter the cell as a xenobiotic, like dithiothreitol (DTT), or may be generated endogenously, has only minor quantitative impact on Cys in the cytosol, most of which are strongly reduced (blue background). The glutathione pool is highly reduced (GSH:GSSG at least 50,000:1), meaning that excess electrons cannot be buffered and reductive stress does not occur in the cytosol, even though a few specific protein thiols will be affected. Instead, thiol molecules may enter other cell compartments, like the ER, where significant amounts of oxidized Cys are present (red) and lead to reductive stress. To alleviate reductant load and safeguard oxidizing compartments, oxidation strategies are likely to include oxidation through thiol peroxidases, such as peroxiredoxins (PRX) and glutathione peroxidase-like (GPXL). Recent data also point towards mitochondrial respiratory electron transport (RET) as a high capacity mechanism for the detoxification of thiols by oxidation, raising the hypothesis of a new mode of interorganellar redox crosstalk. Note the rates by which mitochondria are able to accept electrons from thiols are currently speculative and requires further investigation.

Indeed, A. thaliana mutants lacking electron transport components, such as AOX1a, show transcriptomic responses that resemble those induced by dithiothreitol (DTT) as a potent reductive stress inducer (Giraud et al. 2011) and we have recently found RET mutants to be hypersensitive to DTT (Fuchs et al. 2020a). Plant redox biology has long focused on the electron donor systems of the antioxidant systems, while the availability of electron acceptors has often been taken for granted as long as oxygen is available. However, the exact path and the rate by which electrons reach oxygen is likely to be just as critical, and the different properties of the most obvious options, i.e., H₂O₂ (see Paradigm shifts 1 and 4), ERO (Paradigm shift 8) or indeed the mitochondrion through the RET are likely to play a crucial role of regulation and avoidance of dysfunction (Figure 11).

Outlook

Several of the recent paradigm shifts in plant redox biology illustrated in this review have been driven by methodological developments that have shaped modern biological research in more general, such as in vivo biosensing, functional proteomics and computational modeling. The decisive factor has been the adoption of those approaches by a global thiol redox community of researchers that is more interactive, interdisciplinary and collaborative than ever. Several avenues to the next important discoveries have been outlined here and it is only realistic to predict that our thinking about redox biology will keep on being driven by technical innovations. More refined mass spectrometry has already started to investigate the crosstalk between Cys-based redox switches and other different posttranscriptional protein modifications, and emerging imaging approaches, such as super-resolution light microscopy and cryo-electron microscopy are heralding exciting new insights into the structural organization, but also the in situ dynamics and principles of specificity in redox regulation. To give an example, an expanding new field in redox regulation is the functional switch of enzymes of central metabolism like GAPDH and malate dehydrogenase from catalytic enzyme to regulator of gene expression or translation [reviewed in Selinski and Scheibe (2020)]. Proteins with moonlighting function realize novel mechanisms of redox-dependent regulation.